Vibrational Spectros PDF

Volume 1
Theory and Instrumentation
INTRODUCTION TO THE THEORY AND PRACTICE OF VIBRATIONAL

SPECTROSCOPY
The Historical Development of Experimental Techniques in Vibrational Spectroscopy

Introduction to Vibrational Spectroscopy
Infrared Spectroscopy; Theory
Raman Spectroscopy; Theory
Optical Constants
Vibrational Band Shapes in Condensed Phases
Mid-infrared Spectroscopy of the Condensed Phase
Raman Spectroscopy of the Condensed Phase
High-resolution Fourier Transform Spectrometry of Gases
Instrumentation and Methods for High Resolution Gas Phase Spectroscopy
Raman Spectroscopy of Gases
Near-Infrared Spectroscopy of Synthetic and Industrial Samples
Near-infrared Spectroscopy of Biological Samples
Far Infrared Spectroscopy
INSTRUMENTATION FOR MID- AND FAR-INFRARED SPECTROSCOPY
Resolution and Instrument Line Shape Function

Monochromators for Mid-Infrared Spectroscopy
Infrared Filtometers
Continuous Scanning Interferometers for Mid-infrared Spectroscopy
Instrumentation for Phase- and Sample-Modulation FT-IR Spectrometry
Polarization Interferometers for Infrared Spectrometry
Diode Laser Spectrometers for Mid-Infrared Spectroscopy
Fabry-Perot Interferometers for Mid-infrared Spectroscopy
Instrumentation for Far-Infrared Spectroscopy
Sources for Mid- and Far-infrared Spectrometry
Synchrotron and Free Electron Laser Sources of Infrared Radiation
Detectors for Mid and Far-infrared Spectrometry
Optical Materials for Infrared Spectroscopy
INSTRUMENTATION FOR NEAR-INFRARED SPECTROSCOPY
Monochromators for Near-Infrared Spectroscopy

Near-Infrared Array Spectrometers
The Technology and Applications of Miniature Spectrometers
Tunable Filter and Discrete Filter Near-Infrared Spectrometers
Hadamard Transform Near-infrared Spectrometers
Infrared Emitting Diodes as a Light Source for Near-Infrared Spectroscopy
INSTRUMENTATION FOR RAMAN SPECTROSCOPY
Raman Monochromators and Polychromators

Volume Phase Holographic Optical Elements
Continuous Lasers for Raman Spectrometry
Array Detectors for Raman Spectroscopy
Fourier Transform Near Infrared Raman Spectrometry
Resonance Raman Spectroscopy
UV-Raman Spectrometry
Introduction to Nonlinear Raman Spectrometry
Coherent Anti-Stokes Raman Spectroscopy
Stimulated Raman Spectroscopy
Hyper-Raman Spectroscopy
TIME-RESOLVED SPECTROSCOPY
Fast Time-Resolved Mid-Infrared Spectroscopy Using an Interferometer

Instrumental Aspects of Time-Resolved Spectra Generated Using Step-Scan
Interferometers
Time Resolved Mid-Infrared Spectrometry Using an Asynchronous FT-IR Spectrometer
Nanosecond Time-resolved Dispersive Infrared Spectroscopy
The Use of Infrared Lasers for Fast and Ultrafast Time-resolved Infrared Spectroscopy
DICHROISM AND OPTICAL ACTIVITY SPECTROSCOPY IN VIBRATIONAL

SPECTROSCOPY
Linear Dichroism in Infrared Spectroscopy

Basics of Orientation Measurements in Infrared and Raman Spectrometry
Trichroic Infrared Measurements of Orientation
Vibrational Circular Dichroism Spectroscopy
Raman Optical Activity Spectroscopy
SURFACE-ENHANCED VIBRATIONAL SPECTROSCOPY
Electromagnetic Mechanism of Surface Enhanced Spectroscopy

Surface Enhanced Raman Scattering (SERS)
Surface-Enhanced Infrared Absorption Spectroscopy (SEIRAS)
OTHER INSTRUMENTAL APPROACHES FOR VIBRATIONAL SPECTROSCOPY
Sum-frequency Spectroscopy
Inelastic Electron Tunneling Spectroscopy
High-resolution Energy Electron Loss Spectroscopy (HREELS)
Inelastic Neutron Scattering Spectroscopy
Two-dimensional Vibrational Spectroscopy by Doubly Vibrationally Enhanced Four Wave
Mixing
Infrared Cavity Ringdown Spectroscopy
CALIBRATION PROCEDURES AND STANDARDS FOR VIBRATIONAL SPECTROSCOPY

Wavenumber Standards for the Mid-infrared
Ordinate Scale Standards for Mid-Infrared Spectrometry
Wavenumber Standards for Near-Infrared Spectrometry
Photometric Standards for the Near-infrared
Wavenumber Standards for Raman Spectroscopy
Photometric Standards for Raman Spectroscopy
Volume 5
Applications of Vibrational Spectroscopy in Life, Pharmaceutical and

Natural Sciences
BIOMEDICAL APPLICATIONS
Vibrational Spectroscopy and Pathology

The Role of Near-infrared Spectroscopy in Minimally Invasive Medical Diagnosis
Noninvasive Diagnoses by Near-infrared Spectroscopy
Glucose Measurements by Vibrational Spectroscopy
Vibrational Spectroscopy Applications in Clinical Chemistry
Vibrational Spectroscopic Studies of Microorganisms
Vibrational Spectroscopy in the Detection of Cancer
In-vivo Near-infrared Spectroscopy
Ex-vivo Vibrational Spectroscopy Imaging (biopsy - microscopy imaging)
Functional Infrared Imaging for Biomedical Applications
BIOCHEMICAL APPLICATIONS
Infrared Spectroscopy of Proteins

Raman Spectra of Proteins
Vibrational Spectroscopy of Lipids
Vibrational Spectroscopy of Nucleic Acids
Vibrational Spectroscopy of Carbohydrates and Glycoconjugates
Infrared Spectroscopic Studies of Biological Membranes
Raman Spectroscopy of Viruses and Viral Assemblies
Time-resolved FT-IR Difference Spectroscopy: A Tool to Monitor Molecular Reaction
Mechanisms of Proteins
PHARMACEUTICAL APPLICATIONS
Polymorphs, Solvates and Hydrates

The Analysis of Pharmaceutical Substances and Formulated Products by Vibrational
Spectroscopy
Guidelines for the Development and Validation of Near Infrared (NIR) Spectroscopic
Methods
Applications of Vibrational Spectroscopy to Combinatorial Chemistry
FOOD SCIENCE
Vibrational Spectroscopy of Food and Food Products

Applications of Vibrational Spectroscopy in Brewing
Vibrational Spectroscopy in the Analysis of Dairy Products and Wine
AGRICULTURAL APPLICATIONS
Vibrational Near-infrared SpectroscopyApplication to Grains and Derived Commodities

Applications of Vibrational Spectroscopy to the Analysis and Study of Forages
Vibrational Spectroscopy Imaging of Agricultural Products
GLOSSARIES
Glossary of Terms Used in Biomedical, Biochemical and Pharmaceutical Vibrational

Spectroscopy
Glossary of Terms Used in Vibrational Spectroscopy
Abbreviations and Acronyms Commonly-used in Vibrational Spectroscopy
Volume 2
Sampling Techniques for Vibrational Spectroscopy
MID- AND NEAR-INFRARED TRANSMISSION SPECTROSCOPY
Standard Sampling Techniques for Infrared Spectroscopy

Sampling Techniques in Near-Infrared Transmission Spectroscopy
Long Path Gas Cells
MID-INFRARED EXTERNAL REFLECTION SPECTROSCOPY
External Reflection Spectroscopy

Reflection-Absorption Spectroscopy of Thin-films on Metallic Substrates
Infrared Spectrometry of Thick Organic Films on Metallic Substrates
External Reflection Spectroscopy of Thin Films on Dielectric Substrates
Infrared Reflection-Absorption Spectrometry of Monolayer Films at the Air-Water
Interface
Polarization-Modulation Approaches to Reflection-Absorption Spectroscopy
Fourier Transform Surface Plasmon Resonance
Infrared Spectroscopic Ellipsometry
MID-INFRARED INTERNAL REFLECTION SPECTROSCOPY
Principles, Theory and Practice of Internal Reflection Spectroscopy

Macro and Micro Internal Reflection Accessories
DIFFUSE REFLECTION SPECTROSCOPY
Continuum Theories of Diffuse Reflection

Internal Transmission Spectroscopy
Discontinuum Theory of Diffuse Reflection
Accessories and Sample Handling for Mid-Infrared Diffuse Reflectance Spectroscopy
Optics And Sample Handling for Near-Infrared Diffuse Reflection
Integrating Spheres for Mid- and Near-Infrared Reflection Spectroscopy
OTHER IR SAMPLING TECHNIQUES
Infrared Emission Spectroscopy

Transient Infrared Spectroscopy
Photoacoustic Spectroscopy
Beam Deflection Photothermal Spectroscopy
RAMAN SPECTROSCOPY
The Optical Problem in Raman Sampling

FT-Raman Spectroscopy
SERS-based Raman Probes
Morphology-dependent Raman Measurements
LOW TEMPERATURE AND HIGH PRESSURE SAMPLING TECHNIQUES
Liquefied Gases as Solvents for Vibrational Spectroscopy

Matrix Isolation Spectroscopy
The Diamond Anvil Cell
Vibrational Spectroscopy at High External Pressures
MICROSCOPY
Infrared Microspectroscopy
Fourier Transform Infrared Imaging Microscopy
Near-Infrared Microspectroscopy
Raman Microscopy
Raman Imaging
Near-field Vibrational Spectroscopy
The Use of Near-field Probes for Vibrational Spectroscopy and Photothermal Imaging
DEPTH PROFILING BY VIBRATIONAL SPECTROSCOPY
Depth Profiling by Microspectroscopy

Depth Profiling by ATR
Depth Profiling by Photoacoustic Fourier Transform Infrared Spectroscopy
OPTICAL CONDUITS FOR VIBRATIONAL SPECTROSCOPY
Introduction to Fiber Optics for Spectroscopists

Optical Fibers for Mid-Infrared Spectrometry
Probes Based on Mid-Infrared Transparent Fibers
Sensor Systems Based on Mid-Infrared Transparent Fibers
Fiber Optic Probes for Near-Infrared Spectrometry
Fiber-optic Probes for Raman Spectrometry
Hollow Optical Conduits for Vibrational Spectroscopy
HYPHENATED TECHNIQUES
Gas Chromatography/Fourier Transform Infrared Spectroscopy

Directly-linked Gas ChromatographyInfraredMass Spectrometry (GC/IR/MS)
Supercritical Fluid Chromatography with FT-IR Detection
High Performance Liquid Chromatography-Fourier Transform Infrared Spectroscopy
Flow Injection AnalysisFourier Transform Infrared Spectrometry (FIA/FT-IR)
Raman-detected Capillary Electrophoresis
Thin-layer Chromatography and Vibrational Spectroscopy
Thermal Analysis/Infrared Methods
Thermal AnalysisRaman
ATMOSPHERIC
Applying Extractive FT-IR Spectrometry to Industrial and Environmental Gas Analysis

Open-path Fourier Transform Infrared Spectroscopy
Lidar and Laser Remote Sensing
Volume 3
Sample Characterization and Spectral Data Processing
SPECTRA-STRUCTURE CORRELATIONS
Spectra-Structure Correlations in the Mid And Far Infrared

Spectra-Structure Correlations in the Near-Infrared
Spectra-Structure Correlations in Raman Spectroscopy
Infrared and Raman Spectra of Inorganic and Coordination Compounds
Spectrastructure Correlations: Polymer Spectra
Hydrogen Bonding
Utility of Isotopic Data
Automated Spectral Searching in Infrared, Raman and Near-Infrared Spectroscopy
Expert Systems
GROUP THEORETICAL AND NUMERICAL APPROACHES TO THE CALCULATION OF

VIBRATIONAL SPECTRA
Group Theory and Effects of Symmetry Lowering on Vibrational Spectra

Normal Coordinate Analysis
Calculation of Vibrational Frequencies by HartreeFock-based and Density Functional
Theory
Calculation of Vibrational Frequencies by Molecular Mechanics
Vibrational Intensities: Interpretation and Use for Diagnostic Purposes
DISCRIMINANT ANALYSIS
Chemometrics an Introduction
Discriminant Analysis
Discriminant Analysis by Neural Networks
Classification Methods
TWO-DIMENSIONAL (2D) ANALYSIS
General Theory of Two-dimensional (2D) Analysis

Two-dimensional Correlation Spectroscopy in Vibrational Spectroscopy
SPECTRAL ENHANCEMENT AND BAND RESOLUTION TECHNIQUES
Smoothing and Derivatives in Spectroscopy

Resolution Enhancement Approaches
Curve-fitting: Modeling Spectra
QUANTITATIVE ANALYSIS
Beers Law
Classical Methods of Quantitative Analysis
Quantitative Spectroscopic Calibration
Inverse Least Squares and Classical Least Squares Methods for Quantitative Vibrational
Spectroscopy
Principal Component Regression and Partial Least Squares Regression
Wavelength Selection by Genetic Algorithms
Data Transfer
ANOMALIES, ARTIFACTS AND COMMON ERRORS IN USING VIBRATIONAL

SPECTROSCOPY TECHNIQUES
Mid-infared Spectroscopy: Anomalies, Artifacts and Common Errors

Anomalies in Near-infrared Spectroscopy
Anomalies and Artifacts in Raman Spectroscopy
GLOSSARY
Glossary of Terms used in Chemometrics

Volume 4
Applications of Vibrational Spectroscopy in Industry, Materials and the

Physical Sciences
ANALYSIS AND CHARACTERISATION OF POLYMERS AND RUBBERS
Qualitative and Quantitative Analysis of Polymers and Rubbers by Vibrational

Spectroscopy
Chemical Characteristics of Polymers and Rubbers by Vibrational Spectroscopy
Measurement of the Physical Characteristics of Polymers and Rubbers by Vibrational
Spectroscopy
Vibrational Spectroscopy of Polymer Composites
Vibrational Spectroscopy of Conducting Polymers
Molecular Orientation of Polymers by Infrared Spectroscopy
Infrared Spectroscopy in the Study of the Weathering and Degradation of Polymers
Determination of Optical and Dielectric Properties of Polymers by Vibrational
Spectroscopy
Measurement of the Thermal Properties of Polymers by Vibrational Spectroscopy
Measurement of the Solar Properties of Polymers by Vibrational Spectroscopy
RHEO-OPTICAL MEASUREMENTS OF POLYMERS AND RUBBERS
Rheo-optical Fourier Transform Infrared Spectroscopy of Polymers

Dynamic Infrared Linear Dichroism Spectroscopy
MATERIALS SCIENCE
Vibrational Spectroscopy of Carbon and Silicon Materials

Infrared Characterization of Device Quality Silicon
Raman Spectroscopy of Semiconductors
Vibrational Raman Spectroscopy of High-temperature Superconductors
Vibrational Spectroscopy of Thin Organic Films
Structure and Composition of Glasses and Amorphous Materials by Vibrational
Spectroscopy
SPECTROELECTROCHEMISTRY
Infrared Spectroelectrochemistry
Raman Spectroelectrochemistry
PROCESS VIBRATIONAL SPECTROSCOPY

Process FT-IR and FT-NIR Spectrometry
Raman Process Analysis: Technology and Applications
ATMOSPHERIC AND ASTRONOMICAL VIBRATIONAL SPECTROSCOPY
Astronomical Vibrational Spectroscopy

FT-IR Measurements of Atmospheric Trace Gases and their Fluxes
Passive Remote Sensing by FT-IR Spectroscopy
Remote Raman SpectroscopyProfiling Water Vapor and Aerosols in the Troposphere
Using Raman Lidars
Vibrational Spectroscopy in the Study of Fires
INDUSTRIAL APPLICATIONS OF VIBRATIONAL SPECTROSCOPY
Vibrational Spectroscopy in the Automotive Industry

Vibrational Spectroscopy of Colours, Dyes and Pigments
Vibrational Spectroscopy in the Paint Industry
Vibrational Spectroscopy in the Pulp and Paper Industry
FORENSIC APPLICATIONS OF VIBRATIONAL SPECTROSCOPY
Applications of Raman Spectroscopy to the Identification and Conservation of Pigments

on Art Objects
Applications of Vibrational Spectroscopy in Criminal Forensic Analysis
CATALYSIS
Application of Vibrational Spectroscopy in the Characterization of Oxides and Sulphides

Catalysts
Vibrational Spectroscopy of Zeolites
Vibrational Spectroscopy of Supported Metal Catalysts
Vibrational Spectroscopy of Model Systems for Adsorbed Species on Finely Divided
Metal Catalysts
OTHER APPLICATIONS OF VIBRATIONAL SPECTROSCOPY
Vibrational Spectroscopy of Liquid Crystals

Vibrational Spectroscopy and Supercritical Fluids
Frozen Gas and Organic Matrix Methods in the Study of Organometallic Photochemical
Intermediates
Applications of Vibrational Spectroscopy in the Study of Explosives
Application of Vibrational Spectroscopy to Geology
Applications of Vibrational Spectroscopy in Soil and Environmental Sciences
VIBRATIONAL SPECTROSCOPY IN EDUCATION
Infrared Spectroscopy and Education

Raman Spectroscopy in Undergraduate Education
The Historical Development of Experimental
Techniques in Vibrational Spectroscopy
Norman Sheppard
University of East Anglia, Norwich, UK
1 INTRODUCTION period, up to the year 2000, was marked by several major

experimental advances, which are discussed in the second
This article is entitled the historical development, rather part of the article entitled The Period, since 1945, of Mul-
than the history, of experimental techniques in vibrational tiple Applications of the IR and Raman Spectroscopies.
spectroscopy because the author is not a professional histor- This period is one of which the author has had consider-
ian. Also, this summary account of a very large subject able personal knowledge since starting in 1943 as a research
necessarily depends substantially on a number of more student using IR spectroscopy for wartime research. During
detailed articles that record the primary sources for par- this period, the use of vibrational spectroscopy increased
ticular areas. At the beginning of each section, these other greatly decade by decade so the account given here consti-
sources are listed. The majority of the original references tutes only an introductory outline.
can be found in these; apart from exceptionally important Since the mid-1960s, several additional methods of ob-
ones, the primary references given are those not readily to taining vibrational spectra have been developed, which have
be found in the other sources. important but specialized areas of application. These are
The article is divided into sections. The Laying of the described and evaluated in the third part of the article under
Foundations: The Period to 1945 covers the initial period the heading The Newer Vibrational Spectroscopies.
during which sound experimental techniques and theoretical
understandings were developed for molecular vibrational
spectroscopy in the infrared (IR) and Raman regions. This
2 THE LAYING OF THE FOUNDATIONS:
spans the period between the discovery of the IR region THE PERIOD TO 1945
of the spectrum by Sir William Herschel in 1800 and
2.1 Infrared spectroscopy: the development of
the summary of the achievements of these two principal
experimental techniques
forms of vibrational spectroscopy in the book Infrared and
Raman Spectroscopy of Polyatomic Molecules written by
In this section of the article, which covers only the more
Gerhard Herzberg and published in 1945.1 By the latter
salient advances, I am greatly indebted to the fine historical
date, the interpretation of such spectra had become well
account of the early developments in IR spectroscopy given
established through the application of quantum theory.
by Jones;2 for Raman spectroscopy to reviews by Long3
As a result of technical experimental developments and by Krishnan;4 and, more recently covering dispersion
during the Second World War, IR spectroscopy in particular spectroscopy as a whole up to 1930, to Brands book Lines
was poised in 1945 to make new and major contribu- of Light.5 The above accounts should be consulted for
tions to molecular structural analysis and to the quali- more detailed information.
tative/quantitative analysis of mixtures. This subsequent Vibrational spectra were first studied in the form of absorp-
tion bands in the IR region, which was famously discov-
John Wiley & Sons Ltd, 2002. ered, as the first non-visible region of the electromagnetic
2 Introduction to the Theory and Practice of Vibrational Spectroscopy
Figure 1. Portrait of Sir William Herschel, the discoverer of the

IR region. (By permission of the President and Council of the Figure 2. The apparatus used by Herschel to demonstrate the
Royal Society.) existence of IR radiation. (Phil. Trans. Roy. Soc., 1800, 90
284292, published by permission of the President and Council
of the Royal Society.)
spectrum, by the astronomer Sir William Herschel (Figure 1)
in 1800.6 It is interesting to quote from the introduction to
his second paper. spectrum obtained by placing a glass prism in front of a
slit cut in a window blind. The heat associated with differ-
It is sometimes of great use in natural philosophy, to doubt ent positions in the well-dispersed spectrum displayed on a
of things that are commonly taken for granted; especially horizontal surface was measured by mercury-in-glass ther-
as the means of resolving any doubt, when once it is
mometers with blackened bulbs (Figure 2). Herschel found
entertained, are often within our reach. We may therefore
say, that any experiment which leads us to investigate the that the temperature maximum shown by the thermometers
truth of what was before admitted on trust, may become of occurred just beyond the red end of the spectrum, i.e. at a
utility to natural knowledge. Thus, for instance, when we very different position from that of maximum luminosity,
see the effect of the condensation of the suns rays in the and that heating extended well beyond there.
focus of a burning lens, it seems natural to suppose, that
At first, Herschel proposed that he had measured an
ever one of the united rays contributes its proportional share
to the intensity of the heat which is produced; and we should extension to the visible spectrum that was otherwise cut
probably think it highly absurd, if it were asserted that off by the materials of the eye. But later, in response
many of them had but little concern in the combustion, or to criticisms and on the grounds that it seemed unlikely
vitrification, which follows, when an object is put into that that the human body would respond to the same type
focus. It will therefore not be amiss to mention what gave
of signals by two different senses, he preferred the view
rise to a surmise, that the power of heating and illuminating
objects, might not be equally distributed among the various that heat and light arose from different mechanisms.
coloured rays. The matter remained controversial for 30 years, but was
finally resolved in favor of his first hypothesis by the
This surmise, related to the unwelcome generation of heat demonstration that heat and light show the same type of
together with the light during his astronomical observations interference phenomena.
of the sun, led to Herschels famous experiment that discov- Herschels glass prism transmitted only the very begin-
ered what we now know as the IR region of the spectrum. nings of the IR range, what we now term the near-infrared
He investigated the distribution of heat in the visible solar (NIR) region. For further advances it was necessary to
The Historical Development of Experimental Techniques in Vibrational Spectroscopy 3
develop temperature-measurement methods that were more platinum wire as one arm of a Wheatstone bridge of
sensitive than the mercury-in-glass thermometer, to find which the resistance changed as a function of its change in
materials with better IR transmission properties, and to temperature, a method of heat detection earlier discovered
find sources of heat radiation for laboratory work that were by A.F. Svanberg in 1851. The wire was blackened to
more convenient than sunlight. The first of these objectives absorb the radiation, and its sensitivity was shown to
was realized by the discovery of the thermoelectric effect exceed that of Mellonis thermopile by about three orders
by Seebeck in 1822 who made antimony/bismuth thermo- of magnitude. Langley at this time was Director of the
couples and also an early thermopile by coupling together Allegheny Observatory in Pittsburgh and was particularly
several such junctions. Nobili in 1825 further developed 27- interested in the IR spectrum from the sun. In the course
junction thermopiles, and these were used by his younger of a gruelling expedition to 4000 m on Mount Whitney
colleague Melloni in 1833 to investigate the relative trans- in California with a glass prism spectrometer, he showed
mission of many materials, during which he found rocksalt that the solar flux was much improved at high levels due
to be particularly effective, using undispersed radiation to reduced absorption by the atmospheric gases, water
from a variety of heated bodies. The latter included an vapor and carbon dioxide, as had earlier been predicted by
oil lamp without its glass chimney, heated copper plates Tyndall on the basis of his undispersed measurements on
and platinum spirals, and blackened copper surface held at the atmospheric gases. On his later move to the Smithsonian
100 C by boiling water. Institution in Washington, DC in 1887, Langley (Figure 3)
John Tyndall at the Royal Institution in London, starting built a spectrometer using a 19 cm high rocksalt prism,
in the 1850s, used the thermopile and cells with rocksalt and was the first to continuously record the galvanometer
windows to make systematic measurements of the reduced deflection from the output of the bolometer as a function of
transparency to undispersed IR radiation of a wide variety the angle of refraction11 using a synchronously advanced
of gases and liquids.7 He was a pioneer in speculating photographic plate.
in terms of molecules, and vibrations within molecules,
as the origin of IR absorption. His results (including the
transparency of the gaseous elements O2 , N2 and H2 ) later
found full explanation in terms of individual absorption
bands associated with the molecular vibrational degrees of
freedom of the molecules concerned.
The first measurements of NIR absorption bands were
made by Sir John Herschel, Sir Williams son, in 1840
using solar radiation, a glass prism and the visually per-
ceived evaporation of alcohol on a blackened sheet of paper
as the detector.8 Melloni seemingly first observed absorp-
tion bands in what we now term the mid-infrared (MIR)
region using a prism made from a large natural crystal of
rocksalt for dispersion experiments.9 He showed that hotter
sources gave stronger radiation that was of higher mean
refractivity, i.e. what we would now express as shorter
mean wavelength. After several refinements, Melloni had
by 1850 designed a spectrometer with a slit, a rocksalt lens
to provide an image of the slit, and provision for moving
a slit-shaped thermopile across the spectrum generated by
the fixed-position prism. This apparatus remained the basic
instrument for the study of MIR radiation for nearly four
decades, but it was of limited resolving power because wide
slits were needed to fill the detector with radiation.
Samuel P. Langley decided in 1880 to overcome this
problem by designing a detector of small size for use
with narrow slits, which was also of low heat capacity
Figure 3. Portrait of Samuel P. Langley, initiator of the Langley
in order to obtain the highest rise in temperature from bolometer and the first to measure IR wavelengths (Smithso-
the consequently reduced heat flux.10 This detector, which nian Institution Archives. Record Unit 95, photograph collection,
he termed a bolometer, consisted of a narrow blackened 1850s-. Negative #10 619).
Langley was also responsible for a second fundamental member of Rubens school who had come from Cornell
advance in the IR region in that he was the first to calibrate University, discovered that many inorganic salts, including
spectra in terms of wavelength by measuring the refractive- those used as prism and window materials, give selective
index-versus-wavelength dispersion curve for rocksalt.12 bands of high, metallic-like reflectivity at longer wave-
He used a source of visible sodium D radiation, which lengths than their transmission limits. They described the
also generated a substantial IR continuum and dispersed filtered beams of radiation selected by these means rest-
this radiation with a concave diffraction grating to give strahlen; we now know that they are associated with the
multiple orders of the Na D-line. He projected the mixed fundamental lattice modes of the crystals. Reflection off
radiation onto the slit plane of his rocksalt spectrometer, a succession of such plates effectively gave a pure beam
which separated the IR and visible components and thereby of long-wavelength radiation. The beam-center wavelength
measured the refractive index of rocksalt at a number of was dependent on the substance being used to provide the
known IR wavelengths up to 5 m. With the help of this reststrahlen and the range of wavelengths available within
result, Knut Angstrom, whose father had done fundamental the resultant beam could be measured by dispersion with
work in the visible region, in 1889 and 1893 explored the a blazed diffraction grating of the type developed by
MIR spectra (at poor resolution) of CO, CO2 , CH4 , HCl, R.W. Wood and A. Trowbridge in 1910. The latter reflec-
and some larger organic molecules, to 5 m. He showed that tion gratings were ruled in a soft metal, such as speculum,
the longer wavelength absorptions were stronger than the with a shaped groove designed to increase sensitivity by
shorter wavelength ones, corresponding to what we now selectively reflecting radiation into a particular direction
term the fundamental and overtone regions, respectively, corresponding to the diffraction orders for particular wave-
and that the absorption bands were characteristic of the length ranges. Single orders of diffraction could be isolated
molecules concerned and not of their constituent atoms. either through the use of suitable filters or by predispersion
Over the next 40 years, the main developments in IR by a prism. Although several such gratings are needed to
techniques were made in Germany in the laboratories cover the wide wavelength range transmitted by a single
of Heinrich Rubens in Berlin and Friedrich Paschen in prism, the higher dispersion and resolution provided by the
Tubingen, both having earlier been students of August grating subsequently played an important role in the eluci-
Kundt in Strasbourg. Sometimes in competition, they used dation of the fine structure that was by then being found in
Langley-type methods to plot the dispersion curves of IR gas-phase absorption bands.
transmitting crystals, and by 1908 dispersion curves of flu- After their first use by E. Pringsheim in 1883, surface-
orite (CaF2 ), rocksalt (NaCl), and sylvine (KCl), had been metallized concave mirrors gradually replaced the non-
measured to their effective transmission limits as prisms achromatic rocksalt lenses of the Melloni/Langley spec-
at 9.7, 15.9 and 17.7 m (wavenumbers 1030, 628 and trometers but a quartz lens found specific use in the far-
565 cm1 ), respectively. All these were naturally occur- infrared (FIR) region because of its capacity to transmit
ring crystals. Potassium bromide (transparent to 25 m, and focus radiation of wavelengths beyond 40 m but to
400 cm1 ) only became available in 1930 when John absorb strongly much radiation of shorter wavelengths.
Strong, of Johns Hopkins University, developed methods Further developments in detectors took place between
of growing artificial single crystals of the alkali-metal 1890 and 1910. Compact linear iron/constantan thermopiles
halides. [To complete the picture cesium iodide (transpar- of low heat capacity, competitive with Langleys bolometer,
ent to 50 m, 200 cm1 ) was introduced by E.K. Plyler and were developed by Rubens in 1898. Other thermal detectors
N. Acquista in 1953.] were C.V. Boys use of a thermocouple as the coil of the
On increasing temperature of an IR source, the inten- galvanometer (the radiomicrometer, 1887), and the torsion
sity of the long-wavelength radiation increases, but that at radiometer perfected by Nichols in 1894 from the original
shorter wavelengths increases even more rapidly. Hence design by W. Crookes (1874). The latter consisted of two
the penetration further into the IR region is made dif- balancing vanes attached to a fine vertical torsion wire. The
ficult by the increasing necessity to exclude stray light absorption of heat by one of the vanes led to heat transmis-
of shorter wavelength. This was a serious problem with sion to the surrounding air molecules causing a twisting of
the usual sources for MIR spectroscopy, which had devel- the torsion wire. This was a very sensitive detector, cho-
oped in Rubens laboratory from the Welsbach gas-mantle sen by Coblentz for his famous series of IR experiments to
to the linear-shaped Nernst glower. The Nernst glower, be described below, although it had a longer response time
which was the preferred IR source for many decades, than a bolometer or a thermocouple. The thermocouple sub-
had an emission peak at about 1.5 m and consisted of a sequently became the most widely used detector until well
ceramic tube of rare-earth oxides, mostly ceria, resistively after the Second World War. Developments included the
heated in air to ca. 2000 K. E.F. Nicholls, a postgraduate use of semiconductor elements of improved thermoelectric
power (the HilgerSchwarz thermocouple) and the enclo- structure with a resolution higher than anything possible at
sure of thermocouples in a vacuum jacket with an IR that time in the MIR region. Later, further sensitization of
transmitting window in order to eliminate energy loss by IR plates by suitable dyes made available a wider range
convection and the thermal effects of draughts. of the NIR region for high-resolution measurements of the
The sensitive galvanometer that responds to the signal spectra of gaseous small molecules in what we now know
from the thermocouple has to be comprehensively screened as the overtone region. These advances were exploited
from mechanical and electromagnetic disturbances, and this in the 1930s by Herzberg and by Mecke, who was the
discouraged the use of IR spectroscopy for routine work first to analyze the vibrationrotation fine structure of an
for chemical structural analysis despite the techniques asymmetric top molecule water, H2 O.14
promise in this direction (see below). The galvanometer The NIR region could alternatively, at lower resolution,
was not replaced until electronic means of relaying the be explored using the experimental techniques of MIR spec-
output from the detector to a pen recorder were devel- troscopy as was shown by Coblentz (see below); during the
oped during the 1940s. As resolutions improved, another 1920s J.W. Ellis15 concentrated on such work and correctly
difficulty for routine applications was the complexity of identified absorptions in the 3.4 and 6.8 m (ca. 2950 and
the water vapor vibrationrotation absorption bands in 1450 cm1 ) regions from CH-containing molecules as the
the background spectrum from the spectrometers atmo- fundamentals from which the much weaker overtones and
spheric pathlength on which absorptions from the sample combinations accounted for the principal features of the
were superimposed. However, these experimental difficul- NIR spectra. He made similar identifications for absorptions
ties did not deter physicists interested in the detailed analy- from NH- and OH-containing molecules.
sis of high-resolution spectra of small polyatomic molecules
where these did not overlap with atmospheric absorptions.
2.2 The origins and interpretation of infrared
While progress was being made in the techniques of
spectra
the MIR and FIR regions, an important development also
occurred in the NIR region. This was the discovery in 1880 It was not until the advent of quantum theory in the
by William Abney, who worked in the Chatham School of first few decades of the twentieth century that a general
Military Engineering in England, that photographic silver understanding of the origins of molecular spectra in the
bromide films treated with nitric acid could be prepared visible, ultraviolet (UV) or IR regions could be achieved.
that were sensitive to the NIR radiation transmitted by a Towards the end of the nineteenth century a general view
glass prism. Photographic methods have the capacity to was that absorption or emission bands in the visible region
accumulate signals over the whole range of a spectrum related to resonances involving oscillatory charges within
simultaneously, a property today known as multiplexing, a molecule, perhaps of a bond-vibrational nature. How-
in contrast to the successive measurements at individual ever this could not account for the emission spectra of
wavelengths that are necessary when using a dispersion atoms, even although these gave line spectra that were
spectrometer. With this sensitivity advantage, Abney was simpler than the band spectra of molecules. Soon after
also able to make use of the high resolution associated with the discovery of the electron, H.A. Lorentz and J. Larmor
diffraction gratings designed for visible spectroscopy. In independently suggested in 1897 that visible spectra arise
1881 Abney and R.E. Festing published the NIR spectra of from oscillations of electronic charges within an atomic or
some 50 liquids, mostly organic compounds, between (as molecular framework. P. Drude in 1904 pointed out that
we now know) 0.75 and 1.1 m using the required long the much-reduced intensities (by several orders of magni-
pathlengths of up to 60 cm.13 They showed that certain tude) of even the stronger longer-wavelength IR absorptions
molecules, such as CCl4 and CS2 , did not absorb in this implied that these involved vibrations of heavier entities,
region and concluded that the observed absorptions gave such as atoms within a molecule, as intuitively anticipated
information about the different types of bonds involving by Tyndall 40 years earlier. In the first Solvay Conference in
the light hydrogen atoms (CH, NH, OH, etc.) present in 1911, N. Bjerrum, in the context of line series associated
the molecules studied. They identified certain absorptions with bands within visible molecular spectra, analyzed by
that could be attributed to particular chemical groups such H. Deslandres (18861888) in terms of parabolic functions,
as ethyl and aromatic CH bonds. This seems to have been introduced the idea that molecules are associated with three
the first example of the identification of group-characteristic types of energy of different magnitudes vibrational, rota-
absorptions in any part of the IR region. tional and translational which could give separate features
The use of diffraction gratings by Abneys group also in spectra. Lord Rayleigh had earlier in 1892 predicted, on
gave high resolution and in 1881 he resolved an absorption grounds of classical mechanics, that absorption bands of
band of atmospheric water vapor into its (rotational) fine linear molecules in the gas phase should have rotational
wings giving doublet structures, but at that time IR spectra Herzberg published his famous and comprehensive account
were of inadequate resolution to confirm this. In due course, of the subject in 1945.1
using improved resolution available in Rubens laboratory,
Eva von Bahr in 1913 observed such doublets for CO, CO2 ,
N2 O and HCl, in the latter case with resolved fine struc- 2.3 Applications of infrared spectroscopy
ture. Rayleighs classical theory assumed a continuum of
rotational energies and so quantum theory was required 2.3.1 Atmospheric studies
to account for the fine structure. Indeed, it was probably
From the earliest days of the study of the IR region there
because of the resolved fine structure, already observed in
was a strong interest in the flux of sunlight incident on
the visible spectra analysed by Deslandres, that Rayleigh
the Earth and the absorption of sunlight by the atmosphere.
did not recognize that the different branches constituted the
Poullet studied the total solar flux, measured as heat, as
rotational wings that he sought.
a function of a variety of atmospheric conditions in the
As quantum theory developed, the Schrodinger wave
1830s. This work, and Tyndalls study of the absorption
equation was applied to a series of models that accounted (or lack of absorption) of undispersed IR radiation by the
with great success for the features of MIR spectra as cor- atmospheric gases, led to conclusions about the relationship
responding to transitions between vibrational and more between atmospheric absorption and the Earths climate,
closely spaced rotational energy levels, the latter giving rise including the greenhouse effect. In the 1880s, Abney in
to the fine structure associated with a vibrational absorption the NIR and Langley in the MIR region studied dispersed
in the gas phase; NIR spectra as originating in overtones atmospheric spectra, and each made comparative measure-
or combinations of the vibrational fundamentals, observed ments at ground level and at high altitudes. Paschen and
with measurable intensity through vibrational anharmonic- Rubens extended the spectra of the atmospheric gases to
ity; and FIR spectra as originating in low-frequency vibra- longer wavelengths and discovered the strong absorption
tional fundamentals, e.g. the lattice modes of crystalline by atmospheric water vapor beyond 11 m, which we now
solids, or in pure rotational spectra in the gas phase. know to be from its pure rotation spectrum.
The choice of parameters required to fit the observed
IR frequencies leads, where the amount of experimental
data is adequate, to determinations of the harmonic and 2.3.2 Blackbody radiation and quantum theory
anharmonic force fields associated with bond-stretching, Langleys work in the MIR region, and Rubens study
angle-bending, or torsional vibrations from the vibration of FIR radiation using reststrahlen, played important roles
frequencies; the analysis of line spacings in the rotational in the evaluation of the wavelength profile of blackbody
fine structures leads to the evaluation of the three principal radiation as a function of temperature. It thereby became
moments of inertia of a molecule and, given the help of iso- clear that at shorter wavelengths there were important
topic substitutions, to individual bond lengths and angles. and systematic deviations from Rayleighs theory based
The rotational features in the spectra can be interpreted on classical mechanics. There seems to be little doubt
in even the finest details when taking into account vibra- that personal and cordial contacts between Rubens and
tionrotation interactions. Max Planck had an important influence on the latters
The consideration that transitional dipoles (in classical development of the quantum theory.
language, dipole changes) are required for there to be inter-
action between electromagnetic radiation and the molecular
motions was shown to lead to selection rules that cause
2.3.3 High-resolution vibrationrotation spectra
certain modes of rotations or vibrations to be unobservable The successful use of quantum theory to interpret vibra-
in the IR region. These systematic spectral absences, such tionrotation spectra, and indeed the development from the
as the lack of IR absorption from the vibrations of sym- old to the finally accepted version of that theory, depended
metrical diatomic molecules early observed by Tyndall, in on the resolution of the rotational fine structure in the spec-
turn provided information about the symmetry properties tra of diatomic and small polyatomic molecules. For this
of the molecules. More general considerations of symme- purpose, during the 1910s, H.M. Randall at the Univer-
try matters in vibrational spectroscopy were introduced by sity of Michigan constructed and subsequently developed
C.J. Brester in 1924 and systematized through the use of reflection grating spectrometers, with foreprisms for the
group theory by Wigner in 1930. selection of individual orders of diffraction. The Michi-
The overwhelming success of the quantum-mechanical gan laboratory ruled their own gratings of increasing
analysis of the vibrationrotation spectra of many small sizes for higher performance and in general remained
polyatomic molecules had become clear at the time that at the forefront of the development and applications of
high-resolution IR techniques from then until after the

Second World War. Many distinguished experimental spec-
troscopists received their initial training in Randalls lab-
oratory, including E.F. Barker, W.H. Bennett, J.D. Hardy,
E.S. Imes, C.F. Meyer, the brothers A.H and H.H. Nielsen,
J. Strong, G.B.B.M. Sutherland and N. Wright. Many of
these later set up their own research groups elsewhere. The
experimental work of the Michigan school was supported
by the accomplished theoretician D.M. Dennison, who
did original work on Fermi resonance between fundamen-
tals and overtones and combinations in vibrationrotation
spectra, and the effects of Coriolis vibrationrotation cou-
plings within the absorption bands from degenerate (or
between close-lying) vibrational modes. He also showed
how to interpret the vibrationrotation spectrum of ammo-
nia in terms of its molecular inversion. Other fine contri-
butions to the interpretation of vibrationrotation spectra
were by Harald H. Nielsen and, as noted earlier, by
R. Mecke.
2.3.4 Molecular structure analysis by vibrational

spectroscopy
As we have already seen, Abney13 was the first to associate
individual absorption bands in the NIR spectra with smaller
functional groups within complex organic molecules. It was
to be expected that similar regularities should be present
in MIR spectra, which we now know to have absorption
bands from the fundamental normal modes of vibration. In
1892 Julius recorded the spectra of 20 organic liquids to
the wavelength of 10 m and, for example, found that com-
pounds with methyl groups always gave an absorption band Figure 4. Portrait of William W. Coblentz, whose wide-ranging
at 3.45 m (2900 cm1 ). In 1896 M. Ransohoff concluded studies demonstrated the use of group-characteristic IR absorp-
tions for molecular structural analysis. (Photo courtesy of the
that an absorption band at 3.0 m correlated with the pres- National Institute of Standards and Technology, Technology
ence of the OH group in the spectra of six alcohols. Work of Administration, United States Department of Commerce.)
this type led into the major study of the IR spectra, mostly in
absorption but also in reflection or emission, of a very wide compounds.) Some problems were caused by Coblentzs
range of compounds by W.W. Coblentz at Cornell Univer- limited knowledge of organic chemistry, but he was very
sity (Figure 4), under the supervision of E.L. Nichols. His careful to purify his compounds and his wavelength calibra-
collected results were later published by the Smithsonian tion was accurate so that his collection of spectra have stood
Institution of Washington, DC.16 Coblentz measured the the test of time extremely well. After Coblentzs work the
spectra of 112 organic compounds to 15 m (667 cm1 ), potential usefulness of IR spectroscopy for structural diag-
the transmission limit of his rocksalt prism, with good nosis could not be in doubt. However, the difficulties of the
wavelength accuracy. Examples of his spectra are shown experimental techniques deterred most chemists, even after
in Figure 5. He listed 15 group-characteristic bands for Lecomte at the Sorbonne in Paris had greatly extended the
most types of XH groups (CH3 , CH2 , NH2 , OH, etc.), polar data available from 1924 onwards.17 Lecomte also showed
groupings like NO2 , CN, -SCN and -NCS, and of aromatic that the overall pattern of absorptions, many of which were
rings. Surprisingly he did not note the specificity of the not group-characteristic in nature, could be used as molec-
strong CDO bond-stretching absorptions, perhaps because ular fingerprints with the help of which closely related
of their (still systematic) variations in position in aldehydes, isomeric molecules (e.g. all the isomeric octanes) could be
ketones, carboxylic acids, esters, etc. (This deficiency was separately distinguished. This latter information proved to
soon after made up by W. Weniger who, in 1910, stud- be very valuable for the analysis of hydrocarbon fuels dur-
ied the spectra of a range of oxygen-containing organic ing the Second World War.
100%
90
80
70
60
50
40
30
20
10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
(a)
100%
90
80
70
60
50
40
30
20
10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
(b)
Figure 5. Coblentzs IR spectra of (a) ethylene (ethene) and (b) nitrobenzene. (Published by permission of the Coblentz Society, USA.)
To find group-characteristic absorptions is one thing; to high amplitude for the group motion, the displacements of
interpret them in terms of particular modes of vibration of the other atoms would be of relatively small amplitude. In a
the grouping in question is another. This is best done by series of related molecules, the different attached structures
considering the detailed analyses of the vibrationrotation will cause only minor variations in the group frequency. In
spectra of small molecules that contain the group in the 1930s Mecke developed a notation, n for bond stretch-
question. At first the physicists who obtained such spectra ing, for angle bending and g for out-of-plane modes,
took the view that it was unrealistic to look for group- which could be used approximately to describe the vibra-
characteristic vibrations. This is because, in principle, any tional motions involved in the group-characteristic vibra-
individual normal mode involves the vibrational displace- tional modes. The spectra of the small molecules showed
ments of all the atoms in a molecule with the same fre- that in general normal modes that involved bond stretch-
quency, and with the same phase in the sense that they pass ing were of higher frequencies than those that involved
through the equilibrium positions simultaneously. Never- angle-bending motions.
theless, it was realized that if an isolated group had a natural Before the Second World War, in 1939 several tens
frequency different from those of other atomic groupings to of IR spectrometers were in use, mostly in academic
which it was attached, then within a normal mode that has a research laboratories and a few latterly in industrial or
government laboratories, such as the American Cyanamid some 70 papers had already been published on the subject.
Co. (R.B. Barnes and V.Z. Williams), Dow Chemical Co. Smekal19 in Austria had previously predicted that such
(N. Wright) and the Shell Development Co. (R.R. Brattain an effect was possible on theoretical grounds. Landsberg
and R.S. Rasmussen) in the USA, I.G. Farben (now BASF) and Mandelstam20 in Russia independently observed the
in Germany (E. Lehrer) and the UK Government Labora- effect in light scattering in quartz and published their results
tory (J.J. Fox and A.E. Martin). Wartime analytical applica- shortly after Raman and Krishnans paper in Nature.
tions led to advances in electronic amplification methods, One reason for the rapid and continuing high interest
which made much easier the recording of IR spectra. At in the phenomenon was that the experimental requirements
last it was possible to exploit the full potential of the IR were much more easily met in the visible than in the IR
method, with the result that by 1947 there were more than region. The required intense source of monochromatic light
500 spectrometers in use, mostly for molecular structure was commercially available in the form of mercury dis-
analysis via group frequencies. charge lamps, the dust-free samples for investigation of
about 1 cm3 in volume could be conveniently contained in
glass sample cells, the dispersing element could be either
2.4 The Raman effect and its applications a glass prism or a diffraction grating, and the detector
was the already-available photographic film. The multiplex
The inelastic scattering of photons of light, named the advantage of photographic detection, and the possibility
Raman effect after its experimental discoverer (Figure 6), of integrating the signal through several hours of expo-
provides a second method for studying the frequency ranges sure, overcame the relative weakness of the proportion
associated with molecular vibrations and rotations. Raman of scattered light of changed wavelength that constitutes
and Krishnan18 discovered the effect in Calcutta in 1928 the Raman spectrum. Another advantage of Raman spec-
while studying the scattering of sunlight, made quasi- troscopy was that vibration frequencies could be measured
monochromatic by the use of a filter, by liquids. The interest down to low values close to the exciting line in one
in this discovery was such that by the end of that year, exposure; in comparison its IR counterpart had difficulties
measuring frequencies below the ca. 670 cm1 cut-off of
the usual rocksalt prism. Also, in comparison with the IR,
water was found to have a very weak Raman spectrum
so that such solutions of ionic or polar solutes could be
readily studied in the glass sample cells. A disadvantage
that subsequently emerged was that traces of colored, often
conjugated impurities in organic samples could lead to flu-
orescence that would swamp the weak Raman spectrum.
The purification of samples was found to be essential.
Rasetti, as early as 1929, showed that heavy scattering
near the exciting line could be avoided by using the mercury
resonance line. With this source, the light of unchanged
frequency (the Rayleigh line) could be self-absorbed by
mercury vapor in the optical path after Raman excitation.
He thereby obtained excellent Raman spectra of O2 and
N2 in which the absence or weaker intensities of alternate
rotation lines reflected nuclear spin statistics.
Photographic recording of Raman spectra remained pre-
dominant until the mid-1940s, when D.H. Rank during
the war developed a photoelectric Raman spectrometer
for quantitative hydrocarbon analysis, using a photomul-
tiplier as the detector, in order to improve on the limited
photometric accuracy available from photographic plates.
Photographic and photoelectric spectra of CCl4 are com-
pared in Figure 7.
The vibration or rotation frequencies are measured as dif-
Figure 6. Portrait of Sir C.V. Raman, discoverer of the Raman ferences between the frequencies (conventionally measured
effect (A.C. Cooper Ltd, London). in wavenumber units, cm1 ) of the monochromatic incident
0 = 22 943 cm1 a fundamental and a nearby overtone of the same symmetry,

(435.8 nm)
brought about through cubic terms in the potential function,
by chance arose from the experimental observation that two
rather than one strong Raman feature occurred in the Raman
Mercury
arc lamp spectrum of CO2 .
By 1939, over 1800 papers had been published on the
10 11 12 13 14 15 Raman effect, showing its wide-ranging applications. These
included vibrational studies of smaller molecules, which,
in conjunction with IR data, enabled complete or near-
complete assignments to be made of the frequencies of the
fundamental normal modes. However vibrationrotation
analyses remained principally the province of IR spec-
troscopy because of the weakness of rotational wings in
the Raman spectra. Kohlrausch in Graz, starting in 1929,
measured many Raman spectra of organic liquids and sum-
0 = 19 436 cm1 Argon
ion laser marized the results in several monographs.22,23 The group-
(514.5 nm)
characteristic frequencies in the Raman spectra proved to
be complementary to those in IR spectra, in the sense that
chemical groupings that gave weak or missing bands in the
IR often gave strong Raman features. Examples include
the bond-stretching modes of the nonpolar CDC double
Figure 7. The Raman spectrum of carbon tetrachloride, CCl4 ,
taken by photographic recording (Hg 435.8 nm excitation) and by and CC triple bonds (compared with the polar CDO or
photoelectric recording (Ar 514.5 nm excitation). This figure also CDN bonds, which absorb strongly in the IR) and bond-
demonstrates that virtually identical Raman spectra are obtained stretching modes involving heavy atoms of high polarizabil-
using different exciting lines. (Reproduced by permission of ity. Because of the then greater experimental accessibility,
P.J. Hendra, C. Jones and G. Warnes, Fourier Transform Spec- Raman rather than IR group frequencies tended to be listed
troscopy, Ellis Horwood, Figure 2.11, 30.)
in the physical chemistry textbooks until the 1950s.
light and of the Raman bands. For comparison pur-
poses, this encouraged IR spectroscopists to express their 3 THE PERIOD, SINCE 1945, OF
band positions in terms of wavenumbers rather than in
wavelengths as had been customary. It was immediately MULTIPLE APPLICATIONS OF THE
apparent that Raman spectra showed very different inten- INFRARED AND RAMAN
sity relationships compared with their IR counterparts. The SPECTROSCOPIES
most obvious example was that the vibrations of symmet-
rical diatomic molecules gave strong features in Raman During the Second World War, developments in electronic
spectra but are forbidden for symmetry reasons in the IR. amplification methods gradually eliminated the sensitive
The difference in behavior lies in the fact that the induced galvanometer, with its requirement of high mechanical sta-
re-emission that constitutes the Raman spectrum depends bility, as an intermediate between the detector output and
on the quantum-mechanical equivalents of induced dipole the recording of IR spectra. Secondly, the development
moments that fluctuate with the vibration or rotation fre- of double-beam spectrometers, whereby absorption spec-
quencies. Hence, unlike the IR case, it is the polarizability tra could be plotted automatically as a proportion of the
rather than the intrinsic dipolar properties of molecules that incident energy, eliminated the tiresome problem of having
determined Raman intensities. In an article published in to measure absorption spectra against the rapidly changing
1934, Placzek21 comprehensively laid out the theory for output, in wavelength terms, of the thermal IR sources. This
selection rules, relative intensities and polarizations of fea- was further complicated by the complex spectrum of water
tures in Raman spectra. This included the rule of mutual vapor in the spectrometer pathlength. Once these advances
exclusion between IR and Raman spectra that holds for had made the recording of spectra easy, the fact that spectra
the vibrations of centrosymmetric molecules, and the fact could be obtained from very small samples, irrespective of
that completely symmetrical modes of vibration (i.e. those whether these were in the forms of solids, liquids or gases,
that retain all the symmetry elements of the non-vibrating gave IR spectroscopy major advantages over its Raman
molecule) give polarized Raman bands. Another theoretical counterpart. From then on the IR method became extremely
understanding, that of Fermi-resonance interaction between important for molecular structure analysis and was made
available for wide use first in all organic, and later in inor- The IR source documents for this period are the com-
ganic preparative laboratories. The resulting wide potential prehensive books by Brugel24 published in 1962 covering
usage of IR spectrometers led to their commercial pro- dispersion methods, and by Griffiths and de Haseth25 of
duction so that from 1945 onwards, only spectrometers 1986, which covers IR interferometry; a set of papers by
of exceptional performance were built individually in uni- Miller,26 Wilks, Jr,27 Griffiths28 and Sheppard29 that origi-
versity or industrial laboratories. As a result, most further nated in the Waters Symposium of 1992 on the history of
developments were made by design teams of instrument commercial IR spectroscopy; and a similar paper by Beck-
manufacturers rather than by individuals. In what follows man published in 1977.30 For NIR spectroscopy, a book
there is the space available to mention the names of only a edited by Creaser and Davies in 198831 and a series of
few team leaders. papers in the journal NIR News32 38 provide good coverage.
A second major IR advance occurred in the 1970s when For Raman spectroscopy the reference sources are the
the development and incorporation of computers within books by Brandmuller and Moser39 of 1962, by Gilson and
spectrometers enabled the multiplex and energy-throughput Hendra40 of 1970, by Hendra et al.41 of 1991, and review
advantages of interferometry to be exploited. Fourier trans- articles by Long42 of 1977 and by Hathaway43 of 1971.
form mathematical analysis of the interferogram was used
to derive conventional intensity versus frequency spectra.
Fourier transform spectrometers, as they are now widely 3.1 The evolution of commercial infrared
known, are now standard in the IR region. The much spectrometers
higher sensitivity that resulted, combined with the many
advantages accruing from the possibilities of digital pro- 3.1.1 Dispersion spectrometers
cessing of data, opened up IR spectroscopy for applica- A commercial IR spectrometer, designed by F. Twyman,
tions to many areas previously considered difficult, such as had been put on the market as early as 1913 by Adam
IR microscopy, diffuse-reflection spectroscopy from pow- Hilger Ltd as their Model D83. It subsequently became
dered samples, photoacoustic spectroscopy (PAS), and the available with quartz, NaCl, CaF2 or KBr prisms in a
detection of spectra of chemisorbed species on catalyst Wadsworth minimum-deviation mounting, metal mirrors, a
surfaces. Nernst filament source and a thermopile detector. Neverthe-
Raman spectroscopy also obtained a new lease of life less it seems to have found limited usage. When wartime
from the development in the 1970s of lasers as ideal requirements for analyzing enemy fuels came to the fore
and intense monochromatic sources. Although small samin the early 1940s, Adam Hilger Ltd, in consultation with
ples could then be studied with ease, and although the G.B.B.M. Sutherland and H.W. Thompson of the Univer-
information from the spectra was complementary to that sities of Cambridge and Oxford, respectively, were given
available from the IR, Raman spectra did not even then a contract to produce a more modern recording version.
become as widely used for structural analysis, partially The new and more advanced Model D209 that resulted was
because in the intervening period nuclear magnetic reso- designed by W. Zehden and A.C. Menzies. It became the
nance (NMR) spectra had been developed, which provided first commercially produced double-beam instrument, and
similar structure-analysis information with efficiency. Nev- gave spectra in the form of percentage absorption versus
ertheless, the laser did make Raman spectra available on a wavenumber. The linear wavenumber drive was achieved
more routine basis and also opened up new important appli- by cams cut according to the dispersion curves of the prism
cations, such as the use of tunable dye lasers sources to materials, NaCl or KBr. The spectrometer had two separate
excite high-intensity resonance-Raman spectra from small light beams passing through upper and lower parts of the
biochemical samples. For industrial applications, problems slits and falling on separate HilgerSchwarz thermocouple
of fluorescence from impurities remained a problem. This detectors. The amplified outputs from these were ratioed
was addressed in the 1980s by the use of NIR lasers as using a pen recorder designed by the Cambridge Instrument
excitation sources in a region that is less likely than in Co. in which the output of the sample beam was measured
the visible to excite masking fluorescence spectra. How- against that of the blank beam in a potentiometer. In prac-
ever, Fourier transform methods could once again be used tice it was difficult to keep the two beams in adjustment and
because spectra in the NIR region are detector-noise lim- the spectrometer was more widely used as a single-beam
ited rather than photon shot-noise limited in sensitivity. instrument. Nevertheless it was a substantial pointer to the
This helped to compensate for the reduced scattering at future.
long wavelengths. Additional information about the conse- Wartime contracts, based on requirements for the analy-
quently widened fields of applications is given in separate sis of synthetic rubbers, were given likewise in the USA
sections below. to the National Technical Laboratories (later Beckman
Instruments Inc.) and to the Perkin-Elmer Corporation to mirror-chopper alternately presents one beam or the other to
produce single-beam spectrometers, the Models IR-2 and the detector; a variable aperture linear comb in the blank
12C respectively. The former was based on a design by beam is servo-operated so as to continually reduce the
R.R. Brattain of the Shell Development Company and the alternating current signal to zero and thereby balance
latter on one by R.B. Barnes, V.Z. Williams and colleagues the two beams. The pen records the movement of the
of American Cyanamid. In 1947 Baird Associates marketed aperture and thereby traces out the spectrum in percent-
a double-beam spectrometer, the Model NK-1, that was age absorption terms. The null principle, which had also
the first of this type to achieve substantial sales. It was a been independently and successfully developed by Lehrer44
modification of a double-beam design by N. Wright of the in Germany, became generally accepted for the use in
Dow Chemical Company and was also the first commer- double-beam spectrometers for another 20 years. Figure 8
cial spectrometer to offer calibrated spectra on preprinted compares single- and double-beam spectra of butene-1 pub-
charts. It used a null-type double-beam system in which a lished by Lehrer in 1942.
100%
-Butylene
0%
(a) 1 2 3 4 5 6 7 8 9 10 11 12 13 14
H2O
CO2
Luft
CO2
-Butylene
H2O
1.0 1.2 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
Luft
H2O
H2O
Luft
-Butylene
-Butylene
(b) 5.0 5.5 6.0 6.5 6.5 7.0 7.5 8.0 8.5 9.0 9.5
Figure 8. IR spectra of a-butylene (butene-1) obtained by E. Lehrer, which contrast the great convenience of the first double-beam
spectrum obtained by null beam balancing (upper spectrum) compared with the several sections of the corresponding single-beam
spectrum (lower spectrum below). In order to obtain a spectrum that is the equivalent of the double-beam version, the single-beam
spectrum has to be ratioed at numerous points against the background spectrum (labelled luft), which has rapid variations of intensity
with wavelength resulting from atmospheric absorptions superimposed on the output profile of the source.44
Prism instruments based on the null principle were later The spectrometers listed above used prisms as dispers-
marketed by Perkin-Elmer (the Model 21) and Beckman ing elements, usually of NaCl or KBr. In attempts to
(IR-4) in the USA; by Hilger (H800), and Unicam (SP100) achieve higher performance in resolution through improved
in the UK; and by Jena GmbH (UR-10) and E. Leitz GmbH dispersion, Perkin-Elmer manufactured a few single-beam
in East Germany. The later-appearing Unicam spectrometer, spectrometers, the Model 112, in which the prism was
designed by E.F. Daly, had a rotating optical comb, rather double-passed with a special beam-chopping system to
than a linearly driven one, for greater accuracy plus evacua- select the signal from the double-passed beam, according to
tion of the pathlength of the spectrometer to improve water a design by A. Walsh. However, this was soon overtaken
vapor spectral cancellation. by the advent of cheap replica diffraction gratings in the
The Perkin-Elmer Model 21, produced in the early 1950s early 1950s. In the UK these were produced in plastic on a
and designed by a team under V.Z. Williams, achieved cylindrical spiral master ruling cut by a method devised by
particularly wide sales because of its meticulous design for T. Merton of the UK National Physical Laboratory. A first
ease of operation (Figure 9). It was followed in 1957 by precision-lathe spiral was cut on a cylinder and used to gen-
the much cheaper and extremely successful Perkin-Elmer erate a second more accurate one in which the systematic
Model 137 (the Infracord) designed for widespread use errors of the first were averaged out by the use of a wide
for molecular structure determination by nonspecialists in cork clamp. Plastic pellicules carrying the imprint of the
organic chemistry laboratories. This was also manufactured precision ruling were removed by cutting along the length
in the UK by Perkin-Elmer Ltd. and was the forerunner of the cylinder. The resulting opened-out rectangular plas-
of a series of increasingly versatile low- and medium- tic diffraction grating was attached to a planar mount and
cost spectrometers designed in the UK by M.A. Ford. The front-aluminized. The UK firm of Sir Howard Grubb, Par-
analogous Beckman spectrometers were the Models IR-4 sons and Co., which had earlier manufactured single-beam
(W. Ward and colleagues, 1956) and IR-5 (G.T. Keahl, prism and grating models designed by A.E. Martin, were in
1957). 1955 the first to make available double-beam prism/grating
(a)
M14 M8
M15 W M4
L2 R
M6
S2 M7 I2
M2
M13 S L3 M10
1
S0
Pr M1
M11 M12
I1
M9 M5 L1 C S
M3
1 foot
(b)
Figure 9. (a) Appearance and (b) optical diagram of the Perkin-Elmer Model 21 double-beam spectrometer. S0 , source; M, mirrors
with M7 a semicircular chopping mirror to send the two radiation beams from the source alternately to the spectrometer entrance slit;
L, lenses; Pr, prism; S1 , S2 , entrance and exit slits. 1 foot D 30.48 cm. (Courtesy of the Perkin-Elmer Corporation.)
spectrometers using these replica gratings (the Model GS2).

The USA instrument firms, plus the Applied Physics Corp.
of California (the Cary-White Model 90), were quick to
follow suit using replica gratings supplied by Bausch and
Lomb. The fore-prism was required within the spectrometer
so that the spectrum could be obtained from a single
selected order of the diffraction grating. Later, from about
1960, it became possible to simplify the designs by replac-
ing the fore-prisms by IR filters. The Perkin-Elmer Model
137G for the region to 1330 cm1 , and the Model 237 for
the full rocksalt region, manufactured in the USA and
in the UK, were the first to offer this type of design. The
subsequent Model 457, with a caesium iodide thermocouple
window, provided low-wavenumber operation for inorganic
spectra and became a particularly popular medium-priced
spectrometer.
As IR spectrometers became increasingly used for quan-
titative analysis, limitations became apparent in the photo-
metric accuracy of null-type spectrometers at the extremes
of the percentage absorption scales. Ratio-recording sys-
tems, this time using a single detector with the signals
from the two beams electronically separated, as described
by Zbinden and colleagues in 1949, and by Savitzky and
Halford and Hornig and others in 1950, once again came
into vogue. Each technical advance was incorporated in
due course in moderate-cost and then low-cost models.
The Perkin-Elmer Model 580, manufactured in the UK
in 1975 and designed by M.A. Ford and H.M. Mould, Figure 10. A contemporary photograph of P.B. Fellgett who,
was the first medium-cost ratio-recording spectrometer. The anticipating the future development of computers, in 1951 descri-
Perkin-Elmer Corporation in the USA and their sister firms, bed in principle the multiplex advantage of interferometry. In the
late 1950s the method was put to use by H.A. Gebbie to improve
Perkin-Elmer Ltd. in the UK and Perkin-Elmer GmbH signal/noise in the FIR region, and by P. and J. Connes for high-
in West Germany (the latter manufactured the Model 125 resolution spectroscopy.
series filter-grating spectrometers), together with Unicam
(later Pye-Unicam and then Philips Analytical) in the UK, the combined interferogram by Fourier-transform mathe-
were firms that successfully stayed the course until Fourier- matical analysis. He foresaw the future development of
transform interferometric spectrometers took over the high- fast computers, whereby the computing time could become
performance end of the IR market in the 1970s. Before much less than the time for data acquisition, as leading to
then the most advanced commercially manufactured dis- the possibility of very advantageously analyzing complex
persion IR spectrometer was the Perkin-Elmer Model 180 IR spectra with high sensitivity. If, other factors being
manufactured in the USA (A. Savitzky and R.W. Hannah). equal, an interferogram is measured over the same period
of time as is required to record a dispersion spectrum with
3.1.2 Infrared interferometers; the N resolution elements, then any single element will be
Fourier-transform revolution recorded for an N-times longer period p in the interferogram,
The work of A. Michelson, who used interferometry to giving a signal/noise advantage of N. This gain in sig-
give very high resolution in atomic line spectra, is very nal/noise applies if, as is the case in the IR region, the
well known. It was pointed out in 1952 by P.B. Fellgett45 random fluctuations (noise) in the spectrum originates in
(Figure 10), in his PhD thesis submitted to Cambridge the detector, i.e. is independent of the signal strength (the
University, that in principle interferometry also provides gain does not apply if, as with photon-counting detectors
a multiplex advantage using broad-band sources in that an in the visible region, the noise is proportional to the square
interferogram records the superimposed signals from all the root of the signal). A typical IR spectrum between 4000 and
different wavelengths presented to the detector. The sepa- 400 cm1 with a resolution of 4 cm1 has N D 900 spec-
rate signals at each wavelength can again be recovered from tral elements giving a multiplex gain in signal/noise of 30.
This gain, already substantial, increases with the square root Michelson interferometer the two mirrors in the interfer-
of the resolution. Furthermore Jacquinot46 emphasized the ometer must remain accurately parallel and this is difficult
additional energy-throughput advantage of interferometers to attain for long mirror travel. The Connes overcame this
over dispersion spectrometers based on their circular aper- problem by replacing the plane mirrors with cats-eye opti-
tures in comparison with the entrance slits of the former. cal systems.
Spectral resolution in interferometry is determined by the In commercial terms, the next move was made by Block
reciprocal of the extreme path differences between the two Engineering for whom L. Mertz developed a MIR fast-
beams. It was not until the early 1970s that the develop- scanning Michelson interferometer. The resulting high fre-
ment of minicomputers enabled the practical realization of quencies of the components of the interferogram were such
the multiplex gains predicted by Fellgett. Nowadays, with that analysis could be carried out by use of a conventional
continued advances in computing speed, a 1000-point inter- wave analyzer following digital to analog signal conversion.
ferogram can be analyzed to give a normal intensity versus This spectrometer was however of very limited resolution
wavenumber spectrum in a fraction of a second. (1637 cm1 ) and for this reason did not find favor with
IR interferometry had been developed earlier by J. Strong the professional IR spectroscopists. Mertz later proposed
of Johns Hopkins University in the 1930s as a means of the use of a HeNe laser to accurately record the position
investigating the energy-deficient FIR region. H.A. Gebbie of the moving mirror.
was a member of Strongs group who, in the early 1960s The Block instrument however provided a vital develop-
at the UK National Physical Laboratory, set to work to ment step towards the very successful FTS-14 spectrometer
develop a computer-based Michelson interferometer for use of resolution 4 cm1 (or better) marketed by Block in
in the FIR region (Figure 11). He persuaded the firms 1969 under the Digilab label (Figure 12). T. Dunn and
Grubb-Parsons and Research and Industrial Instruments to R. Curbelo were responsible for the overall design, which
market FIR spectrometers based on his Cube interferom- incorporated a fast-scanning mechanism for the mirror
eter during the 1960s; later in the same decade Coderg in using an air bearing, a fast-response triglycine sulfate (TGS)
France and Polytec in Germany followed suit. These instru- pyroelectric thermal detector and an integrally incorporated
ments collected interferometric data point-by-point on paper Data General 8-kB Nova 16-bit minicomputer with a 14-bit
tape, which was processed, usually overnight, on the local analog to digital converter, backed up by a 500-kB disk
mainframe computer. It was still a very laborious process! memory for the Fourier transformation process. Because,
During this period P. Connes and J. Connes47 in France via the interferometer, the detector receives all frequencies
built a research interferometer for high resolution in the of the incident radiation at a time, its output is potentially
NIR region with a path difference exceeding 1 m, i.e. a of very high signal/noise, which could exceed the dynamic
prospective resolution of <0.01 cm1 . This required the range of the digital computer. The fast-scanning procedure,
computation of very large numbers, which was achieved with its consequent low signal/noise, allows the noise level
by use of the CooleyTukey algorithm for fast Fourier per scan to substantially exceed that of a single bit of the
transformation, introduced in 1965 and based on the use computer output while the magnitude of the signal still falls
of data points that are factors of two. For a conventional within the available dynamic range of the computer. The
coadding of data from many scans is used to give spectra
of high signal/noise.
Other manufacturers such as Idealab, Eocom, Analect,
Nicolet (with D. Mattson as designer who later formed
his own firm of that name), Bio-Rad (which incorpo-
rated Digilab), Bruker, and Perkin-Elmer gradually joined
the market. In most cases they adopted several features
of the DunnCurbelo inspired design, i.e. a Michelson
interferometer, fast scanning, a TGS detector, and a mini-
200 mm
processor (later a microprocessor as incorporated in per-
sonal computers, the power of which have continued to
develop rapidly). The Analect spectrometer had fixed cube-
Figure 11. The Cube interferometer developed by H.A. Gebbie corner mirrors and pathlength scanning by the relative
at the National Physical Laboratory (UK) for FIR spectroscopy. movement of wedge-shaped components of the beam split-
This was marketed within spectrometers by Grubb Parsons and
by the Research and Industrial Instruments Company during ter. Nicolet in its Model 7199 spectrometer incorporated
the 1960s. (Crown copyright, courtesy of the National Physical the Eocom interferometer and later produced the first
Laboratory.) medium-cost Fourier transform spectrometer, the Model
(a)
F1
IR source
Model 296 interferometer
Sample
compartment
TGS
F3 Reference or
M
irr sample cell
or
rB
C holders
rro
Mi
Mirror A Collimated lamp

for visual alignment
F2
(b)
Figure 12. (a) Appearance and (b) optical diagram of the Digilab FTS 14 Fourier transform interferometric IR spectrometer. (Courtesy
of Bio-Rad Laboratories.)
MX-1, in 1979. Perkin-Elmer were late into the Fourier Bomem have remained unique in offering Fourier transform
transform field with their Models 1800 and 1700. The spectrometers of very high resolution (0.002 cm1 ) with
latter was an instrument of high sensitivity and stability dynamic mirror-alignment (H. Buijs) since the mid-1980s.
with a self-aligning interferometer and path-length changes The overall result has been a total transformation of the
brought about by rotational rather than translational motion power and sensitivity of even routine IR spectrometers
(M.A. Ford). Perkin-Elmer were also the first to produce leading to the widespread use of a range of sample-
truly low-cost interferometers, the Model 1600 (R. Hoult) handling methods, to be described below, for which there
in 1987; its interferometer has since seen very widespread had previously been inadequate sensitivity.
usage. Today low-cost NIR instruments are also available. At the higher end of the market, such spectrometers
The IBM version of the Bruker spectrometer (the Model can now be bought with automatically interchanged beam
IR-44) was the first PC-based spectrometer, and Matt- splitters for the NIR, MIR or FIR regions (and for Fourier
son Instruments first introduced Windows -based software. transform Raman spectroscopy, see below) with fast-, slow-
or step-scanning, the latter being of importance for very reviewed by W. Kaye in 1954. The conversion of this tech-
rapid kinetic measurements of repetitive reactions, and also nique from one of occasional use to one of wide importance
for phase-modulation spectroscopy. Specially constructed for quantitative analysis occurred in the early 1970s in the
non-commercial interferometers are now capable of giving context of requirements for analytical figures for the water,
high-resolution vibrationrotation spectra of gases down to protein and oil contents of agricultural products, such as
the natural Doppler-limited linewidths (<0.001 cm1 ) over grains and soya beans. For example, diffuse reflection from
very wide wavenumber ranges. intact grains was found to be very successful as described
by the principal pioneer of the renaissance, K.H. Norris of
3.1.3 Tunable lasers as infrared sources the US Department of Agriculture31,36 and later by Norris
and his collaborator, P.C. Williams of the Canadian Grain
Lasers, with their coherence and narrow-beam properties, Commission.37 Pioneering dedicated instruments involved
are attractive sources, and if they are tunable they can the use of multiple filters, such as the DICKEY-john
eliminate the need for a spectrometer. Unfortunately in the GAC-1 (D. Funk and J. Anson) and the Neotec GQA-1
IR region, where they have been available for the past (R. Rosenthal, I. Landa and D. Webster) and many of these
1020 years, as yet tunability has proved to be possible remain in use; the Technion InfraAlyzer had 19 filters and
over only limited wavenumber ranges. was used worldwide. Today continuous rapid scanning of
At the high-frequency end of the IR region, optical
the NIR region is readily available from numerous instru-
parametric oscillators (pumped nonlinear crystals that can
ment firms, with computer control of data acquisition and
be tuned by crystal orientation or by temperature) allow
analysis. The Neotec 6350 (I. Landa) was the first commer-
tunability down to about 2000 cm1 .
cial digital scanning NIR spectrometer and its successor,
IR-emitting diode lasers can produce radiation over par-
the NIRSystems 6500, is very widely used. Davies48 has
ticular narrow ranges in much of the IR region, the posi-
recently reviewed the large number of spectrometers avail-
tion being determined by the flexible composition of the
able for analytical purposes. The Technion 400 was the first
diode materials available. Although any given diode can
digital filter instrument. Fourier transform spectrometers are
only be tuned over a range of about 100 cm1 , so that
also available as adjuncts to MIR Fourier transform instru-
in effect a diode is chosen to suit the absorption band
ments or as small dedicated instruments but are at present
under investigation, the resolution can be very high (ca.
mostly used for special applications. The NIR method has
0.0001 cm1 ), i.e. better than is needed to measure nor-
become very widely adopted by the food industry and its
mal Doppler line widths in spectra from gases. In addition
usage is spreading out strongly from there into pharmaceuti-
to their use for high-resolution vibrationrotation spec-
cal and general analytical areas, including process analysis.
troscopy, diode lasers have been used for the detection of
Much of the work involves analyses of different mixtures
gas-phase atmospheric pollutants and of species in high-
of known constituents. In this context, NIR was the first
temperature plasma.
spectroscopic field where purely computer methods were
Tunable IR lasers have also been used in conjunction
used to determine which wavelengths were most sensitive
with sum frequency generation (see Section 3.2).
for monitoring each required constituent within particular
types of samples. These did not always turn out to be
3.1.4 Near-infrared spectroscopy the expected overtones of the relevant CH, NH, OH, etc.
The NIR region of the spectrum, by convention taken stretching fundamentals. This field has also seen the first
to be the range between 4000 cm1 (2.5 m) and visible spectroscopic use of multivariate statistical (chemometric)
radiation, principally records much weaker overtones and methods for analyzing sets of related samples where the
combinations of the vibrational fundamentals. These mostly spectra of the individual constituents are not known. Prin-
involve the overtones of bond-stretching modes of XH, XH2 cipal component analysis is used for this purpose.
or XH3 groups (X D C,N,O, etc.) or these in combination In an academic context, Fourier transform techniques
with the angle-deformation modes of the same group. An in conjunction with long-pathlength interferometry give
advantage of the NIR region for analytical purposes is high-resolution vibrationrotation spectra, which, like their
that thick samples of a very wide range of materials can counterparts in the fundamental region, yield detailed
be studied, often without special preparation, by use of information on molecular dimensions. Analysis of the many
transmission or diffuse-reflection methods. overtone and combination vibration transitions in the NIR
Historically, after the advent of photoconductive detec- region has proved to be more straightforward using local-
tors, the region was first investigated as an adjunct to mode rather than normal-mode analysis. In part this is as
visible/UV spectroscopy, e.g. by use of the Cary Model 14 a result of Fermi-resonance complications when analysis is
UV/visible spectrometer. Work in this early period was attempted by the normal mode method.
3.1.5 Far-infrared spectroscopy individual experimental techniques. Salient advances, given

in alphabetical sequence, are described in outline below.
By convention this is usually taken to be the region below Other major advances during this period, which are less
400 cm1 down to the microwave region. It covers the dependent on experimental techniques, and result from joint
higher frequency range of pure rotation gas-phase spectra work involving IR and Raman spectra, are summarized at
(and of analogous motions in condensed phases) and in the end of this article.
this respect blends into microwave spectroscopy at the low
wavenumber end. It also includes the lower wavenumber
vibration frequencies of molecules, mostly of the angle 3.2.1 Attenuated total (internal) reflection
deformation or torsional types. This region requires some-
At an interface between materials of high and low refractive
what different experimental techniques to those used in
index, it is well known that total internal reflection of light
the MIR region. It is one in which thermal blackbody
occurs within the material of high refractive index at angles
sources have low output; for this reason, as described above,
of incidence above a certain value relative to the surface
it played an important historical role in the development
normal. When this occurs the reflected light penetrates into
of Fourier transform methods. Alternative nonequilibrium
the medium of lower refractive index by a distance of the
sources, such as high-pressure mercury arcs with quartz
order of a wavelength, the actual distance being dependent
envelopes, or synchrotron radiation, are used, the latter pro-
on the ratio of the two refractive indices. If a flat surface
viding a collimated beam of particularly high brightness.
(or, for certain purposes, a cylinder) of a high-index mate-
Beam splitters and cell or detector windows are normally
rial, typically germanium, silicon or zinc selenide (or more
of the Mylar -type or high-density polyethylene, respec-
recently diamond for small samples) is in contact with a
tively, whose more intense absorption bands are at higher
material whose absorption spectrum is required, the latters
wavenumbers, although cesium iodide can be used down to
features are recorded as changes in the reflection above
ca. 200 cm1 .
the critical angle as the spectrum is scanned. The atten-
uation of the reflectance is caused by the passage of the
3.2 Additional experimental and theoretical radiation through a thin layer of the low-index material.
techniques in infrared spectroscopy and The spectrum thereby obtained is a good approximation
their applications to a normal absorption spectrum. It can be enhanced in
intensity by multiple reflections of the beam if the attenu-
When commercial IR spectrometers became increasingly ated total reflection (ATR) element is a parallel-sided plate
available after the Second World War, the earlier pio- or cylinder. This technique was independently developed
neering work on group-characteristic IR absorptions by during 1960 and 1961 by N.J. Harrick and J. Fahrenfort,
Abney, Coblentz and others was greatly expanded to pro- respectively. It can be applied to any sample that can form
vide a major advance in chemical structural diagnosis, first close optical contact with the ATR plate. It provides, for
for organic and then for inorganic chemistry. Colthups49 example, a convenient means of obtaining IR spectra of
spectral chart and books by Bellamy50 and by Jones and solutions in water as the latter solvent has a very strong
Sandorfy51 ably systematized the field for the organic absorption spectrum. It finds multiple applications in indus-
chemists; a book by Nakamoto52 played a similar role for try to softer solid materials, such as plastics, pastes and
the inorganic chemists. Courses were set up to train new foods.
spectroscopists in industry, such as R.C. Lords IR Sum-
mer School at Massachusetts Institute of Technology; and 3.2.2 Depth profiling of samples
industrial/academic discussion groups were formed, such
as the Coblentz Society in the USA and the IR Discussion This can be carried out by the use of ATR as the depth
Group in the UK. Once the instrumental difficulties had of penetration of the radiation into the low-index material
been surmounted, IR spectroscopy benefitted greatly from decreases as the angle of incidence of the radiation beam
the fact that absorption spectra could be obtained from small on the internal surface of the ATR plate increases. In
amounts of material, whether in the gaseous, liquid or solid photoacoustic spectroscopy (PAS) the depth of penetration
states. of the recorded signal is greater the slower the chopping
New experimental techniques since 1945 enabled many speed of the incident beam, and hence the penetration depth
different types of problems and samples to be investigated, varies across the spectrum when such spectra are measured
often as a result of the major Fourier transform enhance- by normal Fourier transform infrared (FT-IR) techniques.
ment in general sensitivity. The book by Brugel24 and a The use of a phase-modulated step-scan Fourier transform
more recent one by Willis et al.53 give more details of the spectrometer can overcome this problem.
3.2.3 Detectors comparison with a transmission spectrum from the same

sample can occur in the regions of the strongest absorp-
The thermal detectors most widely used in dispersion
tions where related selective reflection reduces the entry
spectrometers are thermocouples or the Golay pneumatic
of radiation into the particles. (Reflection effects enhance
detector.54 The fast-scanning requirement of Fourier
the strength of absorption bands in transmission spectra,
transform spectrometry requires the use of the faster-
but reduce them in emission, photoacoustic and diffuse-
responding pyroelectric detector. Such thermal detectors
reflection spectra.) The differences with transmission spec-
have the advantage that they respond to radiation of
tra are reduced by mixing the neat samples at 10 wt% in
all wavelengths. Photoconductive detectors, based on
powdered KBr or KCl. A particular application of diffuse
semiconductor materials, are considerably more sensitive
reflection is to the study of catalytic heterogeneous reactions
and rapid in response but, unlike the thermal detectors,
proceeding in situ.
each individual photoconductive material is only sensitive
over a limited frequency range corresponding to the
width of its band gap. Pre-1940 work on the NIR PbS 3.2.6 Digitization of spectra
detector (sensitive down to ca. 3000 cm1 ) was extended
The necessity to incorporate a computer with an interfer-
to the use of PbSe and PbTe (sensitive down to ca.
ometer for Fourier transform purposes led to the realization
2200 and 1700 cm1 , respectively) by O. Simpson and
G.B.B.M. Sutherland in 1952 and by R.A. Smith in 1953. of the many other advantages that can accrue from the
Today the most widely used photoconductive detector in the recording of spectroscopic data by digital rather than analog
MIR region is HgCdTe, which, through the possibility of means. These include the formatting of spectra for repro-
variable composition, can give cut-off positions from 800 duction, the ready conversion of data between transmit-
to 500 cm1 with higher sensitivity for the former. They tance and absorbance, and the subtraction (in absorbance)
require to be operated at liquid nitrogen temperatures. Other or ratioing (in transmittance) of related spectra, the use
Cu- or Hg-doped germanium photoconductive detectors or of the KubelkaMunk function to convert diffuse reflec-
bolometers operate in the FIR region, but need to be cooled tion infrared Fourier transform (DRIFT) spectra to near-
to liquid helium temperatures and shielded from room- absorbance format, and the use of the KramersKronig
temperature radiation by low-temperature shrouds. function to convert reflectance data into refractive index
and absorption index information.
3.2.4 Diamond anvil (high-pressure) cell

3.2.7 Emission spectroscopy
Very high pressures of up to 105 atmospheres can be
applied to small samples, such as crystalline materials, by Such IR spectra of high-temperature flames had been stud-
sandwiching them between IR-transparent diamond anvil ied since the days of Coblentz. However, as the strengths of
plates forced together by hydrostatic or handscrew-and- emissions decrease rapidly with decreasing temperatures, it
lever methods. This type of cell is of importance for needed the sensitivity boost of the Fourier transform method
solid-state research and was developed by E.R. Lippincott before it could be applied to samples at moderate temper-
and colleagues in 1959. atures. It has been applied to the study of heated solids
that cannot conveniently be investigated by absorption, e.g.
coatings on nonplanar metal surfaces, and also to the anal-
3.2.5 Diffuse reflection spectroscopy ysis of hot remote sources such as effluents from smoke
This technique, which can be applied to powdered mate- stacks. Emission in the form of chemiluminescence was
rials without further preparation, had long been used in also used in Nobel Prize winning work by Polanyi and oth-
the visible region but found little use in the IR until the ers for near-transition-state kinetic studies of mixtures of
enhanced sensitivity of Fourier transform methods became reacting gases.55
available. R.R. Willey built a dedicated diffuse-reflection
spectrometer for this purpose in 1976 and M.P. Fuller and
3.2.8 Gas chromatography
P.R. Griffiths developed an accessory for use with standard
Fourier transform spectrometers in 1978. Accessories of This is a large and important subject that has been well
this general type, now manufactured by several companies, described in the book by Griffiths and de Haseth.25 The
have led to wide usage of the technique. The technique fast-scanning attribute of Fourier transform spectrometers
actually aims to measure the absorption by the particu- enables eluents to be identified as they come off the column,
late sample as reflected in an attenuation of the diffusely as first demonstrated by M.J.D. Low and S.K. Freeman in
scattered radiation. Substantial differences in a spectrum in 1967.
into a translucent disk by use of a hydraulic press. This

widely used technique, devised in 1952 by M.M. Stimson,
can, however, be misleading if there is chemical reaction
between the KBr and the sample in question.
3.2.12 Low-temperature matrix isolation

This technique normally uses a solidified inert gas as
the transparent matrix material. It has been widely used
for obtaining vibrational spectra of labile species, such
as free radicals, by their codeposition on a very low-
temperature surface with a large excess of the matrix gas.
The technique was introduced by E. Whittle, D.A. Dows
and G.C. Pimentel in 1954.56
3.2.13 Nujol mulls

This is an alternative way of overcoming scattering by
powdered samples by grinding them, in a miniature pestle
and mortar, in liquid paraffin (Nujol) of similar refrac-
tive index. A few absorption bands of the liquid paraffin
obscure narrow regions of the spectra; these can be regained
by use of the spectroscopically complementary perfluo-
roparaffin. This technique appears to have been introduced
by R.B. Barnes and R.S. McDonald in 1942, although
J. Lecomte had used paraffin wax for this purpose during
the 1930s. The method is widely used and has the advantage
Figure 13. The Perkin-Elmer IR microscope. (Courtesy of that the mulling agent is unreactive.
Perkin-Elmer Ltd.)
3.2.9 Infrared microscopy 3.2.14 Photoacoustic spectroscopy

IR microscopes require reflecting optics in order to avoid In this form of spectroscopy, a chopped beam of radia-
wide-region achromatism. Such microscopes were first built tion is incident on the sample held within a sealed cell
by C.R. Burch in 1947 and manufactured by Perkin-Elmer containing a gas. The fluctuating temperature of the sam-
at an early stage (Figure 13). IR microscopes have become ple caused by the absorption of radiation transfers heat to
very widely used since the onset of FT-IR. They are applied the gas phase and thereby gives rise to pressure fluctua-
to obtain spectra of very small samples or of small parts of tions, which are detected by a microphone. The resulting
larger samples. signals are dependent only on the energy absorbed and,
unlike transmission measurements, do not include a reflec-
tion component. The phenomenon was first observed by
3.2.10 Long-pathlength gas cells
Alexander Graham Bell in 1881. It was revived under the
A three-mirror design for a multiple-path gas cell, devel- name of the spectrophone by M.L. Veingerov and by
oped by J.U. White in 1942 for the detection of species in P.V. Slobodskaya in 1947 and 1948, respectively, in the
low concentrations, has found wide usage in both academic USSR.57,58 They showed that dispersion IR spectra could be
work and atmospheric studies. Overall pathlengths of a kilo- measured in this manner but their particular interests were
metre or more have been achieved. to measure the vibrational lifetimes of the absorption bands
of CO2 through the phase-dependence of the signal gener-
ated with respect to the beam-chopping cycle. The arrival of
3.2.11 KBr disks
FT-IR improved the sensitivity available and led to its more
This technique for improving the quality of spectra from extensive usage as pioneered by M.G. Rockley in 1979.
powdered solids involves mixing them with finely pow- An advantage of the technique is that it requires no sam-
dered IR-transparent KBr followed by pressing the mixture ple preparation, although finely divided samples of large
surface areas give the most efficient transfer of the signal calculated and observed spectral features for the purpose of
to the gas phase. As noted in Section 3.2.2, the different their assignments to particular vibrational modes.
modulation frequencies for different wavelengths provided
by the interferometer under normal Fourier transform oper-
3.2.17 Reflection spectroscopy
ation lead to different signal-penetration depths. Gaseous
samples generate the acoustic signal directly. Specular reflection from flat surfaces can be used to calcu-
late the refractive index and absorption index (absorption
3.2.15 Polarization and polarization-modulation per wavelength) for materials by KramersKronig analy-
spectroscopy sis as shown for the IR region by Robinson and Price61
in 19521953. Although the reflectance spectra of organic
IR polarizers are available for the study of oriented samples materials and some polymers can be very weak, once
such as crystals or polymers. These can be of the type of again the improved sensitivity of FT-IR enable these to
the pile of plates with radiation incident at the Brewster be measured. This is the most direct means of measuring
angle (A.H. Pfund and also A. Elliott and E.J. Ambrose, accurate absorption coefficients by ready calculation from
1947) using materials such as selenium and AgCl. More the derived absorption index. In the transmission method,
compact wire-grid polarizers, in which the grid is incor- both reflection and absorption contribute to the overall
porated in a soft optical material, were later developed. attenuation.
For certain purposes, such as reflection-absorption infrared
spectroscopy (RAIRS) and vibrational circular dichroism 3.2.18 Reflection-absorption infrared
spectroscopy (VCDS), where small differences between spectroscopy and the metalsurface
large signals have to be measured, polarization-modulation selection rule
followed by lock-in amplification is advantageous. Brad-
shaw and Hoffman59 used this method in conjunction with This technique enables spectra to be obtained from thin
dispersion RAIRS and Dignam et al.60 did so earlier using films on metal surfaces. Metals are near-perfect reflectors
the alternative method of IR ellipsometry. A.E. Dowrey and in the IR region and the radiation passes twice through
C. Marcott (1982) applied this method in Fourier transform the film. However the metalsurface selection rule (which,
spectroscopy; for the latter purpose a high-speed photoelas- for film thicknesses that are small compared with the
tic modulation of a linearly polarized beam was required wavelength, enables only vibrations with dipole changes
because of the necessity for the polarization modulation perpendicular to the surface to be observed) implies that
frequency to exceed that of the highest Fourier transform for high sensitivity it is necessary to reflect the beam
modulation rate by an order of magnitude. M. Grosjean and at near-grazing angles of incidence with polarization of
M. Legrand in 1960 achieved this by using an isotropic the radiation in the plane of incidence (R.G. Greenler,
optical element, typically ZnSe, that is alternately com- 1966). The selection rule was first used by Francis and
pressed and expanded by a piezoelectric transducer. Ellison62 in 1958 for flat metal surfaces and shown by
Pearce and Sheppard63 in 1976 to be of importance also
for the study of adsorption on metalparticle catalysts.
3.2.16 Quantum-mechanical calculations of
Although the selection rule is restrictive, it can be turned
vibrational spectra
to advantage to determine the orientation of a molecule
Such ab initio calculations of molecular structures, using with respect to the surface when the orientation of an
the minimum-energy or density-function methods, have absorption band is known with respect to the molecular
been successfully extended to calculating force fields for framework. As pioneered by Chesters64 in 1986, the RAIRS
vibrational motions within the molecule. Although these method is widely used in conjunction with FT-IR sensitivity
first-principle calculations have been only moderately suc- to study chemisorbed monolayers on metal single-crystal
cessful in reproducing individual experimental frequencies surfaces under ultra-high vacuum conditions. Surfactants
this is in large measure due to the fact that they have, on surfaces can be studied without the vacuum requirement
except in the case of the simplest molecules, been limited (see Section 3.2.15).
to the harmonic approximation. As a result the calculated
frequencies are usually higher than the experimental val-
3.2.19 Sum frequency generation
ues. On the other hand, these same methods can now be
used to calculate the approximate dipole or polarizability This is a form of spectroscopy in which two lasers, a fixed-
changes that accompany the vibrations, and hence the rel- frequency visible one and a tunable IR one, combine to
ative intensities of the bands in IR and Raman spectra; give a weak sum frequency beam that can be measured
the intensity plus frequency patterns help in correlating the by efficient detectors in the visible region. Photons are
produced when the two lasers interact on a noncentrosym- The method is increasingly used to characterize biological
metric target sample and when the IR active vibration is also molecules. The photoelastic modulation technique was suc-
Raman active. The method was developed by Y.R. Shen cessfully applied to VCDS using dispersion spectrometry
in 198765,66 and, because no signals are produced from by G. Holzwarth et al. (1974) and using Fourier transform
centrosymmetric media such as liquids, gases and opti- methods by L.A. Nafie et al. (1979).
cally isotropic solids, it provides a selective method for
studying interfaces. Cremer and Somorjai67 have recently
used the method to study species adsorbed on single-crystal 3.3 The Raman spectroscopies from 1945
metal surfaces of catalytic interest under high pressures of
gas, such as is frequently the requirement in real catalytic 3.3.1 Raman spectroscopy
systems.
For the first two decades after the Second World War, unlike
3.2.20 Surface spectroscopy in IR spectroscopy, there were few new developments in
techniques in Raman spectroscopy and much of the work
IR spectroscopy is now widely used to study the spectra done was of an academic nature. However, the introduc-
of monolayers that are physically adsorbed or chemisorbed tion of photoelectric recording of Raman spectra by Rank
on surfaces and, to a lesser extent, the structures of the et al.73,74 in 1942, using the photomultiplier as detector,
bare surfaces. Chemisorption is closely related to cataly- was in due course reflected in a number of commercial
sis and many such studies now routinely use such IR data. spectrometers because of the advantage for quantitative
The first successful spectroscopic experiments in adsorp- intensity measurements. Hybrid spectrometers incorporat-
tion used porous silicon oxide glass, a very high-area ing photographic and photoelectric recording were put on
material with low scattering, as the adsorbent. Studies of the market by Applied Research Laboratories, Hilger and
physical adsorption using this material were made succes- Watts Ltd. and C.A. Steinheil Sohne. Figure 14 shows
sively in the overtone region by Terenin68 in 1940 and in the Hilger spectrometer. The Applied Physics Corp. and
the fundamental region by Sidorov69 in 1954 and, inde- Perkin-Elmer Corp. marketed photoelectric-only spectrom-
pendently, by Yates et al.70,71 in 19551956. Work on eters. The former spectrometer, designed by H. Cary, was
chemisorption on finely divided metals, the metal particles probably the most sophisticated commercially available
being distributed over the higher-area surface of a trans- instrument until the advent of laser excitation. Its appear-
parent oxide, were initiated by Eischens et al.72 in 1954. ance and optical arrangement are shown in Figure 15.
They studied CO adsorbed on several metal catalysts of It used a Toronto-type low-pressure mercury arc for
this type. The weak nature of spectra from even frac- high resolution (H.L. Welsh et al. 1952), a grating double-
tional monolayers ceased to be a problem when Fourier monochromator to minimize stray light from the Rayleigh
transform methods became available. Normally the cat- line, and an image slicer for maximum throughput achieved
alyst samples are studied in the form of pressed disks
but more recently the original powdered catalysts have
been investigated by diffuse reflection, even under reaction
conditions.
The IR study of single monolayers on flat metal surfaces
by RAIRS was initiated in 1959 by Francis and Ellison62
and developed, together with the metal-surface selection
rule, as described above. The spectra from such a simplified
system, particularly when the surfaces of metal single
crystals are involved, have helped to interpret the more
complex spectra obtained on finely divided metals (see also
Section 3.2.15).
3.2.21 Vibrational circular dichroism spectroscopy

Optically active molecules of opposite chirality give identi-
cal IR spectra except when studied with circularly polarized
radiation. Even so the spectra from the two optically active Figure 14. The Hilger and Watts Raman spectrometer for photo-
forms are very similar and are best obtained as differ- graphic or photoelectric operation. (Courtesy of Hilger Analytical
ence spectra using high-frequency polarization modulation. Ltd.)
(a)
Interference
filter
Laser
Image slicer units Reference
beam
monitoring
laser
output
Sample
optics
Grating
Intermediate slit
Beam
Grating switch
Double monochromator
Reference and
main photomultiplier
tubes
(b)
Figure 15. (a) Appearance and (b) optical diagram of the Cary 81L Raman spectrometer. The original model, also widely used,
had Toronto-arc excitation. [(a) Courtesy of Varian Incorporated. (b) Reproduced from T.R. Gilson and P.J. Hendra, Laser Raman
Spectroscopy, Wiley-Interscience, Figure 2.3, 37 (1970).]
by spreading the originally circular image of scattered Focusing Fixed

radiation along a lengthy slit. mirror mirror
Beam Sample 200-mm
Gas lasers operating in the visible region became avail- LN2 - cooled splitter lens
able in the early 1960s and were soon shown, particularly Ge detector Moving
by S.P.S. Porto and his colleagues in 1962 and later, to mirror
be ideal monochromatic sources for Raman spectroscopy. Parabolic
Those favored were the HeNe, ArC or KrC gas lasers, such Dielectric collection
as were early supplied by Coherent Radiation or Spectra filters mirror
Nd/YAG
Physics; more recently the N2 laser has come into use for
laser with
excitation in the UV region. The narrow laser beam enables line filter
much smaller samples of liquids or powdered solids to be
studied with ease. Spatial filters
Perkin-Elmer, with their Model LR-1, were the first to Figure 16. The optical diagram of the pioneering Fourier trans-
produce a laser Raman spectrometer. This used a single form Raman spectrometer, with spectral excitation in the NIR
monochromator with double-pass optics of the Walsh type region, developed by D.B. Chase and T. Hirschfeld. [Reprinted
with permission from J. Am. Chem. Soc., 108, 7485 (1986). Copy-
but later commercial spectrometers, such as the Cary Instru-
right (1986) American Chemical Society.]
ments Model 81L (an adaptation of the earlier Hg-arc spec-
trometer) and the Spex spectrometer (S.P.S. Porto), which
the reciprocal fourth power of the wavelength. The Fourier
became available in 1966, incorporated double monochro-
transform method is applicable in the NIR region because
mators in order to give much improved performance near
the detectors are less efficient and once again noise-limiting.
the exciting line. Other firms such as Coderg, Jarrell-Ash,
An important design requirement is that the intense radia-
Spectra Physics, Hilger and Watts, etc. (for more details
tion from the exciting line is eliminated by use of a very
see the books by Hendra et al.40,41 and the article by
efficient filter. Rabolt and colleagues independently devel-
Hathaway43 ), and later Dilor and Jobin/Yvon, joined the oped a similar spectrometer in 1987.77
market for laser Raman spectrometers. The Coderg and Such Fourier transform-Raman NIR spectrometers, first
Dilor spectrometers were designed by M. Delhaye. Cary with germanium and later InGaAs detectors were soon
and Coderg later introduced triple monochromators, the shown to find useful applications for many types of sam-
Models 82 and T800, respectively. Jobin/Yvon, since joined ples previously plagued by fluorescence. Perkin-Elmer
with Spex and Dilor in S.A. Instruments, particularly pro- announced their Model 1700X (M.A. Ford), developed
moted the use of Raman microscopy for detecting the from a prototype by P.J. Hendra at the University of
spatial distribution of different components in composite Southampton, in 1988. By 1991 Bruker (B. Schrader),
materials. Nicolet, Bio-Rad, and Bomem marketed spectrometers,
A remaining difficulty for use with routine samples in with JEOL contributing later. All these used the same com-
Raman spectroscopy was that small amounts of impu- putation systems as for FT-IR but with detectors and beam
rities often excite fluorescence that swamps the Raman splitters appropriate for the NIR region. By 1999, some
features. Background smoothing does not solve this prob- 1500 Fourier transform-Raman spectrometers had been sold
lem because in the visible region, where photon counting and probably more spectra are recorded on such spectrom-
is used, the noise is proportional to the square root of the eters than on non-Fourier transform instruments because of
overall signal within the bandpass of the slit. However, the the ease and speed of data acquisition.
probability of exciting fluorescence falls off substantially The most recent addition to Raman techniques has been
at long wavelengths. Following an earlier pioneering con- the introduction of diode, or better, the more sensitive
tribution by Chantry et al.75 in 1964 in the days before charge-coupled device (CCD), detectors in arrays of ele-
small computers and improved filters, etc. were available, ments for work in the visible region. This is the semi-
Chase and Hirschfeld76 in 1986 built a very successful conductor/electronic equivalent of the photographic plate
spectrometer using the Nd/YAG (yttrium aluminum gar- and exploits the same multiplex advantage. These array
net) laser at 1.064 m as the excitation source in the NIR detectors hence provide uniquely high sensitivity for obtain-
region, together with highly efficient filters for rejection ing weak spectra, such as are obtained from adsorbed
of contributions to the Raman spectrum of scattering from species or in Raman optical activity. Such spectrometers
the exciting line. The optical layout of their spectrometer are necessarily operated at fixed resolution predetermined
is shown in Figure 16. Fourier transform interferometric by the dimensions and number of elements of the detec-
methods were used to boost sensitivity in order to compen- tor array but this can be as good as 1 cm1 . Instru-
sate for the fact that the intensity of scattering falls off as ments of this type have been marketed by firms such as
Renishaw, Kaiser Optical Systems, and Dilor (S.A. Instru- by R.W. Terhune, P.D. Maker and C.M. Savage in 1965,
ments). Ocean Optics have recently offered dedicated CCD the hyper-Raman bands being associated with a frequency-
instruments for process monitoring. CCD detectors are also doubled exciting line. The selection rules are different from
widely used in Raman microscopy, for example by the elec- those for the normal Raman effect and the phenomenon can
tronics industry. be used to identify modes of vibration that are forbidden in
normal Raman and IR spectra.
3.3.2 Resonance Raman spectroscopy
3.3.5 Inverse Raman spectroscopy
Resonance Raman spectroscopy (RRS) was discovered by
P.P. Schorygin in 1947 and independently by M. Harrand This form of spectroscopy involves irradiating the sam-
in 1948. It provides a great but selective enhancement of ple simultaneously with a strong pulsed monochromatic
intensity of Raman bands, by up to a factor of 106 , when laser and white light continuum. The Raman spectrum is
the exciting line falls within the fringes of an electronic observed as absorptions within the continuum. It was first
absorption band, but before greater overlap causes a full demonstrated by W.J. Jones and B.P. Stoicheff in 1964 but
fluorescence spectrum to be excited. The major exploita- its potential has so far been little exploited because of
tion of RRS occurred much later when tunable dye lasers the difficulties in generating wide enough continua in the
became available in the 1970s so that the optimum wave- required regions.
lengths could be chosen for exciting the spectra.
The theory of RRS was originally given in semiclassical 3.3.6 Stimulated Raman effect
form by Schorygin; the quantum mechanical equivalent was
developed by Behringer in 1958. Which of the Raman lines This form of scattering occurs in a forward direction by
are greatly enhanced depends on the nature of the elec- mutual stimulation when a giant pulse laser excites Stokes
tronic transition, but usually these are from the completely transitions. The effect was discovered by Woodbury and
symmetrical vibrations. The relative intensities within this colleagues in 1962. Although such spectra are intense, with
set are determined by FranckCondon type considerations additional features at multiples of the Raman frequency rel-
and depend on which molecular dimensions are affected ative to that of the exciting line, usually only the strongest
by the electronic transition. Such effects can also lead to Raman line gives stimulated emission and hence the ana-
extended series of overtones associated with the stronger lytical applications are limited.
Raman lines.
Because of its high intensity, RRS has found vari- 3.4 Additional techniques in Raman
ous valuable applications in different fields, such as bio- spectroscopy and their applications
chemistry and medicine for identifying transition-metal-
containing proteins, the study of molecular semiconductors 3.4.1 Raman optical activity
and, for archaeological purposes, the identification of pig-
This requires the detection of the weak differences between
ments and other constituents in paintings and ceramics.
spectra taken with beams that are circularly polarized
in opposite directions. The effect was discussed theoreti-
3.3.3 Coherent anti-Stokes Raman spectroscopy
cally by Atkins and Barron78 in 1969 and by Barron and
(CARS)
Buckingham79 in 1971 and experimentally pioneered by
This technique is dependent on the interaction of two lasers, Barron et al.80,81 in 1973. It finds application in the study
one of fixed and the other of variable frequency, which are of the conformations and (in conjunction with quantum
pulsed and of high intensity. The method was theoretically mechanical calculations) absolute configurations of chiral
suggested by P.D. Maker and R.W. Terhune in 1965 and molecules, including biomolecules.
brought to fruition by R.F. Begley, A.B. Harvey, R.L. Byer
and B.S. Hudson in 1974. Its principal application is in the 3.4.2 Optical fiber sampling
study of the spatial distributions of species in flames and
combustion processes. As Raman spectra are measured in the visible and NIR
regions, there is a considerable advantage in being able to
use optical fibers to transfer scattered light from a sample
3.3.4 Hyper-Raman spectroscopy
at a considerable distance from the spectrometer, and even
This very weak form of Raman spectroscopy, which also to convey the excitation to the sample through the use
depends on the variation of the third-order molecular of concentric cable. This capability has led to the rapid
hyperpolarizability during a vibration, became measurable expansion of the use of Raman spectroscopy for process-
with the advent of high-power lasers. It was discovered control purposes within industry.
3.4.3 Surface Raman spectroscopy wavelengths, more comparable to molecular dimensions,

than the photons in the IR or visible region. In this
Successful Raman spectra have been made of adsorbed
respect they provide additional information, such as the
species on finely divided oxide and oxide-supported metal
complete vibrational dispersion ranges in the spectra of
catalysts. These are very usefully complementary to IR
crystals or their surfaces. Except for specularly reflected
spectra from the same samples because the highly polar
beam of electrons in electron energy loss spectroscopy
oxides have only weak bands in the Raman spectra but
(EELS) (see below), there are no selection rules and rel-
absorb at blackout level in the lower wavenumber ranges ative intensities within the spectra can, in principle, be
of the IR spectra. For example, features can be used calculated from the amplitudes of the atomic motions dur-
to monitor the presence of surface MDO groups (M D ing the vibrations, given the relevant atomic scattering
metal) in spectra from transition metal oxide samples. Also, factors.
acidic sites on oxide catalysts can be monitored with high The reference sources for this section are books edited
sensitivity using a base with strong Raman lines, such as by Willis,87 by Bell and Hair,88 and by Yates and Madey,89
pyridine, as the probe. which cover the various techniques, and by Ibach and
By the use of CCD detection, it is even possible in Mills,90 which specifically covers EELS in more detail.
favorable cases to measure Raman spectra from single Many of the applications of these newer spectroscopies
monolayers on flat surfaces, such as from surfactants, as have been in the study of monolayers on flat surfaces.
demonstrated by Takenaka and Fukuzaki82 in 1979.
3.4.4 Surface-enhanced Raman spectra 4.1 High-resolution electron energy loss

spectroscopy
These greatly enhanced spectra of adsorbates, by factors
of up to 106 , occur on the surfaces of a few metals This is analogous to Raman spectroscopy with a beam of
(notably Cu, Au and Ag) that are in the form of parti- monoenergetic electrons replacing the beam of photons, the
cles with dimensions such that the plasma resonances of energy losses being the difference between the energies
the metal are stimulated by the laser beam. SERS spec- of the incident and reflected or scattered electrons. The
tra were first observed by Fleischmann et al.83 in 1974 first electron energy loss spectra of adsorbed CO were
from adsorbed species on a roughened electrode, and the obtained and interpreted by Propst and Piper91 in 1967.
effect was quantitatively evaluated independently by Jean- They published no further work in the field. In 1970
maire and Van Dyne84 and by Albrecht and Creighton85 H. Ibach reported work on the ZnO surface itself and
in 1977. They can also be obtained by adsorption on by the late 1970s his laboratory and several others were
colloidal metal particles, as shown by Creighton et al.86 doing systematic work on adsorbed species. The EELS
in 1979, or on cold-deposited metal particles. Chemical method has been particularly applied to the interaction of
bonding to the surface can increase the SERS enhance- electron beams with flat single-crystal surfaces of metals
ment, possibly through RRS contributions. The method and of the elemental semiconductors such as silicon. As
has been particularly valuable for studying the spectra of the lattice modes of these substrates are low in frequency
molecules adsorbed on electrodes of the SERS active metals and weak in intensity in EELS, very complete spectra can
in aqueous solutions because of the weakness of the Raman be readily obtained from adsorbed layers. This has been
spectrum of water. It has also found wide use in obtaining the major application of the technique; similar adsorption
the Raman spectra of biomolecules adsorbed on silver col- work on oxide surfaces or compound semiconductors has
loids. Attempts are being made to widen the phenomenon been limited by the strong reflection losses of the substrate
to other transition metals of catalytic interest, such as Pt itself.
and Ni. For electrons reflected in the specular direction, the
vibrational energy losses are principally excited by the
long-range electric field associated with the incoming elec-
4 THE NEWER VIBRATIONAL trons and the same metalsurface selection rule as in
SPECTROSCOPIES RAIRS dominates the spectrum, i.e. only vibrations with
dipole changes perpendicular to the surface are allowed.
These have in common the use of particles, i.e. elec- In comparison with RAIRS, the EELS sensitivity is high
trons, neutrons or He atoms, rather than photons to excite but the resolution attainable is low, rarely better than
molecular vibrational or rotational transitions. Under the 10 cm1 and more routinely 30 cm1 . The two meth-
appropriate experimental conditions the beams involved in ods are therefore complementary but it was in fact the
the reflection or scattering processes have much shorter great success of the EELS method, even with its limited
resolution, that led to the application of Fourier trans- 4.2 Inelastic electron tunneling spectroscopy
form methods by Chesters64 in 1986 to improve the
sensitivity of RAIRS for studying adsorbed monolay- In this form of vibrational spectroscopy the sample in ques-
ers under ultra-high vacuum conditions. As increasing tion is adsorbed on a finely divided oxide layer sandwiched
energy loss causes the reflected electrons to deviate more between two metal electrodes. The effect was discovered
from the specular direction, the specular EELS sensitiv- by chance by Jaklevic and Lambe94 in 1966 when measur-
ity is highest for low-energy losses and, for hydrogen- ing currentvoltage properties of such a metal/oxide/metal
containing species, weakest for the high-energy XH stretch- sandwich at liquid helium temperatures. At particular volt-
ing modes. ages the current was enhanced by tunneling and it was
The closer approach of electrons to the surface leads found that these corresponded to vibrational energy transi-
to a second the impact mechanism for exciting vibrations of pump oil, present as an impurity in the oxide film,
tional spectra. This leads to the scattering of electrons in which served to bridge the energy gap. Jaklevic and Lambe
a variety of directions but, while the intensity in any one went on to show that good vibrational spectra with band-
direction is weak, all the vibrational modes are in principle widths of ca. 10 cm1 could be obtained for other molecules
allowed under this mechanism. Willis and colleagues,92,93 deliberately introduced to freshly prepared oxide films, and
in a study of hydrogen atoms adsorbed on W(100), first that the energies of the effective transitions clearly cor-
observed such off-specular spectra in 1979 and showed that responded to their expected vibration frequencies. Most of
they gave information about additional modes of vibra- these were present in the spectra and resolved, but the inten-
tion with atomic motions parallel to the surface. These sity patterns within the spectra differed substantially from
two excitation mechanisms provide in practice the means those found in IR or Raman spectra. Similar results were
of identifying most of the modes of vibration. The over- later obtained for molecules chemisorbed on metal particles
all spectroscopic patterns have been widely and success- supported on the high-area oxide film. Although the exper-
fully used for deducing the structures of chemisorbed imental arrangements are straightforward, the use of liquid
species of interest with respect to catalysis. Typical on- helium temperatures is essential to give resolvable spectra
and off-specular high-resolution EELS spectra are shown and this limits the applicability of the method to adsorption
in Figure 17. research.
Recently a vibrational spectrum has been obtained by
Stripe et al.95,96 from a single adsorbed molecule in an
1000 1350
analogous experiment where the adsorbate was an ethyne
1130 molecule chemisorbed on a copper (110) crystal with the
Intensity (arbitrary units)
450 other metal electrode the tungsten tip of a scanning tunnel-

CO 2900
ing microscope.
1820 2040 (a)
2500
4.3 Inelastic neutron scattering

2940
1010 Monoenergetic beams of neutrons of thermal energies, from
1150 high-flux reactors or spallation sources, can be inelastically
1420
(b) scattered from the nuclei of a wide range of samples so as to
give energy-loss vibrational spectra. The experimental facil-
0 1000 2000 3000
ities have been increasingly available since the mid-1960s.
Energy loss / cm1
Rather large samples (several grams) are usually required
Figure 17. The electron energy loss vibrational spectrum of the as neutron scattering is weak, and the spectra are more
ethylidyne surface species, M3 CCH3 (M D metal), derived from
readily obtained and of better resolution at smaller energy
the dissociation of ethene, H2 CDCH2 adsorbed at 350 K on a
single-crystal Pt(111) surface: (a) with the electrons collected in transfers. There are no symmetry-based selection rules pro-
the specular reflection direction showing spectroscopic features hibiting particular vibrational transitions, but in practice the
dominated by the dipolar mechanism; and (b) collection at 14 uniquely high scattering by hydrogen nuclei (and the weak
off specular showing additional features excited by the impact scattering by the deuterium isotope) mean that vibrations
mechanism. The dipolar mechanism only excites those modes involving hydrogen nuclei often dominate spectra, and can
that give vibrational dipole moments perpendicular to the metal
surface. [Reproduced from H. Ibach and D.L. Mills, Electron be readily picked out when this element is present. These
Energy Loss Spectroscopy, Academic Press, New York (1982), are not just the bond-stretching and angle-bending modes
by permission of Academic Press.] involving XH bonds themselves, but also skeletal modes
in which the hydrogen atoms ride on the heavier ones. 5.1 Biomolecular spectroscopy
Low-frequency modes that are weak or difficult to obtain in
the FIR region have been frequently identified by neutron Not surprisingly this is now a very large area of research.
scattering. The technique was first successfully applied to Edsall101 was an early pioneer who studied the ionization
the study of the adsorption of hydrogen-containing species of aqueous amino acids as a function of pH using Raman
on the surfaces of high-area adsorbents so that the surface spectroscopy in 1936. In the 1940s and 1950s Darmon and
contributions are comparable with those from the bulk by Sutherland102 studied the IR spectra of amides and peptides,
Renouprez and others.97 Elliott and Ambrose103 used polarized IR spectra to study
the structures of oriented proteins, and R.N. Jones and
others made comprehensive studies of steroids.104 From the
4.4 Inelastic molecular beam scattering 1960s R.C. Lords group studied the spectra of nucleotides
and nucleic acids, and D. Chapman investigated phase
This technique, most frequently in the form of helium-atom changes in lipids and biomembranes.
scattering developed by Toennies98 in 1974, requires the
use of ultra-high vacuum techniques and is highly surface-
specific as the low-energy beams of incident atoms do not
5.2 Crystal spectra
penetrate into the bulk. In both these respects it differs
from, and is therefore complementary to, inelastic neu-
Born and Von Karman105 in 1912 were the first to consider
tron scattering. However, the chosen incident energies used
the vibrations of simple crystalline lattices in connection
are similar to those in inelastic neutron scattering so as to
with the specific heats (heat capacities) of crystals. Dur-
provide similar dispersion information with an energy reso-
ing the 1940s B.D. Saksena, J.P. Mathieu, R.S. Halford,
lution equivalent to a few cm1 . The technique principally
D.F. Hornig, S. Bhagavantam and T. Venkatarayudu bet-
provides information about the phonon-like low-frequency
ween them provided a comprehensive account of the spec-
(usually less than 100 cm1 ) vibrational degrees of free-
troscopic aspects of the vibrations of ionic and molecular
dom in ordered arrays of surface atoms or of adsorbates;
crystals in terms of the numbers of atoms/molecules within,
bulk modes, to the limited extent that they contribute to
and the symmetry properties of, the unit cell. This theoret-
the spectra, give broad background features. The extent to
ical framework has since been widely used, such as by
which parallel modes are excited depends on the degree of
the research groups of S. Califano (organic crystals) or of
corrugation of the surface and on the angle of incidence of
G. Wilkinson (inorganic crystals). Krishnan4 has provided
the particle beam. The surfaces themselves, or adsorption
a review of work until the early 1970s.
on metallic, ionic, or covalent solids can be studied so that,
despite the specialized techniques involved, the information
provided is highly complementary to that provided by other
forms of vibrational spectroscopy. 5.3 Double-minima spectroscopic phenomena
When the potential energy function that controls molecular

5 EPILOG vibrations has close-lying minima that are separated by a
low barrier, the vibrational transitions that occur below
The emphasis in this historical account has been on the barrier are very different in spacings from those that
advances in the experimental techniques within vibrational involve energy levels above the barrier. The early-studied
spectroscopy that have led to new types of experiments. classic case was the inversion of the ammonia molecule
However, since 1945 there have also been advances in studied by D.M. Dennison and others in the 1930s. Since
areas of understanding of spectroscopic phenomena of 1945 such phenomena have been discovered in a number of
great chemical importance that have resulted from extended different vibrational spectroscopic contexts. These include
applications of the traditional IR or Raman techniques, the inversion phenomena of flexible rings (R.C. Lord),
including fields that have flowered from seeds sown in and vibrations of floppy quasi-linear molecules such as
the pre-1945 era. To provide a degree of overall balance, carbon suboxide, OCCCO (P.R. Bunker). Double-minima
a very limited selection of some of the more important situations, mostly asymmetric, are also associated with
of these themes, up to the mid-1970s, is added below. remarkable experimental features in condensed-state spectra
More information about the multitude of present-day appli- involving strong hydrogen bonding (see below) although,
cations of vibrational spectroscopy can be found in two because of intermolecular interactions, the spectroscopic
multivolume series of reviews edited by Clark and Hester,99 phenomena are more easy to recognize than to evaluate
and Durig.100 quantitatively.
5.4 Hydrogen bonding and intermolecular coordinates, sometimes involving electron orbital follow-
interactions ing. M.W. Wolkenstein in the 1940s developed a bond-
polarizability model for interpreting Raman intensities,
Buswell and Rodebush106 and Freymann107,108 in the late which was subsequently developed by L.M. Sverdlov and
1930s were amongst the first to realize that the breadth of L.A. Gribov.
absorption bands associated with OH or NH bond-stretching
vibrations was associated with a hydrogen-bonding interac- 5.7 Internal rotation within molecules:
tion between such an acidic group and another basic atom. conformational (rotational) isomerism
IR spectroscopy in particular has led to a major understand-
ing of this phenomenon from the weakest such interaction When, as in the case of the paraffins or substituted paraffins,
to the strongest where, as in the case of the bifluoride ion, internal rotation leads to different relatively stable con-
FHF , the proton is equally bonded to each basic atom. In formational isomers, then the gas-phase or liquid/solution
between are to be found the medium-strength interactions spectra have additional complexities dependent on the num-
so important in determining the structures of biomolecules ber of such isomers. As Mizushima et al.111,112 showed in
such as proteins and DNA (L. Pauling), and very strong 1936, in such cases the IR or Raman spectra in the crys-
interactions involving double-minima potential wells for talline solid state are much simplified, and can often be fully
the hydrogen atom. The names of B.I. Stepanov (theory), vibrationally analyzed, because they often result from the
D. Hadzi, A. Novak (condensed phases), and D.J. Millen presence of a single conformational isomer. This is spectac-
(gas-phase studies) are some of those associated with ularly the case for the normal paraffins, which have many
this major body of research. Other forms of strong inter- conformations in the liquid state but adopt the single planar
molecular interactions, such as between Lewis acids of zigzag structure in the crystalline state.113 115 Such work
the AlCl3 type and organic bases, have been explored by led to the interpretation of the spectra of polymers, such
J. Lascombes group and others. as by S. Krimm and by G. Zerbi. In many other cases the
crystal spectra, in conjunction with solvent- or temperature-
dependent spectra of gas or liquid phases, have resulted
5.5 Inorganic spectroscopy in the identification of the structures of the different con-
formational isomers and the determination of their relative
General applications in this field took off later than in stabilities. This is exemplified in the work of the groups led
its organic equivalent, although Coblentz16 in reflection, by P. Klaeboe or by J.R. Durig.
and F. Matossi and others in 1926 in transmission,109 did
record the spectra of some minerals early while Duval in
Lecomtes laboratory published some more systematic stud-
5.8 Molecular force fields and vibration
ies in the 1940s.110 The early renaissance of such work
frequencies
after 1945 depended on contributions from groups such
The calculation of molecular vibration frequencies at the
as those of D.B. Powell, F.A. Miller, E.R. Lippincott and
simple harmonic level, much simplified by taking into
D.M. Adams, who developed the spectroscopy of inor-
account the symmetries of the different vibrations as set
ganic coordination compounds. Nakamotos book52 collects
out in the books by Wilson, Jr et al.116 or by Wolken-
together many of the significant publications.
stein et al.,117 became much easier with the development of
rapid computer-based iterative methods to optimize the fit
between experimental and calculated data. T. Shimanouchi
5.6 Intensities in vibrational spectra and his group systematically added non-bonded interactions
to the principally used valence force field. More recently,
From the late 1940s, attention turned to the experimen- as discussed earlier in this article, force fields and har-
tal determination of absolute or relative intensities in IR monic vibration frequencies for small molecules have been
and Raman spectra and their use to determine the mag- calculated from quantum-mechanical first principles with
nitudes of dipole or polarizability changes during molec- considerable success.
ular vibrations. Such phenomena were investigated by
B.L. Crawford, Jr, R.N. Jones, and W.B. Person and also,
from the functional-group point of view, by H.W. Thomp- 6 CONCLUSION
son and by others in the petroleum industry. IR intensities
were rationalized in terms of bond dipoles and their deriva- IR spectroscopy from the late 1940s was the first physi-
tives with respect to bond-stretching and angle-bending cal method to find near-universal application for molecular
structure identification in university, and then in industrial, EELS Electron Energy Loss Spectroscopy
laboratories. It was followed in the 1950s by NMR, and in RRS Resonance Raman Spectroscopy
the 1960s by mass spectrometry. These three spectroscopic SERS Surface-enhanced Raman Spectroscopy
techniques, in conjunction with advances in separation VCDS Vibrational Circular Dichroism Spectroscopy
techniques such as the gas and liquid phase chromatogra- YAG Yttrium Aluminum Garnet
phies, and the later-developed computer-based techniques
for the complementary X-ray diffraction method, revolu-
tionized the scope of chemistry during the second half of the REFERENCES
twentieth century. Because molecular vibration frequencies
are little (and even then informatively) affected by inter- 1. G. Herzberg, Infrared and Raman Spectra of Polyatomic
molecular interactions, vibrational spectroscopy is unique Molecules, Van Nostrand, New York (1945).
in its range of applicability to samples of every type. From 2. R.N. Jones, Analytical Applications of Vibrational Spec-
Sir William Herschels first experiments to today, the vast troscopy A Historical Review, in Chemical, Biological
range of knowledge that has been generated by this field of and Industrial Applications of Infrared Spectroscopy, ed.
research has been dependent on those who, in academia or J.R. Durig, John Wiley & Sons, Chichester, 150 (1986).
in the instrument industry, have developed better and bet- 3. D.A. Long, Internat. Rev. Phys. Chem., 7, 317 (1988).
ter experimental techniques. This historical account salutes 4. R.S. Krishnan, Historical Introduction, in The Raman
their collective achievements. Effect, ed. A. Anderson, Marcel Dekker, Inc., New York,
147, Vol. 1 (1971).
5. J.C.D. Brand, Lines of Light The Sources of Disper-
ACKNOWLEDGMENTS sion Spectroscopy, 18001930, Gordon & Breach, London
(1995).
The author is indebted to a number of friends and col- 6. F.W. Herschel, Philos. Trans. R. Soc. London, 90, 49, 255,
284, 293, 437 (1800).
leagues who have read parts of this article in draft and who
have made important contributions to its content and accu- 7. J. Tyndall, Heat, a Mode of Motion, 6th edition, Long-
mans, Green and Co., London (1880).
racy. These are Professor Foil Miller, the editors (Professor
Peter Griffiths and John Chalmers), Dr R. Norman Jones 8. J.F.W. Herschel, Philos. Trans. R. Soc. London, 130, 52
(1840).
and Professor John Brand, who have seen and commented
on most of the content; Professor Peter Fellgett, who read 9. M. Melloni, Annal. Chim. Phys., 53, 5 (1833); 55, 337
(1835).
the FT-IR section; Dr Michael Ford, who discussed with me
the section on commercial IR spectrometers; Dr Upali Jaya- 10. S.P. Langley, Phil. Mag., 3, 282 (1872).
sooriya and Professor Patrick Hendra, who advised on the 11. S.P. Langley, Phil. Mag., 15, 153 (1883).
Raman spectroscopy section; Mr Tony Davies, who read 12. S.P. Langley, Phil. Mag., 17, 194 (1884); 22, 149 (1884).
the section on NIR spectroscopy; and Professor Michael 13. W. Abney and R.E. Festing, Philos. Trans. R. Soc. London,
Chesters, who reviewed the section on the newer spectro- 172, 887 (1881).
scopies. All remaining errors are my own responsibility. 14. R. Mecke, Naturwissenschaften, 20, 657 (1932).
I am also indebted to the following individuals for
15. J.W. Ellis, Trans. Faraday Soc., 25, 888 (1929).
their help in locating figures for this article: David Bax
of Hilger Analytical Ltd.; Cindy Clark of the National 16. W.W. Coblentz, Investigations of Infrared Spectra, Parts I
to V, Carnegie Institution, Publication no. 35, 65 and 97
Institute of Science and Technology, USA; William Cox of (19051908). Republished by the Coblentz Society and the
Smithsonian Institution Archives, USA; Richard Crocombe Perkin-Elmer Corp., Norwalk, CT (1962).
of Bio-Rad Laboratories; Richard Spragg of Perkin-Elmer 17. J. Lecomte Le Spectre Infrarouge, Les Presses Universi-
Ltd.; and Christine Woollett of the Royal Society. taires, Paris (1928); Spectroscopie dans lInfrarouge, in
Particularly to Norman Jones, and later to John Brand, Handbuch der Physik, Vol. XXVI, Light and Matter II,
we all owe a debt for their pioneering contributions2,5 to Springer-Verlag, Berlin, 244937 (1958).
the early history of the IR and Raman spectroscopies. 18. C.V. Raman and K.S. Krishnan, Nature, 121, 501 (1928).
19. A.G. Smekal, Naturwissenschaften, 11, 873 (1923).
20. G.D. Landsberg and L.I. Mandelstam, Naturwissenschaften,
ABBREVIATIONS AND ACRONYMS 16, 557 (1928).
21. G. Placzek, Rayleigh Streuung und Ramaneffekt, in
CARS Coherent Anti-Stokes Raman Spectroscopy Handbuch der Radiologie, ed. G. Marx, Akad. Verlagsge-
DRIFT Diffuse Reflection Infrared Fourier Transform sellschaft, Leipzig, 205, Vol. 6 (1934).
22. K.W.F. Kohlrausch, Der Smekal-Raman-Effekt, in Struk- Structure, in Chemical Applications of Spectroscopy, ed.
tur und Eigenschaften der Materie, Vols XII and XIX, W. West, Interscience, New York, 247580, Vol. IX (1956).
Springer Verlag, Berlin (1931 and 1938). 52. K. Nakamoto, Infrared and Raman Spectra of Inorganic and
23. K.W.F. Kohlrausch, Ramanspektren, in Hand- und Jahr- Coordination Compounds, Wiley/Interscience, New York,
buch der Chemischen Physik, Vol. 9, Part VI, Akad. Ver- 1st edition (1963) and 4th edition (1986).
lagsgesellschaft, Leipzig (1943). 53. H.A. Willis, J.H. Van Der Maas and R.G. Miller, Labora-
24. W. Brugel, Einfuhrung in die Ultrarotspektroskopie, Dr. tory Methods in Vibrational Spectroscopy, John Wiley &
Dietrich Steinkopff Verlag, Darmstadt (1962). Sons, Chichester (1987).
25. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infra- 54. M.J.E. Golay, Rev. Sci. Instrum., 18, 347, 357 (1947).
red Spectroscopy, in Chemical Analysis, eds P.J. Elving, 55. L.T. Cowley, D.S. Horne and J.C. Polanyi, Chem. Phys.
J.D. Winefordner and I.M. Kolthoff, John Wiley & Sons, Lett., 12, 144 (1971).
New York, Vol. 83 (1986).
56. E. Whittle, D.A. Dows and G.C. Pimentel, J. Chem. Phys.,
26. F.A. Miller, Anal. Chem., 64, 825A (1992). 22, 1943 (1954).
27. P.A. Wilks, Jr, Anal. Chem., 64, 833A (1992). 57. M.L. Veingerov, Zavodskaya Lab., 13, 426 (1947).
28. P.R. Griffiths, Anal. Chem., 64, 868A (1992). 58. P.V. Slobodskaya, Isvest. Akad. Nauk SSSR, Ser. Fiz., 12,
29. N. Sheppard, Anal. Chem., 64, 877A (1992). 656 (1948).
30. A.O. Beckman, Anal. Chem., 49, 280A (1977). 59. A.M. Bradshaw and F.M. Hoffman, Surface Sci., 72, 513
(1978).
31. C.S. Creaser and A.M.C. Davies (eds), Analytical Applica-
tions of Spectroscopy, Royal Society of Chemistry, London 60. M.J. Dignam, B. Rao, M. Moskovits and R.B. Stobie, Can.
(1988). J. Chem., 49, 1115 (1971).
32. A.M.C. Davies, NIR News, 2(2), 12 (1991). 61. T.S. Robinson and W.C. Price, Proc. Phys. Soc. London,
B65, 910 (1952); B66, 969 (1953).
33. K.B. Whetsel, NIR News, 2(3), 4 (1991).
62. S.A. Francis and A.H. Ellison, J. Opt. Soc. Am., 49, 131
34. R. Miller, NIR News, 2(4), 12 (1991).
(1959).
35. K.B. Whetsel, NIR News, 2(5), 12 (1991).
63. H.A. Pearce and N. Sheppard, Surface Sci., 59, 205 (1976).
36. K.H. Norris, NIR News, 3(1), 12 (1992).
64. M.A. Chesters, J. Electron Spectrosc. Relat. Phenom., 58,
37. P.C. Williams, NIR News, 3(2), 3 (1992). 123 (1986).
38. K.H. Norris, NIR News, 4(5), 10 (1993). 65. X.D. Zhu, H. Suhr and Y.R. Shen, Phys. Rev. B, 35, 3047
39. J. Brandmuller and H. Moser, Einfuhrung in die Raman- (1987).
spektroskopie, Dr. Dietrich Steinkopff Verlag, Darmstadt 66. J.H. Hunt, P. Guyot-Sionnest and Y.R. Shen, Chem. Phys.
(1962). Lett., 133, 189 (1987).
40. T.R. Gilson and P.J. Hendra, Laser Raman Spectroscopy, 67. P.S. Cremer and G.A. Somorjai, J. Chem. Soc., Faraday
Wiley-Interscience, London (1970). Trans., 91, 3671 (1995).
41. P. Hendra, C. Jones and G. Warnes, Fourier Transform 68. A.N. Terenin, Zh. Fiz. Khim., 14, 1362 (1940).
Raman Spectroscopy, Ellis-Horwood, New York (1991).
69. A.N. Sidorov, Dokl. Akad. Nauk SSSR, 95, 1235 (1954).
42. D.A. Long, Raman Spectroscopy, McGraw Hill, New York
70. D.J.C. Yates, N. Sheppard and C.L. Angell, J. Chem. Phys.,
(1977).
23, 1980 (1955).
43. C.E. Hathaway, Raman Instrumentation and Techniques,
71. N. Sheppard and D.J.C. Yates, Proc. R. Soc. London A, 238,
in The Raman Effect, ed. A. Anderson, Marcel Dekker
69 (1956).
Inc., New York, 183285, Vol. 1 (1971).
72. R.P. Eischens, W.A. Pliskin and S.A. Francis, J. Chem.
44. E. Lehrer, Zeit. technische Physik, 23, 169 (1942). Phys., 22, 1786 (1954).
45. P.B. Fellgett, PhD thesis, University of Cambridge (1949); 73. D.H. Rank, R.J. Pfister and P.D. Coleman, J. Opt. Soc. Am.,
J. Phys. Radium, 19, 187 (1958). 32, 390 (1942).
46. P. Jacquinot, J. Phys. Radium, 19, 223 (1958). 74. D.H. Rank, R.J. Pfister and H.H. Grimm, J. Opt. Soc. Am.,
47. P. Connes, J. Phys. Radium, 19, 197, 215 (1958). 33, 31 (1943).
48. A.M.C. Davies, NIR News, 10(6), 14 (1999). 75. G.W. Chantry, H.A. Gebbie and C. Hilsum, Nature, 203,
49. N.B. Colthup, J. Opt. Soc. Am., 40, 397 (1950). 1052 (1964).
50. L.J. Bellamy, Infrared Spectra of Polyatomic Molecules, 76. D.B. Chase and T. Hirschfeld, Appl. Spectrosc., 40, 133
1st edition, Methuen, London (1954); 3rd edition, Chapman (1986).
and Hall, London (1975). 77. V.M. Hallmark, C.G. Zimba, J.D. Swalen and J.F. Rabolt,
51. R.N. Jones and C. Sandorfy, The Applications of Infrared Spectroscopy, 2, 40 (1987).
and Raman Spectrometry to the Elucidation of Molecular 78. P.W. Atkins and L.D. Barron, Mol. Phys., 16, 453 (1969).
79. L.D. Barron and A.D. Buckingham, Mol. Phys., 20, 1111 98. J.P. Toennies, Appl. Phys., 3, 91 (1974).
(1971). 99. R.J.H. Clark and R.E. Hester (eds), Advances in
80. L.B. Barron, M.P. Bogaard and A.D. Buckingham, Nature, Spectroscopy, John Wiley & Sons, Chichester (1972,
241, 113 (1973). onwards).
81. L.B. Barron, M.P. Bogaard and A.D. Buckingham, J. Am. 100. J.R. Durig (ed.), Vibrational Spectra and Structure, Else-
Chem. Soc., 95, 603 (1973). vier, Amsterdam (1972, onwards).
82. T. Takenaka and H. Fukuzaki, J. Raman Spectrosc., 8, 151 101. J.T. Edsall, J. Chem. Phys., 4, 1 (1936).
(1979). 102. S. Darmon and G.B.B.M. Sutherland, J. Am. Chem. Soc., 69,
83. M. Fleischmann, P.J. Hendra and A.J. McQuillan, Chem. 2047 (1947).
Phys. Lett., 26, 163 (1974). 103. A. Elliott and E.J. Ambrose, Discuss. Faraday Soc., 9, 246
84. D.L. Jeanmaire and R.P. Van Dyne, J. Electroanal. Chem., (1950).
84, 1 (1977). 104. K. Dobriner, E.R. Katzenellenbogen and R.N. Jones, Infra-
85. M.G. Albrecht and J.A. Creighton, J. Am. Chem. Soc., 99, red Absorption Spectra of Steroids, Interscience, New York
5215 (1977). (1953).
86. J.A. Creighton, C.G. Blatchford and M.G. Albrecht, J. 105. M. Born and Th. Von Karman, Phys. Zeit., 13, 297
Chem. Soc., Faraday Trans. II, 75, 790 (1979). (1912).
87. R.F. Willis (ed.), Vibrational Spectroscopy of Adsorbates, 106. A.M. Buswell and W.D. Rodebush, J. Chem. Phys., 6, 296
Springer-Verlag, Berlin (1980). (1938).
88. A.T. Bell and M.L. Hair (eds), Vibrational Spectroscopies 107. R. Freymann, Ann. Phys., 20, 243 (1933).
of Adsorbed Species, ACS Symposium Series 137, Amer- 108. R. Freymann, Spectre Infrarouge et Structure Moleculaire,
ican Chemical Society, Washington, DC (1980). Gauthier-Villars, Paris (1947).
89. J.T. Yates, Jr and T.E. Madey (eds), Vibrational Spec- 109. C. Shaeffer, C. Bormuth and F. Matossi, Zeit. Phys., 39, 648
troscopy of Molecules on Surfaces, Plenum Press, New (1926).
York (1987).
110. C. Duval and J. Lecomte, Compt. Rend., 215, 525 (1942).
90. H. Ibach and D.L. Mills, Electron Energy Loss Spec-
111. S. Mizushima, Y. Morino and S. Noziri, Sci. Pap., Inst.
troscopy, Academic Press, New York (1982).
Phys. Chem. Res. Tokyo, 29, 63, 188 (1936).
91. F.M. Propst and T.C. Piper, J. Vac. Sci. Technol., 4, 53
112. S. Mizushima and Y. Morino, Proc. Ind. Acad. Sci., 8, 315
(1967).
(1938).
92. W. Ho, R.F. Willis and E.W. Plummer, Phys. Rev. Lett., 40,
113. D.H. Rank, N. Sheppard and G.J. Szasz, J. Chem. Phys., 17,
1463 (1978).
83 (1949).
93. R.F. Willis, Surf. Sci., 89, 457 (1979).
114. J.K. Brown, N. Sheppard and D.M. Simpson, Phil. Trans.
94. R.C. Jaklevic and J. Lambe, Phys. Rev. Lett., 17, 1139 Roy. Soc. London, 247, 5 (1954).
(1966).
115. J.H. Schachtschneider and R.G. Snyder, Spectrochim. Acta,
95. B.C. Stripe, M.A. Rezeal and W. Ho, Science, 280, 1732 19, 117 (1967).
(1998). 116. E.B. Wilson, Jr, J.C. Decius and P.C. Cross, Molecular
96. B.C. Stripe, M.A. Rezeal and W. Ho, Phys. Rev. Lett., 81, Vibrations, McGraw-Hill, New York (1955).
1263; 82, 1734 (1999). 117. M.W. Wolkenstein, M.A. Elyashevich and B.I. Stepanov,
97. A.J. Renouprez, F. Fouilloux, C. Cordurier, D. Tocchetti Kolebanya Molekul (Molecular Vibrations), State Pub-
and R. Stockmeyer, J. Chem. Soc., Faraday Trans. 1, 73, 1 lishers of Technical-Theoretical Literature, Moscow and
(1977). Leningrad (1949).
Introduction to Vibrational Spectroscopy
Peter R. Griffiths
University of Idaho, Moscow, ID, USA
1 INTRODUCTION then ever before. The reason is, of course, largely related
to the development of Fourier transform infrared (FT-IR)
Most chemists receive their first introduction to vibrational spectrometers, which have revolutionized the measurement
spectroscopy by using infrared (IR) spectra to elucidate of IR spectra and allowed IR spectra to be used for much
the structure of organic chemicals. They quickly learn, more than structural elucidation. Let us first consider why
for example, that the CDO stretching vibration of alkyl FT-IR spectrometers have assumed such a position of
esters absorbs at 1735 cm1 or that the out-of-plane CH dominance for the measurement of IR spectra.
bending mode of para-substituted aromatics absorbs at FT-IR survey spectra are often measured at a resolution
810 cm1 . In many large research laboratories, however, of 4 cm1 . When mid-infrared (MIR) spectra between
the structure of complex molecules is now usually found 4000 and 400 cm1 were measured with a prism or grating
by two-dimensional nuclear magnetic resonance (NMR) monochromator, only one 4-cm1 resolution element in the
spectrometry and not by IR. In laboratories where funds 3600 cm1 wide spectral range of interest was measured at
are available to purchase all the necessary equipment for any instant while the remaining 3596 cm1 were not. Thus
structural elucidation of organic molecules, it is probable the efficiency of the measurement is only about 0.1%. It was
that NMR spectrometry, X-ray diffraction and mass spec- typical for survey scans to take several minutes to measure
trometry will take precedence over IR spectrometry for this while the measurement of archival spectra (measured at
12 cm1 resolution) often took at least 30 min.
purpose. Only in polymer laboratories, or laboratories with
In FT-IR spectrometry, all the resolution elements are
a fairly limited budget, is IR spectrometry still used as the
measured at all times during the measurement (the multiplex
primary tool for determining molecular structure.
or Fellgetts advantage). In addition, more radiation can be
As long ago as 1973, Herbert Laitinen, who was the
passed between the source and the detector for each resolu-
editor of Analytical Chemistry at the time, made an anal-
tion element (the throughput or Jacquinot advantage). As
ogy with Shakespeares seven ages of man to define the
a result, transmission, reflection and even emission spec-
seven ages of an analytical instrument.1 These seven ages
tra can be measured significantly faster and with higher
start with the discovery of the principles of the technique
sensitivity than before.
in a physicists laboratory. After characterization and com-
In this Handbook, authors with a wide range of interests
mercialization of the technique, instruments graduate from
will demonstrate how vibrational spectrometry can not only
the research laboratory to routine analytical measurements
be used to measure IR spectra of the type of samples that
until, in their seventh age they are superseded by newer
have been classically investigated by IR spectrometers for
instruments with higher speed, sensitivity, specificity or res-
decades, i.e. gases, liquids, and bulk and powdered solids in
olution. In his 1973 editorial, Professor Laitinen used IR
milligram quantities, but that interfacial species, microsam-
spectrometry to illustrate an instrument in its seventh age.
ples and trace analytes can now be characterized routinely.
With these facts in mind, readers may be surprised to Measurement times have been reduced from minutes to
learn that IR spectrometers are in use in more laboratories fractions of a second; in special cases, reactions taking place
in less than a microsecond can be followed. The physical
John Wiley & Sons Ltd, 2002. properties of materials can be correlated to the molecular
structure by vibrational spectroscopy better than by any instruments. The remainder of this article describes the the-
other analytical technique. Indeed, it is probably true to oretical basis of many of these measurements.
say that, during the 25 years following Laitinens editorial,
IR spectroscopy has entered and passed through its second
childhood. Because of the remarkable advances made in 2 MOLECULAR VIBRATIONS
the performance of FT-IR spectrometers, IR spectrometry
IR and Raman spectra result from transitions between
has matured to the point that it is used for the solution of
quantized vibrational energy states. Molecular vibrations
a variety of problems from the research laboratory to the
can range from the simple coupled motion of the two
manufacturing floor, and sales of IR spectrometers are at
atoms of a diatomic molecule to the much more complex
an all time high.
motion of every atom in a large polyfunctional molecule.
A similar situation exists with Raman spectroscopy. Up
Molecules with N atoms have 3N degrees of freedom,
to the late 1960s, Raman spectra were typically excited
three of which represent translational motion in mutually
by a line from a mercury arc source and measured with a
perpendicular directions (the x-, y- and z-axes) and three
photomultiplier tube (PMT). With the invention of the laser,
represent rotational motion about the x-, y- and z-axes. The
the mercury arc was rapidly superseded by heliumneon
remaining 3N 6 degrees of freedom give the number of
and argon ion lasers. With laser excitation, Raman spectra
ways that the atoms in the molecule can vibrate, i.e. the
of acceptable signal-to-noise ratio could be measured in less
number of vibrational modes.
than 30 min, provided that the sample did not fluoresce.
Each mode involves approximately harmonic displace-
However, most samples encountered in typical industrial
ments of the atoms from their equilibrium positions; for
analytical laboratories do fluoresce to a greater or lesser
each mode, i, all the atoms vibrate at a certain characteris-
extent when illuminated with an intense source of visible tic frequency, ni . The potential energy, V r, of a harmonic
radiation. Thus, despite the complementary nature of IR oscillator is shown as a function of the distance between
and Raman spectra, Raman spectroscopy was rarely used the atoms, r, as the broken line in Figure 1. For any mode
outside of academic laboratories. in which the atoms vibrate with simple harmonic motion,
This situation changed dramatically in 1986 when, in i.e. obeying Hookes law, the vibrational energy states, Vin ,
back-to-back papers in Applied Spectroscopy, Hirschfeld can be described by equation (1):
and Chase2 and Fujiwara et al.3 reported that the 1064-nm
(9400-cm1 ) line of a neodymium-doped yttrium alu- Vin D hni vi C 12 1
minum garnet (Nd : YAG) laser could be used to excite
the Raman spectrum. With a wavelength this long, fluo- where h is Plancks constant, ni is the fundamental fre-
rescence of most samples is minimal. When this laser line quency of the particular mode and vi is the vibrational
is used, the Raman spectrum is found between about 5600 quantum number of the ith mode (vi D 0, 1, 2, . . .). Note
and 9500 cm1 (3800 100 cm1 Raman shift). In this
case, the spectrum could no longer be measured by a PMT,
Harmonic potential
and both groups1,2 used a liquid-nitrogen-cooled germa- (Hooke's law)
nium detector to acquire the spectral data. Fujiwara et al.
obtained their spectra with a high-power pulsed laser and
scanning monochromator, while Chase and Hirschfeld used 0
a continuous-wave laser and a Fourier transform (FT) spec-

trometer, with the latter technique ultimately proving to be Anharmonic
V(r)
the more popular. Only a couple of years later, high-quality potential

near-infrared (NIR) Raman spectra were measured using a (Morse type)
780-nm diode laser and a polychromator equipped with a

silicon-based charge coupled device (CCD) array detector
=2
in measurement times as short as 1 s. Today, both CCD and =1
FT spectrometers are used for the routine measurement of =0
the Raman spectra of most compounds.
As a result of the developments described above, vibra-
tional spectroscopists have access to a wide range of instru- r
ments that are fast, sensitive, easily operated and relatively Figure 1. Potential energy of a diatomic molecule as a function
inexpensive. This Handbook describes many of the measure- of the atomic displacement during a vibration for a harmonic
ments that can be made by these, and other more specialized, oscillator (broken line) and an anharmonic oscillator (solid line).
Introduction to Vibrational Spectroscopy 3
that frequency in units of hertz is usually given the sym- centered and is minimally affected by the nature of the
bol, n. Vibrational frequencies are often given in units of other atoms in the molecule. Thus the observation of
wavenumber, or the number of waves per unit length. The spectral features in a certain region of the spectrum is
most common unit of length is the centimeter, in which case often indicative of a specific chemical functional group in
wavenumber has units of cm1 and is given the symbol, nQ . the molecule. Extensive spectra/structure correlation tables
The energy difference for transitions between the ground (often known as Colthup charts) have been developed to
state (vi D 0) and the first excited state (vi D 1) of most allow chemists to assign one or more absorption bands in
vibrational modes corresponds to the energy of radiation in a given IR spectrum to the vibrational mode(s) associated
the MIR spectrum (4004000 cm1 ). with a certain functional group. These tables may be found
The motion of the atoms during the vibration is usu- in many textbooks on the interpretation of IR spectra.
ally described in terms of the normal coordinate, Q. The The correlation of MIR spectral bands with molecular
molecule is only promoted to the excited state if its dipole structure is described by Shurvell (see SpectraStructure
moment, m, changes during the vibration, i.e. provided that Correlations in the Mid- and Far-infrared).4
@m/@Q 6D 0. For molecules with certain elements of sym- Not all bands are useful for identifying the presence of
metry, some vibrational modes may be degenerate, so that functional groups in the structure of organic molecules.
more than one mode has a given vibrational frequency Other bands involve significant motion of several atoms
while others may be completely forbidden. Thus because and so their frequency varies from one molecule to another
of degeneracy, the number of fundamental absorption bands containing the particular functional group. These modes are
able to be observed is often less than 3N 6. Because rota- useful for distinguishing one molecule from another con-
tion of a linear molecule about the axis of the bond does taining similar functional groups and hence are often known
not involve the displacement of any of the atoms, one of the as fingerprint bands.
rotational degrees of freedom is lost and linear molecules One class of fingerprint bands involves skeletal modes
have an additional vibrational mode. Thus the number of which involve significant displacements of many of the
modes of a linear molecules is 3N 5, so that a diatomic atoms in the molecule. These bands are rarely used to
molecule (N D 2) has a single vibrational mode. indicate the presence or absence of a specific functional
The actual variation of the potential energy as a function group but again may be useful for distinguishing between
of the displacement of the atoms from their equilibrium structurally similar compounds. The vibrational frequency
positions is shown as the solid line in Figure 1. From this of skeletal modes is usually quite low. As a result they
curve it can be seen that equation (1) is only valid for low absorb long wavelength radiation that is often below the
values of the vibrational quantum number, and is not valid cutoff of many MIR detectors. The far-infrared (FIR) region
when vi is large. In practice, Vin must be described using an of the spectrum (10400 cm1 ) is rarely used for structural
anharmonic (Morse-type) potential function. This behavior elucidation, but contains useful information on the vibration
is shown on Figure 1 as the solid line, and the potential of heavy atoms (especially for inorganic compounds) and/or
energy is given to a first approximation by equation (2): weak bonds such as hydrogen bonds. A summary of FIR
2 spectroscopy appears later in this volume (see Far-infrared
Vin D hni vi C 12 C hni xi vi C 12 2
Spectroscopy).5
where xi is the anharmonicity constant. xi is dimension- Every molecule has slightly different vibrational modes
less and typically has values between 0.001 and 0.02 from all other molecules (with the exception of enan-
depending on the mode. tiomers). Thus the vibrational spectrum of a given molecule
If all vibrational modes were strictly harmonic, no tran- is unique and can be used to identify that molecule. IR
sitions involving changes in vi by more than 1 would be and Raman spectra give far more detailed information than
allowed. The effect of anharmonicity is to relax this selec- simply allowing the presence or absence of certain func-
tion rule, i.e. to allow bands caused by jvi j > 1 to become tional groups to be recognized. In the past, many chemists
allowed. Thus overtone (vi D 2, 3, . . .) and combination had a solid understanding of how to interpret IR spectra,
vi D 1, 2, . . . ; vj D 1, 2, . . ., where j represents a dif- but spectral interpretation is starting to become a lost art,
ferent mode, bands commonly appear weakly in the MIR in part because of the emergence of NMR, mass spec-
spectrum of organic compounds along with the bands due trometry and X-ray diffraction for this purpose. Although
to fundamental transitions (vi D 1). molecular orbital programs are starting to permit vibrational
For many vibrational modes, only a few atoms have spectra of quite complex molecules to be calculated, per-
large displacements and the rest of the molecule is almost fect matching of calculated and measured spectra has yet
stationary. The frequency of such modes is characteristic to be achieved. Today, computer-assisted comparison of
of the specific functional group in which the motion is the spectra of unknowns to a large number of reference
spectra in a database (spectral, or library, searching) has 3 RAMAN SPECTROMETRY

become a far more popular way of finding the structure
of a molecule from its vibrational spectrum than manual Raman spectrometry is an alternative, and often comple-
interpretation. Because of the subtle differences between mentary, way of measuring vibrational spectra to IR spec-
the spectra of many compounds, the result of a computer- trometry. In Raman spectrometry, the sample is illuminated
ized spectral search should never be assumed to give the with a monochromatic beam of radiation of wavenumber
true identity of the compound without the operator visu- n0 (typically from some type of laser). All of the photons
ally comparing the best match and the actual spectrum of that interact with the sample cause the potential energy of
the unknown. It should also be noted that the spectrum of a molecule to be raised to a virtual state, hn0 above the
a given molecule depends on its physical state and will ground state. Almost immediately most molecules return to
change depending on its crystalline state, conformation, the ground state through the emission of a photon of the
configuration and method of sample preparation. Thus, for same wavelength as that of the incident photon. Since the
example, only if a polymer sample has the same history energy of the incident and scattered photons are identical,
and has been prepared in the same way as the reference this process is a form of elastic scattering and is often
material can a spectroscopist be certain that the identi- to as Rayleigh scattering. A small fraction of the incident
fication given by a spectral searching program is indeed photons drop back to the first excited vibrational state of
correct. the ith vibrational mode of the molecule, so that the energy
The intensity of fundamental bands in IR spectra is proof the scattered photon is h n0 ni , i.e. it will be observed
portional to dm/dQ2 . For most pure compounds, a sample at a wavenumber of (nQ 0 nQ i ) where nQ i is the wavenumber
thickness of only about 10 m is needed to yield a MIR of the ith vibrational mode. For Raman spectra, the abscissa
spectrum for which the bands are neither saturated (max- is best referred to as wavenumber shift, cm1 . Since the
imum transmittance less than 1%) nor so weak that they energy of the incident and scattered photons are different,
require ordinate expansion. It is often inconvenient and the scattering is inelastic. This process is known as Stokes
sometimes impossible to prepare such thin samples. In these Raman scattering. Since a given molecule has many differ-
cases, measurement of the near-infrared (NIR) spectrum ent vibrational modes, measurement of the spectrum from
may prove to be advantageous. Overtone and combina- nQ 0 to nQ 0 4000 cm1 allows all Raman-active vibrational
tion bands are usually much weaker than the fundamental bands to be measured.
modes from which they are derived. The only exception Molecules that are in an excited vibrational state will
to this rule is when these bands are enhanced by Fermi undergo analogous effects when illuminated by a laser.
resonance, which occurs when an overtone or combina- Again, most of these molecules will undergo Rayleigh
tion band absorbs at approximately the same frequency of scattering, but a small number will return from the virtual
a fundamental mode involving the same atoms. Although state to the ground vibrational state. In this case, which is
many overtone and combination bands absorb in the MIR known as anti-Stokes scattering, radiation will be observed
region, the first and second overtones (vi D 2 and 3) of at a wavenumber of (nQ 0 C nQ i ). In practice, anti-Stokes-
CH, OH and NH stretching vibrations are found above shifted Raman bands are only seen for bands where nQ i is
4000 cm1 , i.e. in the NIR region. In the last two decades, low, usually below 1000 cm1 , because only for these bands
NIR spectrometry has become of tremendous importance, in is the Boltzmann population of the excited vibrational state
large part because of the very weakness of these bands. For populated to greater than 1% at ambient temperature.
samples that are between about 0.1 and 5 mm in thickness, The intensity of bands in the Raman spectrum of a com-
the NIR spectra are often much more appropriate for quan- pound is governed by the change in polarizability, a, that
titative, and sometimes even qualitative, analysis than the occurs during the vibration. The intensity of any band in
corresponding MIR spectra of these samples. Furthermore, the Raman spectrum is given by equation (3):
samples do not have to be mounted in salt cells, sources 2
da
are more intense, and NIR detectors are more sensitive than IRaman D KIL nQ 0 nQ i 4 3
dQ
MIR detectors. Because the bands are derived from just
a few functional groups, NIR spectra are not as easy to where IL is the power of the laser at the sample, (nQ 0 nQ i )
interpret as MIR spectra, but they are very amenable to mul- is the wavenumber at which the band is measured and
tivariate statistical analysis of the type that is now becoming (da/dQ) is the change in polarizability with the normal
common throughout analytical chemistry. Indeed, many of coordinate of the vibration. The value of the constant
these algorithms were originally developed specifically for of proportionality, K, is dependent on the efficiency at
the analysis of NIR spectra. These algorithms are described which Raman-scattered light may be collected, as well as
in Volume 3 of this Handbook. some other factors. A detailed summary of the theory of
Raman spectroscopy by Keresztury may be found later measurement made with a NIR or visible laser of the same
in this volume of the Handbook6 and a summary of the power. Finally, UV detectors are significantly more sensi-
characteristic group frequencies for Raman spectra is to be tive than NIR detectors. Surprisingly, UV-Raman spectrom-
found in the article by Edwards (see SpectraStructure etry is not very popular today, but there is good reason to
Correlations in Raman Spectroscopy) in Volume 3.7 expect that this state-of-affairs will soon change, as can be
Before the mid-1980s, Raman spectroscopy was often seen in the article by Asher later in this volume.8
considered to be less desirable than IR absorption spec- There are many reasons why scientists want to mea-
troscopy for two reasons. Firstly, only one in about 108 sure the Raman spectra of compounds. Firstly, many bands
of the incident photons undergoes Raman scattering. Thus that are weak in the IR spectrum are among the strongest
until about 1985, Raman spectrometry was considered to be bands in the Raman spectrum. For example, the SS and
a relatively insensitive and/or time-consuming technique. CDC stretching bands are often so weak as to be essen-
However, the advent of multiplex (usually FT) and multi- tially unrecognizable in the IR spectrum but stick out like
channel (monochromators with array detectors) techniques the proverbial sore thumb in a Raman spectrum. Secondly,
have greatly increased the sensitivity of modern Raman some Raman bands are found at very characteristic frequen-
spectrometry. Of greater difficulty today is the fact that cies. For instance, monosubstituted aromatic compounds,
many compounds fluoresce when illuminated by visible together with 1,3-disubstituted and 1,3,5-trisubstituted aro-
lasers, such as the radiation from an ArC laser at 488 nm. matics, have a very intense band at 1000 cm1 . Samples
For some molecules the quantum efficiency for fluores- for Raman spectrometry can be mounted in standard glass
cence with 488-nm radiation can approach 100% and it tubes, making sample handling far easier for Raman than
is often greater than 0.001%. Even in the latter case, the for IR spectrometry. Finally, low-frequency bands are far
radiation emitted by luminescence far exceeds the intensity more easily measured by Raman spectrometry than by IR.
of Raman-scattered radiation. Similarly, fluorescence from For FIR spectrometry, the source, beamsplitter and detector
a trace molecule with a quantum efficiency approaching are different from the corresponding components in a MIR
100% can also swamp the Raman signal. spectrometer. On the other hand, most Raman spectrome-
To ameliorate the problem of fluorescence, NIR or ultra- ters readily measure the spectrum down to 100 cm1 , and
violet (UV) lasers can be used to illuminate the sample. By often lower.
using NIR radiation, the wavelengths used to excite Raman For these reasons, analytical Raman spectrometry has
spectra are long enough that most compounds no longer flu- undergone a remarkable rebirth since about 1985. Instru-
oresce. The most popular NIR laser is the Nd : YAG laser ments based on FT techniques and CCD array detectors
emitting at 1064 nm (nQ 0 9400 cm1 ). With this laser, the are commercially available from a large number of vendors
problem changes from fluorescence to detection. Consider in North America, Europe and Asia. Like IR spectrometry,
a Raman band due to a CH stretching mode at 2950 cm1 Raman spectrometry has outlived its seventh age, and prob-
shift. This band must be measured at 94002950 cm1 , ably may be considered to be in its fourth or fifth age after
i.e. 6450 cm1 , which is well below the wavenumber at its second childhood. Certainly it can be said that Raman
which silicon CCD array detectors cut off. Detectors oper- spectrometry should now be a vital weapon in every vibra-
ating in this region of the spectrum are significantly less tional spectroscopists arsenal.
sensitive than CCDs. Furthermore, it can be seen from
equation (3) that IRaman is proportional to nQ 0 nQ i 4 . Thus
a Raman band at 2950-cm1 measured with a 1064-nm 4 VIBRATIONROTATION
Nd : YAG laser would be about 55 times weaker than the SPECTROSCOPY
same band measured with the 488-nm line of an ArC laser
of equal power. Since CCD array detectors do not respond One of the greatest strengths of IR and Raman spectrome-
at 6450 cm1 , it is not surprising that FT techniques have try is their capability of allowing samples in all phases of
been invoked for the measurement of weak Raman signals matter to be studied. Vibrational spectra of gases, liquids
at such long wavelengths. and solids have different characteristics, and it is essential
The use of UV lasers to minimize fluorescence has also that these differences are understood if spectra of materials
been shown to be feasible by a number of workers. In in each state are to be measured optimally and to yield the
this case, advantage is taken of the fact that many spec- greatest amount of information. For example, the spectra of
tra that fluoresce when illuminated by visible radiation do small molecules in the vapor phase show considerable fine
not fluoresce when illuminate by a UV laser. Since IRaman structure because transitions between quantized rotational
is proportional to nQ 0 nQ i 4 , UV-Raman measurements energy levels occur at the same time as vibrational transi-
have a greater intrinsic sensitivity than the corresponding tions. Similar features are rarely seen in the spectra of larger
molecules in the vapor phase (because the individual rota- the transition J D 0 is allowed. The line corresponding to
tional transitions are too close together to be resolved) or a transition with J D 0 is known as the Q branch. Under
any molecule in the liquid state (because collisions occur at low resolution, only a single line can be seen, but at very
a greater rate than the rotational frequency). The full theory high resolution, it is often possible to resolve many lines
of rotationvibration spectroscopy is quite complex, and a in the Q branch corresponding to vi D 1 and J D 0, for
detailed exposition of this subject is beyond the scope of molecules in each rotational state.
this introductory article, but a brief introduction to rota- In practice, molecules are not rigid rotors and centrifugal
tionvibration spectroscopy is given below. forces cause the length of the bond between X and Y to
The simplest vibrationrotation spectra to interpret are increase as the angular velocity of the rotating molecule
those of diatomic molecules. The rotational energy levels of increases. The effect of centrifugal distortion is to increase
diatomic molecules are characterized by a single rotational the moment of inertia, increasing the rotational constant, B,
quantum number, J. If the molecule is assumed to be a rigid at high J. To a first approximation, the effect of centrifugal
rotor, the rotational energy is given by equation (4): distortion is taken care of by adding a second term to
equation (4) to yield equation (9):
EJ D BJ J C 1 4
EJ D BJ J C 1 DJ2 J C 12 9
where B is the so-called rotational constant, given by
equation (5): where D is the centrifugal distortion constant. Usually,
h 0.1 < B < 10 cm1 and D 104 cm1 . Because of the
BD 2 5
8p Ic effect of centrifugal distortion, the spacing of the lines in
where I is the moment of inertia of the molecule and c is the the P branch increases as the distance from nQ 0 increases
velocity of light. The selection rule for transitions between while that of the lines in the R branch decreases.
rotational energy states for linear molecules is given by As molecules become larger and less symmetrical, the
equation (6): spacing of the lines in their vibrationrotation spectra
J D 1 6 decreases and the spectra become more complex. For a
linear molecule such as CO2 , the selection rules for each
so that [equation (7)]: mode depend on the symmetry of the vibration. The sym-
EJ EJ1 D BJ J C 1 BJ J 1 D 2BJ 7 metric stretching mode of CO2 (n1 ) is forbidden in the
IR, since dm/dQ D 0. It is, however, allowed in the
Thus for a rigid rotor the pure rotation spectrum would Raman spectrum. For the antisymmetric stretching mode
be composed of a series of lines with equal spacing of (n3 ), in which the molecule remains linear throughout the
2B cm1 . For most molecules, B is sufficiently small that vibration, the selection rules are the same as for diatomic
the pure rotation spectrum is found in the microwave molecules and hence the IR spectrum has a similar appear-
region of the spectrum; however, for light molecules such ance, (Figure 2a). For the (degenerate) bending mode (n2 ),
as HCl or H2 O, rotational transitions are found as high a vibrational transition without a simultaneous change in J
as 400 cm1 . is permitted, i.e. [equation (10)]
Diatomic molecules, XY, have a single fundamental
vibrational mode, of wavenumber n0 , which is only active in v2 D 1 and J D 0, 1 10
the IR if X 6D Y. (If X D Y, the transition is Raman active.) and a sharp feature corresponding to the sum of all the
For any IR-allowed vibrational transition of a gaseous transitions where J D 0 can be seen at the band center
diatomic molecule, there must be a simultaneous rotational (Figure 2b). The reason that the selection rules are different
transition, i.e. [equation (8)] for these two modes is because different symmetry elements
vi D 1 and J D 1 8 of the linear CO2 molecule are lost during these two
vibrations.
Thus the vibrationrotation spectrum of a rigid diatomic Linear molecules have two equal principal moments
molecule consists of a series of equally spaced lines above of inertia, corresponding to rotation about the center of
and below nQ 0 , corresponding to J D C1 and J D 1, mass about two mutually perpendicular axes, with the
respectively. The series of lines below nQ 0 (J D 1) is third principal moment equal to zero. Nonlinear molecules
known as the P branch of the band, while the series of lines usually have three different moments of inertia. In this case,
above nQ 0 J D C1 is known as the R branch. Because the vibrationrotation spectrum can be very complex, even
J 6D 0, there is no absorption line at nQ 0 . There is one for such a simple molecule as water. The rotational fine
exception to this rule. The fact that nitric oxide has an structure in the IR spectrum of the HOH bending mode
unpaired electron in its ground state leads to the fact that of water is shown in Figure 3.
100 The two molecules whose vibrationrotation spectra

are shown in Figures 2 and 3, CO2 and H2 O, are often
95 encountered as interferences when MIR spectra are mea-
sured (although the rotational lines in the spectrum of CO2
Percent transmission
90
are often unresolved when the spectrometer resolution is
85 4 cm1 or poorer). In fact, it is good practice to eliminate
all traces of these molecules in the beam path of an IR
80 spectrometer by purging the instrument with dry, CO2 -free
air or, better still, pure nitrogen gas as the bands seen in
75
Figures 2 and 3 will often be seen in the spectra. As noted
70 above, for molecules in the liquid state, collisions occur
at a greater rate than their rotational frequency. Thus no
65 rotational fine structure is seen in the vibrational spectra of
2380 2340 2300 2260 species in the liquid state.
(a) Wavenumber /cm1 Raman spectra of gases are less commonly measured than
IR spectra. Nonetheless, they can prove to be equally impor-
tant. For example, unlike the case for IR spectroscopy, the
100 Raman spectra of homonuclear diatomic molecules such
95
as oxygen and nitrogen are allowed. The theory of Raman
90
85 spectroscopy of gaseous molecules is given in the article in
80 this volume of the Handbook by Weber.9
75
70
65
60 5 WIDTHS OF BANDS AND LINES IN IR
55
50
SPECTRA
45
40 5.1 Vibrationrotation spectra of gases
35
720 700 680 660 640 620
The shape and width of individual lines in the vibra-
(b) Wavenumber /cm1 tionrotation spectrum of a gas depend on its pressure, P.
Figure 2. IR vibrationrotation fundamental bands of carbon For gases at low pressure (typically P < 1 Torr), the shape
dioxide: (a) antisymmetric stretching mode (n3 ) for which and width of each spectral line in the band are determined
the selection rule is v3 D 1 and J D 1; (b) bending by the Doppler effect, i.e. by the variation of the speed
mode (n2 ) for which the selection rule is v2 D 1 and J D of each molecule in the direction of the beam. The shape
0, 1.
of a Doppler-broadened line centered at wavenumber nQ 0 is
Gaussian, i.e. the absorbance at any wavenumber nQ is given
100 by equation (11):
95
nQ nQ o 2
90 A nQ D A0 exp 4 ln 2 11
g2D
85
80
where A0 is the absorbance at nQ o . gD is the full width at
75
70
half height (FWHH) of the band, given by equation (12):
65
kT 1/2 n
60 gD D 2 2 ln 2 12
55 m c
50 where k is Boltzmanns constant, T is the temperature
45
(Kelvin) and m is the mass of the molecule. Since m is equal
40
35
to the molecular weight, M, divided by Avogadros number
2000 1900 1800 1700 1600 1500 1400 1300 and frequency, n, is directly proportional to wavenumber,
Wavenumber /cm1 nQ , we have [equation (13)]:
1/2
Figure 3. Vibrationrotation spectrum of the HOH bending gD 7 T
mode of water vapor. D 7.16 10 13
nQ M
For a line in the HOH bending mode of water the IR spectra of liquids is Lorentzian, equation (16):
(M D 18 g mol1 ) at 1500 cm1 and the Doppler width
g2
at room temperature (298 K) is about 0.0044 cm1 . Thus A n D A0 16
an instrument with very high resolution is needed before g2 C 4 nQ nQ o 2
Doppler-broadened spectra can be measured accurately. where g is the FWHH of the band. In practice, the far wings
As the total pressure of the gas is raised above 1 Torr, of bands in the spectra of liquids decay faster than would be
the dominant mechanism of line broadening becomes more given by equation (16). To model the behavior of bands in
dominated by the effect of intermolecular collisions than the spectra of liquids, bands are sometimes expressed as the
by the Doppler effect. The shape of lines in collision- sum of Lorentzian and Gaussian bands as in equation (17):
broadened spectra is Lorentzian, equation (14):
g2L
A nQ D 1 aA0
g2C g2L C 4 nQ nQ o 2
A nQ D A0 14
g2C C 4 nQ nQ o 2 nQ nQ o 2
C aA0 exp 4 ln 2 17
where gC is the FWHH of the collision-broadened line. g2G
gC is directly proportional to the pressure of the gas and where gL is the FWHH of the Lorentzian component, gG
increases with the polarity of each component. There are is the FWHH of the Gaussian component (usually assumed
several mechanisms of collision broadening, each of which to be the same) and a is the Gaussian fraction. For many
leads to a slightly different variation of the broadening liquids, a < 0.1. Although complicated, this form is at least
coefficient with temperature, from T0.5 to T1 . For many analytical, unlike the Voigt profile. Even better descriptions
molecules in air at ambient temperature, the collision- of bands in the spectra of liquids can be made using other
broadening coefficient is approximately 0.1 cm1 atm1 . models, but a full description of these is beyond the scope of
For molecules at pressures between about 1 and 100 Torr, this article. These models have been summarized by Turner
the line width is determined both by Doppler and collision (see Infrared Spectroscopy: Theory) later in this volume
broadening. In this case, the shape is given by a convolu- of the Handbook.10
tion of the shapes given by equations (10) and (13); such a The actual width of absorption bands in the vibrational
shape is known as a Voigt profile. The higher the pressure of spectra of liquids and solutions depends strongly on the
the gas, the greater the contribution of collision broadening rigidity of the part of the molecule where the vibrational
to the Voigt profile. Voigt profiles cannot be expressed ana- motion of the molecule is localized. For example, bands
lytically, but the FWHH is given to a good approximation associated with aromatic groups often have widths between
by equation (15): 2.5 and 5 cm1 whereas the FWHH of bands associated

2 1/2 with the more flexible alkyl chains is often at least 10 cm1 .
gC gC Band widths also depend on the polarity of the solvent, with
gV D C C gD 2 15
2 2 bands of a particular solute in a polar solvent usually being
For large molecules, the spacing of lines in the vibra- broader than the corresponding bands of the same molecule
tionrotation spectrum is usually less than 0.1 cm1 . For in a nonpolar solvent. Since water is the most polar solvent,
such molecules in one atmosphere of air or some other bands of solutes in aqueous solution are usually broader
inert gas, the line spacing is less than gC , so that the lines than in any other solvent. When functional groups such
can never be separated, no matter how high the resolution as OH are hydrogen bonded, the width of the stretching
of the spectrometer and only the band contour can be mea- bands may be greater than 100 cm1 . Because the widths
sured. In this case, the higher the molecular weight of the of most absorption bands are rarely smaller than 4 cm1 ,
molecule, the narrower is the band contour. Nonetheless, it is common for survey spectra of liquids to be measured
for most molecules the FWHH of the band contour is at at a resolution of 4 or 8 cm1 whereas high-quality spectra
least 4 cm1 . For such molecules it is rarely necessary to for archival purposes are usually measured at a resolution
use a resolution much higher than 2 cm1 to measure the of 1 or 2 cm1 .
spectrum with minimal distortion. Few detailed investigations into the shapes and widths
of bands in the spectra of solids have been made but the
bands of solids are usually narrower than the corresponding
5.2 Condensed-phase samples bands of the same molecule in solution because of the
restricted motion of the functional groups in the molecules.
Since there is no rotational fine structure in the vibrational The shapes of MIR absorption bands of solid compounds
spectra of liquids, their spectra are much simpler than those are often represented by equation (17), with no limitation
of gases. To a good approximation, the shape of bands in as to the value of the Gaussian fraction, a.
Because bands in NIR spectra are overtones or com- measurement of IR spectra. On these instruments, the ratio
binations of fundamentals, the widths of these bands are of I nQ and I0 nQ was usually measured, so the usual out-
typically greater than the widths of bands from which they put of these instruments was a transmittance spectrum. In
are derived. For example, the FWHH of the first overtone the 1970s, transmittance spectra were produced in digital
of a CH stretching band is to a first approximation twice form, so they could be quickly converted to absorbance
that of the corresponding fundamental. As a result, most for quantitative applications. FT-IR spectrometers measure
NIR spectra of liquids are measured at significantly lower a signal known as an interferogram, from which a single-
resolution than the corresponding MIR spectrum. beam spectrum is calculated. The intensity of the single-
beam spectrum at any wavenumber nQ is proportional to the
power of the radiation reaching the detector. Thus to mea-
6 QUANTITATIVE CONSIDERATIONS sure the absorbance spectrum of a sample, the ratio of the
single-beam spectra of the sample and background is first
6.1 Beers law calculated (to yield the transmittance spectrum) which is
then converted to absorbance as shown in equation (19). It
The BouguerBeerLambert law (usually abbreviated to should always be remembered that as the first step of many
Beers law ) is the fundamental law of quantitative absorp- of the more popular operations in FT-IR spectrometry,
tion spectroscopy and is derived in all elementary textbooks including spectral subtraction, multicomponent analysis and
on instrumental analysis. The transmittance of any sample spectral searching, the measured spectrum should always be
at wavenumber nQ is given by the ratio of the radiant power converted to absorbance. Many of the practical parameters
emerging from the rear face of the sample at that wavenum- that determine whether Beers law is obeyed in practice are
ber I nQ to the power of the radiation at the front face of discussed in an article by this author in Volume 3 of this
the sample, I0 nQ . The transmittance of a pure sample of Handbook.11
thickness b (cm) at wavenumber nQ is given by Beers law Secondly, the effect of reflection loss at the windows
as equation (18): of the cell has been neglected in the above treatment. The
refractive index, n, of most organic samples and windows
I nQ
T nQ D D exp[a nQ b] 18 is about 1.5, so that the reflectance of the front surface
I0 nQ is about 4%. If Beers law is to be applied accurately,
where a nQ is the linear absorption coefficient (cm1 ) at nQ . the apparent absorbance caused by reflection loss (0.018
The absorbance of the sample at nQ , A nQ , is given by the absorbance units for windows with n D 1.5) should first be
base 10 logarithm of 1/T nQ [equation (19)]: subtracted from the measured absorbance spectrum.
It is also worthwhile to recognize that Raman spec-
1 1 troscopy is a technique that depends on inelastic light
A nQ D log10 D a nQ b 19
T nQ ln 10 scattering, not absorption. Thus Raman spectra do not
where 1/ ln 10 a nQ is the absorptivity at n, a nQ . need to be converted into a parameter that is equivalent
If the sample is a mixture, the absorbance of each com- to absorbance. Raman spectra are single-beam spectra and
ponent, i, at concentration, ci , is given by Beers law as must therefore be corrected for the instrument response
equation (20): function before they can be used for quantitative analysis on
any instrument other than the one on which the calibration
1 data were acquired.
log10 D Ai nQ D ai nQ bci 20
T nQ
with ai nQ having the units of (concentration pathlength)1 .
6.2 Optical constants
For N-component mixtures where more than one com-
ponent absorbs at nQ , the total absorbance is given by
The fundamental parameters that govern the absorption of
equation (21):
N radiation are the real and imaginary components of the
A nQ D [ai nQ bci ] 21 complex refractive index given by equation (22):
iD1
Q nQ D n nQ C ik nQ
n 22
Two important points should be raised. First, the absor-
bance, and not the transmittance, of any component is where n Q nQ is the real refractive index and k nQ is the
proportional to its concentration in the sample. IR spectra imaginary refractive index at nQ . n nQ is often simply called
linear in absorbance are rarely measured directly. Before the refractive index. k nQ is also sometimes known as the
1980, grating monochromators were routinely used for the absorption index and, incorrectly, as the attenuation index.
For any material, n nQ is determined by Snells law. A 7 POLARIZED RADIATION

few materials have no significant absorption in the mid-
and near-infrared. Those materials with low refractive index If a beam of unpolarized electromagnetic radiation is
(1.45
n
1.6) are useful as windows whereas those with being transmitted in the z direction, the amplitudes of the
high refractive index (2.4
n
4.0) are frequently used components of the sinusoidally varying electric field in the
as internal reflection elements. For organic and inorganic x and y planes are identical. When the beam is passed
molecules whose spectra exhibit typical absorption bands, through a polarizer, the component of the electric field in
the refractive index changes across the absorption band. A one plane is transmitted. For unoriented samples such as
typical refractive index spectrum has the appearance shown all gases and liquids and isotropic solids, the absorbance of
in Figure 4(a). This type of variation in the refractive index all bands in the spectrum is independent of the orientation
across absorption bands is known as anomalous dispersion. of the polarizer. If the molecules in a certain sample are
Outside those regions of anomalous dispersion, n is fairly preferentially oriented in a given direction, however, the
constant from the visible through the MIR spectrum. component of the dynamic dipole moment derivative of
each vibrational mode, (dm/dQ), in the direction that the
Like the real refractive index, the imaginary refractive
radiation is polarized will change as the polarizer is rotated.
index is also a dimensionless quantity. For pure materials,
Since the largest intensity of each band in an IR spectrum
k nQ is given by equation (23):
is observed when the beam is polarized in the direction for
a nQ which the change in dipole moment for that vibrational
k nQ D 23 mode is greatest, important information on the orientation
4pnQ
of samples can be derived by installing a polarizer in the
where a nQ is the linear absorption coefficient defined in beam. For example, one of the more important applications
equation (18). A typical absorption index spectrum looks of the use of polarized IR radiation is the measurement of
like an absorbance spectrum (Figure 4b). Whereas the the orientation of the chains in drawn or extruded polymers
transmission spectrum of samples is largely determined by (see, for example the articles by Bouffard and Pezolet13 and
the value of k nQ for each band, both the optical constants, Koenig14 later in this volume). A more esoteric application
n nQ and k nQ , control the reflection of samples, as discussed is the estimation of the angle at which surfactants are
by Claybourn in Volume 2 of this Handbook.12 adsorbed on the surface of water as described by Hoffmann
in Volume 2.15
Several different types of reflection spectroscopy yield
1.8
more information when the radiation is polarized than
Refractive index
1.7 when unpolarized light is used. For example, very thin

1.6 films of molecules adsorbed on the surface of metals only
1.5 absorb radiation polarized parallel to the plane of incidence.
1.4 The reflection of light from bulk samples depends on the
1.3 polarization of the light with respect to the plane of the
1.2 sample. The effective depth into a sample that can be
sensed by internal reflection spectroscopy is also different
4000 3000 2000 1000 for radiation polarized perpendicular and parallel to the
(a) Wavenumber /cm1 surface. Polarized radiation may even be used to eliminate
interference fringes from the spectra of thin polymer films
0.3 when the light is incident at Brewsters angle.
Absorption index
0.2
ABBREVIATIONS AND ACRONYMS
0.1 PMT Photomultiplier Tube
Nd : YAG Neodymium-doped Yttrium
Aluminum Garnet
0.0
4000 3000 2000 1000 REFERENCES

(b) Wavenumber /cm1
Figure 4. (a) Refractive index and (b) absorption index spectra 1. H.A. Laitinen, Anal. Chem., 45, 2305 (1973).
of poly(methyl methacrylate). 2. T. Hirschfeld and B. Chase, Appl. Spectrosc., 40, 133 (1996).
3. M. Fujiwara, H. Hamaguchi and M. Tasumi, Appl. Spec- P.R. Griffiths, John Wiley & Sons, Chichester, 557571,
trosc., 40, 137 (1996). Vol. 1 (2001).
4. H.F. Shurvell, Spectra-structure Correlations in the Mid- and 10. J.J. Turner, Bandwidths, in Handbook of Vibrational Spec-
Far-infrared, in Handbook of Vibrational Spectroscopy, troscopy, eds J.M. Chalmers and P.R. Griffiths, John Wiley
eds J.M. Chalmers and P.R. Griffiths, John Wiley & Sons, & Sons, Chichester, 101127, Vol. 1 (2001).
Chichester, 17831816, Vol. 3 (2001). 11. P.R. Griffiths, Beers Law, in Handbook of Vibrational
5. P.R. Griffiths, Far-infrared Spectroscopy, in Handbook Spectroscopy, eds J.M. Chalmers and P.R. Griffiths, John
of Vibrational Spectroscopy, eds J.M. Chalmers and P.R. Wiley & Sons, Chichester, 22252234, Vol. 3 (2001).
Griffiths, John Wiley & Sons, Chichester, 229239, Vol. 1 12. M. Claybourn, External Reflection Spectroscopy, in Hand-
(2001). book of Vibrational Spectroscopy, eds J.M. Chalmers and
6. G. Keresztury, Raman Spectroscopy: Theory, in Hand- P.R. Griffiths, John Wiley & Sons, Chichester, 969981,
book of Vibrational Spectroscopy, eds J.M. Chalmers and Vol. 2 (2001).
P.R. Griffiths, John Wiley & Sons, Chichester, 7187, Vol. 1 13. T. Buffeteau and M. Pezolet, Linear Dichroism in Infrared
(2001). Spectroscopy, in Handbook of Vibrational Spectroscopy,
7. H.G.M. Edwards, Spectra-structure Correlations in Raman eds J.M. Chalmers and P.R. Griffiths, John Wiley & Sons,
Spectroscopy, in Handbook of Vibrational Spectroscopy, Chichester, 693710, Vol. 1 (2001).
eds J.M. Chalmers and P.R. Griffiths, John Wiley & Sons, 14. J.L. Koenig, Trichroic Infrared Measurements of Orien-
Chichester, 18381871, Vol. 3 (2001). tation, in Handbook of Vibrational Spectroscopy, eds
8. A. Weber, Raman Spectroscopy of Gases, in Hand-book J.M. Chalmers and P.R. Griffiths, John Wiley & Sons, Chich-
of Vibrational Spectroscopy, eds J.M. Chalmers and P.R. ester, 711715, Vol. 1 (2001).
Griffiths, John Wiley & Sons, Chichester, 176195, Vol. 1 15. H. Hoffmann, External Reflection Spectroscopy of Thin
(2001). Films on Dielectric Substrates, in Handbook of Vibrational
9. S.A. Asher, Ultraviolet Raman Spectrometry, in Handbook Spectroscopy, eds J.M. Chalmers and P.R. Griffiths, John
of Vibrational Spectroscopy, eds J.M. Chalmers and Wiley & Sons, Chichester, 10091027, Vol. 2 (2001).
Infrared Spectroscopy: Theory
Derek Steele
Royal Holloway, University of London, Egham, UK
1 INTRODUCTION fields. For vibrational spectroscopy, electric dipole interac-

tions with the electric field are totally dominant, and we
The factors that determine the position and strength of can ignore the magnetic field. This of course is the reverse
absorption in the infrared (IR) are surveyed in this article. of the situation with nuclear magnetic resonance (NMR).
Band shapes in condensed state matter contain information A quantum-theoretical analysis (see later) shows that there
on relaxation processes, although this is difficult to evaluate. must be a change in dipole moment before absorption of
Following a brief survey of this, the classical and then the the radiation may occur. Now the forces and nuclear masses
quantum-theoretical calculation of vibrational frequencies within molecules are such that the vibrational frequen-
are briefly described. Finally the simultaneous excitation of cies lie within the range 1012 1014 Hz. Such frequencies
vibration and rotation in the vapor state is introduced. lie within the IR portion of the spectrum. Because instru-
ments used to measure IR rays measure wavelength, rather
than frequencies, it is normal to describe such radiation
2 FACTORS THAT CONTROL
by its inverse wavelength (wavenumber). The usual unit
ABSORPTION is cm1 . In such terms, fundamental vibrational wavenum-
bers fall below 3700 cm1 . A lower practical limit is usually
Absorption in the IR region arises predominantly from exci-
200 cm1 , although there are modes that involve such weak
tation of molecular vibrations. For noninteracting molecules
forces that the wavenumbers tend towards zero. Such modes
consisting of N atoms there are 3N 6 vibrational degrees
occur in carbon suboxide and diboron tetrafluoride and are
of freedom (3N 5 for a linear molecule), which is
associated with the very weak resistance to torsion about
equivalent to stating that all vibrational distortions can
the nonterminal bonds.
be described as the sum of 3N 6 fundamental vibra-
In diatomic molecules the requirement of a dipole change
tional modes. The forms of these vibrations are adequately
leads to the simple requirement that the molecule must have
described in the classical limit, although quantum theory is
a dipole moment. It follows that homonuclear diatomics do
essential to explain certain phenomena. We shall commence
not absorb, whereas molecules such as HCl absorb strongly.
with a semiclassical approach, and later review the essential
For small displacements of the atoms from their equilibrium
quantum effects. Each of the fundamental vibrational modes
positions, the forces on the nuclei are linearly proportional
will have a frequency determined by the forces exerted on
to the displacement x (Hookes law), and for displacements
the atoms as they move from the equilibrium molecular
from equilibrium these must be towards the rest position
configuration in the manner as required by that mode. If
(i.e. negative). Since force equals mass times acceleration
electromagnetic radiation passes through the material it may
we have:
interact provided that the frequency matches exactly that of
a fundamental mode. Electromagnetic radiation, however, is @V2 @2 x
Force D fx D x D m 1
a pure energy wave, that is it has no inertial mass, and there- @x 2 @t2
fore it can only interact through its electric and magnetic where m is the reduced mass equal to m1 m2 /m1 C m2 ), V
represents the potential energy and t represents time. The
John Wiley & Sons Ltd, 2002. atomic masses are designated m1 and m2 , respectively. The
solution of this equation is deformations. It is not feasible to distort one bond without
affecting the adjacent bond(s). Conservation of momentum
x D xo cos2pnt C q 2
dictates that, for a CH bond, movement of the hydrogen
n 1 f has to lead to an equivalent momentum for the remainder of
D nQ D 3 the molecule. Since no chemical bond is infinitely strong,
c 2pc m
the adjacent bond deforms to some extent, and a coupling of
where q is an arbitrary phase angle, n is the frequency the motions of the two bonds ensues. Clearly the more the
of the vibration and nQ is the wavenumber. It transpires two bonds are in line, then the greater will be the coupling.
that for first-row elements and hydrogen the force con- Also the closer the unperturbed frequencies, then the more
stant, f D @2 V/@x 2 , is reasonably proportional to the bond strongly they will interact. Such couplings will always push
order, and we can take it that for single bonds the fre- the uncoupled frequencies apart. Notice that, even though
quency is inversely proportional, to a first approxima- we cannot identify it here, the CC frequency will also
tion, to the inverse square root of the reduced mass.
be depressed by the CH motion. In C2 HD and C2 D2 the
(The terms wavenumber and frequency will be used inter-
CC mode drops to 1851 and 1762 cm1 , respectively. The
changeably, although one should be aware that they are
decreases are due to the heavier masses of the deuterium
dimensionally different and are strictly proportional only
atoms compared with hydrogen and the unperturbed CD
in as long as the wavelength is measured in a vacuum
wavenumber being near
or in constant atmospheric conditions so that the veloc-

ity of light is constant.) On this basis we can ratio- 3374 C 3287 mCH
nalize the relative positions of many of the bands aris- nQ CD ' D 2245 cm1 4
2 mCD
ing in the spectra of complex molecules. First we note
that CH bonds in saturated aliphatic molecules absorb Since the CD wavenumber is still above the CC value, the
near 2900 cm1 . As the reduced masses of the CH and latter will be depressed substantially more than in C2 H2 . In
CC bonds are 0.9297 and 6.0000 amu, respectively, then methyl ethyne the CC wavenumber rises to 2140 cm1 ,
the CC stretching bands should appear ( if allowed and in dimethyl ethyne it rises further to 2300 cm1 . It
by creation of a polarity gradient in some way!) near might be tempting to consider this as due to a strength-
1140 cm1 . CC bands do center around 1100 cm1 . Car- ening of the triple bond as a result of the inductive effect
boncarbon double and triple bonds of the methyl, but this is totally insignificant. The replace-
p should occur on the
basis of equation (3) near 1140 2 D 1612 and 1140 ment of a hydrogen by a methyl replaces the downward
p
3 D 1974 cm1 , respectively. These are indeed remark- shift of the triple bond wavenumber (due to the higher
ably accurate. As will be seen in SpectraStructure Cor- CH stretching wavenumber) by an upward shift, as the
relations in the Mid- and Far-infrared CDC bonds absorb CCH3 stretch is found at 930 cm1 in the methyl ethyne
in the range 16501600 cm1 . In ethyne the CC vibra- and at 687 and 1126 cm1 in the dimethyl. The extent of
tion is at 1970 cm1 ! With a slightly larger reduced mass, it the interaction is very large, not only between the adjacent
might be anticipated that the carbonyl wavenumber would bond stretches, but also between those of bonds attached to
be just below 1600 cm1 , whereas it occurs in ketones and either side of the ethynic bond. It is easily visualized that
aldehydes near 1720 cm1 . The rationalization of this is that interactions will be greatest when the bonds lie in line. In
there are good arguments for believing that the bond orders Section 7 we will return to this example and demonstrate
of CO bonds are significantly greater than those assigned that the magnitude of the interaction is explained almost
on the simple valency model. exclusively by the kinetic effect. One other aspect of these
It is widely known that the vibrational spectra of mole- vibrations will be introduced here however. Where symme-
cules are highly characteristic. They are often described as try elements exist, the motion of symmetrically equivalent
giving a fingerprint for the molecule. This does not tally atoms (bonds) must bear a well-defined relationship to one
with the idea that the spectrum is the sum of contributions another. Thus when a plane of symmetry exists, bonds on
from each bond and inter bond angle. For example, one opposite sides of the plane must move either in phase or
might expect on that basis that rotational isomers, and out of phase with respect to one another. This gives rise to
indeed positional isomers, would have the same spectra. symmetric and antisymmetric modes. It should be appar-
This is certainly not the case. The distortions of different ent that the vibration described above in which the two
bonds interact. Consider the stretching vibrations of the CH bonds move in phase can give no dipole change.
three molecules ethyne, methyl ethyne and dimethyl ethyne. It cannot therefore give rise to absorption. It does appear
In ethyne there are two CH stretching modes at 3374 and however in the Raman effect, which will be described in
3287 cm1 . As mentioned earlier the CC is at 1974 cm1 . Raman Spectroscopy: Theory. We shall deal briefly with
The major factor is kinetic interaction between the bond symmetry and group theory in Section 3, followed by an
Infrared Spectroscopy: Theory 3
introduction to matrix methods of calculating vibrational element, S, can contribute to cS since the symmetry
frequencies in Section 7. In Section 5 we will return to operation would convert the coordinate into one of an
absorption intensities. In Sections 4 and 5 we shall see how equivalent atom. Thus no coordinates of the hydrogen
anharmonicity in the force field and higher terms in the atoms can contribute to the general character for the
dipole moment expansion lead to overtones and combina- plane of symmetry sV yz. However, for the oxygen
tion bands (two vibrational modes excited simultaneously). z1 transforms into itself, as does y1 , whereas x1 trans-
The higher-amplitude motions resulting from this multiple forms into x1 . The character for sV yz arising from
excitation will require a relaxation of symmetry require- the oxygen atom is then 1 C 1 1 D 1. It can be read-
ments, and this will lead us to consider local modes. ily seen that each nucleus on any plane of symmetry
must contribute C1 to the character. Thus for sV xz
with all three atoms on the plane c[sV xz] D C3.
3 THE EFFECTS OF SYMMETRY ON
Following this logic the character contribution for each
MOLECULAR VIBRATIONS atom lying on a two-fold rotation axis, C2 , is 1, and
for a C3 axis cS is zero, while for an atom at a center
Where any symmetry exists in a molecule, the vibrations
of symmetry, i, ci is 3. Applying this now to H2 O,
will reflect that symmetry in that the vibrational wave
Table 1 is generated. Table 1 shows the character of the set
functions will be symmetric or antisymmetric with respect
of atomic Cartesian coordinates for each symmetry operator
to any simple symmetry element. This statement requires
and also shows the C2V character table.
that a symmetry operation converts any distortion into plus To determine the number of irreducible representations
or minus the distortion of an equivalent bond, or angle. The belonging to each class arising from a representation char-
quadratic energy is then unchanged, as can be seen from the acterized by a set of characters, such as the above c (in other
quadratic expression involving two equivalent distortions, words, to what species the modes belong arising from such
R1 and R2 . as the above set of Cartesian coordinates) the following
2V D fR12 C fR22 C f12 R1 R2 5 formula is used;
1
Here, by symmetry, f1 D f2 D f. For R1 D R2 , n D c ScS 6
h s
this expression is unchanged by interchanging the labels,
namely reflecting through the symmetry plane, as required where h is the order of the group (that is, the number of
by the equivalence of the two bonds. The above has to symmetry elements) and the sum is taken over the set of
be modified slightly for situations when a rotational axis
C3 or higher occurs. For a detailed discussion of groups z1
and their relevance in spectroscopy the reader is referred
to Cotton.1 In view of the limited space available it must O1 x1
be assumed that the reader has some basic knowledge of H2 H3
group theory. The existence of group character tables and
their significance is presumed. In the following the pro- Figure 1. An atom-centered Cartesian coordinate system for H2 O.
cedure will be described for determining the number of
fundamental modes belonging to each symmetry class for a Table 1. The characters of the Cartesian displacement coordi-
given molecule belonging to a given point group. To eval- nates and the valence coordinates of the H2 O molecule. (For
uate the number of modes in each symmetry class, it is convenience, the C2V character table is copied from the appendix.)
first necessary to determine the character for each symme-
Sym. oper. cS
try operator, cS, generated by the nuclear coordinates. Atom
This is then transformed to find the number of irreducible E C2 sV xz sV yz
representations. We shall clarify this by considering H2 O.
O1 , H2 , H3 9 1 3 1
1. Associate a set of Cartesian coordinates with the OH2 , OH3 2 0 2 0
HOH 1 1 1 1
molecule (Figure 1).
2. Determine the appropriate character for each symme- Character table C2V
try operation applied to the x, y, and z coordinates A1 1 1 1 1 Tz x 2 ; y 2 , z2
A2 1 1 1 1 Rz xy
associated with the atomic coordinates. Remember that
B1 1 1 1 1 T x ; Ry xz
by character we mean the trace of the transformation B2 1 1 1 1 T y ; Rx yz
matrix. No coordinate that does not lie on a symmetry
symmetry elements S, and n is the number of irreducible we could generate eight internal coordinate representations.
representations in the species (in this case the number Clearly there must be three combinations of these that are
of vibrations of this species, including null frequencies equivalent to no deformation at all. If we wish to form
associated with translations and rotations). The c S are an idea of nature of the molecular modes it is important
the characters of the irreducible representation . h equals 4 to establish at least to what species these combinations
for the C2V group. Applying equation (6) to the B1 species: belong. The D4h character table is given in Appendix 1.
Let us assign the x and y axes as shown in Figure 2. The
[9 1] C [1 1] C [3 1] C [1 1] characters of the Cartesian displacement coordinates under
nB 1 D D3
4 the symmetry operations are shown in Table 2. Note that
In a similar way it is seen that nA1 D 3, nA2 D 1 and because we are dealing only with the in-plane coordinates
nB2 D 2, and thus the total number is 9. The number of then the characters under the C2 operations are zero (x ! x;
fundamental vibrations however is but three. Included are y ! y). Reducing this representation using equation (6)
the six non-periodic motions of rotation (R) and translation leads to A1g C A2g C B1g C B2g C 2Eu (total 8). (For a brief
(T). Since these always belong to the same symmetry explanation of these symmetry representation symbols, see
classes within the group they are usually included in the the end of Appendix 1.) Rotation about the z axis belongs
to the A2g species and the degenerate translations, Tx and
character tables, and here they are shown in the last column
Ty , are of the Eu . The vibrational species are then A1g C
of Table 1. Subtracting these we are left with 2A1 and 1B1 ;
B1g C B2g C Eu . The characters for the stretches and ring
for example, nB1 2 D 1.
angles are also listed in Table 2. As there are eight valence
We could have used a set of internal valence coordinates
coordinates it is apparent that we have three redundancies.
instead of Cartesian coordinates. The advantage would be
Once again applying the rule for reducing a reducible rep-
that the non-periodic motions are eliminated and the forms
resentation we find the stretches are A1g C B2g C Eu and
of the vibrations are seen. In this case, a suitable set would
the angle deformations are A1g C B1g C Eu . By subtraction
be the two OH stretches and the HOH angle. The char-
of the species of the vibrational fundamentals modes, the
acters for these are also given in Table 1, and it is easily
redundancies are seen to be of A1g and Eu . The sum of
derived that the OH stretches form representations of the
the angular deformations must be zero for a planar closed
A1 and B1 species while the angle bend is an A1 repre-
figure. The other redundancy is more complex and will not
sentation. The difficulty with using internal coordinates is
be considered further here.
that redundancies will always exist except for noncyclic
Before leaving this section on group theory, the signifi-
molecules with no more than two bonds attached to any
cance of the species of the translations Ts (s D x, y, z) and
atom. Thus the sum of the interbond angles about a tetra-
the Cartesian products will be described briefly. Only those
hedron is six, and with the four bonds this makes 10 internal
classes for which translations form representations can be
coordinates. There are, however, only nine vibrational fre-
IR active. The reason for this is that the dipole derivatives
quencies. It can be shown that for tetrahedral angles the sum transform in a parallel manner to simple translations. Thus
of all angular deformations must equal zero. This is a fully for the C2V group it is seen that only A1 , B1 and B2 class
symmetric representation. A set of symmetry coordinates
could be evaluated by taking all the angles and determin- y
ing the number of representations they generate, and then
subtracting one fully symmetric representation. For ring O O
systems the complexity arising from redundancies greatly
increases and the nature of the redundancies can be much x
more difficult to ascertain, although the procedure described
above, comparing the number of representations generated O O
by the Cartesian displacement and the valence coordinates,
Figure 2. Cartesian axis system for planar cyclic C4 .
will permit some idea of the modes to be ascertained. We
will demonstrate the methodology and the problems by ref- Table 2. The characters of the atomic Cartesian displacement
erence to a planar C4 ring system. The symmetry group is coordinates, the CC stretches and the ring angle bends of C4 .
D4h . We shall look only at the in-plane coordinates. In this
case then there must be 2N 3 D 5 modes corresponding E 2C4 C2 2C02 2C002 i 2S4 sh 2sv 2sd
to the 2N degrees of freedom of the nuclei in the molec-
c (a.c.c) 8 0 0 0 0 0 0 8 0 0
ular plane less two translational and one rotational (about
cR 4 0 0 0 2 0 0 4 0 2
an axis perpendicular to the plane) nonperiodic motions. ca 4 0 0 2 0 0 0 4 2 0
There are, however, four bonds and four angles, so that
vibrations can be IR active. For H2 O this happens to be quantum-theoretical result for the mean square amplitude
all three vibrations, but for high-symmetry molecules less of vibration is the same as the classical result provided that
than half may be active. Only those classes to which the we identify the energy as hn. It is
Cartesian quadratic products form representations may be
Raman active (see Raman Spectroscopy: Theory). Note E
x 2 0.5 D 8
these are necessary, but not sufficient, criteria for activity. f
Although symmetry allowed, the activity may be too weak
for absorption or scattering to occur. These points will be For a CH oscillator, taking a wavenumber of 3000 cm1 ,
developed later. this leads to a mean square amplitude of 0.11 A. Such a
In the event that a rigid point group is inappropriate, as figure is about one-tenth of the actual bond length. It can be
for free rotors (e.g. CH3 BF2 ), then the invariance of energy appreciated then that the harmonic requirement of a small
to certain distortions requires that permutation group theory amplitude is far from reality. The amplitude is greatest
be used.2 What we are seeking to explore with group theory for small masses, and therefore anharmonicity is greatest
is the invariance of energy and the consequences to cer- for XH oscillators. For a CH bond the anharmonicity
tain transformations of the molecular configuration. In free constant, x, in equation (7) has a value of near 0.021.
Whereas the observed wavenumber of a CH stretching
rotors, by definition, the energy is unchanged by rotation
in an alkane will be seen about 2900 cm1 , the harmonic
about a bond. This then makes rigid group theory inap-
frequency is greater by 2!x or 126 cm1 . As a result of
propriate. This is an advanced topic for which interested
large anharmonicities, the frequency of the transition from
parties are referred to the original literature.
V D 0 state to V D 1, n0!1 D n 2nx, differs from that of
n1!2 D n 4nx. At room temperature kB T hn, and by
4 QUANTIZATION AND Boltzmanns equation, the population of the V D 1 state
is negligible. However, for such as Cl2 with a wavenum-
ANHARMONICITY IN REAL ber of 565 cm1 about 8% of the molecules are in the
VIBRATING MOLECULES V D 1 state. Due to the anharmonicity, the transitions,
as seen in the Raman effect, will be displaced from the
Real molecules are not perfect harmonic oscillators; fur- 0 ! 1 transition. A much more dramatic example, how-
thermore they deviate in two very important ways from ever, is afforded by uranium hexafluoride. For such a heavy
classical oscillators. By Heisenbergs uncertainty principle, molecule the fundamental modes are sufficiently low3 that
the atoms can never be at rest. Quantum theory leads to the all V D 1 states are substantially occupied at room tem-
result that the minimum energy associated with any vibra- perature. Indeed the number of such modes is such that
tional mode is a half quantum, hn/2 D hc!/2. The second to simulate the experimental band contour requires thou-
factor is that energy levels are quantized so that the popula- sands of multiply occupied states to be considered. The
tion of the vibrational states at equilibrium is determined by positive wavenumber displacements result in the observed
Boltzmann statistics. For a single isolated mode of vibration band maximum of the triply degenerate IR active mode, n3
then the energy, Ei , can be expressed by Plancks law as (F1u ) being several wavenumbers above the V0 ! 1 value.
In low-temperature molecular beam studies, the individual
Ei 1 1 2 excited state transitions become resolved.
D !i Vi C !i xi Vi C
hc 2 2 The large amplitudes of vibration have a further and more
3 significant impact on observed spectra. As will be described
1
C !i yi Vi C C 7 in Section 5, overtones and combination bands occur due to
2
anharmonicity and due to derivatives of the dipole moment
where Vi is the vibrational quantum number associated with with respect to vibrational distortions, higher than linear.
the ith vibration and !, the (harmonic) wavenumber, is n/c. In these cases the increased energy leads to even larger
It follows that even at the absolute zero of temperature amplitudes. The equilibrium symmetry of the molecules
substantial vibrational energy remains, equal to a half now becomes so distorted that the invariance of energy
quantum of each and every fundamental mode. This is far to symmetry operations is destroyed. A methylene group
from insignificant and for methane it amounts to 2.15 kJ in an environment that conserves the equivalence of the
mol1 . This has a substantial impact on the specific heats of two bonds will have vibrations that are symmetric and
solids at low temperatures. Given the energy in an oscillator antisymmetric with respect to the bonds in the limit of
and the restoring force resulting from displacement from small-amplitude vibrations (point group symmetry applies).
equilibrium, it is simple to calculate the mean square As the amplitudes of all the vibrations increase, so the sym-
amplitude and the maximum amplitude of vibration. The metry breaks down and the bonds tend towards independent
oscillators. In this event, the theoretical model needed 3050

becomes that of the local mode.4,5 Application of equa-
tion (7) shows that the separation of states in which one of
two equivalent bonds is doubly excited and that in which
both bonds are singly excited is 2xi !i . The higher transition 2900
will be at twice the observed fundamental frequency (that
is, twice !i 2xi !i , see Figure 3).
/V (cm1)
Of course the transition from symmetry-based modes to
local modes is not abrupt in going from fundamental to 2750
harmonic, and some interaction at the first overtone level is

to be expected. Somewhat surprisingly it has been shown6
A
that the CH bonds of alkanes in the liquid phase show local
mode behavior already at the first overtone level. Spectra 2600
B
of many compounds have been measured up to such as the
tenth overtone level by using long pathlengths and using
techniques such as photoacoustic spectroscopy. In the local
mode, we can calculate the transition wavenumber for the 2450
1 2 3 4 5 6 7
Vth overtone from equation (7) as: Quantum number, V
E0,V Figure 4. Plot of nQ 0!V against V for the two nonequivalent sets
nQ 0!V D D !V !xVV C 1 C higher terms 9 of oscillators in acetone: A, in plane; B, out of plane. The points
hc at V D 1 : 3 are from acetone d5 spectra. [Taken from A. Allen,
A plot of nQ 0!V against V therefore has a gradient of D.C. McKean, J.P. Perchard and M.L. Josien, Spectrochim. Acta
A, 27, 1409 (1971), by permission of Elsevier Science.]
x !. In Figure 4 such a plot is shown for the two sets of
nonequivalent hydrogens in acetone,7 those in the skeletal should note, however, that higher-level interactions lead to
plane and those out of the plane. The wavenumbers for the splittings in the higher overtones.5
out-of-plane are lower as a result of the lone-pair electrons
of the oxygen feeding into the CH antibonding orbital. We
5 INTENSITIES OF ABSORPTION
We shall begin by defining some measures of intensity.

1,1 9000 cm1 In an absorption spectrum we can measure the intensity
= 2 4 + ZPE of the beam when no sample is present and when it is
0,2 2 8880 cm1
= 2 6 + ZPE present. Of course the cell itself leads to attenuation of the
beam and reflection will occur at all interfaces. We shall
assume here that this has been allowed, or corrected, for.
6020 cm1
0,1 = 2 + ZPE Let the intensity of the beam in the absence of the sample
be In0 and with the sample present be In . In the event that
v (r )
the probability of a molecule absorbing that frequency is

constant then
dIn
0,0 3040 cm1 n D al 10
= 2(/2 /4) = ZPE I0
where l is the pathlength through the sample over which the
intensity change is defined and a is a proportionality con-
stant. Integration gives the Lambert law In D In0 expal.
This law is rigorous provided that saturation of the excited
r energy states does not occur. Since relaxation of the vibra-
Figure 3. Plot of vibrational energy versus bond length for a tional energy is fast (see Section 6) this is true unless very
pair of degenerate anharmonic oscillators, showing the relative high laser powers are used a situation that is very difficult
energies in the local mode model of the ground, fundamental,
first harmonic and combination bands. A fundamental harmonic to achieve at the present state of technology of IR lasers. If
wavenumber of 3100 cm1 and an anharmonicity, !x, of 60 cm1 it can be done, then the relaxation times of the excited states
have been assumed. ZPE, zero point energy. could be measured and experiments comparable to those
often carried out in nuclear magnetic resonance (NMR) with terms that, apart from multipliers, are hyi jQp jyf i with
would be possible. In /In0 is known as the transmittance (T) p D 0, 1, 2,. . . . Since wavefunctions are orthogonal the
and log10 (1/T) is the absorbance (A). Note that for conve- p D 0 term disappears. For harmonic oscillators hyi jQjyf i
nience here, we use base 10 rather than natural logarithms. (abbreviated here after to hijQjfi D 0, unless the vibra-
For noninteracting molecules the proportionality constant tional quantum number, V, of state f differs from that of
a can be split so as to incorporate the molecular concen- state i by 1. (For the quantum theory of an harmonic oscil-
tration, a D2 c and then we have the BeerLambert law, lator see Eyring et al.8 ) According to quantum mechanics
log10 In0 /In D2 cl, where 2 is known as the absorptivity. the probability of an induced transition per unit time per
It is essential to note that strictly this is not a law, but unit radiation density between states jii and jfi is given by
a hypothesis, since it requires that there is no interaction Fermis Golden Rule (for a derivation see Heitler9 ). Care
between molecules. This situation rarely prevails even in must be taken in looking at these older texts to remember
gases. When used to measure concentrations it is strongly that following the adoption of the SI system of electromag-
advisable to test for linearity, or better, to calibrate (see netic units electric fields require dividing by a factor of
Beers Law). 4p 20, where 20 is the permittivity of free space.
Even with perfect instrumentation of infinite resolving
power, any band will have a finite band width. As we 2p2
Bif D Bfi D jhfjOer mjiij2 n nf C ni 14
will describe later, this is due to molecular rotation, to h2 20
Doppler shifts, to molecular interactions and to finite life- 2o is the permittivity of free space, eO r is a unit vector in the
times in the vibrationally excited states. The intensity of direction of the electric field vector of the radiation, E, and
a band then requires integration of absorbance over fre- (..) is the Kronecker delta that takes the value zero unless
quency (wavenumber). Although this integrated absorbance its argument is zero, in which case it has a value of unity.10
is widely employed, as we shall show a more fundamental In the above equation the Kronecker delta requires that the
quantity is the integral of absorbance divided by frequency: radiation frequency is exactly equal to the difference in
n energies of the initial and final states divided by Plancks
1 I 1
D loge 0n dnQ constant.
cl I nQ
n The term in Q2 exists for a change in the vibrational
1 I
D loge 0n d loge nQ 11 quantum number of 2, and therefore overtones will occur
cl I if the second derivative of the dipole moment is non-zero.
The integral is over the band. Since the breadths of bands It has been shown, for example, that in aromatics11 and
are generally small compared with the band center fre- in ethenes the second derivatives of the dipole moment
quency, it is a reasonable approximation to take the inte- with respect to movements of the hydrogen atoms out of
grated absorbance A (base e) as equal to n0. the plane of the molecular skeleton are substantial and
It has already been indicated that absorption depends on a responsible for the strong combination bands between 2000
dipole change accompanying excitation of the vibration. We and 1600 cm1 that characterize these types of compounds.
shall now look at this in more detail. In quantum-theoretical Overtones may also exist provided that the wavefunc-
terms the interaction between an electromagnetic beam and tions are nonharmonic. In general, however, they are much
a molecule is proportional to the square of the term weaker than the fundamentals. Note that since the total
molecular wavefunction can be written as a product over
hyi jHjyf i D hyi jm Ejyf i 12 all states, then the condition for combination bands to occur
(simultaneous changes in two different quantum numbers)
jyi i and jyf i are the initial and final state wavefunctions, m is that @2 m/@Qi @Qj 6D 0.
is the molecular dipole moment and E is the electric field Occasionally combination bands appear with consider-
of the incident radiation. H is the Hamiltonian operator, in able strength due to mixing with a nearby fundamental
this case equal to m E. We shall later refer to either side of transition of the same symmetry (Fermi resonance). The
equation (12) as the transition dipole. The dipole can be classic example of this is usually quoted as with the anti-
expanded as a Taylor series in the molecular distortion Q. symmetric CCl stretch (species F2 ) of CCl4 . The IR spec-
@m 1 @2 m 2 trum shows two equally strong bands at 797 and 768 cm1 .
m D m0 C QC Q C higher terms 13 Only one fundamental is expected here, and the splitting is
@Q 2 @2 Q
due to the almost perfect coincidence of the fundamental
Introducing the expansion into the transition integral and energy level with that in which two quanta, one of n1 (A1
recognizing that the derivatives are constants and may at 461 cm1 ) and one of n4 (F2 at 314 cm1 ), are simul-
therefore be taken outside of the integral, then we are left taneously excited. Since the symmetry of the combination
transition (D F2 A1 D F2 ) is of the same symmetry as change expected from the equilibrium charge distribution.
the fundamental, resonance occurs and the resulting states This latter explains the remarkably high intensity associ-
push one another apart and take on mixed character. For ated with movement of a hydrogen atom from the plane
perfect coincidence the new states will contain 50% of of benzene14 or ethene derivatives. As the electrons rehy-
each, and, since the combination transition has negligi- bridize around the carbon (moving from sp2 towards sp3 ),
ble intrinsic intensity, the two transitions have half of the so electrons build up on the opposite side of the skeletal
intensity due to the fundamental. Such resonances are quite plane to the hydrogen, thereby enhancing its positive charge
common and care needs to be exercised, in making vibra- (see Figure 5).
tional assignments, not to assume that the strongest bands Many studies have been made of the absorption intensi-
have to each be due to fundamentals. If the transition ties associated with bands arising from group vibrations.15
is both IR and Raman active, then the same partition- It is recognized that the transition moments are far more
ing of intensity should exist in both spectra. It may also sensitive to neighboring group effects than are the frequen-
be possible to alter one energy state relative to the other, cies themselves, and intensities are not particularly useful
and thereby alter the mixing, by interaction with different in evaluating the concentration of functional groups. How-
solvents. ever, some qualification of this statement is required in view
For an ensemble of randomly oriented molecules, the of the widespread application of principal factor (or com-
relation between the intensity, k , and the dipole derivatives ponent) analysis (PFA) in industrial spectroscopy.16 This
@m/@Qk is obtained by substituting the harmonic oscillator has been mainly used with near-infrared (NIR) spectra,
value of the transition integral but this is probably more to do with ease of generating
suitable spectra than due to fundamental limitations. The
h 0.5
hijQi jji D Vi C 10.5 if Vj D Vi C 1 application to the mid-infrared (MIR) spectra of coals is
8p2 n
an excellent example.17 The relevance to group intensities
otherwise D 0 15 is that PFA is a statistical technique of fitting data to a
into equation (14) followed by averaging over all orienta- progressively increasing number of unknown components
tions, and then finally by summing over all hot bands (all until the fit reaches an acceptable level. As such, it requires
initial states, jii) using Boltzmann statistics. The result of the data to consist of a finite number of linearly additive
this calculation is components. The application to the spectra of very complex
mixtures, such as coals, wheats and petroleum, both refined
Np @m 2 and unrefined, strongly suggests additivity is much better
i D 16
3cni 20 @Qi than used to be believed. This has been rationalized by
As could be seen by balancing the units of both sides demonstrating that group spectra are characteristic provided
of equation (16) the normal coordinates, Qi , are mass that the neighboring groups are the same.6 In this context,
weighted (for each atom by the square root of the atomic the methylene adjacent to a methyl in an n-alkane is quite
mass). This is a complication that will be dealt with later. different from that mid-chain. The CH stretches, for exam-
At present we shall ignore this. ple, are at about 2855 and 2940 cm1 for the methylene
There is considerable interest in the factors that deter- adjacent to methyl, while for the mid chain they are at
mine the dipole derivative. The stretching of a highly polar 2846 and 2920 cm1 .
bond, such as CF, might well be expected to lead to
a large dipole change. In practice, the dipole change is

much greater than would be expected for the charges on
the atoms in their equilibrium positions. Charge flows as
F B F
the nuclei move in this case towards the situation where F B F
F F
the fluorine atom carries a complete extra electron.12 In
certain cases where a trans lone pair of electrons exist, (a)
such as for the hydrogen atom in the aldehyde unit, exten-
sion of the bond results in a flow of electrons from the
lone pair into an antibonding orbital of the CH bond, C
thereby greatly enhancing the dipole change.13 This is C C
C H
C H
the same phenomenon responsible for bond weakening in C
such cases. In bending motions, nonbonding interactions, (b)
as well as rehybridization of the electrons at the nonter- Figure 5. Electron rehybridization from the pz orbital during the
minal atom, cause substantial deviations from the dipole transitions (a) A002 of BF3 and (b) A2u of benzene.
6 BAND SHAPES IN LIQUIDS with the molecule, and, since its value is constant for no
relaxation, Ct becomes equal to m2 hcos qt i where hcos qt i
In low-density vapors an absorption band consists of lines is the average over the ensemble of the cosine of the angle
arising from a large number of transitions arising from through which the molecules rotate and m is the scalar value
rovibrational transitions, perhaps from excited vibrational of m. The exponential in equation (17) can be written as
states as well as the ground state. In high-density fluids the cos2pnQ t C i sin2pnQ t. Since sina D sina for
interaction between different molecules will affect the spec- a symmetric band only the cos term survives in the inte-
trum by causing perturbations of the rovibrational energy gration. By integrating about the band center (note that
states and by reducing the lifetimes of the excited states the band does not have to be symmetric to do this), the
through energy-transfer processes. The individual transi- speed with which the direction of the transition moments
tions initially broaden and then merge into a smooth band become randomized can be found. For our fixed moment
contour. It is sometimes stated that the rotational transitions this gives us the rate at which the memory of the system for
are suppressed, but this cannot be so. The equipartition the- its original orientation is lost. The timescale of the decay
orem states that energy must be equally distributed between will be of the order of the inverse of the band width. Thus
each degree of freedom, and this must extend to the confor a band of half band width 10 cm1 the half-life of the
densed state. It was demonstrated by Gordon18,19 that the decay will be 1/3 1010 10 3 1012 s. This con-
contours still contain the rotational contribution. Thus for a cept of a fixed moment is not realistic, but for the present
diatomic molecule of moment of inertia I, the integrated we shall hold onto this model and examine the effects of
second moment of the intensity ( as defined in equa- two extreme assumptions about the molecular rotation. In
tion 11) about the vibrational band center, divided by the the first it will be assumed that all the molecules rotate
integrated intensity itself is, apart from a vibrational relax- about the same axis at the same speed. Ct will now vary
ation contribution, kB T/IX . IX is the moment of inertia. as [m0]2 cos2pnQ rot t. We can use equation (17) to evalu-
This is the same for vapor and for liquid. It is simply the ate the band shape. It is a property of a FT if say A is the
relaxation of the excited states that has changed. There is FT of B, then B is the FT of A with a change of sign in
a vibrational contribution to the width (see below). It is the argument. We have therefore
possible in certain situations to separate these contribu- 1
tions using Raman spectroscopic methods.18 For a more nQ 0 C nQ D Ct expi2pnQ t dt 18
detailed discussion see the literature.18 20 If we desig- C1
nate the transition dipole moment (see equation 12) of a Substituting for Ct and using the fact that the FT of a
given transition at a time t as mt, then its development cosine is a delta function equal to one when the argument,
as a function of the time lapse t D t00 t0 is related to nQ , is zero and equal to zero otherwise, we find that this
Fourier transform (FT) of the intensity about the band center predicts an infinitely sharp line.
as: Now consider a more realistic model in which Ct
decays exponentially. The FT of an exponential is a Gaus-
Ct00 t0 D hmt00 mt0 i D hm0 mt00 t0 i
C1 sian. Real bands do approximate to this, but dCt/dt
D nQ 0 C nQ expi2pnQ t dnQ goes to zero as the time tends to zero (see, for example,
1 Figure 6). That this must be so follows from the require-
D Ct where t00 t0 D t 17 ment, true absolutely in the classical limit, that all odd
moments of the correlation function of the dipole with
The brackets h. . .i designate an average over all the respect to time must be zero. To see this, consider the dipole
molecules, and since the transition dipole is a vector, this correlation function normalized so that C0 D 1. Let u be
transform yields the ensemble time decay of the transition a unit vector in the direction of the dipole. Then
dipole moment from an arbitrary time of origin. Ct is
known as the dipole correlation function. In the above we Ct hut u0i
D
assume that the system is in equilibrium so that the cor- C0 hu02 i
C1
relation function is independent of the time of origin t0 .
This expression is capable of quite simple interpretation. nQ C nQ 1 C i2pnQ tn /n! dnQ
1
First note that the dipole transition moment, m, is a vec- D C1
tor, which for noninteracting molecules is locked to the nQ C nQ dnQ
1
molecular framework. Assume first of all that the excited
i2ptn
state has a very long lifetime compared with the rota- D Mn 19
tional period of the molecule. The moment mt now rotates nD0,1
n!
0.0 the motion is governed by the rotational energy, then decay

becomes exponential due to collisional processes. In the
0.2 vapor the dipole correlation shows a minimum, perhaps
becoming negative, in a period after which most molecules
0.4 are facing in the opposite direction to that in which they
started (Figure 6). As in magnetic resonance, the decay time
0.6 will consist of isotropic and anisotropic contributions. In
a fluid, the transition dipole will also decay by transfer
0.8 of energy between molecules and by multibody collision
C (t )
deactivation of the vibration. These are responsible for the

1.0 isotropic component. Due to the very short timescale of
(a) these processes, separation of the relaxation mechanisms
1.2 is generally impossible. However, some progress has been
made with laser optical nutation experiments22 and with
1.4 (b)
Raman studies.
1.6
7 CALCULATION OF VIBRATIONAL
1.8
0.0 0.5 1.0 1.5 2.0 2.5 3.0
SPECTRA
Time (ps)
Although energetic aspects of vibrational spectroscopy,
Figure 6. Dipole autocorrelation function of the V11 band of such as ZPE and crystal lattice energies, are strictly
benzene. (a) is the dipole correlation function for benzene
dissolved in cyclohexane and (b) is that for benzene vapor. [Taken quantum phenomena, much can be achieved by classical
from I.R. Hill and D. Steele, PhD thesis, University of London type calculations based on interatomic potential functions.
(1975).] Assuming that the potential energy for deformation about
an equilibrium configuration is strictly quadratic in defor-
where we have used the Fourier series expansion of the mations, then the vibrational frequencies are given as the
exponential in equation (17). Mn is the nth moment of eigenvalues of the matrix product GF where F is the matrix
the normalized intensity with respect to the frequency shift of quadratic force constants and G (so called inverse kinetic
from the band center energy matrix) is such that
C1 @Ri @Rj 1
nQ C nQ nQ n dnQ Gij D x D x, y, z 21
@xk @xk mk
Mn D 1 C1 20 k
nQ C nQ dnQ where R represents an internal coordinate of the molecule,

1
usually chosen from bond stretches, angle bends and
Classically, the odd moments are zero since they have imag- bond torsions. The sum is over all atoms. The method-
inary coefficients. Expressions have been derived for the ology of carrying out these classical calculations is well
second and fourth moments for a number of types of vibra- documented.23,24 One of the principle advantages of such
tional transitions, including symmetric tops and diatomic calculations is that the eigenvectors of GF yield the nature
oscillators.18,19 The second moments are independent of of the molecular vibration. Although it is not possible in
torques, which first appear in the fourth moments. For a the limited space available to deal in detail with calcu-
diatomic absorption band the second moment has a value lational procedures, we shall demonstrate the methodol-
of 2kB T/I, where I is the moment of inertia. Note that it ogy by carrying out calculations on the H2 O molecule.
takes the same value for both gaseous and liquid states. As already seen, all possible internal distortions of the
For the degenerate vibrations of symmetric tops the first- nuclear framework are covered by three valence coordi-
order Coriolis constant (see Section 9) is involved and may nates, two OH stretches and the HOH angle. Using the
drastically increase the moment.21 atomic-centered Cartesian coordinates shown in Figure 1,
Continuing to discount vibrational relaxation and con- the derivatives such as @fOH2 g/@x2 are easily derived
sider the transition moment as being fixed to the molecule, to be as shown in Table 3. For the central atom of
then the correlation function hm.0/ m.t/i will describe the an angle bend, it is easiest to use the fact that the
rotation of the molecule as defined by the transition dipole sum of the derivatives of a given internal coordinate,
axis. In the condensed state after an initial period in which summed over the x or y or z displacements of all the
Table 3. The inverse kinetic (G) and the force constant (F) matrices and their symmetrized forms. (The deformation of the HOH angle
is designated q and the value of the angle is 2a.)
B Matrix
x1 y1 z1 x2 y2 z2 x3 y3 z3
d C sin a 0 C cos a sin a 0 cos a 0 0 0

d sin a 0 C cos a 0 0 0 C sin a 0 cos a
HOH D q 0 0 2 sin a/R cos a/R 0 C sin a/R C cos a/R 0 C sin a/R
G Matrix
d12 d13 da
d12 mH C mO cos2amO sin2a

d13 cos2amO mHC mO sin2a
da sin2a sin2a 2[1 cos2a]mO C 2mH
U Matrix
d12 d13 da
S1 21/2 21/2 0
S2 21/2 21/2 0
S3 0 0 1
F Matrix
d12 d13 da
d12 fR fR/R fR/q

d13 fR/R fR fR/q
da fR/q fR/q fq
G D UGUt
S1 S2 S3
S1 mO [1 C cos2a] C mH 0 21/2 sin2a

S2 0 mO [1 cos2a] C mH 0
S3 21/2 sin2a 0 21/2 mO C mH
F D UFUt
S1 S2 S3
S1 fd C fd/d 0 21/2 fd/q

S2 0 fd fd/d 0
S3 21/2 fd/q 0 fq
atoms, must be zero. The central atom derivatives then M1 is a diagonal matrix, of order three times the num-
will be equal to minus the sum of the derivatives of ber of atoms, in which the entries are the inverse of
the two outer atoms. Defining this matrix @Rj /@xk as B, the atomic masses. For H2 O then we have a G matrix
then the G matrix (elements defined by equation 21) is of order 3. We can simplify this by transforming into a
given by symmetrized basis set. For this case, this would be the
G D BM 1 B t 22 unique set
d12 C d13 interatomic coordinate between nonbonded atoms to bonded

S1 D p
2 internuclear distances and to interbond angles. This is
S2 D q the so-called Urey Bradley force field. The existence of
rotational barriers even in ethane demonstrates the reality
d12 d13
S3 D p of such a model. The development of good quality force
2 fields for molecular dynamics simulations requires such a
where q represents the interbond angle. We can now define model. Interested readers are referred elsewhere.24,26
a transformation matrix, U, as @Sm /@Rn , where Rn desig- Having recognized the sort of force field that is needed
nates a valence coordinate. This is also shown in Table 3. in order to simulate vibrational frequency patterns, the
In symmetrized coordinates the G matrix becomes G D next problem is to obtain some values as input data. It
U GU t . The physical significance of G is that the kinetic was pointed out in Section 2 that force constants were, at
energy is given by 2T D P t GP where P is a column least to a first approximation, characteristic of a bond (or
vector of momenta conjugate to the coordinates S. (See interbond angle). Throughout the past half-century, efforts
Appendix VII of Bright Wilson et al.23 where it is also have been made to derive force fields from experimental
shown that 2T D SP t G1 S.
P It is from this latter relationship frequencies, mean square amplitudes of vibration, vibration
that the name inverse kinetic energy matrix derives.) rotation interaction constants and other data that are force
It is now necessary to give form to F. It is usual to field dependent. These efforts have been largely thwarted
assume that the potential energy for distortions of the through a variety of factors. The problem is easily seen
nuclear framework can be written as a Taylor series expan- by considering a symmetry class to which three vibrations
sion in the internal coordinates. Designating the general belong. Since the force constant matrix is symmetric about
internal coordinate as Rj , then the diagonal, the number of independent force constants is
six. For four vibrations of a class there are 10 constants.
@V 1 @2 V
V D V0 C Ri C Rj Rk More data dependent on the force field are needed. The use
i
@Ri 2 jk @Rj @Rk of isotopically substituted molecules gives extra data, but
in fact the number of new pieces of independent data is
1 @3 V very restricted. The problems here are as follows:
C Rl Rm Rn C 23a
3! lmn @Rl @Rm @Rn
Isotopic substitution leads to a set of vibrational fre-
Provided that interest is solely in the internal distortion quencies that are related to those of the original set by
energy, then V0 can be defined as zero and then ignored. a number of isotopic product rules. We shall return to
For small displacements, Ri , the cubic term, will be small the more important of these later in this section.
compared with the quadratic term and, at least for first- In the calculations we are envisaging the vibrations are
order calculations, this last term listed in equation (23) can assumed to be harmonic. The experimentally observ-
be dropped. If all the coordinates, R, are independent and able data are anharmonic. Some allowance for this may
the distortions are defined from the equilibrium position, be made, especially for CH stretching frequencies, for
then the first derivatives must be zero. Equation 23(a) now which anharmonicity, as we have seen, is of major
reduces to equation 23(b). importance. In general, however, there is an inher-
@2 V ent uncertainty in the precision of the frequencies as
2V D Rj Rk kjk Rj Rk 23b a result of anharmonicity and vibrational perturbation
jk
@Rj @Rk jk effects that requires that the number of pieces of data
where kjk is the quadratic force constant for coordinates j is significantly greater than the parameters (force con-
and k. While this is the assumption that is normally made, stants) to be fitted so that a statistical fit is possible.
it should be recognized that for simple valence coordinates The forms of the vibrational modes rarely involve the
where redundancies exist, the basic assumption of inde- motions of all atoms. In fact experience soon teaches
pendence is invalid. Conversion into a set of symmetry us to recognize that many modes are due, for example,
coordinates in which the redundancies are subsumed avoids to the CH2 (or CD2 ) scissoring motion. If each mode
this problem. Interpretation of the force field in terms of was due to one clearly defined symmetry coordinate,
valence deformations, however, brings us back into this then it would depend solely on the symmetrized force
dilemma. Crawford has considered this problem.25 constant associated with that coordinate. Changes in
Another approach, which has been widely used in the vibrational frequencies on isotopic substitution would
past, particularly by Japanese scientists, is to include linear be entirely explained by the changes in the correspond-
terms in nonbonded interactions, and then to relate the ing G matrix elements that result from changes in the
atomic masses. Although such factoring rarely occurs We must be satisfied with stating these and referring to
to such an extent, the interaction may be insufficient to other texts for proof.23,24 The matrix product GF has a
allow the interaction terms to be evaluated accurately. disadvantage computationally in that it is asymmetric. A
more convenient symmetric matrix with the same eigen-
In recent years, progress has been rapid due to the quality values can be generated as follows. Define D as BM1=2
of prediction possible with quantum-theoretical methods.27 where M1=2 is a diagonal matrix with diagonal elements
This will be dealt with in Section 8. Let us proceed in the equal to the inverse square root of the atomic masses. From
assumption that we have a quadratic force field. To calculate equation (22):
the vibrational frequencies from the GF matrix this must be G D DDt 30
diagonalized by a suitable transformation. This is equivalent
to saying we require the eigenvectors, L, and eigenvalues, Now consider the matrix product Dt FD with eigenvalues
3 where 30 and eigenvectors Y:
GFL D L3 24
Dt FDY D Y30 31
For each eigenvalue lm there will be a column vector of
L, lk , the elements of which give the relative amplitudes Premultiplying equation (31) by D and substituting equa-
of motion of each coordinate used to define G. In this tion (30):
way we can analyze the nature of the molecular vibra- GF.DY/ D .DY/30 32
tions. The eigenvectors lm are related to the vibrational
We see that DY is also an eigenvector matrix of GF with
frequencies by lm D 4p2 n2m . If the G matrix is calculated
eigenvalues l0 . It follows that 3 D 30 and, provided that L
using relative atomic masses and the force constants are
and DY are similarly normalized, that L D DY. As Dt FD is
expressed in mdyn A1 (10 p
18
J A2 ), then the wavenum-
1 symmetric, it can be diagonalized using faster routines. We
ber nQ m /cm D 1302.9 lm .
will also see that Y, which must have the dimension number
The eigenvector matrix, L, requires further specification.
of atomic Cartesian coordinates by number of eigenvectors,
Multiplying any column of L, say lm by a constant, x, will
gives the Cartesian amplitudes when weighted properly.
still lead to a satisfactory eigenvector.
While the B matrix relates the vector of internal valence
GF lm D xlm lm 25 coordinates, R, to the atomic Cartesian displacement coor-
dinates, x (see equation 21), the inverse relationship,
Standard computer routines will usually normalize so that
ltm lm D 1. However, an alternative normalization offers x D AR 33
considerable advantages. Eigenvectors, and the correspond-
is nontrivial as conditions of zero translational and rota-
ing normal modes, are orthogonal. If the atomic displace-
tional momenta must be built in (the so-called Eckart
ment coordinates are weighted by the square root of the
conditions). The matrix A has been shown to be equal to
atomic masses, then representing these mass weighted coor-
dinates as Q, the kinetic energy, T, may be written in the A D M1 Bt G1 34
simple form
2T D Q P tQ
P 26 Substituting into equation (33)
With this form the potential energy, V, becomes x D M1 Bt G1 LQ

D M1=2 M1=2 Bt .Bt /1 MB1 LQ
2V D Qt 3Q 27
D M1=2 D1 DYQ D M1=2 YQ 35
Writing the relationship between the normal coordinates
and the chosen set of internal displacement coordinates, where equation (22) and the relation between the L and Y
R, as matrices have been used. Q denotes a normal coordinate,
R D LQ 28 but can be envisaged as a scaling factor. To assign a numer-
ical value to the Cartesian displacements, the displacement
then it can be proven that the 3 matrix in equation (27) is must be defined more precisely. We can define this as the
the diagonal matrix with entries equal to 4p2 n2 and that the maximum displacement of the oscillator or as the root mean
following matrix relationships ensue: square displacement. We obtained earlier a classical value
for the root mean square amplitude of vibration as a func-
G D LLt 29a
tion of the energy. One basic difference from the quantum
t
L FL D 3 29b model is that in the quantum theory there is always the ZPE
and in that transitions from a series of levels may occur. The eigenvalues and normalized eigenvectors are
Ignoring the latter, but considering the increase in ampli-
tude with excitation of one quantum of vibrational energy, 6.741 0.985 0.169 0
then for a harmonic oscillator the scaling is h/4p2 n1/2 . If 3 D 2.341 l D 0.420 0.908 0 40
allowance is made for hot band transitions the expression 6.414 0 0 1.000
becomes
1/2 By normalization here we mean that for any given column
h hni
coth 36 of l the sum of the squares of the terms equals unity,
8p2 ni kB T
i.e. Ikt Ik D 1. Multiplying to satisfy equation 29(b) can be
carried out for each vector in turn to yield L.
7.1 Revisiting the ethyne spectrum
1.023 0.170 0
L D 0.176 0.368 0 41
We shall now use the theory of this section to prove our 0 0 1.037
earlier assertion that the increase in frequency of the ethynic
CC triple bond is entirely explicable in terms of the mass It is left as an exercize for the reader to verify that equa-
effects on the G matrix. First of all the G matrix for the tions 29(a) and 29(b) are satisfied. The l values translate
stretching vibrations of a linear system XCDCY is easily into nQ of 3383, 1994 and 3300 cm1 . The antisymmetric
obtained by writing down the B matrix for the stretching CH mode has the original assumed bond wavenumber,
of each bond and then applying equation (21). The result whereas the symmetric CH is raised 83 cm1 and con-
is found to be tains significant CC stretching as witnessed by the l12
m C m m 0 value of 0.176. Likewise the CC mode contains a lot of
C X C
CH motion.
GD mC mC C mC mC 37 If the hydrogens are now replaced by carbon atoms we
0 mC mC C mY will get an approximate model of the dimethyl ethyne
in which the force field is held as in ethyne itself and
where mX is the inverse mass of atom X. To give an the methyl hydrogens are ignored. Repeating the calcula-
expression for F we can ignore interaction terms for the tion with the g11 and g22 terms as 2mC the wavenumbers
present purpose and obtain approximate diagonal quadratic change to 2350, 831 and 1299 cm1 . The CC has been
force constants as follows. Take nQ CH as 3300 cm1 , giving pushed up by 356 cm1 (experimental 339 cm1 ) and the
l as 6.414, and the corresponding G element as mC C mH D CC stretching is depressed by 1299 831 D 468 cm1 .
1.07559 amu1 . The force constant needed so that GF D l The observed CC wavenumbers are lower at 1029 and
then is 5.9634 mdyn A1 . For the CC triple bond it is 697 cm1 , but the proportional decrease is well reproduced.
tempting to take nQ CC as 2000 cm1 but this would be found The wavenumber discrepancy is partly due to carbon car-
to yield too large a wavenumber when interaction with the rying the hydrogen atoms, thereby increasing its effective
CH bonds is considered. Instead we will increase the force mass, which could account for one-half of the observed
constant from 14.1 to 16.0 mdyn A1 . Now H D GF is difference (applying equation (3) the wavenumber will be
lower by (12/15)1/2 ). The residual difference arises from
1.07559 0.08333 0
the force constant for the CC bond being lower than that
H D GF D 0.08333 0.16667 0.083333
of the CH bond. Note also that the predicted separation of
0 0.08333 1.07559
the CH modes of ethyne agrees well with the experimen-
5.9634 0 0 tal of 83 cm1 (33733287 cm1 ) proving that these kinetic
0 16.0000 0 38 arguments are quite adequate to explain the experimental
0 0 5.9634 observations.
To close this section the effect of isotopic substitution on
To simplify the calculation the vibrational matrices can be vibrational frequencies will be considered. For an isolated
symmetrized using a U matrix: harmonic oscillator, equation (3) requires that the frequen-
1 1 cies are inversely proportional to the square root of the
p 0 p reduced masses. In a polyatomic system, however, there
2 2
are also bending motions and interaction with rotations and
UD 0 1 0 39

1 1 translations to be considered. We will not seek to prove
p 0 p these, but simply quote the rules.
2 2
7.2 Product rule representation, but a translation along the x-axis transforms
as B3u and thus
Consider two isotopically related molecules and classify the
i nH mD 1/2 114 1/2
vibrations according to the symmetry group that embraces D D 1.389 43
the symmetry of both molecules. For example, for ethene i nD mH 118
d0 and 1,1 ethene d2 the common symmetry is C2V . For It is left to the reader to show that for the B1g species the
a particular symmetry species within this group, the ratio ratio is 1.286.
of the product of the vibrational frequencies for the two
molecules a and b is given by
@q/2 7.3 Sum rule

i nai j maj 1/2 Mb T/2 Ibq
D qDx,y,z 42 There are a number of sum rules, but the most important is
i nbi j mbj 1/2 Ma Iaq
as follows:
where is the product operator with the product of the
argument taken over all values belonging to the species If several isotopically related compounds can be superim-
; the product j is taken over each nonequivalent atom posed with appropriate signs in such a way that the atoms
vanish in all positions, then the corresponding linear com-
of mass mj , raised to the number of representations that bination of the sums of the squares of the frequencies will
atom contributes to species ; M is the molecular mass vanish in the harmonic approximation.
and Iq the moment of inertia about the qth axis, T is the
number of translational degrees of freedom forming repre- As an example we can take ethyne and its mono- and
sentations of the species and @q takes the values of 0 or dideuterated forms. Taking two molecules of the mon-
1 depending on whether rotation about the q axis belongs odeuterated form, rotated with respect to one another by
to the representation in question or not. As an example, 180 , and superimposed, we would have an indistinguish-
consider p difluorobenzene and its d4 isotopomer (see able stack from superimposing a d0 and a d2 molecule.
Figure 7). First note that since the masses appear as sim- Considering just the vibrations along the molecular axis
ple products of one another, all masses cancel in numerator (A1 in C1v ), the wavenumbers (cm1 ) are (C2 H2 ) 3373.7,
and denominator except those for the isotopically replaced 1973.8 and 3287; (C2 D2 ) 2700.5, 1762.4 and 2427; and
atom. The symmetry species is D2h . With the Cartesian dis- (C2 HD) 3334.8, 1851.2 and 2584. With these values the
placement coordinates oriented as in the figure the x and discrepancy in the sum rule is 0.24%.
z coordinates have characters cE D 4 and csxy D 4,
others zero, while for the y coordinates cE D 4 and nQ 2 C2 H2 C nQ 2 C2 D2 nQ 2 C2 HD
D 0.0024
csxy D 4. Applying equation (2) we find that the x and nQ 2 C2 H2 C nQ 2 C2 D2
y coordinates of the hydrogens generate the representa- 44
tions Ag C B1g C B2u C B3u and the z coordinates generate
B2g C B3g C Au C B1u . The molecular masses of the d0 and
d4 molecules, MH and MD , are 114 and 118 amu, respec- 8 QUANTUM-THEORETICAL
tively, and the moments of inertia IX , IY and IZ are 440.7 CALCULATIONS OF SPECTRA
(465.3), 351.7 (357.8) and 89.0 (107.5) amu A2 . The d4 val-
ues are given in parentheses. In the A1g species there are Applied quantum theory has reached the stage where IR and
no translations or rotations (see D2h table in Appendix 1) Raman spectra can be predicted to a very useful degree of
and, as both the x and y coordinates generate representa- accuracy for molecules of quite substantial size. The abil-
tions of this species, the product ratio is mH /mD D 1.997. ity to program the calculations so that force constants and
For the B3u species, only the x coordinates generate a dipole gradients are calculated analytically from the wave-
functions is one of the features that has greatly enhanced
11 12
these computations. The calculations as routinely performed
y7
are of the harmonic wavenumbers and the band intensities.
5 6
10 y2 7 At the Hartree Fock level (no electron configuration inter-
4 1 x7 z7 action allowed for), the wavenumbers are too high when
3 2
x2 z2 compared with experiment due to the restriction of elec-
tron pairing. This leads to dissociation to states in which
9 8 electron pairing is still present, and these are generally of
Figure 7. An atom-centered Cartesian coordinate system for 1,4- much higher energy. Two other factors act in the opposite
difluorobenzene. direction but generally fail to counteract the paired electron
restriction. The lack of electron correlation will lead to correlation. Physicists, however, paid more attention to
shorter bonds and hence higher frequencies and the second an approach based on electron density theory (density
factor is due perhaps to comparing the computed har- functional theory or DFT).32 The essential concept is that
monic wavenumbers with experimental anharmonic values. there is a 1 : 1 correspondence between the electron density
Very good results are obtained simply by multiplying the and the molecular wavefunction. It is inappropriate to
wavenumbers (frequencies) by a scaling factor.28 With a expand on the basis of DFT here. We will simply note
split valence basis, such as 6-31G*, the factor needed is that electron correlation is inherently built in and that
about 0.89. Improved results, with scaling constants nearer the first ionized state is well approximated by the highest
to unity, are obtained with perturbation calculations of the occupied molecular orbital. The latter it will be appreciated
electron correlation correction (Moller Plesset theory),29 but removes much of the problem with the calculation of
these calculations are much more time consuming and are vibrational frequencies in the Hartree Fock methodology.
not feasible with good electron basis sets and more than DFT now permits very good frequencies to be obtained
about 12 atoms. With the 6-31G* basis and use of second- in times almost as short as those needed for Hartree Fock
order Moller Plesset theory, the factor increases to about computations.
0.94.30 Root mean square errors of about 50 cm1 are found Both Raman and IR intensities are much more dif-
at this level. ficult to compute accurately. This is not too surprising
Significant improvements are made by scaling stretching when it is seen that dipole moments are extremely sen-
motions differently from bending motions. This, however, sitive to the basis sets. Although there are cases where the
leads us to a complication that often is not fully appreciated. results are as yet disappointing, DFT appears to be yield-
In virtually all ab initio packages the calculations are carried ing much improved intensities and dipole moments over
out in an inertial Cartesian framework. To appreciate what earlier Hartree Fock and even possibly over Moller Ples-
this means consider a calculation on the H2 O molecule set calculations. Once again, the effect of carrying out the
carried out in the atomic-centered Cartesian framework of calculations in an inertial frame must be taken into account
Figure 1. A movement of the hydrogen atom H3 in the when seeking to interpret the dipole gradients with respect
z- or x-axes will cause the inertial axes to move relative to atomic Cartesian displacements as presented. For H2 O
to the bond OH2 , even though that bond has not been a large value of @m/@y (Figure 1 coordinates) is generated.
distorted. The force constant for displacing H3 then is In a framework locked to the molecule, a displacement in
different from that which exists in a framework locked to the y direction is a pure rotation and indeed it is easy to
the original molecule fixed axes. We would not get the prove that the derivative is the dipole change resulting from
correct wavenumber for the OH stretch by using the force the molecular reorientation. Even for the in-plane motions
constant derived by moving the hydrogen atom along the of the hydrogen atom, large contributions arise from the
OH bond due to this inertial rotation. This molecule is a movement of the inertial axes with respect to the initial
rather extreme example due to the center of gravity lying direction of the molecular dipole. It is easy to visualize
very close to the oxygen atom. Even a small displacement that in the symmetric OH stretch no reorientation of the
of H3 will cause the inertial axes to rotate significantly. inertial axes occurs. However, for the antisymmetric stretch
To evaluate the force field for valence type coordinates the contribution is huge. The necessity of considering the
(stretches, angle bends and torsions), F R , in terms of contribution of the molecular dipole moment to the vibra-
the Cartesian force constant matrix, F C , the A matrix, tional transition moment whenever the vibrations belong
x D AR, has to be invoked. This leads to to the same representation as a rotation was first proposed
and formulated by Crawford.25 Recently the author has pub-
F R D G1 BM1 F C M1 Bt G1 45
lished software for evaluating the correction.33 For further
Generally the results of vibrational frequency and force field information, see Calculation of Vibrational Frequencies
calculations are very satisfactory, although some difficulties by HartreeFock-based and Density Functional Theory.
remain to be resolved. For example, both the experimen-
tal structures and spectra of nitrosyl fluoride are badly
reproduced, even with relatively sophisticated configura- 9 VIBRATIONROTATION SPECTRA
tion interaction calculations.27 This particular case seems
to be typical of the difficulties that arise when a fluorine is When a vibrational quantum state changes, other quan-
attached to an oxygen, which is itself involved in p bonding. tum states (for example, rotational) of lower energy will
Theoretical chemists until about 1980 focused attention also generally change. Thus for a heteronuclear diatomic
almost entirely on the Hartree Fock approaches to quantum molecule the j0i to j1i transition is accompanied by
theory, followed by various methods of including electron a change in rotational quantum number of 1 leading
to the well-known P and R branches. For polyatomic S = = S= = S = = 54

molecules the selection rules required so as to allow the
non-disappearance of the transition integral are more com-
Type A
plex, and for details the reader is referred to a specialist text,
such as that by Allen and Cross.34 In the situation where
individual rovibrational transitions cannot be resolved in an
asymmetric top, then the overall contour can be related to
Type B
the direction of the transition moment with respect to the
inertial axes and to the moments themselves. Thus transi-
tions in which the dipole change occurs along the direction
of greatest moment of inertia are characterized by a strong
Q branch (type C bands), corresponding to no change in the Type C
rotational quantum number change (see Figure 8). When
the dipole change is along the axis of intermediate moment
of inertia, the band contour is characterized by a split Q 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
branch with broad wings (type B). For transition moments Figure 9. Simulated type A, B, and C contours of an asymmetric
along the axis of least moment of inertia, a strong P, Q and top. Only half of the band is shown. Half of the total Q branch
intensity is denoted by a dotted rectangle. [Reproduced from
R structure exists. The contours have been calculated as a
R.M. Badger and L.R. Zumwalt, J. Chem. Phys., 6, 711 (1938)].
function of the moments of inertia and presented in easily S and r are defined by
used figures.35 These can be invaluable in identifying the

symmetry of vibrations when measurements are feasible in 2 1 1 1 1
SD
the gaseous phase (e.g. see Figure 9). IB IA IC IA IC

With laser and high-resolution Fourier instruments, an 1 1
rD IB
increasing amount of detail of rovibrational information is IA IC
becoming available. Here we shall refer to a few general
points. The parameters of fitting include the moments ground state at the absolute zero of temperature each
of inertia. Even with small molecules and isotopically quantum state has a half quantum of energy (ZPE, see
substituted molecules it is very difficult to generate accurate Figure 3). The average of the root mean square bond length
molecular geometries. The inertial constants refer to the during a vibration differs significantly from the equilibrium
ground and vibrationally excited states, but even in the bond length. Ab initio calculations, such as referred to in
the previous section, may be used to estimate the necessary
C corrections.
One feature unique to vibration rotation interaction is due
to Coriolis interaction. There is a component of force on a
moving particle due to rotation about a perpendicular axis
equal to
! ! mv 46
where represents the vector product operator, ! is the
Absorption
C A
molecular angular velocity, and mi and vi are the mass and
B velocity of the ith particle. The vector products in equa-
tion (46) are taken over all particles. Such coupling leads
to interaction between two states, the product of whose
symmetry representations belongs to the same species as
a molecular rotation. In Figure 9 the three fundamental
modes of a linear triatomic molecule are shown with the
atomic motions designated by solid lines. With the direction
800 700 of rotation as shown in Figure 10 the Coriolis forces are in
cm1 the directions of the dotted lines. For the symmetric stretch
Figure 8. Part of the IR absorption spectrum of 2,6-difluoro- the forces are in the direction countering rotation and of no
pyridine as measured in the vapor phase showing typical A, B
and C contours. [Reproduced from R.T. Bailey and D. Steele, direct interest to us. For n2 and n3 , however, the forces cou-
Spectrochim. Acta A, 23, 2997 (1967), by permission of Elsevier ple directly with the other mode. This force has substantial
Science.] impact on the vibrationrotational energy levels when the
Since the Cartesian displacements can be calculated from

Direction of rotation the L matrices, it follows that it is straightforward to
compute the zeta values. For further study of these we
refer to the literature.34,37 The same phenomenon, of course,
Y X Y 1 controls the contrary direction of rotation of air masses in
the northern and southern hemispheres.
In this section we have focused solely on vibrationrota-
2
tion interaction as it directly affects the vibrational spec-
Y X Y
trum. The specialist texts should be consulted for high-
resolution vibrationrotation band structure. There is one
further aspect of Coriolis interaction that has a direct bear-
Y X Y 3
ing on the subject matter of this article. Second-order Cori-
olis interaction between neighboring bands will deplete, or
Figure 10. The Coriolis forces (dotted lines) resulting from the enhance, the intensities of the rotational lines on the sides
fundamental modes of a linear triatomic molecule. closest to one another depending on whether the product
107 of the Coriolis constant and the transition dipoles of the
399
2 two interacting bands is positive or negative. If it is posi-

tive, then the adjacent branches are enhanced. This can be
= 0.75 = 0.6 = 0.5 a very useful source of information on the relative signs
1 of the transition moments, the signs not being determinable
from band intensities alone.38
0 In such a multi-authored text as this, some duplication
11.6 10.9 10.6 is inevitable. This is usually a good thing in that dif-
2
ferent viewpoints are expressed and different approaches
may enlighten the reader. The BeerLambert law (Beers
= 0.25 = 0.0 Observed
1 Law), vibrational calculations (Normal Coordinate Anal-
ysis) and group frequencies (SpectraStructure Corre-
lations in the Near-infrared) are all expounded in more
0 detail in sections of their own.
9.2 8.0 10.5
Figure 11. Computed and observed band contours for the
315 cm1 band of hexafluorobenzene. [Reproduced from D. Steele ABBREVIATIONS AND ACRONYMS
and W. Wheatley, J. Mol. Spectrosc., 32, 265 (1969), by
permission of Academic Press.]
PFA Principal Factor Analysis
vibrational energy states are close to one another. The ZPE Zero Point Energy
extreme case of coupling is when the two vibrations have
exactly the same unperturbed frequencies, in other words,
LIST OF SYMBOLS
they are degenerate. For degenerate vibrations, this can be
a first-order effect and has dramatic impact on the overall A Integrated absorbance
vibrational rotational contour. Simulation of the vibrational Bif Probability of an induced transition per unit time
rotational band contour as functions of the Coriolis con- per unit radiation density
stants have been made by Edgell and Moynihan,36 and c Velocity of light
these can be used to estimate the Coriolis constant from the C Concentration
experimental contour. In Figure 11 we show the computed Ei Energy of ith mode
contours computed for the out-of-plane A2u mode of hex- f Quadratic force constant
afluorobenzene as a function of the Coriolis constant zeta F Quadratic force constant matrix
(z). Zeta can take values between C1 and 1 and is related
G Inverse kinetic energy matrix
to the Cartesian displacement matrix, x by zs D xt Rs x
h Plancks constant
where s D x, y or z and Rx for example is given by
kB Boltzmanns constant

0 0 0 l Path length

R D 0 0 1 47 M Molecular mass
0 1 0 Q Normal mode
R Internal coordinate 18. R.G. Gordon, J. Chem. Phys., 42, 3658 (1965).
V Vibrational quantum number 19. R.G. Gordon, J. Chem. Phys., 43, 1307 (1965).
Integrated absorption intensity 20. B.J. Berne and R. Pecora, Dynamic Light Scattering, John
m Dipole moment Wiley & Sons, Chichester (1976).
n Observed frequency 21. I.R. Hill and D. Steele, J. Chem. Soc., Faraday Trans. 2, 72,
nQ Wavenumber 555 (1975).
r Radiation density 22. R.G. Brewer and R.L. Shoemaker, Phys. Rev. Lett., 27, 631
! Harmonic frequency (1971).
23. E. Bright Wilson, J.C. Decius and P.C. Cross, Molecular
Vibrations, McGraw-Hill, New York (1955).
REFERENCES
24. D. Steele, Theory of Vibrational Spectroscopy, W.B. Saun-
ders Co., Philadelphia (1971).
1. F.A. Cotton, Chemical Applications of Group Theory, 3rd
edition, John Wiley & Sons, New York (1990). 25. B. Crawford, Jr, J. Chem. Phys., 20, 977 (1952).
2. H.C. Longuet Higgins, Mol. Phys., 6, 445 (1963). 26. L.A. Gribov and W.J. Orville-Thomas, Theory and Methods
of Calculation of Molecular Spectra, John Wiley & Sons,
3. H.H. Claassen, B. Weinstock and J.G. Malm, J. Chem. Phys.,
Chichester (1988).
25, 426 (1956).
27. W.J. Hehre, L. Radom, P.V.R. Schleyer and J. Pople, Ab
4. I.M. Mills and A.G. Robiette, Mol. Phys., 56, 743 (1985).
Initio Molecular Orbital Theory, John Wiley & Sons, New
5. M.S. Child, Acc. Chem. Res., 18, 45 (1985). York (1985).
6. M.E. Parker, D. Steele and M.J.C. Smith, J. Phys. Chem. A, 28. C.E. Blom and C. Altona, Mol. Phys., 31, 1377 (1976).
101, 9618 (1997).
29. C. Mller and M.S. Plesset, Phys. Rev., 46, 618
7. H.G. Kjaergaard and B.R. Henry, J. Chem. Phys., 94, 5844 (1934).
(1991).
30. J.A. Pople, A.P. Scott, M.W. Wong and L. Radom, Israel J.
8. H. Eyring, J. Walter and G.E. Kimball, Quantum Chem- Chem., 33, 345 (1993).
istry, Wiley & Sons, Chapman & Hall (1944).
31. P.N. Noble, J. Phys. Chem., 95, 4695 (1991).
9. W. Heitler, The Quantum Theory of Radiation, 3rd edition,
Oxford University Press, Oxford (1954). 32. E.J. Baerends and O.V. Gritsenko, J. Phys. Chem. A, 101,
5383 (1997).
10. P.W. Atkins, Molecular Quantum Mechanics, 2nd edition,
Oxford University Press, Oxford, 47 (1983). 33. T.A. Ford and D. Steele, Spectrochim. Acta, A55, 2823
11. F.E. Dunstan and D.H. Whiffen, J. Chem. Soc., 5221 (1960). (1999) (corrigendum 56, 1851, 2000).
12. J.W. Russell, C.D. Needham and J. Overend, J. Chem. Phys., 34. H.C. Allen and P.C. Cross, Molecular Vib-Rotors, John
45, 3383 (1966). Wiley & Sons, Chichester (1963).
13. A. Kindness, D.C. McKean and D. Stewart, J. Molec. Struct., 35. R.M. Badger and L.R. Zumvalt, J. Chem. Phys., 6, 711
224, 363 (1990). (1938).
14. D. Steele and W. Wheatley, J. Mol. Phys., 32, 265 (1969). 36. W.F. Edgell and R.E. Moynihan, J. Chem. Phys., 45, 1205
(1966).
15. A.S. Wexler, Appl. Spectrosc., 1, 29 (1967).
37. D. Papousek and M.R. Aliev, Molecular Vibrational- Rota-
16. H. Martens and T. Naes, Multivariate Calibration, John
tional Spectra, Studies in Physical and Theoretical Chem-
Wiley & Sons, Chichester (1989).
istry, Elsevier, Vol. 17 (1982).
17. P.M. Fredericks, J.B. Lee, P.R. Osborn and D.A.J. Swinkels,
Appl. Spectrosc., 39, 303, 311 (1985). 38. I.M. Mills, Pure Appl. Chem., 11, 325 (1965).
APPENDIX 1
Labelling of the Cartesian axes used to define internal Co., 1945). For many symmetry groups this suffices. How-
nuclear distortions is quite arbitrary in the absence of spe- ever, in other cases choices still exist and confusion in
cific artificial rules. If different axes are chosen by different vibrational labelling has resulted. For example, three differ-
people then confusion as to class species nomenclature fol- ent authors have used the three possible sets of axes labels
lows. One rule that is universally followed is that molecular for p-difluorobenzene. This results in the same vibrations
axes shall be labelled so as not to conflict with the group being assigned, for example, to B1g , B2g and B3g sym-
character tables listed in Herzbergs book Infrared and metry species of the D2h symmetry group. This chaotic
Raman Spectra of Polyatomic Molecules (Van Nostrand state of affairs led to the compiling of a special report in
which recommendations were made to standardize spec- Recommendation 5(b). For planar D2 h molecules the
tral notation [R.S. Mulliken, J. Chem. Phys., 23, 1997 x-axis shall be chosen perpendicular to the plane of the
(1955)]. Some of the more pertinent recommendations are molecule and the z-axis so that it passes through the greatest
reproduced here. number of atoms, or if this rule is not decisive, so that it cuts
Recommendation 3. Species symbols and their definitions the greatest number of bonds.
as given by Herzberg shall be adopted as standard for Recommendation 5(c). For planar D4 h and D6 h systems
nonlinear molecules. Thus for the C2V group the z-axis the C2 and C2 operations shall be distinguished by specify-
is always chosen as the C2 axis so that we have symmetry ing that (a) the C2 axes shall pass greater number of atoms
planes sxz and syz . or (b) in case specification (a) is not decisive, shall intersect
Recommendation 4. Every author shall define his species the greater number of bonds.
nomenclature it terms of the specific geometry of the mole- Recommendation 5(d). sn and sd operations shall be
cules in each paper whenever confusion may occur. He shall distinguished by specifying that (a) the sn planes shall pass
make an effort to follow previous usage if such confusion through the greater number of atoms which are in the square
exists and there is no strong reason to change. or hexagonal array or (b) in case specification (a) is not
Recommendation 5(a). For planar C2 V molecules the x- decisive, shall intersect the greater number of bonds among
axis shall always be chosen perpendicular to the plane of such atoms.
the molecule. (The z-axis is always taken as the C2 axis If recommendations 5(a) to 5(d) do not remove ambigui-
following Herzberg.) ties recommendation 4 must always be adhered to.
Groups Cs Ci and Cn (n D 2, 3, 4, 5, 6)
Cs E sxy
A0 1 1 Tx ; Ty ; Rz x 2 ; y 2 ; z2 ; xy
A00 1 1 Tz ; Rx ; Ry xz; yz
Ci E i
Ag 1 1 Rx ; R y ; R z x 2 ; y 2 ; z2 ; xy; xz; yz
Au 1 1 Tx ; Ty ; Tz
C2 E C2
A 1 1 Tz ; Rz x 2 ; y 2 ; z2 ; xy
B 1 1 Tx ; Ty ; Rx ; Ry xz; yz
C3 E C3 C23 e D exp2pi/3
A 1 1 1 Tz ; Rz x 2 C y 2 ; z2
1 e e
E (Tx , Ty ); (Rx , Ry ) x 2 y 2 , xy; xz, yz
1 e e
C4 E C4 C2 C34
A 1 1 1 1 Tz ; Rz x 2 C y 2 ; z2
B 1 1 1 1 x 2 y 2 ; xy
1 i 1 i
E (Tx , Ty ); (Rx , Ry ) (xz, yz)
1 i 1 i
C5 E C5 C25 C35 C45 e D exp2pi/5
A 1 1 1 1 1 Tz ; Rz x 2 C y 2 ; z2
1 e e2 e2 e
E1 (Tx , Ty ); (Rx , Ry ) (xz, yz)
1 e e2 e2 e
1 e2 e e e2
E2 x 2 y 2 , xy
1 e2 e e e2
C6 E C6 C3 C2 C23 C56 e D exp2pi/6

A 1 1 1 1 1 1 Tz ; Rz x 2 C y 2 ; z2
B 1 1 1 1 1 1
1 e e 1 e e
E1 (Tx , Ty ); Rx , Ry ) (xz, yz)
1 e e 1 e e
1 e e 1 e e
E2
1 e e 1 e e x 2 y 2 , xy
Groups Cnv (n D 2, 3, 4, 5, 6)
C2v E C2 sv xz sv yz

A1 1 1 1 1 Tz x 2 ; y 2 ; z2
A2 1 1 1 1 Rz xy
B1 1 1 1 1 Tx ; Ry xz
B2 1 1 1 1 Ty ; Rx yz
C3v E 2C3 3sv
A1 1 1 1 Tz x 2 C y 2 ; z2
A2 1 1 1 Rz
E 2 1 0 (Tx , Ty ); (Rx , Ry ) x 2 y 2 , xy; xz, yz
C4v E 2C4 C2 2sv 2sd
A1 1 1 1 1 1 Tz x 2 C y 2 ; z2
A2 1 1 1 1 1 Rz
B1 1 1 1 1 1 x2 y 2
B2 1 1 1 1 1 xy
E 2 0 2 0 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
C5v E 2C5 2C25 5sv a D 72
A1 1 1 1 1 Tz x 2 C y 2 ; z2
A2 1 1 1 1 Rz
E1 2 2 cos a 2 cos 2a 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
E2 2 2 cos 2a 2 cos a 0 x 2 y 2 , xy
C6v E 2C6 2C3 C2 3sv 3sd
A1 1 1 1 1 1 1 Tz x 2 C y 2 ; z2
A2 1 1 1 1 1 1 Rz
B1 1 1 1 1 1 1
B2 1 1 1 1 1 1
E1 2 1 1 2 0 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
E2 2 1 1 2 0 0 (x y 2 , xy)
2
Groups Cnh (n D 2, 3, 4, 5, 6)
C2h E C2 i sh
Ag 1 1 1 1 Rz x 2 ; y 2 ; z2 ; xy
Bg 1 1 1 1 Rx ; R y xz; yz
Au 1 1 1 1 Tz
Bu 1 1 1 1 Tx ; Ty
C3h E C3 C23 sh S3 S35 e D exp2pi/3

A 0
1 1 1 1 1 1 Rz x 2 C y 2 ; z2
1 e e 1 e e
E0 (Tx , Ty ) x 2 y 2 , xy
1 e e 1 e e
A00 1 1 1 1 1 1 Tz
00 1 e e 1 e e
E (Rx , Ry ) (xz, yz)
1 e e 1 e e
C4h E C4 C2 C34 i S43 sh S4
Ag 1 1 1 1 1 1 1 1 Rz x 2 C y 2 ; z2
Bg 1 1 1 1 1 1 1 1 x 2 y 2 ; xy
1 i 1 i 1 i 1 i
Eg (Rx , Ry ) (xz, yz)
1 i 1 i 1 i 1 i
Au 1 1 1 1 1 1 1 1 Tz
Bu 1 1 1 1 1 1 1 1
1 i 1 i 1 i 1 i
Eu (Tx , Ty )
1 i 1 i 1 i 1 i
C5h E C5 C25 C35 C45 sh S5 S57 S53 S59 e D exp2pi/5
A0 1 1 1 1 1 1 1 1 1 1 Rz x 2 C y 2 ; z2
1 e e2 e2 e 1 e e2 e2 e
E01 (Tz , Ty )
1 e e2 e2 e 1 e e2 e2 e
1 e2 e e e2 1 e2 e e e2
E02 (x 2 y 2 , xy)
1 e2 e e e2 1 e2 e e e2
A00 1 1 1 1 1 1 1 1 1 1 Tz
1 e e2 e2 e 1 e e2 e2 e
E001 (Rx , Ry ) (xz, yz)
1 e e2 e2 e 1 e e2 e2 e
1 e2 e e e2 1 e2 e e e2
E002
1 e2 e e e2 1 e2 e e e2
C6h E C6 C3 C2 C23 C56 i S35 S65 ah S6 S3 e D exp2pi/6
Ag 1 1 1 1 1 1 1 1 1 1 1 1 Rz x 2 C y 2 ; z2
Bg 1 1 1 1 1 1 1 1 1 1 1 1
1 e e 1 e e 1 e e 1 e e
E1g (Rx , Ry ) (xz, yz)
1 e e 1 e e 1 e e 1 e e
1 e e 1 e e 1 e e 1 e e
E2g (x 2 y 2 , xy)
1 e e 1 e e 1 e e 1 e e
Au 1 1 1 1 1 1 1 1 1 1 1 1 Tz
Bu 1 1 1 1 1 1 1 1 1 1 1 1
1 e e 1 e e 1 e e 1 e e
E1u (Tx , Ty )
1 e e 1 e e 1 e e 1 e e
1 e e 1 e e 1 e e 1 e e
E2u
1 e e 1 e e 1 e e 1 e e
Groups Dn (n D 2, 3, 4, 5, 6)
D2 V E C2 z C2 y C2 x

A 1 1 1 1 x 2 ; y 2 ; z2
B1 1 1 1 1 Tz ; Rz xy
B2 1 1 1 1 Ty ; Ry xz
B3 1 1 1 1 Tx ; Rx yz
D3 E 2C3 3C02
A1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 Tz ; Rz
E 2 1 0 (Tx , Ty ); (Rx , Ry ) (x 2 y 2 , xy); (xz, yz)
D4 E 2C4 C2 2C02 2C002
A1 1 1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 1 1 Tz ; Rz
B1 1 1 1 1 1 x2 y 2
B2 1 1 1 1 1 xy
E 2 0 2 0 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
D5 E 2C5 2C25 5C02 a D 72
A1 1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 1 Tz ; Rz
E1 2 2 cos a 2 cos 2a 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
E2 2 2 cos 2a 2 cos a 0 (x 2 y 2 , xy)
D6 E 2C6 2C3 C2 3C02 3C002
A1 1 1 1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 1 1 1 Tz ; Rz
B1 1 1 1 1 1 1
B2 1 1 1 1 1 1
E1 2 1 1 2 0 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
E2 2 1 1 2 0 0 (x y 2 , xy)
2
Groups Dnh (n D 2, 3, 4, 5, 6)
D2h Vh E C2 z C2 y C2 x i sxy sxz syz

Ag 1 1 1 1 1 1 1 1 x 2 ; y 2 ; z2
B1g 1 1 1 1 1 1 1 1 Rz xy
B2g 1 1 1 1 1 1 1 1 Ry xz
B3g 1 1 1 1 1 1 1 1 Rx yz
Au 1 1 1 1 1 1 1 1
B1u 1 1 1 1 1 1 1 1 Tz
B2u 1 1 1 1 1 1 1 1 Ty
B3u 1 1 1 1 1 1 1 1 Tx
D3h E 2C3 3C02 sh 2S3 3sv
A01 1 1 1 1 1 1 x 2 C y 2 ; z2
A02 1 1 1 1 1 1 Rz
E0 2 1 0 2 1 0 (Tx , Ty ) (x 2 y 2 , xy)
A001 1 1 1 1 1 1
A002 1 1 1 1 1 1 Tz
E00 2 1 0 2 1 0 (Rx , Ry ) (xz, yz)
D4h E 2C4 C2 2C02 2C002 i 2S4 sh 2sv 2sd

A1g 1 1 1 1 1 1 1 1 1 1 x 2 C y 2 ; z2
A2g 1 1 1 1 1 1 1 1 1 1 Rz
B1g 1 1 1 1 1 1 1 1 1 1 x2 y 2
B2g 1 1 1 1 1 1 1 1 1 1 xy
Eg 2 0 2 0 0 2 0 2 0 0 (Rx , Ry ) (xz, yz)
A1u 1 1 1 1 1 1 1 1 1 1
A2u 1 1 1 1 1 1 1 1 1 1 Tz
B1u 1 1 1 1 1 1 1 1 1 1
B2u 1 1 1 1 1 1 1 1 1 1
Eu 2 0 2 0 0 2 0 2 0 0 (Tx , Ty )
D5h E 2C5 2C25 5C02 sh 2S5 2S53 5sv a D 72
A01 1 1 1 1 1 1 1 1 x 2 C y 2 ; z2
A02 1 1 1 1 1 1 1 1 Rz
E01 2 2 cos a 2 cos 2a 0 2 2 cos a 2 cos 2a 0 (Tx , Ty )
E02 2 2 cos 2a 2 cos a 0 2 2 cos 2a 2 cos a 0 x 2 y 2 , xy
A001 1 1 1 1 1 1 1 1
A002 1 1 1 1 1 1 1 1 Tz
E001 2 2 cos a 2 cos 2a 0 2 2 cos a 2 cos 2a 0 (Rx , Ry ) (xz, yz)
E002 2 2 cos 2a 2 cos a 0 2 2 cos 2a 2 cos a 0
D6h E 2C6 2C3 C2 3C02 3C002 i 2S3 2S6 sh 3sd 3sv
A1g 1 1 1 1 1 1 1 1 1 1 1 1 x 2 C y 2 ; z2
A2g 1 1 1 1 1 1 1 1 1 1 1 1 Rz
B1g 1 1 1 1 1 1 1 1 1 1 1 1
B2g 1 1 1 1 1 1 1 1 1 1 1 1
E1g 2 1 1 2 0 0 2 1 1 2 0 0 (Rx , Ry ) (xz, yz)
E2g 2 1 1 2 0 0 2 1 1 2 0 0 (x 2 y 2 , xy)
A1u 1 1 1 1 1 1 1 1 1 1 1 1
A2u 1 1 1 1 1 1 1 1 1 1 1 1 Tz
B1u 1 1 1 1 1 1 1 1 1 1 1 1
B2u 1 1 1 1 1 1 1 1 1 1 1 1
E1u 2 1 1 2 0 0 2 1 1 2 0 0 (Tx , Ty )
E2u 2 1 1 2 0 0 2 1 1 2 0 0
Groups Dnd n D 2, 3, 4, 5, 6
D2d Vd E 2S4 C2 2C02 2sd

A1 1 1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 1 1 Rz
B1 1 1 1 1 1 x2 y 2
B2 1 1 1 1 1 Tz xy
E 2 0 2 0 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
D3d E 2C3 3C02 i 2S6 3sd
A1g 1 1 1 1 1 1 x 2 C y 2 ; z2
A2g 1 1 1 1 1 1 Rz
Eg 2 1 0 2 1 0 (Rx , Ry ) (x 2 y 2 , xy); (xz, yz)
A1u 1 1 1 1 1 1
A2u 1 1 1 1 1 1 Tz
Eu 2 1 0 2 1 0 (Tx , Ty )
D4d E 2S8 2C4 2S83 C2 4C02 4sd

A1 1 1 1 1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 1 1 1 1 Rz
B1 1 1 1 1 1 1 1
B2 1 1
p 1 1
p 1 1 1 Tz
E1 2 2 0 2 2 0 0 (Tx , Ty )
E2 2 p 0 2 p0 2 0 0 (x 2 y 2 , xy)
E3 2 2 0 2 2 0 0 (Rx , Ry ) (xz, yz)
D5d E 2C5 2C25 5C02 i 3
2S10 2S10 5sd a D 72
A1g 1 1 1 1 1 1 1 1 x 2 C y 2 ; z2
A2g 1 1 1 1 1 1 1 1 Rz
E1g 2 2 cos a 2 cos 2a 0 2 2 cos a 2 cos 2a 0 (Rx , Ry ) (xz, yz)
E2g 2 2 cos 2a 2 cos a 0 2 2 cos 2a 2 cos a 0
A1u 1 1 1 1 1 1 1 1
A2u 1 1 1 1 1 1 1 1 Tz
E1u 2 2 cos a 2 cos 2a 0 2 2 cos a 2 cos 2a 0 (Tx , Ty ) (x 2 y 2 , xy)
E2u 2 2 cos 2a 2 cos a 0 2 2 cos 2a 2 cos a 0
5
D6d E 2S12 2C6 2S4 2C3 2S12 C2 6C02 6sd
A1 1 1 1 1 1 1 1 1 1 x 2 C y 2 ; z2
A2 1 1 1 1 1 1 1 1 1 Rz
B1 1 1 1 1 1 1 1 1 1
B2 1 1
p 1 1 1 1
p 1 1 1 Tz
E1 2 3 1 0 1 3 2 0 0 (Tx , Ty )
E2 2 1 1 2 1 1 2 0 0 (x 2 y 2 , xy)
E3 2 0 2 0 2 0 2 0 0
E4 2 1
p 1 2 1 1
p 2 0 0
E5 2 3 1 0 1 3 2 0 0 Rx , Ry (xz, yz)
Groups Sn (n D 4, 6, 8)
S4 E S4 C2 S43
A 1 1 1 1 Rz x 2 C y 2 ; z2
B 1 1 1 1 Tz x 2 y 2 ; xy
1 i 1 i
E (Tx , Ty ); (Rx , Ry ) (xz, yz)
1 i 1 i
S6 E C3 C23 i S65 S6 e D exp2pi/3
Ag 1 1 1 1 1 1 Rz x 2 C y 2 ; z2
1 e e 1 e e
Eg (Rx , Ry ) (x 2 y 2 , xy); (xz, yz)
1 e e 1 e e
Au 1 1 1 1 1 1 Tz
1 e e 1 e e
Eu (Tx , Ty )
1 e e 1 e e
S8 E S8 C4 S83 C2 S85 C34 S87 e D exp2pi/8

A 1 1 1 1 1 1 1 1 Rz x 2 C y 2 ; z2
B 1 1 1 1 1 1 1 1 Tz
1 e i e 1 e i e
E1 (Tx , Ty )
1 e i e 1 e i e
1 i 1 i 1 i 1 i
E2 (x 2 y 2 , xy)
1 i 1 i 1 i 1 i
1 e i e 1 e i e
E3 (Rx , Ry ) (xz, yz)
1 e i e 1 e i e
Miscellaneous groups
Td E 8C3 3C2 6S4 6sd

A1 1 1 1 1 1 x 2 C y 2 C z2
A2 1 1 1 1 1
E 2 1 2 0 0 (2z2 x 2 y 2 , x 2 y 2 )
T1 , F1 3 0 1 1 1 (Rx , Ry , Rz )
T2 , F2 3 0 1 1 1 (Tx , Ty , Tz ) (xy, xz, yz)
O E 8C3 3C2 6C4 6C02
A1 1 1 1 1 1 x 2 C y 2 C z2
A2 1 1 1 1 1
E 2 1 2 0 0 (2z2 x 2 y 2 , x 2 y 2 )
T1 , F1 3 0 1 1 1 (Rx , Ry , Rz );
(Tx , Ty , Tz )
T2 , F2 3 0 1 1 1 (xy, xz, yz)
Oh E 8C3 3C2 6C4 6C02 i 8S6 3sh 6S4 6sd
A1g 1 1 1 1 1 1 1 1 1 1 x 2 C y 2 C z2
A2g 1 1 1 1 1 1 1 1 1 1
Eg 2 1 2 0 0 2 1 2 0 0 2z2 x 2 y 2
x2 y 2)
T1g , F1g 3 0 1 1 1 3 0 1 1 1 Rx , Ry , Rz
T2g , F2g 3 0 1 1 1 3 0 1 1 1 xy, xz, yz
A1u 1 1 1 1 1 1 1 1 1 1
A2u 1 1 1 1 1 1 1 1 1 1
Eu 2 1 2 0 0 2 1 2 0 0
T1u , F1u 3 0 1 1 1 3 0 1 1 1 Tx , Ty , Tz
T2u , F2u 3 0 1 1 1 3 0 1 1 1
C 1v E 2Cf 1s v
A1 D C 1 1 1 Tz x 2 C y 2 ; z2
A2 D 1 1 1 Rz
E1 D 2 2 cos f 0 (Tx , Ty ); (Rx , Ry ) (xz, yz)
E2 D 2 2 cos 2 f 0 (x 2 y 2 , xy)
E3 D 2 2 cos 3 f 0

D1h E 2Cf 1s v i 2Sf 1C02

A1g D C
g 1 1 1 1 1 1 x 2 C y z ; z2
A2g D
g 1 1 1 1 1 1 Rz
E1g D g 2 2 cos f 0 2 2 cos f 0 Rx , Ry xz, yz
E2g D g 2 2 cos 2 f 0 2 2 cos 2 f 0 x y 2 , xy
2

A1u D C
u 1 1 1 1 1 1 Tz
A2u D
u 1 1 1 1 1 1
E1u D u 2 2 cos f 0 2 2 cos f 0 Tx , Ty
E2u D u 2 2 cos 2 f 0 2 2 cos 2 f 0

Raman Spectroscopy: Theory
Gabor Keresztury
Hungarian Academy of Sciences, Budapest, Hungary
1 INTRODUCTION of electromagnetic radiation and quantum theory. More

in-depth treatments are available elsewhere.5,6
As has been shown in previous articles, the most straightfor-
ward way of observing vibrational transitions of molecules
is through absorption (or emission) of infrared (IR) radia- 1.1 The Raman effect phenomenological
tion. The interaction of matter with electromagnetic radi- description
ation is subject to selection rules, hence not all vibrations
may give rise to absorption (or emission) bands in the IR When a transparent liquid, gas or crystal is illuminated with
spectrum. The prerequisite of IR activity is the change in a beam of monochromatic light of wavenumber nQ 0 , most
dipole moment of the molecule during a given normal mode of the incident light is transmitted without change, while
of vibration. Because of this the stretching vibration of a a small portion of it is scattered within the whole solid
homonuclear diatomic molecule (such as H2 , O2 , etc.) or angle. (Scattering refers to light deflected from the origi-
the totally symmetric vibrations of larger molecules having nal direction of propagation of the incident light.) Spectral
a center of symmetry, as well as any other vibrations that analysis of the scattered light shows that, in addition to
do not generate a vibrating electric dipole, cannot absorb scattering without change of wavenumber of the incident
or emit electromagnetic radiation; these vibrations are said light (Rayleigh scattering), it contains also discrete com-
to be IR inactive. In order to detect such vibrations, it ponents of altered wavenumber. In general, there are pairs
is necessary to consider the inelastic scattering of light of new lines appearing in the spectrum at wavenumbers
a phenomenon usually referred to as Raman scattering positioned symmetrically with respect to the Rayleigh line,
(described in Russian literature as combination scattering). i.e. nQ 0 D nQ 0 nQ M , where the wavenumbers nQ M correspond
This is governed by different selection rules compared to to transitions between rotational or vibrational (rarely elec-
the absorption or spontaneous emission of light. tronic) energy levels of molecular systems. The appearance
The possibility of light scattering with significant change of altered frequencies (wavenumbers) in scattered light
of frequency was first predicted on theoretical grounds by is called the Raman effect or Raman scattering, after its
Smekal1 in 1923. The first observations of the phenomenon discoverer.
were made in liquids by Raman and Krishnan,2 then in Raman scattering is a very feeble effect; it is always
crystals by Landsberg and Mandelstam,3 both in 1928. The accompanied with Rayleigh scattering with an intensity usu-
first comprehensive treatise of the underlying theory was ally 35 orders of magnitude greater. Note, however, that
offered by Placzek4 in 1934. Rayleigh scattering itself is only about 104 103 of the
This article gives a short general description of the intensity of the incident exciting radiation. The new com-
phenomenon and properties of radiation by electric dipoles, ponents appearing in the spectrum of the scattered radiation
then the vibrational Raman effect is described covering the at shifted wavenumbers are termed Raman lines or Raman
most important aspects in terms of both classical theory bands, and collectively they are referred to as the Raman
spectrum. The Raman bands at wavenumbers less than the
John Wiley & Sons Ltd, 2002. exciting wavenumber (i.e. nQ 0 nQ M ) are referred to as Stokes
Stokes Anti-Stokes
Raman intensity
O O
Coumarin
50
105
4000 1500 1000 500 0 +500 +1000 +1500 +4000

s (cm1)
15 400 18 000 19 000 20 000 21 000 23 400

(cm1) absolute
650 550 500 430

(nm)
Figure 1. A typical Raman spectrum of a polycrystalline substance (coumarin) showing the Rayleigh line, and both the Stokes and
anti-Stokes sides of the Raman spectrum on wavelength, absolute wavenumber, and relative wavenumber (or Raman shift) scales.
(Reproduced from Schrader7 .)
lines, whereas those appearing at higher wavenumbers (i.e. and its wavelength is several orders of magnitude greater
nQ 0 C nQ M ) as anti-Stokes lines. than the scattering particles (typically molecules).
As an example, the Raman spectrum of an organic com-
pound (coumarin) is shown in Figure 1, with the intense
Rayleigh line positioned in the middle, at nQ 0 , and the Stokes 1.2 Interaction of light and matter:
and anti-Stokes lines to the left and to the right of it, absorption/emission and scattering of light
respectively.7 Note the typical intensity relationships, with
the anti-Stokes lines being considerably weaker and quickly According to the classical theory, the most efficient source
diminishing in intensity with increasing nQ M as compared to of electromagnetic radiation is an oscillating electric dipole:
their Stokes counterparts. The frequency shifts in the Stokes the intensity of its radiation is several orders of magnitude
and anti-Stokes Raman spectra usually correspond to the greater than that of other sources of radiation, such as oscil-
frequencies of the normal modes of molecular vibrations; lating magnetic dipoles, electric quadrupoles, or higher-
thus the shift values are more convenient for characterizing order multipoles. Thus in order to understand the origin
the band positions than the absolute wavenumbers them- of emission/absorption or scattering of light by molecules,
selves. In view of all this, in most applications of Raman we must consider the distribution of electric charges within
spectroscopy only the more readily observable Stokes side the molecule and establish whether there is a permanent or
of the Raman spectrum is considered, plotted in a relative induced electric dipole that could oscillate when modulated
wavenumber scale, as nQ 0 nQ 0 D nQ M , in the range of Raman by the normal vibrations.
shifts from 4000 to 0 cm1 (with the upper and lower limits An electric dipole formed by a pair of point charges q
depending on the actual instrument). In terms of absorption and Cq separated by a distance r is characterized by its
spectroscopy, this corresponds to the IR and far-infrared dipole moment vector m, defined as
(FIR) regions, i.e. to vibrational and rotational transitions
of molecules, including the lattice vibrations of crystals. m D qs 1
It should be noted that, when applying lasers as sources
of incident exciting radiation, Raman spectroscopy is not where s is a vector pointing from q to Cq. If such a dipole
limited to the examination of transparent samples, although oscillates with a frequency n (corresponding to circular
the theory of ordinary Raman scattering considers the case frequency ! D 2pn, or wavenumber nQ D n/c, where c is
when the exciting light (typically visible or near-infrared the speed of light), then it emits electromagnetic radiation
(NIR) light) is not absorbed by the sample to be studied of the same frequency.
Raman Spectroscopy: Theory 3
In a polyatomic molecule, an electric dipole is formed z

if the center of positive charges and the center of negative
charges do not coincide. Thus, depending on the symmetry
E
of the molecule at the equilibrium configuration of atomic
Direction of
nuclei, the molecule may or may not have a permanent propagation
electric dipole moment. But irrespective of this, the dipole
moment may change during certain normal vibrations as r
the nuclei are displaced from their equilibrium positions.

This change is periodical in time, with the frequency of the y
given normal vibration nM , i.e. x
Figure 2. Coordinate system and relative orientation of vectors
m D m0 cos 2pnM t 2 m (at the origin) and E (at a distance r from the origin).
where m0 is the amplitude vector of the oscillating dipole.

(Note that m0 is not the permanent dipole moment, and where nQ is the wavenumber of the radiation, m0 is the
m is only the difference between the instantaneous and magnitude of the oscillating dipole, e0 is the permittiv-
permanent dipole moment vectors, whereas the latter may ity of the medium, and q is the angle the dipole makes
or may not differ from zero; that is, it is only the dipole with the direction of propagation considered (see Figure 2),
contributed by vibrations of the nuclei that matters.) This whereas the electric field vector E, being perpendicu-
oscillating dipole of the molecule is capable of producing lar to the direction of propagation, lies in the common
electromagnetic radiation of frequency nM ; it is also capa- plane of the dipole vector and the chosen direction of
ble of absorbing radiation of the same frequency, which propagation.
will excite the given molecular vibration. Emission and The time-averaged value of the total energy density of
absorption bands associated with molecular vibrations can radiation at a given position along the propagation direction
be observed in the IR region of the spectrum. (see Figure 3) is
Scattering of light by a molecule is associated with oscil-
lations of an induced electric dipole. Electric dipoles may p2 nQ 4 20 sin2 q
be induced in a molecule by external electric fields, either rN D 4
2e0 r 2
static or alternating. The external electric field will polarize
the molecule: it will shift the negatively charged electrons (Note that the nQ 4 factor should be replaced by nQ 3 in
and positively charged nuclei in opposite directions. If this this and all related equations when using photon-counting
induced dipole oscillates, it can produce electromagnetic detection.) The distribution of the energy density has axial
radiation. In the case of a static external field, the induced symmetry with the rotation axis running along the dipole.
dipole oscillates only with the frequencies of the normal It follows from equation (4) that the energy density of
vibrations of the molecule, just like permanent dipoles. radiation is maximal in the equatorial plane around the
However, in the case of an oscillating external field, such dipole (q D p/2), and it decreases toward the poles and
as a beam of visible light, the induced dipole follows the drops to zero at q D 0.
alternating electric field of the radiation and it is also modu-
lated by the vibrations of the nuclei. As a result, it oscillates
at the frequency of the electric field as well as at combina- Direction of
z propagation
tion or beat frequencies of the external field frequency and E
the frequencies of normal vibrations, radiating at all these
frequencies.
xy
1.3 Radiation of an oscillating dipole
The amplitude E0 of electric field intensity (E) of the

radiation produced by the oscillating dipole at a distance
r is given by Figure 3. Angular distributions of the amplitude of E (broken
pnQ 2 0 sin q line) and of the radiant intensity, I (solid line), of the oscillating
E0 D 3 electric dipole m.
e0 r
The total power radiated by the dipole in the whole respectively. [Note that tensors are operators describing
solid angle is given by physical properties responsible for the connection between
vectorial quantities. They are usually represented by matri-
4p3 cnQ 4 20
D 5 ces. Thus a tensor of rank two can be represented by a
3e0 two-dimensional (eventually, a 3 3) matrix whose ele-
and the radiant intensity, I, of the dipole within the given ments are specified by a symbol having two indices (e.g.
element of solid angle d in a particular direction defined aij ); a tensor of rank 3 is represented by a three-dimensional
by the angle q (see Figure 3) is array of matrix elements (e.g. bijk ), and so forth.] The polar-
izabilities can be regarded as the measure of the flexibility
d p2 cnQ 4 20 sin2 q
Iq D D 6 of the electron cloud: the ease with which the electron cloud
d 2e0 of the molecule can be deformed or displaced to produce
an electric dipole under the influence of the external elec-
1.4 The scattering cross-section tric field. The nonlinear terms in equation (9) are usually
so small compared to the linear term that they do not play
The above quantities can be used to characterize the inten- a role in normal, linear Raman scattering. If we restrict the
sity of scattered radiation and the scattering efficiency of discussion to the linear term in equation (9), i.e.
the scattering particle as well. For the latter purpose it is m0 D a E 10
customary to take the ratio of the total scattered power
to the irradiance J of the incident radiation to obtain the then it can be written in the form of three linear equations,
so-called scattering cross-section s:
m0x D axx Ex C axy Ey C axz Ez
m0y D ayx Ex C ayy Ey C ayz Ez 11
sD 7
J m0z D azx Ex C azy Ey C azz Ez
This has the dimension of area, as the total scattered power corresponding to the matrix multiplication
is proportional to the irradiance of incident radiation. This 0
quantity refers to scattering in the whole solid angle around mx axx axy axz Ex
the scattering center. The fraction scattered into a given m0y D ayx ayy ayz Ey 12
direction within a small element of solid angle d is m0z azx azy azz Ez
called the absolute differential scattering cross-section or
where the nine coefficients aij are the components of the
scattering coefficient:
polarizability tensor a. As this and similar tensors play key
ds d/d roles in Rayleigh and Raman scattering, some important
D 8
d J properties of them will be discussed briefly.
The polarizability tensor can be described by a real,
symmetric matrix, where all aij D aji ; thus it has at most
2 CLASSICAL TREATMENT OF six independent components: three diagonal and three off-
RAYLEIGH AND VIBRATIONAL diagonal. This matrix is only necessarily symmetric in
RAMAN SCATTERING the case of nonresonant Raman scattering; for the case of
resonance Raman scattering (RRS) see Resonance Raman
2.1 The induced dipole moment and the Spectroscopy.
polarizability tensor The polarizability tensor of any molecule can be repre-
sented graphically as an ellipsoid having, in general, three
The classical theory of Rayleigh and Raman scattering is different half-axes. Although the shape of the polarizabil-
based on the concept that scattered light is generated by ity ellipsoid of a molecule is independent of the choice
oscillating electric dipoles induced by the electric field of reference coordinate system, the actual values of tensor
of incident (exciting) radiation. The relation between the elements depend on the orientation of axes. With sym-
induced dipole moment vector m0 and the electric field vector metric molecules it is good practice to choose the system
E can be written in the form of the following power series: axes in accordance with the symmetry elements of the
m0 D aE C 12 bEE C 16 gEEE C 9 molecule (to orient them along the axes of symmetry, or
perpendicularly to the plane of planar molecules, etc. as
where a is the polarizability, b the hyperpolarizability, and per the recommendations of the Joint Commission for
g the second hyperpolarizability of the molecule. Polar- Spectroscopy8 ). If so oriented, the axes of reference coin-
izabilities a, b, and g are tensors of rank 2, 3, and 4, cide with the principal axes of the polarizability ellipsoid,
in which case the polarizability tensor takes on a simpler, where aij is a component of the polarizability tensor, with
diagonal form, when all off-diagonal elements vanish (i.e. aij 0 being its value at the equilibrium configuration;
aXY D aYZ D aZX D 0), and the lengths of the half axes of Qk is the kth normal coordinate associated with vibra-
1/2 1/2 1/2
the ellipsoid are: aXX , aYY , and aZZ . tion of wavenumber nQ k ; the subscripts 0 refer to derivatives
It should be mentioned that, although the individual taken at the equilibrium configuration. (For a definition of
components of the polarizability tensor change on rotation normal coordinate see Calculation of Vibrational Fre-
of axes, certain combinations of them remain invariant. quencies by Molecular Mechanics.) In order to further
One such invariant is the so-called mean polarizability a, simplify the treatment, the double harmonic approxima-
defined as tion is adopted that neglects both mechanical and electrical
a D 13 axx C ayy C azz , 13 anharmonicity: on the one hand, instead of equation (18),
linear dependence of polarizability on each vibrational coor-
The other is the anisotropy g, dinate (for small amplitude vibrations near the equilibrium)
g D 12 [axx ayy 2 C ayy azz 2 is assumed, i.e.

@a
C azz axx 2 C 6a2xy C a2yz C a2zx ] 14 ak D a0 C Qk 19
@Qk 0
These invariants of the polarizability tensor acquire spe-
On the other hand, the normal vibrations are treated as being
cial importance when, instead of space-fixed molecules,
harmonic, i.e.
scattering from freely rotating or randomly reorienting
Qk D Qk0 cos 2pcnQ k t 20
molecules are considered. The power of the scattered radia-
tion is then determined by the space averages of the squares Let us consider now the time dependence of the induced
of the polarizability components. It can be shown that all electric dipole moment m0 , generated under the influence
significant terms contained in these averaged squares can of the electric field of incident radiation of wavenumber
be expressed through the mean polarizability a and the nQ 0 (that is normally much higher than nQ k ) in a vibrating
anisotropy g in the following way: molecule. The variation of the electric field strength with
45a2 C 4g2 time is given by
a2xx D a2yy D a2zz D 15
45 E D E0 cos 2pcnQ 0 t 21
2
g
a2xy D a2yz D a2zx D 16 Inserting equations (19)(21) into equation (10) gives, for
15
the kth vibration,
45a2 2g2
axx ayy D ayy azz D azz axx D 17 @a

45 m0 D a0 E0 cos 2pcnQ 0 t C
The space averages of the remaining quadratic terms (the @Qk 0
cross-products involving off-diagonal elements of the polar- Qk0 E0 cos 2pcnQ 0 t cos 2pcnQ k t 22
izability tensor) are all zero.6
Trigonometric transformation yields

0 1 @a
2.2 Time dependence of the induced dipole in m D a0 E0 cos 2pcnQ 0 t C Qk0 E0
2 @Qk 0
vibrating molecules
[cos 2pcnQ 0 nQ k t C cos 2pcnQ 0 C nQ k t] 23
The time dependence of the induced dipole moment defined
by equation (10) in a vibrating molecule is now considered, The three cosine functions having three different argu-
where the polarizability tensor may be modulated by the ments in this equation mean that the induced dipole oscil-
normal vibrations. lates with three distinct frequencies simultaneously, there-
The dependence of the molecular polarizability on the fore it generates radiation at nQ 0 and also at wavenumbers
normal coordinates can be expressed in the form of the shifted by nQ k . The first term describes Rayleigh scatter-
Taylor series ing observable at nQ 0 , whereas the second and third terms
account for Stokes Raman and anti-Stokes Raman scattering
@aij
at nQ 0 nQ k and nQ 0 C nQ k , respectively. These so-called beat
aij D aij 0 C Qk
k
@Qk 0 frequencies are produced when the dipole oscillating at nQ 0
2
is modulated by the molecular vibration at wavenumber nQ k .
1 @ aij It can be concluded that the classical theory successfully
C Qk Ql C 18
2 k,l @Qk @Ql 0 describes the frequency relationships of vibrational Raman
scattering. It shows that the Raman shift is independent of Raman activity: those vibrations for which a0k D 0 (meaning
the frequency (wavelength) of the incident radiation. Thus, that all tensor components are zero) are inactive in Raman
whether one excites a molecule with the green or blue laser scattering, and those normal vibrations for which at least
line of an ArC ion laser or a NIR Nd : YAG (yttrium alu- one component of a0k differs from zero, i.e.
minum garnet) laser, one will obtain the same wavenumber
pattern (but not necessarily the same relative intensities, 0 @aij

aij k D 6D 0 25
see later), with Raman shifts characteristic of the scattering @Qk 0
molecule.
are Raman active.
The relative intensities and polarization properties of
2.3 Intensities, directional and polarization spectral lines in scattered radiation depend on different
properties factors, such as the scattering geometry, the orientation of
the molecule with respect to the incident light, and the state
According to equation (23), the oscillation amplitudes of of polarization of the incident light. Determination of the
the induced dipole moment are directly proportional to the state of polarization of the scattered radiation is of great
amplitude of the electric field vector of the incident radi- importance because it can be correlated with the symmetry
ation E0 . In addition, the intensity of Rayleigh scattering of the scattering species and the symmetry of the individual
depends on a0 , the polarizability of the molecule at the vibrational modes.
equilibrium nuclear configuration, whereas that of Raman The scattering geometry is defined by the relative orien-
scattering (both Stokes and anti-Stokes) is governed by the tation of the directions of illumination and observation in
so-called derived polarizability tensors, a0k : a laboratory-fixed coordinate system according to Figure 4,

where the scattering molecule is located at the origin O. By
@a
a0k D 24 convention, the direction of illumination (by a monochro-
@Qk 0
matic laser light) is set along the positive z axis (vertical),
These reflect the sensitivity of molecular polarizability to whereas the scattered light is observed along the direction
changes of nuclear configuration along the normal coordi- Ow making an angle q with the z axis (in the xz plane). The
nate of vibration. plane containing the direction of propagation of the incident
Considering the properties of the derived polarizability light and the direction of observation is called the scatter-
tensor (sometimes also referred to as the Raman tensor), ing plane (in this case the xz plane) that serves as a plane
there are similarities but also notable differences in com- of reference when the state of polarization of the incident
parison to those of a0 . The invariants of the derived polar- and scattered light is specified. For instance, the incident
izability tensor (its mean value a0 and its anisotropy g0 ) are beam is said to be parallel polarized when its electric vec-
defined by analogy to those of a0 , in accordance with equa- tor is parallel to the scattering plane (i.e. when Ex 6D 0 and
tions (13) and (14). Both a0 and a0 are symmetric tensors Ey D 0), and it is said to be perpendicularly polarized when
(at least, as far as nonresonant Raman scattering is con- its electric vector is perpendicular to the scattering plane
cerned), but although the components (aij 0 of a0 can only (Ex D 0 and Ey 6D 0).
be positive and its diagonal terms always differ from zero We are interested in the radiant intensity I of the light
(which ensures that the Rayleigh line is always present in scattered at an angle q and in its polarization characteristics
the spectrum of the scattered light), the components (a0ij k described by the so-called depolarization ratio r. The latter
of a0k can be positive, negative or zero. Consequently, the quantity can be defined in different ways. For the case of
mean value of the derived polarizability tensor a0 may also illumination with radiation polarized perpendicular to the
be zero for certain normal vibrations. This means that the
derived polarizability tensor can no longer be represented z
as a real ellipsoid, because some of its axes may become
w
zero or imaginary. Another important difference is that the
orientation of principal axes of the derived polarizability
ellipsoids pertaining to the various normal modes of vibra-
tion may differ from that of the equilibrium polarizability O
ellipsoid. y
These differences in properties bring about differences in x Ey
Ex
the directional and polarization properties of Rayleigh and
Raman scattering. It is also clear from equation (23) that Figure 4. Orientation of axes for definition of illumination and
the derived polarizability tensor serves as the measure of observation geometry, where xz is the scattering plane.
scattering plane, it is denoted r? and defined as calculation (details given by Long6 ) for Raman scattering
? of perpendicularly polarized incident radiation at 90 scat-
Ijj q
r? q D ? I q
26 tering geometry (for an unspecified kth normal mode of
? vibration) we obtain
where the superscript preceding I refers to polarization of p
the incident radiation (perpendicular to the scattering plane
?
I? / nQ 0 nQ k 4 a0yy 2 Qk0
2
J 29
2
for both ? Ijj and ? I? ), and the subscript following I stands p
for the direction of polarization of the scattered light (paral-
?
Ijj / nQ 0 nQ k 4 a0zy 2k Qk0
2
J 30
2
lel and perpendicular, respectively, to the scattering plane).
Similarly, if the electric vector of incident light were where the minus sign refers to Stokes and the plus sign
polarized parallel to the scattering plane, the corresponding to anti-Stokes Raman scattering, and J is the irradiance of
depolarization ratio rjj would be defined as the incident (exciting) radiation. Although these expressions
do not fully describe the real dependence of observed
jj
I? q intensities on the wavenumber of the scattered radiation
rjj q D jj I q
27
jj (e.g. the Stokes/anti-Stokes intensity ratios which can be
correctly accounted for only by quantum mechanics), they
In fact, the common governing principle in equat-
are still useful and lead to the correct result when they are
ions (26) and (27) is that the intensity of depolarized scat-
used to calculate the depolarization ratios.
tering (i.e. the scattering of altered polarization) is ratioed
For this purpose, the space averages of the squares of
against the intensity of scattering with unaltered polariza-
the derived polarizability components in equations (29) and
tion, the states of polarization being considered with respect
(30) are replaced by the corresponding expressions given
to that of the incident light. This allows for a simpler (but
in terms of the invariants a0 and g0 (by analogy with
less rigorous) definition of the depolarization ratio as
equations (15) and (16)) to yield
I? q
rq D 28 p 45a 0 2
C 4 g0 2

Ijj q ?
I? / nQ 0 nk 4 k k 2
Qk0 J 31
2 45
where subscripts ? and jj refer to the mutual orientation
of the electric vectors of the incident radiation and the p g0 2
k
scattered light. Irrespective of the scattering geometry, this
?
Ijj / nQ 0 nk 4 2
Qk0 J 32
2 15
simpler notation is often used in applied spectroscopy, but
the current discussion reverts to the notation introduced in The ratio of ? Ijj to ? I? then gives the well known expres-
equation (26). sion of depolarization ratio for the case of plane polarized
In the practice of Raman spectroscopy two different incident radiation:
scattering arrangements have become widely used: (a) 90 p ?
Ijj p/2 3g0 2k
scattering (q D p/2), corresponding to illumination along r? D? D 33
the positive z axis (Figure 4) and observation along the 2 I? p/2 45a0 2k C 4g0 2k
positive x axis; and (b) 180 scattering or back-scattering If the sample is illuminated with natural (i.e. nonpolarized)
(q D p), when illumination is along the positive z axis and light using the same scattering geometry as above, and the
observation is in the negative direction of the z axis. Note scattered light is analyzed with parallel and perpendicular
that although the convergence of the focused laser beam settings of the polarizer to measure n Ijj and n I? , respec-
used for illumination is negligible, the scattered radiation is tively, then the corresponding depolarization ratio can be
usually collected within a fairly large solid angle around the calculated from a similar expression:
selected nominal direction of scattering, so as to increase
the observed signal. p n
Ijj p/2 6g0 2k
rn D D 34
For samples consisting of freely rotating molecules (such 2 n I p/2
? 45a0 2k C 7g0 2k
as low-pressure gases) or an assembly of randomly orient-
ing molecules (as in most liquids) and illumination with lin- Equations (33) and (34) show that the depolarization
early polarized light, the radiant intensity for both Rayleigh ratios of Raman bands depend on the mean values of the
and Raman scattering can be calculated using equation (6). derived polarizability tensor and its anisotropy. As both a0
This is achieved by substituting the appropriate frequencies or g0 can be zero in some cases, the intervals of the possible
and the corresponding components of the amplitude of the depolarization ratio values are
induced dipole moment from equation (23), then averaging p 3
over all orientations of the molecule. After performing this 0 r? 35
2 4
and p and g0 D 0, when equation (33) yields r? D 0 and the

6
0 rn 36 observed band is considered completely polarized.
2 7
Although we have considered only the case of 90 scattering As an example, the low-frequency part of the two differ-
above, the results can be extended for the case of 180 ently polarized Raman spectra of liquid chloroform, CHCl3 ,
scattering. More precisely, depolarization results are, in is shown in Figure 5. The strongest two bands at 669 and
principle, the same for 180 scattering when using polarized 368 cm1 are almost completely polarized, so they must
exciting radiation (i.e. for r? p) as for 90 scattering with belong to the totally symmetric a1 species of the C3v
incident light perpendicularly polarized with respect to the point group. Considering their wavenumbers, they can be
scattering plane r? p/2: assigned to the symmetric CCl3 stretching (ns CCl3 ) and
symmetric CCl3 deformation (s CCl3 ) vibrations, respec-
3
0 r? p 4
37 tively, whereas the two depolarized bands (with r D 0.75)
belong to the degenerate nas CCl3 (761 cm1 ) and CH
Note that neither 90 scattering with parallel polarized
(263 cm1 ) vibrations.
incident light, nor 180 scattering with natural incident light
Since the advent of lasers, polarized laser beams are
can be used to determine the depolarization ratios, because
almost exclusively used as the source of illumination for
they would give values of rjj p/2 D 1 and rn p D 1 for
the measurement of Raman scattering, so the depolariza-
all observed bands in randomly oriented samples.
tion ratio r? is much more often used than rn , the latter
The actual values of r? and rn are determined by the
symmetry properties of the derived polarizability tensor, assuming illumination with natural light. However, there
that reflect the symmetry of the corresponding normal mode is a way of determining rn experimentally with the use of
of vibration. Thus, the following cases can be distinguished: plane polarized exciting light but no polarization analyzer in
the path of the scattered radiation, as follows. Illumination
1. For nontotally symmetric vibrations a0 D 0 and g0 6D 0, is along the z axis (as shown in Figure 4) and the scattered
so that r? D 3/4 rn D 6/7 and such Raman bands intensity is detected in the direction of the x axis without
are said to be depolarized. polarization. In the first measurement the electric vector of
2. For totally symmetric vibrations both a0 or g0 differ the laser beam is polarized parallel to x (parallel to the scat-
from zero, thus r? < 3/4 rn < 6/7 and such bands tering plane), in which case the measured intensity can be
are said to be partially polarized. described as the sum of parallel and perpendicular scattered
3. At an extreme, for totally symmetric vibrations of intensities, jj Ijj p/2 Cjj I? p/2. For the second measure-
molecules belonging to cubic point groups a0 6D 0 ment the plane of polarization of the incident light is turned
669
368
Raman intensity
263
761
(a) Parallel polarized
(b) Perpendicularly polarized
800 600 400 200

Raman shift / cm1
Figure 5. Part of the polarized Raman spectra of liquid chloroform, recorded with 180 scattering for determination of depolarization
ratios. The scattered light is analyzed with the polarizer set (curve a) parallel and (b) perpendicular to the polarization direction of the
incident radiation.
parallel to the y axis when the total intensity measured is in the round brackets, because they define the component
?
Ijj p/2 C? I? p/2. This can be done conveniently by of the scattering tensor being measured. Further details of
inserting a half-wave plate into the path of the laser beam, single crystal Raman studies are provided by Gilson and
which will turn the plane of polarization by 90 . Then the Hendra,10 Turrell,11 and Wilkinson.12
ratio of these two measured intensity values is taken which
gives:
3 PARTIAL QUANTUM MECHANICAL
jj
Ijj p/2 Cjj I? p/2 6g0 2k TREATMENT OF RAYLEIGH AND
D
? I p/2 C? I p/2
jj ? 45a0 2k C 7g0 2k
p RAMAN SCATTERING
D rn 38
2 3.1 Refinement of the classical description
i.e. rn can be measured and equation (36) can be used also
in the case of plane polarized exciting radiation. According to the basic principles of quantum mechanics,
Returning to the interpretation of the measured depolar- the energy associated with electronic, vibrational and rota-
ization ratios, we have seen that in samples consisting of tional degrees of freedom of a molecule can assume values
randomly orienting molecules it is possible to distinguish only from a discrete set, namely the quantized energy lev-
between Raman active vibrations belonging to totally sym- els corresponding to the possible stationary states of the
metric and nontotally symmetric species only. A practical molecule. As well as their energies, these states are char-
difficulty is that for some polarized bands the depolariza- acterized by a specific set of quantum numbers describing
tion ratio r? , although smaller than 0.75, may occasionally the level of excitation of each quantized motional degree
approach this value very closely, hence the intensity mea- of freedom, and by a corresponding wavefunction, .
surements should be of fairly high quality to make the Using the terminology of quantum mechanics, radiation
distinction. This may involve calibration of the depolar- is absorbed or emitted by a molecular system as the result
ization measurement or the use of an internal standard. of an upward or downward transition between two energy
levels. The radiation absorbed or emitted is also quantized,
with the energy enclosed in discrete photons that can alter-
2.4 Polarization measurements in single crystals natively be viewed as electromagnetic waves. The loss or
gain of energy by the molecule, E, is equivalent to the
It should be mentioned that the derived polarizability ten- energy of electromagnetic radiation emitted or absorbed
sors show different patterns of entries for vibrations belong- (Figure 6). This energy is directly proportional to the fre-
ing to different symmetry species; so, if these patterns can quency or wavenumber of radiation,
be determined, all Raman active symmetry species can be E D hn D hcnQ 39
distinguished from each other. Such a possibility is pro-
vided by Raman polarization measurements performed on where h is the Planck constant (h D 6.62608 1034 J s),
single crystals, where the molecules have fixed orienta- c is the speed of light (in vacuum c D c0 D 2.99792458
tions within the sample. In this case the relative magnitudes 108 m s1 ), n is the frequency, and nQ is the wavenumber of
of all six independent components of the derived polariz- the radiation.
ability tensor can be determined for each Raman line by In IR absorption or emission, there is a direct transition
direct measurement at suitable orientation of the crystal and between two vibrational energy levels, most often between
appropriate polarization settings. In single crystal measure- the vibrational ground state (v D 0) and the first excited
ments it is more convenient to use a system of reference state (v D 1). These transitions are simple one-photon pro-
axes tied to the crystal axes rather than laboratory-fixed cesses one photon is absorbed or emitted during the
coordinates. In this case the nomenclature recommended transition. In contrast to this, both Rayleigh and Raman
by Damen et al.9 is used to describe the different arrange- scattering involve two almost simultaneous transitions pro-
ments. This involves four symbols, a(bc)d in general, where ceeding via virtual states in which one photon of the
a stands for the propagation direction of the incident radi- incident radiation is annihilated and another photon, either
ation, b for the orientation of the electric vector of the of the same energy (Rayleigh scattering) or of lower energy
incident radiation, c for the orientation of the electric vec- (Stokes Raman) or higher energy (anti-Stokes Raman), is
tor of scattered light (e.g. the setting of the polarization created. (The term virtual state refers to a transition state
analyzer), and d for the direction of propagation of the which does not correspond to an eigenstate of the molecule;
scattered light. It follows from the definition that the most so it is only an imaginary state, a practical convenience by
essential part of the notation is the two symbols enclosed which the energy exchange between the radiation field and
Excited electronic level a final state f induced by radiation of wavenumber nQ 0 :
[m0 ]fi D hyf jajyi i E0 42

Virtual
levels where yf and yi are vibrational wavefunctions (but occa-
sionally can be considered to be vibrationalrotational or
Energy
pure rotational wavefunctions as well), and a is the polar-

h(0 + M) izability tensor. As the frequency dependence of Raman
h0 h0 h0 h0 scattering was obtained correctly from the classical the-
h(0 M)
ory, we are now particularly interested in determining the
Vibrational
energy levels: intensity of scattered radiation for a given vibrational tran-
=1 sition. In this case we can avoid consideration of the time
h dependence and examine only the transition moment ampli-
=0 tudes; then yf and yi are considered as time-independent
IR Rayleigh Stokes Rayleigh Anti-Stokes
vibrational wavefunctions.
absorption scattering Raman scattering Raman
and emission scattering scattering The Raman selection rules (i.e. the Raman activity of
Figure 6. Diagram of transitions between vibrational energy lev- transitions) and the intensity of scattered radiation will
els corresponding to the processes of IR absorption/emission, and be determined by the properties of the transition polariz-
Rayleigh and Raman scattering. ability tensor (sometimes referred to as Raman tensor or
scattering tensor) [a]fi , whose components [axx ]fi , [axy ]fi ,
the molecule during the scattering process can be split up etc., are the matrix elements of the polarizability tensor for
into two one-photon transitions.) the transition f i. Thus it is necessary to examine the
To refine the classical description of Raman scattering we dependence of the transition polarizabilities on the vibra-
proceed by using the results of a partial quantum mechan- tional normal coordinates in a manner similar to classical
ical treatment in which the molecule is treated quantum mechanics.
mechanically and the radiation classically. Then, for the By neglecting electrical and mechanical anharmonicity
description of the interaction of light with matter, electro- again and exploiting the properties of the harmonic oscil-
magnetic radiation is considered as a source of perturbation lator wavefunctions, it is obtained that only those matrix
of the molecular system. This partial quantum mechanical elements differ from zero for which one vibrational quan-
treatment yields correct results for most spectroscopic pro- tum number changes by unity (i.e. vf k D vi k 1); all other
cesses including absorption, emission and also Rayleigh and vibrational quantum numbers remain unchanged (i.e. vf k D
Raman scattering. vi k , for all j 6D k). Then for a typical matrix element of the
It is anticipated that any direct transition between two transition polarizability associated with Stokes Raman scat-
energy levels is accompanied by emission or absorption of tering (for vf k D vi k C 1) at Raman shift nQ k we obtain (in
radiation if the dipole moment of the transition (which is molecule fixed axes x, y, and z),
called the transition moment) is nonzero:
i
1 @axy
[axy ]vf vi D vk C 1 bvk
2
O ji i 6D 0
mfi D hf jm 40 @Qk 0
1
where i and f are the wavefunctions of the initial and D vik C 1 2 bvk a0xy k 43
O is the dipole moment oper-
final states, respectively, and m
ator. In equation (40) the so-called Dirac bracket notation Similarly, for anti-Stokes scattering (when vf k D vi k 1),
is used to abbreviate the expression
1 @axy 1
[axy ]vf vi D vik 2 bnk D vik 2 bnk a0xy k 44
@Qk 0
hf jOji i f O i dt 41
where bv2k D h/8p2 cnQ k is the quantum mechanical analog
where the integral is over all the coordinate space. For of the amplitude Qk of the classical oscillator. It is clear
absorption or emission mO corresponds to the permanent from the last two equations that for such transitions to be
electric dipole operator, whereas for light scattering m O is Raman active it is also required that at least one component
the induced dipole moment operator. of the derived polarizability tensor be nonzero, just as in
The quantum mechanical equivalent to the amplitude of classical theory.
the induced dipole of equation (10) is the transition moment A very significant difference between the classical and
amplitude concerning a transition from an initial state i to quantum mechanical treatments is that in the latter there is
a dependence on the quantum number vk , which has impor- 3.2 Raman scattering from absorbing samples:
tant consequences concerning the intensities of Raman scat- resonance Raman scattering
tering. When we consider scattering from an assembly of
N molecules, the initial distribution of molecules among The previous discussion was based on polarizability theory
the vibrational states at temperature T must be taken assuming that the scattering sample is transparent to the
into account (utilizing the Boltzmann distribution law) to exciting laser light. This corresponds to the requirement
establish the fraction of molecules capable of a given that the frequency of the incident radiation (n0 ) should be
transition. Thus, introducing the appropriate changes into much higher than the frequencies of vibrational and rota-
equations (31) and (32) obtained from classical theory, tional transitions (nM ), but also much lower than those of
we obtain for scattered intensity of Stokes lines (at 90 any electronic transitions. For a given compound this con-
scattering): dition determines the wavenumber range extending from
p NIR to some visible wavenumbers that can be safely
? hNgk nQ 0 nQ k 4 [45a0 2k C 4g0 2k ] used to excite Raman scattering. However, as the intensity
I? D J 45
2 8e20 cnQ k [1 exphcnQ k /kT] 45 of Raman scattering is proportional to the fourth power
p hNgk nQ 0 nQ k 4 g0 2k of the absolute wavenumber of scattered light, a signifi-
?
Ijj D 2 Q [1 exphcn
J 46 cant improvement in Raman scattering efficiency can be
2 8e0 cnk Q k /kT] 15
expected when higher exciting wavenumbers are used. Thus
where gk is the degeneracy of the kth vibration (all other on theoretical grounds, excitation with the green or blue
symbols have been defined earlier). The corresponding lines of an Ar ion laser is to be preferred to the NIR radi-
equations for anti-Stokes lines can be obtained by sub- ation of a Nd:YAG laser, as they are expected to generate
stituting nQ 0 C nQ k 4 for nQ 0 nQ k 4 , and [exphcnQ k /kT 1] 1620 times greater Raman intensities.
for [1 exphcnQ k /kT]. Note also that according to these However, when the frequency of the incident light
equations the intensity of Raman scattering depends on the approaches that of an electronic transition or falls inside
number of molecules (or scattering centers N) in the scat- an electronic absorption band of the sample, resonances
tering volume and on the intensity of the exciting radiation occur that may considerably enhance the intensities of cer-
J. By dividing equations (45) and (46) by J we arrive at tain Raman lines. This effect is called resonance Raman
the corresponding expressions for the absolute differential scattering (RRS) and is due to the coupling of electronic
cross-section that characterizes the scattering efficiency of and vibrational transitions. This process can be described
the sample. theoretically by time-dependent perturbation theory, show-
Obviously, according to these relationships, the wave- ing that if a resonant frequency is approached, the induced
number dependence of relative intensities of Raman lines dipole becomes abnormally large. Resonance enhancement
differ very strongly from those obtained from classical of scattering is accompanied, of course, by absorption
theory. However, as has been anticipated, quantum theory of both the incident and the scattered radiation, whereas
leads to the same depolarization ratios as classical theory, excitation of fluorescence is not excluded either. Because
i.e. equation (33) is still valid, which can be verified by of these competing processes, observation of resonance
taking the ratio of the right-hand side of equation (46) to Raman spectra may need some expertise. The topic of reso-
that of equation (45). nance Raman scattering is described in Resonance Raman
The most pronounced difference between the classical Spectroscopy.
and quantum mechanical treatments of Raman scattering is
found in the relative intensities of the corresponding Stokes
and anti-Stokes Raman lines: 4 VIBRATIONAL SELECTION RULES

CONTRASTING IR AND RAMAN
IStokes InQ 0 Qnk nQ 0 nQ k 4 hcnQ k
D D exp 47 SPECTROSCOPY
Ianti-Stokes InQ 0 CQnk nQ 0 C nQ k 4 kT
In addition to correctly accounting for the observed The foregoing discussion has shown that molecular vibra-
intensity relationships, this equation allows determination tions are responsible for the appearance of absorption or
of the sample temperature (by contact-free measurement) emission bands in the IR spectrum and for scattering of
provided that the measured Raman intensities are corrected light at shifted frequencies. We have also touched on the
for instrument response. Note that equations (45)(47) are question of activity of normal modes in a general sense,
only correct when measuring the scattered power; in the without analyzing the conditions of activity for each nor-
case of photon counting (employed in many instruments) mal mode separately. Recall that classical theory has led to
the fourth-power dependency reduces to a third-power one. the following criteria:
1. A given normal vibration of a molecule may appear polarizability along this direction differs from that along
in the IR spectrum if at least one component of the any perpendicular directions (which are indistinguishable),
electric dipole moment of the molecule changes dur- but the mean polarizability will still be positive. Although it
ing this vibration, i.e. when the value of a dipole is not easy to predict the actual change of polarizability with
moment derivative, taken at the equilibrium, differs internuclear distance in different directions, for qualitative
from zero:
predictions one may use the analogy between polarizability
@i and the volume occupied by the electron cloud as the
6D 0 48
@Qk 0 bond is stretched, the volume of the electron cloud increases
2. A given normal vibration of a molecule may appear which will certainly affect its polarizability as well. Quan-
in the Raman spectrum if at least one component of tum chemical calculations confirm that for all homonuclear
the polarizability tensor changes during this vibration, molecules (@a/@Q0 > 0; thus, the vibrations of X2 type
i.e. when the value of a derived polarizability tensor molecules are Raman active.
component differs from zero: In heteronuclear diatomic molecules (XY type) the dipole

moment goes through a maximum if the internuclear dis-
@aij tance changes from zero to infinity, but the position of the
6D 0. 49
@Qk 0 maximum differs from the equilibrium distance. According
to this, the dipole moment derivative may be positive or
According to quantum theory, oscillation of a permanent negative, but generally different from zero, which means
or induced dipole does not produce any radiation. For that the vibration is IR active. Regarding the change of
radiation to be emitted or absorbed, a transition should polarizability, the same arguments hold as for homonuclear
occur between different vibrational energy levels of the molecules above (see also Table 1), so this vibration will
molecule (Figure 6) which may be concomitant with loss be Raman active as well.
or gain of energy in the form of electromagnetic radiation.
The classical analog of excitation of a molecule to a
higher vibrational energy level is vibration with increased 4.2 Molecules with a center of symmetry and the
amplitude. With this, we can accept the conditions of principle of mutual exclusion
IR and Raman activity in general as formulated above
and examine whether they are fulfilled in the case of In polyatomic molecules, as the number of constituting
particular molecular vibrations having different symmetry atoms and the complexity of normal vibrations increases,
properties. it becomes ever more complicated to follow the change
of dipole moment and polarizability with normal coordi-
nates. The situation is still simple enough in the case of
4.1 Diatomic molecules triatomic molecules where two typical examples are worth
examining linear and bent XY2 molecules.
First we examine the fundamental transitions of diatomic Carbon dioxide is a good example of a linear molecule
molecules which have only one mode of vibration, the which has no permanent dipole moment due to the symmet-
stretching of bond length. It is necessary to distinguish rical arrangement of electric charges. This molecule has
between X2 type homonuclear and XY type heteronuclear three normal vibrations shown schematically in Table 2:
diatomic molecules since the former have a center of a symmetrical stretching (Q1 ), an asymmetrical stretch-
symmetry, and consequently no dipole moment, whereas ing (Q2 ), and a degenerate pair of bending vibrations
the latter are polar molecules with an asymmetric electron (Q3 ). The center of symmetry is preserved in the molecule
distribution which creates a permanent dipole. The factors during the symmetrical stretching vibration (Q1 ), thus dur-
determining IR and Raman activity of the vibrations in these ing this vibration the molecule behaves as a homonuclear
molecules are compared in Table 1. diatomic molecule the vibration is IR inactive but Raman
During small-amplitude vibrations of a homonuclear active.
diatomic molecule, the symmetry of the electron distribu- However, when the molecule is distorted along the asym-
tion is retained and no dipole is generated. Consequently, metric normal coordinates Q2 or Q3 , the symmetry of
@m/@Q0 D 0, so this vibration cannot be observed in the the distribution of charges is destroyed and dipoles are
IR (it is IR inactive). To judge its Raman activity, it is generated. In case of the asymmetric stretching vibra-
necessary to examine the dependence of polarizability on tion (Q2 ) the dipole oscillates along the bond direction,
internuclear distance. At equilibrium, the molecule must whereas during the bending vibration (Q3 ) it oscillates in
have a nonzero polarizability, with one axis of the polar- a perpendicular direction. These two asymmetric vibrations
izability ellipsoid oriented along the bond direction. The are therefore IR active. The molecule at the equilibrium
Table 1. Factors determining the IR and Raman activity of vibrations of homonuclear and heteronuclear
diatomic molecules.
Parameter Type of molecule

X2 XY
XX stretching XY stretching
Normal mode of vibration
Dipole moment 0
Variation of dipole moment with normal coordinate
Q Q
Dipole moment derivative (at equilibrium)

= 0 0
Q Q
0 0
Infrared activity No Yes
Shape of polarizability ellipsoid
Variation of polarizability with normal coordinate (schematic)

Q Q
Polarizability derivative
0 0
Q Q
0 0
Raman activity Yes Yes
position has nonzero polarizability which may change little D6h point group, which can be proved by depolarization
during the asymmetric vibrations, as these distortions do not measurements.
really alter the volume of the molecule. But even if there is The occurrence of mutual exclusion demonstrates the
some change, these asymmetric vibrations are, in fact, anti- complementary nature of IR and Raman spectroscopy best
symmetric, which means that the distortion of polarizability of all. In many other cases, when there is partial or even
ellipsoid during the first half period of vibration would be full overlap between the frequencies allowed in IR and
the mirror image of that during the second half period. The Raman spectra, there may still be very great differences
variation of polarizability is thus symmetric about the equi- in relative band intensities. Transitions giving rise only
librium position (see the corresponding graphs in Table 2) to hardly detectable weak features in either the IR or the
and the orientation of axes of the ellipsoid do not change Raman spectrum may be strong or at least much easier to
either. So, its gradient (the polarizability derivative) at the identify in the other.
equilibrium is zero and, consequently, these vibrations are In a bent triatomic molecule of XY2 type, the two XY
inactive in Raman scattering. bonds are also equivalent by symmetry, but there is no
Thus we have arrived at a notable result that in this center of symmetry here the molecule belongs to the C2v
molecule there is no overlap in the fundamental vibrational point group and all the three normal vibrations of it prove to
frequencies between IR and Raman spectra. In other words, be both IR and Raman active. This result can be obtained by
there is mutual exclusion in IR absorption and Raman the method employed above, i.e. by evaluating the dipole
scattering, which is a property of molecules possessing a moment derivatives and the polarizability derivatives for
center of symmetry. As an example of a larger polyatomic each normal mode of vibration. However, as this method
molecule having center of symmetry, the IR and Raman requires the knowledge of each normal mode as well as the
spectra of liquid benzene are compared in Figure 7. The two contribution of the displacement of each atom to changes of
strongest Raman lines at 3062 and 993 cm1 correspond to dipole moment and polarizability, this becomes an insoluble
totally symmetric vibrations belonging to a1g species of the task with larger molecules.
Table 2. Factors determining the IR and Raman activity of vibrations for a linear YXY molecule.
Parameter YXY
Normal mode of vibration Symmetric stretch, Q1 Asymmetric stretch, Q2 Bending, Q3
Dipole moment 0
Variation of dipole moment with normal coordinate z z x y
Q1 Q2 Q3
Dipole moment derivative (at equilibrium) z x

= 0 0 0
Q1 Q2 Q3
0 0 0
Infrared activity No Yes Yes
Shape of polarizability ellipsoid
Variation of polarizability with normal coordinate

Q1 Q2 Q3
Polarizability derivative
0 = 0 =0
Q1 Q2 Q3
0 0 0
Raman activity Yes No No
100
80
Transmittance (%)
60
40
20
0
3000 2500 2000 1500 1000 500
(a) Wavenumber / cm 1
Raman intensity
3000 2500 2000 1500 1000 500

(b) Raman shift / cm 1
Figure 7. (a) Infrared and (b) Raman spectra of liquid benzene (illustrating the noncoincidence of fundamental transitions active in the
two spectra).
4.3 Polyatomic molecules consideration is known, all we have to do is to check

whether any of the dipole moment or polarizability ten-
To simplify the procedure, and thereby make larger mole- sor components belong to that very symmetry species. At
cules more tractable, it is necessary to revert to the most this step it is convenient to use the character tables contain-
general form of the selection rules derived from the quan- ing this information, which are readily available in several
tum mechanical treatment, as follows: monographs (as well as article Infrared Spectroscopy:
Theory in this Handbook ) for all point groups.7,13
1. A transition from vibrational state vi to vf (denoted This procedure is now illustrated on the planar four-
vf vi ) is IR active if at least one of the three vec- atomic molecule formaldehyde, H2 CO. Based on its sym-
tor components (matrix elements) of the type [mx ]k fi metry at equilibrium, the molecule belongs to the C2v point
differs from zero; group, and it has 3N 6 D 6 vibrational degrees of free-
2. A transition vf vi is Raman active if at least one dom. Let us put the molecule into a Cartesian coordinate
of the six tensor components of the type [axy ]kfi differs system in accordance with the recommended practice:8 the
from zero. twofold rotation axis of symmetry C2 running along the
For the fundamental transition of the kth vibration of a direction of the CDO bond and bisecting the HCH angle
molecule this requires evaluation of integrals of the type coincides with the z axis; the molecule lies in the yz plane
(corresponding to the syz plane of symmetry) so that the x
[mx ]k fi D hy1 Qk jmx jy0 Qk i 50 axis (and the sxz plane of symmetry) is perpendicular to the
and plane of the molecule. Now we determine the distribution of
[axy ]k fi D hy1 Qk jaxy jy0 Qk i 51 vibrational degrees of freedom among the symmetry species
(the structure of the irreducible representation) using the
respectively, to see whether they differ from zero. This is information available in the character table of the C2v point
fulfilled if the triple products under the integrals belong group (Table 3), and apply the well established methods of
to a representation whose structure contains the totally group theory.13,14 The following result is obtained:
symmetric species.
Fortunately, there is a fairly simple practical solution to H2 CO D 3a1 C b1 C 2b2 52
this task that makes use of the symmetry properties of the
This information is added to the last column of Table 3:
transition dipole moment and the transition polarizability
three normal vibrations are assigned to a1 , one to b1 , and
matrix elements. According to this, the above integrals
two to b2 species, numbered consecutively from Q1 to Q6 .
differ from zero only if at least one component of the
For this small, symmetric molecule it is still easy enough
dipole moment (e.g. x ) or one component of the transition
to predict the approximate shape of normal coordinates or
polarizability tensor (e.g. axy ) have the same symmetry
normal modes of vibration (but note that it is not necessary
species as the normal coordinate Qk . This follows from
to know them in order to determine their IR and Raman
the fact that the vibrational wavefunction y0 of the ground
activity). For H2 CO they are depicted in Figure 8 and can
state (vk D 0) is always totally symmetric, whereas that of
be described roughly as follows:
the excited state y1 (vk D 1) has the same symmetry as the
normal coordinate Qk .6 Q1 (a1 ) practically pure (>99%) symmetric CH2
The application of these general selection rules requires stretching mode;
knowledge of the equilibrium symmetry (the point group) Q2 (a1 ) predominantly CDO stretching (80%) com-
of the molecule and the distribution of normal modes bined with CH2 bending (20%);
among the symmetry species (the irreducible representa- Q3 (a1 ) predominantly CH2 bending (80%) com-
tions). Once the symmetry species of the vibration under bined with CDO stretching (20%);
Table 3. Character table of point group C2v , with extension for the fundamental vibrations of formaldehyde, H2 CO.
C2v symmetry species Symmetry operations Activity H2 CO

normal modes, Qk
E C2v z sxz syz T R IR [m]k Raman [a]k
a1 1 1 1 1 z z axx , ayy , azz Q1 , Q 2 , Q3

a2 1 1 1 1 Rz axy
b1 1 1 1 1 x Ry x axz Q4
b2 1 1 1 1 y Rx y ayz Q5 , Q6
z
O O Q4 , Q5 , and Q6 are depolarized (r? D 0.75). In addi-
O
x y
tion, the three bands belonging to the totally symmetric
C C C
H H H H H H species a1 are expected to be noticeably stronger than the
Q1(a1) Q2(a1) Q3(a1) others, because the diagonal elements of transition polar-
izability tensors are usually significantly greater than the
+ off-diagonal elements.
O O O
C C C
+H H+ H H H H
Q4(b1)
4.4 Overtones and combinations
Q5(b2) Q6(b2)
Figure 8. Approximate forms of normal vibrations of formalde- It is stated above that in the double harmonic approxima-
hyde, H2 CO, with indication of their symmetry species (in paren- tion of molecular vibrations only fundamental transitions
theses). with vk D 1 are allowed in the IR and Raman spectra.
In real molecules, however, mechanical anharmonicity can
Q4 (b1 ) out-of-plane bending mode (100%); lead to the appearance of overtones and combinations as
Q5 (b2 ) almost pure (>99%) asymmetric CH2 stretch- well. Furthermore, if electrical anharmonicity is also taken
ing; into account, then the third (and possibly further) terms in
Q6 (b2 ) almost pure (>99%) CH2 rocking vibration. equation (18) describing the dependence of polarizability
The description is based on calculated total energy distri- on normal coordinates must be considered. This introduces
butions among the cited symmetry coordinates. additional induced dipoles with wavenumbers nQ 0 2nQ k and
The assignment of vibrations to specific symmetry spe- nQ 0 nQ k nQ l ), etc., corresponding to overtones and com-
cies and inspection of the corresponding rows of the char- binations of the fundamental vibrational frequencies. The
acter table immediately leads to information about their IR intensities of these transitions would depend on second-
or Raman activity: we just have to look for components of (and possibly higher)-order derived polarizability tensors
the dipole moment and for components of the polarizability of the type
2
2
tensor. Thus finding z (or z) in the symmetry species of Q1 , @ a @ a

Q2 , and Q3 means that these vibrations have nonzero tran- , , etc. 53
@Qk2 0 @Qk @Ql 0
sition moments (namely, in the z direction), so they are IR
Similar considerations are valid for activities of over-
active. Finding axx , ayy , and azz in the same row means that
tones and combinations in IR spectra where the intensities
these vibrations change all three diagonal elements of the
depend on the values of second order transition moments
polarizability tensor as well, so they are also Raman active.
of the type
Mode Q4 will have transition moment along the x axis (x ) 2
2
and will have a nonzero value for axz . Normal modes Q5 @ m @ m

, , etc. 54
and Q6 will change the dipole moment along the y axis and @Qk2 0 @Qk @Ql 0
have nonzero values for ayz . Consequently, all six normal The activity of overtones and combinations can be eval-
modes of this molecule are both IR and Raman active. uated on the basis of the selection rules formulated in the
When we try to locate the fundamental transitions (vk D most general way in equations (49) and (50), which are
1 vk D 0) in the measured spectra and assign them to independent of assumptions about mechanical or electri-
specific symmetry species, we would like to use, in addi- cal anharmonicity. The procedure is thus similar to that
tion to information about coincidence or noncoincidence followed with fundamental transitions (if the transition
of frequencies observed in IR and Raman spectra, the starts from the ground state) one has to determine the
polarization characteristics indicated in the character table symmetry species of the final vibrational state and check
as well. For IR absorption or emission bands this would the activity of that species. If the symmetry species of
require the use of oriented samples such as oriented single the fundamentals are already known, those of the over-
crystals or molecules aligned in anisotropic media.11,15 tones or combinations of nondegenerate vibrations are
For samples consisting of freely rotating or randomly obtained using the appropriate character tables and calculat-
oriented molecules, only some part of this information can ing the direct products of the characters of the fundamentals
be utilized when interpreting the measured Raman depolar- involved. (The symmetry species of overtones of degenerate
ization ratios, as discussed in connection with the classical vibrations are discussed and tabulated in the monograph by
description. This amounts to distinguishing between totally Herzberg;16 see also Resonance Raman Spectroscopy.
symmetric and nontotally symmetric transitions only. In the We can illustrate this with the character table of point
example of formaldehyde, the Raman bands belonging to group C2v (Table 3). It is easy to see that a binary combi-
Q1 , Q2 , and Q3 must be polarized (r? < 0.75), whereas nation of any symmetry species with itself yields the totally
symmetric species; for example, b2 b2 D a1 for 2nQ 5 or 3. G. Landsberg and L. Mandelstam, Naturwiss., 16, 557, 772
nQ 5 C vQ 6 of formaldehyde. In this point group, the first over- (1928).
tone of any normal vibration will be totally symmetric. In 4. G. Placzek, Rayleigh-Streuung and Raman-Effekt, in
contrast, the combination nQ 4 C nQ 6 (b1 b2 D a2 ) belongs to Handbuch der Radiologie, ed. E. Marx, Akademische
Verlag, Leipzig, Vol. VI, Part 2, 205374 (1934).
the a2 species, thus it may appear in the Raman but not in
the IR spectrum. 5. A. Anderson (ed.) The Raman Effect: Principles, Marcel
Dekker, New York, Vol. 1 (1971).
A noteworthy case is that of the linear triatomic mole-
cules examined in Table 2. We have seen that due to the rule 6. D.A. Long, Raman Spectroscopy, McGraw-Hill, New York
(1977).
of mutual exclusion, asymmetric vibrations with wavenum-
bers nQ 2 and nQ 3 (normal modes Q2 and Q3 ) are inactive in 7. B. Schrader General Survey of Vibrational Spectroscopy in
Infrared and Raman Spectroscopy Methods and Applica-
the Raman spectrum. However, the wavefunctions of their tions, ed. B. Schrader, VCH, Wieinheim, 761
overtones, 2nQ 2 and 2nQ 3 become symmetric with respect to (1995).
the center of symmetry, thus the overtones turn out to be 8. Report on Notation for the Spectra of Polyatomic Molecules,
Raman active. This can also be explained within the frames J. Chem. Phys., 23, 1997 (1955).
of the classical picture in that the first derivative of polar- 9. T.C. Damen, S.P.S. Porto and B. Tell, Phys. Rev., 142, 570
izability for these vibrations (taken at equilibrium) is zero, (1966).
whereas the second derivative is not. 10. T.R. Gilson and P.J. Hendra, Laser Raman Spectroscopy,
Finally, it needs mentioning that overtones and combina- Wiley, London (1970).
tions are relatively rarely observed in nonresonant Raman 11. G. Turrell, Infrared and Raman Spectra of Crystals, Aca-
scattering because, even if allowed by the selection rules, demic Press, London (1972).
these bands are invariably much weaker than the fundamen- 12. G.R. Wilkinson, Raman Spectra of Ionic, Covalent, and
tals. The reason for this can be traced back to the extent Metallic Crystals, in The Raman Effect: Applications, ed.
of mechanical and electrical anharmonicity responsible for A. Anderson, Marcel Dekker, New York, Vol. 2, 811987
the activity of these transitions. As most normal vibrations (1973).
are only slightly anharmonic, the amplitudes of the anhar- 13. W.G. Fateley, F.R. Dollish, N.T. McDevitt and F.F. Bentley,
monic (higher-order) terms in the generalized Fourier series Infrared and Raman Selection Rules for Molecular and Lat-
expression of the time dependence of normal coordinates tice Vibrations: The Correlation Method, Wiley-Interscience,
New York (1972).
are very small compared to that of the harmonic term.17
14. I. Hargittai and M. Hargittai, Symmetry through the Eyes of
a Chemist, VCH, New York (1987).
ABBREVIATIONS AND ACRONYMS 15. J. Michl and E.W. Thulstrup, Spectroscopy with Polar-
ized Light Solute Alignment by Photoselection, in Liq-
uid Crystals, Polymers, and Membranes, VCH, New York
RRS Resonance Raman Scattering
(1986).
YAG Yttrium Aluminum Garnet
16. G. Herzberg, Molecular Structure and Molecular Spectra.
II. Infrared and Raman Spectra of Polyatomic Molecules,
D. Van Nostrand, New York, 125131 (1945).
REFERENCES
17. G. Herzberg, Molecular Structure and Molecular Spectra.
II. Infrared and Raman Spectra of Polyatomic Molecules,
1. A. Smekal, Naturwiss., 11, 873 (1923). D. Van Nostrand, New York, 204205, 245246
2. C.V. Raman and K.S. Krishnan, Nature, 121, 501 (1928). (1945).
Theory of Vibrational Spectroscopy
of Polymers
Chiara Castiglioni
Politecnico di Milano, Milano, Italy
1 INTRODUCTION amorphous bands is in some cases a really hard

task, involving great effort from the point of view
Polymers are relatively complex materials, at of both theoretical modeling and experimental
least from the point of view of their structure spectroscopy.
and morphology. These factors certainly have a Considering the problems outlined above, we
profound influence on the choice of the theoretical will start with a more detailed definition of
approaches suitable for the interpretation and/or polymer structures, and with a qualitative des-
the prediction of their vibrational spectra. For cription of the available tools for theoretical
this reason a preliminary discussion about the modeling of the vibrational properties of some
structural characteristics of macromolecules is prototype structures.
required in order to define the suitable tools for a
correct treatment of any specific case. (i) Crystalline polymers: The historically more
The first step is to distinguish between crys- relevant structure, (at least in relation to
talline and amorphous polymers. While, from the modeling), is that of the linear homo-poly-
point of view of modeling this distinction leads mer, in a crystalline domain. The (appar-
to the use of different theoretical approaches, ently) simplest and maybe over-exploited
the co-presence of two phases (namely crys- case is that of polyethylene. It represents a
talline and amorphous domains) in real, semi- sort of text-book reference example for any
crystalline materials often presents a nontrivial theoretical approach to modeling the vibra-
problem when comparing predicted and exper- tions of crystalline polymers. What makes
imental spectra. The features observed can be crystalline polymers so appealing from the
described as the superposition of the vibrational theoretical point of view is the fact that their
transitions of at least two different species, one characteristic packing in 3-D crystals gives
of them showing conformational disorder at the rise to structures where a single polymer
molecular level. The assignment of crystallinity chain belongs to several unit cells of the
bands, regularity bands, defect bands, and crystal. In other words, in most cases, one
Vibrational Spectroscopy of Polymers: Principles and Practice. Edited by Neil J. Everall, John M. Chalmers and
Peter R. Griffiths. 2007 John Wiley & Sons, Ltd. ISBN: 978-0-470-01662-6.
2 Vibrational Spectroscopy of Polymers
of the three crystal axes (usually referred between adjacent chains, which are usually
to as the c axis) can also be viewed as held together by relatively weak intermolec-
an axis of the polymer chain. This makes ular van der Waals forces, (see for instance
extended sequences within the polymer chain in Figure 1 the sketch of the crystal struc-
1-D crystals, a fact that necessarily implies a ture of polyethylene). As will be shown in
regular chain conformation along the whole the following sections, a first approach to
crystalline domain. As a consequence, a rela- modeling the vibrational dynamics of a crys-
tively small portion of the polymer chain talline polymer chain can be successfully
(with a well defined geometry, and a few made in terms of vibrational dynamics of
degrees of freedom in the vibrational space) a 1-D crystal. This approach benefits from
is needed in order to describe the whole all the advantages of the powerful formalism
chain by means of a translation opera- of solid-state physics, based on the concept
tion along the chain axis. This observa- of crystal phonons. Notice that in the case of
tion is even more relevant if one considers an isolated polymer chain, this treatment is
the fact that polymer crystals are highly further simplified, since only phonons propa-
anisotropic; along the c axis one encounters gating along the chain axis have to be treated.
strong covalent bonds that link the chem- Indeed, the q phonon wave-vector for a 1-
ical units forming the chain; on the other D crystal is by definition parallel to the
hand, moving along either of the other two crystal axis. In most cases, in 3-D polymer
fundamental directions, (namely along the crystals the effect of the crystal field on
a and b axes of the crystal), one jumps phonons with wave-vectors along the chain
H H
C
H H
C
l
C
(a)
(b)
H
C
H
H a
C
(c) H
Figure 1. Sketch of the structure of: (a) a poly(ethylene) lamellar crystal (the vertical lines represent segments of the
molecular chains which fold at the surface) and (b), (c) of its crystalline unit cell.
Theory of Vibrational Spectroscopy of Polymers 3
axis results in a relatively small perturbation appears to be an insoluble problem. Each

to the predictions of the 1-D model, that is, polymer coil is a very large molecule, char-
the intrachain potential dominates over inter- acterized by 3N-6 degrees of freedom in
chain effects. vibrational space (where N, the number of
atoms in the molecule, can easily reach the
(ii) A completely different description can be value of 103 104 ). Moreover, with regard to
made for the amorphous phase. Chain struc- its exact conformation, each coil is unique.
tures typical of amorphous polymers are In other words, the conformation of each
encountered in several different cases: flexible chain is defined in terms of the
sequence of values taken up by the torsional
polymers that are intrinsically unable to angles around each CC bond. This sequence
form crystalline domains and form amor- is different from chain to chain. In spite
phous solids. This happens, for instance, of the apparent impossibility of predicting
in the presence of chemical or structural the vibrational behaviour of a real poly-
defects (as in the case of random co- meric material in the amorphous phase, it
polymers), for linear polymers affected by is actually possible to obtain some useful
random inversion of the link of asym- results, since the polymer chain is made up
metric monomers, (e.g., head-to-head ins- of many units that are chemically identical
tead of headto-tail linking), for atactic (the monomer units). Moreover, the great
polymers, or for polymers with non- variety of conformations that can be found
controlled cis and trans configurations of in a statistical ensemble of polymer coils is
CC double bonds in the main chain. obtained by sampling relatively few choices
polymers with bulky lateral groups are of the torsional angles: for instance, in the
able to reach only a modest degree of case of polymethylene chains, a realistic
crystallinity, due to their very slow kine- description of the chain conformation can be
tics of crystallization; made in terms of trans (T = 180 ), gauche+
polymeric materials obtained by rapid (G+ = 60 ), gauche (G = 60 ) values
quenching from the melt can form amor- of the torsional angles. In these cases,
phous materials; due to the absence of long range interac-
semi-crystalline polymers above their tions between nonbonded atoms along the
melting temperature and polymers in solu- chain, a meaningful set of different con-
tions often take up the characteristic formers of a relatively short chain can be
conformations of the amorphous phase. taken as a satisfactory starting model for
the interpretation of the spectral features
Many flexible linear chains (for instance shown by a statistical ensemble of polymer
polymers with CC single-bond sequences coils.
and CO single bonds in their backbone)
adopt coiled conformations in the amor- (iii) There are several cases that can be regarded
phous (or melt) state, and in solution. Theo- as hybrid cases, since they show some
ries treating these phases in terms of statis- aspects characteristic of the two ideal situ-
tical (random) coils1 have been developed ations described by (i) and (ii). We can
in the past, which are able to account for summarize them as follows:
many of the peculiar physical, structural
and mechanical properties of such materials. (a) 1-D polymer crystals in the absence of
At first sight, from the standpoint of the 3-D crystalline structures: these struc-
theoretical prediction of vibrational spectra, tures can be found, for instance, in
the case of a disordered polymeric phase smectic phases, where orientational order
at the molecular level is reached (in with their chain axis perpendicular to the
this case the chains preferentially choose lamellae surfaces, where chain folding
extended conformations). This is the case takes place, the 1-D crystal that describes
for quenched isotactic polypropylene, for the polymer chain in its straight confor-
which evidence of regular conforma- mation between the two folding planes is
tions, (originating from the well-known a nano-crystal (Figure 1(a)). In Section 4
31 helix characteristic of the 3-D crys- we will discuss how confinement in
tals) has been demonstrated, on the nano-domains affects the vibrational
basis of the analysis of the vibrational spectra, which show non-negligible devi-
spectra.2,3 Several polymers, spun and ations from what is predicted by the
drawn to produce fibers (with character- model of a perfect, infinite 1-D crystal.
istic orientational order on both molec- (c) Localized defects. Several relevant stu-
ular and macroscopic scales), take up dies have been devoted to the identi-
extended conformations of the chains4 : fication of spectroscopic markers that
their vibrations can be theoretically trea- are able to reveal defects in polymer
ted as phonons of a 1-D crystal. Other chains. These defects range from chem-
examples include liquid crystalline poly- ical, (e.g., oxidation and other chemical
mers, which present a rod-like struc- modifications by degradation, intrinsic
ture of the chain and can be modeled defects introduced during the polymer
under the hypothesis of 1-D transla- synthesis, such as chain branching, end
tional symmetry of the chain, which groups, . . .), to structural defects, and
occurs even in the liquid phase. Gener- in particular conformational defects. The
ally speaking, all the polymers made by prediction of the effect of various kinds
rigid segments forming a linear back- of conformational defect on a previ-
bone are much more similar to a 1-D ously conformationally regular chain6 12
crystal than to a random coil, even in the has been one of the most fruitful lines
amorphous phase. Because of the weak of research in the field of theoretical
interactions between different chains, the modeling of the vibrational dynamics
structure and the intramolecular potential of polymers. This activity, in spite of
of the single chain dominates the vibra- the relatively modest computational re-
tional dynamics; this fact has been of sources available at those times, reached
great help in the theoretical studies on a high level of accuracy and provided the
conducting polymers, where the presence tools for fundamental work13 15 that was
of a system of conjugated electrons is able to clarify, for instance, the thermal
at the origin of the quasi planar, straight evolution of polymethylenic chains in
conformation of the chains. terms of well defined changes of the
(b) Nanocrystals. A regular chain in a 3-D chain structure at any stage of a thermal
crystal has several features that cannot treatment.
be accounted for by a rigid application
of the phonon theory, which is devel- 2 CALCULATION OF
oped under the hypothesis of an infinitely
extended crystal, the only model for VIBRATIONAL SPECTRA:
which translational symmetry can be THEORETICAL BASIS
rigorously claimed.5 It is known that AND METHODS
certain polymers form thin lamellar crys-
tals with a thickness in the 520 nm Simulation of spectra (or simply calculation of
range. Since the chains are arranged the vibrational observables) is not only a practice
of great importance from a fundamental point band broadening, intra-molecular or intermolec-

of view, but it is also a powerful tool that ular interactions, etc.) are relevant, and specific
assists the spectroscopist in the interpretation of simulation of band shapes can be carried out.
experimental spectra. Indeed, when the simula- A synthetic spectrum can be directly comp-
tions favourably compare with experimental find- ared with the experimental one in order to gain
ings we have a proof of the correctness of an idea of the quality of the prediction, before
the (approximate) equations used, which means proceeding to the analysis of the eigenvectors,
that the relevant physics is well described by which is of great help in order to establish the
the mathematical model. Moreover, a satisfac- vibrational assignment of the observed features.
tory quantitative agreement validates the use This process allows one to associate an infrared
of the theoretical parameters adopted for the or Raman band to a given normal mode of
prediction (e.g., vibrational force constants, dipole vibration, which in turn helps assignment of
derivatives, electro-optical parameters,. . .) for the the transition to localized or collective motions.
physical characterization of the molecule under Moreover one identifies the atoms or the chem-
study. Accurate calculations help translation of ical groups involved in the vibration and the
relatively complex experimental findings into kind of deformations they undergo, namely the
physico-chemical information such as structure extent of bond stretching, the bending of valence
and symmetry of the molecule, effects of chemical angles, and torsional displacements. Last but not
substitution on the molecular structure, intramole- least, a vibrational assignment immediately allows
cular potential energy and electronic structure, assignment of the symmetry species to which the
phases occurring in the sample, orientation etc. mode belongs. In several cases, the prediction of
A simulation of the vibrational spectra of a infrared and Raman intensities is a difficult task
molecule usually entails a calculation that aims (see below): in these cases, the comparison is
to produce the following information: no longer made between the synthetic and the
experimental spectrum. Nevertheless, the simple
comparative analysis of theoretical and experi-
vibrational frequencies and associated eigen-
mental vibrational frequencies can be exploited
vectors (i.e., normal modes of vibration);
as a very useful tool to understand the observed
symmetry classification of vibrational modes,
features.
and selection rules for infrared and Raman
Conceptually, the vibrational spectrum of a
activities;
macromolecule (a polymer chain) is not different
magnitude of the dipole transitions (associated
from that of any other molecule, although it is
with the infrared absorption intensities), and of
necessary to adjust the theoretical tools needed
Raman cross-sections.
for the prediction of the frequencies, and the
calculation of the intensities. While we refer the
The prediction of frequencies and intensities reader elsewhere16 21 for a comprehensive treat-
generates a simulated spectrum, which can be ment of polymer spectroscopy, in the following
plotted as a series of bands (a Lorentzian band text we will illustrate the fundamental points rele-
shape is used most of the time). The maximum vant to frequency and intensity predictions, which
of each Lorentzian curve is fixed at the value are common in the case of small molecules and
of predicted frequency, and its integrated area polymers.
is fixed according to the computed transition
intensity. Band widths can be fixed for all
the transitions, or adapted in such a way to 2.1 Calculation of frequencies
obtain a better qualitative agreement with the
experimental spectrum. Sometimes very subtle The calculation of vibrational frequencies is
aspects related to band shapes (e.g., origin of usually made for the isolated molecule, (i.e., in
vacuo), under the hypothesis that the intramolec- with:

ular potential is well described by the harmonic R = LR Q (4)
approximation, that is, by a quadratic function
of the vibrational degrees of freedom (Carte- In equations 3 and 4 the symbols have the same
sian or internal displacements of the nuclei, meaning as in equations 1 and 2, provided the
taken from a well defined equilibrium molecular Cartesian coordinates are changed into internal
geometry). Following the classical treatment,22 24 coordinates. Moreover, notice that the use of
which starts from a set of differential equa- generalized coordinates (Rt ) requires a suitable
tions describing the time evolution of degrees definition of the matrix of the kinetic energy GR ,
of freedom of the nuclei obtained according to which now takes the place of the M1 matrix. The
the Lagrangian formulation, the 3N-6 vibrational GR matrix can be built for any molecule following
frequencies are obtained through diagonalization the treatment proposed by Wilson,22 provided
of the dynamical matrix, which can be written the atomic masses and molecular geometry at
on the basis of the 3N atomic Cartesian displace- equilibrium are known.
ment coordinates (xi ), according to the eigen- From the above equations it is clear that the
values equation: prediction of vibrational frequencies of a molecule
requires the knowledge of several parameters,
M1 FX LX = LX (1) namely the set of the molecular force constants
(FX R
ij or Fij ) and the geometrical parameters. There-
where M1 is the diagonal matrix containing the fore an accurate knowledge of the molecular bond
inverse of the atomic masses, FX is the Cartesian lengths and angles (obtained by experiment or
force constant matrix (FX ij = ( V/xi xj )0 ), is
2
computation) is a prerequisite for the vibrational
the diagonal matrix of the frequency parameters analysis.
k = k2 = 4 2 c2 k2 (k = wavenumber, usually At least in principle, the parameters required are
expressed in cm1 ). The k-th column of the eigen- very many in the case of a large molecule (macro-
vector matrix LX k describes the amplitudes of molecule); moreover, the dimension of the math-
oscillation of the set of Cartesian coordinates ematical problem increases, (e.g., the size of the
during the k-th normal mode, with harmonic time dynamical matrix to be diagonalized increases),
evolution (Qk = Ak cos(k t + k ), where k is a as the size of the molecule increases.
phase constant; Ak is the amplitude of the vibra- However, it must be stressed that these prob-
tion). This concept is expressed in a compact way lems can be easily solved for a number of real
by the relationship: polymers: first of all, since the polymer units are
chemically identical, many force constants can be
X = LX Q (2) set to identical values when they involve coor-
dinates of the same kind belonging to different
Notice that the solution of equation 1 gives 6 structural units (monomeric units) in the polymer.
vanishing k parameters, which correspond to the Notice that this procedure can be unambiguously
6 non-vibrational degrees of freedom (translations defined only if valence coordinates are explicitly
and rotations of the molecule as a whole) in the introduced, that is, equation 2 is used.
Cartesian space. As for the large dimension of the dynamical
In a similar way, following the formalism problem, two observations ought to be made:
described by Wilson22 it is possible to write
the dynamical problem on the basis of internal (1) These calculations were very difficult in the
(valence) coordinates (Rt ): in this case one obtains early stages of polymer dynamics, when the
the equation: computational resources were extremely poor:
in 1980 the diagonalization of a 100 100
GR FR LR = LR (3) matrix was considered a huge vibrational
problem, and, in order to treat giant struc- dealing with phonons of inorganic 3-D crys-
tures with conformational defects, ingenious tals, which are always based on force fields
powerful mathematical algorithms, based on expressed in Cartesian coordinates.
the Negative Eigenvalue Theory (NET),25 27 (ii) Vibrational force fields can also be obtained
were developed and used (see for instance, from first principles calculations of the intra-
references 1621) in those years. The compu- molecular potential. With his pioneering
tational cost of such big problems appear work, Pulay opened this new field of applica-
today to be very modest, especially if comp- tion of quantum chemistry, obtaining accurate
ared with the demands of high level first force fields for several organic molecules,
principles calculations (e.g., ab initio molec- in the frame of ab initio HartreeFock
ular orbital or density functional theory calcu- molecular orbitals calculations.28 At that time
lations), which are routinely performed for this research activity was restricted to a few
molecules of non negligible size (4050 experts, since in many cases the comparison
heavy atoms). between predicted and experimental results
(2) As will be made clear in Section 3, because of were not so immediate (e.g., relatively comp-
translational symmetry, the dynamical prob- lex scaling procedures were necessary). In
lem for a conformationally regular polymer, contrast, the high level of approximation
(treated as a 1-D crystal), can be reduced to that can be reached today allows one to
several small problems with the dimension obtain good vibrational force field and accu-
of the number of degrees of freedom of rate predicted frequencies, even without any
the unit cell. An even more efficient use of scaling procedures. A general improvement
symmetry can be introduced in the presence in frequency prediction can be obtained with
of a screw axis, taking explicit account of the use of density functional theory (DFT)
the helical symmetry and thus reducing the methods,29 provided that suitable functionals
dimension of the dynamical problem to the and large basis sets of atomic orbitals are
number of degrees of freedom of the minimal chosen. High-level DFT calculations often
structural unit, which usually coincides with allow one to predict frequencies that fit
the monomeric unit. the experimental ones with deviations not
exceeding 1015 cm1 . The more popular
The vibrational force fields can be derived quantum chemical programs yield as a stan-
following two different approaches: dard output the harmonic vibrational potential
(in Cartesian coordinates), frequencies and
(i) empirical force constants can be used, which normal modes which are directly calculated
in most cases are derived from the refine- according to equation 1 by the same program.
ment on the experimental frequencies of a The results can be routinely visualized thanks
common set of force constants for a family to a graphical interface, which directly shows
of molecules containing the same chem- the animation of normal modes.
ical groups. A polymer naturally finds in
its oligomers the more suitable partners for In the case of relatively small molecules,
the construction of a common empirical quantum chemical predictions can be directly
force field. Also, this procedure is practi- used to guide the interpretation of infrared and
cally affordable only if one treats the vibra- Raman spectra. However, for the case of a
tional problem on the basis of internal valence polymer this still remains a nontrivial task,
coordinates, a fact that generally makes the due to the prohibitively large dimension of the
approach to polymer crystals quite different quantum chemical problem. Three ways can be
from that followed by solid-state physicists followed in order to take advantage of a quantum
chemical calculation of the spectroscopic response In the simple case of a transmission experiment
of a polymer: with a sample in solution (i.e., randomly oriented
molecules), the absolute infrared absorption inten-
(i) First principles calculations can be perfor- sity of the i-th transition is defined as:
med on small model molecules (oligomers),

and then the theoretical predictions can be 1 I0
transferred or extrapolated to the limiting Ai = band ln d
Cl I
case of an infinitely long polymer chain. This 2 (5)
M
technique has been followed by Hirata et al.,
=K
who built one of the best force fields for Qi
polyacetylene30,31
(ii) Quantum codes suitably derived in order where C is the concentration of the sample, l the
to treat translational symmetry can be used length of the cell, I0 and I are respectively the
for the description of regular chains as 1- incident and transmitted light intensities. K is a
D crystals. The available programs allow constant whose value depends on the units used.
one to obtain phonon dispersion curves with The intensity of the Raman line associated with
frequencies calculated under the harmonic the i-th transition is given by:
approximation, based on Cartesian force
fields.
Ii = kI0 (s )4 |( )i |2 (6)
(iii) Sometimes semi-empirical treatments give
,
better results than full ab initio predictions.
In these cases, only some physical quantities

are obtained on the basis of quantum theory, ( )i = 1i |Qi |0i
while several other parameters are fitted Qi 0
to the experimental data. This technique 1 (7)
h 2
has been applied for instance in the case =
of polyconjugated polymers32 and carbon Qi 0 8 2 ci
nanotubes.33 In these cases, long range inter-
actions between stretching coordinates of where s is the wavenumber of the scattered
the conjugated CC bonds were successfully beam and i is the vibrational wavenumber, h
predicted on the basis of the very simple the Plank constant, and c the speed of light in
Huckel theory. vacuo. Notice that the above definition applies
to an experiment where all the light scattered by
the sample is collected and measured. Usually
2.2 Calculation of intensities Raman experiments are performed according to
standard experimental geometries such as 90 or
Under the double (mechanical and electrical) 180 scattering, and the appropriate expression for
harmonic approximation, the problem of the Raman intensity has to be defined accordingly.
prediction of infrared and Raman intensities can These expressions look like equation (6) but the
be reduced to the calculation of the relevant elec- term ( /Qi ) is replaced by the appropriate
trical parameters responsible for the infrared and combination of its invariants.22,24
Raman fundamental transitions, namely M /Qi As in the case of frequencies, vibrational inten-
(the derivative of the molecular electrical dipole sities can be calculated in two different ways,
moment with respect to the normal vibration Qi ) namely through empirical electrical parameters
and /Qi (the derivative of the molecular as well as by fully theoretical (first principles)
polarizability tensor with respect to the normal treatments. For a comprehensive discussion of
vibration Qi ).22 24 the theory of molecular infrared intensities, and
of the methods available for parametrization and kind are not included yet in standard quantum
simulation, we address the reader to a recent chemical software.
review by Gussoni et al.34 Only recently, computational codes were devel-
In this brief discussion we only mention the use oped in order to solve typical solid state prob-
of infrared electro-optical parameters,35,36 which lems focused on the calculation of vibrational
for many years have been used with success not intensities under the hypothesis of translational
only as parameters for the prediction of spectra, symmetry.47 So far, however, the field of applica-
but also as physical quantities, (i.e., localized tion remains restricted to a few cases tackled by
atomic charges and charge fluxes), able to give the very few researchers who personally devel-
a meaningful picture of the electron distribution oped the software. What today is routinely afford-
in a given molecule (or chemical group), and of able is the calculation of vibrational intensi-
its mobility during vibrational displacements.37 ties of model molecules (oligomers), a practice
Examples of successful application of these that can be helpful in order to detect trends
parameters to the study of polymers are in intensity and/or to define transferable elec-
presented in references 3841. Raman electro- trical parameters, useful for predictive purposes
optical parameters (bond polarizabilities and in the case of polymers. As an example we can
their derivatives) have been developed for quote a recent study on Raman intensities of
the interpretation of Raman intensities.42 44 In polyenes carried out by high-level DFT calcula-
contrast to the case of infrared spectroscopy, the tions; this work has been of great help in under-
application of this model remained restricted to standing the origin of the Raman response of
a few cases and its development was severely polyacetylene.48,49
delayed. These facts are simply the consequence
of at least two problems, namely the lack of
3 PREDICTION OF THE
reliable data for absolute Raman cross-sections
and, the fact that Raman intensities depend on PHONONS OF A 1-D CRYSTAL
a collective response of the electronic cloud, As pointed out in the introductory section of
which in several cases cannot be modeled this chapter, from the standpoint of a theoret-
by localized and transferable parameters. It ical treatment of vibrations, a polymer does not
is however pleasing to notice that the recent differ from any other molecule, but for the large
great improvement reached in the prediction dimensions of its vibrational problem. However,
of the Raman intensities by first principles a different theoretical approach is adopted for a
calculations has revitalized the research on Raman linear chain showing regular conformation, since
electro-optical parameters; bond polarizabilities translational symmetry can be used in this case
for families of molecules have been derived from in order to develop an elegant formalism that
theoretically predicted Raman intensities.45,46 On drastically simplifies the treatment and lowers the
this basis, optimized sets of Raman parameters dimension of the dynamical problem. The phonon
that are useful for the prediction of the Raman theory of a perfect crystal, starting from the
response of complex molecules or solids have simple example of a linear chain (i.e., 1-D crystal)
been proposed. Although this procedure has not made up by identical atoms linked to first neigh-
yet been applied to polymers, it seems that in bours by harmonic springs is widely treated in
the future this may be a valuable approach. On any classical text-book of solid-state physics50,51 :
the other hand, a fully first principles prediction this formalism is at the basis of any treatment
of infrared and Raman intensities of a polymeric of phonons in polymers.52 54 However, when
material can be, at least in principle, obtained with dealing with real polymers, an important advan-
the use of codes where the translational symmetry tage can be obtained if the theoretical treatment
is implemented. However, calculations of this is built in such a way as to take explicit account
of the screw axis symmetry, which often occurs Another important point that sometimes makes
in polymers with regular conformations. This can the treatment of the vibrations of polymers
be better realized if one considers the fact that formally different from that followed by the
usually the minimal structural unit, (often coin- common approaches of solid-state physics is
cident with the monomeric unit), of a crystalline the use of internal (valence) coordinates, which
polymer does not coincide with its translational yield more easily transferable parameters and
unit. Indeed in most cases several monomeric intramolecular potentials that are invariant under
units related to each other by a screw-symmetry screw-symmetry operations.
operation form the translational repeat unit of the For the above reasons, the treatment of the
1-D crystal. For instance, in the case of isotactic phonons of polymers illustrated in this section
polypropylene, (CH2 CHCH3 )x , the translational will refer to the general case of an isolated regular
unit of the 31 helix contains 3 monomeric units, polymer chain (NM helix), whose geometry can
each of which can be considered as an elemen- be built by application of a screw-symmetry
tary structural unit linked to the adjacent units by operation to a reference chemical unit of p
elemental steps of the screw operation (rotation atoms. Moreover, internal coordinates will be
of = 2/3 about the chain axis, followed by a chosen as the basis set for the description of the
translation of c/3 along the same axis, where c vibrational space.
is the length of the translational unit cell). In this Let s(, L) be the screw-symmetry operation,
case one can handle the vibrational problem on the defined by a rotation of an angle = 2M/N
basis of the minimal structural unit, reducing the followed by a translation along the polymer axis
degrees of freedom from 81 (translational unit) to by L = |c|/N , where c represents the elemen-
27 (structural unit). For polymers with larger unit tary translation along the crystal axis. Let Ri
cells the effect may be even more drastic; consider describe the i-th internal displacement coordinate
for instance the case of the 157 helix of crys- belonging to the n-th chemical unit.
talline polytetrafluoroethylene, (CF2 )x , where use In the harmonic approximation, the intramolec-
of the translational symmetry alone would imply ular potential energy of the polymer chain can be
a problem with 15 9 = 135 degrees of freedom, expressed as a quadratic form as follows:
which can be lowered to 9 if one works with the
1
i,k
CF2 group as the elementary structural unit.55,56
n n
Moreover, a treatment based on the structural V = (FR )nn
ik Ri Rk (8)
2 n,n
unit allows one to handle simultaneously different
polymorphs with different conformations, namely
In equation 8 the index i ranges over all the
other helices characterized by a different rota-
internal degrees of freedom of the unit n (i. e.,
tional angle, such as the 136 helix or the trans-
3p degrees of freedom, in the absence of local
planar modification of the same (CF2 )x polymer.
redundancies between internal coordinates), the
An even more impressive case is that of carbon
index k ranges over all the internal degrees of
nanotubes, which can be regarded as complex
freedom of the unit n , and the sum over n and
polymers generated by a bundle of helices based
n is extended to units over the whole polymer
on the characteristic unit cell of graphene (two
chain.
carbon atoms). Using the full screw-symmetry,
The invariance of the second derivatives of
the vibrational problem can be reduced to the
the intramolecular potential with respect to the
degrees of freedom of a structural unit containing
order of derivation and the periodicity of the chain
only two atoms; moreover, it can be treated with
implies respectively:
a common formalism for any nanotube with its
characteristic diameter and chiral angle, that is,
ik = (FR )ki ;
(FR )nn nn
the structure of the relevant equations can be (9)
s
ik = (FR )ik = (FR )ki
(FR )nn s
settled once and for all.33
where s = (n n) represents the distance of each choice of . Frequencies values are given
interaction. by the determinantal equation:
The use of the relationship described by equa-
tions 9 into 8 leads: |GR ()FR () 2 ()E| = 0 (12)

i,k
i,k
In equation 12
2V = (FR )0ik Rin Rkn + (FR )sik Rin Rkn+s

n n,s
GR () = (GR )0 + (GR )s eis + (GR )s eis

i,k s

s
+ (FR )ski Rin Rkns (10) FR () = (FR )0 + (FR )s eis
(FR )s eis +
s s
n,s (13)
and E is the unit tensor of dimension 3p 3p.
It can be seen from equation 10 that the Let us point out a few important points related
intramolecular interactions can be grouped into to the above equations:
three different terms which contain: (i) diagonal
and interaction force constants relative to internal (i) Equation 12 is of 3p-th degree in 2 ; there
coordinates all belonging to the same (n-th) struc- are 3p characteristic roots 2 = 4 2 c2 2
tural unit, namely force constants labelled with for each value of the phase difference .
the superscript 0; (ii) interaction force constants The dispersion relation = f () is then a
between pairs of internal coordinates belonging, multiple-valued function with 3p branches.
respectively, to the n-th unit and to that at distance For a chain containing W = N Nt struc-
s, (s positive screw steps); and (iii) interaction
tural units (Nt being the number of the trans-
force constants between pairs of internal coordi-
lational units in the 1-D crystal), the appli-
nates belonging respectively to the n-th cell and
cation of the cyclic Born and von Karman
that at distance s (s negative screw step). The
condition57 yields the explicit expression for
kinetic energy 2T of the infinite chain can be
the allowed phase factors, namely:
written according to the same formalism. The
Lagrange equations of motion can be obtained
M
starting from the above expression of the poten- = 2 r
N Nt
tial (kinetic) energy. In this way a system of (14)
(3p W ) (W = number of structural (chemical) M
= 2 r, r = integer value.
units in the polymer, W ) second order W
differential equations in the unknown Rin+s is
obtained. In order to obtain the expected 3p W solu-
Solutions can be found of the form: tions, the integer values of r have to be chosen in
the range (W/2) < r (W/2) giving values
Rkn+s = Ak exp[i(s t)] (11) in the range M < M. Any choice of r
outside this interval does not give any new solu-
In equation 11 Ak is independent of n, is the tion of the dynamical problem, since both and
so called phase shift between equivalent (by the eigenvectors are periodic functions of , with
screw operation) internal coordinates belonging to period 2. However, due to this periodicity it
two adjacent roto-translational units, and is the can be realized that for any value of (chosen
vibrational frequency. Substitution of equation 11 according to equation 14), such that || exceeds
into the system of differential equations leads to , (this happens, according to equation 14 for a
3p simultaneous linear equations in the unknown helix where the translational unit corresponds to
amplitudes whose nontrivial solutions are given more than one turn of the helix, namely for M
by the 3p values of , which can be obtained for = 1), there exists a value in the range <
such that the corresponding solutions of (iii) Notice that among the 3p W roots calcu-
the dynamical problem are identical. Clearly lated for each value, according to equa-
can be written in the form = 2 m, where tion 12, 4 vanishing roots ( = 0) are expec-
m is the suitable integer value. Moreover, due ted, corresponding to the cyclic polymeric
to the mirror symmetry of the phonon disper- redundancies that occur among the internal
sion curves with respect to the y axis ( axis), coordinates of the 1-D crystals. These roots,
(which is a consequence of the degeneracy in the in a treatment based on Cartesian displace-
energy for phonon with phases and ), only ments, would correspond to the 4 nonvi-
the positive half of the dispersion curve is usually brational degrees of freedom of the nuclei,
plotted. In conclusion, phonon dispersion curves namely the three pure translations and the
are usually reported only for positive values and only allowed rotation (around the chain axis)
are restricted to the interval 0 . of the polymer chain as a whole. As a conse-
(ii) Equation 11 parallels the more popular exp- quence, two of the 3p branches always reach
ression50,51 for the vibrational displacements zero for = 0; these branches are commonly
known as acoustical branches. Depending on
associated with a crystal phonon, as obtained
the geometry of the chain, it may happen that
working on the basis of the translational cell
onea or both acoustical branches reaches zero
(unit vector a) of the lattice:
again for those = 0 values which corre-
spond to q = 0. The other 3p-2 branches are
Rin+u = Ai exp[i(q a u t)] (15) optical branches, since for particular values
of (depending on the geometry of the
where q is the phonon wave-vector taken chain) they can give rise to spectroscopic
in the first Brillouin zone of the reciprocal activity in the infrared and/or in the Raman.
lattice. While it is trivial to define a phase (iv) Some comments ought to be made about
associated with the expression reported the phase dependent force constant matrix
described in equation 13. Formally, the
in equation 15, (namely = q a), attention
sum over s runs over distant structural
must be paid if a correspondence between
units, but for non-conjugated polymers force
the value which appears in equation 11
constants describing interactions between
and the value of of equation 14 has to
coordinates belonging to structural units at
be stated. This correspondence can be imme-
a distance larger than one or two helix
diately obtained considering the fact that steps (s larger than 1 or 2 in equation 13)
represents the phase difference between are usually very small and are neglected.
equivalent coordinates Rk belonging to adja- A completely different case is that of
cent structural units along the helix, while polyconjugated polymers, for which the
is the phase difference between equivalent prediction of accurate dispersion relations
coordinates Ri belonging to adjacent trans- would require that long range interactions
lational unit cells. Considering that N steps are included (see Vibrational Spectroscopy
along the helix corresponds to an elementary of Conducting Polymers: Theory and
translation, we immediately obtain the result: Perspective).
= N. For instance, for a 21 helix (e.g., (v) General selection rules for optical transitions
the case of the poly(ethylene)) the value = can be easily worked out on the basis of
/2 corresponds to = and the value the translational symmetry. If we consider
= corresponds to = 2 0, namely fundamental transitions (i.e., those allowed in
to q = 0. This last point is very relevant when the double harmonic approximation), conser-
selection rules for optical transitions have to vation of momentum implies50,51 that only
be worked out (see v). phonons with wave-vector q = 0 can be
excited by interaction with the radiation field. reference 0-th unit. More formally:
According to what was previously stated (see
point (ii)), only phonon transitions corre- Rkn = A0k exp[i(n t)] (16)
sponding to = 0 and = 2M/N are opti-
cally allowed. On the other hand, in the case The vibrational modes of the polymer described
of a breakdown of the hypothesis of a strict according to equation 16 are amplitude waves
double harmonicity, selection rules relax, with frequency , travelling along the polymer
since the momentum conservation can be chain. The displacement A0k , which characterizes
guaranteed by combination of wave-vectors the coordinate Rk0 belonging to the 0-th unit
of the several phonons involved (e.g., the at time t = 0, is found for the corresponding
simultaneous activation of two phonon with coordinate Rkn belonging to the unit n at the time
wave-vectors q and q conserves the total t = n/. In the language of solid-state physics
momentum). these kinds of vibrations are called phonons of the
(vi) Selection rules dictated by the translational 1-D crystal.
symmetry represent the necessary but not Notice moreover that the two counter-propag-
sufficient conditions for infrared and/or ating travelling waves Q|| and Q|| at the same
Raman activity of the fundamental transi- frequency , but with opposite values of the phase
tions. The full symmetry of the polymer factor can be combined as: Q+ = 1/21/2 (Q|| +
will determine the activity in the vibra-
Q|| ) and Q = 1/21/2 (Q|| Q|| ) in order
tional spectra of the phonons with specific
to obtain stationary waves in the polymer chain.
symmetry. Notice that phonons at q = 0
The construction of stationary waves from two
(which are indeed the only optically active
degenerate counter-propagating phonons is the
ones in the first order vibrational spectra,
procedure that allows one to compare phonons
see point iv) can be treated in terms of
with normal modes of vibrations in molecules or
the symmetry of the point group, which is
oligomers, which indeed are necessarily stationary
isomorphous with the polymer line group.
waves due to the confinement within a molecule
Symmetry and momentum conservation thus
of finite size (see Section 4).
explains why the vibrational spectra of poly-
An alternative procedure to that outlined above
mers with highly symmetric and small struc-
takes explicitly into account the roto-translational
tural units, for example, poly(ethylene), are
symmetry by the introduction of symmetry ada-
so simple, showing very few lines both in the
pted vibrational coordinates, which transform
infrared and in the Raman.
according to the character table of the cyclic
group which contains all the symmetry operations
Equation 11 states that, during a given polymer associated to the screw axis, as illustrated in
vibration characterized by an allowed frequency references 53, 54.
, the amplitude of the displacement of a given Symmetry coordinates Sk , for any allowed
vibrational coordinate Rk (belonging to a given value can be defined as linear combinations of
structural unit) at a given time t undergoes the the equivalent internal coordinate belonging to the
same displacement as that of a unit at a distance different structural units, where the coefficients
s, provided that it is multiplied by a phase factor are given by the suitable phase factors. According
exp(is ). to the symbols previously introduced:
Since the displacement Ak is independent of
n, we can arbitrarily choose a reference unit (0- 1
th unit) along the polymer: accordingly we will Sk = exp[i(n)]Rkn ;
W (17)
relate any vibrational displacement in a structural 2M 2M
unit different from 0 to the displacement in the with = ( )r = ( )r
N Nt W
It can be shown that, because of the invari- n are linked to the chain only from one side,
ance of the intramolecular potential and kinetic while in a chain characterized by fixed ends
energy under screw-symmetry operations, the these two atoms are linked to masses of infinite
introduction of the above set of symmetry coor- value by the same springs which describe first
dinates allows one to factorize the dynamical neighbor interactions along the whole chain. In
matrix (dimension (3p W ) (3p W )) into other words, the problem is no longer represented
W blocks, each of them corresponding to a given by a cyclic dynamical matrix, which applies to
value. In this way one immediately reaches the the infinite case, characterized by the Born and
eigenvalue equation 12. The eigenvectors give the von Karman boundary conditions. The first impor-
relative amplitudes of any Sk coordinate involved tant consequence is that vibrations of these model
in the phonon; Sk can be in turn projected onto oligomers are not described as travelling waves
internal coordinates, yielding the amplitude wave (phonons), but are necessarily stationary waves,
described above. that is, normal modes. However, in the case of
In the frame of the theory of the phonons finite chains it is also possible to handle the math-
of 1-D crystals illustrated above, several empir- ematical problem in an analytic way; it can be
ical valence force fields were developed together shown that the oligomer frequencies exactly fit the
with calculations of the phonon branches for dispersion relation obtained for the parent poly-
a variety of crystalline polymers. Among mers (see Figure 2) according to the relationship:
these we can mention isotactic poly(propylene),
poly-(vinylchloride), poly(ethylene terephtalate),
trans-poly(isoprene), trans-poly(butadiene) and
j
j = 2 sin (18)
poly(tetrafluoroethylene). References to the orig- 2
inal works, together with illustrations of the
predictions obtained, and a critical comparison where = F /m, and j are phase factors that
between theoretical results and experimental find- have to be selected in a suitable way according to
ings can be found in reference 54. the model chosen:
j
4 POLYMER AND OLIGOMERS: j = with j = 0, 1, . . . n 1 (19)
n
DISPERSION CURVES AND
FREQUENCIES OF FINITE
LINEAR CHAINS
2(K/m)1/2
In classical books52,54 dealing with polymer vibra-
s=1
tions, a chapter is usually devoted to the analytic s=2
s=3
description of the longitudinal normal modes of s=4

a finite linear chain formed by n identical atoms s=5
of mass m, linked each other by ideal springs of s=6
force constant F . Free ends
s=7
The dynamical matrix, which has to be diag- Fixed ends
s=8
onalized, is identical to that of the ideal infi- 0 q /c
nite case (linear polymer), with the only excep-
tion of the matrix elements describing interac- Figure 2. Dispersion relations of the frequencies of the
longitudinal phonons of a linear chain with one atom as
tions that involve end atoms. These interactions the basis. Squares and circles represent the frequencies of
differ according to the model adopted: in a model model oligomers with 8 atoms, under the hypothesis of
chain characterized by free ends, atoms 1 and free and fixed ends, respectively (see text).
for the case of free ends where j is an integer ranging from 1 to n 1

and n is the number of unit cells contained in
j L(L = an). is the characteristic wavelength of
j = with j = 1, 2, . . . n (20)
(n + 1) the stationary wave and q its corresponding wave-
vector. The discussion above, simply based on a
for the case of fixed ends linear chain with only one atom per unit cell, can
The normal vibrations (of the oligomer) are be generalized to real polymer structures whose
described as stationary waves which can be built chemical unit carries p degrees of freedom. In this
as a combination of two phonons of the corre- case, each suitable choice of the phase j would
sponding ideal polymer, travelling in opposite extract from the phonon dispersion curves of the
directions, i.e. characterized by wave-vectors qj parent polymer the characteristic frequencies of
and q j respectively (j = qj a, a = interatomic the oligomer considered.
spacing). An appropriate initial phase has to be Since translational symmetry does not hold
included in the solution (eigenvector) in order to for oligomers, all the frequencies selected with
obtain the correct description of displacements at the procedure described above can be (in prin-
the end of the chain (this factor is indeed different ciple) observed in the infrared and Raman spectra.
in the case of the fixed ends model and of the However, selection rules determined by the point
free ends model).52 In Figure 3(a) sketch of the symmetry of the oligomer will determine the
eigenvectors is reported for the case of a free ends activity of any normal mode in the vibrational
oligomer with eight atoms. spectra.
The choice of a suitable value for j , as deter- As a consequence, a series of bands, which
mined by boundary conditions for the oligomer, are indeed located on the dispersion curves of
can be derived with the same rationale that allows the parent polymers, appears in the vibrational
one to select the stationary waves characteristic of spectra of the molecular species. These series
a cavity of fixed length L. In this case (/2) j = of bands are usually referred as sequences
L, that is q = (2/) = (2j/2L) = (2j/2na), or progression bands. Sequences of bands
related to relatively steep phonon branches give
rise to a very characteristic pattern in the vibra-
= 18 d tional spectra, formed by families of lines with
8
increasing intensity, which usually merge into a
= 18 d
7 stronger band, corresponding to the only allowed
optical phonon of the polymer (q 0 phonon)
= 18 d
6 (see for instance in Figure 4 the CH2 rocking
= 18 d
sequence as observed58 60 in the infrared spec-
5
trum of a crystalline sample n-C24 H50 ). The
= 18 d observation of sequences in oligomeric samples
4
provides information that is very useful for
= 18 d the construction of the phonon branches of the
3
polymer. A great contribution to the study of the
= 18 d physics of classical polymers was given as early
2
as 1963 by Snyder and Shachtschneider,58 60 who
= 18 d applied these concepts in their detailed study of
the dynamics of poly(ethylene) and of many n-
Figure 3. Longitudinal vibrations of a linear chain of alkanes in the crystalline phase.
8 masses (Bond length = d) with fixed ends. The real
displacements of the masses can be obtained by turning For very long chain molecules, the breakdown
the vectors by 90 clockwise, as indicated for the mode of the translational selection rule is usually effec-
with = (18/8)d. tive only in the activation of phonons near to
Wavenumber (cm1)
a consequence of some geometrical confine-
ment, can show, in some cases, the activation of
1100 1000 950 900 850 800 750 700
100 some q = 0 phonons in its vibrational spec-
trum. This is exactly what happens for the so
80 called longitudinal accordion modes (LAM)61 63
of poly(ethylene), located along the longitudinal
% Transmission
60 acoustic phonon branch, where the translational

symmetry is broken by chain folding.
40 LAM modes of linear alkanes are easily recog-
nizable in the Raman spectrum; for n-alkane
20 molecules, (n even), if j represents the number
of nodes, because of symmetry, only modes with
0
odd j are Raman active, and the Raman inten-
9 10 11 12 13 14 sity strongly decreases when j increases. LAM-1
Wavelength (microns) has received most of the attention since it gives
Figure 4. The CH2 rocking sequence as observed58 60 rise to the strongest line in the Raman spectrum,
in the infrared spectrum of a crystalline sample n-C24 thus becoming a very useful spectroscopic probe.
H50 . The strong band at about 720 cm1 corresponds to Mizushima and Shimanouchi61,62 have shown that
a phonon of the 1-D crystal (poly(ethylene)) at q 0. if a segment of a trans-planar poly(methylene)
molecule is represented by an elastic rod of length
the zone-centre (q 0 phonons). This can be L, density and Young modulus E, the LAM-1
easily realized considering the shape of the normal frequency [cm1 ] can be related to the observed
modes of the molecule: these modes can be clas- frequency by:
sified according to the number of nodes of the 1/2
1 E
associated stationary wave. For instance, in the (LAM-1) = (21)
case with fixed ends, the mode with the longest 2Lc
(lowest q) allowed (/2 = L) is characterized Once the numerical values of E and are
by only two nodes at the boundaries (end of the known for a given polymer, the length L of
chain): in this case the phase difference between the molecule (between folds) can be immediately
amplitudes of vibration for atoms belonging to derived from the observation in the Raman of their
different repeating units of the chain is always less (LAM-1). On the other hand, if a sequence
than 180 . As a result, the contribution by each of several LAM-j modes can be experimentally
unit to the change of the total molecular dipole observed for a given finite chain, part of the
(for the infrared) and to the molecular polariz- acoustical branch can be fitted to the experimental
ability (for the Raman) is added to that of the data, and the elastic modulus of the polymer can
others, (i.e., the effect of each unit is coopera- be derived.
tive). Modes with higher q values are described LAM modes have been used in structural
as an amplitude wave with more than 2 nodes. In determinations of many molecular segments. As
this case we have several pairs of units along the an example we can mention the contribution
chain that vibrate with the opposite phase, giving of LAM spectroscopy to the study of the
rise to contributions to the total dipole (polariz- (orthorhombic pseudo-hexagonal melt)
ability) that cancel each other. This mechanism of phase transition of n-nondecane (taken as
cancellation is more and more efficient as the prototype of a class of molecules).15 At T <
number of nodes increases and as the dimension 0 C the observed value (LAM-1) = 124.5 cm1
of the chain increases. indicates that n-nonadecane has 19 carbon
On the other hand, comparison with the case atoms in a trans-planar geometry. Near T =
of the oligomers suggests that a polymer, as 27.78 C, (LAM-1) = 123.2 cm1 , the chain is
still straight, but the weak surface interactions (ii) Long range interactions. The intramolecular
at either end of the chain (interlamellar forces) potential may be affected by the length of the
further weaken, at T > 32 C the molecule chain. In this case the force constants are no
collapses in a disordered structure and LAM- longer transferable from oligomers to poly-
1 disappears, leaving a broad scattering near mers and are, at least in principle, different
220 cm1 . This feature has been ascribed to at different sites along the chain. This effect
skeletal motions of the coiled chains and is is usually found in systems characterized by
referred to as pseudo-LAM or D-LAM.64 long range interaction force constants; this
The observation of LAM modes has also happens, for instance, in the case of organic
contributed greatly to the study of the folding of materials containing conjugated electrons,
polymer chains during the formation of lamellae. with electronic structure strongly sensitive to
As early as 1978 Snyder et al.65 were able to the delocalization path available (and hence
modify the existing Raman spectrometers in order to the size of the crystal). As a consequence,
to reach extremely small Raman shifts (reaching one finds a continuous change in the values
3 cm1 ) for the recording of LAM modes of spec- of some force constants often accompanied
imens of polyethylene as extruded or as bulk crys- by an increase of the distance of interac-
tallized materials. From the shape of the observed tion as a function of the size of the crystal.
LAM-1 band, the authors were able to obtain The signature of this phenomenon in the
a quantitative distribution of lengths of straight Raman spectrum is a frequency shift of some
chain segments existing in the polymer lamellae. characteristic Raman band, with respect to
Even though it is very powerful, the treatment the frequency of the parent (ideally infi-
of vibrational modes of polymer and oligomers nite) crystal. Systematic dispersion of the
described above has some weakness and limita- Raman frequency of the strongest Raman
tions, which we now briefly discuss. active modes has been observed for linear
polyenes with increasing chain length.66,67
(i) End defects. The ends of a real linear polymer These frequencies do not reach the limit of
(and of its oligomers) are functional groups the characteristic frequencies of polyacety-
different from the ones that form the chain: lene even for very long chains. The effect
masses and force constants involving end described above for the case of a 1-D crystal
group should be chosen according to the real cannot be simply explained by the fact that
chemical nature of the group. As a conse- normal modes of finite size chains correspond
quence of the presence of end defects in to phonons with small (but finite) q values.
polymers, we expect to find specific signals Due to the large frequency shifts observed,
in the vibrational spectra; characteristic group this would imply having very steep optical
frequencies corresponding to normal modes phonon branches (near q = 0) for the parent
localized on end groups (end group modes) crystal. For this reason a change of some
can indeed be detected in the case of rela- relevant force constants (e.g., a decrease of
tively short chains. End group modes are the relevant force constant with increasing
few and can be easily located in the spec- chain length) has been proposed67 in order
trum because their relative intensity decreases to account for the observed phenomena. In
when the number of repeated units in the Vibrational Spectroscopy of Conducting
chain increases. These signals are often easily Polymers: Theory and Perspective of this
detectable in the infrared spectrum, where book a thorough discussion of the conse-
they are routinely used for an estimation of quence of the delocalization of electrons
the average chain length in low molecular in polyconjugated polymers and oligomers
weight polymer. (Some examples are reported on intramolecular potential and spectroscopic
in references 1621). features is presented.
5 REGULARITY AND a doublet is indeed due to the fact that any q = 0

phonon of a given regular chain (e.g., in phase
CRYSTALLINITY BANDS
CH2 rocking of polyethylene) can couple with
One of the more debated questions concerning the same phonon of the second regular chain
the structural diagnosis of polymer materials via belonging to the same cell, leading to two combi-
vibrational spectroscopy concerns the possibility nations where the two chains vibrate either in-
of establishing whether the material is crystalline phase or out-of-phase. These two combinations
to some extent. According to the above discus- are still q = 0 phonons of the whole 3-D crystal
sion it should be clear that a description of the (i.e., phonons at point of the first Brillouin
polymer vibrations in terms of a 1-D crystal, Zone of the 3-D reciprocal lattice) and are both
(the single polymer chain in a regular confor- potentially active in the vibrational spectra. Obvi-
mation), is often a good approximation for the ously, the point symmetry of the crystal will
phonons propagating along the polymer axis. We dictate symmetry selection rules, which estab-
have already noticed that selection rules dictated lish whether the two components are simulta-
by translational invariance allow the activity of a neously active in the infrared (and/or Raman)
relatively small number of transitions that corre- spectrum. As an illustration of this point, we
spond to excitation of q = 0 phonons of the 1-D show in Figure 515 the thermal evolution of bands
crystal in the infrared and/or Raman spectra to be belonging to the rocking sequence observed in
predicted. This is the reason why a spectral pattern the case of n-nonadecane while going from the
characterized by few sharp lines (regularity bands) crystalline orthorhombic phase to a pre-melting
is often reported as evidence of the presence of phase, usually referred to as the phase. The
a large amount of crystalline material. However, main characteristic of this phase is the appearance
it is important to stress the fact that the appear- of disorder in the crystalline plane orthogonal to
ance of a spectral pattern that can be interpreted as the chain, accompanied by a substantial conser-
being due to a 1-D crystal is not a sufficient condi- vation of the orientational and conformational
tion for a safe diagnosis regarding the presence of order of the chains; regularity bands are indeed
3-D crystals. The observation of transitions that observed both in the crystalline and the phase,
can be accounted for on the basis of the dynamics but in the last case the crystal splitting, which
of a 1-D crystal only allows one to state that in is the signature of the regular packing in the 3-
the sample under investigation the polymer chain D crystal, is gradually lost as the temperature
has a regular conformation. On the other hand, for increases. On the basis of these and other obser-
several cases of flexible polymers, regular confor- vations obtained by following the evolution of the
mations occur mostly in 3-D crystals; this is the vibrational spectra with temperature, models have
reason why in these cases the observation of regu- been proposed for the description at the atomic
larity bands indicates that 3-D crystalline domains scale of the mechanism that prepares the melting
are present in the sample. of polyethylene lamellar crystals.15
Fortunately, in some cases some additional The theoretical prediction of all the phonons of
spectral features can be observed, which are direct a 3-D polymeric crystal can be done following the
evidence of the occurrence of a crystalline phase. general approach reported in solid-state physics
These signals arise because of intermolecular textbooks, based on Cartesian displacement coor-
interactions between chains belonging to the same dinates.50,51 This approach is suitable for any kind
crystalline unit cell. This happens, for instance, in of crystals, ranging from molecular crystals to
the case of the orthorhombic crystal of polyethy- covalent crystals to metals. In spite of the fact
lene, which shows the phenomenon of crystal that the use of valence coordinates provides a
field splitting that is very evident for the rocking description of the potential energy through param-
and the bending bands observed in the infrared eters with a more direct physical meaning, the
spectrum. The splitting of a regularity band into calculations in three dimensions are indeed greatly
2 C to that described for the 1-D polymer crystal, but

now the phonon wave-vector has three compo-
10 C
nents (along the three fundamental axes of the
reciprocal lattice) and requires the definition of
three different phase factors (1 , 2 , 3 ) which
14 C allow one to relate the displacement in a given
cell (n1 , n2 , n3 ) to those of the reference unit cell.
16 C The phase factor j is related to the appropriate
phonon wave-vector of the 3-D lattice by the
17 C relationship j = qj aj = qj a j aj , aj and a j
being respectively the j th unit vectors in the
direct and in the reciprocal lattice of the crystal.
19.5 C
Phonon dispersion branches are usually plotted
along some selected directions (usually charac-
21 C terized by high symmetry) in the first Brillouin
TG Zone of the crystal. In particular, dispersion rela-
22 C tions plotted for q wave-vectors in the direction
c immediately give an idea of the effect of the
23.5 C crystal field on phonons propagating along the
polymer chains, i.e. on the phonons characteristic
25 C
of the 1-D crystal68,69 representing the regular
3000 3000
26 C
28 C
2000 2000
28.5 C
1500 1500
Wavenumbers (cm1)
29 C
1000 1000
500 500
780 760 740 720 700
n (cm1)
Figure 5. Thermal evolution of infrared bands belonging
to the rocking sequence of n-nonadecane while passing
0 j p 0 q.c p
from the crystalline orthorhombic phase to a pre-melting (a) (b)
phase and to the melt.15 The arrow indicates a band asso-
ciated with some conformational disorder (trans-gauche Figure 6. Phonon dispersion curves of a single poly
defects) that affects chain ends in the -phase. (ethylene) chain. (a) Dispersion curves plotted versus the
phase factor , relating displacements in adjacent struc-
tural (CH2 ) units; (b) dispersion curves plotted versus the
simplified working with Cartesian displacement phase factor = q.c, relating the atomic displacements
coordinates. Obviously the formalism is similar in adjacent translational (CH2 CH2 ) units.68, 69
chain (Figure 6). In Figure 7 it can be seen that 6 CONFORMATIONAL DEFECTS

in the case of poly(ethylene)70 calculations for IN OLIGOMERS AND
the 1-D case result in small but non-negligible
splitting of each phonon branch into two different POLYMERS
branches, as expected for a crystal where two As already mentioned, powerful structural diag-
chains form the basis of the unit cell of the 3- nostic tools have been developed starting from the
D crystal. Crystal field splittings can be predicted analysis of conformationally disordered oligomers
considering q = 0 phonons (corresponding to the and polymer chains.
phase factors = 0 and , when phonons are In particular, frequency markers of confor-
built on the basis of the minimal structural unit mational defects (short sequences of CC bonds
as in the plot of Figure 7) in agreement with the involving G+ and G torsional angles) local-
experimental findings (see above). ized on a mainly trans-planar polymethylene chain
have been clearly identified.6 12 Calculations on
model molecules10 12 showed that conforma-
3000 tional defects give rise to very peculiar normal
n6
modes which can be either localized on the
2900 n1
group of atoms that carries the conformational
2800 defect, (localized defect modes), or are dynami-
n2 cally coupled with characteristic vibrations, (i.e.,
1500
cm1
n3 phonons of an ideally infinite chain) of the trans-

1400 planar CH2 sequence (resonance modes) [see for
n7
1300 instance references 1621]. The availability of a
1200
reliable force field for polymethylene chains and
of empirical electro-optical parameters for CH2
1100
0 p and CH3 groups in different conformational envi-
ronments made it possible to give a fundamental
1200
contribution to the discussion on the existence
n4
1100 of tight folding on the surfaces of the lamellae
1000 in crystalline polyethylene samples.71 73 In this
case the existence of a tight, regular loop asso-
900
ciated with chain re-entry along the [200] crys-
800 n8 talline planes was demonstrated by theoretical
700 modeling71 73 of the cyclic molecule C32 H64 . This
molecule contains two loops formed by exactly
cm1
600
the same sequence of torsional angles (GGTGG)
500 as required by the [200] tight folding. With
400 the help of the theoretical frequency prediction
n5
following the GF Wilson formalism22 using a very
300
reliable empirical force field for n-alkanes,10 12
200 n9 characteristic CH2 wagging and rocking normal
100 modes mainly localized on the loops were iden-
tified; their infrared intensity was predicted by
0
0 f p transferring the appropriate electro-optical param-
Figure 7. Phonon dispersion curves for crystalline (3D) eters. The favorable comparison with the experi-
poly(ethylene) plotted in the direction 0-c , according to mental spectrum of the model molecule allowed
the force field from reference 70. unambiguous identification of the markers for a
(a)
Exp
Calc
GG Exp
Calc
GTG
GTG
GGTGG
Frequency Frequency
(b) cm1 1376 1352 1328 1304 cm1 738 714 690
Figure 8. Theoretical modeling and experimental spectra of the cyclic molecule C32 H64 : (a) sketch of the molecular
structure; (b) region 13801304 cm1 ; (c) region 750690 cm1 .71 73
tight fold amongst other experimental spectral sample after melting and re-crystallization at
features (Figure 8). After that, these same markers ambient pressure. After this treatment, one expects
were located in the infrared spectrum of a to find the usual morphology of a semi-crystalline
poly(ethylene) sample (Figure 9), which consisted polymer, characterized by thin lamellae. In partic-
of single crystals (sectored lamellae) obtained by ular, one expects a marked increase of folded
very careful crystallization from dilute solution. (conformationally defective) chains arising from
These findings give some evidence for the exis- the increased relative amount of lamellar surfa-
tence on the surface of a non-negligible amount ces and from the amorphous phase localised in
of chains that show regular tight re-entries.71 73 the interlamellar regions. From the point of view
Another impressive application of infrared of spectroscopy this would result in a marked
spectroscopy to the detection of conformational increase of the characteristic bands associated
defects is that reported in reference 74, where the with the presence of gauche defects, as indeed has
infrared spectrum of a poly(ethylene) sample crys- been observed. Notice however that the diagnoses
tallized under high pressure, (namely a sample described in the two examples illustrated above
characterized by extended chain crystalline would be impossible without a preliminary iden-
domains,75 with very thick lamellae) was com- tification (through models) of the precise spectral
pared with the spectrum obtained for the same regions where specific defect markers appear.
To conclude this discussion on the spectro-

scopic detection of conformational defects we
have to briefly return to the problem of the spec-
troscopic recognition and characterization of the
amorphous phases. Since the chemical groups and
a
the intermolecular bonds are the same both in
amorphous polymers and in their crystalline (1-
D or 3-D) forms, in many practical cases it is
b
not simple to clearly distinguish the spectroscopic
features associated with the two phases. Indeed,
signals arising from the crystals and from the
amorphous often occur at the same frequencies.
The recognition of regular conformations is
1400 1375 1350 1325 1300 1275 1250 1225 1200 difficult, especially when the monomer contains
Wavenumbers (cm1) many atoms: in this case the effect of the
Figure 9. Comparison between the infrared spectrum of coupling between the different chemical units in
the [200] surface of poly(ethylene) lamellae (spectrum a), the polymer, which is responsible for the appear-
and the infrared spectrum of the cyclic molecule C32 H64 ance of well dispersed phonon branches, has a
(spectrum b) in the region 14001200 cm1 71 73 . relatively modest importance with respect to the
effects of the interactions between atoms that
belong to the same chemical unit. Polymers with
An even finer structural diagnosis can be
flat phonon dispersion branches generally do not
obtained if one takes advantage of the peculiar
allow an easy and safe identification of regularity
local defect modes, which are obtained by selec-
bands, and for the same reason a diagnosis of
tive deuteration of CH2 units at well defined
chain regularity using oligomers is not assisted
positions along a polymethylene chain. Mass
by the detection of sequences of bands.
defects are responsible for the occurrence of As a practical rule, one can always assume
vibrations at energies in completely free regions that the presence of a well defined environment
of the density of phonon states characteristic of (guaranteed by a regular conformation of the
the non-deuterated trans-planar chain. For this chain and by a regular packing of chains in
reason these modes are usually referred to as the crystalline phase) usually gives well defined
gap-modes. Several CD2 gap-modes are very and sharp bands in the spectra. The intra- and
sensitive to the conformational environment, (i.e., inter-chain disorder, which is characteristic of the
they change their frequency), as proven by calcu- amorphous phase, is often responsible for band
lations on selectively deuterated n-alkanes. An broadening, the recognition of which can be of
investigation based on this rationale studied the help in the diagnosis.
structural changes that happen in the -phase of As a proof of the above statement, we can cite
n-nonadecane.15 The synthesis of two different several cases where outdoor aging of polymer
samples consisting of C19 H40 chains labelled by samples was monitored by means of spectroscopy
a CD2 group respectively in position 2 and 10 in conjunction with other characterization techni-
proved the presence of end-TG defects during ques.76,77 One of the effects of polymer degra-
the pre-melting phase, thus indicating that the dation by aging is the relative increase of the
disorder mainly starts from lamellar surfaces. percent crystallinity, which can be ascribed to
Only at the melting temperature were signals two different phenomena: i) the easier attack of
detected which were related to gauche conforma- the amorphous regions by oxygen, due to the
tions involving the bulk of the chains, (and then higher free volume in the disordered phases; ii)
also CD2 units in position 10). re-crystallization phenomena, accompanying the
Absorbance
1900 1800 1700 1600 1500 1400 1300 1200 1100 1000
Wavenumbers (cm1)
Figure 10. Infrared absorption spectra obtained from the surfaces of a thick film of ethylene-carbon monoxide
copolymer: (a) pristine un-aged sample; (b) sample withdrawn after 7944 h of outdoor exposure (lower surface),
(c) sample withdrawn after 7944 h of outdoor exposure (upper surface)76 .
natural thermal cycles suffered by the sample amorphous phase, in the spectra of semicrystalline
during prolonged outdoor exposure. Spectroscopy solid samples.
was used in these cases to follow the structural Another indicator, which is of great help for the
changes: it is particularly noteworthy that changes characterization of partially or fully amorphous
of the shape of the main bands were observed. As samples, is the knowledge of specific markers,
an example of this effect we report in Figure 10 which indicate the presence of defects that cannot
the evolution of the infrared carbonyl stretching exist in the crystal. These markers can be iden-
band in a sample of the alternating copolymer tified by simulation of the spectra of different
between ethylene and carbon monoxide (chemical conformational isomers, modelled by oligomers
unit: CH2 CH2 CO). This band corresponds to a containing a few chemical units characteristic of
localised group frequency of the CO double bonds the polymer. This can now be also done with
and appears as a relatively broad band in the pris- the help of first principles (ab initio or DFT)
tine, semicrystalline sample. As the time of expo- calculations.78
sure of the sample increases we can observe that
the peak becomes sharper and sharper, because
of the decrease of the contribution of the amor- 7 FIRST PRINCIPLES AND
phous phase, whose decrease is also confirmed by
thermal analysis of the samples at different aging
SEMI-EMPIRICAL
times. A further nice proof of the correctness of CALCULATIONS OF POLYMER
these conclusions is the direct comparison with PHONONS
the infrared spectrum of the polymer in the melt
phase, showing a broad C=O band with a band As already discussed in Section 2, while an
shape similar to the component assigned to the accurate simulation of molecular spectra can at
present be obtained by high-level first principles calculations with 6-31G* and 3.21G basis sets) for
calculations, their application to polymers is still polyene chains with increasing length from buta-
limited to a few and particular cases. On one hand, diene to tetra-deca-heptaene.30,31 A careful anal-
simulation of oligomers considered as models for ysis of the evolution of individual CC stretching
polymers can be performed as a practical tool force constants, (both diagonal and interactions
that can support band assignment and structural at any distance along the chain), showed a trend
diagnosis for polymers. One example is reported from which values for a chain with infinite
in reference 79, where the effect on the Raman length have been extrapolated. The force field so
spectrum of the presence of a peroxide group in a obtained allows calculation of the phonon disper-
chain of perfluoroether units is investigated. The sion curves for an ideally infinite poly(acetylene)
study is based on quantum chemical DFT calcu- (1-D crystal), which compares favourably with
lations, carried out on short chains. The results the data obtained from neutron scattering exper-
obtained were used to validate the experimental iments. As observed by the authors,30,31 in this
procedure developed to obtain an estimate of the case the Raman frequencies of oligomeric poly-
amount of peroxide groups in the polymer. The enes do not lie on the phonon dispersion branches
diagnosis was made by comparing the intensity obtained for poly(acetylene). This fact is justi-
of the strong Raman band involving a normal fied30,31 as being due to large chain-length
mode localised on the OO bond, (as confirmed dependence of the skeletal stretching force cons-
by theoretical eigenvectors), and bands mainly tants, that is, force constants obtained for short
ascribed to COC symmetric stretching of the
to medium chains cannot reliably be transferred
ether groups.
to longer chains. Notice that the consequence
In the case of -bonded polymers, even in the
of this chain-length dependence results, in the
early days of polymer dynamics good theoretical
frame of Effective Conjugation Coordinate (ECC)
predictions of phonons were reached by trans-
theory,67 in the softening of the collective force
ferring the valence force field refined by the
constant FR - (see Vibrational Spectroscopy of
experimental frequency data of oligomers.58 60
Conducting Polymers: Theory and Perspective
This also implies that first principles intramolec-
ular potentials could be successfully transferred in this book).
from the oligomers to the polymer, in order A very recent application of density func-
to obtain the phonon frequencies for the infi- tional theory to the direct prediction (i.e., without
nite regular chain, following the procedure illus- extrapolation procedures) of polymer phonons
trated in Section 3. The problem becomes much has been proposed for the case of polyynes.80
more intriguing in cases characterized by long The infinite polyyne is a chain (1-D crystal)
range dynamic interactions, as for instance when consisting of a linear sequence of CC bonds.
dealing with chains containing conjugated There are two possible structures of this polymer:
bonds. This is certainly the reason why most i) a metallic chain, where all CC bonds are
of the calculations that appear in the literature identical; this structure can be defined in the
dealing with quantum chemical approaches to chemical language as cumulenic and its bonds
polymer phonons are applied to cases of polycon- are seen as a sequence of perfect conjugated
jugated systems. Calculations on model molecules double bonds; ii) a second structure can be
with increasing chain length allow one to obtain obtained if Peierls distortion81 of the metallic
an insight into the conjugation length dependence chain is allowed: in this case two energetically
of the relevant parameters. Following this line, the indistinguishable equilibrium structures (dimers)
best valence force field (amongst those now avail- are obtained, formed by alternation of longer
able) for poly(acetylene) has been obtained on the (quasi single) and shorter (quasi triple) CC bonds.
basis of high level quantum chemical predictions The two structures correspond to two oppo-
(ab initio second order Moller-Plesset perturbation site phases of the bond alternation, that is,
they correspond to triple-single and to single- dimerization, R, defined as the difference

triple sequences, respectively. These dimerized between the equilibrium bond length of a quasi
structures show the opening of a band gap single and a quasi triple adjacent CC bonds. This
between occupied () electronic states and unoc- is what we have carried out in the frame of the
cupied ( ) states: they are indeed semiconduc- density functional perturbation theory (DFPT)86
tors. Finite size carbon chains, with acetylenic based on pseudo-potential theory, as implemented
CH bonds as end groups, can be considered as by PWscf code.87 Phonons have been calculated
the natural oligomers of this polymer; these are for the infinite carbon chain at several fixed values
the chemical species for which it is important of R, ranging from the equalized, metallic
to obtain a reliable knowledge and interpreta- chain with R = 0, to a highly alternated Peierls
tion of the spectroscopic features. Indeed, the chain ( R = 0.142A).
spectroscopic detection of polyynes of different The evolution with R of the phonon disper-
length is relevant to the study of nanostructured sion curves is presented in Figure 11. What is
carbon materials, since these chemical species clearly shown in this plot is that as the metallic
are expected in the initial stage of formation of limit is approached, the optical branch near q = 0
fullerenes and nanotubes.82,83 Moreover, bands becomes steeper and steeper. This effect is a
ascribed to the presence of linear carbon chains direct consequence of the increase of the elec-
have been detected in the Raman spectra of tron conjugation along the chain, which causes
films of amorphous carbon obtained by newly a systematic increase in the values of the long
developed deposition techniques from supersonic range CC stretching interaction force constants.
beams of carbon clusters.84,85 In Figure 11 values of the frequencies computed
New insight into the physics of finite length for several polyyne oligomers (treated at the same
polyynes and a rationalization of their Raman level of theory) are placed on the theoretical
spectra can be obtained through the calculation phonon branches of the polymer, at the proper
of the phonon dispersion curves of an infi- values of the phonon phase j , chosen according
nite polyyne chain at different degrees of bond to equation 20.
2500 2400
2200
2000
Phonon wavenumber (cm1)
Phonon wavenumber (cm1)
2000
1500
1800
Cumulene
1600 opt. geom. (0.038)
1000 0.062
0.069
1400 0.080
Cumulene 0.099
opt. geom. (0.038) 0.142
500 0.062 40 CC units
0.069 1200 30 CC units
0.080 20 CC units
0.099 10 CC units
0.142 8 CC units
0 1000
0 0.2 0.4 0.6 0.8 1 0 0.05 0.1 0.15 0.2
Phonon wave vector (units of p/c) Phonon wave vector (units of p/c)
Figure 11. (a) Longitudinal optical phonon branches of carbon linear chains for different R values (from R = 0
A to R = 0.142 A). (b) Comparison with vibrational frequencies of finite chains characterized by different lengths
(from 8 to 40 CC units).80
The results obtained indicate that: determination of suitable empirical intramolec-

ular potential in terms of harmonic valence force
For any real conjugation length (determined constants.
by the real size of the polyyne chain), a Phonon dispersions and the associated density
different force field is appropriate. It can be of vibrational states can be validated by compar-
put in correspondence with an ideally infi- ison with data from neutron scattering experi-
nite chain with a given degree of conjuga- ments, which probe phonons corresponding to
tion, associated with a given value of the forbidden (q = 0) optical transitions. Another
R parameter. source of information about q = 0 phonons is the
The effect of the electron-phonon coupling, spectroscopic observation of second order tran-
which is responsible for the softening of the sitions involving more than one phonon. These
longitudinal CC stretching mode, is modu- transitions are allowed in terms of momentum
lated by a relevant structural parameter, conservation provided that the sum of the wave-
namely by the degree of bond alternation of vectors of the phonons involved vanishes. Second
order transitions are expected, (and are indeed
the chain, R. The value of this parameter, in
responsible for relatively weak features in infrared
the case of short chains, is determined by the
and Raman spectra), whenever the hypothesis of
confinement of electrons due to the finite
double harmonicity is relaxed.89 On the other
length of the molecule.
hand, the calculation of phonon branches allows
The Kohn anomaly for the optical branch of
one to safely state the number and the frequen-
the infinite polyyne chain can be predicted at
cies of regularity bands, corresponding to the
the transition between a semiconducting and a
fundamental q = 0 transitions, for which spectro-
metallic chain.88
scopic activity is determined by symmetry selec-
tion rules. The recognition of the regularity
bands in the experimental vibrational spectra is
8 CONCLUSIONS a powerful tool for a diagnosis of the regu-
larity of the polymer conformation in real mate-
In the above paragraphs we have illustrated the rials. Sometimes the occurrence of crystal split-
theoretical approaches available for the interpreting, which is predicted by theory in the case
tation and, in some cases, for the prediction of the of crystal packing with at least two polymer
spectroscopic response of polymeric materials. It chains in the unit 3-D cell, gives experimental
should be clear that the regular structure taken by evidence of the presence of 3-D crystalline
a polymer chain in the crystalline phase allows a domains.
simple and rigorous mathematical treatment of its It is worthwhile noting that rapid progress in
vibrations, based on the translational symmetry of terms of computational resources, and the recent
the crystal. Moreover, the theory of phonons in a development of suitable computer codes, make
1-D crystal, (which indeed describes the isolated it presently possible to carry out first princi-
regular polymer chain) can be improved, taking ples calculations for the prediction of the vibra-
into account also interchain interactions between tions of polymers. This can be firstly done by
adjacent chains, packed in a 3-D crystal. On the extrapolating the results obtained for molecular
basis of these models, phonon dispersion rela- species (oligomers); moreover, in the case of crys-
tions (usually based on empirical force fields) for talline polymeric materials, calculations could be
several polymers have been successfully calcu- directly carried out for the polymer by means of
lated. In this frame, the knowledge of normal Quantum codes able to account for the transla-
vibrations and frequencies of several oligomers, tional symmetry.87,90
which can be put in correspondence with phonons First principles predictions (e.g., with Hartree-
of the parent polymer, can be of great help for the Fock and density functional theory (DFT)
mehods) of vibrational spectra of molecules compute phonon dispersions and in some cases
(oligomers) can be performed at different levels also probabilities of vibrational transitions (see for
of accuracy according to theory chosen. In instance Ref. 87, 97, 98). However, in spite of the
the framework of LCAO (linear combination fact that the use of these approaches for the study
of atomic orbitals) calculations, the use of the of crystalline polymers is appealing and rela-
631G** or of a more extended basis set is tively straightforward, so far applications of these
recommended if a quantitative agreement with the approaches for calculations of infrared or Raman
experimental data is required. This is different spectra (frequencies and intensities) of polymers
from HF (HartreeFock) calculations, where are not reported in the scientific literature.
quantitative predictions of vibrational frequencies As already mentioned throughout this chapter,
can be reached only after some scaling of successful first principles predictions of the vibra-
the theoretical results28 ; in the case of DFT tional spectra of polymers are still restricted
calculations a suitable choice of the functional to the use of the theoretical parameters (force
allows one to reach an excellent numerical constant and electro-optical parameters) previ-
agreement between predicted and experimental ously obtained from quantum chemical calcula-
frequencies. BLYP and B3LYP functional91 93 tions on suitable series of oligomers.30,31,46
are widely and successfully employed for Some more words can be said about simula-
simulations of vibrational spectra; recently it tions applied to the case of disordered polymeric
has been shown80 that GGA (generalized materials, (e.g., in the case of flexible polymer
gradient approximation) functionals such as PBE chains in the amorphous phase). The theoretical
(PerdewBurkeErnzerhof)94 provide even better study of conformational defects in model struc-
results for the case of conjugated organic organic tures (studied by empirical treatments or with first
molecules. principles approach) allows assignment of spec-
Several software programs (both commercial troscopic defect markers, which are very useful
and freely available) provide the tools for for structural diagnosis. Regarding this point, it
routine calculations of molecular spectra; one is important to recall that especially in the study
of the most widely used commercial programs of defects the search for reliable spectroscopic
is GAUSSIAN95 ; among the freely available markers requires also the modeling of the inten-
ones we can cite GAMESS.96 A completely sity of the relevant infrared and Raman transi-
different situation is that of first principles spectra tions. Localized defects modes which are practi-
prediction for crystalline systems (treated taking cally silent in the spectra are useless for practical
into account the translational periodicity). Several purposes; for this reason the use of transferable
quantum chemical commercial programs (for electro-optical parameters for the prediction of
instance CRYSTAL90 and the last edition of the intensity of defects bands in poly(methylene)
GAUSSIAN95 ) allow one to treat translational chains turned out to be of great help.
symmetry. However the codes needed for the Several important questions, which may be very
solution of the vibrational problem are not relevant to the field of polymer spectroscopy,
yet included in GAUSSIAN95 ; CRYSTAL90 have been deliberately omitted in this chapter. In
provides tools for the calculations of the particular, no mention is made of the problem of
vibrational frequencies, however algorithms for the theoretical modeling of intermolecular inter-
the calculation of the strength of vibrational actions, as, for instance, specific interactions in
transitions for the prediction of Raman intensities polymer blends, interactions between a polymer
have not yet been included. film and a substrate, or interactions with the
On the other hand, when changing the basis surrounding solvent in polymer solutions. In spite
set from atomic orbitals to plane waves, several of the lack of systematic theoretical studies,
programs developed by the scientific community experimental correlations derived from series of
of solid-state physicists are available and able to spectroscopic measurements on samples under
different experimental conditions provided reli- 2. G. Natta, M. Peraldo and P. Corradini, Atti Accad.
able tools for successful diagnostic studies.99,100 Naz. Lincei, Rend. Classe Sci. Fis. Mat. Nat., 26,
On the other hand, the study of complex systems 14 (1959).
consisting of several molecules interacting with 3. G. Zerbi, M. Gussoni and F. Ciampelli,
an environment characterized by well defined Spectrochim. Acta, 23, 301 (1967).
physical-chemical properties represents at present 4. S. Frisk, R.M. Ikeda, D. Bruce Chase, A.
the frontier of first principles calculations.101 It Kennedy and J. Rabolt, Macromolecules, 37,
6027 (2004).
seems clear also that polymer science and spec-
troscopy will benefit from future developments 5. C.E. Bottani, C. Castiglioni and G. Zerbi, Raman
Scattering in Nanostructures, in Encyclopedia
along this line of research.
of Nanoscience and Nanotechnology, ed H.S.
Nalwa, American Scientific Publishers, Stevenson
Ranch (2004).
ACKNOWLEDGMENTS 6. G. Zerbi, L. Piseri and F. Cabassi, Mol. Phys., 22,
241 (1971).
I am grateful to Professor Giuseppe Zerbi who,
during the wrinting of this chapter, put at my 7. G. Zerbi, M. Sacchi, Macromolecules, 6, 692
(1973).
disposal his experience and the results of his
research in the field of polymer spectroscopy. 8. A. Rubcic, G. Zerbi, Macromolecules, 7, 754
(1974).
9. A. Rubcic, G. Zerbi, Macromolecules, 7, 759
END NOTES (1974).
10. R.G. Snyder, J. Chem. Phys., 47, 1316 (1967).
a.
This happens in the presence of screw axis 11. R.G. Snyder, M.W. Poore, Macromolecules, 6,
different from s(21 ). In these cases the two rigid 708 (1973).
translations in the two directions orthogonal to the 12. J.L. Koenig, Appl. Spect. Rev., 4, 233 (1971).
chain axis are degenerate. 13. G. Zerbi, G. Minoni and M.P. Tulloch, J. Chem.
Phys., 78, 5853 (1983).
14. R. Piazza and G. Zerbi, Polymer, 23, 1921 (1982).
ABBREVIATIONS AND
15. G. Zerbi, R. Magni, M. Gussoni, K. Holland-
ACRONYMS Moritz, A.B. Bigotto and S. Dirlikov, J. Chem.
Phys., 75, 3175 (1981).
DFPT Density Functional Perturbation Theory
16. S. Krimm, Fortschr. Hochpol. Forsch., 2, 51
DFT Density Functional Theory (1960).
ECC Effective Conjugation Coordinate
17. G. Zerbi, Molecular Vibrations of High
G Gauche Polymers, in Applied Spectroscopy Reviews,
GGA Generalized Gradient Approximation ed A.D. Brame, Marcel Dekker, New York, 193,
HF HartreeFock Vol. 2 (1969).
LAM Longitudinal Accordion Mode 18. G. Zerbi, Molecular Dynamics and Vibra-
LCAO Linear Combination of Atomic Orbitals tional Spectra of Polymers, in Vibrational
NET Negative Eigenvalue Theory Spectroscopy-Modern Trends, eds A.J. Barnes
PBE PerdewBurkeErnzerhof and W.J. Orville-Thomas, Elsevier, New York,
T Trans 378, Chapter 24 (1977).
19. G. Zerbi, Vibrational Spectroscopy of Very Large
Molecules, in Advances in Infrared and Raman
REFERENCES spectroscopy, eds R.J.H. Clark and R.E. Hester,
Wiley, Heyden, 301, Vol. 11 (1984).
1. P. Flory, Statistical Mechanics of Chain 20. G. Zerbi ed, Modern Polymer Spectroscopy,
Molecules, Interscience (1969). Wiley-VCH, Weinheim (1999).
21. G. Zerbi, C. Castiglioni and M. Del Zoppo, 38. S. Radice, N. Del Fanti, C. Castiglioni, M. Del
Structure and Optical Properties of Polyconju- Zoppo and G. Zerbi, Macromolecules, 27, 2194
gated Molecules from their Vibrational Spectra: (1994).
Techniques and Data. A Guided Tour, in Elec- 39. S. Radice, M. Pianca, E. Barchiesi and J.
tronic Materials. The Oligomer Approach, eds Abusleme, J. Fluorine Chem., 58, 333 (1992).
G. Wegner and K. Mullen, VCH, Weinheim,
345402, Chapter 6 (1999). 40. M. Rui, M.N. Ramos, C. Castiglioni, M. Gussoni
and G. Zerbi, Mol. Cryst. Liq. Cryst., 197, 445
22. E.B. Wilson, Jr, J.C. Decius and P.C. Cross, (1990).
Molecular Vibrations: the Theory of Infrared and
Raman Vibrational Spectra, Mc Graw-Hill, New 41. C. Castiglioni, M. Gussoni, J.T. Lopez Navarrete
York (1955). and G. Zerbi, Solid State Commun., 65, 625
(1987).
23. S. Califano, Vibrational States, John Wiley &
Sons, New York (1976). 42. M.W. Wolkenstein, Comp. Rend. Acad. Sci.
URSS, 30, 791 (1941).
24. L.A. Woodward, Introduction to the Theory
43. D.A. Long, Proc. R. Soc. London, A217, 203
of Molecular Vibrations and Vibrational Spec-
(1953).
troscopy, Clarendon Press, Oxford (1972).
44. L.S. Mayants and B.S. Averbukh, Theory and
25. P. Dean, Rev. Mod. Phys., 44, 1881 (1972).
Calculation of Intensity in Vibrational Spectra of
26. A.A. Maradudin, E.W. Montroll and G.H. Weiss, Molecules, Nakua, Moscow (1971).
Solid State Phys., 3, (Suppl.), 1 (1963).
45. K.S. Smirnov, D. Bougeard, J. Raman Spectrosc.,
27. M. Tasumi and G. Zerbi, J. Chem. Phys., 48, 3813 37, 100 (2006).
(1968). 46. K.S. Smirnov, D. Bougeard and P. Tandem, J.
28. G. Fogarasi and P. Pulay, Annu. Rev. Phys. Chem., Phys. Chem. A, 110, 4516 (2006).
35, 191 (1984). 47. L. Wirtz, M. Lazzeri, F. Mauri and A. Rubio,
29. R.G. Parr and W. Yang, Density Functional Phys. Rev., B, 71, 241402 (2005).
Theory of Atoms and Molecules, Oxford 48. C. Castiglioni, M. Tommasini and G. Zerbi,
University Press, New York (1989). Philos. Trans. R. Soc. London, Ser. A, 362, 1
30. S. Hirata, H. Yoshida, H. Torii and M. Tasumi, J. (2004).
Chem. Phys., 103, 8955 (1995). 49. A. Bianco, M. Del Zoppo and G. Zerbi, J. Chem.
31. S. Hirata, H. Torii and M. Tasumi, J. Chem. Phys., Phys., 120, 1450 (2004).
103, 8964 (1995). 50. N.W. Ashcroft and N.D. Mermin, Solid State
32. L. Piseri, R. Tubino, L. Paltrinieri and G. Physics, International Editions, Philadelphia
Dellepiane, Solid State Commun., 46, 183 (1988). (1976).
33. E. Di Donato, M. Tommasini, C. Castiglioni and 51. C. Kittel, Introduction to Solid State Physics,
G. Zerbi, Phys. Rev. B, 74, 184306 (2006). 8th edition, John Wiley & Sons, Hoboken (2004).
34. M. Gussoni, C. Castiglioni and G. Zerbi, 52. R. Zbinden, Infrared Spectroscopy of High
Vibrational Intensities: Interpretation and use Polymers, Academic Press, New York (1964).
for Diagnostic Purposes, in Handbook of 53. J.C. Decius and R.M. Hexter, Molecular
Vibrational Spectroscopy, eds J.M. Chalmers Vibration in Crystals, McGraw-Hill, New York
and P.R. Griffiths, John Wiley & Sons, Chichester (1977).
(2002).
54. P.C. Painter, M.M. Coleman and J. Koenig,
35. L.A. Gribov, Intensity Theory for Infrared The Theory of Vibrational Spectroscopy and its
Spectra of Polyatomic Molecules, Consultants Application to Polymeric Material, John Wiley
Bureau, New York (1964). & Sons, New York (1982).
36. A.J. van Straten and W.M.A. Smit, J. Mol. 55. G. Zerbi and M. Sacchi, Macromolecules, 6, 692
Spectrosc., 65, 202 (1977). (1973).
37. M. Gussoni, C. Castiglioni and G. Zerbi, Phys. 56. G. Masetti, F. Cabassi, G. Morelli and G. Zerbi,
Chem., 88, 600 (1984). Macromolecules, 6, 700 (1973).
57. M. Born and T. von Karman, Phys. Z., 13, 297 77. R. Gallo, L. Brambilla, C. Castiglioni and
(1912). F. Severini, J . Macromol. Sci., Part A, 43, 535
58. R.G. Snyder and J.H. Shachtsneider, Spectrochim. (2006).
Acta, 19, 85 (1963). 78. M.C. Rumi and G. Zerbi, Chem. Phys., 242, 123
(1999).
59. J.H. Shachtsneider and R.G. Snyder, Spectrochim.
Acta, 19, 117 (1963). 79. S. Radice, P. Toniolo, E. Barchiesi, P.A. Guarda,
M. Tommasini and C. Castiglioni, J. Fluorine
60. R.G. Snyder and J.H. Shachtsneider, Spectrochim. Chem., 125, 151 (2004).
Acta, 21, 169 (1965).
80. A. Milani, M. Tommasini, M. Del Zoppo, C.
61. S.I. Mizushima and T. Shimanouchi, J. Am. Castiglioni, G. Zerbi, Phys. Rev. B, 74 153418
Chem. Soc., 71, 1320 (1949). (2006).
62. R.J. Shaufele and T. Shimanouchi, J. Chem. 81. R.E. Peierls, Quantum Theory of Solids,
Phys., 47, 3605 (1967). Clarendon Press, Oxford (1955).
63. T. Shimanouchi and M. Tasumi, Indian J. Pure 82. R.J. Lagow, J.J. Kampa, H. Wei, S.L. Battle, J.W.
Appl. Phys., 9, 958 (1971). Genge, D.A. Laude, C.J. Harper, R. Bau, R.C.
64. R.G. Snyder, J. Chem. Phys., 76, 3921 (1982). Sevens, J.F. Haw and E. Munson, Science, 267,
362 (1995).
65. R.G. Snyder and J.R. Scherer, J. Polym. Sci., 16,
83. A.A. Shvatsburg, R. Hudgins, P. Dugourd, R.
1593 (1978).
Gutierrez, T. Frauenheim and M.F. Jarrold, Phys.
66. H.E. Schaffer, R.R. Chance, R.J. Silbey, K. Knoll Rev. Lett., 84, 2421 (2000).
and R.R. Schrock, J. Chem. Phys., 94, 4161 84. L. Ravagnan, F. Siviero, C. Lenardi, P. Miseri,
(1991). E. Barborini, P. Milani, C.S. Casari, A. Li Bassi
67. M. Gussoni, C. Castiglioni, G. Zerbi, Vibra- and C.E. Bottani, Phys. Rev. Lett., 89, 285506
tional Spectroscopy of Polyconjugated Materials: (2002).
Polyacetylene and Polyenes, in Spectroscopy 85. C.S. Casari, A. Li Bassi, L. Ravagnan, F. Siviero,
of Advanced Materials, eds R.J.H. Clark and C. Lenardi, P. Miseri, G. Buongiorno, C.E.
R.E. Hester, John Wiley & Sons, 251, Chapter Bottani and P. Milani, Phys. Rev., B, 69, 075422
5 (1991). (2004).
68. L. Piseri and G. Zerbi, J. Mol. Spectrosc., 26, 254 86. S. Baroni, S. de Gironzoli, A. Dal Corso and
(1968). P. Giannozzi, Rev. Mod. Phys., 73, 515 (2001).
69. L. Piseri and G. Zerbi, J. Chem. Phys., 48, 3561 87. S. Baroni, A. Dal Corso, S. De Gironzoli,
(1968). P. Giannozzi, C. Cavazzoni, G. Ballabio, S. Scan-
70. M. Tasumi and S. Krimm, J. Chem. Phys., 46, (2), dolo, G. Chiaraotti, P. Focher, A. Pasquarello,
755 (1967). K. Laasonen, A. Trave, R. Car, N. Marzari,
A. Kokalj, 2006, http://www.pwscf.org/.
71. H.P. Grossmann and H. Boelstler, Polym. Bull.,
88. W. Kohn, Phys. Rev. Lett., 2, 393 (1959).
5, 175 (1981).
89. M.C. Rumi and G. Zerbi, Chem. Phys. Lett., 242,
72. G. Zerbi and M. Gussoni, Polymer, 21, 1129 639 (1995).
(1980).
90. V.R. Saunders, R. Dovesi, C. Roetti, R. Orlando,
73. S.J. Spells, S.L. Organ, A. Keller and G. Zerbi, C.M. Zichovich-Wilson, N.M. Harrison, K. Doll,
Polymer, 28, 697 (1987). B. Civalleri, I.J. Bush, Ph. DArco, M. Llunell,
74. E. Agosti, G. Zerbi and I.M. Ward, Polymer, 33, 2003, CRYSTAL2003 1.0, http://www.crystal.
4219 (1992). unito.it/.
75. B. Wunderlich, Macromolecular Physics, Aca- 91. A.D. Becke, Phys. Rev., A, 38, 3098 (1988).
demic Press, New York, Vol. 1 and 2 (1973). 92. C. Lee, W. Yang and R.G. Parr, Phys. Rev., B 37,
76. F. Severini, R. Gallo, L. Di Ladro, M. Pegoraro, 785 (1988).
L. Brambilla, M. Tommasini, C. Castiglioni and 93. A.D. Becke, J. Chem. Phys., 98, 56485652
G. Zerbi, Polymer, 42, 3609 (2001). (1993).
94. J.P. Perdew, K. Burke and M. Ernzerhof, Phys. M. Torrent, G. Zerah, M. Mikami, Ph. Ghosez,
Rev. Lett., 77, 3865 (1996). M. Veithen, J.-Y. Raty, V. Olevano, F. Bruneval,
95. M.J. Frisch, et al., Gaussian 03, Revision L. Reining, R. Godby, G. Onida, D.R. Hamann,
C.02, Gaussian, Inc., Wallingford (2004), and D.C. Allan. Z. Kristallogr. 220, 558562
http://www.gaussian.com/citation g03.htm. (2005).
96. M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. 99. Y. Hu, P.C. Painter and M.M. Coleman,
Elbert, M.S. Gordon, J.H. Jensen, S. Koseki, Macromol. Chem. Phys., 201, 470477 (2000).
N. Matsunaga, K.A. Nguyen, S. Su, T.L. 100. D.B. Chase and J.F. Rabolt, Macromolecules, 37,
Windus, M. Dupuis and J.A. Montgomery, 877 (2004).
J. Comput. Chem., 14, 13471363 (1993), 101. R. Cammi, B. Mennucci and J. Tomasi,
http://www.msg.ameslab.gov/GAMESS/ Computational Modelling of Solvent Effects
GAMESS.html. on Molecular Properties: An Overview of the
97. X. Gonze, J.-M. Beuken, R. Caracas, F. Detraux, Polarizable Continuum Model (PCM) Approach,
M. Fuchs, G.-M. Rignanese, L. Sindic, in Reviews in Computational Chemistry:
M. Verstraete, G. Zerah, F. Jollet, M. Torrent, Reviews of Current Trends, ed F. Leszczynski,
A. Roy, M. Mikami, Ph. Ghosez, J.-Y. Raty and World Scientific Publishing Company, Singapore,
D.C. Allan. Comput. Mater. Sci., 25, 478492 Vol. 8 (2003).
(2002).
98. X. Gonze, G.-M. Rignanese, M. Verstraete,
J.-M. Beuken, Y. Pouillon, R. Caracas, F. Jollet,
Optical Constants
John E. Bertie
University of Alberta, Edmonton, Canada
1 INTRODUCTION The optical constants are properties of the bulk phase,

not properties of individual molecules in the phase. For
The optical constants are the real, n, and imaginary, k, parts anisotropic materials their values differ for different direc-
of the complex refractive index tions, which complicates their determination and discus-
sion. Anisotropic materials will not be discussed in this
O nQ D nnQ C iknQ
n 11 3 chapter.
p The behavior of n and k through an absorption band
where i D 1 [1]. The use of complex quantities results is illustrated in Figure 1 for the CO stretching band of
from the description of the electric field of electromag- liquid methanol at 25 C.9 The k spectrum shows an absorp-
netic radiation as E D E0 exp[ik .r !t]. The imaginary tion band. The real refractive index decreases gently as
refractive index, k, is also called the absorption index [2]3 the absorption is approached from high wavenumber then
and describes absorption of radiation. In regions of no decreases sharply immediately to high wavenumber of the
absorption k D 0 and the real refractive index, n, is the k peak. In this case, and for other strong absorptions, n
ratio of the velocity of light in vacuum to the velocity of becomes less than 1 immediately above the peak. As the
light in the material. wavenumber decreases through the peak n rises sharply,
The optical constants fully describe the optical behav- but then decreases again, initially sharply but increasingly
ior of materials. They are different for different materials, gently, as the wavenumber decreases below the peak. Pro-
different temperatures and pressures, different phases of vided no other absorptions interfere, n is higher below the
the same material and different wavenumbers, nQ [3]. They peak than at the same distance above the peak.
are, thus, important fundamental properties of matter. The The change of n with wavenumber is called the dis-
result of any optical measurement can be calculated if the persion of the refractive index and the behavior shown
optical constants of the sample are known at all wavenum- in Figure 1 is called anomalous dispersion. Note that the
bers of interest and the experimental configuration can be weak absorption near 1115 cm1 in Figure 1 causes its
described mathematically. Conversely, the optical constants own weak anomalous dispersion. The other type of dis-
of a sample can be calculated from any well-defined opti- persion is normal dispersion, in which n decreases slowly
cal measurements. Two measurements are required at each with decreasing wavenumber. This occurs in regions of
wavenumber since there are two optical constants, n and k, no absorption that are far removed from major absorp-
to be determined. Under some circumstances one of these tion. Colorless compounds show normal dispersion in the
measurements can be replaced by the KramersKronig visible region and the available data can be fitted to equa-
relations, which, as described later, allow the n and k spec- tion (2).5,10 The first three terms in equation (2) are due to
tra to be interconverted provided the measurements cover absorption in the ultraviolet (UV) and X-ray regions, while
a sufficiently wide wavenumber range and n is known at the last term is due to infrared (IR) absorption and can
one wavenumber close to that range. usually be neglected except for hydrogen-bonded liquids.10
a2
John Wiley & Sons Ltd, 2002. nnQ D ao C a2 nQ 2 C a4 nQ 4 2
nQ 2
n k
0.6 (a)
2
1.6
0.3 C6H6
n
1.2
H2O
1
k
0.0
1200 1000 (b)
(cm1)
0.8
Figure 1. Anomalous dispersion: the real, n, and imaginary, k,
refractive indices of liquid CH3 OH9 at the CO stretching band.
0.4
For gases, n is very close to 1 because the molecules are H2O
far apart. For standard air at 1 atm and 15 C, n decreases
slightly from 1.000272984 to 1.000272599 between 0.0
5000 cm1 and 10 cm1 ,11 as calculated from Edlens C6H6
formula.12 Even for dipolar gases like hydrogen chloride 5000 4000 3000 2000 1000 0
at 1 atm and 0 C the variation of n through the fun- (cm1)
damental HCl stretching band is only 0.0005 about Figure 2. The real (a) and imaginary (b) refractive index spectra
n 1.00045.13 Thus, only a measure of absorption is usu- of liquid benzene16,17 and water18 at 25 C.
ally needed to describe the optical behavior of a gas.
For liquids and solids both n and k are required. In non-hydrogen-bonded and nonpolar materials in that n
regions where a material does not absorb, n is determined changes little through the IR, except for sharp regions
by the polarization of the material by the electric field of of anomalous dispersion, and also changes little down to
the electromagnetic radiation. The material is not polarized zero frequency. The out-of-plane hydrogen vibration near
if the frequency of the radiation is extremely high, because 673 cm1 is one of the more intense vibrations in molecular
its electrons cannot follow the changing electric field of the compounds and is notable for its large effect.
radiation, and n D 1 at nQ D 1. As the frequency decreases The facts that n is not near unity and undergoes marked
from infinity, n decreases below 1 until the highest energy changes through the IR indicate that both n and k are
absorption is reached and then increases through the anoma- needed to describe the IR optical behavior of liquids and
lous dispersion of this absorption and the other absorptions solids. A frequently observed manifestation of this need is
in the X-ray and UV regions. For colorless compounds the shown in Figure 3, which shows the transmission spectrum
visible is a region of normal dispersion in which most liq-
of a 50 m thick KBr cell full of chlorobenzene21 and the
uids and solids formed by light elements have refractive
indices between 1.3 and 1.614,15 due to the polarization of 0.3
the electron cloud, i.e. to electronic polarization. Further, C6H5Cl(l)
for most colorless compounds, n is less than 0.03 smaller
at 5000 cm1 than in the visible.10
0.2
Absorbance
Figure 2 shows IR spectra of n and k for two liq-

uids, benzene16,17 and water,18 at 25 C; the UV optical
constants of water have been illustrated elsewhere.19 The
broad, intense absorption by water causes n to vary almost 0.1
continually through four regions of anomalous dispersion
between 5000 and 1 cm1 , where n D 4.9. Below 1 cm1
n continues to increase to 9 near 0.03 cm1 (1 GHz)20 Empty cell
0.0
due to the reorientational relaxation, then n remains near 2000 1800 1600
9 to zero frequency. This behavior illustrates the disper- (cm1)
sion of n of hydrogen-bonded materials, although water Figure 3. Experimental absorbance spectra of a 50-m cell full
is an extreme example. The dispersion of n of ben- of chlorobenzene21 and the same cell full of air, both referenced
zene in Figure 2 illustrates the much gentler behavior for against the empty instrument.
Optical Constants 3
spectrum through the same cell empty. Interference fringes The relation between the imaginary refractive index
are visible for the empty cell but not for the full cell, due and the decadic molar absorption coefficient or molar
to the different refractive indices of air, chlorobenzene and absorptivity, Em , follows from this result [4]. Thus,
KBr. These fringes show clearly that the spectrum of the
empty cell cannot be used to remove cell effects from the 4pnQ k
Em D 7
spectrum of the cell full of liquid, and that both the n and 2.303c
k are needed to analyze such spectra. The cell path length A10
can be determined to 1% from the fringes. where Em D 7a
c

I
and A10 D log10 7b
2 THE RELATION OF OPTICAL Io
CONSTANTS TO OTHER INTENSITY Here c is the amount concentration, i.e. the mole concen-
QUANTITIES tration, and A10 is the decadic absorbance3 and is called
simply the absorbance in this article.
Maxwells equations1,2,4 of electromagnetism applied to The absorption coefficients are proportional to nQ k. The-
an isotropic dielectric material yield equation (3) for the oretically this means that k is a measure of the transition
electric field of the radiation. probability while the absorption coefficients are measures
of the energy absorbed. This also means that k D 0.5,
E D Eo exp2pnQ kr exp[i2pnQ nr 2pft] 3 which is a fairly large value for the IR, corresponds to
a linear absorption coefficient a of 628.3 cm1 at nQ D
where nQ is the vacuum wavenumber, r is the position vector,
100 cm1 , 6283 cm1 at nQ D 1000 cm1 , and 21 991 cm1
f is the frequency, t is the time and n and k are the optical
at nQ D 3500 cm1 . Hence the path lengths required for
constants. In this formulation, the wave travels along r
10% transmission, D 2.303/a, are 36.6 m at 100 cm1 ,
with the electric vector perpendicular to r and perpendicular
3.66 m at 1000 cm1 , and 1.05 m at 3500 cm1 .
to the magnetic vector H. Note that for anisotropic materials
Note that in equations (6b) and (7b) the intensities I and
the situation is more complex but the same result is obtained
Io must be corrected for reflection and other cell effects so
for waves that travel along a principal axis of the optical
that they differ solely due to absorption.
indicatrix that is fixed by symmetry.4
The intensity of radiation is the energy crossing unit area The complex dielectric constant eO D e0 C ie00 is an inten-
in 2 sity quantity widely used by physicists. The real part e0
p unit time and is given by I D njEo j /2Zo , where Zo D is called the dielectric constant and the imaginary part e00
eo /mo , the square root of the ratio of the permittivity to
the permeability of vacuum. Thus, electromagnetic theory is often called the dielectric loss [5]. From the definitions
gives equation (4) for the intensity in a beam after it has of electromagnetism,1 4 the relative magnetic permeability
traveled distance r through a material from the position in equals 1 for nonmagnetic materials, and for such materials
the material where the intensity was Io .
eO D nO 2 e0 D n2 k 2 and e00 D 2nk 8
I D Io exp4pnQ kr 4
Chemists are accustomed to dielectric constants only for
Comparison of equation (4) with equation (5), the cor- radiowaves and microwaves, but the complex dielectric
responding equation in terms of the linear absorption constant is used for all wavenumbers and the imaginary part
coefficient a, yields equation (6) for the relation between describes absorption. At wavenumbers where there is no
the imaginary refractive index and the linear absorption absorption, i.e. where e00 D 0, e0 describes the polarization
coefficient. of the material by the electromagnetic wave and may be
measured as the capacitance of a capacitor full of dielectric
I D Io expar 5
relative to its capacitance when evacuated. The relation
a D 4pnQ k 6 between the real and imaginary parts of the dielectric
Ae constant and refractive index of methanol-d4 22 is clear in
where aD 6a Figure 4, where the upper curve in each box is the dielectric

constant and the lower curve is the refractive index.
I
and Ae D ln 6b Note that for an isolated band during anomalous disper-
Io
sion the difference between the maximum and minimum
Here Ae is the Napierian absorbance3 and is the path values of e0 equals the peak value of e00 , and that e0 values
length. can be negative.
2.5 RnQ D rO nQ rO nQ 10a

(a)

rO nQ D rnQ eifQn 10b

2.0 rnQ D RnQ 10c

and ln rO nQ D 12 ln[RnQ ] C ifnQ 10d
1.5
The conditions for validity of the KramersKronig trans-
form are met by 12 ln[RnQ ] 12 lnR1 D 12 ln[RnQ /R1 ]
n
and the phase shift fnQ , so equation (9) is also valid when
1.0
1.2
nnQ -n1 is replaced by 12 ln[RnQ /R1 ] and k is replaced by
(b) fnQ .23 27

The most important applications of the KramersKronig
0.8 relations for this article are to the optical constants and
to the reflectance and phase change on reflection. The
greatest problem in their application lies in the fact that
0.4 the integrals in equation (9) are over all wavenumbers from
zero to infinity and data are never available for the entire
k range.
0.0 For the interconversion of n and k, algorithms have been
2000 1000 0
(cm1)
presented that provide 0.1% accuracy in the calculated n
and 0.2% in the calculated k for the hypothetical case that
Figure 4. The real (a) and imaginary (b) dielectric constants and
both n and k are known from zero to infinity.28,29 Two of
refractive indices of CD3 OD at 25 C.22
these algorithms use equation (9) directly and make use of
McLaurins expansion to obtain the principal value, while
3 THE KRAMERSKRONIG others use the speed of the fast Fourier transform to carry
out the equivalent Hilbert transform. One of the conditions
TRANSFORMS of validity of the KramersKronig transform is that the real
and imaginary parts of the complex function must be even
The KramersKronig transforms,1,2,7,23 27 are based on the and odd functions, respectively, of wavenumber [8]. This
principle of causality [6] and connect the real and imagi- is also the requirement for the KramersKronig transform
nary parts of many complex quantities in physics [7]. The to be equivalent to the Hilbert transform, which is, in turn,
transforms are valid for any complex quantity that meets equivalent to two successive Fourier transforms. Thus the
certain conditions that are discussed23,24 and summarized25 speed of the fast Fourier transform can be used provided
elsewhere. These conditions are met by the optical con- that appropriate manipulation of the data arrays occurs
stants of insulators,1,24,25 so n and k are connected by the between the two Fourier transforms.29
KramersKronig transforms, equation (9a) and (9b). In practice, n and k are not known from zero wavenum-
1 ber to infinity. There is nearly always a lack of infor-
2 nQ knQ
nnQ a n1 D P dnQ 9a mation at both high and low wavenumbers. The usual
p o nQ 2 nQ 2a situation for IR spectroscopists is that n is known in the
1
2nQ a nnQ n1 visible, n2 D e0 is known at microwave frequencies, and
knQ a D P dnQ 9b values of k are available for use in equation (9a) from
p o nQ 2 nQ 2a
measurements between about 5000 cm1 and 400 cm1 .
where P means that the principal value of the integral is In such situations the lack of high wavenumber data can
taken to resolve the singularity at nQ D nQ a . be dealt with for colorless liquids and solids in the fol-
The conditions are also met by the complex dielectric lowing way. The real refractive index is well known at
constant,1,23,24 and equation (9a) and (9b) also interrelate e0 several visible wavenumbers for most common liquids and
and e00 when n is replaced by e0 and k is replaced by e00 .24 solids.10,14,15 These values are rarely affected by the IR
The amplitude and the phase change on reflection are also absorption above the fourth decimal place10 so are due to
linked by the KramersKronig transforms. The reflectance, electronic polarization to the third decimal place or bet-
RnQ , the complex coefficient of reflection, rO nQ , and the ter. They can be fitted to equation (2) and extrapolated to
phase shift on reflection, fnQ , are connected by give nel at 8000 cm1 .10 The absorption between 5000 cm1
equation (10). and 8000 cm1 is too weak to cause significant change
Optical Constants 5
in n, so k can be set to zero in this range. The upper 4 DETERMINATION OF THE OPTICAL
integration limit in equation (9a) is made 8000 cm1 and CONSTANTS OF ISOTROPIC
n1 is replaced by nel (8000 cm1 ). In the authors lab-
oratory, the upper limit of measurement is 6250 cm1 , MATERIALS
and the results obtained in this way indicate that the error
in n is not above 0.003. Further, because normal disper- There are two optical constants to be determined so two
sion reduces nel by 0.003 through the IR, this error can measurements are needed at each wavenumber. Alterna-
be reduced further if the extrapolated value of nel nQ a is tively, one of these measurements may be replaced by the
used at each wavenumber nQ a instead of the constant value KramersKronig transform and a value of n at a single
of nel 8000 cm1 .10 The lack of data to low wavenum- wavenumber. These requirements can be met in a variety
of ways and many different methods have been used. The
ber is potentially more serious, but the error in n can
methods used before 1970 and the results obtained have
be reduced by a factor of 10 simply by joining the last
been reviewed.35 This section describes methods in cur-
point in the k spectrum to zero at 0 cm1 with a straight
rent use. Specifically, they are external reflection, which is
line.29 Even greater improvement is possible if the spectrum
the best method for very strong absorption by nonvolatile
between the low wavenumber limit and zero is known qual-
materials, ATR spectroscopy, which is far more sensitive
itatively and can be added approximately to the measured
than external reflection, ellipsometry, which should perhaps
k spectrum.
be more widely used, dispersive Fourier transform spec-
The KramersKronig transform between 12 ln[RnQ /R1 ]
trometry, which has been widely used in the far-IR, and
and the phase change on reflection fnQ has been studied
transmission spectroscopy, which is the most suitable for
extensively because the transform is not the same for all of
weak absorption.
the different cases of reflection.30,31 Equation (11) applies
to external reflection at normal incidence.
1
1 RnQ a 2 nQ fnQ 4.1 From external reflection measurements
ln D P dnQ 11a
2 R1 p 0 nQ 2 nQ 2a
1 1 The earliest method used to determine IR optical cons-
2nQ a 2
ln[RnQ /R1 ] tants26,27 was external reflection spectroscopy at near-
fnQ a D P dnQ 11b
p 0 nQ 2 nQ 2a normal incidence, which usually includes incident angles
up to 15 . The reflectance spectrum, RnQ , from the surface
The KramersKronig transform has also been applied to
of the liquid or solid is measured, and the KramersKronig
p-polarized30 and s-polarized30,32 external reflection and
transform of equation (11b) is used to calculate the phase
to p-polarized30,33,34 or s-polarized25,30,32 attenuated total
change spectrum, fnQ . The optical constants are calculated
reflection (ATR), both at non-normal incidence [9]. In most at each wavenumber through equation (13).1,36
applications to ATR the refractive index of the internal
p
reflection element has been assumed to be independent of 1R 2 R sin f
wavenumber.30 34 However, the real case, in which the nD p kD p 13
1 C R 2 R cos f 1 C R 2 R cos f
refractive index of the internal reflection element changes
with wavenumber, has been considered for s-polarized The method remains valuable for involatile samples that
ATR25 and gave equation (12) instead of equation (11b) absorb strongly. Modifications recognize the use of non-
for the phase change. normal incidence and the need to use different algebra
for s- and p-polarized measurements.30 34 The sensitivity
fs nQ a D f0s nQ a C fs nQ a 12a of the method has been enhanced slightly30,37 by cover-
nQ u 1 ing the sample with an AgCl plate, but ATR methods
2 nQ 2 ln[Rs nQ ] give far greater sensitivity. IR optical constants have also
f0s nQ a D P dnQ 12b
p 0 nQ 2 nQ 2a been determined by measuring external reflectance at two
fs nQ D fo nQ f0s nQ u 12c angles of incidence without using the KramersKronig
transform.36
[n2o nQ sin2 qo n2 nQ u ]
fo nQ D p 2 arctan 12d
no nQ cos qo
4.2 From ellipsometry
where no nQ is the refractive index of the window material,
nQ u is chosen to be in a region of constant reflectance at high The name ellipsometry38,39 comes from the fact that plane
wavenumber, and qo is the angle of incidence. polarized light becomes elliptically polarized after reflection
at non-normal incidence. The ellipticity, , is measured path shift and the amplitudes of the interferograms.1,42
as the difference between the phase shifts on reflection The method requires a parallel-faced sample that does not
of s- and p-polarized light, fp fs . Knowledge of and absorb completely and whose thickness can be measured.
of the square root of the ratio of the p- and s-polarized It was, therefore, initially used largely for solids but over
reflectances, tan y D Rp /Rs , allows the optical con- the past 30 years has also yielded accurate far-IR optical
stants to be calculated. In ellipsometry the optical constants constants of a large number of simple liquids.43
are determined from measurements at the wavenumber of
interest without use of the KramersKronig transform.
A very clear and concise introductory account has been 4.4 From attenuated total reflection
given by Roseler and Korte.40 Four intensity (i.e. single- measurements
beam) spectra are measured, with one polarizer after the
sample set to 45 for all four spectra and a second polarizer External reflection methods are insensitive to moderate
before the sample set to 0 for the first spectrum, 45 for the to weak absorption. Internal reflection occurs when the
next, 90 for the next and 135 for the last. The ellipsometric sample is in optical contact with a nonabsorbing material of
variables and the dielectric constants are then determined higher refractive index through which the radiation travels
through equation (14). before and after reflection. The internal reflection spectrum
I90 I0 is largely determined by the real refractive index if the
cos 2y D (14a) angle of incidence is less than the critical angle, and by
I90 C I0
the imaginary refractive index if the angle of incidence
I45 I135 exceeds the critical angle. In the latter case total internal
sin 2y cos D (14b)
I45 C I135 reflection occurs if the sample does not absorb, and ATR44

cos2 2y sin2 2y sin2 or frustrated total internal reflection45 occurs if the sample
0 2 2
e D no sin qo 1 C tan qo does absorb [10]. The sensitivity near to the critical angle
1 C sin 2y cos 2
is far greater than in external reflection, but declines as the
(14c)
difference between the incident and critical angles increases
sin 4y sin in either direction. The sensitivity can be increased by
e00 D no sin qo tan qo 2 (14d)
1 C sin 2y cos 2 multiple ATR,45,46 i.e. by repeating the initial reflection
Here qo is the angle of incidence and no is the real refractive several times.
index of the incident medium. The optical constants are Optical constants have been obtained from ATR or mul-
calculated from e0 and e00 by the reverse1,25 of equation (8). tiple ATR measurements either by measurement of the
These equations of ellipsometry apply to both external reflectance, i.e. Rs , Rp , Rs /Rp or a known combination of
reflection and ATR.40 them, at two or more angles of incidence,6,19,46 53 or at
Most IR ellipsometry today is done with the Fourier a single angle of incidence and calculation of the phase
transform spectrometer.39,40 It is perhaps the main method shift through the KramersKronig transform.19,31 34,44,53
used for inorganic compounds while the other methods The main difficulty with these methods is that accurate
described here have been used mainly for organic com- results require use of collimated incident light striking the
pounds. The use of IR lasers for ellipsometry has also been reflecting interface at an angle that is known to minutes
described.41 of arc.
A method of calibrated ATR spectroscopy has been used
for liquids to reduce the experimental difficulty of setting
4.3 From dispersive Fourier transform the incident angle to high accuracy, by using the optical
measurements constants of benzene16,54,55 to calibrate the apparatus.56 58
The method was developed for the Spectra-Tech CIRCLE
In dispersive42 or asymmetric1 Fourier transform spec- cell, a multiple ATR accessory in which the radiation under-
trometry an interferogram is recorded with a Michelson goes several reflections at close to 45 incidence inside a
interferometer with no sample in the instrument, followed cylindrical ZnSe internal reflection element whose sides
by an interferogram with the sample in one arm of the are surrounded by the sample. The effective number of
interferometer. The presence of the sample in one arm reflections at an assumed 45.00 incidence is calibrated
shifts the zero-path-difference position and, because the real from measured spectra of benzene. To determine the opti-
refractive index of the sample changes with wavenumber, cal constants of a liquid, an approximate k spectrum is
increases the asymmetry of the interferogram. The opti- calculated from the pATR spectrum [11] and the corre-
cal constants of the sample are calculated from the zero sponding n spectrum is calculated via the KramersKronig
Optical Constants 7
transform, equation (9a). These approximate n and k spec- 1. An approximate n spectrum is calculated from the k
tra are refined by calculating the pATR spectrum from spectrum through equation (9a).
them through Fresnels equations,1,4,5 comparing this with 2. The n and k spectra are used with Fresnels equa-
the measured spectrum, reducing k at those wavenum- tions,1,4,62 the n spectrum of the windows nw nQ , and
bers for which the calculated spectrum is higher than the the path length to calculate the transmission spectrum.
observed spectrum and vice versa, and repeating the cycle 3. The k spectrum is adjusted to improve the fit between
to convergence.56 58 The refinement procedure fails for the calculated and measured transmission spectra.
very strong absorptions because for large k the pATR spec-
trum does not always increase with k,59 but the method is The transmission method is particularly valuable for
suitable for any absorption weak enough to be measured weak and moderately strong bands, and is limited only by
in the CIRCLE cell58 and has been applied to water18 and the need for transmission cells with path lengths down to
methanol.9,22,60 A superior refinement procedure that does <1 m for the strongest IR bands.73 Cells with path lengths
not fail is to calculate the phase shift, f, spectrum from of several millimeters are also needed to obtain accurate k
the original approximate n and k spectra, calculate the k values in the baseline regions.65,68 72
It has been found recently that for all but the strongest
spectrum that yields the observed pATR spectrum with the
bands the correction for reflection effects can usually be
n spectrum and this f spectrum, recalculate the n spectrum
simplified without significant degradation of accuracy.74
via equation (9a), and iterate to convergence.59
When the refractive indices of sample and window are
An algorithm has been described25 for s-polarized ATR
between 1.3 and 1.6 the contribution from the airwindow
at 45 , and for calibrated CIRCLE cell measurements, that
and samplewindow interfaces to the experimental absor-
incorporates all recent improvements and is believed to
bance of the cell plus sample is given to good accuracy by
be the best currently available. The optical constants are
the spectrum of log10 f2nw nQ /[n2w nQ C 1]g. This spec-
calculated through equations (12) and (15), in which qo and
trum is subtracted from the experimental absorbance spec-
no are the angle of incidence and the refractive index of
trum of the full cell, the baseline is adjusted to the position
the internal reflection element, followed by the reverse of
indicated by spectra through very thick cells, and the k
equation (8).
spectrum is calculated through equation (6).74
e0 1 Rs 2 4Rs sin2 fs The Fortran programs for the computation of optical
2
D sin2 qo C cos2 qo p 15a constants from transmission spectra and from CIRCLE
no 1 C Rs 2 Rs cos fs 2
p cell ATR spectra are available [12], as are optical constant
e00 2 sin fs 1 Rs Rs spectra of several compounds in digital form.
D 4 cos qo p 15b
n2o 1 C Rs 2 Rs cos fs 2
The optical constants are then refined by the above proce- 5 USE OF OPTICAL CONSTANTS IN
dure through the phase shift, with the k spectrum extended
SPECTROSCOPY
linearly to zero at 0 cm1 before the n spectrum is calcu-
lated through equation (9a). The optical constants can be used to calculate any intensity
quantity and the result of any well-defined optical exper-
iment. This is illustrated for external reflection, internal
4.5 From transmission measurements reflection and ATR in Figure 5. The figure shows the s-
polarized reflectance, Rs , from the airC6 H6 or ZnSeC6 H6
Transmission measurements have been widely used to interface between 1550 cm1 and 950 cm1 . The spectra
determine optical constants.16,21,36,54,55,61 74 Reflection were calculated through Fresnels equation4,56 from the n
effects cause serious errors for very strong bands36,61 and k spectra of C6 H6 16,55 and the n spectrum of ZnSe.75
but this problem can be completely solved by applying For the ZnSeC6 H6 interface, incident angles of 10 , 35 ,
Fresnels equations1,4,5 to the interfaces. An algorithm 45 and 80 were used and the latter three are labeled A in
to achieve this was developed62,63 during the 1970s and Figure 5. The points made earlier about the dependence of
applied to solids64 and to many liquids.54,63,66,67 The current the internal reflection spectrum on angle of incidence, qo ,
version65,74 has been applied to liquids.68 72 The algorithm are clear in the figure. The critical angle is 39 . Thus the
allows consideration of the convergence and polarization of 35 spectrum resembles a reflection spectrum while the 45
the light beam.62 An approximate k spectrum is first calcu- ATR spectrum resembles a transmission spectrum. The 80
lated from the measured transmission spectrum of the cell ATR spectrum also resembles a transmission spectrum but
full of liquid and is used in the following three-step iterative the peaks are far weaker than in the 45 spectrum so the
process to convergence: sensitivity to weak bands is poorer than at 45 .
A 80
1.00 1.5
A 45 EA
0.75
1.0 A
Rs
0.50
A 35 0.5
pATR
0.25 k
0.0
E 45 3500 3250 3000 2750
0.00
1400 1200 1000
1.5
(cm1)
Figure 5. s-Polarized reflection spectra of liquid C6 H6 at 25
calculated from the optical constants.16 Curves marked A and 1.0
the thin line second from the bottom show the internal reflectance
through ZnSe at 80 , 45 , 35 and 10 . The curve marked E 45
and the bottom curve show the external reflectance in air at 45 0.5
and 10 .
The 10 internal reflection spectrum is the thin line 0.0

just above the bottom spectrum. It resembles a reflection 1500 1250 1000 750
spectrum but is far weaker than the 35 spectrum. It is,
however, about twice as strong and sensitive to weak 1.5 EA
bands as the spectrum of 10 external reflection from the A pATR
airC6 H6 interface (Figure 5). This result shows the benefit
noted above of covering near-normal-incidence external 1.0
reflection samples with a high-refractive-index window.
Figure 5 also shows the spectrum of 45 external reflec-
0.5
tion from the airC6 H6 interface. It is labeled E 45 and k
comparison with the 45 ATR spectrum shows the large
gain in sensitivity provided by ATR. 0.0
1075 1050 1025 1000 975
Another use of the optical constant spectra is shown
in Figure 6, which contains spectra of several intensity (cm1)
quantities of CH3 OH. The intensity quantities are the imag- Figure 6. Experimental absorbance, EA, absorbance, A, and
inary refractive index, k, the absorbance, A, of a 5 m pATR spectra of CH3 OH at 25 calculated from the optical
constants9 as noted in the text, together with the k spectrum. The
thick sample of CH3 OH calculated from k through equa- sequence of curves is the same in all boxes except that the pATR
tion (6) and (6a) then divided by 2.303, the experimental curve crosses the EA and A curves on the right-hand side of the
absorbance, EA, of 5 m of CH3 OH in a cell with KBr middle box.
windows, and the pATR calculated as log10 of the atten-
uated total reflectance after two 45 reflections from the spectrum obtained as log10 of the transmission spec-
ZnSeCH3 OH interface of light that is 50% s-polarized and trum of the cell full of sample is called the experimental
50% p-polarized with retention of polarization between the absorbance spectrum and the word absorbance is used
reflections. The latter conditions are those for the CIRCLE only to describe absorption by the sample. The difference
cell.56 between the experimental absorbance and absorbance is
The term absorbance spectrum is widely used for clear in Figure 6, and is not independent of wavenumber.
log10 of a transmittance spectrum and also for log10 The lower part of Figure 6 shows the region of the intense
of an ATR spectrum. In the former case it contains win- CO stretching band, where the EA is higher at the ends
dow effects, specifically effects of reflection at the window of the range but essentially coincides with the absorbance
surfaces and of vignetting by the window. Absorbance at the peak. Clearly a low estimate of the absorbance
is defined3 and used in quantitative work as solely due peak height would be obtained by subtracting from the EA
to absorption, and it is important in quantitative work to peak the reflection correction estimated from the baseline.
use terms precisely. Thus, in the authors laboratory, the Figure 6 also shows that this error is smaller in regions of
Optical Constants 9
weaker absorption, and in fact it can be ignored for weak formula when it is applied to nonabsorbing materials.5 It
absorption. can be used at all wavenumbers if the dielectric constant and
The difference between the absorbance spectrum, A, polarizability are complex so that they describe absorbing
and the pATR spectrum is larger and even more impor- materials.31,76,77 If aO is multiplied by the Avogadro con-
tant (Figure 6). The peak wavenumber in the pATR spec- stant to obtain the molar polarizability, aO m , the real and
trum is always lower, by 9 cm1 for the CO stretch, imaginary molar polarizabilities, a0m and a00m , are calculated
23 cm1 for the OH stretch and 3 cm1 for the CH through equation (17a) and (17b) [14].
stretches. Further, the pATR intensities decrease markedly
relative to those in the absorbance spectrum with increasing 3Vm e0 1e0 C 2 C e002
wavenumber. The absorbance spectrum is a measure of the a0m D 17a
4p e0 C 22 C e002
energy absorbed, i.e. of the transition probability multiplied
by the energy per transition, hnQ , while the pATR spectrum, 9Vm e00
a00m D 17b
like the k spectrum, is a measure of the transition proba- 4p e0 C 22 C e002
bility. For these reasons the use of the term absorbance for
log10 of the attenuated total reflectance is strongly dis- where Vm is the molar volume. The local susceptibility,76
O D C0 C iC00 , is the molar polarizability per unit volume,
C
couraged, and the term pATR is recommended instead [13].
The final point to note in Figure 6 is that the peak and is more useful for comparing different materials.
wavenumbers and bandshapes in the k spectrum are essen- Figure 7 shows the a00m , e00 and k spectra of the strongest
tially the same as in the absorbance and experimental absorption of liquid C6 H6 , the band near 673 cm1 .16
absorbance spectra and also, but not shown, in the molar
absorption coefficient spectrum. This is in spite of the fact 16
that the latter three intensity quantities are weighted by (a)
wavenumber relative to k. 12
6 USE OF OPTICAL CONSTANTS TO 8
OBTAIN MOLECULAR PROPERTIES

4
Like the dielectric constants, molar absorption coefficient
and absorbance, n and k, are properties of the bulk 0
material. They reflect properties of the molecules in the (b)
anisotropic environment created by their near neighbors 4
as they are modified by long-range effects in the bulk. In
order to determine the properties of the molecules in their
anisotropic environment, it is necessary to remove these
long-range effects. This is achieved by determining the rela- 2
tion between the local electric field, Eloc , that acts on a

molecule and the macroscopic electric field, E, that acts on
the bulk material. A general relation between the two is 0
only available for an isotropic material. The relation must
(c)
be determined individually for anisotropic materials.
1.0
For isotropic materials the required relation is the Lorentz
local field,5,76 Eloc D E C 4p/3P , where P is the polar-
ization of the material. This local field yields equation (16)
as the relation between the optical constants, dielectric 0.5
constants, number density of molecules, N, and mean com-
plex polarizability of a molecule in the liquid, aO nQ D
a0 nQ C ia00 nQ .
0.0
eO nQ 1 nO 2 nQ 1 4 700 675 650
D 2 D pNaO nQ 16 (cm1)
eO nQ C 2 nO nQ C 2 3
Figure 7. (a) The imaginary molar polarizability, a00m , (b) dielec-
Equation (16) is known as the ClausiusMossotti equation tric constant, e00 , and (c) refractive index, k, spectra of the most
when it is used at zero frequency, and the LorentzLorenz intense band of liquid benzene at 25 C.16
with poor precision.77 Such wavenumber differences have

0.8 (a)
been reported for the most intense bands of four iso-
topic forms of methanol,9,22,60 C6 H6 ,76 C6 D6 ,68 C6 H5 D,69
C6 F6 ,50 CHCl3 ,31 acetone,78 C6 H5 Cl78 and CH2 Cl2 .78 For
medium and weak bands the peak wavenumber differ-
0.4
ences between the a00m , e00 and k spectra are less than
0.2 cm1 , as has been described for four isotopic forms
of methanol,9,22,60 C6 D6 ,68 and C6 H5 D.69
According to theory,31,77 the peak wavenumber in the
0.0 nQ a00m spectrum is the wavenumber of the molecule in its
0.8 (b) liquid environment that is appropriate to compare with
wavenumbers in the gas phase. For all but broad bands
this is within 0.2 cm1 of the peak wavenumber in the
a00m spectrum. It can be expected, therefore, that this is
0.4 also the correct wavenumber to be compared with Raman
shifts. It has been shown,78 however, that there is lit-
tle general consistency in such IRRaman comparisons,
undoubtedly because of the same fundamental physics that
0.0
causes noncoincidence between isotropic and anisotropic
1.0 (c) Raman bands.78
Em
Figure 7 also shows that the a00m band is far more sym-
k metrical than the e00 and k bands. This, again, has been
observed for many compounds9,22,31,49 52,60,68,69,76 78,80
0.5 and is expected from theory.31,77 Liquid a00m bands are
found68,69,77,80 to be very close to classical damped har-
monic oscillator (CDHO) bands (equation 18), unless the
vibration involves a hydrogen-bonded hydrogen atom, as
0.0 in Figure 8, in which case the bands resemble Gaussian
3500 3000
bands, usually sums of Gaussian bands.80
(cm1)
Figure 8. (a) The imaginary molar polarizability, a00m , (b) dielec- Sj j nQ
a00m D 18
tric constant, e00 , and (c) molar absorption coefficient, Em , and nQ 2j nQ 2 C 2j nQ 2
refractive index, k, spectra of the intense OH stretching band of
liquid H2 O at 25 C.18 The Em ordinate has been scaled for simple NA m2j /3
comparison. where Sj D 18a
4p2 c2
and j is the full width at half-height (FWHH).
Figure 8 shows the same spectra of the broad and intense
The CDHO model, and other theoretical models, of IR
OH stretching band of liquid water.18
spectra give polarizability spectra in the first instance. The
Figure 7 shows that the peak wavenumbers differ in the
CDHO model gives the a00m spectrum as a sum of CDHO
three spectra, being 676 cm1 in the a00m spectrum, 671 cm1
bands.77 Such models then compute the e00 and e0 spectra
in the e00 spectrum and 673 cm1 in the k spectrum. The
and the n and k spectra from the a00m and a0m spectra.77
peak wavenumbers in the absorbance spectrum, the exper-
Neighboring bands influence each other in the k and e00
imental absorbance spectrum and the molar absorption
spectra if they are sufficiently close and intense, so it is only
coefficient spectrum are the same as in the k spectrum.
the a00m spectrum that is theoretically composed of a sum of
Similarly in Figure 8, the peak wavenumbers are 3427 cm1 individual bands.77 The a00m spectrum is, therefore, the spec-
for a00m , 3389 cm1 for e00 , 3406 cm1 for k and in this case trum from which molecular properties should be obtained.
slightly higher for the molar absorption coefficient, Em , at
3413 cm1 due to the extreme breadth of the band.
Theoretically,77 the peak wavenumber is expected to 6.1 Use of optical constants to obtain integrated
decrease in the sequence a00m > k > e00 for all strong bands. absorption intensities
The effect is related to the band intensity, which can be
calculated from the difference between the squares of the Integrated absorption intensities of molecular vibrations
peak wavenumbers in the a00m and e00 spectra but only in isotropic condensed phases are obtained, completely
Optical Constants 11
corrected for long-range effects, from the imaginary molar integration limits. Otherwise these methods are accurate
polarizability spectrum through equation (19a) and are for weak bands, but for strong bands it is impossible to
related to molecular properties through equation (19b): determine accurately the hypothetical value of n required
for the correction for condensed-phase effects through
Cj D nQ a00m nQ dnQ 19a equation (21b). The use of Cj is far more accurate.77
The traditional method of comparing condensed-phase
NA p NA p intensities with gas-phase intensities is through the
Cj D gj nQ j jRj j2 D gj m2j 19b
3hco 24hco2 PoloWilson equation:
where NA is Avogadros number, h and co are Plancks con- 2
Aj,liq 1 n2 C 2 nQ j,liq jRj,liq j2
stant and the velocity of light in vacuum and, for vibration D
j, gj is the degeneracy, nQ j is the wavenumber, jRj j2 is the Aj,gas n 3 nQ j,gas jRj,gas j2
square of the transition moment and m2j is the square of the 2
1 n2 C 2 m2j,liq
derivative of the dipole moment with respect to the normal D 22
n 3 m2j,gas
coordinate. Equation (19b) assumes that all hot bands of the
transition are included in the band envelope, and nQ j jRj j2 is This has the disadvantages of needing the hypothetical
really the weighted sum of the terms for each such tran- value of n discussed above, and of being numerically differ-
sition. The second equality in equation (19b) assumes the ent for each band of each compound as well as for different
double harmonic approximation. compounds. The molar polarizability provides a cleaner
In practice the integrated intensities have been way of relating liquid and gas intensities. Equation (19b)
determined by fitting the a00m spectrum with CDHO for the liquid and equation (21b) for the gas yield equa-
bands,68,69,77,81,82 or a mixture of CDHO and Gaussian tion (23):
bands for hydrogen-bonded systems.80 For the CDHO band
of equation (18), Cj D Sj p/2. Cj,liq nQ j,liq jRj,liq j2 m2j,liq
p For the Gaussian band of 8p2 D D 23
equation (20), Cj D nQ j j Hj fp/[4 ln2]g. Aj,gas nQ j,gas jRj,gas j2 m2j,gas

00 4 ln2 2
Thus, if the nQ j jRj j2 product and m2j are the same for gas
am D Hj exp nQ nQ j 20 and liquid phases, we have the simple result that Aj,gas D
2j
8p2 Cj,liq D 78.96Cj,liq for all bands of all compounds.
Thus, the integrated intensities are determined directly This equation has been applied to C6 H6 ,81 C6 D6 68 and
from the fitting parameters, which eliminates the usually C6 H5 D69 with the result that the CH stretching bands are
impossible task of integrating over a sufficiently wide range clearly about half as intense in the liquid as in the gas,
that 99% of the intensity of the bands is recovered. In but otherwise there is little evidence of different intensities
this way integrated intensities have been determined and in the two phases. Equation (23) has also been applied
published for four isotopomers of methanol,80 C6 H6 ,81 to methanol80 to compare experimental intensities for the
C6 D6 ,68 C6 H5 D69 and C6 H5 Br.82 liquid with ab initio intensities for the isolated molecule.
Integrated intensities can also be determined from the The results suggested that the phase dependence of IR
imaginary dielectric constant, e00 , spectrum77 and from intensities may not be as large as is frequently thought
the k spectrum.77 In the latter case the molar absorption except for the XH stretching vibrations of hydrogen-
coefficient Em is calculated through equation (7) and used bonded hydrogen.
in equation (21a), and the integrated intensity is related to
molecular properties through equation (21b).
Aj D 2.303 Em nQ dnQ 21a
2 CDHO Classical Damped Harmonic Oscillator
8p3 NA 1 n2 C 2
Aj D gj nQ j jRj j2
3hco n 3
2 NOTES
pNA 1 n2 C 2
D gj m2j 21b
3co2 n 3
[1] The complex refractive index has also been defined
Here n is the hypothetical value that n would have at as n C ik4 and n1 C ik/.5,6 For E D E0 exp[i!t
wavenumber nQ j if the band were not present. The use k .r], n C ik is changed to n ik.7,8
of e00 or Em requires direct integration so needs wide [2] Other names have also been used for k.
[3] In this article the term wavenumber means the 2. F. Wooten, Optical Properties of Solids, Academic Press,
vacuum wavenumber of the radiation. New York (1972).
[4] The symbol Em is used for the molar absorption 3. I. Mills, T. Cvitas, K. Homann, N. Kallay and K. Kuchitsu,
coefficient instead of the recommended3 e to avoid Quantities, Units and Symbols in Physical Chemistry, 2nd
edition, Blackwell Scientific Publications, Oxford (1993).
confusion with the dielectric constant.
[5] In this article dielectric constants eO , e0 and e00 are 4. G.R. Fowles, Introduction to Modern Optics, Holt Rinehart
& Winston, New York (1975).
dimensionless quantities that are correctly called the
relative dielectric constants in the rationalized MKS 5. M. Born and E. Wolf, Principles of Optics, 6th edition,
Pergamon Press, Oxford (1984).
system3 of SI units. To obtain the dielectric constants
6. A.C. Gilby, J. Burr, Jr, W. Krueger and B. Crawford, Jr, J.
in SI units the values should be multiplied by eo D
Phys. Chem., 70, 1525 (1966).
8.8542 1012 F m1 .
7. T.S. Moss, Optical Properties of Semi-conductors, Butter-
[6] Causality means that a system cannot respond to an
worth Publications Ltd, London (1959).
action until after the action has occurred.
8. O.S. Heavens, Optical Properties of Thin Solid Films,
[7] There is a large literature on this subject so citations Dover Publications, Inc., New York (1965).
are chosen to guide the reader to the current practice
9. J.E. Bertie, S.L. Zhang, H.H. Eysel, S. Baluja and M.K.
and the fundamental issues. Ahmed, Appl. Spectrosc., 47, 1100 (1993).
[8] It is sometimes said that the KramersKronig rela-
10. J.E. Bertie and Z. Lan, J. Chem. Phys., 103, 10 152 (1995).
tions are not applicable to certain absorption band-
11. A.R.H. Cole, Tables of Wavenumbers for the Calibration
shapes, such as the Gaussian. This is never true
of Infrared Spectrometers, 2nd edition, Pergamon Press,
provided that the spectrum of the imaginary part is Oxford, 206210 (1977).
constructed to be an odd function of wavenumber. 12. B. Edlen, J. Opt. Soc. Am., 43, 339 (1953).
[9] No distinction is made between ATR and total internal
13. J.E. Chamberlain, F.D. Findlay and H.A. Gebbie, Appl. Opt.,
reflection. The only difference between them is that 4, 1382 (1965).
k D 0 for total internal reflection. In s-polarization the
14. J. Timmermans, Physico-chemical Constants of Pure Organic
electric vector of the radiation, E, is perpendicular to Compounds, Elsevier, Amsterdam, 98, Vol. 2 (1965).
the plane of incidence and, therefore, parallel to the
15. E.W. Washburn (ed.), International Critical Tables of Numer-
reflecting surface. In p-polarization E is in the plane of ical Data, Physics, Chemistry and Technology, McGraw-Hill,
incidence and at the angle of incidence to the reflecting New York, 1216, 3462, Vol. 7 (1930).
surface. 16. J.E. Bertie, R.N. Jones and C.D. Keefe, Appl. Spectrosc., 47,
[10] Harrick and Fahrenfort appear to have realized the 891 (1993).
potential of ATR, or frustrated total internal reflection 17. M.N. Afsar, Proc. IEEE, 67, 1460 (1970).
as Harrick called it, at essentially the same time. 18. J.E. Bertie and Z. Lan, Appl. Spectrosc., 50, 1047 (1996).
Harrick went on to develop it for penetrating surface
19. V.M. Zolotarev and A.V. Demin, Opt. Spectrosc. (USSR), 43,
layers while Fahrenfort pursued it for determining 157 (1977).
optical constants.
20. U. Kaatze and V. Uhlendorf, Z. Physikal. Chem. Neue Folge.,
[11] The pATR spectrum is log10 of the ATR spectrum. 126, 151 (1981).
[12] From ftp.chem.ualberta.ca/pub/jbertie or http://www. 21. J.E. Bertie, R.N. Jones and Y. Apelblat, Appl. Spectrosc., 48,
ualberta.ca/jbertie/jebhome.htm 144 (1994).
[13] pATR is derived by analogy with pH, and p means 22. J.E. Bertie and S.L. Zhang, J. Chem. Phys., 101, 8364 (1994).
log10 in both cases.
23. J.S. Toll, Phys. Rev., 104, 1760 (1956).
[14] The molar polarizability defined here has the units
24. M. Cardona, Optical Constants of Insulators, in Opti-
of volume, usually cm3 . It has been termed the
cal Properties of Solids, eds S. Nudelman and S.S. Mitra,
polarizability volume,3 and its units should be con- Plenum Press, New York, 137151 (1969).
sidered to be 4peo cm3 , so that 1 cm3 1.112650
25. J.E. Bertie and Z. Lan, J. Chem. Phys., 105, 8502 (1996).
1016 F m2 .3
26. T.S. Robinson, Proc. Phys. Soc., London, B65, 910 (1952).
27. T.S. Robinson and W.C. Price, Proc. Phys. Soc., London,
B66, 969 (1953).
REFERENCES
28. K. Ohta and H. Ishida, Appl. Spectrosc., 42, 952 (1988).
29. J.E. Bertie and S.L. Zhang, Can. J. Chem., 70, 520 (1992).
1. E.E. Bell, Optical Constants and their Measurement, in
Handbuch der Physik, ed. S. Flugge, Springer-Verlag, 30. J.S. Plaskett and P.N. Schatz, J. Chem. Phys., 38, 612 (1963).
Berlin, 158, Vol. XXV/2a (1967). 31. M.J. Dignam, Appl. Spectrosc. Rev., 24, 99 (1988).
Optical Constants 13
32. J.A. Bardwell and M.J. Dignam, J. Chem. Phys., 83, 5468 57. J.E. Bertie, H. Harke, M.K. Ahmed and H.H. Eysel, Croat.
(1985). Chem. Acta, 61, 391 (1988).
33. K. Yamamoto and H. Ishida, Spectrochim. Acta, 50A, 2079 58. J.E. Bertie, S.L. Zhang and R. Manji, Appl. Spectrosc., 46,
(1994). 1660 (1992).
34. M.J. Dignam and S. Mamiche-Afara, Spectrochim. Acta, 59. J.B. Huang and M.W. Urban, Appl. Spectrosc., 46, 1666
44A, 1435 (1988). (1992).
35. R.P. Young and R.N. Jones, Chem. Rev., 71, 219 (1971). 60. J.E. Bertie and S.L. Zhang, Appl. Spectrosc., 48, 176 (1994).
36. H.D. Downing and D. Williams, J. Geophys. Res., 80, 1656 61. S. Maeda and P.N. Schatz, J. Chem. Phys., 35, 1617 (1961),
(1975). and citations therein.
37. D.W. Barnes and P.N. Schatz, J. Chem. Phys., 38, 2662 62. J.P. Hawranek, P. Neelakantan, R.P. Young and R.N. Jones,
(1963). Spectrochim. Acta, 32A, 75, 85 (1976).
38. D.E. Aspnes, The Accurate Determination of Optical proper- 63. J.P. Hawranek and R.N. Jones, Spectrochim. Acta, 32A, 99,
ties by Ellipsometry, in The Handbook of Optical Constants 111 (1976).
of Solids, ed. E.D. Palik, Academic Press, Boston, 89112, 64. G.K. Ribbegard and R.N. Jones, Appl. Spectrosc., 34, 638
Chapter 5 (1985). (1980).
39. A. Roseler, Infrared Spectroscopic Ellipsometry, Akademie- 65. J.E. Bertie, C.D. Keefe and R.N. Jones, Can. J. Chem., 69,
Verlag, Berlin (1990). 1609 (1991).
40. A. Roseler and E.H. Korte, J. Mol. Struct., 349, 321 (1995). 66. W. Wrzeszcz, B. Czarnik-Matusewicz and J.P. Hawranek,
41. Y. Feiner and U.P. Oppenheim, Infrared Phys., 33, 289 (1992) Pol. J. Chem., 72, 302 (1998).
42. J. Chamberlain, The Principles of Interferometric Spec- 67. W. Wrzeszcz, B. Czarnik-Matusewicz and J.P. Hawranek, J.
troscopy, John Wiley & Sons, Chichester, 189215, Chap- Mol. Struct., 322, 181 (1994), and citations therein.
ter 7 (1979). 68. J.E. Bertie and C.D. Keefe, Fresenius J. Anal. Chem., 362,
43. J.R. Birch, A Bibliography on Dispersive Fourier Transform 91 (1998).
Spectrometry, 1963 to 1988, NPL Report DES93, National 69. J.E. Bertie, Y. Apelblat and C.D. Keefe, J. Mol. Struct.,
Physical Laboratory, Teddington (1989). 550551, 135 (2000).
44. J. Fahrenfort, Spectrochim. Acta, 17, 698 (1961). 70. J.E. Bertie, R.N. Jones, Y. Apelblat and C.D. Keefe, Appl.
45. N.J. Harrick, Phys. Rev. Lett., 4, 224 (1960). Spectrosc., 48, 127 (1994).
46. J. Fahrenfort and W.M. Visser, Spectrochim. Acta, 18, 1103 71. J.E. Bertie, Z. Lan, R.N. Jones and Y. Apelblat, Appl. Spec-
(1962). trosc., 49, 840 (1995).
47. A.C. Gilby, J. Burr, Jr and B. Crawford, Jr, J. Phys. Chem., 72. C.D. Keefe and J. Pittman, Appl. Spectrosc., 52, 1062 (1998).
70, 1520 (1966). 73. R.P. Young and R.N. Jones, Can. J. Chem., 47, 3463 (1969).
48. G.M. Irons and H.W. Thompson, Proc. R. Soc., 298A, 160 74. J.E. Bertie and Y. Apelblat, Appl. Spectrosc., 50, 1039
(1967). (1996).
49. C.E. Favelukes, A.A. Clifford and B. Crawford, Jr, J. Phys. 75. W.L. Wolfe and G.J. Zissis, The Infrared Handbook, Office
Chem., 72, 962 (1968). of Naval Research, US Dept of the Navy, Washington, DC,
50. T. Fujiyama and B. Crawford, Jr, J. Phys. Chem., 72, 2174 7-65 (1978).
(1968). 76. A.A. Clifford and B. Crawford, Jr, J. Phys. Chem., 70, 1536
51. T. Fujiyama and B. Crawford, Jr, J. Phys. Chem., 73, 4040 (1966).
(1969). 77. J.E. Bertie, S.L. Zhang and C.D. Keefe, J. Mol. Struct., 324,
52. E. Remsberg, D. Lavery and B. Crawford, Jr, J. Chem. Eng. 157 (1994).
Data, 19, 263 (1974). 78. J.E. Bertie and K.H. Michaelian, J. Chem. Phys., 109, 6764
53. V.M. Zolotarev, Sov. J. Opt. Technol., 43, 494 (1976). (1998).
54. D.C. Cameron, T.G. Goplen and R.N. Jones, Appl. Spec- 79. H. Torii and M. Tasumi, J. Phys. Chem., B102, 315 (1998).
trosc., 34, 657 (1980). 80. J.E. Bertie and S.L. Zhang, J. Mol. Struct., 413, 333 (1997).
55. J.E. Bertie, C.D. Keefe and R.N. Jones, Tables of Intensities 81. J.E. Bertie and C.D. Keefe, J. Chem. Phys., 101, 4610
for the Calibration of Infrared Spectroscopic Measurements (1994).
in the Liquid Phase, Blackwell Science Ltd, Oxford (1995). 82. C.D. Keefe, L.A. Donovan and S.D. Fleet, J. Phys. Chem.,
56. J.E. Bertie and H.H. Eysel, Appl. Spectrosc., 39, 392 (1985). A103, 6420 (1999).
Bandwidths
James J. Turner
University of Nottingham, Nottingham, UK
1 INTRODUCTION of the idea of dephasing, described in Section 1.4.

Most theories of band shapes assume that the band is
1.1 General homogeneous, but the possibility of inhomogeneous
broadening is extremely important, and in Section 1.5 the
In vibrational spectroscopy both Raman and infrared Kubo model of describing the difference between them is
(IR) attention is traditionally concentrated on the position outlined.
and intensity of spectral bands. As this Handbook testifies, In Section 2 we consider in detail the three main contri-
enormous chemical insight can be obtained from an under- butions to the homogeneous bandwidths: 2.1, vibrational
standing of these features. In condensed phases, the shapes relaxation; 2.2, pure dephasing, both direct and in-
of individual bands are often ignored, except for complaints direct; 2.3, rotation. We then return in Section 3 to a
about broad bands hiding important information! However discussion of how to distinguish, experimentally, between
the band shape is an important spectral feature, since the homogeneous and inhomogeneous broadening; Section 3.1
band shape contains information about the dynamics of the describes hole-burning experiments, and Section 3.2 dis-
system. cusses at some length the fairly recent sophisticated vibra-
There has been an immense amount of work on the the- tional echo experiments which, for the first time, allow a
oretical treatment of band shapes, but this article adopts clear distinction to be made.
a fairly simple approach, concentrating on what kind of The application of dynamic NMR (nuclear magnetic res-
factors determine the shapes of IR and Raman vibra- onance) to the rates of chemical processes was established
tional bands in solution, ignoring small molecules where in the early days of chemical NMR. It is often asked
some remnant of the gas phase rotational structure is often whether or not vibrational spectroscopy could be used in a
retained. We shall also mostly ignore intermolecular cou- similar manner, albeit on much faster timescales. Section 4
pling between identical molecules, which can complicate looks at this question in some detail. In Section 4.1 we
matters in pure liquids and concentrated solutions. For sim- have a reminder of the fundamental Bloch equations and
plicity we shall also not describe the method of moments. their application to chemical exchange via NMR. This is
Some other articles discuss some of the issues in more followed in Section 4.2 by a description of the modifica-
detail. tion of these equations for vibrational spectroscopy, and
In Section 1.2 it is shown how the band shape and the possible application to very fast exchange. Since it has
dynamics are related by the Fourier transform relation- been argued, convincingly, that there are serious problems
ship between frequency and time. In Section 1.3 correlation in the application of the Bloch equations to vibrational spec-
functions, introduced into vibrational spectroscopy by Gor- troscopy, in Sections 4.2.1 and 4.2.2 we look at situations
don, are briefly described; the functions provide a powerful where Bloch equations are likely to work. In Section 4.3 we
method of understanding how interactions in solution con- consider pseudocollapse, where apparently convincing
tribute to the bandwidth. This leads to the introduction coalescence, equivalent to NMR, can be misleading. Finally
Section 4.4 looks in detail at examples where there is defi-
John Wiley & Sons Ltd, 2002. nite evidence of exchange effects on vibrational spectra, and
tries to answer the question Under what circumstances do

simple Bloch equations give a sensible answer to very fast
1
chemical exchange?
e|t |/T2
F (t )
1.2 Fourier transforms
The general mathematical relationships for Fourier trans-

formation (FT) are equations (1) and (2):1 3
C1
0
p D Fxe2pipx dx 1 (a) t 0 +t
1
C1
Fx D pe2pipx dp 2
1
2T2
In these relationships, p and x must be conjugate vari-
ables; the product px must, of course, be dimensionless.
Since, in spectroscopy, we are concerned with the conju- 1/T2
()
gate variables frequency and time, the FT relationships are
equations (3) and (4):
C1 C1
2pint
n D Fte dt Ft D ne2pint dn
1 1
3 0
C1 C1
(b) 0 +
1 i!t i!t
I! D Fte dt Ft D I!e d! jtj/T2
Figure 1. (a) Plot of Ft D e . (b) Plot of the Fourier trans-
2p 1 1
4 form [n] of Ft a Lorentzian of FWHM D 1/pT2 and
maximum 2T2 .
where n is frequency in cycles s1 , ! is in radians s1 , and
t is time in seconds; in this article we shall use n or ! as
appropriate. Thus the spectrum (i.e. plot of n against n, determines the dynamics? To quote Tokmakoff and Fayer,4
or I! against !), is the FT of a time-dependent function Vibrational line shapes in condensed phases contain all of
Ft, and vice versa. Consider a simple example; if Ft the details of the interactions of a normal mode with its
has an exponential form, which is quite common, such as environment. The best way of describing this interaction
equation (5): is via the correlation function approach, introduced into
Ft D ejtj/T2 5 vibrational spectroscopy in a series of classic papers by
where jtj represents the modulus of t, and T2 is chosen for Gordon.5
reasons that will be clear shortly, then the FT of Ft is
given by equation (6):
C1 1.3 Correlation functions
n D ejtj/T2 e2pint dt
1 Correlation functions6 are usually defined in the time
2T2 domain as an ensemble average of the product of two time-
D 6 dependent quantities evaluated at different times [equa-
2pT2 n2 C 1
tion (7)]:
which describes a spectrum whose shape is a Lorentzian Ct D hAt0 Bt0 C ti 7
of FWHM (full width at half maximum) intensity equal to
1/pT2 cycles s1 (D 2/T2 radians s1 ), and the intensity at where t0 is the time at the start of the measurement, and
the maximum (n D 0) is 2T2 . This is illustrated in Figure 1. the brackets imply the ensemble average (i.e. over all
Thus if T2 is 1 ps, the Lorentzian band has a FWHM of molecules). However we are usually mostly interested in
3.2 1011 s1 10 cm1 . (Some of the subtle points in FT seeing how the value of a single property is correlated with
relationships are considered in Appendix 1.) time [equation (8)]:
Thus the shape of a spectral band is determined by
the time-dependent dynamics of the system. But what Ct D hAt0 At0 C ti 8
Bandwidths 3
which is the autocorrelation function. It is usual to assume The first term describes the orientational properties of the
we are starting at time t0 D 0, so equation (8) reduces to dipole moment, and we can define a rotational correlation
equation (9): function, C1r [equation (17)] (the reason for choosing C1r ,
Ct D hA0Ati 9 rather than simply Cr , will be clear shortly):
which is often normalized by dividing by [A0]2 . C1r t D hu0 uti 17
The normalized correlation time is defined by equa-
Suppose the dipole moment of an individual molecule has
tion (10): 1 a certain direction at time t D 0. Then after a long time and
tc D Ct dt 10 many collisions, it is equally likely to be pointing in any
0
direction, and hence the value of C1r t will tend to zero
To quote Gordon Qualitatively, a correlation function as t tends to infinity.
describes how long some given property of a system persists The second term is a mixture of vibrational and orien-
until it is averaged out by the microscopic motion of the tational effects; we can define it as Cv,1r . If the coupling
molecules in the system. between them is small, then the contributions can be aver-
The relationship between the spectrum and the correla- aged separately [equation (18)], i.e.
tion function is given by equation (11):
C1 Cv,1r t D hx0xtu0 uti D hx0xtihu0 uti
n D Cte2pint dt 11 D Cv tC1r t 18
1
i.e. in equation (3) Ft becomes Ct. where hx0xti is defined as Cv t.
For IR spectroscopy of isotropic samples which for The spectrum is given by the FT of Cm t [equa-
this article means liquids, solutions and powders Gor- tion (19)], i.e.
don showed that the autocorrelation function is given by n D FT[Cm t] D FT[C1r t] C FT[Cv t C1r t]
equation (12): 19
The first part gives the pure rotation spectrum [equa-
Cm t D hm0 mti 12
tion (20)]:
where m is the transition dipole moment, and hence [equa- 1r n D FT[C1r t] 20
tion (13)]: which, if C1r t decays exponentially, will be a Lorentzian.
C1 The second gives the vibrational part modified by the
n D hm0 mtie2pint dt 13 rotation, i.e. v,1r ; however, the FT of a product of two
1
functions is the convolution of the FTs of each component
For the simple case of a diatomic molecule, following [equation (21), see Appendix 2], i.e.
Bulanin et al.,7 we can expand the dipole moment as a
series in the vibrational coordinate x [equation (14)]: v,1r n D FT[Cv t C1r t] D v n 1r n 21

dm where the convolution is represented by the asterisk (*). If
mt D m0 t C xt C 14 each contribution, Cv t and C1r t, decays exponentially,
dx 0
then the resulting band shape is a Lorentzian with overall
where m0 is the permanent dipole moment, and the vector FWHM equal to the sum of FWHM of each component
designation (m0 ) arises because of its orientation. It is (see Appendix 2). The FT of the overall Lorentzian band
convenient to introduce the unit vector ut along the (v,1r ) will have an exponential decay with a time constant
direction of the dipole moment, and rewrite the equation we shall call T2 [equation (22)]:
as equation (15):
Cv,1r t D Cv t C1r t D Aet/T2 22
dm
mt D m0 tut C xtut C 15
dx 0 and the FWHM is 1/pT2 (see Section 1.1).
For polyatomic molecules the treatment is similar except
The correlation function is thus [equation (16)]: that dm/dx0 x is replaced by dm/dQ0 Q, where Q is the
normal coordinate for the mode giving rise to the IR band.
Cm t D hm0 mti D hm20 ihu0 uti
2 We have written the above for IR spectroscopy. Of course
dm there are equivalent correlation functions for Raman spec-
C hx0xtu0 uti C 16
dx 0 troscopy, which will involve, instead of dipole moments,
polarizability terms.5,10 The primary difference from IR is where D 1 for IR and D 2 for Raman. If the molecule is
that there are both isotropic and anisotropic contributions spherically symmetric, Dor will be independent of direction,
to the Raman scattering. The isotropic part [isot n] is the and the contribution to the vibrational line width from the
vibrational contribution only [equation (23)], i.e. orientation, or , is given by equation (31):
isot n D FT[Cv t] 23 1

or D
ptor
which means that molecular rotation does not make any 2Dor 6Dor
contribution to the vibrational bandwidth in the isotropic D for IR and for Raman 31
p p
spectrum. (This assumes there is no intermolecular coupling
between identical molecules.) As pointed out by Bulanin,8 That is, 1r n corresponds to a band of FWHM 2Dor /p,
a striking example of this is provided by the vibrational and 2r n corresponds to a band of FWHM of 6Dor /p.
fundamental of liquid N2 with FWHM of 0.067 cm1 .9
The anisotropic part (anis n) is given by equation (24):
1.4 Dephasing
anis n D FT[C2r t] C FT[Cv t C2r t] 24
If a spectral band is a Lorentzian with FWHM D 1/pT2 ,
which is rather similar to equation (19) except that C1r has then its FT is an exponential with a decay constant of
been replaced by C2r . Hence the spectrum is described by T2 (Section 1.2). T2 is called the dephasing time; it is
equation (25): instructive to see where this term comes from. We shall
concentrate on a purely classical picture; equivalent quan-
anis n D 2r n C v 2r n 25 tum mechanical pictures are less easy to grasp intuitively.
(A pictorial model11 and a more mathematical treatment12
Thus there is a pure rotational part and a vibrational
have been published.)
part, as before. The isotropic and anisotropic parts can be
Returning to the diatomic molecule of Section 1.3, and
obtained experimentally. If polarization perpendicular to the
ignoring the rotational effects on the correlation func-
scattering plane is called V, and in the plane, H, then it can
tion, we can focus on the purely vibrational part, Cv t D
be shown [equations (26) and (27)] that
hx0xti. For a diatomic harmonic oscillator of frequency
VV n D isot n C 43 anis n 26 n0 and amplitude A [equation (32)]:
VH n D HV n D HH n D anis n 27 xt D A cos2pn0 t C f 32
Rearranging equations (26) and (27) yields equations (28) where f represents the phase. If we imagine instant exci-
and (29): tation of many oscillators, all with identical amplitudes,
frequencies and phases (for the moment we shall assume
isot n D VV n 43 VH n D v n 28 f D 0) and further assume that nothing affects these three
anis n D VH n D v n 2r n 29 parameters, then the correlation function will be given by
equation (33):
(ignoring the pure rotation contribution).
hx0xti D hA cos0 A cos2pn0 ti
Thus from the Raman scattering it is possible to obtain
the vibrational spectrum without any rotational contribu- D NA2 cos2pn0 t 33
tion; the rotational contribution to the vibrational spectrum
can be obtained (without measuring the pure Raman rota- where N is the number of excited molecules. Thus the cor-
tional spectrum) by first obtaining v n from isot and relation function is a non-decaying oscillation. The FT will
then deconvoluting this from anis . Bartoli and Litovitz give an infinitely narrow line at n0 with integrated intensity
exploited this in elegant early papers.10 proportional to the number of excited molecules. Now we
It is interesting to see how the rotational contributions ask what will be the effect of changes in time with any
will differ in IR absorbance and in Raman scattering. In of the three parameters. If interaction with the bath of sol-
the simplest case of small step diffusion, the rotational vent molecules causes changes in the phases of individual
correlation time, tor , is related to the orientational diffusion oscillators (i.e. in f), then interference between the oscilla-
constant, Dor , by equation (30): tors will cause the correlation to decay, eventually to zero.
This will cause the FT to give rise to a band with a width
tor D [ C 1Dor ]1 30 dependent on the rate at which the phase coherence is lost,
Bandwidths 5
and with a shape which depends on the form of the decay. in other words tc is the characteristic time that frequency
Similarly if the bath causes changes in the frequencies of fluctuations persist. The Kubo parameter, k, is defined by
the individual oscillators, the phase coherence is lost and equation (39):
the correlation function will decay, giving a contribution to k D tc 39
the band broadening. The contribution from frequency and
phase change is called pure dephasing (T2 ).12 and we are particularly interested in situations where k 1,
If an individual molecule suddenly loses its excitation and k 1. If k 1, is much greater than 1/tc , i.e.
(i.e. A goes to zero) then its contribution to hx0xti dis- the spread of perturbed frequencies is much greater than
appears. If many molecules do this at some average time T1 the inverse of the characteristic time that the fluctuations
after the excitation, then they cause the correlation function persist. This is slow modulation, and the intensity distri-
to decay with a time that depends on T1 , and hence also bution reflects directly the distribution of the modulation;
contribute to the broadening. Assuming exponential decays thus the spectrum consists of the overlapping of the static
in each case, and still ignoring rotation, then more detailed distribution of frequencies. The spectral band will, usually,
analysis shows that the contribution to the Lorentzian band- have a Gaussian profile and the FWHM will be . This
width () is given by equation (34): is inhomogeneous broadening.
If k 1, tc becomes very small and any modulation !
1 1 1 hardly lasts for any significant time, so the fluctuation is
D D C 34
pT2 2pT1 pT2 smoothed out and the spectral band becomes sharp around
the average frequency !0 ; this is motional narrowing.
We return to this in Section 2. In the limit the band approaches a Lorentzian profile with
However, before looking at such processes in detail, there FWHM of 2 tc . This is homogeneous broadening. The
is an important question: is the spectral band under investi- width in the homogeneous case is much less than for the
gation broadened homogeneously or inhomogeneously? inhomogeneous case since 2 tc D k .
At intermediate values of k it is not so easy to distin-
guish between homogeneous and inhomogeneous broaden-
1.5 Homogeneous and inhomogeneous ing. Also since tc will depend on temperature, for the same
broadening system homogeneous broadening may change to inhomo-
geneous with temperature and vice versa.
To appreciate the distinction between homogeneous and We return to the very important question of homoge-
inhomogeneous broadening, it is useful to describe it in the neous and inhomogeneous broadening shortly, but for the
manner first introduced by Kubo;13 it is also well described moment, assuming that a spectral band is homogeneous,
by Wood and Strauss.14 (We use ! rather than n, partly we consider in more detail the different contributions to its
because this was how Kubo described it.) Suppose there breadth.
is an oscillator whose frequency is randomly modulated
because of some dephasing interactions. We can write
equation (35): 2 CONTRIBUTIONS TO
!t D !0 C !t 35 HOMOGENEOUS LINEWIDTH
where !0 , the static part, is selected so that the mean of !, In the Markovian limit, in which quantities are described
the fluctuating part, is zero. by independently relaxing correlation functions, the total
The modulation amplitude , is defined by equation (36): correlation function decays exponentially at a rate of 1/T2 ,
where T2 is the dephasing time. (A Markovian process
2 D h[!t]2 i 36
is one where the probabilities of any events in a system
i.e. it is the root mean square of the frequency fluctuation. depend on the state of the system at that time, and not on
The correlation function of the modulation is defined by its previous history.) The contribution to the total bandwidth
equation (37): () by the various terms is given by equation (40):
1 1 1
1 D D C C or 40
Ct D h!t!0i 37 pT2 2pT1 pT2
2
In this equation, T1 is the excited vibrational state relaxation
and hence the correlation time by equation (38):
time, T2 is the pure dephasing time, and these two terms
1
make up the vibrational contribution to the bandwidth (i.e.
tc D Ct dt 38 from Cv t). or is the contribution to the bandwidth from
0
orientational relaxation effects [i.e. from C1r t or C2r t].

0.75 J IR pump A 3
The overall decay time T2 is the homogeneous dephasing 0.4
time (see Section 1.4).
0.2
2.1 Vibrational relaxation time T1
The lifetime of a molecule in an excited vibrational state

0.0
v D 1, relaxing to v D 0, is defined as T1 . Since the value Time delay (ps)
of the vibrational quantum number changes, this is an 4.0 ps
inelastic process. T1 processes are caused by anharmonic 8.4 ps
0.2
coupling of the vibrational mode to the bath. The bath
16.8 ps
includes other vibrational modes of the solute molecule and
50.3 ps
of the solvent, and the low frequency continuum of inter-
molecular solvent modes. If the route is via intramolecular 0.4 100.5 ps
vibrational relaxation, we have IVR; via external vibrational 167.5 ps
relaxation to the solvent we have EVR.15,16 The relaxation
268 ps
can be exponential, but can be highly non-exponential; non- 0.6
exponentiality is very striking if the up-pumping also leads
to population of levels higher than v D 1; we shall ignore 1990 1980 1970 1960 1950 1940
most of such complications. Under the simplest relaxation,
Probe wavenumbers /cm 1
with a lifetime for the excited vibrational state (v D 1) of
Figure 2. Spectral evolution of the W(CO)6 transient spectra
T1 , the contribution to the bandwidth is the familiar term,
with 1983 cm1 pump frequency showing bleach recovery at
1/2pT1 . The best way of obtaining the value of this term is 1982 cm1 and transient absorption kinetics for v D 1 and v D 2
to measure it directly, by either of two methods. In a one- at 1967 and 1950 cm1 , respectively. [Reproduced from Arrivo
color experiment, an IR laser pumps the molecule from et al. (1995)18 by permission of Elsevier Science.]
v D 0 to v D 1; this bleaches the v D 0 to v D 1 absorp-
tion, and the decay is followed by monitoring the recovery of 1.6 1010 cycles s1 to 3.2 108 cycles s1 (0.5 to
of this bleach. In a two-color experiment, with separate 0.01 cm1 ). Since metal carbonyl bandwidths under these
pump and probe IR lasers, it is possible to monitor both the conditions vary from, say 5 to 20 cm1 , the bandwidth
bleach in the v D 0 ! v D 1 absorption and the decay in contribution from relaxation is negligible. However there
the v D 1 ! v D 2 absorption. will be conditions under which the contribution is highly
The range in T1 values covers some 14 orders of significant; we return to this point later. Complications can
magnitude.16 In liquid N2 , it is 1 s,17 making a contri- also arise from orientational contributions; we return to this
bution to the Raman line width of only 5 1012 cm1 , point in Section 2.3.
which was why we could use liquid N2 as an example of
a very narrow band (0.067 cm1 ), uncomplicated by any
rotational contribution, when considering isotropic Raman
scattering in Section 1.3. In strongly interacting systems, 2.2 Pure dephasing T2
such as ions in polar solvents, it can be 10 ps or less (10 ps
corresponds to a bandwidth of 0.5 cm1 ). Pure dephasing describes the adiabatic modulation of the
Some of the most interesting experiments have been vibrational energy levels of a transition caused by ther-
performed on transition metal carbonyls (e.g. Cr(CO)6 ), mal fluctuations of its environment. Measurement of this
particularly by Heilweil and colleagues.18,19 Figure 2 shows quantity provides detailed insight into the fast dynamics
experiments on the t1u n(CO) band of W(CO)6 in n-hexane. of its environment. Usually the dephasing is due to short
The absorption (v D 0 ! v D 1) at 1983 cm1 is bleached range, and nearly hard-core intermolecular interactions.15
by the IR pump laser, tuned to this frequency; this is shown One mode of intermolecular interaction can be via res-
to recover, while the transient absorption (v D 1 ! v D 2) onant energy transfer which can occur between identical
at 1967 cm1 and the weaker feature (v D 2 ! v D 3) at molecules close together in solution; this dephasing contri-
1952 cm1 decay. In general, for metal carbonyls in non- bution can be eliminated by diluting the solution (see Doge
polar solvents at room temperature, T1 values range from and Yarwood11 for a discussion of these points). More fre-
10 to 500 ps, thus making a contribution to the line width quently we are concerned with interaction of the vibrational
Bandwidths 7
mode with the thermalized reservoir of the bath of solvent specifically the appearance of the fundamental. There are
molecules. different ways of treating the interaction between modes A
Since, in general, these interactions do not change the and B, the details of which need not concern us. With some
energy level, they are described as elastic. In principle, assumptions, of which the two most important are that the
the way to estimate these effects is to devise a model coupling is weak (i.e. that ! !A ), and that the process
which will encapsulate the intermolecular interactions; in is Markovian, which means in this case23 that the proba-
other words calculate the appropriate correlation function. bilities of excitation and de-excitation of the exchanging
FT will then provide the spectral contribution from these mode depend only on the state of the system and not on
interactions. In practice, this can be an extremely difficult its previous history, the different approaches give the same
task. This is partly because of the difficulty of correctly answers. The position of the fundamental is shifted by an
setting up the model, and partly because to make the calcu- amount given by equation (41):
lations tractable, several assumptions have to be made. For
instance when considering induced contributions which !eEB /kT
! A D 41
make unallowed bands allowed there are cross terms in 1 C !2 t2
the correlation function which are almost impossible to dis- where EB is the energy of mode B and t is the lifetime of
entangle. It is outside the scope of this article to consider mode B.
these effects in detail. So the apparent frequency of mode A, !0A , is given by
These types of interactions have been described as equation (42):
direct.20 But there is another type of interaction, described !0A D !A C !A 42
as indirect,20 in which the most important dephas-
ing mechanism involves interaction with a specific low- The shape of the fundamental band is given by equa-
frequency mode, usually intramolecular, but not necessarily tion (43):
so. We raise this here because it has important consequences I!0A
I! D 43
for a later section. The simplest model21 is illustrated 1 C ! !0A 2 T02
2
in Figure 3. (More sophisticated treatments are listed in
the references on page 2097 of Hsu and Skinner.22 ) The where [equation (44)]
high-frequency mode, A, has a frequency !A when the 1 1 !2 teEB /kT
low-frequency mode, B, is in the ground vibrational state 0 D C 44
T2 T2 1 C !2 t2
vB D 0, but, because of anharmonicity, has a frequency
!A C ! when mode B is in vB D 1, i.e. the hot-band in which T2 is the contribution to the dephasing in the
has a different frequency from the fundamental. If the fre- absence of the A/B exchange. Thus the fundamental band
quency of mode B, !B , is sufficiently low compared with is now a Lorentzian with FWHM equal to 1/pT02 cycles s1
kT, then both the fundamental and the weaker hot-band will or 2/T02 radians s1 . The position and width of the band are
be seen. However if interaction between mode B and the functions of temperature and lifetime of B. In particular
environment causes rapid excitation and de-excitation of both position and width display an apparent activation
this mode then this will have an effect on the spectrum, energy which is equal to EB .
nA nB
1 1
2.3 Rotational contribution 0or
1 0
In IR, this term usually arises from the physical rotation
of the dipoles. It can arise from more subtle effects and
it is probably better to define the term in the following
A A +
way: orientational relaxation is defined as any process
which causes the loss of angular correlation of an ensemble
0 1 of dipoles. For Raman scattering there is an analogous
0 0 B polarization contribution. The contribution from rotation
A B depends on how the rotation occurs.24 At the end of
Section 1.3, we considered how the rotational correlation
Figure 3. Illustration of anharmonic coupling between a high-
frequency mode A, and a low-frequency mode B. If the molecule time, tor , and the reorientational diffusion constant, Dor ,
is in Bs ground state, the frequency of A is !A ; if in the excited are related for the simple model of small step diffusion.
vibrational state of B, the frequency of A is !A C !. This led to the conclusion that for spherically symmetrical
molecules, the contribution to the bandwidth is given by obtained by the appropriate deconvolution. The general
equation (31): principles behind pumpprobe polarization spectroscopy
1 are given by Fleming.26
or D
ptor
2Dor 6Dor 3 SEPARATION OF HOMOGENEOUS
D for IR and for Raman 31
p p AND INHOMOGENEOUS
In practice the ratio is seldom as large as 3, but is usually BROADENING
>1. If the molecule is a symmetric top then for Dor there
will be Dparallel and Dperpendicular , so that the contribution will We now return to the problem of homogeneous versus
be different for different bands. With molecules of lower inhomogeneous band broadening, introduced in Section 1.5.
symmetry the situation becomes very complicated. Clarke24 This topic might seem esoteric, but it is often assumed
has discussed at length the various models of rotational dif- implicitly, if not explicitly, that in room temperature liquid
fusion, and their subsequent contributions to the bandwidth. phases, the vibrational spectra are dominated by homo-
This has the interesting corollary that where we assume geneous broadening. Since practically all theories about
homogeneous broadening and where T1 is relatively long, band shapes address only homogeneous broadening, com-
if the same mode has the same bandwidth in IR and Raman, parison with observed spectra become meaningless if the
the orientational contribution must be small, and hence the band has substantial inhomogeneous contribution. We now
bandwidth is determined by pure dephasing, T2 . consider two experimental approaches to the problem, hole-
Another approach to rotational diffusion is to use NMR burning and vibrational echo experiments.
or dielectric relaxation data to obtain values for Dor , and
then compare with vibrational data. For instance Wang and
Schwartz25 have studied 17 O relaxation data for W(CO)6 in 3.1 Hole-burning
several solvents to obtain rotational correlation times, tor ,
Some of the most striking examples of the proof of
via equation (45):
inhomogeneous broadening come from very low tempera-
12p2 ture matrix isolation experiments. Under these conditions tc
[T1Q 17 O]1 D c2 tor 45
125 is very long, hence k 1, and hole-burning experiments
where T1Q is the quadrupole relaxation rate, and c is the can be done. The principle of these experiments is to use
17
O quadrupole constant (in cycles s1 ). This gives tor an IR laser, tuned to a narrow frequency range, to burn a
and hence from equations (30) and (31) the orientational homogeneous hole in an inhomogeneous IR band, and
contributions to the bandwidths. For instance, from the then to record the overall spectrum showing the hole.
relaxation data, or for the t1u mode of W(CO)6 in CDCl3 This is elegantly illustrated in experiments by Gunthard
is calculated to be 0.6 cm1 ; since the IR bandwidth is and colleagues27,28 on the n17 band of 1,2-difluoroethane
18 cm1 , the conclusion is that for this mode of this trapped in solid Ar, Kr or N2 matrices. Figure 4 shows the
molecule in this solvent, the contribution to the line width is result of one experiment in Ar at 6 K. Using an IR diode
dominated by pure dephasing. (Note, this is again assuming laser for both spectroscopy and hole-burning, a very narrow
that the bandwidth is not dominated by inhomogeneous
1
broadening.) By contrast in n-hexane, or is 0.8 cm1 , out
of a total bandwidth of 4 cm1 ; we shall return to this
Transmittance
molecule later.
However, in principle, the most elegant method of obtain- 0.21 cm1
ing orientational relaxation is via laser pumpprobe experi- 0.006
ments, as outlined in Section 2.1. If the pump and probe
lasers are polarized parallel to each other, and the v D
1 ! v D 2 absorption is monitored by the probe pulse,
then the signal will decrease by two routes: decay from 0
970 980 990 1000
v D 1 ! v D 0; and also by rotation of those molecules,
Laser current (mA)
still in v D 1, whose transition moment turns so that the
molecules are blind to the probe polarization. If the pump Figure 4. IR laser diode spectrum of one component of the matrix
split n7 band of trans-CH2 FCH2 F in an Ar matrix at 6 K, taken
and probe pulses are at 54.7 (the magic angle ) to after 6 min irradiation at the arrowed frequency (1047.550 cm1 ).
each other, then the orientational contribution to the decay [Reproduced from Poliakoff and Turner (1980)29 by permission
disappears, and thus the orientational contribution can be of Academic Press.]
Bandwidths 9
homogeneous hole is burned in the inhomogeneous band, 1.8

itself rather narrow (0.2 cm1 ). [We need to enter a caveat
here. On some time scales, a molecule will sample all the
positions that make up the inhomogeneous envelope; this 1.6
effect is known as spectral diffusion.30 It will be fast in

liquids but in a matrix at 6 K it will be extremely slow. Thus
Absorbance
1.4
although in principle the hole may be somewhat wider than
the true homogeneous line width (see text below for another
subtlety), the experiment described here is almost certainly 1.2
reflecting the homogeneity.]
Experiments with polarized IR show that there is no free
1.0
or hindered rotation of the molecule in the matrix, and so
the homogeneous hole, which is a Lorentzian, is caused
by either, or both, vibrational relaxation (T1 term) or pure 0.8
dephasing (T2 term). The line width of the hole varies with 931.0 931.1 931.2 931.3
temperature: from 0.005 cm1 at 0 K (extrapolated value) cm1
to 0.05 cm1 at 20 K; above 20 K the hole disappears. The Figure 5. IR spectrum of n3 mode of SF6 in Xe crystal. Concen-
model assumes that the hole is caused by slight changes of tration Xe/SF6 D 108 ; temperature D 14 K; thickness of sample D
the rotational configuration of the molecule, i.e. a subtle site 1 cm. The circles are calculated points for a Lorentzian band with
change. The pure dephasing is modeled on an interaction FWHM D 0.028 cm1 , convoluted with the instrument lineshape.
[Reproduced from Swanson et al. (1986)33 by permission of Else-
between the molecule and the matrix; this will extrapolate to vier Science.]
zero at 0 K, so that the width at 0 K is determined by just the
relaxation term, implying a lifetime of 3 ns. The overall bandwidth, and its changes with temperature, derive from
inhomogeneous line width is attributed to slightly differ- pure dephasing, which is probably determined by bulk and
ent environments in the matrix. (There is a subtlety here: local phonon modes coupling with the n3 mode.
the width of the hole is actually twice the homogeneous With the development of picosecond IR lasers, it became
width because the resulting spectrum is the convolution of possible to probe hole-burning in fluid environments.
the homogeneous and inhomogeneous band shapes, in this These have yielded a great deal of information about
case of a Lorentzian with itself, resulting in a hole of twice the interaction in solution. We show just one recent
the width.27,28 ) intriguing example.35 Figures 6(a) and 6(b) show what
As pointed out by Cho and Strauss,31 IR hole-burning happens when dilute solutions of hydrogen-bonded
experiments have been carried out on isolated molecules, 1 : 1 pyrroletetrahydrofuran and 1 : 1 methanolpyridine
impurities doped in ionic crystals, and polymers; the hole- complexes in CCl4 are subjected to picosecond IR pulses.
burning is usually caused by a photophysical process. In Figure 6(a), even though the IR pulse is only 8 cm1
If the inhomogeneous broadening is caused by irregulari- wide, the whole of the NH stretching mode, centered
ties in the matrix, is there any way in which the environment at 3350 cm1 , is bleached. The Fourier transform infrared
can be made more uniform, and hence remove the inhomo- (FT-IR) spectrum of the complex best fits a Lorentzian
geneity? Based on earlier experiments by Holland et al.,32 (Figure 6a), so the conclusion is that the whole band is
which showed that SF6 is slightly soluble in solid Xe, Swan- homogeneously broadened. A new, short-lived band, the
son et al.33 looked at the high resolution IR spectrum of the beginning of which can be seen at the lower frequency
n3 mode of SF6 in solid Xe over the temperature range 14 K end of the spectrum, is generated. The bleaching is due to
to 160 K (see Figure 5). Normally SF6 in frozen gas matri- the excitation of v D 0 ! v D 1, and the lower frequency
ces shows the effects of a great deal of inhomogeneous band is due to the v D 1 ! v D 2 absorption, which is
broadening and site splittings.34 However in the later at lower frequency than the 0 ! 1 absorption because
experiments,33 in which the sample was obtained by slowly of anharmonicity. In contrast, in Figure 6(b) only the
freezing a solution in liquid Xe to obtain a single crystal, central part of the band is bleached, implying that the
the band is Lorentzian and homogeneously broadened. The overall band is inhomogeneous, and this is also implied
FWHM changes from 0.028 cm1 at 14 K to 0.5 cm1 at by the fact that the FT-IR shows a better fit to a
160 K. Rotation is frozen out so there is no contribution Gaussian. For the latter case, the authors attribute the hole-
to the bandwidth from or ; time-resolved IR measurements burning to the presence in solution of long-lived (>2 ps)
show that the n3 energy relaxation is slow (T1 > 108 s) H-bonded complex geometries, giving rise to different
contributing only 0.0005 cm1 to the bandwidth. Thus the absorption frequencies, and hence an inhomogeneous
0.10 largely academic since there was no way of demon-

Transient data
strating conclusively what are the individual contribu-
0.08 FT-IR data tions of homogeneous and inhomogeneous broadening.
Recent beautiful experiments, particularly by Fayer and
Lorentzian fit
0.06 colleagues,4,15,30,36 38 and by Berg and colleagues,39 43
Gaussian fit based on the theory by Loring and Mukamel,44 show how
In(T /T0)
0.04 this distinction can be made. In this very important paper,44

it was shown that previous attempts to separate homoge-
0.02 neous and inhomogeneous contributions by Raman pulse
experiments in fact do not do so. This only became obvi-
0.00 ous when the process was formulated in terms of correlation
functions. In both IR and Raman spectroscopy the best way
0.02 to separate these effects is via vibrational echo exper-
3200 3300 3400 3500 3600 iments, which are analogous to the spin echo technique
(a) Wavenumber /cm1 in magnetic resonance first introduced in 195045 and later
extended to the optical regime.46,47 Both the theory and
Transient data experiments for these methods are complex, although the
0.05
IR experiment is intuitively more obvious, and below is
FT-IR data
a quote taken from a review by Tokmakoff and Fayer,4
0.04
Gaussian fit where the authors give an analogy to IR echo experi-
ments, following on from Hahn (reproduced from Tok-
0.03 Lorentzian fit
makoff and Fayer4 by permission of the American Chemical
In(T /T0)
Society):
0.02
To obtain a physical feel for the manner in which the echo

0.01 experiment can reveal homogeneous fluctuations in spite
of a broad inhomogeneous spread of transition frequencies,
0.00 consider the following foot race. Initially all the runners
are lined up at the starting line. At t D 0 the starting gun
3000 3100 3200 3300 3400 3500 3600 3700 (analogous to the first IR pulse) is fired, and the runners take
(b) Wavenumber /cm1 off down the track. After running for some time, the faster
runners have pulled out in front, and the slower runners
Figure 6. (a) Transient IR spectrum after IR excitation of the are somewhat behind. The runners are no longer in a line
NH-stretching mode absorption (at 3345 cm1 ) of dilute 1 : 1 because of the inhomogeneity in their speeds. At time t,
hydrogen-bonded pyrroletetrahydrofuran complexes in CCl4 at the gun is fired again (analogous to the second laser pulse),
295 K. (b) Transient IR spectrum after IR excitation of the and everyone turns round and runs back toward the starting
OH-stretching mode absorption (at 3350 cm1 ) of dilute 1 : 1 line. If each runner maintains a constant speed, out and
hydrogen-bonded methanolpyridine complexes in CCl4 at 295 K. back, then all the runners will cross the starting line exactly
[Reproduced from Arrivo et al. (1999)35 by permission of Gordon
in line again. At the starting line, the group is rephased; the
and Breach.]
inhomogeneity in speeds has been nullified. If the runners
do not run at exactly constant speeds, but each runner has
spread of frequencies. The authors also suggest that a some fluctuation in speed about his average (homogeneous
deeper understanding would require femtosecond IR pulses fluctuations), then the runners will not cross the starting
or photon echo experiments, a topic which is addressed line exactly in a line. There will not be perfect rephasing.
in Section 3.2. In this kind of experiment the burning is A snapshot of the group as it crosses the starting line
directly caused by the v D 0 ! v D 1 excitation; of course will reveal the small fluctuations in the runners speeds, in
no distinction can be made between the various possible spite of the large inhomogeneous distribution of speeds. In
the same manner, the vibrational echo experiment reveals
contributions to the homogeneous bandwidth. We now turn
the fluctuations in the vibrational transition frequency in
to experiments which can do precisely this. spite of the large inhomogeneous distribution of vibrational
energies.
3.2 Vibrational echo experiments
With this background we can now describe the IR
Until recently this problem of distinguishing between the echo experiment, again following closely the description
various contributions to homogeneous broadening was in Tokmakoff and Fayer.4
Bandwidths 11
3.2.1 IR echo measurements fluctuate. Thus, at 2t there is not perfect rephasing. As

t is increased, the fluctuations produce increasingly large
Figure 7 schematically represents the experiment. A accumulated phase errors among the microscopic dipoles,
picosecond IR pulse puts each solute molecules vibration and the size of the echo is reduced. A measurement of
into a superposition state, which is a mixture of the v D 0 the echo intensity versus t, the delay time between pulses,
and v D 1 vibrational levels. Each vibrational superposition is called an echo decay curve. Thus the echo delay is
has a microscopic electric dipole associated with it. This related to the fluctuations in the vibrational frequencies,
dipole oscillates at the vibrational transition frequency. not the inhomogeneous spread in frequencies. The Fourier
Immediately after this pulse, all the microscopic dipoles transform of the echo decay is the homogeneous line shape.
oscillate in phase. Because there is an inhomogeneous [Note, as long as the homogeneous contribution to
distribution of vibrational transition frequencies, the the bandwidth is much smaller than the inhomogeneous
individual dipoles oscillate with some distribution of contribution, the echo decay is given by equation (46):
frequencies. Thus the initial phase relationship is very
4t
rapidly lost. This is referred to as the free induction It D I0 exp 46
T2
decay; Fourier transform of this decay will give the overall
band shape, including inhomogeneity. After a time, t, a i.e. the decay rate is four times the homogeneous dephas-
second pulse traveling along a path making an angle of q ing time, because the echo rephases at 2t and the inten-
with that of the first pulse, passes through the sample. This sity decays twice as fast as the depolarization. As the
second pulse changes the phase factors of each vibrational homogeneous contribution increases, compared with the
superposition state in a manner that initiates a rephasing inhomogeneous contribution, the rephasing occurs at times
process. At time t after the second pulse, the sample between t and 2t. In the extreme of a completely homoge-
emits a third coherent pulse of light. The emitted pulse neous band, the echo decay has a decay constant of 2/T2 ,
propagates along a path that makes an angle 2q with the rather than 4/T2 .]
path of the first pulse. This third pulse is the vibrational To provide an example36 where there is an enormous dif-
echo. It is generated when the ensemble of microscopic ference between homogeneous and inhomogeneous broad-
ening, Figure 8 shows the echo decay for the antisymmetric
dipoles is rephased at time 2t. The phased array of
CO stretching mode of Rh(CO)2 (acac) in DBP glass at
microscopic dipoles behaves as a macroscopic oscillating
3.4 K. The IR laser pulses were provided by the Stan-
dipole, which generates an IR pulse of light. A free
ford free electron laser, although recent experiments have
induction decay (inhomogeneous frequency distribution)
used a table-top set up.48 The decay is exponential, with
again destroys the phase relationships, so only a short pulse
a decay constant of 23.8 ps (so T2 D 95.2 ps), and FT
of light is generated. The rephasing at time 2t has removed
gives the homogeneous bandwidth of 0.11 cm1 , to be con-
the effects of the inhomogeneous broadening. However,
trasted with the bandwidth measured by conventional IR
fluctuations due to coupling of the vibrational mode to
spectroscopy of 15 cm1 . Thus the band in the glass
the heat bath (solvent) cause the oscillation frequencies to
is massively inhomogeneously broadened. [The reason for
choosing metal carbonyls for this and related experiments
Sample
Infrared is that the n(CO) IR bands are extremely intense, usually
pulse 2 rather narrow, and well-understood, so that dilute solutions
can be used; this avoids complications due to energy trans-
Infrared fer between molecules.]
pulse 1
Echo (Note, if the pulse width exceeds the anharmonic shift
pulse
of the appropriate vibrational mode then the echo decay
displays beats, in which the inverse of the time separation
of the beats gives the anharmonicity.36 38 )
The most interesting experiments employing IR pho-
Pulse 1 Pulse 2 Echo pulse ton echoes have involved the triply degenerate t1u n(CO)
Figure 7. Schematic diagram of the IR vibrational echo experi- stretching mode of W(CO)6 in the three glass-forming
ment. Two picosecond IR pulses, tuned to the vibrational solvents, 2-methylpentane (2-MP), 2-methyltetrahydrofuran
transition frequency, enter the sample crossed at a small angle, (2-MTH) and DBP, over a wide temperature range. In the
q. Because of wave vector matching, the echo pulse emerges
three glasses, below the glass transition temperatures, the
from the sample in a unique direction, 2 q. Pulses 1 and 2 are
separated by time t. The echo is formed at time 2t after pulse 1. absorption spectrum bands are massively inhomogeneously
[Reproduced from Tokmakoff and Fayer (1995)4 by permission broadened and the overall total homogeneous bandwidth
of the American Chemical Society.] (1/pT2 ) is well described by a (temperature)2 power law.
350
1
300
250
0.1
200
Linewidth (GHz)
150
Signal (a.u.)
0.01
0 40 80 120
Time (ps) 100
50
0
0 50 100 150 200 250 300
Temperature (K)
Figure 9. ( ) Temperature dependence of the conventional
absorption line width of the t1u CO stretching mode of W(CO)6
in 2-MP. () Homogeneous bandwidths (since the homogeneous
bandwidth is much less than the inhomogeneous bandwidth,
40 20 0 20 40 60 80 100 120 140 It D I0 exp4t/T2 is applied; see text). () Homogeneous
Time (ps) bandwidths (since homogeneous inhomogeneous bandwidth,
Figure 8. Vibrational echo decay data of the antisymmetric CO It D I0 exp2t/T2 is applied); this shows that above 250 K
stretch of Rh(CO)2 acac (where acac represents acetylacetonato) in the band is entirely homogeneously broadened. ( ) (200 K) Decay
DBP (dibutyl phthalate) at 3.4 K. The decay is single exponential. function between the two extremes; the band is an equal mixture
The inset is a semilog plot of the data and fit; see text. (a.u. D of homogeneous and inhomogeneous broadening. [Reproduced
arbitrary unit.) [Reproduced from Rector and Fayer (1998)36 by from Tokmakoff and Fayer (1995)4 by permission of the American
permission of the American Institute of Physics.] Chemical Society.]
total homogeneous bandwidth (0.05 cm1 ) is dominated

At room temperature, the width of the band in 2-MP is
by vibrational relaxation. The absorption spectrum at 10 K
entirely due to homogeneous broadening, whereas in DBP
has a bandwidth of 10 cm1 , i.e. massive inhomoge-
it is still extensively inhomogeneously broadened. In fact
neous broadening. At room temperature, the homogeneous
these were the first experiments to show conclusively intrin-
bandwidth, which accounts for the whole bandwidth, is
sic inhomogeneous broadening for a room temperature dominated by pure dephasing. The orientational contribu-
liquid. Extensive studies were carried out on W(CO)6 in 2- tion is always modest, but is most significant at 100 K.
MP. Figure 9 compares the absorption bandwidth and the A surprising feature is that the orientation contribution fol-
homogeneous bandwidth over a wide temperature range. lows a straight line, even in the glass. This is because,
To separate the different contributions to the bandwidth, for this molecule, the effect comes not from ordinary rota-
T1 and tor need to be determined. This was done by tion, which would be very small or zero in the glass, but
pumpprobe experiments, as described in Section 2.3.37 from a rotation of the dipole through the three degenerate
With the pump and probe pulses polarized parallel, the motions of the t1u mode. Intriguingly the rotational contri-
decay curves were bi-exponential due to contributions from bution at room temperature (0.7 cm1 ) is very similar to
both vibrational relaxation and orientation; the latter effect the value obtained (0.8 cm1 ) in the quadrupole relaxation
was eliminated by setting the polarizations of pump and experiments25 in room temperature in n-hexane.
probe at the magic angle. Thus the contributions from Two further points are noteworthy. The slow decrease in
vibrational relaxation and reorientation can be subtracted the T1 contribution with increase in temperature is counter-
from the total homogeneous bandwidth to give the pure intuitive, but explanations have been offered. The pure
dephasing contribution. Figure 10 shows the fascinating dephasing term behaves in a manner at odds with prac-
results (note the loglog scales). At the lowest tempera- tically all the theoretical models, which predict much more
tures, the contribution from pure dephasing and orientation sensitivity to viscosity than the experiments determine.
become negligible (both must be zero at 0 K), and the Discussion of this is outside the scope of this article.
Bandwidths 13
same solvent.36 The difference may lie in the formally triple

100 Total linewidth degeneracy of the t1u mode of W(CO)6 , but there is clearly
Pure dephasing much still to be learned in this area.
Homogeneous linewidth contributions (GHz)
Orientational A French group50 has also used a free electron laser

30 Lifetime to examine pumpprobe and IR echoes for W(CO)6 in
2-MP, n-hexane, and CCl4 at room temperature. The
results are in qualitative agreement with the experiments
10
of Heilweil and colleagues18,19 and with Fayer and co-
workers4,36 38 . In each case the t1u n(CO) stretching mode is
homogeneously broadened. In CCl4 , T2 1 ps, equivalent
3 to a homogeneous bandwidth of 10 cm1 , compared with
a measured absorption bandwidth of 10.6 cm1 , perhaps a
somewhat surprising result. In 2-MP and n-hexane, T1
1
6070 ps, with a very short component of 24 ps, which
is assigned to the t1u ! eg phonon-induced population
transfer. Preliminary results in solids suggest, as expected,
0.3 a high degree of inhomogeneous broadening.
Tg It is worth adding that for small molecules, either as
liquids or in solution, the orientational contribution to the
0.1
10 20 30 50 100 200 300 bandwidth will be large, perhaps even as great as pure
Temperature (K) dephasing.43 This is partly why Raman studies, which,
Figure 10. Loglog plot of the different contributions to the under some circumstances, show no rotational effects, have
homogeneous width of the t1u CO mode of W(CO)6 in 2-MP. Tg been popular. We consider these next.
is the glass transition temperature. [Reproduced from Tokmakoff
and Fayer (1995)4 by permission of the American Chemical
Society.] 3.2.2 Raman echo experiments
The earliest Raman echo experiments,39 43,51 based on the
As mentioned above, some experiments have more proposal of Loring and Mukamel,44 employed a seventh
recently been carried out36 on the antisymmetric n(CO) order (three-dimensional) method. The vibration is initially
stretching mode of Rh(CO)2 (acac) in the highly viscous excited via stimulated Raman by a pair of pulses (at
glass-forming solvent DBP. Since the mode is not degener- frequencies !L and !S , where !L > !S ). Dephasing occurs
ate, there is no opportunity for the rotational effect observed for a time t1 , after which a second pair of pulses (also at !L
with W(CO)6 and so, in this solvent, it is reasonable to and !S ) reverses the phase of the coherent vibration. The
assume that the orientational contribution is zero, and this size of the coherence is monitored by coherent anti-Stokes
is confirmed by polarized pumpprobe experiments. Thus scattering from a fifth pulse at t2 !L . Phase-matching
pure dephasing is easily obtained by subtracting the T1 conditions apply [equation (47)]:
contribution from the total homogeneous bandwidth. One
interesting result is that the best explanation of the pure ke D kL3 C 2kL2 kS2 kL1 kS1 47
dephasing is that the high frequency n(CO) mode is anhar- where ke is the wave vector of the echo signal, kL1 , kL2 and
monically coupled to a low-frequency n(RhC) mode (see kL3 are the wave vectors of the higher frequency pulses and
Section 2.2). Thus above 20 K, the dephasing is expo- kS1 and kS2 are the wave vectors of the lower frequency
nentially activated with energy E 400 cm1 ; there is an pulses. Full details are given elsewhere.39 43,51 In this way
RhC mode at 404 cm1 . From Section 2.2, and assuming it was shown that the symmetric methyl stretch of liquid
! D 20 cm1 , and !B D 404 cm1 , then t 0.75 ps. This CH3 CN is homogeneously broadened, which contradicts
value of t is very short and perhaps is not related directly some earlier views and supports others.39 43 Similarly
to the simple bath-inspired excitation and de-excitation of the CN stretch of liquid benzonitrile is homogeneously
the RhC mode, but rather to the resonant transfer time broadened.51 In a 1 : 1 mixture of CH3 I and CDCl3 , the
for the excitation to move to a neighboring molecule; but symmetric methyl stretch of CH3 I is inhomogeneously
there are problems with this interpretation.49 Without going broadened;40 this, not surprisingly, is due to concentration
into detail, the variation in the pure dephasing contribu- fluctuations in the first solvation shell, and hence is not an
tion to the two carbonyls is quite different, even though inherent property of either pure compound.
the pure dephasing is for a CO antisymmetric vibration In these systems the depolarization ratios are small so that
stretching mode of both molecules at 2000 cm1 in the the Raman cross-section is isotropic, which in turn means
that there is no rotational contribution to the line shape extremely narrow (1 cycles s1 3 1010 cm1 !). If
(Section 1.3). Thus the homogeneous dephasing time (T2 ) the rate of exchange A ! B D kA (lifetime D tA D 1/kA ),
is given approximately by equation (48): and B ! A D kB (lifetime D tB D 1/kB ), then the Bloch
1 1 1 equations are modified to take account of the exchange
D C 48 by including appropriate terms.55,56 The line shape of the
T2 2T1 T2
resulting spectrum, as a function of n, is given by the real
and hence the contribution to the bandwidth by equa-
part of G in equation (50):
tion (49):
1 1 1 k A C k B C a A pA C a B pB
D D C 49 GDC 50
pT2 2pT1 pT2 aA k B C aB k A C aA aB
To separate T2 into its components, it would have been
necessary to do separate IR laser pumpprobe experiments where C is an amplitude constant, pA and pB are the
to determine T1 , although in most of the experiments it was fractional populations of nuclei in sites A and B, and the a
reasonable to assume that the bandwidth contribution from terms are given by equation (51):
T1 was small. 1 1
It has subsequently been shown that two-dimensional (i.e. aA D 2pinA n C ; aB D 2pinB n C
T2A T2B
fifth order) Raman echoes can, in principle, also provide 51
detailed information about interactions in solution.52 More- If there are equal populations of A and B, and kA D kB D
over Mukamel and co-workers53 have shown recently that k, tA D tB , with t defined by t D tA tB /tA C tB D 1/2k,
both T1 and T2 can be determined in a two-dimensional
then [equation (52)]:
Raman experiment, i.e. without the necessity of doing a
pumpprobe experiment to determine T1 . 2 C t a A C a B
Thus these new echo experiments, which it must be GDC 52
aA C aB C 2taA aB
added are technically very difficult, are really beginning
to probe the intimate dynamics of solution interactions On separation of real and imaginary parts, and assuming
and hence the subtle contributions to the band shapes in T2A D T2B D T2 , then the absorption is given by equa-
condensed phases. tion (53):
P1 C t/T2 C QR
vn D C 53
P2 C R2
4 CAN VIBRATIONAL SPECTROSCOPY
where, with n D 0.5nA C nB n, and n D nA nB ,
MIMIC DYNAMIC NMR? P D t[1/T22 4p2 n2 C p2 n2 ] C 1/T2 , Q D 2ptn
and R D 2pn[1 C 2t/T2 ].
To some extent the previous parts of this article have Figure 11 shows the appearance of the spectrum for
been leading up to this question. The application of various parameter values. This result is well known and
dynamic NMR to rapid chemical processes was established similar diagrams appear in every NMR text. However,
in the early days of NMR. Before asking the question anticipating the later discussion, some points are worth
whether there is a possibility of a vibrational spectroscopy emphasizing:
equivalent of course on a different timescale it is worth
reminding ourselves of the basic NMR equations.54 1. If the exchange is very slow (k ! 0), the spectrum
consists of two Lorentzians centred at nA and nB , each
with FWHM given by equation (54):
4.1 Bloch equations for dynamic NMR
1 k
FWHM D C in c s1 54
The simplest situation is where there is exchange in solution pT2 p
between hydrogen nuclei in two sites or positions (A i.e. the original Lorentzians have been broadened by
and B), which have different chemical shifts but with an amount D k/p cycles s1 .
the coupling constant equal to zero. In the absence of 2. If the exchange is very fast (k ! 1), the spectrum
exchange, there are two NMR signals at nA and nB ; from shows a single Lorentzian at the mid-point with FWHM
the Bloch equations, with the usual relaxation terms, these given by equation (55):
signals are Lorentzians with FWHM equal to 1/pT2A and

1/pT2B . Since, for protons, T2 is usually of the order 1 p
FWHM D C n2 55
of seconds, the bands in the absence of exchange are pT2 2k
Bandwidths 15
density matrix approach must be used;54 but basically the

philosophy of the Bloch model is applied.
Since the frequency and time dependence must be related
by the FT relationships [equations (3) and (4)], an alterna-
tive approach to the Bloch equations is to use correlation
functions (Section 1.3). This was first demonstrated by
Anderson59 and by Kubo.60,61 It is frequently convenient
to incorporate the density matrix formulation in the corre-
lation function approach.62,63 In Appendix 364 is shown a
simple example employing these methods.
4.2 Bloch equations for vibrational spectroscopy
The non-exchange Bloch equations can be redrafted for

optical spectroscopy.65,66 It is then necessary to include
terms for the exchange process, and this has been done,
via the density matrix method, in an important paper by
MacPhail and Strauss.67 (These authors were not the first
to use, either directly or by implication, similar equations,
but this paper points out some of the errors in previous
applications.) For IR spectroscopy, for an isotropic sample
(i.e. liquid or polycrystal), the spectrum (in radians s1 ) for
a similar situation to the NMR case above, is described by
the real part of G in equation (56):
NtA
GD [aB tB C 1IA C K2 aA tA C 1IB
pD1 C K
C 2KIA IB 1/2 cos g] 56
where N is the total number of molecules, D D [aA tA C

Figure 11. NMR band shapes calculated for an uncoupled two- 1aB tA C 1 1], K is the equilibrium constant for the
site exchange involving two spin 1/2 nuclei, with n D 20 cycles exchange A $ B; IA and IB are the intensities of bands A
s1 , T2 D 0.5 s, and t D 5, 0.1, 0.05, 0.0225, 0.0113, 0.005, and B, per individual molecule, without any exchange and
0.001 and 0.00005 s. [Reproduced from Sandstrom (1982)54 by [equations (57) and (58)]:
permission of Academic Press.]

1 1
i.e., the faster the exchange becomes, the nearer does aA D i! !A C D 2pin nA C
T2A T2A
the spectrum become to a Lorentzian with FWHM 57
of (1/pT2 ).
1 1
3. At intermediate rates of exchange, the spectrum con- aB D i! !B C D 2pin nB C
sists of the two individual lines broadening and moving T2B T2B
together, plus an extra increase in intensity between the 58
peaks. It is important to emphasize this point because, Apart from the cos g term (g is the angle between mA
in the early days of NMR, before the advent of dynamic and mB ), this equation is identical in form to the NMR
NMR computer programs, approximate estimates of the equation. As pointed out,67 the reason there is no similar
exchange rate were obtained from the ratio of the inten- cos g term in the NMR equation is that the large external
sity in the center to the intensity at the maxima,57 or magnetic field keeps the nuclei lined up regardless of
at nA and nB .58 whether they are in site A or B, so that g D 0 and cos g D 1.
If we make similar approximations to the NMR case, i.e.
This simple Bloch method will not work for systems IA D IB , K D 1 (implying tA D tB D 2t), T2A D T2B D
which have to be described quantum mechanically, i.e. T2 , then the spectrum is given by the real part of G in
when spinspin coupling is not zero. In such cases the equation (59). (Note, in the NMR literature t is defined as
[tA tB /tA C tB ], so when tA D tB , t D tA /2. In MacPhail processes. However, in vibrational spectroscopy, in order
and Strauss,67 , when tA D tB , t is defined as tA . We use t D for the rate to be within reasonable limits, to pro-
tA /2 to give direct compatibility, and hence the equations duce coalescence the bands are limited to being, say,
look slightly different from those in the original paper.) 10 cm1 apart. Unfortunately in liquid media vibrational
bands are not narrow compared with 10 cm1 , so that
1 C cos g C taA C aB
GDB 59 obtaining an accurate picture of the coalescence is very
aA C aB C 2taA aB difficult.
which, if mA and mB are in the same direction, i.e. Returning to the Bloch model. Since there are many
g D 0 , is identical to equation (52), and hence the same examples in the literature, which, either directly or indi-
approximations will apply as for NMR. In particular, for rectly, apply the vibrational Bloch equations to chemical
reasons which will be apparent later, there will be a build- exchange,68 80 we shall return to the extremely important
up of intensity in the middle, over and above the simple point about their applicability in due course. However we
broadening and moving together of two bands. may first ask if there are any circumstances where the Bloch
MacPhail and Strauss67 also derive the corresponding equations will be applicable. We suggest there are two: pure
equations for non-resonant Raman spectroscopy, for both dephasing, which mimics chemical exchange, and electron
isotropic and anisotropic scattering; for isotropic Raman transfer processes.
scattering there is no angle term, but there is for
anisotropic scattering. 4.2.1 Band collapse due to pure dephasing
However the major statement that MacPhail and Strauss
make is that, for chemically exchanging species, the mod- In Section 2.2 on pure dephasing we described the anhar-
ified Bloch equations are not applicable! For Bloch equa- monic interaction of a low-frequency mode with a high-
tions to be applicable, the transfer between the two sites frequency fundamental, which results in the broadening
must be fast enough to cause no change in the phase of (and shift) of the fundamental. An interesting variation
the density matrix elements during the transfer. What this on this has been described by Abbott and Oxtoby,81
means in practice is that the transit time (tt ), i.e. the time and, slightly differently, by Marks et al.82 Following the
taken for the molecule to jump from site to site, must be approach of Abbott and Oxtoby,81 we suppose that two
very short compared with the reciprocal of the spread of fre- high frequency fundamentals !A and !B lie close together
quencies characterizing the reaction. Suppose the two NMR (!A !B D !), and that to simplify the calculations,
proton peaks for A and B in Section 4.1 are 10 cycles s1 assume both interact with one low-frequency mode, !i , as
apart; the time scale is thus 1/10 s. Since tt may be of shown in Figure 12. It is assumed that thermal activation of
the order of 1012 s, the conditions are (reassuringly!) eas- the low-temperature mode causes dephasing via fluctuating
ily met. However if two IR peaks are 10 cm1 apart, the anharmonic terms of the form of equation (60):
timescale is 3 1012 s, which puts it in the range of tt . Vanh D FAB tqA qB C FiiAB tqi2 qA qB 60
For the reaction to affect the vibrational spectrum, the reac-
tion time, tr (rate constant D 1/tr ) must be short enough to where the force constants FAB and FiiAB are time dependent
affect the vibrational spectrum. We can get an idea of what due to the low-frequency lattice modes; the qs are the
this means by supposing that the two bands 10 cm1 apart appropriate normal coordinates. It has also been assumed
are each 5 cm1 FWHM in the absence of exchange. Ignor- that terms that are responsible for the additional intensity
ing any other effects of change with temperature on the between bands (see the NMR discussion above) are ignored.
spectrum, we suppose that an increase in FWHM of 1 cm1
due to exchange can be detected. From equation (54), Interaction via i
this corresponds to k/p 1 cm1 D 3 1010 cycles s1 , so
tr 1011 s. For such a very rapid exchange process, the
barrier to exchange must be very low. This means that, in
principle, it may be impossible to disentangle the exchange
contribution to the band broadening from other contribu- B B B
tions to the band broadening. The reactive motion is A A +A
correlated with the non-reactive motions.
There is a practical point. In NMR spectroscopy, the A B
narrow bands associated with non-exchanging spectra can Figure 12. Representation of the interaction, via the low-
be far apart, and still easily coalesce, because the rate frequency mode !i , of the two close-lying fundamentals !A
constant for coalescence is within normal chemical and !B .
Bandwidths 17
Assuming the Markovian approximation (bath relaxation is

fast compared with vibrational relaxation) then for one of
the fundamentals the shift in frequency (!A ) is given by
equation (61): 15 K
!"2 eEi /kT

!A D A 61
1 C !2 "2
and the effect on the bandwidth (1/pT20 ) by equation (62): 13 K
1 1 "eEi /kT
D C A 62
T02 T2 1 C !2 "2
These equations look very similar to those given in
Section 2.2 for pure dephasing involving coupling to a 12 K
low-frequency mode. There are however some subtle dif-
ferences. In equations (41) and (44), ! is the difference
in frequency between fundamental and hot-band caused by
anharmonicity, whereas in equations (61) and (62), ! is
11 K
the difference in frequency between fundamentals. In equa- 0.17
tions (41) and (44), t is the lifetime of the low-frequency D F
Absorbance
mode, whereas in equations (61) and (62), " is the decay C G
E
time (assumed exponential) for correlation functions of the 0.07
type hFiiAB 0FiiAB ti which in a liquid may well be as A B
9K
short as 1013 s. Perhaps the most striking thing about this
0.03
approach is its prediction that, as the temperature is raised,
770.0 771.0 772.0 773.0 774.0
the two fundamentals will broaden and merge together.
cm1
(Note, since the model specifically leaves out additional
intensity terms, it does not predict that there should be Figure 13. 80 SeF6 in Kr matrix, deposited at 20 K. IR spectrum of
extra intensity between the bands.) matrix split n1u SeF stretching band as a function of temperature.
[Reproduced from Jones and Swanson (1981)83 by permission of
The two methods81,82 were initially used to explain
the American Institute of Physics.]
the behavior of solid durene. However, one of the most
striking applications comes from matrix isolation studies. (FranckCondon), and hence the phase of the vibration will
Figure 13 shows the IR spectrum in the region of the matrix be maintained during the electron jump. Thus the Bloch
split t1u n(SeF) stretching band of 80 SeF6 in Kr, over the condition is likely to be fulfilled. We describe two sets of
temperature range 9 K to 15 K.83 Although bands A, B, C, D experiments that have shown this effect. The first84 exam-
and G are hardly affected by the temperature rise, it is clear ined the IR spectrum of [Fe3 O(OOCCMe3 )6 (C5 H5 N)3 ]
that bands E and F coalesce to a single band, analogous from 80 to 410 K. The core structure of the complex con-
to NMR collapse. Following the model of Abbott and sists of a central O atom surrounded by three Fe atoms,
Oxtoby,81 Jones and Swanson83 used the simplest modified with oxidation states (III), (III) and (II). Thus the electron
optical Bloch equations to fit the experimental behavior of transfer process is:
bands E and F. This gives Ei D 31 cm1 and " D 4 ps; the
energy falls within the range of matrix host phonon bands. Fe(III)Fe(III)Fe(II) ! Fe(III)Fe(II)Fe(III), etc.
As pointed out,14 this use of Bloch equations is justified
because the vibrational exchange dephasing involves no Two of the bands in the IR spectrum can be assigned to
physical exchange between sites, which is what we usually the two components of the antisymmetric in-plane stretch
imply when describing chemical exchange. What therefore of the Fe3 O unit with C2v symmetry. If all the Fe atoms
of an exchange that is more chemically significant? become equivalent (D3h symmetry) then these two bands
should coalesce to a single band, which is the degenerate
stretch of the D3h unit. In fact over the temperature range
4.2.2 Effect of electron transfer on vibrational
examined, the two bands broaden and move together, while
bandwidths
other IR bands are hardly affected; the temperature range
The rate constants for electron transfer (1/tr ) vary over is too small to observe actual coalescence. Applying the
many orders of magnitude. However, the transit time (tt ) is Bloch model to the increase in breadth of one of the bands
essentially zero compared with the time of nuclear motions (see above) gives a rate constant (1/tr ) of 5 1011 s1 .
In the second example, Ito et al.85 synthesized some 4.4 Genuine chemical exchange effects on
complicated Ru complexes, but whose core features for vibrational spectra
our purposes can be described as (OC)[Ru3 ]L[Ru3 ](CO).
Here [Ru3 ] represents a core structure with a central O 4.4.1 Exchange effect on spectral profile
atom surrounded by three Ru atoms with different groups It was noted above that in the Bloch equation solutions
bridging the Ru atoms; each [Ru3 ] group also has one CO and in correlation function equivalent solutions for both
group attached to a particular Ru atom; L is a bridging NMR and vibrational spectroscopy, one striking feature is
ligand. If the mean oxidation state of the two [Ru3 ] that the effect of exchange on spectra does not simply
groups is the same, the n(CO) IR bands of each of the lead to the broadening and eventual coalescence of two
two CO groups occur at the same frequency; this applies bands. There is in addition, in the simplest model, extra
to both [Ru3 ]/[Ru3 ] and to [Ru3 ] /[Ru3 ] , where in the intensity between the bands; in the NMR case this is
latter case both [Ru3 ] units have been electrochemically clearly shown in Figure 11. What about the IR? This
reduced by one unit. In fact, for a typical complex the is most strikingly shown by comparing the behavior of
n(CO) frequencies are at 1938 and 1889 cm1 , respectively. tricarbonyl(h4 -norbornadiene)iron (1) and tricarbonyl(h4 -
However in several of the mixed complexes, [Ru3 ] /[Ru3 ], butadiene)iron (2).
there is a single n(CO) band at the mid point, presumably
CO CO
because of very rapid electron transfer [Ru3 ] /[Ru3 ] !
[Ru3 ]/[Ru3 ] . Applying a Bloch analysis gives electron CO CO
Fe Fe
transfer rate constants of 1011 to 1012 s1 . CO CO
(1) (2)
4.3 Pseudocollapse CO
CO
Before considering genuine cases of the effect of chemical Fe
CO
exchange on vibrational spectra we should first consider
(3)
pseudocollapse. This phrase was coined14 to describe a
situation which looks, at first sight (but see later), exactly We shall return to these, and tricarbonyl(h4 -1,5-
like an IR equivalent of NMR collapse, but in fact is not. cyclooctadiene) (3), in more detail shortly. For the moment
The phenomenon is best understood from a diagram (see we note that all structures have a plane of symmetry (Cs
Figure 14). We suppose an XY group in some molecule point group), and hence the n(CO) IR spectra of both (1)
can shift position most obviously by a rotation from one and (2) (and 3) show a high-frequency band (a0 ), and two
position in a molecule to another from site A to site lower-frequency bands (a0 and a00 ) close together. If there is
B. Depending on circumstances, in particular the details very rapid intramolecular exchange among the CO groups
of the potential functions, anharmonicity and relaxation in either (1) or (2), then the vibrational potential function
speed, the spectra might appear as in Figure 14(b). Again develops three-fold symmetry (C3v point group) and the
depending on the detailed conditions14 , raising the temper- spectrum will consist of a high-frequency band (a1 ) and a
ature may cause the two bands in Figure 14(bii) to broaden degenerate low-frequency band (e). In other words the low-
and move closer together. Thus the behavior seems to frequency pair (a0 and a00 ) will collapse to a single band (e).
mimic NMR band collapse, under circumstances in which NMR evidence89,90 shows that for (2) the intramolecular
there is no exchange of XY between sites A and B; in exchange is far too slow to affect the vibrational spectrum,
other words the spectral behavior is entirely intrawell, but there are good theoretical reasons91 for supposing that
with no need for interwell/site-jumping motion. It is the exchange energy barrier in (1) is far lower than in (2).
argued convincingly14 that several of the examples which Figure 15 shows the effect of temperature on the lower
claim the observation of dynamic Raman/IR band col- pair of n(CO) IR bands for (1) and (2). For (2), as the
lapse, analogous to NMR, in fact are best explained as temperature is raised, there is some effect on the spectrum
pseudocollapse. Recent examples where the possibility of slight shift and broadening presumably due to the kind
pseudocollapse is suggested include perfluoropolyethers of effects discussed previously, i.e. change in degree of
with carboxylic end groups,86 and sodium hydrogen bis(4- homogeneity/inhomogeneity, change in contributions to
nitrophenoxide)dihydrate.87,88 However, in both cases the dephasing and rotational contribution. However, the total
authors suggest that real exchange is occurring in addition spectrum is easily modeled on the overlapping of two
to pseudocollapse. How then can one be convinced of gen- bands. It is clear that for (1), extra intensity is observed, and
uine chemical exchange? We take this up in the next section. this is readily attributed to the fact that genuine exchange
Bandwidths 19
rAB rAB
Energy
0A 1A 0B 1B
vt = 2 n2A n2B vt = 2
vt = 1 n1A n1B vt = 2
vt = 0 n0A n0B vt = 0
Vtor A B
0 180
(a) Reaction coordinate
(i)
0A 0B
1A 1B
2A
2B
3A 3B
(ii)

A B
(b)
Figure 14. (a) Schematic representation of the potential function for restricted rotation of the XY group between sites A and B,
assumed to be 180 apart. The torsional levels are denoted by nt D 0, 1, 2, etc, and n1m represents the population of level m for site
1. For convenience the ground-state potential is drawn symmetric, with the difference in frequency in the two sites arising from the
potential function differences in the excited vibrational state of XY. nA0 is the frequency of the XY stretch in site A with the XY group
in level 0, etc. Ignoring splitting of levels due to degeneracy, there will also be transitions such as nAB
r , which represent transitions from
energy levels well above the barrier, i.e. from free rotational levels. These have the superscript AB, since in the simplest approximation
they will have the same frequency, which will be the mean of nA0 and nB0 . (b) Schematic representation of the spectra to be expected
for the potential function in (a). A simple assumption is that, because of anharmonicity, the excited-state potential is such that nA0 , nA1 ,
nA B B B
2 , etc., and n0 , n1 , n2 , etc., differ in frequency by a small but constant increasing or decreasing amount. If the individual lines can be
resolved then the spectrum will be as shown in (i), where the intensity of each line is determined only by the Boltzmann distribution.
If the individual lines cannot be resolved then the spectrum may appear as shown by the dotted line. If rapid Redfield-type relaxation
occurs, then the spectrum will have the appearance of (ii). [Reproduced from Turner et al. (1995)64 by permission of the American
Chemical Society.]
with the appropriate extra intensity, in the middle.

If the system is sufficiently stable, further temperature
rise will lead to this coalesced band becoming nar-
(i) (i)
rower (see Figure 11). As has been pointed out,14 the
chemical exchange rate cannot usually be varied over
Absorbance
Absorbance
a range wide enough to see the complete exchange
collapse of well-separated vibrational bands. However
here are two examples which seem to satisfy these
(ii) (ii) conditions:
1. The II stretching vibration of I2 in heptane at 213 K
can be observed in the resonance Raman spectrum,
for both free I2 (211.3 cm1 ) and for I2 complexed
to benzene (207.8 cm1 ).93 As the temperature is
2010 1992 1974 2010 1970 1930
raised the bands coalesce, giving, at 293 K, a sin-
(a) Wavenumber /cm1 (b) Wavenumber /cm1
gle band with FWHM D 5.2 cm1 . On raising the
Figure 15. (a) Low-frequency pair of n(CO) bands of (2) in temperature to 349 K, the single band narrows to
liquid Kr/Xe solution at (i) 122 K, and (ii) 193 K. The upper
spectra are experimental; the lower curves are simulated spectra 4.0 cm1 .
based on a mixed Lorentzian/Gaussian. (b) Low-frequency pair of 2. In the spectrum of (1) described above, the FWHM
n(CO) bands of (1) in liquid Kr/Xe at (i) 112 K, and (ii) 177 K. of the coalesced band in supercritical Xe at 25.5 C is
The upper spectra are experimental; the lower curves are the best 15.3 cm1 , which narrows to 13.4 cm1 at 100 C.94
simulated spectra based on a mixed Lorenztian/Gaussian. The
thick black lines indicate the extra intensity in the experimental
spectra, greater than that predicted on the basis of simple band 4.4.2 Theory
overlap. [Reproduced from Turner et al. (1995)64 by permission
of the American Chemical Society.] As stated in Section 1.2, we know the FT relationships
between the time and frequency domains [equations (3)
between sites is occurring. We note that no model, e.g. and (4)]. We also know (Section 1.3), that, for IR, Ft
pseudocollapse, that does not include genuine exchange, is the autocorrelation function for the dipole moment oper-
will predict the extra intensity feature in the spectrum of ator; for Raman the autocorrelation function involves the
(1). Moreover there is no way that the Oxtoby dephasing corresponding polarizability operator. Thus, in principle, if
model could explain these results.64 (It should however there is some way of calculating the autocorrelation func-
be noted that if an exchanging system involves the angle tions for each stage of a reaction, and as a function of
between the dipoles being 90 , then cos g D cos 90 D 0, temperature, we shall obtain the vibrational spectrum for
and there will be no exchange contribution.) We thus the reacting system under all conditions. We shall then be
argue that evidence of extra intensity is conclusive in a position to compare this result with that obtained by
proof of genuine exchange; absence does not preclude it. using some model involving Bloch equations. Not surpris-
However we must enter a caveat: if, in an exchanging ingly this ideal is not obtainable for real systems and a
system, the temperature is sufficiently high to populate number of different authors have approached the problem
levels above the barrier height, then there may well be somewhat differently. This is not the place to go through
extra intensity in the middle, arising in Figure 14 from these models in detail; interested readers should go to the
transitions corresponding to nAB
r . This point has been made
original literature. However a few comments are in order.
by, amongst others, Cavagnat et al.92 , in a vibrational study Cavagnat and Lascombe95 examined the IR and Raman
of the fast conformational dynamics of cyclopentene and spectra of C6 H5 CHD2 and C6 D5 CHD2 in the aliphatic CH
derivatives. But if the requirement is to distinguish between stretching mode range, over the temperature range 17 to
inter and intra-well dynamics, then exchange occurring 165 K. Based on Robertson and Yarwoods96 treatment of
because of population of levels above the barrier is certainly hydrogen bonding, the authors considered two theoretical
not intra-well. Note, the extra intensity argument does models involving the anharmonic coupling of the CH
not lead to a quantitative estimate of kinetics and barrier stretch with the torsional CHD2 mode. The first assumes
heights, because there will be other contributions to the Markovian jumps from one equilibrium torsional position to
change in band shape. another (i.e. analogous to NMR jumps); the second assumes
There is a further important point. We have seen that true variation in the CH stretch as the molecule oscillates
exchange between two sites will lead, as the temperature within the torsional potential well. It is concluded that
is raised, to a coalescence of the two vibrational bands, the bands broaden more than can be explained by the
Bandwidths 21
jump model alone, and hence some combination of the barriers, such that U kB T. At room temperature this
two models is necessary to explain the data, exactly as means that the barrier height will have to be less than
described by Strauss. In a related study on the CH stretches 0.6 kcal mol1 (1 cal D 4.18 J) for the Bloch equations not
of the CH3 groups of some long-chain alkanes, MacPhail to be usable. However it is not clear if the approximations
et al.97 used the Redfield relaxation method to show that made in the calculations actually mimic systems of real
the coalescence of the IR bands can be explained by the chemical interest.104
variation in CH force constant with motion within the In mixtures of pyridine and ethanol, the frequency of
torsional potential well, i.e. there is no need for these the Raman n1 band of pyridine is sensitive to whether
systems to invoke jumps between equilibrium positions. the pyridine is, or is not, complexed to ethanol. Raising
Using the methods of Mori98 and Zwanzig99 , MacPhail the temperature leads to band coalescence.105,106 Bearing
and Strauss67 set up the appropriate correlation functions in mind all the problems,67 the authors conclude that a
for the two-site exchange system, assuming that !q D !0 C window of opportunity exists for extracting exchange
!Q, i.e. that the frequency of the oscillator (!q ) varies information, when the exchange rate is 1011 1012 s1
from a fixed frequency (!0 ) by an amount that depends on and the frequency difference between the uncoalesced
the difference in frequency between the two sites (!) and bands is 25 cm1 . Related experiments have also been
the reaction coordinate (Q). Thus, in principle, both intra- described.107 109
well and inter-well effects on the frequency are taken into In a quite different approach, Weiss110 112 has attempted
account. It is shown that under certain approximations, of to deduce the intra-well effects on the inter-well dynam-
which the most important is equivalent to the statement that ics by examining some very simple theoretical mod-
vibrational phase does not change during a jump, the model els. For example, in one model the molecule YXY0 is
reduces to the Bloch equation approach. assumed to have a potential barrier between two struc-
The isotropic Raman spectra of the OH stretching vibra- tures involving different YXY0 bond angles. Assuming
tion of ethanol in n-decane changes from a broad band particular values for force constants, bond lengths, masses
showing evidence of splitting at 293 K, to a single, nar- and LennardJones potentials for the interaction with sol-
rower, band at 423 K.100 The authors suggest that this vent, the dipole correlation function is calculated (as a
difference arises because, at the lower temperature, the function of barrier height) and then Fourier transformed
OH stretches for the two torsional isomers of ethanol to give the predicted IR spectra. The conclusion is that
are separate but unresolved, but that at the higher tem- the spectra obtained could all be fitted with Bloch
perature the rate of torsional isomerization is so fast that lineshapes.
the OH stretch shows one band, i.e. once again analo- It will be clear from what is described above, that the
gous to NMR exchange behavior. In an attempt to take situation is complicated. Suffice it to say that, in the present
account of the intra-well contribution to the change in spec- authors view, the jury is still out, but that the restrictions
trum, it is assumed that the change in OH band profile of on the extraction of meaningful kinetic data from collapsing
tertiary butanol, which shows no evidence for inter-well vibrational spectra are probably not as severe as postulated
exchange, will be a good model for the intra-well contri- by Strauss.
bution in ethanol. With the experimental data for butanol
incorporated in the behavior of ethanol, the Anderson cor-
4.4.3 Occams razor
relation function method59 is used to calculate a theoretical
spectrum. A good fit is obtained with tr D 4.4 1012 s This section will concentrate on compounds of type (1),
at 293 K and 2.0 1012 s at 423 K. It should be noted (2) and (3), since they provide the clearest evidence of the
that this calculation is based on the assumption of similar match between theory and experiment. It was first noted
contributions to intra-well features in ethanol and butanol; some years ago, by Grevels and colleagues,91,113 that (1)
however, as the torsional barrier is lowered in going from and (3) displayed the behavior described in Section 4.4.
butanol to ethanol, the torsional potential function will The coalescing bands were fitted by a conventional mod-
change, so this assumption must be treated with caution ified Bloch equation to give an activation energy for
(but see later). the intramolecular exchange of 1 to 1.5 kcal mol1 , and
In a series of papers, Bratos and colleagues101 103 have with the rate of scrambling of the CO groups varying
attempted to calculate the correlation functions for isotropic from 0 at 140 C to 9.6 1011 s1 at C20 C. The
Raman scattering for some real systems. Using slightly h4 -norbornadiene complexes with Ru(CO)3 and Os(CO)3 ,
different assumptions in the various cases, they reach the and (h4 -cyclobutadiene)Fe(CO)3 were shown to behave
important conclusion that the jump models (i.e. the Bloch similarly114 ; there was also evidence for fast intramolecular
approach) are really only inadequate for extremely low CO scrambling in Os4 (CO)12 .115 A detailed study94 of (1)
and (2), including isotopic substitution, was performed in CO group positions occurs at temperatures as low as 28 K.
liquid noble gases, which permit a wide range of tempera- This argues for a very low barrier. In the second part,
tures and in which the bands in the absence of any exchange there are two stages. The first stage involves deriving
are quite narrow. Again the best explanation of the coa- the appropriate IR correlation function in the absence
lescing spectra involved dynamic exchange. Reassuringly a of exchange, including terms which on FT give rise
density functional study116 of the barrier height to exchange to both Lorentzian and Gaussian bands; this is because
in (2) gave 9.4 kcal mol1 , compared with the experimen- at low temperature the n(CO) bands appear Voigt-like.
tal value of 9.1 kcal mol1 determined by NMR. This adds The experimental behavior, with temperature, of the non-
confidence to the same calculation116 for (1), which gives exchanging (2) is then used as input to predict the behavior
0.5 kcal mol1 , hence suggesting very rapid exchange, con- of (3) without exchange (this is similar to the example
sistent with Grevelss observations, and adding further given above100 where the behavior of butanol is used
contradiction to the argument117 that the behavior could as input in the study of ethanol). The exchange term is
be explained by pseudocollapse. now added to the correlation function, with assumptions
The last word on molecules of this type has surely been equivalent to those that permit use of the Bloch equation.
provided in the paper by Grevels and colleagues in 1998118 The next stage is to take account of the fact that the
on the behavior of tricarbonyl(h4 -1,5-cyclooctadiene)iron n(CO) spectrum derives from coupled CO motions and
(3). The paper has two parts. In the first part, hence that the transfer of dipole moment on exchange is
low-temperature, isotopic (13 CO), matrix photochemical not simple (contrast this with the XY system in Figure 14).
experiments show conclusively that the exchange between The final result is shown in Figure 16. The agreement
between experiment and calculation is so good that
they are indistinguishable. An Eyring plot gives H D
A 0.7 kcal mol1 and S D 0.3 cal K1 mol1 . Applying
Occams razor we conclude that if such perfect agreement
is best explained on what is tantamount to a Bloch equation
Absorbance
A calculation, then perhaps this method works reasonably

A
well. We do not need to use enormously more complex
models to explain the data.
5 CONCLUSIONS
(a) We started this article with the intention of outlining
some of the intriguing factors that influence the shape of
A simple vibrational bands in condensed systems. We have
ignored many problems, such as the influence of isotopes,
Fermi resonance and internal rotation, as well as the very
A
Absorbance
A important topic of band shapes of adsorbed molecules,

which can provide information about the dynamics of the
adsorbate. There is also the interesting problem of the
shapes of bands in the vibrational spectra of electronically
excited states, which again provide information about
the interaction in the excited state between molecule
and solvent. It seemed appropriate to concentrate on the
simple issues, themselves complicated enough, and to then
2050 2020 1990 1960 1930 1900 turn to the still unresolved question of the possibility
(b)
/cm1 of using vibrational spectra in a manner equivalent to
Figure 16. (a) The n(CO) bands in the variable-temperature IR dynamic NMR.
spectra of tricarbonyl(h4 -1,5-cyclooctadiene) (3), recorded in 2-
MP at 293 K (bold line) and 273/253/233/213/193/173/153/133 K.
(b) Simulation of the bands of (3) using rate constants for the
CO site exchange process ranging from 1.5 1011 s1 at 133 K
ACKNOWLEDGMENTS
to 1.54 1012 s1 at 293 K (bold line) with various spectral
parameters.118 [Reproduced from Grevels et al. (1998)118 by I am particularly grateful for considerable help with this
permission of the American Chemical Society.] article to Professor Michael Fayer (Stanford University),
Bandwidths 23
Professor Jeanne McHale (University of Idaho) and Pro- 14. K.A. Wood and H.L. Strauss, J. Phys. Chem., 94, 5677
fessor Jack Yarwood (Sheffield Hallam University). Others (1990).
who have provided valuable assistance include Professor 15. A. Tokmakoff, D. Zimdars, R.S. Urdahl, R.S. Francis, A.S.
Mark Berg (University of North Carolina), Professor Mike Kwok and M.D. Fayer, J. Phys. Chem., 99, 13 310 (1995).
Chesters and Dr Katharine Reid (both of University of Not- 16. J.C. Owrutsky, D. Raftery and R.M. Hochstrasser, Annu.
tingham). Rev. Phys. Chem., 45, 519 (1994).
17. C. Manzares and G.E. Ewing, J. Chem. Phys., 69, 2803
(1978).
ABBREVIATIONS AND ACRONYMS 18. S.M. Arrivo, T.P. Dougherty, W. Tandy Grubbs and E.J.
Heilweil, Chem. Phys. Lett., 235, 247 (1995).
DBP Dibutyl Phthalate 19 E.J. Heilweil, R.R. Cavanagh and J.C. Stephenson, Chem.
EVR External Vibrational Relaxation Phys. Lett., 134, 181 (1987).
FT Fourier Transformation 20. S.F. Fischer and A. Lauberau, Chem. Phys. Lett., 55, 189
IVR Intramolecular Vibrational Relaxation (1978).
2-MP 2-Methylpentane 21. S. Marks, P.A. Cornelius and C.B. Harris, J. Chem. Phys.,
2-MTH 2-Methyltetrahydrofuran 73, 3069 (1980).
NMR Nuclear Magnetic Resonance 22. D. Hsu and J.L. Skinner, J. Chem. Phys., 83, 2097 (1985).
23. R.M. Shelby, C.B. Harris and P.A. Cornelius, J. Chem.
Phys., 70, 34 (1979).
REFERENCES 24. J.H.R. Clarke, Band Shapes and Molecular Dynamics in
Liquids, in Advances in Infrared and Raman Spec-
1. J.F. James, A Students Guide to Fourier Transforms, troscopy, eds R.J.H Clark and R.E. Hester, Heyden, Lon-
Cambridge University Press, Cambridge (1995). don, 109193, Vol. 4 (1978).
2. N.F. Barber, Experimental Correlograms and Fourier Trans- 25. S.P. Wang and M. Schwartz, J. Mol. Liq., 47, 121 (1990).
forms, Pergamon, Oxford (1961). 26. G.R. Fleming, Chemical Applications of Ultrafast Spec-
3. R.N. Bracewell, The Fourier Transform and Its Applica- troscopy, Oxford University Press, New York (1986).
tions, 2nd edition, McGraw-Hill, New York (1986). 27. M. Dubs and Hs.H. Gunthard, Chem. Phys. Lett., 64, 105
4. A. Tokmakoff and M.D. Fayer, Acc. Chem. Res., 28, 437 (1979).
(1995).
28. M. Dubs, L. Ermanni and Hs.H. Gunthard, J. Mol. Spec-
5. R.G. Gordon, Adv. Magn. Reson., 3, 1 (1968). trosc., 91, 458 (1982).
6. J. McHale, Molecular Spectroscopy, Prentice Hall, New 29. M. Poliakoff and J.J. Turner Infrared Laser Photochemistry
York, Chapter 5 (1998). in Matrices in Chemical and Biochemical Applications
7. M.O. Bulanin, N.D. Orlova and G.Ya. Zelinka, Spectra and of Lasers, ed. C.B. Moore, Academic Press, New York,
Intermolecular Interactions in Cryosystems, in Molecular 175216, Vol. V (1980).
Cryospectroscopy, eds R.J.H. Clark and R.E. Hester, John 30. A. Tokmakoff, R.S. Urdahl, D. Zimdars, R.S. Francis, A.S.
Wiley & Sons, Chapter 3 (1995). Kwok and M.D. Fayer, J. Chem. Phys., 102, 3919 (1995).
8. M.O. Bulanin, Liquefied Gases as Solvents for Vibrational 31. H.-G. Cho and H.L. Strauss, J. Chem. Phys., 98, 2774
Spectroscopy, in Handbook of Vibrational Spectroscopy, (1993).
eds J.M. Chalmers and P.R. Griffiths, John Wiley & Sons,
32. R. Holland, W.B. Maier, S.M. Freund and W.H. Beattie, J.
Chichester, 13291341, Vol. 2 (2002).
Chem. Phys., 78, 6405 (1983).
9. M.J. Clouter, H. Kiefte and R.K. Jain, J. Chem. Phys., 73,
33. B.I. Swanson, L.H. Jones, S.A. Ekburg and H.A. Fry, Chem.
673 (1980).
Phys. Lett., 126, 455 (1986).
10. F.J. Bartoli and T.A. Litovitz, J. Chem. Phys., 56, 404, 413
34. B.I. Swanson and L.H. Jones, High Resolution IR Stud-
(1972).
ies of Site Structure and Dynamics for Matrix Isolated
11. G. Doge and J. Yarwood, Infrared and Raman Studies Molecules, in Vibrational Spectra and Structure, ed.
on Molecular Dynamics in Liquids, in Spectroscopy J.R. Durig, Elsevier, Amsterdam, Vol. 12 (1983).
and Relaxation of Molecular Liquids, eds D. Steele and
J. Yarwood, Elsevier, Amsterdam, Chapter 6 (1991) 35. S.M. Arrivo, V.D. Kleiman, W.T. Grubs, T.P. Dougherty
and E.J. Heilweil, Laser Chem., 19, 1 (1999).
12. S.F. Fischer and A. Lauberau, Chem. Phys. Lett., 35, 6
(1975). 36. K.D. Rector and M.D. Fayer, J. Chem. Phys., 108, 1794
(1998).
13. R. Kubo, Stochastic Theory of Line-shape Relaxation,
in Fluctuation, Relaxation and Resonance in Magnetic 37. A. Tokmakoff and M.D. Fayer, J. Chem. Phys., 103, 2810
Systems, ed. D. Ter Haar, Oliver and Boyd, Edinburgh, (1995).
2368 (1962). 38. Y.S. Bai and M.D. Fayer, Chem. Phys., 128, 135 (1988).
39. D. Vanden Bout, L.J. Muller and M. Berg, Phys. Rev. Lett., 68. M.M. Kreevoy and C.A. Mead, J. Am. Chem. Soc., 84, 4596
67, 3700 (1991). (1962).
40. L.J. Muller, D. Vanden Bout and M. Berg, J. Chem. Phys., 69. M.M. Kreevoy and C.A. Mead, Disc. Faraday Soc., 39, 166
99, 810 (1993). (1965).
41. D. Vanden Bout, J.E. Freitas and M. Berg, Chem. Phys. 70. A.K. Covington, A.J. Tait and Lord Wynne-Jones, Disc.
Lett., 229, 87 (1994). Faraday Soc., 39, 172 (1965).
42. M. Berg and D. Vanden Bout, Acc. Chem. Res., 30, 65 71. R.A. MacPhail, R.G. Snyder and H.L. Strauss, J. Am. Chem.
(1997). Soc., 102, 3976 (1980).
43. M.A. Berg, Vibrational Dephasing in Liquids: Raman Echo 72. B. Cohen and S. Weiss, J. Phys. Chem., 72, 6804 (1980);
and Raman Free-induction Decay Studies, in Ultrafast 74, 3635 (1981).
Infrared and Raman Spectroscopy, ed. M.D. Fayer, Marcel
Dekker, New York (2001). 73. B. Cohen and S. Weiss, J. Chem. Phys., 87, 3606 (1983).
44. R.F. Loring and S. Mukamel, J. Chem. Phys., 83, 2116 74. H. Strehlow, I. Wagner and P. Hildebrandt, Ber. Bunsen-
(1985). Ges. Phys. Chem., 87, 516 (1983).
45. E.L. Hahn, Phys. Rev., 80, 580 (1950). 75. M. Eskola, L. Keskisaari and F. Stenman, J. Mol. Struct.,
175, 383 (1988).
46. N.A. Kurnit, I.D. Abella and S.R. Hartmann, Phys. Rev.
Lett., 13, 567 (1964). 76. G. Eaton, A.S. Pena-Nunez and M.C.R. Symons, J. Chem.
47. I.D. Abella, N.A. Kurnit and S.R. Hartmann, Phys. Rev., Soc., Faraday Trans. I, 84, 2181 (1988).
141, 391 (1966). 77. R.K. Pomeroy, J. Organomet. Chem., 383, 387 (1990).
48. K.D. Rector, D.E. Thompson, K.A. Merchant and M.D. 78. M. Besnard, M.I. Cabaco and J. Yarwood, Chem. Phys. Lett.,
Fayer, Chem. Phys. Lett., 316, 122 (2000). 198, 207 (1992).
49. D.W. Oxtoby, Ann. Rev. Phys. Chem., 32, 77 (1981). 79. D. Cavagnat, R.M. Cavagnat, J.C. Cornut and S. Banisaeid-
50. C. Crepin, M. Broquier, H. Dubost, D. LHermite, A. Tra- Vahedie, J. Phys: Condens. Matter, 4, 4205 (1992).
mer, J.P. Galaup, J.M. Berset and J.M. Ortega, Laser Chem., 80. G. Wilson, L. Hecht and L.D. Barron, Biochemistry, 35,
19, 65 (1999). 12 518 (1996).
51. R. Inaba, K. Tominaga, M. Tasumi, K.A. Nelson and 81. R.J. Abbott and D.W. Oxtoby, J. Chem. Phys., 70, 4703
K. Yoshihara, Chem. Phys. Lett., 211, 183 (1993). (1979).
52. S. Mukamel, A. Piryatinski and V. Chernyak, Acc. Chem. 82. S. Marks, P.A. Cornelius and C.B. Harris, J. Chem. Phys.,
Res., 32, 145 (1999). 73, 3069 (1980).
53. V. Chernyak, A. Piryatinski and S. Mukamel, Laser Chem.,
83. L.H. Jones and B.I. Swanson, J. Chem. Phys., 74, 3216
19, 109 (1999).
(1981).
54. J. Sandstrom, Dynamic NMR Spectroscopy, Academic
Press, London (1982). 84. R. Wu, S.K. Arap Koske, R.P. White, C.E. Anson, U.A.
Jayasooriya and R.D. Cannon, J. Chem. Soc., Chem. Com-
55. H.S. Gutowsky, D.W. McCall and C.P. Slichter, J. Chem. mun., 1657 (1994).
Phys., 21, 279 (1953).
85. T. Ito, T. Hamaguchi, H. Nagino, T. Yamaguchi, H. Kido,
56. H.M. McConnell, J. Chem. Phys., 28, 430 (1957). I.S. Zavarine, T. Richmond, J. Washington and C.P. Kubiak,
57. M.T. Rogers and J.C. Woodbrey, J. Phys. Chem., 66, 540 J. Am. Chem. Soc., 121, 4625 (1999).
(1962). 86. V. Doan, R. Koppe and P.H. Kasai, J. Am. Chem. Soc., 119,
58. V.S. Dimitrov, Org. Magn. Reson., 6, 16 (1974); 8, 132 9810 (1997).
(1976).
87. M. Kreevoy, S. Marimanikkuppam, V.G. Young, Jr, J. Ba-
59. P.W. Anderson, J. Phys. Soc. Jpn., 9, 316 (1954). ran, M. Szafran, A.J. Schultz and F. Trouw, Ber. Bunsen-
60. R. Kubo, J. Phys. Soc. Jpn., 9, 935 (1954). Ges. Phys. Chem., 102, 370 (1998).
61. R. Kubo, Nuovo Cimento Suppl., 6, 1063 (1957). 88. M.M. Kreevoy and V.G. Young, Can. J. Chem., 77, 733
62. A.G. Redfield, Adv. Magn. Reson., 1, 1 (1965). (1999).
63. C.S. Johnson, Adv. Magn. Reson., 1, 33 (1965). 89. L. Kruczynski and J. Takats, J. Am. Chem. Soc., 96, 932
(1974).
64. J.J. Turner, C.M. Gordon and S.M. Howdle, J. Phys. Chem.,
99, 17 532 (1995). 90. L. Kruczynski and J. Takats, Inorg. Chem., 15, 3140 (1976).
65. S. Mukamel, Principles of Nonlinear Optical Spectroscopy, 91. F.-W. Grevels, J. Jacke and K. Seevogel, J. Mol. Struct.,
Oxford University Press, New York, 181 (1995). 174, 107 (1988).
66. W.H. Flygare, Molecular Structure and Dynamics, Pren- 92. D. Cavagnat, S. Banisaeid-Vahedie, L. Lespade and
tice Hall, New York, 444 (1978). S. Rodin, J. Chem. Soc., Faraday Trans, 88, 1845 (1992).
67. R.A. MacPhail and H.L. Strauss, J. Chem. Phys., 82, 1156 93. M. Besnard, N. del Campo, R.M. Cavagnat and J. Las-
(1985). combe, Chem. Phys. Lett., 162, 132 (1989).
Bandwidths 25
94. J.J. Turner, F.-W. Grevels, S.M. Howdle, J. Jacke, M.T. 107. J. Yarwood, Chem. Phys. Lett., 208, 557 (1993).
Haward and W.E. Klotzbucher, J. Am. Chem. Soc., 113, 8347
108. E. Zoidis, J. Yarwood, Y. Denten and M. Besnard, Mol.
(1991).
Phys., 85, 385 (1995).
95. D. Cavagnat and J. Lascombe, J. Chem. Phys., 76, 4336
(1982). 109. T. Tassaing, M. Besnard and J. Yarwood, Chem. Phys. Lett.,
267, 496 (1997).
96. G.N. Robertson and J. Yarwood, Chem. Phys., 32, 267
(1978). 110. S. Weiss, Mol. Phys., 76, 669 (1992).
97. R.A. MacPhail, R.G. Snyder and H.L. Strauss, J. Chem. 111. H. Stassen and S. Weiss, J. Mol. Liquids, 62, 185 (1994).
Phys., 77, 1118 (1982).
112. S. Weiss, M. Buchner and Th. Dorfmuller, Mol. Phys., 81,
98. M. Mori, Prog. Theor. Phys., 33, 423 (1965). 327 (1994).
99. R. Zwanzig, J. Chem. Phys., 33, 1338 (1960).
113. F.-W. Grevels, J. Jacke, W.E. Klotzbucher, C. Kruger,
100. W. Richter, D. Schiel and W. Woger, Mol. Phys., 60, 691 K. Seevogel and Y.-H. Tsay, Angew. Chem., Int. Ed. Engl.,
(1987). 26, 885 (1987).
101. S. Bratos, G. Tarjus and P. Viot, J. Chem. Phys., 85, 803
114. F.-W. Grevels, J. Jacke, R.E.D. McClung, G. Russell,
(1986).
J. Schrickel and J. Takats, Inorg. Mechan. Disc. Group, P22
102. P. Viot, G. Tarjus, D. Borgis and S. Bratos, J. Chem. Phys., (1993).
90, 7022 (1989).
115. V.J. Johnston, F.W.B. Einstein and R.K. Pomeroy, Organo-
103. P. Viot, S. Bratos and G. Tarjus, J. Mol. Liquids, 43, 93 metallics, 7, 1867 (1988).
(1989).
104. H.L. Strauss, personal communication (1990). 116. M. Buhl and W. Thiel, Inorg. Chem., 36, 2922 (1997).
105. M.I. Cabaco, M. Besnard and J. Yarwood, Mol. Phys., 75, 117. H.L. Strauss, J. Am. Chem. Soc., 14, 905 (1992).
139, 157 (1992). 118. F.-W. Grevels, K. Kerpen, W.E. Klotzbucher, R.E.D.
106. M. Besnard, M.I. Cabaco and J. Yarwood, Chem. Phys. Lett., McClung, G. Russell, M. Viotte and K. Schaffner, J. Am.
198, 207 (1992). Chem. Soc., 120, 10423 (1998).
APPENDICES
APPENDIX 1 FOURIER TRANSFORM is real because [equation (63)]:
e2pint D [cos2pnt i sin2pnt] 63

As stated in Section 1.2, the relationships between fre-
quency (in cycles s1 ) and time (in s) are: and, since Ft is symmetrical about t D 0, in the integra-
C1 C1 tion the sine term (the imaginary term) vanishes.
n D Fte2pint dt Ft D ne2pint dn It is interesting to see what happens if the exponential
1 1 is only drawn from t D 0 to t D 1; that is a single one-
3
C1 C1
sided decay starting at t D 0, Ft D et/T2 . In this case the
1 integral is complex, and the FT is given by equation (64):
I! D Ftei!t dt Ft D I!ei!t d!
2p 1 1
4 1
n D
The text also points out that if Ft has the exponen- 2pin C 1/T2
tial form T2 2piT22 n
Ft D ejtj/T2 5 D 64
2pT2 n2 C 1
where jtj represents the modulus of t, then the FT is Since the experimental spectrum is real we need the real
C1 part of the FT [equation (65)]:
n D ejtj/T2 e2pint dt T2
1 Ren D 65
2pT2 n2 C 1
2T2
D 6 which, not surprisingly, is just a Lorentzian of the same
2pT2 n2 C 1
FWHM (1/pT2 ), but with a maximum equal to T2 ,
which is a Lorentzian of FWHM 1/pT2 , and intensity at compared with the double-sided function which gives a
the maximum n D 0 of 2T2 (Figure 1). The function n Lorentzian of maximum 2T2 . It is thus important to know
whether or not the functions that are to introduce the time

dependency are independent of time reversal, i.e. whether 1/ 0
or not Ft is equal to Ft. In practice, these functions
(correlation functions see Section 1.3) are very nearly
F (t )
time reversible; any lack of time symmetry arises from
t
quantum mechanical effects which are usually small.6 We
shall assume that Ft D Ft
A Gaussian band shape is often encountered in vibra-
tional spectroscopy. A typical spectral profile is that of (a)
equation (66):
n D expn2 /t2 66
2T2
which has FWHM of 1.665t.
To obtain the time dependent part, we take the FT
1 /T2
[equation (67)]:
()
C1
Ft D expn2 /t2 e2pint dt
1
D tp1/2 expp2 t2 t2 67 0

(b) 0 0
which is another Gaussian of FWHM 1.665/pt.
Figure 17. (a) Plot of Ft D expt/T2 cos 2n0 t from t D 0
However in all the cases so far, the centers of the onwards. (b) Plot of FT of Ft, after adding together the two
Lorentzian and Gaussian bands are at n D 0, which is, bands centred at n0 and Cn0 .
of course not what happens in vibrational spectroscopy.
But if we superimpose an oscillation of frequency n0 on,
for example, the decaying exponential, the time dependent APPENDIX 2 CONVOLUTIONS
function is given by equation (68):
Various contributions to the shape of a spectral band,
Ft D et/T2 cos 2pn0 t 68 including any from the instrumental slit width, are actually
convoluted together. The formal definition of a convolution
which is shown in Figure 17(a). Assuming that for the of two functions is given by equation (71) (with frequency
decay Ft D Ft, i.e. that the function from t D 0 to as an example):
1 is a mirror image of t D 0 to C1, the FT gives C1
equation (69): An D 1 n0 2 n n0 dn0 71
1
T2
nC1
1 D 69 This is conventionally written as equation (72):
2pT2 n0 n2 C 1
i.e. strictly speaking two Lorentzians centred at n0 and An D 1 n 2 n 72
Cn0 , each with FWHM of 1/pT2 . Since we cannot distin- There is a very important relationship between the indivi-
guish between n0 and Cn0 , we add the two sides together dual n terms and their Fourier transforms, Ft: if
(i.e. assume n0 D Cn0 ), to obtain equation (70), shown in "t is the Fourier transform of the convolution An, then
Figure 17(b) [equation (73)]:
2T2 "t D F1 t F2 t 73
nC1
0 D 70
2pT2 n0 n2 C 1
i.e. the Fourier transform of the convolution is the product
which is simply the same result as ejtj/T2 , except that the of the Fourier transforms of the components.
center of the band is now at n D n0 , rather than n D 0, which Suppose a spectrum is the convolution of two Lorentzians
is more in keeping with experiments. L1 (n) and L2 (n), with L1 D a/[2pa n2 C 1] and L2 =
In discussing vibrational band shapes it is usual to assume b/[(2pb n2 C 1], and hence with FWHM D 1/pa and
that an oscillation is superimposed on the decay pattern, and 1/pb, respectively. The FT of L1 D F1 t D et/a and
to concentrate on the part of the time dependent function of L2 D F2 t D et/b , so F1 F2 D et[aCb/ab] . The FT
that determines the shape of the band. of this is [ab/a C b]/f4p2 [ab/a C b]2 n2 C 1g, which is
Bandwidths 27
a Lorentzian of FWHM of a C b/pab D 1/pa C 1/pb. requires the use of density matrices; correlation functions
That is, the final spectrum is a Lorentzian with half width make the problem more tractable.59 61 That this more
equal to the sum of the halfwidths of the two Lorentzians complex approach reduces to the conventional Bloch
convoluted together. Of course this relationship depends on analysis in simple cases can be shown for the system AB
the two contributing functions being Lorentzian; if one or of two uncoupled spin 1/2 nuclei.62 64
both is Gaussian or partly Gaussian then the convolution In the notation used by Redfield62 and by Johnson,63 the
will be more complicated, and such a convoluted function general equations are [equation (77)]:
is called a Voigt function.
A classic example of the use of the relationships in draa0 2pEa0 Ea

Di raa0 C Raa0 bb0 rbb0 77
equations (72) and (73) is in obtaining the true spectrum dt h bb0
[Tn] from an experimental spectrum [n] where
the slit function [Sn] of the spectrometer is known. As In these equations, raa0 is the density matrix element
stated above, Tn is convoluted with Sn to give n, connecting levels a and a0 , which have energies 2pEa /h
equation (74): and 2pEa0 /h, so that 2pEa0 Ea /h D !aa0 . The R terms
n D Tn Sn 74 are the matrix elements of the relaxation matrix. They are
defined in terms of the spectral densities (J) [equation (78)]:
and it is very difficult to obtain Tn.

However, if Ft, t and t are the FTs of n, Raa0 bb0 D Jaba0 b0 !a0 b0 C Jaba0 b0 !ab a0 b0 Jgbga !gb
Tn and Sn, then [equation (75)]: g

Ft D t t 75 ab Jga0 gb0 !gb0 78

g
so [equation (76)]:

Ft The Js in turn are related to the appropriate correlation
Tn D FT oft D FT of functions via equation (79):
t
C1
FT of n
D FT of 76 Jaba0 b0 ! D 1/2 hajH1 tjbihb0 jH1 t C tja0 i
FT of Sn 1
Hence we obtain the true spectrum via the experimental expi!t dt 79
spectrum n and the slit function Sn.
Another, and very important, application of these rules We now wish to consider which terms will be rele-
is to the technique of resolution enhancement. This is vant to chemical exchange. As an example, it is fairly
described in great detail in the article by Saarinen and straightforward62,63 for the two-site fast exchange. The
Kauppinen (see Resolution Enhancement Approaches). energy levels are limited to a, a0 !aa0 D !A and b,
b0 !bb0 D !B . In this case the required density matrix is
given by equation (80):
APPENDIX 3 REDFIELD RELAXATION
METHOD Rabab D 2Jaabb 0 Jaaaa 0 Jbbbb 0
1 !A !B 2 80
(From Turner et al.64 by permission of the American D
8 k
Chemical Society.)
Section 4.1 describes the modified Bloch equation This makes a contribution to the bandwidth of (p/2k)(n)2 ,
approach to chemical exchange for an uncoupled two- exactly the same [equation (55)] as for fast exchange using
proton NMR example. Inclusion of spinspin coupling Bloch equations.
Mid-infrared Spectroscopy of the Condensed Phase
John M. Chalmers
The University of Nottingham, Nottingham, UK
1 INTRODUCTION AND SCOPE the rotational levels are too close to resolve.1,2 On con-
densation from a gas to a liquid, a rotationalvibrational
As pointed out in the next article in this handbook (see contour reduces to a single band with an approximately
Raman Spectroscopy of the Condensed Phase), by Neil Lorentzian contour, which narrows, intensifies and becomes
Everall, much of the discussion on condensed phase spectra more Gaussian in shape on solidification. Figure 1, repro-
is common to both mid-infrared and Raman spectra. Indeed, duced from a book published in 1963,1 illustrates the effect
it is intended that the reader interested in condensed phase on an absorption band contour on going from the gas to
examinations by vibrational spectroscopy techniques should liquid phase.
read both this article and its Raman counterpart. The lat- Some intermolecular interactions, such as hydrogen bond-
ter was composed first, and this article seeks to minimize ing or solutesolvent interactions, will likely have signifi-
duplication (some will be inevitable) of what is presented cant effects on the detail of a spectrum. Association will lead
in the corresponding discussion of Raman spectroscopy. Its to position shifts and changes in intensities for some bands,
purpose is to complement the Raman article, concentrat- and, in the case of hydrogen bonding, the appearance of new
ing more on the peculiarities associated with mid-infrared bands1 (see Section 3.4).
spectroscopy of condensed phases.
The references cited in this article are far from com- 3 LIQUID- AND SOLID-PHASE
prehensive, but inserted merely as examples of sources for MID-INFRARED SPECTRA
more detailed information.
This section will discuss briefly many of the condensed-
phase molecular arrangements that may (or may not) influ-
2 DIFFERENCES BETWEEN LIQUID- ence significantly the vibrational spectrum recorded from a
AND GAS-PHASE SPECTRA sample.
The absorption bands of a substance in the liquid state

are markedly different from those of the material in the 3.1 General
gaseous state. The fundamental reason is that in the vapor
It is commonly said that the mid-infrared (and Raman)
phase, molecules are free to rotate and rotational fine
spectrum of a material is a unique fingerprint of that
structure may be observed. With increasing pressure, the
material. This is a little overstated, and only half the truth.
number of molecular collisions increases, with the con-
It is, in fact, a unique fingerprint of the specimen of
sequence that bands in the gas phase become broadened
that material being examined, under the conditions and at
and the fine structure is lost, so that only the contours of
the time at which the spectrum was recorded. This is a
the rotationalvibrational bands are observed. For heavy
particularly important consideration for materials that may
molecules, again only the contours may be observed, since crystallize, especially into more than one form.
The mid-infrared spectrum recorded from a solid material
John Wiley & Sons Ltd, 2002. is composed of absorption features characteristic of the
R branch P branch
and Hydrates). Both these publications are extensively
referenced, and are recommended to the reader particularly
interested in more detailed information and examples on
(a)
polymorphism and its effects on vibrational spectra.
As a consequence of the different packing, each polymor-
phic form will have a unique vibrational spectrum, although
in some cases the distinctions between each spectrum may
only be slight, involving only small variations in some peak
R branch P branch positions, peak shapes and/or peak intensities. Since such
properties as dissolution rate, which is of particular impor-
tance in the pharmaceutical industry, can vary dramatically
(b) between polymorphs, their characterization and analysis is
Absorption
very important. Vibrational spectroscopy techniques mid-

max infrared, Raman, near-infrared are key tools, since they
can be used to distinguish between polymorphic forms and,
in some cases, assay mixtures. However, one must exer-
cise caution, since interconversion between forms can occur
using different sampling procedures (see Polymorphs, Sol-
vates and Hydrates and references therein).
(c) Figure 2 shows the FT-IR spectra of two enantiotropic
polymorphs of delavaridine mesylate,4 the anhydrate forms
max
Wavelength /m
Figure 1. Example of the transition of an absorption band profile
Form XII
of a compound from (a) gas phase at pressure p1 to (b) gas phase
at pressure p2 > p1 to (c) liquid. [Reproduced from Rao (1963),1
by kind permission of Academic Press.]
material, in its physical and morphological states at the

time the spectrum was recorded, influenced by the method
of sample preparation, presentation and perhaps its contain-
Transmittance
ment, and its environment, e.g. temperature and pressure. Form VIII
There are many examples within the pages of this handbook
that discuss experimentally induced temperature, pressure
and/or time-dependent changes in spectral features. And, a
few examples of non-deliberately induced time-dependent
changes in solid samples are given in the article in Vol-
ume 3, Mid-infrared Spectroscopy: Anomalies, Artifacts
and Common Errors.
Form XI
3.2 Polymorphism
Many compounds organic, inorganic, polymeric in the

solid state may crystallize into two or more different forms. 3800 2800 2000 1800 1600 1400 1200 1000 800 600
This phenomenon, in which the same molecules in the
Wavenumber /cm1
solid state pack differently, is known as polymorphism. A
review3 of the analysis of organic polymorphism, published Figure 2. FT-IR spectra of delavaridine mesylate forms VIII, XI
and XII. Spectra have been offset for clarity. [Reproduced from
in 1995, gives a much more detailed introduction to poly- Bergren et al. (1996),4 copyright (1996, John Wiley & Sons,
morphism. The author of that review has also contributed Inc.). Reprinted by permission of Wiley-Liss, Inc., a subsidiary of
an article to this handbook (see Polymorphs, Solvates John Wiley & Sons Inc.]
Mid-infrared Spectroscopy of the Condensed Phase 3
VIII and XI, analyzed as dispersions in mineral oil. The 1.0

spectra are clearly distinct. The spectrum of the other form
0.8
shown, which is more correctly identified as a pseudopoly-
Transmittance
morph, is that of form XII, which exists as a crystalline 0.6
hydrate.4 Examples of polymorphism in organic polymers
have been discussed in other articles in this Handbook (see 0.4
SpectraStructure Correlations: Polymer Spectra and 0.2 Enol
Qualitative and Quantitative Analysis of Polymers and
Keto
Rubbers by Vibrational Spectroscopy). 0.0
4000 3500 3000 2500 2000 1500 1000 500
Wavenumber /cm1
3.3 Tautomerism Figure 3. Infrared spectrum of ethyl propionylacetate showing
presence of both keto and enol tautomers.
In tautomerism there may be an equilibrium between two Oxindole molecule tautomers:
or more structural isomers. Each of these will have a dis-
tinctive spectrum. In the solid state, one of the tautomers
is usually stabilized, and tautomerism is rarely observed.
C C C
However, in the liquid state or solution, while one may N O N OH N OH
H H
predominate, a slight change in a condition such as temper- Form (I), lactam Form (II), lactim Form (III)
ature, solvent or pH may shift the equilibrium. Conversion
between the forms may be rapid and reversible if the equi- Cyanuric acid, tautomeric structure:
librium is disturbed or disrupted. The predominant form or
H
change of equilibrium can be characterized readily from OH N OH O N O
their vibrational spectra through well-defined functional
group vibrations. For instance, the equilibrium mixture of N N HN NH
acetoacetic esters in the liquid or solution contain both ethyl
OH O
acetoacetate, a b-ketonic acid, and b-hydroxycrotonic ester Amido Imidol
(Scheme 1).
Mercaptothiazoline (thiazoline-2-thiol) tautomers:
CH3.CO.CH2.CO2C2H5 CH3.C=CH.CO2C2H5
H
OH N N
H2C H2C
Keto form Enol form C S C SH
H2C H2C
S S
Scheme 1
Thioamide Thio
Observable readily in the Raman spectrum will be

Azobenzene derivative, tautomeric structures:
bands due to both the CDO and CDC entities, while
the infrared spectrum will feature clearly both the CDO R1 OH R1 O
and OH functionality. Figure 3, the mid-infrared spec-

trum of ethyl propionylacetate, shows both tautomeric
forms.5
Awareness of tautomerism clearly has importance to N N HN N
Azo Hydrazo
reaction monitoring and kinetics studies, but should also
be borne in mind when trying to pattern match a recorded Scheme 2
liquid or solution spectrum with a solid-phase fingerprint
library spectrum.
The examples of acetoacetic esters and ethyl propionyl- 3.4 Hydrogen bonding
acetate are of ketoenol tautomerism. Other tau-
tomeric pairs include the lactamlactim, thiothioamide, Although many carboxylic acids may exist as dimers in
amidoimidol and azohydrazone rearrangements. Exam- the vapor state below 100 C, at higher temperatures in the
ples are shown in Scheme 2. The spectrum of each will vapor state or very dilute solution in non-polar solvents (e.g.
feature absorption bands that are characteristic of the func- CCl4 ) molecules with OH and NH groups generally show
tional groups present in the particular tautomer. clear, well-defined characteristic stretching vibration (nstr )
infrared absorption bands at high wavenumbers (i.e. OH,

ca. 3600 cm1 ; NH, ca. 3500 cm1 ). As their solution
concentration is increased, or the solvent changed from (a)
non-polar to polar, or in the solid state, these groups may
undergo associative interaction to form hydrogen bonds.6
This may be depicted as:
Transmittance
(b)
AH----B
A hydrogen bond is formed when an AH group (proton
(c)
donor) interacts with an acceptor group B, and the AH
band is weakened but not broken.6,7 The most common
donor groups are OH (e.g. carboxyl, hydroxyl) and NH
(d)
(e.g. amine, amide). Others, such as SH or PH, gen-
erally give rise to weaker hydrogen bonds, and therefore 3600 3400 3200
induce a less dramatic effect in terms of spectral changes. s (cm 1)
The acceptor groups are typically an unshared electron pair
on an electronegative atom, such as an O, or the p electrons Figure 4. Infrared spectra of the OH stretch of methanol in CCl4
solution and condensed phases. (a) CH3 OH in CCl4 very dilute,
of a multiple bond system.6,7
25 C; (b) CH3 OH pure liquid, 25 C; (c) CH3 OH glass, 20 K;
The spectral changes introduced by hydrogen bonding (d) CH3 OH crystal, 20120 K. [Reproduced from The Hydrogen
are similar for organic and inorganic substances, in solid, Bond by George C. Pimentel and A.L. McClellan 1960 by
liquids, solutions (and gases), in complexes, polymers, George C. Pimentel and Audrey L. McClellan. Used with the
biopolymers, chelates, etc.6 Some of these may be sum- permission of W.H. Freeman and Company.6 ]
marized as:
The most prominent and widely studied feature is the
1. weakening of the AH bond through hydrogen bonding nstr AH by infrared spectroscopy. Indeed infrared spec-
causes a shift in the nstr AH (and its overtones) to troscopy was described by Pimentel and McClellan6 in
lower wavenumber. These shifts vary in range from 1960 as providing a definitive criterion for the detection
ca. 30 to several hundred cm1 ; of H-bonds and direct evidence for the detection of the
2. the nstr shifted absorption band is broadened, and the role of the proton in the association.
observed feature may encompass overlap from several The AH stretching vibration is generally very weak
hydrogen-bonded species, with the half-band width of in the Raman spectrum and by comparison little studied.
the absorption envelope extending over several hundred While both intermolecular and intramolecular hydrogen
cm1 ; bonds may be observed in the infrared spectrum, for an
3. whereas with hydrogen bonding the integrated intensity equivalent nstr shift, the changes in band intensity and
of the fundamental AH stretching band increases by width are generally much smaller for intramolecular bonds.
as much as an order of magnitude or more, that of the At low concentration solutions, hydrogen-bonded spectral
overtones in the near-infrared spectrum decreases only behavior is retained with intramolecular bonded species,
slightly; while inter-association may be lost.6 Studies of intramolec-
4. since deformation modes of the AH group are con- ular bonding have been very informative in the characteriza-
strained by hydrogen bonding, and hence their force tion of enol conformation in solution containing b-diketones
constants increased, they are shifted to high wavenum- and similar.
bers. Although, the magnitude of this shift is appre- Temperature, pressure, solvent and concentration changes
ciably smaller than for the nstr , and no substantial can all have dramatic effects on the hydrogen-bonded nstr
broadening or intensity changes are observed; profile in the infrared spectrum. See, for example, Figures 5
5. new vibrational modes associated with the H----Bstr- and 6. A strong, sharp band near 3600 cm1 indicates a free
etching vibration and deformation are found in the OH in the condensed phase, such as that observed in the
low-wavenumber (far-infrared) region; infrared spectrum of 2,6-di-tertiary butyl phenol, where the
6. vibrational modes of the proton acceptor, B, may OH group is sterically hindered by the bulky tertiary butyl
be shifted to higher or lower wavenumbers, but by groups.
amounts much less than for the donor AH vibrations. A much more comprehensive discussion of hydrogen
Figure 4 compares a CCl4 solution at a low enough bonding and its influences on vibrational spectra may be
concentration that intermolecular hydrogen bonding is min- found in another article (Volume 3) in this Handbook (see
imal with other condensed-phase spectra of methanol.6 Hydrogen Bonding).
Wavenumber / cm1
3704 3448 3226 3030
(f)
9.4 M
(e)
80
1.88
(d) 70
3623
Transmittance
Absorptivity (l mol1 cm1)

(c)
60 9.4.103
4.7 102
50
(b) 9.4 102 4.7 101
40
3342
(a)
30
4.7 101
20
3500 3400 3300

10 1.88 9.4 102
Wavenumber /cm1 4.7 102
Figure 5. Infrared spectra in the NH stretching region of
0
N-methylacetamide in various solvents (0.4 M at 30 C): 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.4
(a) CH2 ClCH2 Cl; (b) CHBr3 ; (c) CHCl3 ; (d) CH3 CCl3 ; (e) CS2 ; Wavelength / m
(f) CCl4 . [Reproduced from The Hydrogen Bond by George
C. Pimentel and A.L. McClellan 1960 by George C. Pimentel Figure 6. Infrared absorption spectra of cyclohexanol solution at
and Audrey L. McClellan. Used with the permission of varying concentrations in CCl4 at ca. 20 C in the fundamental
W.H. Freeman and Company.6 ] OH stretching region. [Reproduced from Vinogradov and Linnell
(1971).7 ]
3.5 Stereoisomers
3.5.2 Geometric or cistrans isomers
3.5.1 Optical isomers (enantiomers) and racemates
Many geometrical stereoisomers occur because of absence
Since the arrangement of neighboring atoms in a pair of of rotation about carboncarbon double bonds, CDC,
enantiomers is the same, their mid-infrared (and Raman) which give rise to cis and trans isomers. Since these
spectra are identical and indistinguishable from that of the molecules have distinct geometries, and are therefore nei-
racemate.5 However, these optical isomers (enantiomers) ther mirror images nor superimposable, each isomer has a
are non-superimposable, since their atomic arrangements distinct vibrational spectrum.5 An example has been con-
are mirror images. They may be distinguished by their sidered for the case of such structures in polybutadienes in
effect (rotation) on plane polarized light, which can be another article in this handbook (see SpectraStructure
measured using a polarimeter. Correlations: Polymer Spectra).
The d (dextrorotatory) form causes the plane of polariza- For symmetrical structures such as trans-1,2 dihalo-
tion, when viewed facing the beam, to be rotated clockwise ethenes then the CDC stretching vibration is infrared inac-
(C) direction; the l (levorotatory) enantiomer rotates the tive and not observed. It is, however, observed in the Raman
plane of polarization in a counter-clockwise () direction, spectrum of the compounds.
when viewed similarly. In a racemic sample, equal amounts
of both the d and l molecules are present. This () mixture,
known as a racemate, may have different physical proper- 3.6 Absorbed water, hydrates and water of
ties to the pure enantiomers. crystallization
The specialized mid-infrared technique of vibrational cir-
cular dichroism (VCD) may be used to interrogate the polar- In addition to most living materials, many other solid
ization sensitivity (handedness) of molecules with chiral materials inorganic and co-ordination compounds, clath-
centers (see Vibrational Circular Dichroism). rates, minerals, organic polymers, for example contain
water. This may be present, for example, as an integral primitive unit cell. The number of new bands, in principle,
part of a structure, such as water of crystallization in inor- can equal the number of molecules in the unit cell, although
ganic compounds, located in frameworks, such as those of not all may be mid-infrared active, and some may be degen-
the aluminosilicates, or form the framework, such as in erate. It results from in-phase and out-of-phase or partly
clathrate structures, or held in equilibrium (absorbed) within out-of-phase motions of the molecules within the unit cell.
a thermoplastic polymer. An example, that of the CH2 rocking mode in long alka-
In the mid-infrared region, lattice water exhibits both an nes, is discussed for polyethylene in another article in this
antisymmetric and symmetric OH stretch, both of which handbook (see SpectraStructure Correlations: Polymer
have high absorptivities, and a HOH bending. The former Spectra).
appears as bands in the region 37003200 cm1 , while the
latter occurs in the range 16401600 cm1 . The deforma-
tion mode is absent in hydroxo (OH ) compounds. The 3.8 Disordered states defect crystals and
stretches may give rise to very narrow, sharp bands when amorphous and glassy solids
present as essentially isolated entities, e.g. non-hydrogen
bonded. Defect-induced and doping pure crystals is a special-
Figure 78 compares the mid-infrared spectra of silica ized subject, pertinent nowadays especially to solid-state
gel, (SiO2 )n .xH2 0, and kaolin clay, Al2 O3 .2SiO2 .2H2 O. physics and outside the scope of this introductory arti-
The figure illustrates clearly two very different absorption cle, but substitution perturbations will introduce spectral
patterns associated with the presence of water within these changes and features that aid understanding particularly of
two inorganic compounds. phonon modes.9,10
As geometric order and spatial regularity decrease, the
number of resolvable distinct bands in a mid-infrared spec-
3.7 Crystals lattice modes, factor group trum of a solid tends to decrease until it resembles that of
splitting, phonons and surface modes the liquid or melt state. A fundamental absorption band will
broaden, because of the lack of orientational and positional
Lattice (external mode) vibrations, both bulk (phonons) and order, as the range of the localized molecular environment
surface modes, of ionic crystals occur almost exclusively in increases. (Although, as highlighted in Section 3.4 of the
the far-infrared (<400 cm1 ) region, and consequently are counterpart Raman article, see Raman Spectroscopy of
therefore not strictly for consideration within the context the Condensed Phase, persisting short-range order that
of this article. For the purpose of introduction, these trans- is below the detection limit of other techniques such as
latory and rotational (librational) motions are adequately X-ray diffraction, can still exist that gives rise to spe-
described in this handbook in the complementary articles cific features in the infrared spectrum. Thus, for example,
by Everall (see Raman Spectroscopy of the Condensed the mid-infrared spectrum of atactic polypropylene, which
Phase) and Griffiths (see Far-infrared Spectroscopy), and is often rubbery or waxy, see SpectraStructure Cor-
further information may be found in the references cited relations: Polymer Spectra, is not identical to that of
within each. amorphous polypropylene.)
For crystals that contain molecular groups, rather than The spectral changes involved with a transition from the
just atoms, factor group splitting of fundamental bands may crystalline through the glassy to the amorphous state may
be observed. Static field (or site group) splitting is a conse- be considered simplistically as a gradual progression. There
quence of the influence the surrounding lattice exerts on the is a loss of features arising from long-range intermolecular
molecule.9,10 Lowering of the local symmetry (compared effects and high symmetry through a phase lacking sym-
with the gas phase) can lead to shifting in frequency of metry, but with long-range effects, to a phase in which the
non-degenerate or splitting of degenerate internal modes. In molecular entities are randomly oriented and dominated by
addition, inactive fundamental modes may become active. short-range order influences. As pointed out in the article
For example,2,11,12 in the mid-infrared spectrum of CaCO3 by Cox and Armstrong (see Structure and Composition of
in aragonite, in which the carbonate ion is in a site of Glasses and Amorphous Materials by Vibrational Spec-
Cs symmetry, n1 (1087 cm1 ) is active and both n3 and troscopy), whereas in a crystal, one might only have to
n4 are split. However, in calcite, in which the carbonate deal with a narrow distribution of force constants, for a
ion has D3 symmetry, n1 is inactive, and n2 (879 cm1 ), glassy material a much broader distribution must be con-
n3 (ca. 1450 cm1 ) and n4 (710 cm1 ) are predicted as sidered. The general result is a simple broadening of the
single bands11,12 (see Figure 8). Correlation field splitting absorption bands in the vibrational spectrum of a glass in
may occur for both degenerate and non-degenerate inter- comparison with its crystalline analog, which is proceeded
nal vibrations when two or more molecules co-exist in the by further broadening and loss of spectral definition in the
2.7 3.0 3.5 4.0 4.5 5.0
0.0
0.1
Transmission
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.5
3800 3600 3400 3200 3000 2800 2600 2400 2200 2000
(a) Wavenumber /cm1
2.7 3.0 3.5 4.0 4.5 5.0
0.0
0.1
0.2
Transmission
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.5
3800 3600 3400 3200 3000 2800 2600 2400 2200 2000
(b) Wavenumber /cm1
Figure 7. Mid-infrared transmission spectra in the OH stretching region of (a) silica gel and (b) kaolin clay. [Reproduced from Nyquist
and Kagel (1997)8 by kind permission of Academic Press.]
100 of mid-infrared radiation, i.e. from about 2 m to 15 m.)

Also, important for polymeric films, fibers and the like, may
80 be the extent of any molecular alignments (orientation).
% Transmission
For reflection studies, a samples refractive index becomes

60
a major property that additionally needs to be taken into
40
consideration. In anisotropic materials, this may be different
in all three of the samples geometric axes.
20 For non-continuous solids, e.g. powders, particle size
and, particularly for external reflection studies, packing
0 density both exert a major influence on the profile and
4000 3000 2000 1500 1000 500
relative intensities of bands observed in a samples mid-
(a) Wavenumber /cm1 infrared spectrum.
100 For quasi-liquid systems, e.g. dispersions, slurries,
latexes, suspensions, and many commercial formulations,
80 if the particle diameter approaches 2 m or greater the
effects of scattering may be observed. Also, in studies using
% Transmission
60 internal reflection spectroscopy techniques, one must be

wary of preferential migration (even sticking) of the solid
40 constituent to the internal reflection element surface. This
might, for example, occur with cells from a culture medium.
20 In this case, however, this effect might be exploited
advantageously in some biomedical studies.
0
4000 3000 2000 1500 1000 500
(b) Wavenumber /cm1 4.2 Powders: particle size and packing density
Figure 8. Mid-infrared transmission spectra of CaCO3 minerals:
(a) calcite; (b) aragonite. [Reproduced from Infrared Transmis- Many mid-infrared sample presentation methods are con-
sion Spectra of Carbonate Minerals, C.G. Jones and B. Jackson, cerned with analyzing powders or powder-like materials,
Chapman & Hall, London (1993). 1993 The Natural History
Museum/Trustees of the National Museums of Scotland.] whether by transmission spectroscopy with the specimen
prepared as a mull or as an alkali-halide disk, or by dif-
amorphous state. (The liquid-to-glass-to-crystal transition fuse reflection or photoacoustic spectroscopy. The quality
spectra of methanol in the region 36003100 cm1 shown of the mid-infrared spectrum recorded in terms of both its
in Figure 4 exemplify some of these effects.)6 relative and overall band intensities, spectral contrast and
sometimes background profile will depend on the size, dis-
tribution and packing density of the powder particles.13,14
4 SAMPLE FORM Many of the effects that differences in these parameters
produce, particularly extremes, on mid-infrared spectra are
In contrast to the previous section, in this section we will highlighted and examples given in another article in this
consider briefly the influences that a samples bulk physical handbook (see Mid-infrared Spectroscopy: Anomalies,
form and its physical environment may impose on a mid- Artifacts and Common Errors), and will therefore not
infrared spectrum. be repeated. Suffice it to say here that, for optimizing spec-
tral purity and contrast then, the largest size of any particle
should be less than 10 m, and preferably below 2 m, i.e.
4.1 General of a size that is less than the wavelength of mid-infrared
radiation.
For continuous solids, the key physical properties that Coarse particles will cause excessive scatter of the radia-
influence the appearance of a mid-infrared spectrum recor- tion, and in transmission measurements will give rise to
ded in transmission will clearly be the specimens thickness, a poor (sloping) background profile that decreases with
but also whether the specimen is of uniform thickness, decreasing wavenumber. Also, since there is an associated
whether it is has a matte or reflective surface(s), and dispersion (anomalous dispersion) of the samples refrac-
whether it is scattering or non-scattering. (Each property tive index through an absorption band, distorted (derivative-
variation is considered in the context of the wavelength like) band shapes may be observed. In transmission, these
are a consequence of the Christiansen effect, while in homogeneous sample with a surface that is optically flat
transflection and diffuse reflection measurements they occur relative to the wavelength of the incident radiation is
as a consequence of front surface reflections. More details composed almost entirely of mirror-like (Fresnel, spec-
may be found in Mid-infrared Spectroscopy: Anoma- ular) reflection. The front-surface reflection spectrum of
lies, Artifacts and Common Errors and elsewhere in this a polished (non-scattering) surface will depend on the
handbook. absorption characteristics of the material. For an organic
material, first-derivative-like bands appear as a consequence
of the dispersion in the refractive index as a consequence of
4.3 Temperature and pressure effects absorption known as anomalous dispersion; see, for exam-
ple, External Reflection Spectroscopy in this handbook.
Not considering extremes, the intensity of a band observed
For ionic crystals such as NaCl or KBr, there is a narrow
in the mid-infrared spectrum of a sample depends on
frequency range, usually in the far-infrared region, over
both the samples temperature and pressure, even if no
which they are totally or nearly totally reflecting.9,10 This
phase transitions or conformational changes occur. With
selective reflection of radiation in the vicinity of very strong
decreasing temperature, the rate of decay of the rotational
absorption bands is known as reststrahlen (residual rays).10
correlation function decreases, and the band will sharpen,
For rough (matte) surface or particulate samples, other
that is its band width will narrow and its intensity increase. contributions should be added to the specular reflection
Similarly, with increasing temperature, disorder increases, spectrum. These include rays that have penetrated the
and the band will broaden and its intensity decrease. A sample and been attenuated by absorption processes before
detailed discussion of band widths has been given in this emerging in random directions from the sample. This
handbook by Turner (see Bandwidths). process is commonly referred to as diffuse reflection (and
Similarly, with increasing pressure, the system becomes occasionally as KubelkaMunk reflection). The sum of
more constrained, and bands will again sharpen. Pressure- the specularly and diffusely reflected components has been
induced absorption band shifts in the mid-infrared spectrum called volume reflection; see, for example, Nyquist et al.8
of a solid may be a consequence of a phase change or and Vibrational Spectroscopy of Polymer Composites by
deformation in a crystal lattice15 or may simply be a result Cole in this handbook. Figure 919 compares the reflection
of the changing dielectric constant of the sample. The shift spectra recorded from a polished surface of gypsum and
in internal modes is small, 10 cm1 per 10 kbar pressure.15 calcite with reflection spectra recorded from samples of
Larger shifts are observed for external modes or hydrogen- each of neat grains of differing sizes. Also displayed for
bonded bands. (As with Raman spectroscopy see the next comparison are transmission KBr disk spectra. (Nowadays,
article pressure profiles have been built up by measuring since the common utilization of Fourier transform infrared
shifts in pressure-sensitive bands in the lubricant spectrum (FT-IR) spectrometers for recording mid-infrared spectra,
measured in situ by mid-infrared microscopy).16,17 diffuse reflection has become more synonymous with the
sampling practice of recording a reflection spectrum from
4.4 Effects of stress and strain a particulate-like sample dispersed at low concentration
(5 w/w%) in a finely ground, non-absorbing matrix.)
Strain, like compression, will change the force constants Matte surfaces, at a level appropriate to mid-infrared
and equilibrium state of a solid sample and induce band wavelengths, can be beneficial in negating the formation
shifts, and perhaps conformational changes that give rise of interference fringes in thin, continuous samples, see
to new bands at the expense of others. Thus the frequency Section 5.2.
and shape of skeletal vibrations are stress-dependent, and,
for instance, many studies of deformation of polymers
have been undertaken.18 Additionally, the molecules may 5 TRANSPARENT SOLIDS
reorient and align preferentially in the direction of the
Continuous, non-scattering solids, whether amorphous or
applied stress (see Section 5.1). Several dynamic elongation
crystalline, in which the molecules are axially ordered may
studies are discussed in Rheo-optical Fourier Transform
exhibit orientation effects, that is their spectrum may show
Infrared Spectroscopy of Polymers in this handbook.
differences in relative band intensities depending on how
they are aligned within the spectrometer (see Section 5.1).
4.5 Matte and polished surfaces Mid-infrared spectra recorded from thin, visually trans-
parent or translucent solid sample specimens often have
In the regions where there is no absorption, the reflec- superimposed on them a sinusoidally varying interference
tion of mid-infrared radiation from an optically thick pattern (see Section 5.2).
10
Wavenumber /cm1
Wavenumber /cm1
2400 2200 2000 1800 1600 1400 1200 1000
1800 1600 1400 1200 1000 800
1.0 100 0.6 60
Transmission
80 in KBr disk
0.8 0.4 40
Reflectance
Transmission
of polished
in KBr disk
surface 60
% Transmission
0.6 0.2 Expanded reflectance 20
Reflectance
Reflectance of polished surface
40
Reflectance
0.0 0
%Transmission
0.4 5 6 7 8 9
0.008 Wavelength / m
20
1 1: 250 1200 m
0.2 2: 74 250 m
0.006 3: 0 74 m
0.0 0 4: <5 m
5.5 6 7 8 9 10 11 12 14 2
Wavelength / m
1 1: 250 1200 m 0.004
0.03
2: 74 250 m
Introduction to the Theory and Practice of Vibrational Spectroscopy
2 3: 0 74 m
Reflectance
0.02 4: <5 m 3
3
0.002 4
4
0.01
Reflectance
0.00 0.000
1800 1600 1400 1200 1000 800 2400 2200 2000 1800 1600 1400 1200 1000
Figure 9. (a) Calcite mid-infrared spectra: top, KBr disk transmission spectrum and a reflectance spectrum from a polished sample; bottom, reflectance spectra from matte
surfaces of various grain size fractions. (b) Gypsum mid-infrared spectra: top, KBr disk transmission spectrum and reflectance spectra (one expanded) from a polished
sample; bottom, reflectance spectra from matte surfaces of various grain size fractions. [Reproduced from Vincent and Hunt (1968)19 by kind permission of the Optical
Society of America.]
5.1 Molecular orientation 1.2
In an anisotropic sample, such as an oriented polymer
Absorbance
1.0
film or birefringent crystal, then the alignment of the
sample axes with respect to the geometric axes of the 0.8
spectrometer may have a significant effect on the rel-
0.6
ative intensities of the bands observed in the recorded
spectrum. This was of particular concern when dispersive 0.4
mid-infrared spectrometers dominated the field, since the 2200 2100 2000 1900 1800
grating element imparted considerable differences in the Wavenumber /cm1
response characteristic (throughput) to vertically or hor-
Figure 10. Linearly polarized radiation FT-IR absorbance spectra
izontally polarized mid-infrared radiation. Consequently, recorded from a 6 : 1 uniaxially drawn polyethylene tube sample.
even if non-polarized radiation was used, for a highly ori- The dashed plot (parallel view) was generated with the electric
ented sample, such as a uniaxially oriented polymer film, vector of the infrared radiation aligned along the sample draw
then differing relative band intensities might be observed axis; the solid line (perpendicular view) was generated with the
depending on whether the sample was mounted in the electric vector of the infrared radiation aligned perpendicular to
the sample draw axis.
spectrometer with its draw axis vertical or horizontal with
respect to the sampling plane. For this reason (and in
many studies of molecular orientation using polarized radi- element will experience differing levels of signal as a
ation), the recommended practice was therefore to mount consequence of the differing levels of signal attenuation
the specimen such that one of its principal orientation by the sample. The most likely consequence of this is
axes was at 45 to the vertical. The intrinsic polarization to affect the equivalence of stray light on the recorded
response difference is nowhere near as pronounced with spectrum, such that quantitative measurements may be
mid-infrared FT-IR spectrometers as with grating spectrom- called in to question. (Some effects of stray light are dis-
eters, but some variation may still be present; therefore, cussed in Mid-infrared Spectroscopy: Anomalies, Arti-
it is still best practice to mount an oriented sample at
facts and Common Errors in this handbook.) Minimal
45 to the sampling plane geometric axes. (In some cir-
wedging may however be introduced sometimes (in qual-
cumstances, such as the examination of single crystals,
itative studies) to negate the appearance of interference
orientation effects may become more evident in studies
fringing.
undertaken with FT-IR microscopy, since the averaging
Continuous, non-scattering solid specimens that are uni-
(smearing of the effect) associated with a bulk sample
formly thick may induce multiple passing of the infrared
is lost.)
beam, when examined in a transmission measurement.
For an oriented sample, studied using linearly polarized
These samples often appear visually transparent or translu-
radiation, the intensity of an absorption band in its mid-
cent. As shown in Figure 11, the infrared beam is caused
infrared spectrum (as with a Raman band, see Raman
to double-pass the specimen, such that a path difference
Spectroscopy of the Condensed Phase) will depend on
is induced between the primary transmitted beam and that
its alignment with respect to the direction of polarization
which emerges after it has undergone double internal reflec-
of the infrared beam. In the infrared, if the direction of the
tion within the sample. These coherent beams will sum
dipole moment change associated with a particular vibration
to generate a sinusoidal interference pattern that is most
is aligned preferentially with the direction parallel with the
noticeable in regions of low absorption. Constructive inter-
electric vector of the polarized radiation, then the band will
ference occurs when the pathlength difference equals an
be more intense than it if it is aligned perpendicularly to this
integer number of wavelengths; the two beams will interfere
direction (see Figure 10). As a consequence mid-infrared
spectroscopy (along with other vibrational spectroscopy destructively when the path difference is an odd multiple
techniques) may be used to probe the state of molecular of half the wavelength (see Mid-infrared Spectroscopy:
orientation within anisotropic samples. Anomalies, Artifacts and Common Errors). Typical of
samples in which this phenomenon may be observed in their
mid-infrared spectrum are thin polymer films and alkali-
5.2 Sample thickness and channel (interference) halide disks, for a sample thickness in the range 1300 m.
fringing (The inverse or complementary pattern exists between the
primary beam reflected from the front surface and the beam
If a continuous sample such as a film is excessively that is reflected back from the back surface of the sam-
wedged, then different areas on the spectrometer detector ple.) The interference fringe spacing is governed by the
Sample
ACKNOWLEDGMENT
Io
I H.F. Shurvell provided the data for Figure 3.

(a) FT-IR Fourier Transform Infrared
VCD Vibrational Circular Dichroism
100
80
REFERENCES
% Transmission
60
1. C.N.R. Rao, Chemical Applications of Infrared Spectro-
40 scopy, Academic Press, Inc., New York (1963).
20 2. B. Schrader (ed.), Infrared and Raman Spectroscopy, VCH
Verlagsgesellschaft mbH, Weinheim (1995).
0 3. T.L. Threlfall, Analyst, 120, 2435 (1995).
4000 3500 3000 2500 2000 1500 1000 500
4. M.S. Bergren, R.S. Chao, P.A. Meulen, R.W. Sarver, M.A.
(b) Wavenumber /cm1
Lyster, J.L. Havens and M. Hawley, J. Pharm. Sci., 85(8),
Figure 11. (a) Schematic showing the process leading to the gen- 834 (1996).
eration of interference fringes in the mid-infrared spectrum of a 5. J.B. Lambert, H.F. Shurvell, D.A. Lightner and R.G. Cooks,
continuous solid sample. I0 represents the incident infrared beam; Organic Structural Spectroscopy, Prentice-Hall, New York
I is the transmitted beam. (Also shown is the equivalent pro- (1998).
cess in reflection, which is 90 out of phase with the transmitted
sinusoidal fringe pattern.) (b) An infrared transmission spectrum 6. G.C. Pimentel and A.L. McClellan, The Hydrogen Bond,
of a polymer (polypropylene) thin film showing interference W.H. Freeman & Co., San Francisco (1960).
fringes. 7. S.N. Vinogradov and R.H. Linnell, Hydrogen Bonding, Van
Nostrand Reinhold, New York (1971).
8. R.A. Nyquist and R.O. Kagel, Infrared Spectra of Inorganic
equation: Compounds (380045 cm1 ), Academic Press, New York,
N Vol. 4 (1997).
tD 1 9. P.M.A. Sherwood, Vibrational Spectroscopy of Solids,
2nnQ 2 nQ 1
Cambridge University Press, Cambridge (1972).
where t is the sample thickness, n its refractive index, and N 10. A Finch, P.N. Gates, K. Radcliffe, F.N. Dickson and
the number of complete fringes between the wavenumbers F.F. Bentley, Chemical Applications of Far Infrared Spec-
nQ 2 and nQ 1 , where nQ 2 > nQ 1 . troscopy, Academic Press, London (1970).
11. C. Pettinari and C. Santini, IR and Raman Spectroscopy of
Inorganic, Coordination and Organometallic Compounds, in
6 SUMMARY Encyclopedia of Analytical Science, ed. A. Townshend,
Academic Press, London, 10211034 (1999).
The aim of this article has been to present a brief overview 12. K. Nakamoto, Infrared and Raman Spectra of Inorganic and
of some of the effects and considerations necessary to an Coordination Compounds, Part A: Theory and Applications
in Inorganic Chemistry, 5th edition, J. Wiley & Sons, Inc.,
understanding of condensed-phase mid-infrared spectra. As New York (1997).
mentioned at the outset, in should be read in conjunction
13. P.R. Griffiths and M.P. Fuller, Mid-infrared Spectrometry of
with its Raman and near-infrared counterparts in this Powdered Samples, in Advances in Infrared and Raman
handbook. Many of the discussions have been necessarily Spectroscopy, eds, R.J.H. Clark and R.E. Hester, Heyden &
concise and simply presented, since more comprehensive Son Ltd., London, 63129, Vol. 9 (1982).
explanations of many of the effects may be found in 14. J.M. Chalmers and M.W. Mackenzie, Solid Sampling Tech-
other articles within this handbook, or the references niques, in Advances in Applied Fourier Transform Infrared
they cite. Spectroscopy, ed. M.W. Mackenzie, J. Wiley & Sons,
Chichester, 105188 (1988).
To reiterate the message for the Raman spectra of
condensed-phases (see Raman Spectroscopy of the Con- 15. R. Whyman, Techniques for Vibrational Spectroscopic Mea-
surements under High Pressures, in Laboratory Meth-
densed Phase), in addition to chemical structure, sample ods in Vibrational Spectroscopy, eds, H.A. Willis, J.H. van
morphology will be key to deciding the appearance of the der Maas and R.G.J. Miller, J. Wiley & Sons, Chichester,
mid-infrared spectrum of a condensed-phase sample. 281307 (1987).
16. P.M. Cann and H.A. Spikes, STLE Tribol. Trans., 34(2), 248 18. H.W. Siesler and K. Holland-Moritz, Infrared and Raman
(1991). Spectroscopy of Polymers, Marcel Dekker, Inc., New York
17. P.M. Cann, B.P. Williamson, R.C. Roy and H.A. Spikes, J. (1980).
Phys. D: Appl. Phys., 25(1A), A124 (1992). 19. R.K. Vincent and G.R. Hunt, Appl. Opt., 7(1), 53 (1968).
Raman Spectroscopy of the Condensed Phase
Neil J. Everall
ICI plc, UK
1 INTRODUCTION AND SCOPE bands can be analyzed to yield information on molecular

dynamics specifically, the time taken for molecules to
The factors involved in obtaining and interpreting Raman reorient themselves in solution (the rotational autocorre-
spectra of liquids and solids differ significantly from those lation function).1 This can provide useful information on
pertaining to gas phase spectroscopy. This article introduces intermolecular interactions.
some of the main issues that affect Raman studies of the For noninteracting, randomly aligned molecules in the
condensed phase. This is not an exhaustive listing, or even liquid phase, the molecular point group determines the
a particularly detailed treatment of any given topic; most Raman activity of any given vibration. In other words, only
of the issues raised here are discussed in detail elsewhere intramolecular symmetry is important in determining selec-
in this Handbook. Note also that this section does not tion rules (although the intensity of the band depends on
consider instrumental aspects of Raman sampling this is the nature of the vibrating chemical bonds). Prediction of
also dealt with elsewhere in the Handbook (see Sampling band activity is straightforward using the tools of group
Considerations for Raman Spectroscopy). theory.2 When molecules interact strongly, for example
To provide a flavor for the topics covered in this section, through hydrogen bonding, the molecular symmetry is per-
Table 1 summarizes the most important differences between turbed and selection rules can change. In particular, the
the vibrational spectra of solids, liquids and gases, most vibrations of the functional groups involved in the inter-
of which are discussed below. Note that many of these action are strongly perturbed, meaning that bandshapes,
differences are common to both Raman and IR (Infrared) positions and intensities can all change. IR and Raman
spectra. spectroscopy are therefore powerful tools for studying inter-
molecular interactions.
2 DIFFERENCES BETWEEN
LIQUID AND GAS PHASE 3 SOLIDS
SPECTRA (SEE TABLE 1) 3.1 General
Molecules in the vapor phase give pure rotational, and rota- The situation becomes much more complex in solids, where
tionalvibrational, Raman spectra containing sharp lines atoms, ions or molecules may be distributed with either
due to each rotational energy level. For small molecules, no long-range order (amorphous), perfect order in one,
analysis of gas phase spectra gives detailed quantitative two or three dimensions (crystalline), or a mixed state
information on bond lengths and bond angles. Molecules in (semicrystalline). For example, polymers often exhibit the
liquids and solutions are hindered rotors and do not show latter morphology, with parts of a chain residing in ordered
discrete rotational levels we observe relatively broad crystals, and parts being randomly aligned, in both intra-
bands with little fine structure. However, the shape of these and intermolecular senses. Both IR and Raman spectros-
copies are extremely sensitive to the state of order in
John Wiley & Sons Ltd, 2002. solids.
Table 1. Summary of the main differences between the vibrational spectra of gaseous and condensed phases.
Sample form Main features
Gas Noninteracting free rotors, with vibrational selection rules determined by point group
of molecule.
Pure rotation and rotationvibration spectra, with discrete rotational/vibrational
levels.
Spectra give quantitative details on bond lengths, angles etc.
Liquid (solution, melts) Hindered rotations no discrete rotational/vibrational levels.
Band activity determined by point group.
Bandshapes related to molecular dynamics (reorientation) rather than intramolecular
geometry.
Random alignment of molecules polarization analysis gives restricted information
on symmetry of vibration and relative magnitude of tensor elements.
Can show strong intermolecular interactions (e.g. H-bonding, dipole-dipole, etc).
Solid (amorphous) Strong intermolecular interactions.
Lack of long-range periodicity broad bands.
Boson peaks (low-frequency scattering) can be used to measure short-range
amorphous domain sizes.
Solid (crystalline) Long-range periodicity in one, two or three dimensions yields sharp bands.
Band activity determined by space group; point group symmetry is lowered, giving
band splitting, new bands etc.
Different crystal forms (polymorphs) generally give different spectra.
Precise molecular alignment polarization analysis gives greater detail on vibrational
symmetry and tensor elements.
Vibrating species are coupled over long ranges to give quantized, cooperative
vibrations (phonons). Phonon frequency depends on phasing of vibrations and
strength of coupling (dispersion curve).
Crystal size influences band activity, frequency and intensity (both intrinsic and
observed).
Solid (semicrystalline, e.g. polymers) Long-range periodicity in one, two or three dimensions yields sharp bands for
crystalline phase.
Generally, crystallinity is less than 100% so spectra have mixed
amorphous/crystalline characteristics.
Molecular conformation differs in amorphous and crystalline regions, and generally
gives rise to different spectra.
Band activity determined by line group for extended chains and space group for
chains packed in crystals. Chain packing usually has less effect on spectrum than
chain conformation.
Polarization analysis of oriented systems aids band assignment to symmetry classes.
Imperfect molecular orientation may be quantified on a statistical basis (orientational
averages) using polarized intensities.
3.2 Crystals: lattice modes, sites, factor groups, In a crystal, the vibrational selection rules depend upon
and phonons the symmetry and occupancy of the unit cell, as well
as the molecular symmetry (Figure 1). In general, when
In addition to perturbed intramolecular vibrations discussed a molecule is placed at a lattice site of lower symme-
below, we also observe lattice and librational modes, where try than the point group, the molecule will be perturbed
whole molecules vibrate and twist about their equilibrium and more bands can appear due to removal of vibra-
lattice positions. These modes tend to occur at low fre- tional degeneracy (band splitting), or from activation of
quencies, so Raman spectroscopy is particularly suited to previously silent modes. These effects can be predicted
their observation. Raman bands can also be obtained from by considering how the molecular symmetry correlates
crystals containing no discrete molecules in these cases, under the site symmetry, with the strength of the inter-
atoms or ions are distributed on a lattice and only lattice molecular interactions governing the magnitude of the
modes are observed. effects.
Raman Spectroscopy of the Condensed Phase 3
There are several approaches for calculating vibrational

Gas or Site Factor Phonon dispersion selection rules for crystals; detailed procedures have been
liquid effect effect given in a classic text.4
3.3 Polymorphism
Energy / cm1
Sometimes a compound can adopt more than one crys-

tal form (polymorphism). For molecular crystals, poly-
morphism sometimes involves rather subtle differences in
molecular packing which only slightly perturb the spec-
trum. For example, Figure 2(a) shows the Raman spec-
tra of two crystal forms of terephthalic acid (TA); note
how the low-wavenumber spectrum easily differentiates the
polymorphs I and II.5 Because no sample preparation is
0 Phase angle required, the TA powder can be examined in situ, in its con-
Figure 1. Schematic representation of energy of a doubly degen- tainer this is important in this case as even slight pressure
erate molecular vibration in different environments: gas or liquid can convert the material between the two forms.
phase, in a crystal with site symmetry which lifts the degeneracy, Raman studies of polymorphism can be of great com-
factor splitting (coupling of vibrations of two molecules in the unit mercial importance. For example, controlling the TA crystal
cell), and the phonon dispersion curve arising from coupling of
form is important as the handling characteristics of the bulk
vibrations in adjacent cells. For weak coupling the vibrational fre-
quency is independent of the phonon phase angle, and for normal
optical spectroscopy phonons are only detected with zero phase
angle.
In addition to site symmetry, we also have to consider (Form I)

Intensity
coupling of the vibrations between molecules in the same

(Form II)
unit cell. The intermolecular coupling causes each of
the vibrations of the molecule to be split into as many
components as there are molecules in the unit cell (factor Pressed
group splitting). Finally, all of the unit cells in a crys- Form I
tal can couple to form a phased array of vibrators; these
60 80 100 120 140
cooperative vibrations are termed phonons. The phase angle
(a) Raman shift /cm1
between corresponding vibrators in adjacent cells can vary
from 0 to 180 , and the phonon frequency depends on
the phase angle, to an extent determined by the strength
of the coupling. The phonon frequencyphase relationship
is described by the so-called dispersion curve;3 weak cou- Anatase
pling means that the phonon frequency is independent of
Intensity
the phase, and the dispersion curve is flat. For optical spec- Rutile
troscopies such as IR or Raman, only phonons with a zero
phase angle can be observed, and so the shape of the dis-
persion curve is normally irrelevant (but see Section 4.2.2
below). This selection rule greatly simplifies the spectrum,
because the many phonons with a nonzero phase angle 100 200 300 400 500 600 700
become silent, and we are left with just a few relatively (b) Raman shift /cm1
sharp bands, having the frequency of the phonons at zero Figure 2. Illustrations of detection of polymorphism using
phase angle. The site, factor and phonon effects are sum- Raman spectroscopy. (a) Low-frequency Raman bands of
marized in Figure 1, which illustrates schematically the terephthalic acid polymorphic forms I and II. Note that light
pressure on form I converts it into a mixture of I and II. (b) Raman
changes which can occur to a doubly degenerate vibrational spectra of the anatase and rutile forms of TiO2 . The two forms
mode when two molecules are incorporated into a unit cell are readily distinguished, and low levels of anatase are readily
of low symmetry. detected in rutile-rich mixtures.
powders are very different for the two polymorphs. In the low-wavenumber (below 50 cm1 ) Raman scattering; Achi-
pharmaceutical industry, polymorphism often determines bat et al. have inferred the existence of approximately 5 nm
drug activity, and measurement and control can be critical.6 blobs in amorphous poly(ethylene terephthalate) (PET) and
In the chemical industry, on-line Raman spectroscopy has poly(methymethacrylate), based on the presence of the so-
been used to control the production of titanium diox- called Boson peak near a 10 cm1 Raman shift.13 This peak
ide, where the compositional balance of two crystal forms is assigned to the vibration of a discrete domain, in direct
(anatase and rutile) is critical.7,8 This is a good example analogy with the spectra of inorganic glasses the funda-
of a system where there is no discrete molecular vibrat- mental vibrational frequency of the domain is proportional
ing unit. In such cases, changing the lattice can grossly to v/r, where v is the velocity of sound in the blob, and
perturb its vibrations, and Raman spectroscopy is then r is its size.14 It turns out that the size is very close to
tremendously sensitive to polymorphism throughout the the entanglement distance in the melt from which the poly-
spectrum (Figure 2b). As a third example, Raman spec- mer solidified. In fact, as well as Boson peaks, amorphous
troscopy is used extensively in the semiconductor industry materials also exhibit rather broad Raman features below
as a quality assurance technique to monitor the morphol- 200 cm1 , the interpretation of which is far from straight-
ogy of diamond-like carbon coatings which are applied as forward and outside the scope of this paper.13
protective layers.9,10
4 INFLUENCE OF SAMPLE FORM

3.4 Disordered solids
4.1 General
In the extreme case of a disordered material such as an
inorganic glass or an amorphous polymer, translational
Because it is often possible to obtain Raman data without
periodicity is very low and a wide range of local molec-
sample pretreatment, it is tempting to neglect the influence
ular environments exists, giving rise to very broad fea-
of the bulk sample form on the resultant spectrum. Pow-
tures in place of the sharp bands observed from crys-
ders, pellets, single crystals, mouldings, films and fibres
tals. Figure 3 illustrates this by comparing the spectra of
can all be simply illuminated in a laser beam and the scat-
fused silica and crystalline quartz many of the crystalline
tered radiation collected to yield a high-quality spectrum.
bands are present in the amorphous spectrum, but they are
However, Raman band intensities, shapes and positions can
greatly distorted. However, even amorphous materials can
be influenced significantly both by the form of the sample,
exhibit some short-range order (i.e. approximately nanome-
and how it is presented to the spectrometer. The user must
ter scale), and Raman spectroscopy has been employed to
be aware of these effects, particularly when carrying out
study these disordered systems.11,12 For example, the size
quantitative work.
of these semi-ordered domains can be assessed by analyzing
Silica 4.2 Particle size and shape
4.2.1 Size-induced activation of forbidden bands

Particle size and shape can have a major influence on the
Raman intensity
Quartz Raman spectra of powders. The exact details depend on

the size of the particle relative to the optical wavelength,
and whether particles are single crystals or agglomera-
tions of crystallites. For simplicity, we first consider the
effect of the size of discrete crystals. Strictly speaking,
the simple group theory rules for calculating band activ-
ity in crystals only hold for infinite lattices, which have the
500 1000 1500 2000 2500 full translational symmetry of the space group. When crys-
Raman shift /cm1 tals become sufficiently small, their symmetry is no longer
described accurately by the space group, and so previously
Figure 3. Comparison of Raman spectra of SiO2 in its amorphous
(fused silica) and crystalline (quartz) forms. The amorphous spec- inactive (i.e. symmetry-forbidden) bands can sometimes be
trum has greatly broadened bands due to the range of molecular observed. By considering the size-induced change in crystal
environments and lack of long-range order. symmetry, it is possible to predict which bands may become
I(110 cm1)/I(441 cm1) if the dispersion curve is not flat, the Raman band will be
broadened and shifted according to the dispersion gradi-
ent and the uncertainty in the phase angle. This topic has
been the subject of much study, and is well understood on
a quantitative basis such that the bandshape can be pre-
Raman intensity
1/crystal size (nm) dicted, given a knowledge of the dispersion curve. As a

0.02 0.04 0.06 0.08 0.10 0.12 0.14 result, phonon confinement effects have been used widely
to quantify crystal size and perfection in nanophase materi-
als, for example in semiconductor and inorganic chemistry
research.3,18 The effects can be relatively large; for exam-
ple, the 142 cm1 phonon of bulk anatase (TiO2 ) shifts
Reducing crystal size to 150 cm1 , and broadens from 8 cm1 to 18 cm1 ,
when the crystal size is reduced from 25 nm to 6 nm
100 200 300 400 500 600 700
(only about 6 unit cells in the latter case).18
Raman shift /cm1
Figure 4. Measuring crystal size in nanophase rutile. A band near
110 cm1 correlates with the reciprocal crystal size. 4.2.3 Surface modes
activated by this so-called linear-q effect, although the Surface modes are only observed with ionic crystals that are
analysis is far from trivial.3 significantly smaller than the wavelength of the incident
We would expect the onset of these crystal size effects light. With these systems, the polarization charge on the
to occur once a significant fraction (say 10%) of the atoms, crystal surface can cause the appearance of new modes,
ions or molecules reside at a surface rather than in the termed surface phonons, which occur at higher frequency
bulk. Rough calculations indicate that this situation arises than that observed with large crystals. (The terminology
for crystal diameters of the order of 10 nm or less, i.e. can be a little misleading in that a given crystal often
the so-called nanophase regime. As an example, Figure 4 supports multiple surface modes, but only the higher order
shows the Raman spectra of nanophase rutile as a func- modes are actually localized near the crystal surface
tion of crystal size. In this case, a previously inactive mode the lowest-order mode has uniform polarization throughout
near 110 cm1 is observed, and its Raman intensity is pro- the crystal.) Because the local field depends on the shape
portional to the inverse crystal diameter.15 This mode can of the particles, the powder density, and the dielectric
be reasonably assigned to the Raman-inactive B1u mode, constant of surrounding medium, the Raman spectrum is
which has previously been observed at around 110 cm1 in also sensitive to these parameters.19 For example, surface
large crystals using inelastic neutron scattering.16 Its inten- phonons have been observed in Ge microcrystals of 100 nm
sity grows in proportion to the surface to volume ratio, or lower diameter; they grow in intensity as crystal size
hence the 1/d size dependence (d, diameter). Note how is reduced, and shift to lower frequency as the dielectric
the intensity grows rapidly for crystal sizes below 15 nm, constant of the surrounding medium increases.20 A detailed
in good agreement with the threshold estimate above. The discussion of surface modes and related phenomena has
study of graphitic carbon is another classic example of been given elsewhere.21,22
crystal sizing, where the Raman-silent A1g mode becomes It is worth noting that surface modes are much more com-
activated with a 1/d intensity response.17 monly observed in mid-IR spectroscopy, simply because
they occur for micron-scale rather than nanophase crystals.
The IR spectroscopist needs to be aware of their existence
4.2.2 Phonon confinement as they complicate the analysis of ionic crystal spectra
As well as the activation of forbidden bands, we can also obtained in the usual manner (e.g. thin films, salt discs,
observe vibrations with a nonzero phase; this is manifested mulls). Ruppin and Englman have given a comprehensive
by band broadening and shifting, the magnitude of which discussion of these effects and showed how historical IR
is governed by the slope of the phonon dispersion curve data could be reinterpreted in light of this theory.21
near zero phase angle (Figure 1). This phonon-confinement
effect is a manifestation of the uncertainty principle local-
4.2.4 Lattice defects
izing a phonon in a sufficiently small region of space
introduces an uncertainty into its momentum, which cor- The introduction of defects into a crystal lattice gives rise
responds to an uncertainty in its phase angle. Therefore, to many of the effects noted above the lattice periodicity
phonons with a nonzero phase can become allowed and, is disrupted, activating forbidden modes, and phonons are
confined, causing band shifting and broadening. At high approach could be invalid if the particle sizes of the individ-
defect levels, spectra of the defects themselves become ual components varies from sample to sample the particle
apparent. Raman spectroscopy is widely used to analyze sizes in a calibration set must be similar to those of the real
defect structures in semiconductors,3 and it has also been samples, or the intrinsic Raman intensity of each component
used to characterize defects in insulators such as titanium will not be constant.
dioxide.23 Finally, as particle size drops significantly below 1 m, it
is generally observed that the intrinsic scattering efficiency
4.2.5 Size effects in polycrystalline particles falls dramatically for nonsurface modes thus 10 m par-
diffuse reflection ticles will usually have a much stronger intrinsic Raman
The effects described above only arise in Raman spec- spectrum than 0.01 m particles of the same material.26 The
troscopy when examining exceedingly small crystals, the same general effect occurs in absorption spectroscopy.
formation of which requires special conditions (mechanical
grinding rarely suffices). The study of nanophase systems
is very important, but many Raman spectroscopists are 4.3 Stress and strain
unlikely to encounter such systems, either as discrete crys-
tals or agglomerates. Unfortunately, this does not mean that If a solid sample is stressed and strained, there will be
those working outside the nanophase regime can completely changes in the force constants and equilibrium geome-
neglect size effects, because the overall scattering intensity try, which together determine the vibrational frequencies
can still be strongly dependent on particle size and packing. of the molecule and/or lattice. If this yields a signifi-
This is because both the laser beam and the Raman radiation cant, measurable bandshift, it is possible to calibrate the
will suffer multiple elastic scattering events in a polycrys- band position by straining a test sample by a known
talline powder, so the observed Raman intensity will depend amount, and then use Raman spectroscopy to measure the
upon the diffuse reflectivity of the sample. Schrader has band position in situ within a material of interest. For
summarized the treatment of Raman scattering using an example, Raman microscopy has been used to measure
extended KubelkaMunk theory, which relates forward and the stress in fiber-reinforced composites and in epitax-
backscattered Raman intensities to the incident laser power, ially grown semiconductors,3 and to map these on the
sample thickness, sample absorption, elastic scattering coef- micron scale near interfaces and defects. Perhaps surpris-
ficient and Raman scattering coefficient.24 The treatment is ingly, molecules in pressurized liquids can be perturbed
quite complex, but the basic results are: sufficiently to cause band shifts and removal of degeneracy
(band splitting) this provides a Raman probe for hydro-
1. the Raman intensity decreases as particles get smaller; static pressure and its effect on molecular geometry. Studies
2. for samples with a low absorption coefficient, the are usually carried out in diamond anvil cells on bulk
Raman/Rayleigh ratio is maximized by using coarse samples,27 but Raman microscopy has been ingeniously
powders in a forward scattering arrangement; applied to measure and map pressure fields in lubricated
3. for strong absorbers, one should use coarse powders rolling contacts.28
and backscattering geometry.
Irrespective of the validity of statement 2 above, most

modern instruments are configured to work solely in
5 TRANSPARENT SOLIDS
backscattering geometry. This configuration is very sim- MOLECULAR ORIENTATION
ple to implement and align, and allows easy analysis of
transparent samples without flooding the spectrometer with Transparent solids (glasses, single crystals, amorphous or
unshifted laser light. Hendra et al. have demonstrated the small-crystallite plastics) pose different problems to the
validity of statement 1 above, by comparing the Raman analyst. If the sample is isotropic (no preferred molecular
intensity of powders with different particle sizes.25 orientation) the spectrum should be reproducible, irrespec-
Clearly, absolute Raman intensities from powders have tive of how it is illuminated. The intensity of Raman bands
little analytical use because they will be influenced by par- from samples which show either perfect or partial molecu-
ticle size, and likewise the use of a powder reference as an lar orientation (e.g. single crystals versus drawn polymers)
intensity standard is not recommended. Quantitative mea- will depend strongly on the alignment of the polarized laser
surements of heterogeneous powders should be made using field relative to the sample optical axes. Figure 5 illustrates
band ratio methods; these necessitate measuring relative this effect by comparing the spectra of a drawn PET fiber
concentrations of two or more components by ratioing dis- obtained with the laser polarized parallel and perpendicu-
crete Raman bands arising from each component. Even this lar to the draw axis. Clearly, the relative band intensities
of absorber concentration and optical properties.31 Impor-

V tantly, the analysis shows that for a sample that absorbs at
both laser and Raman wavelengths, 180 backscattering is
H greatly preferred if the absorber molecule is also the Raman
Raman intensity
Laser scatterer of interest. If 90 scattering is employed, it is nec-

beam essary to carefully optimize the absorber concentration so
as to maximize its Raman signal, whereas 180 scattering
efficiency reaches a plateau once the absorber reaches a suf-
VV
ficient concentration. In this case, the laser beam will not
penetrate deeply into the sample, so the analysis will be
HH
specific to the sample surface. In this case (and the analo-
gous situation of a highly turbid sample) one should use
900 1000 1100 1200 1300 1400 1500 1600 1700 1800 the widest aperture collection lens available in order to
Raman shift /cm1 obtain the highest Raman signal. With transparent samples,
it can be shown that the signal is fairly independent of the
Figure 5. Effect of molecular orientation and laser polarization
on the spectrum of a highly drawn PET fiber. The term IJ denotes f-number of the collection lens because the probed sample
the polarization of the laser (I) and the Raman analyser (J) volume remains almost constant.32
respectively (the fiber is held vertical (V) in both cases). Note how Self-absorption is also very important in near-infrared
the relative intensities are strongly affected by the polarization. (NIR)-excited Fourier transform (FT) Raman spectroscopy,
The 1612 cm1 band is due to a para-substituted aromatic ring where the NIR absorption of the sample attenuates the
vibration. When the laser and Raman polarizations are parallel to
the C1 C4 ring axis the intensity is strong; if laser and Raman Raman bands. Because the NIR absorption bands can be
scatter are polarized perpendicular to the axis it is weak. The data relatively narrow, only a few of the Raman bands might
show that for the fiber, the ring axes (and hence polymer chains) be affected, causing variations in relative band intensities.
tend to align in the fiber draw direction. For example, this was manifested in the FT Raman spec-
troscopy of tetrahydrofuran, where simply changing the
are very sensitive to orientation, which makes quantitative position of laser focus within the sample by 2 mm caused
analysis of chemical composition difficult. However, the the relative intensity of the 917 cm1 band to vary by a
polarized intensities can be used to quantify the degree of factor of three.33,34 This effect can be modeled, but will
molecular orientation in partially oriented systems, or to obviously lead to potential irreproducibility in band inten-
measure the relative magnitudes of Raman tensor compo- sities in FT Raman spectra.35,36 Clearly, it is necessary to
nents with single-crystal data.29 In some cases, multivariate consider the NIR absorption spectrum of the sample when
calibration methods can be used to simultaneously account building quantitative models based on FT Raman band
for the effects of molecular orientation and other parame- intensities.
ters for example, it has been shown how a single partial
least-squares model can be used to measure crystallinity in
polymers that have widely differing states of orientation.30 6.2 Heating, degradation and fluorescence
Condensed samples are prone to heating in the laser beam,

6 SOME ARTEFACTS THAT CAN ARISE even if they are only slightly colored. In liquids this can
WITH CONDENSED PHASES cause bubbling, reduction in spectral quality, and even sam-
ple decomposition. In solids, subtle phase changes such as
6.1 Self-absorption/color crystal form or conformational modifications can occur. In
addition, laser-induced heating can induce subtle, reversible
If a sample possesses strong absorption bands in the abso- effects such as band broadening and shifting, as is found
lute wavelength region of the Raman scatter, the Raman with carbon.37 Such changes are sometimes hard to avoid
scattered radiation will be attenuated. Sometimes the effect because the intensity at the laser focus can be very high,
is obvious, and expected hence strongly colored samples particularly in Raman microscopy measurements. So, even
are likely to absorb Raman scatter excited by a visible if a solid sample does not visibly melt, burn or otherwise
laser. The absorption bands of colored samples are gen- indicate damage, it is not safe to assume that the laser has
erally broad and so a large region of the Raman spectrum not modified the sample if possible, spectra should be
is suppressed. Turrell has analyzed this effect in detail and recorded using different laser powers to check for changes.
shown how the attenuation can be quantified as a function Note that it is possible to measure the temperature at
the laser focus using the Stokes/anti-Stokes intensity ratio, 4. W.G. Fateley, F.R. Dollish, N.T. McDevitt and F.F. Bentley,
thereby checking that the sample does not get too hot.38 Infrared and Raman Selection Rules for Molecular and Lat-
tice Vibrations: The Correlation Method, Wiley Interscience,
Finally, we note that if laser light is absorbed by
New York (1972).
the sample, this often leads to relatively intense fluores-
5. N. Everall, Industrial Applications of Raman Spectroscopy
cence, which can severely degrade the signal-to-noise ratio
in An Introduction to Laser Spectroscopy, eds D.L. An-
(S/N) of the Raman spectrum. Fluorescence can be min- drews and A.A. Demidov, Plenum, New York, 115131
imized using a number of techniques to either clean up (1995).
the sample chemically (i.e. remove the fluorescor), or to 6. M.C. Davies, J.S. Binns, C.D. Melia and D. Bourgeois, Spec-
suppress the fluorescence signal by distinguishing Raman trochim. Acta A, 46, 277 (1990).
and fluorescence signals on the basis of polarization or 7. J.P. Besson, P.W.B. King, T.A. Wilkins, M. McIvor and
temporal properties. The many possible approaches have N. Everall, European Patent Application, EP 0 767 222 A2
been reviewed in detail elsewhere,39 and an excellent (1996).
example of effective temporal discrimination has recently 8. N. Everall, B. King and I. Clegg, Chemistry in Britian, (July),
appeared.40 However, by far the most successful univer- 4043 (2000).
sal approach is to shift the excitation laser to a longer 9. M.S. Dresselhaus, G. Dresselhaus, M.A. Pimenta and P.C.
wavelength (usually about 785 nm or 1064 nm) to avoid Eklund, Raman scattering in carbonaceous materials in
exciting electronic transitions. These two approaches are Analytical Applications of Raman Spectroscopy, ed. M.J.
discussed in detail elsewhere in this Handbook (see Con- Pelletier, Blackwell Science, Oxford, 367426 (1999).
tinuous Lasers for Raman Spectrometry, Array Detec- 10. P. Dhamelincourt and S. Nakashima, Applications to mate-
tors for Raman Spectroscopy, Fourier Transform Near- rials science in Raman Microscopy: developments and
Applications, eds G. Turrel and J. Corset, Academic Press,
infrared Raman Spectroscopy, FT-Raman Spectroscopy London, 255257 (1996).
and Mid-infrared Spectroscopy: Anomalies, Artifacts
11. R. Shuker and R.W. Gammon, Phys. Rev. Lett., 25, 222
and Common Errors). (1970).
12. R. Shuker and R.W. Gammon, J. Chem. Phys., 55, 4784
(1971).
7 CONCLUSIONS
13. T. Achibat, A. Boukenter, E. Duval, G. Lorentz and S. Etie-
nne, J. Chem. Phys., 95, 2949 (1991).
This brief review outlines the complications and effects
14. E. Duval, A. Boukenter and B. Champagnon, Phys. Rev.
that may be encountered when moving from gaseous to
Lett., 56, 2052 (1986).
condensed-phase Raman studies. The main point to bear
15. N. Everall, J. Mater. Sci., in preparation (2001).
in mind is that the sample form can have a very large
effect on Raman band intensities, shapes and positions, and 16. J.G. Traylor, H.G. Smith, R.M. Nicklow and M.K. Wilkin-
son, Phys. Rev. B., 3, 3457 (1971).
these factors must be taken into account when interpreting
spectra, particularly when drawing quantitative conclusions. 17. F. Tuinstra and J.L. Koenig, J. Chem. Phys., 53, 1126
(1970).
These factors are not merely an inconvenience they also
mean that Raman spectroscopy is very a powerful tool for 18. S. Kelly, F.H. Pollak and M. Tomkiewicz, J. Phys. Chem. B,
101, 2730 (1997).
studying the structure of condensed phases.
19. T.P. Martin and L. Genzel, Phys. Rev. B., 8, 1630 (1973).
20. S. Hayashi, Solid State Commun., 44, 75 (1982).
ABBREVIATIONS AND ACRONYMS 21. R. Ruppin and R. Englman, Rep. Progr. Phys., 33, 149
(1970).
PET Poly(ethylene terephthalate) 22. L. Genzel and T.P. Martin, Surf. Sci., 34, 33 (1973).
TA Terephthalic Acid 23. J.C. Parker and R.W. Siegel, J. Mater. Res., 5, 1246
(1990).
24. B. Schrader, Tools for Infrared and Raman Spectroscopy,
REFERENCES in Infrared and Raman Spectroscopy: Methods and Applica-
tions, ed B. Schrader, VCH, Weinheim, 138143 (1995).
1. R.G. Gordon, J. Chem. Phys., 43, 1307 (1965). 25. P. Hendra, C. Jones and G. Warnes, Fourier Transform
2. F.A. Cotton, Chemical Applications of Group Theory, Raman Spectroscopy, Ellis Horwood, New York, 152155
Wiley Interscience, New York (1990). (1991).
3. F.H. Pollak, Characterisation of Semiconductors by Raman 26. G.J. Rosasco, Raman Microprobe Spectroscopy, in Advan-
Spectroscopy, in Analytical Raman Spectroscopy, eds J.G. ces in Infrared and Raman Spectroscopy, eds R. Clark and
Grasselli and B.J. Bulkin, Wiley, New York, 137 (1991). R.E. Hester, Heydon, London, Vol. 7, 223283 (1980).
27. D.J. Gardiner, Non-standard Physical and Chemical 33. N. Everall and J. Lumsdon, Vib. Spectrosc., 2, 257 (1991).
Environments, in Practical Raman Spectroscopy, eds D.J. 34. C.J. Petty, Vib. Spectrosc., 2, 263 (1991).
Gardiner and P.R. Graves, Springer-Verlag, Berlin, 103111
(1989). 35. B. Schrader, Tools for Infrared and Raman Spectroscopy,
in Infrared and Raman Spectroscopy: Methods and Applica-
28. D.J. Gardiner, M. Bowden and P.R. Graves, Philos. Trans. R.
tions, ed. B. Schrader, VCH, Weinheim, 153156 (1995).
Soc. London Ser. A, 320, 295 (1986).
36. N. Everall, J. Raman Spectros., 25, 813 (1994).
29. D.I. Bower, J. Polym. Sci., Polym. Phys. Edn., 10, 2135
(1972). 37. N. Everall and J. Lumsdon, J. Mater. Sci., 26, 5269 (1991).
30. N. Everall, P. Tayler, J.M. Chalmers, D. MacKerron, R. Fer- 38. B.J. Kip and R.J. Meier, Appl. Spectros., 44, 707 (1990).
werda and J.H. van der Maas, Polymer, 35, 3184 (1994). 39. N. Everall, Design and performance analysis of a picosec-
31. G. Turrell, Raman Sampling, in Practical Raman Spec- ond pulsed Raman spectrometer for fluorescence rejection in
troscopy, eds D.J. Gardiner and P.R. Graves, Springer- Raman spectroscopy, PhD thesis, University of Durham, UK
Verlag, Berlin, 3137 (1989). (1986).
32. N. Everall, H. Owen and J. Slater, Appl. Spectrosc., 49, 610 40. P. Matousek, M. Towrie, A. Stanley and A.W. Parker, Appl.
(1995). Spectrosc., 53, 1485 (1999).
High-resolution Fourier Transform Spectrometry of
Gases
Guy Guelachvili and Nathalie Picque
Universite Paris-Sud, Orsay Cedex, France
1 INTRODUCTION stepping-mode approaches. Section 4 is devoted to a short

discussion, not intended to be comprehensive, of some of
This article mainly focuses on the experimental aspects the important scientific results that may be obtained in
of high-resolution spectroscopy as applied to molecular high-resolution vibrational spectrometry using FTS. Finally,
gas studies. The intent is not to give a comprehensive some interesting recent developments for FTS are discussed
review of the topic. Instead we will illustrate, through briefly in Section 5.
selected examples, the main characteristics and peculiar-
ities of high-resolution vibrational spectroscopy of gases,
2 HIGH-RESOLUTION VIBRATIONAL
and we will also consider some of the present trends. By
high-resolution spectra, we mean spectra for which the SPECTRA
recorded lines are shaped largely by the Doppler-limited
line profile, and not by the instrument line shape function. High-resolution vibrational spectroscopy for gas phase stud-
This is obtained, for instance in absorption by a gas at ies is mostly concerned with the rotational fine structure of
low pressure, with an instrumental resolution of the order the vibration of the molecule.2 Vibrationrotation spectra
consist of many individual lines that are mostly shaped by
of 5 to 10 103 cm1 in the 5000 cm1 (2 m) region.
the Doppler effect under the usual low-pressure conditions.
Laser and Fourier transform spectroscopies are modern and
These lines may be observed either in absorption or in emis-
complementary research tools, both of which have made
sion. They result from the detection of photons emitted by
tremendous contributions to the field of high-resolution
the various types of source under study (see for instance
vibrational spectroscopy of gases. However, we will con-
review papers in the literature3 5 ).
centrate in this article on results obtained with Fourier
Comments on common features encountered when
transform spectroscopy (FTS). The reader will easily find
recording high-resolution absorption spectra may be made
all complementary aspects dealing with different spectro-
using Figure 1, which shows a typical single-beam FT
scopic accessories and techniques in this Handbook.1
infrared spectrum in the region of 3 m.6 The absorption
Section 2 gives an introduction to the general appear-
reveals the vibrational n3 band of 12 CH4 observed under
ance of high-resolution absorption and emission Fourier
a pressure of the order of 40 000 Pa (300 Torr) at room
transform (FT) spectra. Emphasis is given to the discussion
temperature in a multi-reflection White-type cell adjusted
of the determination of spectral intensities. After a short
to an absorption path equal to 4 m.
comparison between the respective demands of record-
Figure 1 reveals first a wavenumber-dependent transmis-
ing low- and high-resolution FT spectra, Section 3 will sion of the instrument; at 2400 cm1 the flux of photons
focus on the specific advantages of the two most com- reaching the detector is almost three times greater than that
mon FT recording procedures: the fast-scanning and the at 3200 cm1 . The profile of the spectrum is due mainly to
the joint effects of the source, the interferometer transmis-
John Wiley & Sons Ltd, 2002. sion, the detector sensitivity, the optical filter used to restrict
the spectrum as they would if induced by a window as dis-

cussed above, they are much more pernicious and hard to
deal with using the usual phase correction softwares. These
fringes seriously perturb the spectrum and prevent the pre-
cise measurement of the position and the intensity of the
lines.7 Wedged beamsplitters may solve this problem.
The experimental determination of the intensity of the
vibrationrotation lines depends on whether they are obser-
ved in emission or in absorption. In the latter case, the
measurement of narrow lines may be made relative to the
local background. It gives the relative number of photons
absorbed by a known number of molecules and conse-
quently the absolute line intensity. To be accurate, this
2300 2700 3100 comparison should be done on an exact intensity scale.
Wavelength /cm1 Obtaining a correct zero-energy axis is the most crucial
Figure 1. Infrared absorption due to CH4 . The spectrum reveals part. The spectrum in Figure 1 benefits from having an eas-
different features relatively common to all absorption spec- ily determined zero-energy axis. It is easily verified because
tra: non-uniform background, channeled spectrum (see text for many lines are saturated, i.e. totally absorbing all the pho-
details). Unapodized resolution is equal to 3.5 103 cm1 .
tons near the line center and, therefore, transmitting no
the spectral domain of the measurement, and the detection energy. When no saturated lines occur in the recorded spec-
technique used to record the interferogram signal. This rela- tral range, such a check is not possible. It may be replaced
tively smooth profile may not be troublesome in the process by the observation of a spectral region where no energy is
of extracting the parameters from each single absorption expected, for example below or above the cut-off of an opti-
line, which is made relative to the local background. Since cal filter. Departure from a correct intensity measurement
the noise is normally constant all along any Fourier trans- is most often due to nonlinearity in the detector response or
form spectrum, the absorption lines with the same intensity the electronics used to amplify the recorded interferogram.8
will be measured with signal-to-noise ratio (S/N) propor- The measurement of the intensity of the vibrationrota-
tional to their surrounding background level. To obtain a tion profiles observed in emission may also be perturbed by
flat background, a reference spectrum, with no absorbing nonlinear detection effects, which give rise, because of the
gas, is often recorded to undertake a ratioing procedure. different nature of the interferograms recorded in absorption
The periodic modulation of the background is another and in emission, to perturbations of different nature in the
commonly encountered spectral feature. This modulation, spectra. Measuring the relative or absolute line intensity
which has a period of about 20 cm1 , may be observed is not as straightforward in emission as it is in absorp-
at both high and low wavenumber in Figure 1 where the tion. One can no longer measure the absorption of photons
density of lines is not too high. This channel spectrum by an appropriate comparison to a local available back-
may have several origins. Real optical effects, for instance ground level. Instead one counts the photons emitted from
intensity modulation of the background due to interference a standard light source, since knowledge of the spectral
effects, may cause it when the light traverses an unwedged transmission of the whole experimental set-up is needed.
window. In this case, it is just an additional effect con- It is obtained by using a calibrated light source, such as a
tributing to the non-uniform intensity of the single-beam blackbody at a given temperature, the spectrum of which
spectrum, as discussed in the previous paragraph. Chan- is recorded under exactly the same experimental conditions
nel spectra may also be due to parasitic signals added to as that of the sample. Relative intensities from one pro-
the interferogram. A pulse in the interferogram will profile to the other are then measurable. To obtain absolute
duce a sinusoidal variation within the spectrum. In the intensities, knowledge of the absolute number of photons
reverse situation, a periodic variation in the interferogram reaching the detector is needed and the efficiency of the
due, for instance, to mechanical vibrations will produce transfer optics between the source and the detector must
localized perturbations in the spectrum. In this case, the therefore be determined exactly.
additional fringes will be easily identified because they will Figure 2 shows a portion of a plasma emission spectrum
also be present in the zero-energy regions of the spectrum. with transitions mainly due to the HCN molecule.9 Note
Another source of channel spectra is interference effects in that the weak efficiency of the formation of the HCN
the beam-splitter of the interferometer. This effect is far molecule in the plasma leads to the fact that the S/N
more troublesome because, although the fringes appear in is poor.
High-resolution Fourier Transform Spectrometry of Gases 3
In practice, all the instrumental difficulties related to

high-resolution FT steady-state spectroscopy have been
well overcome. Home-made instruments have been con-
structed for many years.11 22 Spectrometer manufactur-
ers are also producing interferometers that are able to
deliver high-quality, high-resolution spectra.23 26 These
instruments are used under various conditions, in the labora-
tory, under telescopes,27 32 on aircraft,33 on balloon-borne
platforms,34 36 or on satellites.37,38
Recording high-resolution spectra is done using two
traditional approaches, generally known as rapid-
3200 3240 3280 3320 3360 3400 scanning (continuous-scanning) and stepping-mode
(often called step-scan or step-and-integrate). In
Wavelength /cm1
both methods, which have also been used for low-
Figure 2. HCN plasma emission spectrum. Unapodized resolu- resolution spectroscopy, the direct current component of
tion is equal to 4.5 103 cm1 .
the interference record is eliminated in order to improve
the measurement of the alternating current, or modulated,
3 RECORDING HIGH-RESOLUTION component, i.e. the interferogram. In the rapid-scanning
mode, path difference is varied linearly to produce an
MOLECULAR FT SPECTRA audio-frequency signal, which is electronically filtered and
recorded. The signal is also modulated in the stepping-mode
Ideally, one should not find much instrumental difference
approach, by a periodic variation of the path difference
between the respective recording demands of low- and high-
around the position defining each step. This is generally
resolution interferograms. The need for quality is normally
called internal or phase modulation. Comparison of the two
not resolution-dependent. It is more related to the nature of
the problem to be solved from the recorded spectra. The recording modes may be summarized briefly as follows.
quality of a FT spectrum is determined by the accuracy of The rapid-scanning mode has the following relative
the measurement of the interferograms, with respect to both advantages over the stepping-mode. Commercial Fourier
the sample and the path difference.10 Once high accuracy spectrometers incorporating rapid-scanning interferometers
is obtained for a mirror travel of 1 mm or so, it should not, have been available for a long time. Conversely, high-
in principle, be technically difficult to extend it to higher resolution stepping-mode interferometers are most often
distances. home-built. With the rapid-scanning approach the path dif-
However, low-resolution spectra are intended for record- ference ideally varies linearly with time. (This is not impera-
ing band contours. As a consequence, the recorded shapes tive, see for instance Brault.39 ) The engineering requirement
of the lines in vibrationrotation spectra, which are a con- is then to control the velocity of a mirror instead of its posi-
volution of the theoretical and the instrument line shape tion, which is somewhat easier. Since increasing the scan
functions, are usually governed by the theoretical line shape speed reduces the S/N, the signal can also be digitized with
rather than the instrument resolution. With high-resolution an available standard 16-bit analog-to-digital converter.
spectra, the requirements are much more exacting. All the Improvement of the dynamic range is further obtained by
disturbing instrumental effects, such as imperfect optical co-additions. It is also easier to obtain a higher modulation
misalignment, become much more significant and deleteri- frequency of the interferogram resulting in improved sig-
ous. Spectral distortion can affect both the position and nal detection relative to the stepping-mode. Spectra using
the shape of the lines to the point that these parameters the rapid-scanning mode may be measured in a few minutes.
may not be measurable with adequate accuracy and preci- This answer, even with a poor S/N, may be very useful when
sion. Also, to maintain a high modulation efficiency with testing source conditions. Since the S/N may be improved
an optical path difference of several meters, the available by co-adding successive full resolution interferograms, it is
etendue must be kept very small, so the signal, and hence also possible to retain or not, according to specific require-
the S/N, will be much lower than for low-resolution mea- ments, each individual interferogram before the co-addition.
surements. In addition, at high resolution, the number of The rapid-scanning approach always utilises the phase cor-
recorded data points usually exceeds 106 . The recording rection of the interferogram to remove from the spectra
system must be able to deal with large data files measured the systematic errors due to the optical imperfections of
over long recording times, and consequently it must be very the interferometers and to the experimental sampling pro-
reliable. cedures. Phase correction softwares are efficient.
On the other hand, with the stepping-mode, the interfer- spectra are generally concerned with measurements of
ogram is recorded in only one run during a time that is the molecules with a small number of atoms (not more than 10).
product of the total number of samples (data points) and These molecules may be stable or not, radical or ion. They
the integration time (which may be adjustable) spent dur- are most often encountered and observed in the gas phase.
ing each stop. Consequently, high-resolution information Observations of astronomical objects37,42 51 as well as
about the spectrum is obtained only when the experiment the observation of planetary29,44,52 62 and earth31,37,63 105
is completed. A malfunction occurring during the data atmospheres have been made for over 30 years. Laboratory
recording results in a fatal error. The resulting truncated spectra were recorded and analyzed for the interpretation of
interferogram provides the full spectrum but with a limited the astrophysical results,49,106 123 as well as for the under-
(lower) resolution. It is impossible to retrieve the high path standing of the molecular spectra of the planetary and earth
difference missing part of the interferogram, which must atmospheres.109,114,124 158 Intensive molecular studies are
be entirely recorded again. Stepping-mode is consequently presently being made. A quick survey of the papers pub-
very demanding as far as the exact and permanent con- lished on high-resolution Fourier spectra between 1997 and
trol of the phase is concerned. Also, as discussed above, 2000 readily confirms this. Studies are concerned with all
it may be difficult to achieve a high frequency for the types of molecules, many of which are listed below:
internal modulation, because of the mass of the moving
parts. It is interesting to note that, in spite of all these 1. Diatomic molecules: ArHC ,159 AsI,160 BaI,161 BeD,162
inconveniences, the first high-resolution FT spectra ever BeH,163 BiO,164 Br2 ,165 CH,121 CO,104,166 HCl,167
obtained11 were made with a stepping-mode interferometer, He2 ,168 HfCl,169 I2 ,170 IBr,171 IrN,172 KRb,173 LiD,174
which does have some specific advantages. For instance, LiH,174 N2 ,175 NbN,123 NH,122 NiCl,176 NO,177
unlike in the fast-scanning mode, careful electronic filtering O2 ,178 180 OsN,181 PbF,182 Rb2 ,183 RbCs,184 RuN,185
of the optical frequencies is not necessary, since all spec- SbBr,186 SbCl,186 SbF,186 SbI,186 ScI,187 SiD,188
tral elements are modulated with the same frequency. The SiH,188 SO,189 SrI,190 TaO,191 TiCl,192 TiO,49 ZrCl.193
number of recorded samples may be more easily restricted 2. Triatomic molecules: ClBS (seven isotopomers),194
to its optimum value. The stepping mode also makes use of CO2 ,195 197 OC18 O,198 D2 Se,199,200 FCN,201 H2 17 O,202
phase correction, although this may not be needed for opti- H2 O,202,203 HCN,9 D13 C15 N,204 HCSi,205 HDS,206
cally symmetric interferometers, since it is possible to start HDSe,199 HNC,9,207 HOCl,208 HSi79 Br,209 NH2 ,210
the interferogram at the exact path difference position.12 NO2 ,211,212 O3 ,213 217 16 O16 O17 O,216 16 O17 O16 O,216
16 17 17 218 17 16 17 219 18
Being able to remain at a given position may also be O O O, O O O, O3 ,217 OCS,220,221
18 222 223 224,225
useful to momentarily stop the interferogram as long as OCS, OCSe, SO2 .
needed, in order to wait for the availability of the source 3. Polyatomic molecules: 79 BrNO2 ,226 81 BrNO2 ,226
11
(for instance of a star occulted by a cloud). These charac- B2 H6 ,227 13 CHD3 ,228 CDF3 ,229 13 CH3 OH,230
teristics opened the way to some interesting spectroscopic CD3 OD,120 C2 D2 ,231 C2 F6 (perfluoroethane),232 C3 H4
applications, such as the investigation of the seismology of (propyne),233 CC4 F6 (cis-perfluorocyclobutane),232
planets and stars,40 or the accurate control of the tuning C6 H6 ,234 CF3 Cl (Freon 13),235 CF3 OCF2 H,236
of diode laser emission.41 The most important advantage CH35 Cl3 ,237 CH2 CO (ketene),238 CH2 F2 ,239
of the stepping-mode is the orthogonality of the time and CH3 CAs,240 13
CH3 D,241 CH3 OCF2 CF2 H,236
242 243 244
path difference dimensions. In other words, in contrast with CH3 OD, CHD3 , ClNO2 , ClONO2 ,167 CO-
245
rapid-scanning, while stepping, it is possible to investigate ortho-D2 (weakly bound complex), Cyanamide,246
247 35 248
time-dependent phenomena without any variation of the D2 CO, D3 Si Cl, HFC-134 (1,1,2,2-tetrafluoro-
path difference. ethane),249 HFC-143a (1,1,1-trifluoroethane),249
15 250 251
H NO3 , H2 O2 , H3 CF,252 H3 SiI,253 HNO3 ,254
13
248 255
HSiCl2 , HSiD3 , HSnD3 ,255 NH3 ,256 ND3 ,257
4 HIGH-RESOLUTION VIBRATIONAL NO3 , N2 O4 , (NO)2 ,260 262 (O2 )2 ,263 P4 ,264 SF6 ,265
258 259
SO2 F2 ,266 SbH3 ,267 SeCF2 ,268 SiH2 D2 ,269 SiH3 CN,270
FTS APPLICATIONS SiH3 CN,270 SiH4 ,271 120 SnD4 ,272 trans-DCOOD,273
trans-HCOOD,273 water cluster.274
For about three decades, high-resolution FTS has pro-
duced a large number of results. Industrial applications,
metrology, studies of earth and planetary atmospheres, 5 TRENDS
stars and interstellar medium, and obtaining an improved
understanding of molecular theory are among the many rea- Several instrumental developments that make use of
sons for undertaking these measurements. High-resolution new recording procedures, new light sources, and new
mixture of Ar and H2 and He. Figure 3(a) shows a non-

selective spectrum, whereas only those lines emitted by
ionized species are recorded in Figure 3(b), where ArHC
ArH + Ar is present and the atomic line of Ar has been eliminated.
2-1 R(7) Both spectra were recorded simultaneously. Other types
of discrimination that take advantage of, for example, the
(a) lifetime or the paramagnetism of the species of interest are
also possible.279,280
(b) Time-resolution is becoming increasingly important in
high-resolution experiments.281 284 Figure 4 shows an
example of time-resolved spectra, essentially centered
3 3 3 C
2601.0 2601.5 2604.0 2604.5 2605.0 on some B03 u B g and B g A u rovibronic
Wavelength / cm1 transitions of N2 . Other trends for high-resolution
vibrational spectroscopy may be also foreseen through
Figure 3. Due to velocity modulation, lines from only ionic
species are retained in the selective spectrum (see text for details).
work coupling FTS with other techniques, such as
(a) Non-selective spectrum; (b) selective spectrum. Unapodized synchrotron radiation,171,177,285,286 molecular beams,4 intra
resolution is equal to 4 cm1 . cavity laser absorption spectroscopy287 289 and laser-
induced fluorescence.3,290
ways of coupling high-resolution spectrometers with
complementary experimental approaches are on the ACKNOWLEDGMENTS
horizon. Specific modulations of the signal are now
practised for only recording selected classes of spectral Careful reading of the manuscript and helpful remarks by
transitions. For instance, molecular or atomic velocity John Chalmers and Peter Griffiths are much appreciated.
discrimination through Doppler shift has been demonstrated
with both a rapid-scanning instrument275,276 and with
a stepping-mode interferometer.277,278 Figure 3 shows a ABBREVIATIONS AND ACRONYMS
portion of two emission spectra illustrating this type
of selectivity. The source is a discharge tube with a S/N Signal-to-noise Ratio
5600 10600
6506 6508 6510 6512
(a) (c)
B 3u - B 3g v = 0 transitions
e
m
Ti
5800 6750 6509 6510

(b) (d) cm1
Figure 4. N2 time-resolved spectra from a pulsed microwave excited plasma. 128 spectra were recorded, each of them covering
the near-infrared range 550011 000 cm1 , with a Doppler-limited resolution equal to 14 103 cm1 . The time resolution is 1.6 s.
Spectrum number 27 is shown in (a). The underlined part is scale-expanded in (b), where the vertical thick arrow indicates a small
portion (about 1/1000) centered around 6509 cm1 . A small portion is expanded further in (c). (d) is a time-resolved representation
giving all 128 spectra every 1.6 s. The spectral range is restricted to the portion underlined in (c). More details may be found in Picque
and Guelachvili.284
REFERENCES 19. J. Verges, Studies of Collision Processes by Fourier Trans-

form Spectroscopy, Ann. Phys., 8, 187 (1983).
1. J.M. Chalmers and P.R. Griffiths (eds), Handbook of 20. J.J. Plateaux, D. Courtois, A. Delahaigue, A. Barbe
Vibrational Spectroscopy, John Wiley & Sons, Chichester, and P. Jouve, Microcomputer Controlled High Resolution
Vols 14 (2001). Fourier Transform Spectrometer, Rev. Phys. Appliquee, 21,
239 (1986).
2. G. Herzberg, Infrared and Raman Spectra, Van Nostrand
21. B. Carli, M. Carlotti, F. Mencaraglia and E. Rossi, Far-
Reinhold, New York (1945).
infrared High-resolution Fourier Transform Spectrometer,
3. J. Verges, C. Amiot, R. Bacis and A.J. Ross, Laser Induced Appl. Opt., 26, 3818 (1987).
Fluorescence and High Resolution Fourier Transform Spec- 22. V. Dana, L. Henry and A. Valentin, POF3 n4 Band Analy-
trometry, Spectrochim. Acta, Part A, 51, 1191 (1995). sis, J. Phys., 48, 1135 (1987).
4. M. Herman, R. Georges, M. Hepp and D. Hurtmans, High 23. D.J.W. Kendall, H.J. Buijs and J.W.C. Johns, A Very High
Resolution Fourier Transform Spectroscopy of Jet-cooled Resolution Fourier Transform Spectrophotometer System for
Molecules, Int. Rev. Phys. Chem., 19, 277 (2000). Doppler-limited Studies, J. Mol. Struct., 79, 39 (1982).
5. P.F. Bernath, Infrared Emission Spectroscopy, Annu. Rep. 24. M. Birk and M. Winnewisser, The RotationVibration
Progr. Chem., Sect C., 96, 177 (2000). Spectrum of Gaseous Cyanamide H2 NCN, Chem. Phys.
6. D.C. Benner, V.M. Devi, M.A.H. Smith, N. Picque and Lett., 123, 382 (1986).
G. Guelachvili, J. Quantitat. Spectrosc. Radiat. Transfer, to 25. E.H. Fink, H. Kruse, D.A. Ramsay and M. Vervloet, An
be published (2001). Electric Quadrupole Transition: the b1 C 1
g a g Emission
7. G. Guelachvili, Distortions in Fourier Spectra and Diag- System of Oxygen, Can. J. Phys., 64, 242 (1986).
nosis, in Spectrometric techniques, ed. G. Vanasse, Aca- 26. T. Ahonen, S. Alanko, V.M. Horneman, E. Koivusaari,
demic Press, New York, 162, Vol. 2 (1981). R. Paso, A.M. Tolonen and R. Anttila, A Long Path Cell
8. G. Guelachvili, Distortion Free Interferograms in Fourier for the Fourier Spectrometer Bruker IFS 120 HR: Applica-
Transform Spectroscopy with Nonlinear Detectors, Appl. tion to the Weak n1 C n2 and 3n2 Bands of Carbon Disulfide,
Opt., 25, 4644 (1986). J. Mol. Spectrosc., 181, 279 (1997).
27. P. Connes, Infrared Stellar Spectroscopy, in Theory and
9. N. Picque and G. Guelachvili, Emission Spectra of
Observation of Normal Stellar Atmospheres, Proceedings
HCN/HNC in the 25 m Range of Astrophysical Interest,
of 3rd HarvardSmithsonian Conference on Stellar Atmo-
Spectrochim. Acta, Part A, 56, 681 (2000).
spheres, ed. O. Gingerich, MIT Press, Cambridge, MA, 323
10. G. Guelachvili, High-accuracy Doppler-limited 106 Sam- (1969).
ples Fourier Transform Spectroscopy, Appl. Opt., 17, 1322 28. P. Connes, Astronomical Fourier Spectroscopy, Quart. J.
(1978). Royal Astronom. Soc., 14, 288 (1973).
11. J. Pinard, Development of a Very High Resolution Fourier 29. P. Connes and G. Michel, High-resolution Fourier Spectra
Transform Spectrometer, Ann. Phys., 4, 147 (1969). of Stars and Planets, Astrophys. J., 190, L29 (1974).
12. J. Connes, H. Delouis, P. Connes, G. Guelachvili, J.P. 30. J.P. Maillard and G. Michel, A High Resolution Fourier
Maillard and G. Michel, Spectroscopie de Fourier avec Transform Spectrometer for the Cassegrain Focus at the
transformation de 106 points, Nouv. Rev. Opt. Appliquee, CFH Telescope, in Instrumentation for Astronomy with
1, 3 (1970). Large Optical Telescopes. Proceedings of IAU Colloquium,
13. P. Connes, High Resolution Fourier Spectroscopy, in ed. C.M. Humphries, Reidel, Dordrecht, 213222, Vol. 67
Optical Instruments and Techniques, Oriel Press, (1982).
Newcastle upon Tyne, 17 (1970). 31. J.M. Flaud, C. Camy-Peyret, J.W. Brault, C.P. Rinsland and
14. G. Guelachvili and J.P. Maillard, Fourier Spectroscopy D. Cariolle, Nighttime and Daytime Variation of Atmo-
from 106 Samples, in Aspen International Conference on spheric NO2 from Ground-based Infrared Measurements,
Fourier Spectroscopy, eds G.A. Vanasse, A.T. Stair, Jr and Geophys. Res. Lett., 15, 261 (1988).
D.J. Baker, Air Force Cambridge Res. Labs, Bedford, MA, 32. E. Serabyn, E.W. Weisstein and D.C. Lis, FTS Atmo-
151161 (1971). spheric Transmission Measurements and Observations of
15. G. Guelachvili, Near Infrared Wide-band Spectroscopy with Planetary Atmospheres, in Physics and Chemistry of Inter-
27-MHz Resolution, Appl. Opt., 16, 2097 (1977). stellar Molecular Clouds, eds G. Winnewisser and G.C.
Pelz, Springer-Verlag, Berlin, 377379, Vol. 393 (1995).
16. P. Luc and S. Gerstenkorn, Fourier Transform Spectroscopy
33. B. Carli, P. Ade, M. Carlotti, U. Cortesi, A. Gignoli,
in the Visible and Ultraviolet Range, Appl. Opt., 17, 1327
P. Hamilton, M. Lanfranchi, C. Lee, A.R. MacKenzie and
(1978).
A. Phillips, Minor Constituent Concentrations Measured
17. J. Kauppinen, Working Resolution of 0.010 cm1 Between from a High Altitude Aircraft Using High Resolution Far-
20 cm1 and 1200 cm1 by a Fourier Spectrometer, Appl. infrared Fourier Transform Spectroscopy, J. Atmos. Chem.,
Opt., 18, 1788 (1979). 35, 273 (2000).
18. J.W. Brault, Fourier Transform Spectrometry in Relation to 34. B. Carli, F. Mencaraglia and A. Bonetti, Submillimeter
Other Passive Spectrometers, Philos. Trans. R. Soc. London, High-resolution FT Spectrometer for Atmospheric Studies,
Ser. A, 307, 503 (1982). Appl. Opt., 23, 2594 (1984).
35. J.C. Brasunas, V.G. Kunde and L.W. Herath, Cryogenic 51. B. Mosser, J.P. Maillard and D. Mekarnia, New Attempt at
Fourier Spectrometer for Measuring Trace Species in the Detecting the Jovian Oscillations, Icarus, 144, 104 (2000).
Lower Stratosphere, Appl. Opt., 23, 4964 (1988). 52. L.D. Kaplan, J. Connes and P. Connes, Carbon Monoxide
36. C. Camy-Peyret, Balloon-borne Infrared Fourier Trans- in the Martian Atmosphere, Astrophys. J., 157, l187 (1969).
form Spectroscopy for Measurements of Atmospheric Trace 53. C. De Bergh, M. Vion, M. Combes, J. Lecacheux and J.P.
Species, Spectrochim. Acta, Part A, 51, 1143 (1995). Maillard, New Infrared Spectra of the Jovian Planets from
37. C.B. Farmer, High Resolution Infrared Spectroscopy of the 12000 to 4000 cm1 by Fourier Transform Spectroscopy. II.
Sun and the Earths Atmosphere from Space, Mikrochim. Study of Saturn in the 3n3 CH4 Band, Astron. Astrophys.,
Acta, 3, 189 (1987). 28, 457 (1973).
38. S. Calcutt, F. Taylor, P. Ade, V. Kunde and D. Jennings, 54. J.P. Maillard, J. Crovisier, T. Encrenaz and M. Combes,
The Composite Infrared Spectrometer, J. British Inter- The Spectrum of P/Halley between 0.9 and 2.5 Microns,
planet. Soc., 9, 381 (1992). in 20th ESLAB Symposium on the Exploration of Halleys
Comet. Proceedings of the International Symposium (ESA-
39. J.W. Brault, New Approach to High-precision Fourier
SP-250), eds B. Battrick, E.J. Rolfe and R. Reinhard, ESA,
Transform Spectrometer Design, Appl. Opt., 35, 2891
Paris, 359364 (1986).
(1996).
55. F. Billebaud, J.P. Maillard, E. Lellouch and T. Encrenaz,
40. J.P. Maillard, Seismology with a Fourier-transform Spec-
The Spectrum of Mars in the (1-0) Vibrational Band of
trometer: Applications to Giant Planets and Stars, Appl.
CO, Astron. Astrophys., 261, 647 (1992).
Opt., 35, 2734 (1996).
56. G.L. Bjoraker, H.P. Larson and V.G. Kunde, The Abun-
41. A. Valentin, C. Nicolas, L. Henry and A.W. Mantz, Tun-
dance and Distribution of Water Vapor in Jupiters Atmo-
able Diode Laser Control by a Stepping Michelson Interfer-
sphere, Astrophys. J., 2, 1058 (1987).
ometer, Appl. Opt., 26, 41 (1987).
57. B. Mosser, D. Mekarnia, J.P. Maillard, J. Gay, D. Gautier
42. P. Connes, J. Connes, R. Bouigue, M. Querci, J. Chauville
and P. Delache, Seismological Observations with a Fourier
and F. Querci, Spectra of Red M and S Stars between 4000
Transform Spectrometer: Detection of Jovian Oscillations,
and 9000 cm1 (2.5 to 1.1 ). I. General Description, Ann.
Astron. Astrophys., 267, 604 (1993).
Astrophys., 31, 485 (1968).
58. E.W. Weisstein and E. Serabyn, Detection of the 267 GHz
43. E.F. Montgomery, P. Connes, J. Connes and F.N. Ed- J D 1-0 Rotational Transition of PH3 in Saturn with a New
monds, Jr, The Infrared Spectrum of Arcturus, Astrophys. Fourier Transform Spectrometer, Icarus, 109, 367 (1994).
J. Suppl., 19, 1 (1969).
59. D.A. Naylor, G.R. Davis, M.J. Griffin, T.A. Clark,
44. W.A. Traub, N.P. Carleton, P. Connes and J.F. Noxon, The D. Gautier and A. Marten, Broad-band Spectroscopic
Latitude Variation of O2 Dayglow and O3 Abundance on Detection of the CO J D 3-2 Tropospheric Absorption in the
Mars, Astrophys. J., 229, 846 (1979). Atmosphere of Neptune, Astron. Astrophys., 3, L51 (1994).
45. L. Dame, L. Acton, M. Bruner, P. Connes, T. Cornwell, 60. A. Coustenis, E. Lellouch, J.P. Maillard and C.P. McKay,
B.H. Foing, J. Heyvaerts, P. Lemaire, M. Martic, R. Muller, Titans Surface: Composition and Variability from the Near-
J. Porteneuve, C.T. Roca, J. Riehl, R. Rutten, M. Sechaud, infrared Albedo, Icarus, 118, 87 (1995).
P. Smith, A.P. Thorne, A.M. Title, J.C. Vial, H. Visser and
G. Weigelt, A Solar Interferometric Mission for Ultra- 61. S.M. Lederer, M.S. Marley, B. Mosser, J.P. Maillard, N.J.
high Resolution Imaging and Spectroscopy: SIMURIS, Adv. Chanover and R.F. Beebe, Albedo Features and Jovian
Space Res., 11, 383 (1991). Seismology, Icarus, 114, 269 (1995).
46. B.H. Foing, L. Dame, A.P. Thorne and P. Lemaire, The 62. E.W. Weisstein and E. Serabyn, Submillimeter Line Search
Imaging Fourier Transform Spectrometer for the SIMURIS in Jupiter and Saturn, Icarus, 123, 23 (1996).
Mission, Adv. Space Res., 11, 387 (1991). 63. J.E. Harries, B. Carli and M.J. Bangham, Very High Res-
47. R. Farrenq, G. Guelachvili, A.J. Sauval, N. Grevesse and olution Fourier Spectroscopy of the Atmosphere in the
C.B. Farmer, Improved Dunham Coefficients for CO from Submillimetre Region, in European Conference on Precise
Infrared Solar Lines of High Rotational Excitation, J. Mol. Electrical Measurement, IEEE, London, 7172 (1977).
Spectrosc., 149, 375 (1991). 64. B. Carli, F. Mencaraglia and A. Bonetti, Fourier Spec-
48. D.A. Simons and J.P. Maillard, Fourier Imaging Spec- troscopy of the Stratospheric Emission, Int. J. Infrared
troscopy of the Galactic Center, Astronom. Soc. Pacific Millimeter Waves, 1, 263 (1980).
Conf. Ser., 102, 232 (1996). 65. B. Carli, Fourier Transform Spectroscopy in the Submil-
49. R.S. Ram, P.F. Bernath, M. Dulick and L. Wallace, The limeter Region, Proc. SPIE, 289, 394 (1981).
A3 X3 System (Gamma Bands) of TiO: Laboratory and 66. C.B. Farmer, High Resolution Spectroscopy of the Earths
Sunspot Measurements, Astrophys. J. Suppl. Ser., 122, 331 Upper Atmosphere from Space, Proc. SPIE, 289, 299
(1999). (1981).
50. J.L. Lemaire, D. Field, J.P. Maillard, G. Pineau des Forets, 67. C. Camy-Peyret, J.M. Flaud, L. Delbouille, G. Roland,
E. Falgarone, F.P. Pijpers, M. Gerin and F. Rostas, High J.W. Brault and L. Testerman, Quadrupole Transitions of
Resolution Fourier Transform Spectroscopy of H2 IR Emis- the 1 from 0 Band of N2 Observed in a High Resolution
sion in NGC 7023, Astron. Astrophys., 349, 253 (1999). Atmospheric Spectrum, J. Phys. Lett., 42, L279 (1981).
68. R. Zander, G.M. Stokes and J.W. Brault, Simultaneous Statistical Analysis of the Errors, J. Geophys. Res., 94,
Detection of FC-11, FC-12 and FC-22, through 8 to 13 16 365 (1989).
Micrometers IR Solar Observations from the Ground, Geo- 82. M.T. Coffey, W.G. Mankin and A. Goldman, Airborne
phys. Res. Lett., 10, 521 (1983). Measurements of Stratospheric Constituents over Antarctica
69. C.P. Rinsland, R.E. Boughner, J.C. Larsen, G.M. Stokes in the Austral Spring, 1987. 2. Halogen and Nitrogen Trace
and J.W. Brault, Diurnal Variations of Atmospheric Nitric Gases, J. Geophys. Res., 94, 16 597 (1989).
Oxide: Ground-based Infrared Spectroscopic Measurements 83. W.C. Maguire, V.G. Kunde and L.W. Herath, Infrared
and their Interpretation with Time-dependent Photochemical Emission High Spectral Resolution Atlas of the Stratospheric
Model Calculations, J. Geophys. Res., 89, 9613 (1984). Limb, Appl. Opt., 6, 1048 (1989).
70. L. Wallace, J.W. Brault, M. Brown and W. Livingston, 84. C.P. Rinsland, R. Zander, J.S. Namkung, C.B. Farmer and
Characteristics of Water Vapor over Kitt Peak as R.H. Norton, Stratospheric Infrared Continuum Absorp-
Determined from FTS data, Publ. Astron. Soc. Pacific, 96, tions Observed by the ATMOS Instrument, J. Geophys.
836 (1984). Res., 94, 16 303 (1989).
71. C.B. Farmer, High Resolution Infrared Spectroscopy from 85. G.C. Toon, C.B. Farmer, P.W. Schaper, J.F. Blavier and
Space: Preliminary Results of the ATMOS Experiment on L.L. Lowes, Ground-based Infrared Measurements of Tro-
Spacelab-3, Proc. SPIE, 553, 87 (1985). pospheric Source Gases over Antarctica during the 1986
72. M.G. Baldecchi, A. Bonetti, B. Carli, M. Carlotti, F. Men- Austral Spring, J. Geophys. Res., 94, 11 613 (1989).
caraglia and E. Rossi, High Resolution Measurement of the 86. C.P. Rinsland, L.R. Brown and C.B. Farmer, Infrared Spec-
Far-infrared Stratospheric Spectrum, Proc. SPIE, 553, 218 troscopic Detection of Sulfur Hexafluoride SF6 in the Lower
(1985). Stratosphere and Upper Troposphere, J. Geophys. Res., 95,
73. J.H. Park, R. Zander, C.B. Farmer, C.P. Rinsland, J.M. Rus- 5577 (1990).
sell, III, R.H. Norton and O.F. Raper, Spectroscopic Detec- 87. P. Demoulin, C.B. Farmer, C.P. Rinsland and R. Zander,
tion of CH3 Cl in the Upper Troposphere and Lower Strato- Determination of Absolute Strengths of N2 Quadrupole
sphere, Geophys. Res. Lett., 13, 765 (1986). Lines from High-resolution Ground-based IR Solar Obser-
74. C.P. Rinsland, R. Zander, C.B. Farmer, R.H. Norton, L.R. vations, J. Geophys. Res., 96, 13 003 (1991).
Brown, J.M. Russell, III and J.H. Park, Evidence for the 88. J.H. Park and B. Carli, Spectroscopic Measurement of
Presence of the 802.7-cm1 Band Q branch of HO2 NO2 HO2 , H2 O2 , and OH in the Stratosphere, J. Geophys. Res.,
in High Resolution Solar Absorption Spectra of the Strato- 96, 22 535 (1991).
sphere, Geophys. Res. Lett., 13, 761 (1986).
89. C.P. Rinsland, A. Goldman, M.A.H. Smith and V.M. Devi,
75. S.T. Massie, J.A. Davidson, C.A. Cantrell, A.H. McDaniel, Measurements of Lorentz Air-broadening Coefficients and
J.C. Gille, V.G. Kunde, J.C. Brasunas, B.J. Conrath, W.C. Relative Intensities in the H2 16 O Pure Rotational and n2
Maguire, A. Goldman and M.M. Abbas, Atmospheric Bands from Long Horizontal Path Atmospheric Spectra,
Infrared Emission of ClONO2 Observed by a Balloon-borne Appl. Opt., 30, 1427 (1991).
Fourier Spectrometer, J. Geophys. Res., D12, 14 806 (1987).
90. C.P. Rinsland, R. Zander, A. Goldman and F.J. Murcray,
76. B. Carli and J.H. Park, Simultaneous Measurement of The Fundamental Quadrupole Band of 14 N2 Line Positions
Minor Stratospheric Constituents with Emission Far-infrared From High-resolution Stratospheric Solar Absorption Spec-
Spectroscopy, J. Geophys. Res., 93, 3851 (1988). tra, J. Mol. Spectrosc., 148, 274 (1991).
77. B. Carli, F. Mencaraglia, M. Carlotti, B.M. Dinelli and 91. C.P. Rinsland, M.R. Gunson, J.C. Foster, R.A. Toth, C.B.
I. Nolt, Submillimeter Measurement of Stratospheric Farmer and R. Zander, Stratospheric Profiles of Heavy
Chlorine Monoxide, J. Geophys. Res., 93, 7063 (1988). Water Vapor Isotopes and CH3 D from Analysis of the
78. A. Goldman, F.J. Murcray, F.H. Murcray, D.G. Murcray ATMOS Spacelab 3 Infrared Solar Spectra, J. Geophys.
and C.P. Rinsland, Quantification of Several Atmospheric Res., 96, 1057 (1991).
Gases from High Resolution Infrared Solar Spectra Obtained 92. C.P. Rinsland, J.S. Levine, A. Goldman, N.D. Sze, M.K.W.
at the South Pole in 1980 and 1986, Mikrochim. Acta, 2, Ko and D.W. Johnson, Infrared Measurements of HF and
409 (1988). HCl Total Column Abundances above Kitt Peak, 197790:
79. V.G. Kunde, J.C. Brasunas, W.C. Maguire, J.R. Herman, Seasonal Cycles, Long-Term Increases, and Comparisons
S.T. Massie, M.M. Abbas, L.W. Herath and W.A. Shaffer, with Model Calculations, J. Geophys. Res., 96, 15 523
Measurement of Nighttime Stratospheric N2 O5 from Infra- (1991).
red Emission Spectra, Geophys. Res. Lett., 11, 1177 (1988). 93. C. Camy-Peyret, J.M. Flaud, F. Karcher, J.P. Meyer,
80. R. Zander, C.P. Rinsland, C.B. Farmer, J. Namkung, R.H. D. Murcray, F. Murcray and J. Van Allen, Stratospheric
Norton and J.M. Russell, III, Concentrations of Carbonyl H2 18 O and H2 17 O Measurements from Balloon-borne Infra-
Sulfide and Hydrogen Cyanide in the Free Upper red Spectra, Ann. Geophys., 10, 267 (1992).
Troposphere and Lower Stratosphere Deduced from 94. C.P. Rinsland, M.R. Gunson, M.C. Abrams, L.L. Lowes,
ATMOS/Spacelab 3 Infrared Solar Occultation Spectra, J. R. Zander, E. Mahieu, A. Goldman, M.K.W. Ko, J.M. Rod-
Geophys. Res., 93, 1669 (1988). riguez and N.D. Sze, Heterogeneous Conversion of N2 O5
81. M. Carlotti, A. Barbis and B. Carli, Stratospheric Ozone to HNO3 in the Post-Mount Pinatubo Eruption Stratosphere,
Vertical Distribution from Far-infrared Balloon Spectra and J. Geophys. Res., 99, 8213 (1994).
95. M. Carlotti, P.A.R. Ade, B. Carli, P. Ciarpallini, U. Cortesi, 105. W.G. Mankin, M.T. Coffey, A. Goldman, M.R. Schoeberl,
M.J. Griffin, G. Lepri, F. Mencaraglia, A.G. Murray, I.G. L.R. Lait and P.A. Newman, Airborne Measurements of
Nolt, J.H. Park and J.V. Radostitz, Measurement of Strato- Stratospheric Constituents over the Arctic in the Winter of
spheric HBr Using High Resolution Far Infrared Spec- 1989, Geophys. Res. Lett., 17, 473 (2000).
troscopy, Geophys. Res. Lett., 22, 3207 (1995). 106. A. Bernard, R. Bacis and P. Luc, Fourier Transform Spec-
96. C.P. Rinsland, M.R. Gunson, M.C. Abrams, L.L. Lowes, troscopy: Extensive Analysis of the A 2 to X2 C and B2 C
R. Zander, E. Mahieu, A. Goldman and F.W. Irion, April to X2 C Systems of Yttrium Oxide, Astrophys. J., 227, 338
1993 Arctic Profiles of Stratospheric HCl, ClONO2 , (1979).
and CCl2 F2 from Atmospheric Trace Molecule Spec- 107. C. Amiot and J. Verges, Fourier Spectroscopy of the 12 C13 C
troscopy/ATLAS 2 Infrared Solar Occultation Spectra, J. and 13 C2 BallikRamsay System, Astrophys. J., 263, 993
Geophys. Res., 100, 14 019 (1995). (1982).
97. A.Y. Chang, R.J. Salawitch, H.A. Michelsen, M.R. Gunson, 108. C. Amiot and J. Verges, Fourier Spectroscopy of the 12 C13 C
M.C. Abrams, R. Zander, C.P. Rinsland, J.W. Elkins, G.S. and 13 C2 Phillips System, Astron. Astrophys. Supp. Ser., 51,
Dutton, C.M. Volk, C.R. Webster, R.D. May, D.W. Fahey, 257 (1983).
R.S. Gao, M. Loewenstein, J.R. Podolske, R.M. Stimpfle,
D.W. Kohn, M.H. Proffitt, J.J. Margitan, K.R. Chan, M.M. 109. G. Guelachvili, D. de Villeneuve, R. Farrenq, W. Urban and
Abbas, A. Goldman, F.W. Irion, G.L. Manney, M.J. New- J. Verges, Dunham Coefficients for Seven Isotopic Species
church and G.P. Stiller, A Comparison of Measurements of CO, J. Mol. Spectrosc., 98, 64 (1983).
from ATMOS and Instruments Aboard the ER-2 Aircraft: 110. S.J. Daunt, A.K. Atakan, W.E. Blass, G.W. Halsey, D.E.
Halogenated Gases, Geophys. Res. Lett., 23, 2393 (1996). Jennings, D.C. Reuter, J. Susskind and J.W. Brault, The
98. M.M. Abbas, H.A. Michelsen, M.R. Gunson, M.C. Abrams, 12 Micron Band of Ethane: High-resolution Laboratory
M.J. Newchurch, R.J. Salawitch, A.Y. Chang, A. Goldman, Analysis with Candidate Lines for Infrared Heterodyne
F.W. Irion, G.L. Manney, E.J. Moyer, R. Nagaraju, C.P. Searches, Astrophys. J., 280, 921 (1984).
Rinsland, G.P. Stiller and R. Zander, Seasonal Variations 111. D. Boudjaadar, J. Brion, P. Chollet, G. Guelachvili and
of Water Vapor in the Lower Stratosphere Inferred from M. Vervloet, Infrared Emission Spectra of Five v D 1
ATMOS/ATLAS-3 Measurements of H2 O and CH4 , Geo- Sequence Bands of the Free Radical NH in its X 3 State,
phys. Res. Lett., 23, 2401 (1996). J. Mol. Spectrosc., 119, 352 (1986).
99. M.R. Gunson, M.M. Abbas, M.C. Abrams, M. Allen, L.R. 112. D.E. Jennings, A. Weber and J.W. Brault, FTS-Raman
Brown, T.L. Brown, A.Y. Chang, A. Goldman, F.W. Irion, Flame Spectroscopy of High-J lines in H2 and D2 , J. Mol.
L.L. Lowes, E. Mahieu, G.L. Manney, H.A. Michelsen, M.J. Spectrosc., 126, 19 (1987).
Newchurch, C.P. Rinsland, R.J. Salawitch, G.P. Stiller, G.C. 113. F. Holland, M. Winnewisser and J.W.C. Johns, High Res-
Toon, Y.L. Yung and R. Zander, The Atmospheric Trace olution FT-IR Spectrum of Tricarbon Disulfide, SCCCS,
Molecule Spectroscopy (ATMOS) experiment: Deployment Mikrochim. Acta, 2, 417 (1988).
on the ATLAS Space Shuttle Missions, Geophys. Res. Lett.,
23, 2333 (1996). 114. R.M. Sergent, V.T. Nguyen, I. Rossi, N. Lacome, G. Tarr-
ago and A. Levy, Pure Rotational Spectrum of PH3 in the
100. F.W. Irion, M.R. Gunson, C.P. Rinsland, Y.L. Yung, M.C. 10100 cm1 Range: Collisional Broadening Coefficients by
Abrams, A.Y. Chang and A. Goldman, Heavy Ozone H2 and He, Mikrochim. Acta, 1, 227 (1988).
Enrichments from ATMOS Infrared Solar Spectra,
Geophys. Res. Lett., 23, 2377 (1996). 115. M. Horani and M. Vervloet, Improved Determination of the
Ground State Molecular Constants of the CSC Cation to Aid
101. M.C. Abrams, A.Y. Chang, M.R. Gunson, M.M. Abbas, Possible Astrophysical Detection, Astron. Astrophys., 256,
A. Goldman, F.W. Irion, H.A. Michelsen, M.J. Newchurch, 683 (1992).
C.P. Rinsland, G.P. Stiller and R. Zander, On the Assess-
ment and Uncertainty of Atmospheric Trace Gas Burden 116. Y. Kasai, K. Obi, Y. Ohshima, Y. Hirahara, Y. Endo,
Measurements with High Resolution Infrared Solar Occul- K. Kawaguchi and A. Murakami, Laboratory Detection of
tation Spectra from Space by the ATMOS Experiment, C5 S by Pulse-discharge-nozzle Fourier Transform Micro-
Geophys. Res. Lett., 23, 2337 (1996). wave Spectroscopy, Astrophys. J. Lett., 410, L45 (1993).
102. B.M. Dinelli, L.G. Lepri, M. Carlotti, B. Carli, F. Men- 117. D. Cossart, M. Horani and M. Vervloet, The A 2 1
caraglia, M. Ridolfi, I.G. Nolt and P.A.R. Ade, Measure- X 2 C and B 2 C A 2 1 Electronic Transitions of CSC
ment of the Isotopic Ratio Distribution of HD16 O and H2 16 O by High Resolution Fourier Transform Spectroscopy, AIP
in the 2038 km Altitude Range from Far-infrared Spectra, Conf. Proc., 312, 367 (1994).
Geophys. Res. Lett., 24, 2003 (1997). 118. C.V.V. Prasad and P.F. Bernath, Fourier Transform Spec-
103. J.W. Harder and J.W. Brault, Atmospheric Measurements troscopy of the Swan d3 g a3 u ) Jet-cooled C2 Mole-
of Water Vapor in the 442-nm Region, J. Geophys. Res., cule, Astrophys. J., 426, 812 (1994).
102, 6245 (1997). 119. J.P. Maillard, Recent Results in Astronomical Fourier
104. C.P. Rinsland, R.J. Salawitch, M.R. Gunson, S. Solomon, Transform Spectroscopy, Spectrochim. Acta, Part A, 51,
R. Zander, E. Mahieu, A. Goldman, M.J. Newchurch, F.W. 1105 (1995).
Irion and A.Y. Chang, Polar Stratospheric Descent of 120. I. Mukhopadhyay, K.V.L.N. Sastry and M. Winnewisser,
NOy and CO and Arctic Denitrification During Winter Measurement of Microwave Transitions in Methanol-D4 ,
19921993, J. Geophys. Res. Atmos., 104, 1847 (1999). Spectrochim. Acta, Part A, 54, 1375 (1998).
121. P.N. Ghosh, M.N. Deo and K. Kawaguchi, Vibrational 137. H. Burger, P. Schulz, E. Jacob and M. Fahnle, The Fourier
Transition Moment of the CH Radical Determined from the Transform IR Spectra of ClF and BrF, Z. Naturforsch., A:
HermanWallis Effect, Astrophys. J., 525, 539 (1999). Phys. Sci., 41, 1015 (1986).
122. R.S. Ram, P.F. Bernath and K.H. Hinkle, Infrared Emis- 138. L.R. Brown, C.B. Farmer, C.P. Rinsland and R.A. Toth,
sion Spectroscopy of NH: Comparison of a Cryogenic Molecular Line Parameters for the Atmospheric Trace
Echelle Spectrograph with a Fourier Transform Spectrome- Molecule Spectroscopy Experiment, Appl. Opt., 26, 5154
ter, J. Chem. Phys., 110, 5557 (1999). (1987).
123. C. Cottaz, I. Kleiner, G. Tarrago, L.R. Brown, J.S. Mar- 139. A. Amrein, H. Hollenstein, P. Locher, M. Quack, U. Sch-
golis, R.L. Poynter, H.M. Pickett, T. Fouchet, P. Drossart mitt and H. Burger, Analysis of the n4 and n1 Bands
and E. Lellouch, Line positions and intensities in the 2n2 /n4 of CF3 Cl Measured by Supersonic Free-jet FTIR Spec-
vibrational system of 14 NH3 near 57 m, J. Mol. Spec- troscopy, Chem. Phys. Lett., 139, 82 (1987).
trosc., 203, 285 (2000). 140. J. Ballard, W.B. Johnston, M.R. Gunson and P.T. Wassell,
124. C. Camy-Peyret, J.M. Flaud, G. Guelachvili and C. Amiot, Absolute Absorption Coefficients of ClONO2 Infrared
High Resolution Fourier Transform Spectrum of Water Bands at Stratospheric Temperatures, J. Geophys. Res., 93,
between 2930 and 4255 cm1 , Mol. Phys., 26, 285 (1973). 1659 (1988).
125. J.M. Flaud and C. Camy-Peyret, The 2n2 , n1 and n3 Bands 141. J.B. Burkholder, P.D. Hammer, C.J. Howard and A. Gold-
of H2 16 O. Rotational Study of the (000) and (020) States, man, Infrared Line Intensity Measurements in the v D 0-1
Mol. Phys., 26, 811 (1973). Band of the ClO Radical, J. Geophys. Res., 94, 2225 (1989).
126. J.M. Flaud, C. Camy-Peyret, J.P. Maillard and G. Guelach- 142. B. Carli, M. Carlotti, B.M. Dinelli, F. Mencaraglia and
vili, The H2 O Spectrum between 4200 and 5000 cm1 , J. H. Park, The Mixing Ratio of the Stratospheric Hydroxyl
Mol. Spectrosc., 65, 219 (1977). Radical from Far Infrared Emission Measurements, J. Geo-
127. J.M. Flaud, C. Camy-Peyret, J.Y. Mandin and G. Guelach- phys. Res., 94, 11 049 (1989).
vili, H2 16 O Hot Bands in the 6.3 m Region, Mol. Phys., 143. J.H. Park, B. Carli and A. Barbis, Stratospheric HBr Mix-
34, 413 (1977). ing Ratio Obtained from Far Infrared Emission Spectra,
128. J.P. Maillard, M. Cuisenier, P. Arcas, E. Arie and C. Amiot, Geophys. Res. Lett., 16, 787 (1989).
Infrared Spectrum and Molecular Constants of CO2 in the 144. A. del Olmo, C. Domingo, J.M. Orza and D. Bermejo, FT
1.41.7 m Atmospheric Window by Very High Resolution Intracavity Laser Spectroscopy: the BX Transition of Cl2 ,
Fourier Transform Spectroscopy, Can. J. Phys., 58, 1560 J. Mol. Spectrosc., 145, 323 (1991).
(1980). 145. C. Chackerian, Jr, D. Goorvitch, A. Benidar, R. Farrenq,
129. M. Carlotti, G. Di Lonardo, L. Fusina, A. Trombetti, G. Guelachvili, P.M. Martin, M.C. Abrams and S.P. Davis,
A. Bonetti, B. Carli and F. Mencaraglia, The High- Rovibrational Intensities and Electric Dipole Moment Func-
resolution Spectrum of Ozone between 8 and 150 cm1 , tion of the X 2 Hydroxyl Radical, J. Quantitat. Spectrosc.
J. Mol. Spectrosc., 107, 84 (1984). Radiat. Transfer, 48, 667 (1992).
130. N. Lacome, A. Levy and G. Guelachvili, Fourier Trans- 146. A. Barbe, S. Bouazza, J.J. Plateaux and M. Jacon, The
form Measurement of self-, N2 -, and O2 -broadening of N2 O 3n3 C 2n2 Band of Ozone: Line Positions and Intensities,
Lines: Temperature Dependence of Linewidths, Appl. Opt., J. Mol. Spectrosc., 162, 335 (1993).
23, 425 (1984). 147. L. Rosenmann, J.M. Hartmann, J.J. Plateaux and A. Barbe,
131. C. Camy-Peyret, J.M. Flaud, J.Y. Mandin, J.P. Chevillard, Absorption in the Q-branch of the n1 C n2 C n3 Band of O3 ,
J. Brault, D.A. Ramsay, M. Vervloet and J. Chauville, The J. Quantitat. Spectrosc. Radiat. Transfer, 50, 233 (1993).
High-resolution Spectrum of Water Vapor Between 16 500 148. M.C. Abrams, S.P. Davis, M.L.P. Rao, R. Engleman, Jr and
and 25 250 cm1 , J. Mol. Spectrosc., 113, 208 (1985). J.W. Brault, High-resolution Fourier Transform Spectro-
132. B. Carli, M. Carlotti, B.M. Dinelli, F. Mencaraglia and scopy of the Meinel System of OH, Astrophys. J. Suppl.
S. Piccioli, Synthetic Spectra for the Analysis of Far IR Ser., 93, 351 (1994).
Stratospheric Emission Measurements, Proc. SPIE, 553, 149. J. Orphal, M. Morillon-Chapey and G. Guelachvili, The
3765 (1985). High-resolution Infrared Spectrum of Nitryl Chloride: Rota-
133. J.W.C. Johns, High-resolution Far-infrared (20350-cm1 ) tional Analysis of the 35 ClNO2 n4 Band around 6 m, J.
Spectra of Several Isotopic Species of H2 O, J. Opt. Soc. Mol. Spectrosc., 165, 315 (1994).
Am. B, 2, 1340 (1985). 150. A. Barbe, O. Sulakshina, J.J. Plateaux, A. Hamdouni and
134. J.W. Brault, Fourier Transform Spectrometry at the Natio- S. Bouazza, High-resolution Infrared Spectra of Ozone in
nal Solar Observatory: Recent Work on the OH Radical, the 23002600 cm1 Region, J. Mol. Spectrosc., 170, 244
Proc. SPIE, 553, 97 (1985). (1995).
135. J. Ballard and W.B. Johnston, Self-broadened Widths and 151. A. Barbe, S. Mikhailenko, V. Tyuterev, A. Hamdouni and
Absolute Strengths of 12 CH4 Lines in the 13101370 cm1 J.J. Plateaux, Analysis of the 2n1 C 2n2 C n3 Band of
Region, J. Quantitat. Spectrosc. Radiat. Transfer, 36, 365 Ozone, J. Mol. Spectrosc., 171, 583 (1995).
(1986). 152. M.T. Bourgeois, J.M. Flaud, A. Barbe, L. Regalia, C. Camy-
136. J.H. Park and B. Carli, Analysis of Far-infrared Emission Peyret and J.J. Plateaux, The n1 C n3 Bands of 16 O18 O18 O
Fourier Transform Spectra, Appl. Opt., 25, 3490 (1986). and 18 O16 O18 O, J. Mol. Spectrosc., 172, 552 (1995).
153. A. Saouli, H. Bredohl, I. Dubois and A. Fayt, FT Infrared 169. R.S. Ram, A.G. Adam, A. Tsouli, J. Lievin and P.F. Ber-
Spectra of ClCN between 1200 and 5000 cm1 and Global nath, High-resolution Spectroscopy and Ab Initio Calcula-
Rovibrational Analysis of the Main Isotopomers, J. Mol. tions on HfCl, J. Mol. Spectrosc., 202, 116 (2000).
Spectrosc., 174, 20 (1995). 170. D. Inard, D. Cerny, M. Nota, R. Bacis, S. Churassy and
154. K. Yoshino, J.R. Esmond, J.E. Murray, W.H. Parkinson, V. Skorokhodov, E0C C 00
g ! A 1u and E0g ! B 1u Laser-
A.P. Thorne, R.C.M. Learner and G. Cox, Band Oscillator induced Fluorescence in Molecular Iodine Recorded by
Strengths of the Herzberg I Bands of O2 , J. Chem. Phys., Fourier-transform Spectroscopy, Chem. Phys., 243, 305
103, 1243 (1995). (1999).
155. L.R. Brown, M.R. Gunson, R.A. Toth, F.W. Irion, C.P. Rins- 171. B. Nelander, V. Sablinskas, M. Dulick, V. Braun and P.F.
land and A. Goldman, 1995 Atmospheric Trace Molecule Bernath, High Resolution Far Infrared Spectroscopy of IBr
Spectroscopy (ATMOS) Linelist, Appl. Opt., 35, 2828 Using a Synchrotron Source, Mol. Phys., 93, 137 (1998).
(1996).
172. R.S. Ram, J. Lievin and P.F. Bernath, Emission Spec-
156. F. Frichot, N. Lacome and J.M. Hartmann, Pressure and troscopy and Ab Initio Calculations on IrN, J. Mol. Spec-
Temperature Dependences of Absorption in the n5 R Q0 trosc., 197, 133 (1999).
Branch of CH3 Cl in N2 : Measurements and Modeling, J.
Mol. Spectrosc., 178, 52 (1996). 173. C. Amiot, J. Verges, C. Effantin and J. d Incan, The KRb
(2) 1 Electronic State, Chem. Phys. Lett., 321, 21 (2000).
157. J.W.G. Seibert, M. Winnewisser, B.P. Winnewisser and
F. Bickelhaupt, Rotational and Rovibrational Spectrum of 174. M. Dulick, K.Q. Zhang, B. Guo and P.F. Bernath, Far- and
C15 NC15 N the Region of the Bending Modes n4 , n5 , the Mid-infrared Emission Spectroscopy of LiH and LiD, J.
Combination Band n4 C n5 and the Fermi Interacting Modes Mol. Spectrosc., 188, 14 (1998).
n3 , 2n4 , J. Mol. Struct., 376, 229 (1996). 175. A. Faye, Q. Kou, R. Farrenq and G. Guelachvili, High-
158. V. Dana, J.Y. Mandin, M.Y. Allout, A. Perrin, L. Regalia, resolution Fourier Transform Spectroscopy of 14 N2 : Analysis
A. Barbe, J.J. Plateaux and X. Thomas, Broadening Param- of the (1-0), (2-1) Bands of the B 3 g W 3 u System, J.
eters of NO2 Lines in the 3.4 m Spectral Region, J. Mol. Spectrosc., 197, 147 (1999).
Quantitat. Spectrosc. Radiat. Transfer, 57, 445 (1997). 176. T. Hirao, C. Dufour, B. Pinchemel and P.F. Bernath,
159. N. Picque and G. Guelachvili, ArHC near 5 m with High Laser-induced Fluorescence and Fourier Transform Spec-
Resolution Double Modulation FTS, Vib. Spectrosc., 19, troscopy of the [21.9] 2 5/2 X 2 3/2 (21 910 cm1 ) and
295 (1999). the [21.9]2 5/2 [0.16]2 5/2 (21 750 cm1 ) Transitions of
NiCl, J. Mol. Spectrosc., 202, 53 (2000).
160. M. Beutel, K.D. Setzer and E.H. Fink, The b 1 C b0C !
X 3 (X1 0C , X2 1) and a 1 a2 ! X2 1 transitions of 177. T. Imajo, K. Yoshino, J.R. Esmond, W.H. Parkinson, A.P.
AsI, J. Mol. Spectrosc., 194, 250 (1999). Thorne, J.E. Murray, R.C.M. Learner, G. Cox, A.S.C. Che-
ung, K. Ito and T. Matsui, The Application of a VUV
161. R.F. Gutterres, J. Verges and C. Amiot, High-resolution Fourier Transform Spectrometer and Synchrotron Radiation
Study of the BaI A Electronic State, J. Mol. Spectrosc., Source to Measurements of: II. The 1, 0 Band of NO, J.
200, 253 (2000). Chem. Phys., 112, 2251 (2000).
162. C. Focsa, P.F. Bernath, R. Mitzner and R. Colin, Fourier
178. S.M. Newman, I.C. Lane, A.J. Orr-Ewing, D.A. Newnham
Transform Emission Spectroscopy of the A 2 X 2 C
and J. Ballard, Integrated Absorption Intensity and Einstein
Transition of BeD, J. Mol. Spectrosc., 192, 348 (1998).
Coefficients for the O2 a 1 g X 3 g (0,0) Transition:
163. C. Focsa, S. Firth, P.F. Bernath and R. Colin, Fourier A Comparison of Cavity Ringdown and High Resolution
Transform Emission Spectroscopy of the A 2 X 2 C Fourier Transform Spectroscopy with a Long-path Absorp-
System of BeH, J. Chem. Phys., 109, 5795 (1998). tion Cell, J. Chem. Phys., 110, 10 749 (1999).
164. O. Shestakov, R. Breidohr, H. Demes, K.D. Setzer and E.H. 179. K. Yoshino, A.P. Thorne, J.E. Murray, K. Ito, T. Matsui,
Fink, Electronic States and Spectra of BiO, J. Mol. Spec- K.W.S. Leung, A.S.C. Cheung and T. Imajo, The Applica-
trosc., 190, 28 (1998). tion of a VUV-FT Spectrometer and Synchrotron Radiation
165. C. Focsa, H. Li and P.F. Bernath, Characterization of the Source to Measurements of the NO and O2 Bands at 295 K,
Ground State of Br2 by Laser-induced Fluorescence Fourier Phys. Chem. Earth Part C, 25, 199 (2000).
Transform Spectroscopy of the B 3 0Cu X 1 Cg System, 180. K. Yoshino, J.R. Esmond, W.H. Parkinson, A.P. Thorne,
J. Mol. Spectrosc., 200, 104 (2000). R.C.M. Learner, G. Cox and A.S.C. Cheung, Fourier Trans-
166. N. Picque, G. Guelachvili, V. Dana and J.Y. Mandin, Abso- form Spectroscopy and Cross Section Measurements of the
lute Line Intensities, Vibrational Transition Moment, and Herzberg III Bands of O2 at 295 K, J. Chem. Phys., 112,
Self-broadening Coefficients for the 3-0 Band of 12 C16 O, J. 9791 (2000).
Mol. Struct., 517, 427 (2000). 181. R.S. Ram, J. Lievin and P.F. Bernath, Fourier Transform
167. S. Payan, C. Camy-Peyret, P. Jeseck, T. Hawat, G. Durry Emission Spectroscopy and Ab Initio Calculations on OsN,
and F. Lefevre, First Direct Simultaneous HCl and ClONO2 J. Chem. Phys., 111, 3449 (1999).
Profile Measurements in the Arctic Vortex, Geophys. Res. 182. K. Ziebarth, K.D. Setzer, O. Shestakov and E.H. Fink,
Lett., 25, 2663 (1998). High-resolution Study of the X2 2 3/2 ! X1 2 1/2 Fine
168. C. Focsa, P.F. Bernath and R. Colin, The Low-lying States Structure Transitions of PbF and PbCl, J. Mol. Spectrosc.,
of He2 , J. Mol. Spectrosc., 191, 209 (1998). 191, 108 (1998).
183. C. Amiot, O. Dulieu and J. Verges, Resolution of the 200. O.N. Ulenikov, G.A. Onopenko, N.E. Tyabaeva, H. Burger
Apparent Disorder of the Rb2 A 1 1u OC 3 C
u and b u Ou and W. Jerzembeck, Isotope Substitution in Near Local
Spectra: A Case of Fully Coupled Electronic States, Phys. Mode H2 X Molecules: The n1 and n3 Bands of D2 Se, J.
Rev. Lett., 83, 2316 (1999). Mol. Spectrosc., 197, 100 (1999).
184. C.E. Fellows, R.F. Gutterres, A.P.C. Campos, J. Verges and 201. A. Farkhsi, H. Bredohl, I. Dubois, F. Remy and A. Fayt,
C. Amiot, The RbCs X 1 C Ground Electronic State: New FT Infrared Spectra of FCN from 1200 to 1800 cm1 and
Spectroscopic Study, J. Mol. Spectrosc., 197, 19 (1999). from 2800 to 7000 cm1 and Global Rovibrational Analysis
185. R.S. Ram, J. Lievin and P.F. Bernath, Fourier Transform of the Main Isotopomers, J. Mol. Spectrosc., 201, 36 (2000).
Infrared Emission Spectroscopy and Ab Initio Calculations 202. C. Camy-Peyret, J.M. Flaud, J.Y. Mandin, A. Bykov,
on RuN, J. Chem. Phys., 109, 6329 (1998). O. Naumenko, L. Sinitsa and B. Voronin, Fourier-transform
186. M. Beutel, K.D. Setzer and E.H. Fink, The b 1 C b0C ! Absorption Spectrum of the H2 17 O Molecule in the
X 3 X1 0C , X2 1 and a 1 a2 ! X2 1 Transitions of SbF, 971111 335 cm1 Spectral Region: The First Decade of
SbCl, SbBr, and SbI, J. Mol. Spectrosc., 195, 147 (1999). Resonating States, J. Quantitat. Spectrosc. Radiat. Transfer,
187. F. Taher, C. Effantin, A. Bernard, J. dIncan and J. Verges, 61, 795 (1999).
The B 1 ! X 1 C System of Scandium Monoiodide, J. 203. M. Carleer, A. Jenouvrier, A.C. Vandaele, P.F. Bernath,
Mol. Spectrosc., 189, 220 (1998). M.F. Merienne, R. Colin, N.F. Zobov, O.L. Polyansky,
188. R.S. Ram, R. Engleman and P.F. Bernath, Fourier Trans- J. Tennyson and V.A. Savin, The Near Infrared, Visible,
form Emission Spectroscopy of the A 2 X 2 Transition and Near Ultraviolet Overtone Spectrum of Water, J. Chem.
of SiH and SiD, J. Mol. Spectrosc., 190, 341 (1998). Phys., 111, 2444 (1999).
189. K.D. Setzer, E.H. Fink and D.A. Ramsay, High-resolution 204. W. Quapp, M. Hirsch, G.C. Mellau, S. Klee, M. Winne-
Fourier-transform Study of the b 1 C ! X 3 and wisser and A. Maki, Climbing the Bending Vibrational
a 1 ! X3 Transitions of SO, J. Mol. Spectrosc., 198, Ladder in D13 C15 N by Hot Gas Emission Spectroscopy,
163 (1999). J. Mol. Spectrosc., 195, 284 (1999).
190. A. Bernard, C. Effantin, J. d Incan, A. Topouzkhanian and 205. R. Cireasa, D. Cossart, M. Vervloet and J.M. Robbe,
G. Wannous, Laser-excited Fluorescence Spectra of Stron- Fourier Transform Emission Spectrum of the HCSi Radi-
tium Monoiodide, J. Mol. Spectrosc., 195, 11 (1999). cal, A 2 C X 2 i Transition, J. Chem. Phys., 112, 10 806
191. R.S. Ram and P.F. Bernath, Fourier Transform Emission (2000).
Spectroscopy of TaO, J. Mol. Spectrosc., 191, 125 (1998). 206. O.N. Ulenikov, G.A. Onopenko, I.M. Olekhnovitch,
192. R.S. Ram and P.F. Bernath, Fourier Transform Emission S. Alanko, V.M. Horneman, M. Koivusaari and R. Anttila,
Spectroscopy of the [12.8] 2 a 2 System of TiCl, J. High-resolution Fourier Transform Spectra of HDS in the
Mol. Spectrosc., 195, 299 (1999). Regions of the Bands n1 and 2n1 /n2 C n3 , J. Mol. Spectrosc.,
189, 74 (1998).
193. R.S. Ram and P.F. Bernath, Fourier Transform Emission
Spectroscopy of the [7.3] 2 a 2 and [9.4] 2 a 2 207. M. Nezu, T. Amano and K. Kawaguchi, Transition Dipole
Systems of ZrCl, J. Mol. Spectrosc., 196, 235 (1999). Moments for the Vibrational Fundamentals of HNC Deter-
mined from the HermanWallis Effect, J. Mol. Spectrosc.,
194. D. McNaughton and L.R. Tay, The High-resolution FTIR
192, 41 (1998).
Spectrum and Equilibrium Structure of Chloro(sulphido)-
boron, J. Mol. Spectrosc., 202, 89 (2000). 208. J.M. Flaud, M. Birk, G. Wagner, J. Orphal, S. Klee and
W.J. Lafferty, The Far Infrared Spectrum of HOCl: Line
195. A. Campargue, D. Bailly, J.L. Teffo, S.A. Tashkun and V.I.
Positions and Intensities, J. Mol. Spectrosc., 191, 362
Perevalov, The n1 C 5n3 dyad of 12 CO2 and 13 CO2 , J. Mol.
Spectrosc., 193, 204 (1999). (1998).
196. R. Rodrigues, K.W. Jucks, N. Lacome, G. Blanquet, J. Wal- 209. A. Ceausu, G. Graner, H. Burger, E.B. Mkadmi, P. Pracna
rand, W.A. Traub, B. Khalil, R. Le Doucen, A. Valentin, and W.J. Lafferty, High-resolution Study of the First
C. Camy-Peyret, L. Bonamy and J.M. Hartmann, Model, Stretching Overtones of H3 Si79 Br, J. Mol. Spectrosc., 192,
Software, and Database for Computation of Line-mixing 47 (1998).
Effects in Infrared Q Branches of Atmospheric CO2 I. 210. I.H. Bachir, T.R. Huet, J.L. Destombes and M. Vervloet,
Symmetric Isotopomers, J. Quantitat. Spectrosc. Radiat. A Combined Analysis of Laser Optogalvanic and Fourier
Transfer, 61, 153 (1999). Transform Emission Spectra of NH2 near its Barrier to
197. R.J. Kshirsagar, L.P. Giver and C. Chackerian, Rovibra- Linearity, J. Mol. Spectrosc., 193, 326 (1999).
tional Intensities of the 000 3 100 0 dyad Absorption 211. J. Orphal, S. Dreher, S. Voigt, J.P. Burrows, R. Jost and
Bands of 12 C16 O2 , J. Mol. Spectrosc., 199, 230 (2000). A. Delon, The Near-infrared Bands of NO2 Observed by
198. R.J. Kshirsagar, L.P. Giver, C. Chackerian and L.R. Brown, High-resolution Fourier-transform Spectroscopy, J. Chem.
The Rovibrational Intensities of the 2n3 Band of 12 C16 O18 O Phys., 109, 10 217 (1998).
at 4639 cm1 , J. Quantitat. Spectrosc. Radiat. Transfer, 61, 212. T.M. Stephen, A. Goldman, A. Perrin, J.M. Flaud, F. Keller
695 (1999). and C.P. Rinsland, New High-resolution Analysis of the
199. W. Jerzembeck, H. Burger, J.M. Flaud and P. Arcas, The 3n3 and 2n1 C n3 Bands of Nitrogen Dioxide NO2 by
n2 Bands of the Deuterated Species D2 Se and HDSe, J. Fourier Transform Spectroscopy, J. Mol. Spectrosc., 201,
Mol. Spectrosc., 197, 215 (1999). 134 (2000).
213. A. Barbe, A. Chichery, V.G. Tyuterev, S.A. Taskhun and 226. J. Orphal, The n1 Bands of 79 BrNO2 and 81 BrNO2 , J. Mol.
S.N. Mikhailenko, Infrared High-resolution Spectra of Struct., 517, 181 (2000).
Ozone in the Range 55005570 cm1 : Analysis of n2 C 5n3
227. R.L. Sams, T.A. Blake, S.W. Sharpe, J.M. Flaud and W.J.
and n1 C n2 C 4n3 Bands, J. Phys. B: At. Mol. Opt. Phys.,
Lafferty, High-resolution Infrared Study of the n14 , n17 , and
31, 2559 (1998).
n18 Bands of 11 B2 H6 and 10 B11 BH6 , J. Mol. Spectrosc., 191,
214. A.J. Bouvier, D. Inard, V. Veyret, B. Bussery, R. Bacis, 331 (1998).
S. Churassy, J. Brion, J. Malicet and R.H. Judge, Contribu- 228. O.N. Ulenikov, G.A. Onopenko, S. Alanko, M. Koivusaari
tion to the Analysis of the 3 A2 X 1 A1 Wulf Transition and R. Anttila, First High-resolution Study of the 13 CHD3
of Ozone by High-resolution Fourier Transform Spectrome- Infrared Spectrum: Rotational Analysis of the Ground State
try, J. Mol. Spectrosc., 190, 189 (1998). and the Fundamentals n5 and n3 /n6 , J. Mol. Spectrosc., 191,
215. A. Chichery, A. Barbe, V.G. Tyuterev and J.J. Plateaux, 9 (1998).
Analysis of High Resolution Measurements of the n1 C 5n3 229. P. Pracna, S. Urban, F. Kolar, J. Cosleou, J. Demaison,
Band of Ozone: Coriolis Interactions with the 6n3 and P. Paplewski and H. Burger, Rovibrational Spectroscopy of
3n1 C n2 C 2n3 Bands, Mol. Phys., 94, 751 (1998). the n2 D 1, and n3 D 1, and n2 D n3 D 1 Levels of CDF3 ,
216. C.P. Rinsland, J.M. Flaud, A. Goldman, A. Perrin, C. Camy- J. Mol. Struct., 517, 119 (2000).
Peyret, M.A.H. Smith, V.M. Devi, D.C. Benner, A. Barbe, 230. A. Predoi-Cross, R.M. Lees and J.W.C. Johns, Fourier
T.M. Stephen and F.J. Murcray, Spectroscopic Parame- Transform Infrared Spectroscopy and Vibrational Coupling
ters for Ozone and its Isotopes: Current Status, Prospects in the OH-bending Band of 13 CH3 OH, J. Mol. Spectrosc.,
for Improvement, and the Identification of 16 O16 O17 O and 191, 348 (1998).
16 17 16
O O O Lines in Infrared Ground-based and Stratospheric
Solar Absorption Spectra, J. Quantitat. Spectrosc. Radiat. 231. M. Herman, M.I. El Idrissi, A. Pisarchik, G. Di Lonardo
Transfer, 60, 803 (1998). and L. Fusina, The Vibrational Energy Levels in Acetylene.
III. 12 C2 D2 , J. Chem. Phys., 108, 1377 (1998).
217. A.J. Bouvier, V. Veyret, I. Russier, D. Inard, S. Churassy,
232. J. Ballard, R.J. Knight and D.A. Newnham, Infrared
R. Bacis, J. Brion, J. Malicet and R.H. Judge, A Compar-
Absorption Cross-sections and Integrated Absorption
ative Rotational Analysis of the 000 Bands of the 3 A2
Intensities of Perfluoroethane and cis-Perfluorocyclobutane,
X 1 A1 Wulf Transition for the Isotopomers 16 O3 and 18 O3
J. Quantitat. Spectrosc. Radiat. Transfer, 66, 199 (2000).
of Ozone by High Resolution Fourier Transform Spectrom-
etry, Spectrochim. Acta, Part A, 55, 2811 (1999). 233. A. Campargue, E. Bertseva, G. Graner and M. Herman,
High-resolution Absorption Spectroscopy of the 3n1 and
218. A. Perrin, J.M. Flaud, A. Valentin, C. Camy-Peyret, N. Gba-
3n1 C n3 Bands of Propyne, J. Mol. Spectrosc., 201, 156
guidi and A. NGom, The n1 and n3 Bands of the 16 O17 O17 O
(2000).
Isotopomer of Ozone, J. Mol. Struct., 517, 157 (2000).
234. G. Di Lonardo, L. Fusina, G. Masciarelli and F. Tullini,
219. A. Perrin, J.M. Flaud, A. Valentin, C. Camy-Peyret and Integrated Band Strengths of Benzene Vapour in the
H. Gbaguidi, The n1 and n3 bands of the 17 O16 O17 O Iso- 6001900 cm1 Region, Spectrochim. Acta, Part A, 55,
topomer of Ozone, J. Mol. Spectrosc., 200, 248 (2000). 1535 (1999).
220. S. Naim, A. Fayt, H. Bredohl, J.F. Blavier and I. Dubois, 235. P. Stoppa and A. Gambi, High-resolution FTIR Spec-
Fourier Transform Spectroscopy of Carbonyl Sulfide from trum of Freon-13 at 1890 cm1 , J. Mol. Struct., 517, 209
3700 to 4800 cm1 and Selection of a Line-pointing Pro- (2000).
gram, J. Mol. Spectrosc., 192, 91 (1998).
236. A.E. Heathfield, C. Anastasi, J. Ballard, D.A. Newnham
221. E. Rbaihi, A. Belafhal, J. Vander Auwera, S. Naim and and A. McCulloch, Integrated Infrared Absorption Co-
A. Fayt, Fourier Transform Spectroscopy of Carbonyl Sul- efficients of CF3 OCF2 H and CH3 OCF2 CF2 H at 297, 253
fide from 4800 to 8000 cm1 and New Global Analysis of and 213K, J. Quantitat. Spectrosc. Radiat. Transfer, 59, 91
16 12 32
O C S, J. Mol. Spectrosc., 191, 32 (1998). (1998).
222. T. Strugariu, S. Naim, A. Fayt, H. Bredohl, J.F. Blavier and 237. J. Pietila, V.M. Horneman, R. Anttila, B. Lemoine, F. Ray-
I. Dubois, Fourier Transform Spectroscopy of 18 O-enriched naud and J.M. Colmont, The Perpendicular Fundamental n5
Carbonyl Sulfide from 1825 to 2700 cm1 , J. Mol. Spec- of Chloroform 12 CH3 35 Cl3 : High Resolution Infrared Study
trosc., 189, 206 (1998). of the n5 Band Together with the Millimetre-wave Rotational
223. M. Litz, H. Burger, L.S. Masukidi, A. Fayt, J. Cosleou, Spectrum, Mol. Phys., 98, 549 (2000).
P. Drean, L. Margules and J. Demaison, Mid- and Near- 238. L. Nemes, D. Luckhaus, M. Quack and J.W.C. Johns,
infrared Spectra, Millimeterwave Spectra, and Global Anal- Deperturbation of the Low-frequency Infrared Modes of
ysis of 16 O12 C80 Se, J. Mol. Spectrosc., 196, 155 (1999). Ketene CH2 CO, J. Mol. Struct., 517, 217 (2000).
224. P.M. Chu, S.J. Wetzel, W.J. Lafferty, A. Perrin, J.M. Flaud, 239. K.M. Smith, G. Duxbury, D.A. Newnham and J. Ballard,
P. Arcas and G. Guelachvili, Line Intensities for the 8-m A High-resolution Analysis of the nu(3) and nu(9) Absorp-
Bands of SO2 , J. Mol. Spectrosc., 189, 55 (1998). tion Bands of Difluoromethane, J. Mol. Spectrosc., 193, 166
225. G. Stark, P.L. Smith, J. Rufus, A.P. Thorne, J.C. Pickering (1999).
and G. Cox, High-resolution Photoabsorption Cross-section 240. P. Drean, H. Burger, J. Demaison and J.E. Boggs, Struc-
Measurements of SO2 at 295 K between 198 and 220 nm, ture, Ab Initio Harmonic Force Field and Infrared Spectrum
J. Geophys. Res. Planets, 104, 16 585 (1999). of CH3 CAs, J. Mol. Struct., 486, 51 (1999).
241. O.N. Ulenikov, G.A. Onopenko, N.E. Tyabaeva, R. Anttila, 256. I. Kleiner, L.R. Brown, G. Tarrago, Q.L. Kou, N. Picque,
S. Alanko and J. Schroderus, Rotational Analysis of the G. Guelachvili, V. Dana and J.Y. Mandin, Positions and
Ground State and the Lowest Fundamentals n3 , n5 , and n6 , Intensities in the 2n4 /n1 /n3 Vibrational System of 14 NH3 near
of 13 CH3 D, J. Mol. Spectrosc., 201, 9 (2000). 3 m, J. Mol. Spectrosc., 193, 46 (1999).
242. I. Mukhopadhyay, P.K. Gupta, G. Moruzzi, B.P. Winne- 257. M. Snels, L. Fusina, H. Hollenstein and M. Quack, The n1
wisser and M. Winnewisser, High Resolution Fourier Trans- and n3 Bands of ND3 , Mol. Phys., 98, 837 (2000).
form Far Infrared Spectroscopy of CH3 OD, J. Mol. Spec-
258. K. Kawaguchi, T. Ishiwata, E. Hirota and I. Tanaka,
trosc., 186, 15 (1997).
Infrared Spectroscopy of the NO3 Radical, Chem. Phys.,
243. O.N. Ulenikov, G.A. Onopenko, N.E. Tyabaeva, S. Alanko, 231, 193 (1998).
M. Koivusaari and R. Anttila, Precise Study of the Lowest
259. M. Hepp, R. Georges, M. Herman, J.M. Flaud and W.J.
VibrationRotational Bands n5 and n3 /n6 of the CHD3
Lafferty, Striking Anharmonic Resonances in N2 O4 : Super-
Molecule, J. Mol. Spectrosc., 186, 293 (1997).
sonic Jet Fourier Transform Spectra at 13.3, 7.9, 5.7 and
244. J. Orphal, M. Morillon-Chapey, S. Klee, G.C. Mellau and 3.2 m, J. Mol. Struct., 517, 171 (2000).
M. Winnewisser, The Far-infrared Spectrum of ClNO2
Studied by High-resolution Fourier-transform Spectroscopy, 260. P. Asselin, P. Soulard and N. Lacome, The NO Dimer: Jet-
J. Mol. Spectrosc., 190, 101 (1998). cooled Study of n1 and n5 Transitions, J. Mol. Spectrosc.,
190, 274 (1998).
245. A.R.W. McKellar, High-resolution Infrared Spectrum and
Energy Levels of the Weakly Bound Complex, CO-orthoD2 , 261. A. Dkhissi, N. Lacome and A. Perrin, The NO Dimer
J. Chem. Phys., 112, 9282 (2000). II. Analysis of the n1 Band, J. Mol. Spectrosc., 194, 156
(1999).
246. G. Moruzzi, W. Jabs, B.P. Winnewisser and M. Winne-
wisser, Assignment and Power Series Analysis of the FIR 262. A. Benidar, R. Georges, R. Le Doucen, J. Boissoles,
Fourier Transform Spectrum of Cyanamide Using a Multi- S. Hamon, A. Canosa and B.R. Rowe, Uniform Supersonic
molecule Ritz Program, J. Mol. Spectrosc., 190, 353 (1998). Expansion for FTIR Absorption Spectroscopy: The n5 Band
of (NO)2 at 26 K, J. Mol. Spectrosc., 199, 92 (2000).
247. A. Perrin, J.M. Flaud, A. Predoi-Cross, M. Winnewisser,
B.P. Winnewisser, G. Mellau and M. Lock, New High- 263. D.A. Newnham and J. Ballard, Visible Absorption Cross
resolution Analysis of the n3 , n4 , and n6 Bands of D2 CO Sections and Integrated Absorption Intensities of Molecular
Measured by Fourier Transform Spectroscopy, J. Mol. Spec- Oxygen O2 and O4 , J. Geophys. Res. Atm., 103, 28 801
trosc., 187, 61 (1998). (1998).
248. J.E. Gadhi, H. Burger and E.B. Mkadmi, Fourier Trans- 264. V. Boudon, E.B. Mkadmi, H. Burger and G. Pierre, High-
form Spectroscopy of the n3 and n6 Bands of D3 Si35 Cl: resolution Fourier Transform Infrared Spectroscopy and
Rovibrational Constants of the n D 0, n3 D 1, and n6 D 1 Analysis of the n3 Fundamental Band of P4 , Chem. Phys.
States, J. Mol. Spectrosc., 191, 311 (1998). Lett., 305, 21 (1999).
249. K. Smith, D. Newnham, M. Page, J. Ballard and G. Dux- 265. V. Boudon, M. Hepp, M. Herman, I. Pak and G. Pierre,
bury, Infrared Absorption Cross-sections and Integrated High-resolution Jet-cooled Spectroscopy of SF6 : The n2 C
Absorption Intensities of HFC-134 and HFC-143a Vapour, n6 Combination Band of 32 SF6 and the n3 Band of the Rare
J. Quantitat. Spectrosc. Radiat. Transfer, 59, 437 (1998). Isotopomers, J. Mol. Spectrosc., 192, 359 (1998).
250. F. Keller, A. Perrin, J.M. Flaud, J.W.C. Johns, Z. Lu and 266. K. Sarka, J. Demaison, L. Margules, I. Merke, N. Heine-
E.C. Looi, High-resolution Analysis of the n6 , n7 , n8 , and king, H. Burger and H. Ruland, Centrifugal Distortion
n9 Bands of H15 NO3 Measured by Fourier Transform Spec- Analysis of a Near-spherical Top, SO2 F2 : The First Deter-
troscopy, J. Mol. Spectrosc., 191, 306 (1998). mination of all Six Quartic Centrifugal Distortion Con-
251. S. Klee, M. Winnewisser, A. Perrin and J.M. Flaud, Abso- stants for an Asymmetric Top, J. Mol. Spectrosc., 200, 55
lute Line Intensities for the n6 Band of H2 O2 , J. Mol. (2000).
Spectrosc., 195, 154 (1999). 267. L. Fusina, G. Di Lonardo and P. De Natale, The Ground
252. D. Papousek, M. Winnewisser, S. Klee, J. Demaison and State Spectroscopic Parameters and Molecular Geometry of
P. Pracna, Simultaneous Rovibrational Analysis of the n2 , SbH3 , J. Chem. Phys., 109, 997 (1998).
n3 , n5 , and n6 Bands of H3 13 CF, J. Mol. Spectrosc., 192, 268. H. Burger, J. Demaison, P. Drean, W. Jerzembeck, I. Merke
220 (1998). and W. Stahl, High Resolution Infrared, Microwave, and
253. J.X. Cheng, D. Wang and G. Graner, High-resolution Millimeter Wave Spectra of SeCF2 Structure Determination
Infrared Spectrum of H3 SiI in the n1 /n4 Region Near Assisted by Ab Initio Calculations, Ber. Bunsen-Ges. Phys.
2200 cm1 , J. Mol. Spectrosc., 190, 240 (1998). Chem., 102, 872 (1998).
254. A. Perrin, J.M. Flaud, F. Keller, A. Goldman, R.D. Blather- 269. M. Rotger, V. Boudon, B. Lavorel, S. Sommer, H. Burger,
wick, F.J. Murcray and C.P. Rinsland, Analysis of the J. Breidung, W. Thiel, M. Betrencourt and J.C. Deroche,
n8 C n9 Band of HNO3 , Line Positions and Intensities, and Ab Initio Calculations and High-resolution Spectroscopy
Resonances Involving the v6 D v7 D 1 Dark State, J. Mol. of the Bending Pentad of SiH2 D2 in the 1016 m Region,
Spectrosc., 194, 113 (1999). J. Mol. Spectrosc., 192, 294 (1998).
255. H. Burger, W. Jerzembeck, H. Ruland and L. Halonen, 270. D. Priem, J. Cosleou, J. Demaison, I. Merke, W. Stahl,
High Resolution FTIR Study of the n5 Bands of HSiD3 W. Jerzembeck and H. Burger, Analysis of the Rotational
and H120 SnD3 , J. Mol. Spectrosc., 189, 8 (1998). Spectra of SiH3 CN and its Isotopomers: Experimental and
Ab Initio Determinations of the Dipole Moment and the 281. G. Durry and G. Guelachvili, N2 (B-A) Time-resolved
Structure, J. Mol. Spectrosc., 191, 183 (1998). Fourier Transform Emission Spectra from a Pulsed Micro-
271. M. Terki Hassaine, C. Claveau, A. Valentin and G. Pierre, wave Discharge, J. Mol. Spectrosc., 168, 82 (1994).
Analysis of the Infrared Fourier Transform Spectrum of 282. L. Letendre, D.K. Liu, C.D. Pibel, J.B. Halpern and H.L.
the Spectra of Silane in the Range 29303300 cm1 , J. Dai, Vibrational Spectroscopy of a Transient Species
Mol. Spectrosc., 197, 307 (1999). through Time-resolved Fourier Transform Infrared Emission
272. M. Halonen, L. Halonen, H. Burger and W. Jerzem- Spectroscopy: The Vinyl Radical, J. Chem. Phys., 112, 9209
beck, Vibrational Energy Localization in the Stretching (2000).
Vibrational (1000 A1 /F2 ), (2000 A1 /F2 ), and (3000 A1 /F2 ) 283. R.A. Loomis, J.P. Reid and S.R. Leone, Photofragmenta-
Band Systems of 120 SnD4 , J. Chem. Phys., 108, 9285 tion of Ammonia at 193.3 nm: Bimodal Rotational Distribu-
(1998). Q J. Chem. Phys.,
tions and Vibrational Excitation of NH2 A,
273. O.I. Baskakov, H. Burger and W. Jerzembeck, The 112, 658 (2000).
Coriolis-coupled States n7 D 1 and n9 D 1 of trans- 284. N. Picque and G. Guelachvili, High-information Time-
HCOOD and of trans-DCOOD, J. Mol. Spectrosc., 193, 33 resolved Fourier Transform Spectroscopy at Work, Appl.
(1999). Opt., 39, 3984 (2000).
274. L.M. Goss, S.W. Sharpe, T.A. Blake, V. Vaida and J.W. 285. K. Yoshino, P.L. Smith, W.H. Parkinson, A.P. Thorne and
Brault, Direct Absorption Spectroscopy of Water Clusters, K. Ito, The Combination of a VUV Fourier-transform Spec-
J. Phys. Chem., 103, 8620 (1999). trometer and Synchrotron Radiation, Rev. Sci. Instrum., 66,
275. X. Hong and T.A. Miller, Velocity Modulated Fourier 2122 (1995).
Transform Emission as a Plasma Diagnostic and a 286. R. Paso and V.M. Horneman, High-resolution Infrared
Spectroscopic Tool, J. Chem. Phys., 101, 4572 (1994). Measurement of the n10 Band of Propyne-d1 Using a Syn-
276. X.C. Hong and T.A. Miller, Determination of Concentra- chrotron Radiation Source, J. Mol. Spectrosc., 172, 536
tions via the Diminution of Doppler Shifts by Radiation (1995).
Trapping, Chem. Phys., 228, 131 (1998). 287. C. Domingo, A. del Olmo, R. Escribano, D. Bermejo and
277. N. Picque, Wide-band Spectroscopic Investigation of the J.M. Orza, Fourier Transform Intracavity Laser Absorption
State-to-state Dependence of the ArHC Ion Average Mobil- Spectra of the 6n1 Band of CHD3 , J. Chem. Phys., 96, 972
ity in a Ar/He Plasma, Chem. Phys. Lett., 310, 183 (1992).
(1999). 288. K. Strong, T.J. Johnson and G.W. Harris, Visible Intracav-
278. N. Picque and G. Guelachvili, High-resolution Multimodu- ity Laser Spectroscopy with a Step-scan Fourier-transform
lation Fourier-transform Spectroscopy, Appl. Opt., 38, 1224 Interferometer, Appl. Opt., 36, 8533 (1997).
(1999). 289. J.X. Cheng, H. Lin, S.M. Hu, S.G. He, Q.S. Zhu and
279. G. Guelachvili, Selective Detection of Paramagnetic Species A. Kachanov, Infrared Intracavity Laser Absorption Spec-
by High-information Fourier-transform Spectrometry, J. troscopy with a Continuous-scan Fourier-transform Interfer-
Opt. Soc. Am. B, 3, 1718 (1986). ometer, Appl. Opt., 39, 2221 (2000).
280. P.A. Martin and G. Guelachvili, Modulation and Selective 290. M. Nela, D. Permogorov, A. Miani and L. Halonen,
Detection of Transient Species in High Resolution FTS, Vibration-rotation Fluorescence Spectra of Water in the
Spectrochim. Acta, Part A, 51, 1117 (1995). Ground Electronic State, J. Chem. Phys., 113, 1795 (2000).
Instrumentation and Methods for High-resolution
Gas-phase Spectroscopy
Don McNaughton
Monash University, Clayton, Australia
1 INTRODUCTION 0.001
0.008
The term high-resolution spectroscopy requires definition
FWHH/cm1
because it can have different meanings depending on the M = 30 amu
0.006
background of the vibrational spectroscopist and the exper-
M = 50 amu
imental work envisaged. For the gas phase it is best defined 0.004
as the resolution required to resolve the rotational structure
M = 100 amu
of vibrational bands and apart from the spectra of very 0.002
light molecules and diatomics this requires a resolution of
better than 0.05 cm1 . Although specialized techniques per- 0.000
500 1000 1500 2000 2500 3000 3500 4000
mit sub-Doppler spectroscopy, the normal resolution limit
Wavenumber/cm1
is determined first by line broadening due to the Doppler
Figure 1. Dependence of Doppler width on wavenumber and
effect and second by the experimental considerations of molecular mass.
pressure broadening. The Doppler line width, expressed as
full width at half-height (FWHH) of the absorption line is carried out between 0.001 and 0.05 cm1 . Pressure broad-
given by ening depends on the mean time between collisions in a
gaseous sample and parameters for pressure broadening for
nN 8NkT ln 2 1/2
FWHH D 2nN D 1 a range of small molecules are given by Rothman et al.,1
c M
but in general to achieve the lower resolution limits gas
where N is Avogadros number, k is the Boltzmann con- pressures of <1 Torr (1 Torr D 133.3 Pa) are required.
stant, T is the temperature, c is the speed of light, nN is
the reciprocal wavelength in wavenumber units (cm1 )
and M is the molecular mass in atomic mass units. The 2 FOURIER TRANSFORM INFRARED
Doppler width thus depends on the gas temperature, the SPECTROSCOPY
molecular mass and the spectral region. Figure 1 shows the
wavenumber dependence of the Doppler width for a range Prior to the 1980s, broad-band high-resolution spectroscopy
of molecular masses at 298 K. The linear relationship with was carried out with instruments based on spectral disper-
temperature leads to halving of the line width with a halv- sion using gratings. For high resolution this required mas-
ing of temperature but with, of course, a resultant reduction sive instruments, very long scan times and careful calibra-
of vapor pressure and hence signal strength. For Doppler- tion to achieve useful chart spectra which were then labori-
limited work, high-resolution spectroscopy is thus generally ously assigned. The advent of the Fourier transform infrared
(FT-IR) technique with its throughput and multiplex advan-
John Wiley & Sons Ltd, 2002. tages and internal calibration, in addition to triggering a
resurgence in analytical infrared (IR) spectroscopy, also Michelson interferometer and the following discussion will
made possible the development of high-resolution instru- be mostly based around this instrument, which can be
ments that rely on the same advantages. Thus all of operated from 10 cm1 through the IR, visible and into the
the theoretical and instrumental considerations outlined in ultraviolet (UV) region.
Continuous Scanning Interferometers for Mid-infrared
Spectrometry also apply to high-resolution instruments,
although a number of these considerations are of greater 2.1 Optical considerations
importance when considering high-resolution instruments.
A schematic of the major commercial high-resolution FT-IR 2.1.1 Instrument line shape and resolution
instrument, the Bruker IFS 120HR (Bruker Optik GmbH,
Ettlingen, Germany), is shown in Figure 2. The unapodized resolution of a Fourier transform (FT)
Apart from the long interferometer arm and the physical instrument is given by
size of the instrument, it is a very similar diagram to 1 1
that detailing a standard FT-IR instrument based on the nN D D 2
OPD 2L
S4
AC
Scanner M1 S3
BS
S2
S1
M2
Source
S1S3 Internal sources compartment
S4 External source
AC Aperture changer FC
FC Filter changer
BS Beamsplitter AC
M1M2 Retroreflectors
D1D6 Detectors
Front Back
Sample
channel channel
compartment
D3 D4
D1 D2
D5 D6
Detector
compartment
Figure 2. Schematic of Bruker IFS 120HR FT-IR spectrometer.

Instrumentation and Methods for High-resolution Gas-phase Spectroscopy 3
where L is the travel length of the moving mirror and a motor-driven pulley system that is dynamically aligned
OPD is the optical path difference. In practice, a finite using the signal of an HeNe laser beam that is indepen-
interferogram is achieved by convolving the spectrum at dent of the calibration laser path. The problems associated
infinite path difference with the instrument line shape (ILS) with keeping the mirror aligned over this distance, together
function, which for a boxcar cutoff is the sinc function. This with the sheer physical dimensions of such an instrument,
leads to a resolution given by ensure that most other instruments are not designed to
achieve a resolution of 0.001 cm1 . Those that do approach
0.61
nN D 3 this limit, e.g. the Bomem DA3-002 (Bomem, Inc., Que-
OPD bec City, Canada) does so by doubling the beam along the
but with side lobes that are up to 22% of the main same path length, thus restricting the arm of the interfer-
lobe amplitude as shown in Figure 3. This phenomenon ometer to 2.5 m but resulting in increased optical losses
is known as leakage. in the double-pass system. Air bearings and the simple
In practice, the side lobes due to leakage must be reduced devices used in some low-resolution instruments are not
and to do this the interferogram must be truncated less appropriate for high-resolution work, although the very first
abruptly with a function that results in less intense side high-resolution instrument introduced by Nicolet Instru-
lobes. The apodized (or feet removed) resolution is further ments, Inc. (Madison, WI, USA), capable of an unapodized
reduced in this process and so the optimum resolution is resolution of 0.04 cm1 , operated with an air bearing. Other
not usually achieved. For high-resolution gas-phase work, high-resolution instruments are evacuable to ensure the
a useful compromise is to convolve the spectrum with a removal of water vapor and carbon dioxide. Water vapor
trapezoidal function with only the last four points used to in particular results in a large number of strong vibra-
cut off the interferogram less abruptly. This reduces the feet tionrotation and rotation lines and in the far-IR region in
considerably with minimum line broadening. The line width particular, lines in large regions of the spectrum are easily
and line shape are always a function of both the ILS and swamped by water absorptions. Bomem, who introduced
the natural line width and natural line shape (NLS), which the DA-002, the other major commercial high-resolution
depends on the Doppler effect and pressure broadening as instrument in the early 1980s solved some of the prob-
discussed above, and to achieve the NLS the ILS must be lems associated with mirror travel by having a dynamically
small compared with the NLS. aligned vertical interferometer arm driven by a pulley sys-
The above discussion means that for a theoretical boxcar tem. This generated a high rather than a long instrument
apodized resolution of 0.001 cm1 , the moving mirror must and to achieve <0.0024 cm1 the beam had to be doubled
travel >3 m. The Bruker IFS 120HR, which is presently through the interferometer.
the major commercial high-resolution instrument, has a
maximum mirror travel of 3.5 m and so theoretically 2.1.2 Aperture and resolution
achieves a resolution of <0.001 cm1 . This resolution is
achieved in the very far-IR region but Doppler broadening The finite aperture (or image stop) of an FT-IR instrument
ensures that in the mid-IR region the actual line widths are introduces oblique rays in the parallel beam which lead
>0.001 cm1 . The drive mechanism for the instrument is to the introduction of circular fringes at the interferometer
at finite path difference. Oversize image stops thus result
in a loss of signal that is most extreme at maximum
path difference and this effectively further apodizes the
interferogram, leading to a consequent loss of resolution.
The diameter of the image stop required to attain a desired
resolution is wavenumber dependent and determined by the
L sin(2vL)
0.61/L focal length of the instruments collimating mirror. The
(2vL)
diameter d is given by

nN
d D 2F 4
nN
where F is the focal length of the parabolic collimating
mirror. For the Bruker IFS120 the focal length is 42 cm and
Figure 4 shows the relationship between aperture size and
wavenumber for a range of resolutions. Thus for very high
Figure 3. ILS of an FT-IR spectrometer the sinc function. resolution in the mid-IR region it is necessary to use very
10 single-scan spectrum taken at an unapodized resolution of

0.002 cm1 requires 5 min to collect. Even with highly
sensitive detectors and long path lengths, it is usually nec-
Aperture diameter/mm
8
essary to co-add 20 or more scans to attain a good S/N and
6 so a typical spectrum at very high resolution takes 2 h.
v = 0.1 cm1 One final but essential optical component for high-
4 resolution FT-IR spectroscopy is an optical filter set, which
has two functions. For photon noise-limited detectors inclu-
2 v = 0.01 cm1 sion of cutoff or band-pass filters reduces the photon noise
on the detector and for helium-cooled far-IR bolometers
v = 0.001 cm1
0
cooled cutoff filters are essential to reduce thermal noise
4000 3500 3000 2500 2000 1500 1000 500 that would otherwise saturate the detector. The second func-
Wavenumber/cm1 tion of filters is to allow undersampling, as discussed below.
Figure 4. Relationship between aperture size and wavenumber
for a range of resolutions.
2.2 Data handling
small apertures with consequently small throughput. For
far-IR spectroscopy much larger apertures and consequently
In addition to leakage, use of the discrete Fourier transform
greater throughput are possible.
(DFT) leads to spectral aliasing and the picket fence effect,
where the spectrum appears to the viewer as though seen
2.1.3 Other optical components through a grill. For an explanation of the picket fence
To obtain the high resolving power and retain throughput, effect, see Herres and Gronholtz.2 A DFT of an N-point
the collimating mirrors in high-resolution instruments are interferogram yields N/2 meaningful output points, with
of large diameter and long focal length. The Bruker the last N/2 points yielding the mirror image along the
IFS120 has an off-axis parabolic reflector of F D 42 cm wavenumber axis folded around the Nyquist number as
with a diameter 90 mm compared with typical low- shown in Figure 5(a). The folding or Nyquist wavenumber
resolution parabolic reflectors of F 15 cm and diameters is given by
1
of 2030 mm. As a consequence of this, all the mirrors nN f D 5
2x
and hence the instrument bench are considerably larger and
of course considerably more expensive than those of low- where x is the sampling interval. This mirror symmetric
resolution instruments. With the large decrease in signal-to- spectrum is repeated endlessly along the wavenumber axis
noise ratio (S/N) with increased resolution it is also essential as shown in Figure 5(b), hence the term aliasing. If the
to use highly sensitive detectors such as liquid nitrogen- spectrum contains a nonzero component above the folding
cooled mercury cadmium telluride (MCT)InSb detectors wavenumber, this will fold back and appear at the wrong
in the mid-IR region and liquid helium-cooled bolometers position. FT-IR software therefore has to choose the maxi-
in the far-IR region. To calibrate spectra taken at high mum sample spacing for which no overlap occurs in order
resolution, it is necessary to use a temperature-stabilized to minimize the interferogram size. In high-resolution work
single-mode HeNe calibration laser with a frequency sta- this becomes extremely important and it is necessary to
bility of 1 part in 107 . This results in a high wavenumber decrease the interferogram size even further by undersam-
precision that is not necessary for low-resolution work pling. Undersampling is achieved if the lower limit of the
where much cheaper calibration lasers can be employed. spectrum is not zero, as in Figure 5(c), and the lower limit
This laser is also used to align dynamically the pulley is related to the upper folding limit by
driven scanning retro-reflector mirror with the speed of
n1
the mirror drive dependent on the detector response time. nN fl D nN fu n D 1, 2, 3, . . . 6
Deuterated triglycine sulfate (DTGS) detectors, for exam- n
ple, have a slow response time which limits the scanning where the upper folding wavenumber nfu must be a natural
mirror speed to 0.25 cm s1 , and this combined with their fraction of the calibration laser wavenumber. An increase
low sensitivity makes them essentially useless for serious in the sample spacing by a factor n will then cause the
high-resolution spectroscopy. CuGe bolometers also have aliases to fill the wavenumber range below the lower fold-
a slow response time but their inherent sensitivity makes ing limit and, provided that spectral filters are incorporated
them useful at low wavenumber values. The dynamic align- to ensure zero spectrum in this area, the spectrum can be
ment restricts the scanning speeds to 1.25 cm s1 and a calculated from the reduced size alias spectrum as shown in

vf mode gaps or narrow tunable ranges. Tunable diode laser
spectroscopy (TDLS) systems are the most commonly used
systems for high-resolution gas-phase work because diodes
(a) 0 N/2 N are available to cover most parts of the IR spectrum from
300 to 3000 cm1 . The instrumentation of TDLS is covered
elsewhere.
Other tunable laser sources are available for high-
(b) 0 N 2N
resolution gas-phase work but generally are restricted in
the bandwidth that can be covered and require dedicated
laser laboratories and an increased commitment to laser
vf l = 43 vfu maintenance. The book by Hollas3 describes a number of
(c) 0
vf l
vfu these systems, but a brief survey of the available techniques
follows.
Nonlinear effects lead to a number of methods of gen-
erating tunable IR radiation using nonlinear crystals and
(d) 0
vfu/4
vfl
vfu high-power lasers and two of these are used extensively.
Difference frequency laser systems, based on mixing a
Figure 5. Folding and aliasing in FT-IR spectroscopy. (a) Both
the spectrum and mirror image emerge from the DFT; (b) aliasing tunable single-mode dye laser with a single-frequency
the spectrum and mirror image is endlessly repeated; (c) spectrum argon ion laser in a temperature-controlled lithium nio-
is zero above and below upper and lower wavenumber; (d) bate crystal, have been available for some time and yield
undersampling produces aliases from which the true spectrum can 45452380 cm1 tunable radiation. The optical paramet-
be calculated.
ric oscillator (OPO) produces tunable radiation with the
Figure 5(d). A detailed explanation of aliasing and under- frequency output dependent on the frequency of a high-
sampling can be found in a publication from Bruker.2 power source laser and the nature and tunability of the
Undersampling allows the size of the high-resolution inter- nonlinear crystal, and the latest OPOs provide a tunable
ferogram to be reduced to a point where the Fourier source down to 2000 cm1 with line widths as small as
transformation can be carried out in the control computer 0.020.004 cm1 .4 7
memory. Color center or F-center lasers generally operate in the
The picket fence effect arises if the interferogram con- near-IR region but radiation down to around 2800 cm1
tains spectral features that do not coincide with the sampling is achievable with very narrow line widths of the order of
points of the DFT. For broad features in a spectrum this 106 cm1 and so the color center laser complements TDLS
is not so important but for the highly resolved compo- by covering the range above 3000 cm1 that is inaccessible
nents in a high-resolution gas-phase spectrum the output to TDLS.
appears as though one is looking at the spectrum through Microwave or millimeter-wave sideband lasers based on
a picket fence or grill. This can be overcome by oversam- microwave tuning of a single-frequency CO2 laser pro-
pling with its longer collection times and large data files vide a tunable radiation source covering the region around
or by adding zeros to the end of the interferogram prior to the laser wavenumber values of 1041 and 943 cm1 with
Fourier transformation. The latter is termed zero filling and a spectral resolution of 104 cm1 . The tuning range
effectively interpolates the spectrum between data points. depends on the microwave source but is usually small.
For high-resolution gas-phase spectra a zero filling factor Similar systems in which a CO2 laser is used to pump a
of 8 is appropriate. far-IR laser provide tunable sources in the 10200 cm1
range. This range is achieved using a large number of
molecules as the lasing medium with each molecule pro-
3 LASER TECHNIQUES viding 5 cm1 range around the lasing line. This latter
technique is used mainly for rotational spectroscopy but
Owing to their high photon flux, lasers can be many orders has also been used for pure IR spectroscopy of gases. The
of magnitude more sensitive than FT methods and the lines wavenumber ranges covered by these sideband techniques
from a laser may be much narrower, allowing resolutions are also accessible using the same IR and far-IR lasers by
at least a factor of 10 better than FT-IR. The major dis- using an electric or magnetic field to tune the energy lev-
advantages of laser-based high-resolution spectroscopy are els of the molecule under study. In the far-IR region this
the long scanning times, the need for calibration gases and is termed laser magnetic resonance (LMR) and a magnetic
etalons and incomplete coverage of spectral bands due to field is used for tuning. The molecules under study must
have a magnetic moment in order to be tunable and so detector must be mounted externally to the FT-IR system.
the technique is useful mainly for species with unpaired Such an external arrangement is still limited to a relatively
electrons. In the mid-IR region the tuning is done with an short pathlength and requires a large sample volume.
electric field usually using CO, N2 O or CO2 laser lines. An alternative method that minimizes the sample volume
This technique is termed laser Stark spectroscopy and is is a multiple traversal cell that can be mounted vertically
useful for any species with a permanent electric dipole. above the normal sample chamber or horizontally using a
set of purpose-designed transfer optics. There are a number
of designs available with the most common being the
4 COUPLING OF EQUIPMENT TO White cell9 and the Herriot cell.10 Such cells are reasonably
HIGH-RESOLUTION efficient using silver-coated mirrors and up to 20 or even 40
SPECTROMETERS multiple passes of the beam give a pathlength of between
4 and 100 m depending on the size of the cell and the
4.1 Signal-to-noise ratio refocusing mirrors. For atmospheric work, specialized cells
with pathlengths >1 km long have been built.11
Particularly for FT-IR spectroscopy, but also for laser spec- For gas-phase laser spectroscopy, the IR beam is usually
troscopy, it is often important to increase the S/N of spectra a very small diameter collimated beam and simple flat
in order to allow analysis, especially for weak absorbers mirrors can sometimes be used to attain long pathlengths in
where the necessity of reducing the sample pressure below relatively simple optical arrangements and with very small
1 Torr can lead to extremely weak signals. Signal averag- sample volumes.
ing is the first methodology used to do this but it is limited
by the square root law, which places a time ceiling on just
how long it is worth averaging data. The general method of 4.2 Spectral simplification
experimentally increasing the S/N is to increase the sample
pathlength and take advantage of Beers law, which can Even at 0.0010.005 cm1 resolution, the spectra of rel-
be simply achieved by increasing the length of the absorp- atively simple molecules consist of a large number of
tion cell. To do this to any significant extent with an FT-IR overlapping lines that are often intractable when analysis
instrument requires the use of a parallel beam rather than is attempted. Much of this spectral overlap is due to the
the normal focused beam that is available in the instru- large number of vibrational and rotational energy states that
ment sample chamber. Figure 6 shows an example of such are populated at room temperature and a number of tech-
a system where a 90 mm collimated beam is taken out of niques have been developed to reduce this spectral clutter.
the 120HR instrument before entering the instrument sam- Simple cooling of the absorption cell reduces the number
ple chamber and then directed down a long absorption cell of thermally populated states but for most molecules the
with a sample pathlength of 3 m.8 In such systems the amount of cooling and hence simplification are severely
Valve Detector
To beamsplitter compartment
Pressure
gauge Sample inlet
Pyrex glass cell
Oven
Bellows
Inlet
Valve Polythene
window
Pump MCT & InSb

Conventional exhaust detectors
sample
compartments Liquid nitrogen trap Si-B bolometer
Figure 6. Long single-path length cell coupled to a Bruker IFS 120HR instrument.
Sample inlet
Detector Nozzle
sample compartment
Nozzle horizontal
Brker IFS 120HR
position adjustment height

Supporting stage adjustment
M2 Thermocouple and B Detector
M1 heating leads
B M4 Mirror assembly
F2
translation and swivel
F1 KBr
M3 T
window
Cryopump
B vanes
Pressurized Pressure
cushion gauge
Interaction zone
Filter
Rotary
pump Compressed
air
External adjustment of
vertical and horizontal slant
for 1st and last mirror
Figure 7. Jet nozzle expansion coupled to a Bruker IFS 120HR instrument using a multipass system. [Reproduced by permission of The
Royal Society of Chemistry from D. McNaughton, D. McGilvery and E.G. Robertson J. Chem. Soc., Faraday Trans. 90, 10551071
(1994).]
restricted by condensation. An alternative method is to 0.125

entrain the molecule of interest in a supersonic jet expansion
Absorbance
by expanding the gas from a high-pressure reservoir through

a small-diameter nozzle into a vacuum. Since the 1980s this
technique has been coupled with a large range of spectro-
scopic techniques and the review by Quack12 discusses the
theory and applications. By entraining the sample gas in
an expansion with a monatomic gas, temperatures down 0.025
to 1 K have been achieved. For spectral simplification this
(a) 2270 2230 2190 2160
is too cold and simple expansion of undiluted gas gives
rotational temperatures of 1080 K depending on the size
of the molecule. The vibrational temperatures achieved are
Absorbance
typically one to two orders of magnitude greater and are

mode dependent. The instrumentation coupled with FT-IR
is wide ranging and a review by Herman et al.13 details the
progress in this area and describes fully the methodology.
Most workers in this area arrange for the free jet super- (b)
sonic expansion to intersect with the focused beam of
Absorbance
the FT-IR spectrometer. This gives a very short path

length but for systems with large pumping capacities the
S/N is adequate. Figure 7 shows the system developed by
McNaughton et al.14 based around a relatively cheap cryo-
pump and combining a multipass operation with a nearly
parallel beam of IR radiation. For a full description of this 2240 2230 2220 2210 2200
system, see McNaughton et al.14 (c) Wavenumber/cm1
Figure 8 shows typical spectra for a small molecule
Figure 8. Jet nozzle spectrum of N2 O: (a) gas cell, 298 K; (b) jet
(N2 O) and comparison of Figure 8(a) for N2 O gas at nozzle, Trot 20 K, Tvib 204 K; (c) heated jet nozzle, Trot
room temperature with the jet nozzle expansion spectra 24 K, Tvib 490 K.
in Figure 8(b) shows considerable rotational cooling to pulsed-jet system combined with a multipass system based
around 20 K and enough vibrational cooling (Tvib 204 K) on a Herriot cell to investigate radicals. The Herriot cell
to remove the hot band. Figure 8(c) shows the jet nozzle arrangement allows for the intersection of the supersonic jet
spectra with a heated nozzle (200 C) where the hot band with a continuously focused beam of IR radiation. TDLS is
due to the bending mode again becomes apparent but with inherently much more sensitive than FT-IR, the IR beam can
a similar rotational temperature to that in Figure 8(b). be more easily manipulated and a much greater throughput
Figure 9 shows the results of a study on Freon 12 of sample gas can be handled because the pulse frequency of
(CCl2 F2 , 75-71-8), where the cooling out of hot bands and the pulsed beam can be matched to the modulation frequency
reduction in the rotational envelope is immediately apparent. of the system. These three advantages allow for TDLS to be
The hot bands apparent in Figure 9(a) have disappeared adapted to a much wider range of problems and it can be used
entirely in the jet-cooled spectra in Figure 9(b) and the to observe clusters and van der Waals complexes where it is
rotational envelope is smaller and individual rotational lines necessary to entrain samples in a monatomic gas to achieve
are resolved. This simplification allows the assignment of much lower temperatures in the gas expansion.16 Luckhaus
what is at first examination an intractable spectrum and after et al.17 employed a pulsed-jet system with a continuously
analysis a spectral prediction using the derived rotational scanning FT-IR spectrometer in order to attain higher sample
constants and a rotational temperature of 40 K generates throughput. They found that the spectral artifacts generated
the simulated spectrum in Figure 9(c) that is a good match in such a system cancel rapidly after averaging over sev-
with that observed. eral scans and this opens up a greater range of expansion
Diode laser spectroscopy has also made use of these super- conditions for FT techniques. Sideband lasers,18 color cen-
sonic expansion techniques and Davies15 has developed a ter lasers19 and difference frequency lasers20 have also been
coupled successfully to molecular beam expansions and pro-
vide a powerful way of obtaining highly resolved spectra
of a range of molecules, molecular ions and van der Waals
Transmittance
clusters.
To obtain spectral simplification, supersonic expansions
do not allow for easy manipulation of temperatures and also
provide non-Boltzmann thermal distributions. An alterna-
tive method for cooling samples whilst retaining them in
the gas phase is the collisional cooling or enclosive cool-
(a) 1140 1150 1160 1170 1180 ing technique. The technique requires that the condensable
sample gas be introduced into a higher pressure of non-
condensable gas in a cooled cell. The sample gas is then
Transmittance
rapidly cooled to the temperature of the system through

molecular collisions and remains in the gaseous state until
a collision with the cell wall. Newnam et al.21 successfully
coupled a collisional cooling cell to an FT-IR spectrometer
and reported sample temperatures down to 77 K. Baurecker
(b) et al.22 coupled such cells to TDLS and achieved simi-
lar temperatures. With such systems a Boltzmann thermal
population is achieved and the temperature can be easily
Transmittance
changed. Systems capable of approaching temperatures of

4 K are under development.
4.3 Computer manipulation

1150 1155 1160 1165 1170
Typically a high-resolution spectrum of a single vibrational
(c) Wavenumber/cm1
band contains hundreds and often thousands of individ-
Figure 9. n8 band of Freon 12 (CCl2 F2 ): (a) 298 K; (b) jet cooled ual vibrationrotation lines. This provides a considerable
at 0.0034 cm1 resolution; (c) simulated band profile at 40 K.
[Reproduced by permission of The Royal Society of Chem- challenge in data handling. First, the spectrum must be
istry from D. McNaughton, D. McGilvery and E.G. Robertson J. calibrated using the accurately known lines of a calibra-
Chem. Soc., Faraday Trans., 90, 10551071 (1994).] tion gas. Even for FT-IR spectroscopy this is necessary in
0.0761
0.0374
0.0012
1270 1260 1250 1240 1230 1220 1210
Wavenumber/cm1
Figure 10. High-resolution spectrum of the n6 C n8 region of diacetylene obtained by a sample of 12 CH4 13 CH4 (4 : 1) sample mixture.
[Reproduced by permission of Plenum Press from R. Tay, G. Metha, F. Shanks and D. McNaughton, Struct. Chem., 6, 4755 (1995).]
The LoomisWood approach essentially replots the spec-

trum using a list of peak centers and intensities and is based
on treating every molecule as a linear molecule. Thus every
series in a spectral band can be predicted using a least-
squares fit to a polynomial expansion where the original
coefficients of the polynomial are educated guesses made by
B A C the spectroscopist. In this replotted spectrum, related lines
appear as vertical or near vertical series down a computer
screen which can be quickly marked and assigned with
quantum numbers and refitted to the polynomial expan-
sion. Figure 10 shows a spectrum of a sample containing
a number of isotopomers of diacetylene (mainly all 12 C
and 13 C singly substituted), which is a complicated set of
Figure 11. LoomisWood plot (using MacLoomis) of the dia- lines where apart from lines emanating from the ground
cetylene spectrum in Figure 10. The band substructure now state of the major isotopomer it is difficult to pick out the
appears vertically down the computer screen. [Reproduced by per- patterns. After processing in the manner above, the spec-
mission of Plenum Press from R. Tay, G. Metha, F. Shanks and
D. McNaughton, Struct. Chem., 6, 4755 (1995).]
trum is replotted in LoomisWood format in Figure 11 and
assignment25 becomes a pattern recognition task provided
order to achieve precise line positions because of variation that the spectroscopist has a good grasp of the necessary
in ILSs. The calibration laser frequency is usually adjusted theory.
within the software in order to reproduce the known line
positions of the calibration gas prior to running peak pick
routines that automatically produce a list of line center 5 SPECIALIST TECHNIQUES AND
wavenumber values, line widths and line intensities. For INSTRUMENTS
laser techniques the calibration process is usually more
tedious. Assignment of the spectrum is then usually car- Many of the laser techniques mentioned above can be
ried out using computer-assisted methods that permit quick described as specialist in that they require considerable
and reliable assignment. In modern high-resolution gas- instrumental expertise to be successful. They do, how-
phase IR spectroscopy these techniques are becoming an ever, provide ways of narrowing lines below the Doppler
integral part of the necessary instrumentation. The most limit and obtaining spectra of species that are normally not
common technique is the LoomisWood technique and obtainable with straightforward techniques, and the follow-
both Macintosh-23 and PC-24 based programs are available. ing discusses some useful aspects of these techniques.
5.1 Sub-Doppler spectroscopy at specific detection frequencies exist in Europe29,30 and

the USA.31,32 Rather than using the sun as a source of
Molecular beams such as those discussed above provide radiation, a number of the balloon-borne instruments use
a means of reducing Doppler broadening by ensuring that long-path absorption by hanging a reflector optic up to 1 km
the molecules are all travelling in the same direction with below the balloon gondola. Instruments purpose designed
the same velocity. If the probe radiation is perpendicu- for exploring planetary atmospheres are also available or
lar to the beam then line-width reduction is achieved. For under construction.32
supersonic expansions as described above there is still an
angular divergence of the molecules and to achieve a line-
width reduction it is necessary to select out the central ABBREVIATIONS AND ACRONYMS
collinear beam by skimming the beam. This requires a
large pumping capacity and considerably reduces the avail- LMR Laser Magnetic Resonance
able sample, making laser spectroscopy the only viable NLS Natural Line Shape
technique. McIlroy et al.26 and Stone et al.18 described sub- OPD Optical Path Difference
Doppler spectroscopy using this technique coupled with OPO Optical Parametric Oscillator
sideband and color center lasers, respectively. An alter- TDLS Tunable Diode Laser Spectroscopy
native way of obtaining sub-Doppler resolution is Lamb
dip spectroscopy or spectral hole burning, described by REFERENCES
Hollas,27 where transitions are observed only in molecules
having a zero velocity component in the direction of the 1. L.S. Rothman, R.R. Gamache, A. Goldman, L.R. Brown, R.A.
probe radiation. This method allows for spectral line widths Toth, H.M. Pickett, R.L. Poynter, J.-M. Flavd, C. Camry-
to be the same as that of the probe radiation, provided that Peyret, A. Barbe, M. Husson, C.P. Rinsland and M.A.H.
the line width is limited only by the Doppler effect, the Smith, Appl. Opt., 26, 4058 (1987).
source power is strong enough to saturate the transition 2. W. Herres and J. Gronholtz, Understanding FTIR Data Pro-
and the radiation is tunable over a very small range. Lamb cessing, Parts 1 and 2, Bruker Analytische Messtechnik,
dip spectroscopy is therefore limited to a few of the laser Karlsruhe (1991).
techniques described above. 3. J.M. Hollas, High Resolution Spectroscopy, John Wiley &
Sons, Chichester, 522548 (1998).
4. G.W. Baxter, H.D. Barth and B.J. Orr, Appl. Phys. B: Lasers
5.2 Field instruments Opt., 66, 653 (1998).
5. G.W. Baxter, Y. He and B.J. Orr, Appl. Phys. B: Lasers Opt.,
High-resolution spectroscopy of gases is of considerable 67, 753 (1998).
importance in atmospheric science for baseline monitor- 6. A. Bohren and M.W. Sigrist, Infrared Phys. Technol., 38, 423
ing and pollution monitoring. For field work the instrument (1997).
design has to be different to that of a conventional lab- 7. J. Raffy, T. Debuisschet, J.P. Pocholle and M. Papuchon,
oratory instrument and both Bruker (120 M, 0.004 cm1 ) Appl. Opt., 33, 985 (1994).
and Bomem (DA8, 0.004 cm1 ) supply modular and rugged 8. D. McNaughton, High Resolution FTIR Spectroscopy of
high-resolution FT-IR instruments capable of operation in Short Lived Molecules, in Fourier Transform Spectro-
balloons or aircraft. These instruments and more conven- scopy, ed. J.A. de Haseth, AIP Conference Proceedings
tional instruments that use solar radiation as a source to 430, American Institute of Physics, Woodbury, NY, 110127
monitor atmospheric species require collecting optics rather (1998).
than a conventional source. The US National Solar Obser- 9. J.M. Hollas, High Resolution Spectroscopy, John Wiley &
vatory has available a 1-m pathlength FT-IR instrument Sons, Chichester, 7576 (1998).
housed totally within an evacuated chamber28 for solar 10. J.D. McManus, I.L. Kababian and M.S. Zahniser, Appl. Opt.,
source work. This instrument achieves 0.005 cm1 reso- 34, 3336, (1995).
lution in the IR region by a folding of the beam path to 11. J. Ballard, K. Strong, J.J. Remedios, M. Page and W.B. John-
double the OPD. ston, J. Quant. Spectrosc. Radiat. Transfer, 52, 677 (1994).
Diode laser instruments are also extensively used in 12. M. Quack, Annu. Rev. Phys. Chem., 41, 839 (1990).
atmospheric work and are more appropriate for airborne 13. M. Herman, R. Georges, M. Hepp and D. Hurtmans, Int. Rev.
and balloon work because the instruments can be made Phys. Chem., 19, 277 (2000).
more compact and lighter than FT-IR instruments, where 14. D. McNaughton, D. McGilvery and E.G. Robertson, J. Chem.
long mirror travels are required. A number of lightweight Soc., Faraday Trans., 90, 1055 (1994).
instruments based around a number of diodes operating 15. P.B. Davies, Spectrochim. Acta, Part A, 55, 1987 (1999).
16. G. Winnewisser, T. Drascher, T. Giesen, I. Pak, F. Schmul- 27. J.M. Hollas, High Resolution Spectroscopy, John Wiley &
ling and R. Schieder, Spectrochim. Acta, Part A, 55, 2121 Sons, Chichester, 5455 (1998).
(1999).
28. US National Solar Observatory, Kitt Peak National Observa-
17. D. Luckhaus, M. Quack, U. Schmitt and M.A Suhm, Ber. tory, Tucson, AZ, http://www.noao.edu/noao/staff/plymate/fts/
Bunsen-Ges. Phys. Chem., 99, 457 (1995). aboutFTS.html.
18. S.C. Stone, L.A. Phillips, G.T. Fraser, F.J. Lovas, L.H. Xu 29. F.G. Wienhold, H. Fischer, P. Hoor, V. Wagner, R. Konig-
and S.W. Sharpe, J. Mol. Spectrosc., 192, 75 (1998). stedt, G.W. Harris, J. Anders, R. Grisar, M. Knothe, W.J.
19. D.G. Lancaster, D. Richter, R.F. Curl and F.K. Tittel, Appl. Riedel, F.-J. Lubken and T. Schilling, Appl. Phys. B, 67, 411
Phys. B, 67, 339 (1998). (1998).
20. M. Ishiguro, T. Imajo, K. Harada, M. Matsubar, K. Tanaka 30. H. Fischer, J. Bonifer, J.P. Burrows, D. Klemp, U. Par-
and T. Tanaka, Chem. Phys. Lett., 263, 629 (1996). chatka, J. Roths, C. Schiller, T. Zenker, R. Zitzelsberger
21. D. Newnam, J. Ballard and M. Page, Rev. Sci. Instrum., 66, and G.W. Harris, Development and Application of Multi-
4475 (1995). laser TDLAS Instruments for Ground-based, Shipboard and
Airborne Measurements of Trace Gases in the Troposphere
22. S. Baurecker, F. Tacher, C. Weitkamp, W. Michaelis and in Instrument Development for Atmospheric Research and
H.K. Cammenga, J. Mol. Struct., 348, 237 (1995). Monitoring, Transport and Chemical Transformation of Pol-
23. B.P. Winnewisser, J. Reinstadtler, K.M.T. Yamda and J. Beh- lutants in the Troposphere, eds J. Bosenberg, D. Brassington
rend, J. Mol. Spectrosc., 136, 12 (1989). and P.C. Simon, Springer, Berlin, Vol. 8 (1997).
24. D. McNaughton, D. McGilvery and F.S. Shanks, J. Mol. 31. Atmospheric Technology Division (ATD) of the National
Spectrosc., 149, 458 (1991). Center for Atmospheric Research (NCAR), Boulder, CO,
25. R. Tay, G. Metha, F. Shanks and D. McNaughton, Struct. http://www.atd.ucar.edu/index.html.
Chem., 6, 4755 (1995). 32. Atmospheric Laser Spectroscopy Group of Jet Propulsion
26. A. McIlroy, D.J. Nesbitt, E.R.T. Kerstel, B.H. Pate, K.K. Laboratory, Pasadena, CA, http://laserweb.jpl.nasa.gov/
Lehmann and G. Scholes, J. Chem. Phys., 100, 2596 (1994). instrumentation.htm.
Raman Spectroscopy of Gases
Alfons Weber
National Institute of Standards and Technology, Gaithersburg, MD, USA and National Science Foundation,
Arlington, VA, USA
1 INTRODUCTION in Raman spectroscopy there have been many books and

review articles dealing with different aspects of this topic.
Since the discovery of the Raman effect, the study of In particular, a summarized version of Placzeks article1 in
the properties of gaseous molecules by means of Raman contemporary notation and in SI units is now available.4,5
spectroscopy has been characterized by feverish activity Raman spectroscopy of gases is principally concerned with
alternating with periods of soporific repose. For a few the study of molecular structure and as such is comple-
years after the discovery of the effect, the Raman spectra mentary to infrared (IR) absorption spectroscopy. Because
of many diatomic, linear, and symmetric top polyatomic of different selection rules that apply to Raman scattering
molecules were investigated. These studies laid the ground and IR absorption, an integration of Raman spectroscopic
work for the development and confirmation of the theory data with those obtained by IR spectroscopy is desirable
of the Raman effect in molecules. After the publication of but requires special care, for the quality of these data are
Placzeks treatise1,2 and Placzek and Tellers article3 on the not always equivalent in their precision and accuracy.
rotational structure of Raman bands, very little further work, This article presents a broad overview of accom-
either experimental or theoretical, was done on this subject. plishments in spontaneous and nonlinear Raman spec-
In the early 1950s new developments in instrumentation, troscopy and concludes with a brief discussion of several
such as the water-cooled mercury arc lamp, the multiple applications.
traversal scattering cell, and the introduction of photoelec-
tric recording spectrometers invigorated the field into a new
spurt of activity. Many additional molecules in the gaseous 2 SPONTANEOUS RAMAN
state were now accessible to the study of their pure rota- SPECTROSCOPY
tional and rotationvibrational Raman spectra and work in
this field continued until the limits imposed by the width of 2.1 High-resolution rotationvibration spectra
the mercury exciting line resulted in only marginal advances
of providing useful Raman spectroscopic data. The inven- Until the late 1970s only fixed-wavelength/frequency
tion of the laser, especially the argon ion laser, as well as sources, be they classical mercury arc lamps or lasers,
major developments in instrumentation, the most recent of were available for the study of Raman spectra. The spec-
which is the charge-coupled device (CCD) detector (for the tra were due to spontaneous or incoherent scattering, the
various aspects and developments of instrumentation see scattering intensity being proportional, among other fac-
Part 4 of this volume), and the subsequent developments tors, to the intensity of the exciting light. For low gas
of nonlinear Raman spectroscopic techniques, prevented pressures the resulting spectra were limited in resolution
the total decline of Raman spectroscopy of gases into by the width of the exciting line or, if narrow-line laser
an historical curiosity. Responding to the revived interest sources were used, by Doppler and pressure broadening of
the transitions. Even with the largest grating spectrographs
John Wiley & Sons Ltd, 2002. available, the best resolution that was attained was never
better, i.e. smaller, than ca. 0.05 cm1 . Such resolution each degenerate vibrational mode with quantum number vt ,
is inadequate when compared with the resolution attained jlt j D vt , vt1 , vt2 , . . . 1 or 0 where t is the mode label.
by high-resolution IR spectroscopy (see High-resolution For linear molecules the general rotational selection rules
Fourier Transform Spectrometry of Gases and Instru- for Raman scattering are
mentation and Methods for High-resolution Gas-phase
J D 0, 2 if l D 0 2
Spectroscopy). In spite of these shortcomings, however, a
large amount of useful data was obtained with spontaneous J D 0, 1, 2 if l 6D 0 3
Raman spectroscopy. These accomplishments have been
described in many review articles and books.6 9 We men- and
tion just four specific achievements with techniques dating /
C ! s !
/ a 4
from the pre-laser era (spectral resolution 0.4 cm1 ). The that is, positive rotational levels do not combine with neg-
structure of the ethane molecule (staggered or eclipsed) ative ones and symmetric rotational levels do not combine
has been a problem since the early 1930s. The D3d (stag- with antisymmetric ones. J D J0 J00 where the single
gered) ground state structure of C2 H6 was unambiguously prime designates the upper and the double prime the lower
established by the observation of R- and S-branch transi- of the two combining states. In the expressions for the
tions in the rotationvibrational Raman bands of gaseous wavenumbers of the rotational lines of a rotationvibration
ethane;10 13 the former discrepancy between the precision band the prime symbols are omitted and the J-symbol is
X-ray determined CC bond length for crystalline ben- always that of the lower state. The rotationvibrational
zene and that determined by pure rotational Raman spec- Raman shifts nQ for linear molecules are represented by
troscopy of the vapor was resolved in favor of the Raman the equation
result;14 17 pressure broadening coefficients for the rota- 0
tional Raman lines and vibrational frequency perturbations nQ D nQ 0 C B0 [J0 J0 C 1 l02 ] DJ [J0 J0 C 1 l02 ]2
of several diatomic and linear molecules were accurately 00
B00 [J00 J00 C 1 l002 ] C DJ [J00 J00 C 1 l002 ]2
determined;18 20 and a large body of data on the Raman
scattering cross-sections for the vibrational modes of many 5
polyatomic molecules was determined.21 These were fur- where nQ 0 is the wavenumber for the pure vibrational tran-
ther developed and extended with argon laser exitation.22 sition. For nondegenerate fundamental bands, representing
Raman spectra have conventionally been obtained in the stretching vibrations, l0 D l00 D 0.
90 scattering configuration (see Part 4 of this volume for For degenerate fundamental bands, representing bending
discussions of experimental techniques). The spectrometer vibrations, l0 D 1, l00 D 0 and each upper-state rotational
was generally a prism or grating spectrograph, or record- level is split into a doublet (l-type doubling) on account of
ing spectrometer. Examples of such spectra are shown the rotationvibration interaction. The members of these
in Figures 1 and 2, which show the pure rotational and doublets are labelled by the letters e and f (in the older
the n1 rotationvibration spectra of the linear molecule literature these labels were c and d). The magnitude of
cyanogen, C2 N2 . A linear molecule composed of N atoms the level splitting is given by nQ D qJJ C 1 , where q
has 3N-5 fundamental normal modes of vibration. These is the l-type doubling constant, which is of the order
are grouped into those where the atoms move along the 104 105 cm1 . The constants B and DJ are different for
internuclear axis (parallel vibrations) and those in which these two sets of levels.
the atomic displacements are perpendicular to that axis For pure rotational spectra in the ground vibrational state
(perpendicular vibrations). Parallel vibrations are nonde- the Raman displacements (shifts) form an S-branch whose
generate, while perpendicular vibrations are degenerate (for wavenumbers are obtained from equation (5) by setting
details about vibration and rotation vibration spectra see the nQ 0 D 0, J0 D J00 C 2, l0 D l00 D 0, B0 D B00 , and DJ0 D DJ00 .
literature).23,24 The rotational term values FJ , neglecting These (Stokes or anti-Stokes) Raman shifts are represented
higher power terms in JJ C 1 , are given by by the equation

3 3 3
FJ D B[JJ C 1 l2 ] DJ [JJ C 1 l2 ]2 1 nQ D 4B 6DJ J C 8DJ J C 6
2 2
Here B is the rotation constant and DJ the centrifugal dis- For pure rotational Raman transitions within a degenerate
tortion constant. The quantity J is the quantum number vibrational state (pure rotational hot bands) we have l0 D
for the total angular momentum, exclusive of nuclear spin, l00 D 1. The selection rules J D C1 and J D C2 give,
while l is the quantum number of the vibrational angular respectively, the equations for the R- and S-branch lines
momentum. For parallel bands l D 0 while for perpendic-
ular bands l 6D 0. l is a signed quantum number and for nQ R D 2BJ C 1 4DJ JJ C 1 J C 2 7
Raman Spectroscopy of Gases 3
Anti-Stokes A4880 Stokes

60 55 50 45 40 35 30 25 20 15 10 5 S(J)
5 10 15 20 25 30 35 40 45 50 55 60 65
40 30 20 10 0 10 20 30 40
(a) Raman displacement /cm1
60 62
64 66
63 68
74
65 70
74 d
72
67 75 c 15
75 d
69
76
71
73 78
75
80
77
79
25
J= 60 65 70
35 40 45 50
(b) Raman displacement /cm1
Figure 1. Photographically recorded pure rotation spectrum of cyanogen, C2 N2 . (a) Gas pressure 1.33 104 Pa (100 torr), exposure
time 10 min. (b) Stokes side of the high-J region of the spectrum recorded in 6.5 h at a gas pressure of 6.65 103 Pa (50 torr). The
resolved l-type doublets in the S-brach of the n5 , as well as the unresolved S-branch lines of the 2n5 excited states, are identified.
(Reproduced by permission from I-Yan Wang and A. Weber, J. Chem Phys. 67, 3084 (1977).)
and O- and S-branches (J D 2 and J D C2, respectively);

3 3 3 0 9
nQ S D 4B J C 8DJ J C 1 J C 2 J C 8 nQ OJ ,SJ D nQ 0 C B B 00
DJ0 DJ00
2 2 4 16
where the constants B and DJ are those appropriate for the
3
split rotational levels (e or f) of the degenerate vibrational C B0 C B00 DJ0 C DJ00 m
2
state.
The equations for the Raman shifts for the lines in a 1 11
C 0 00
B B DJ DJ m2
0 00
fundamental rotationvibration band are obtained by the 4 2

same procedure. Thus, for a parallel band: 1 1
Q-branch, J D 0; C DJ0 C DJ00 m3 DJ0 DJ00 m4
2 16
10
nQ QJ D nQ 0 C B0 B00 JJ C 1
where m D 2J C 1 for the O-branch and m D 2J C 3 for
DJ0 DJ00 [JJ C 1 ]2 9 the S-branch, and where nQ 0 is the wavenumber for the
(1 + 25 25)
(1 + 15 15)
(1 + 14 14)
(1 + 315 315) (1)
O(J)1 = 70 60 50 40 30 20 10 20 30 40 50 60 70 = S(J)1
O(J)( +11 ) = 40 30 20 10 20 30 =S(J)( +11 )
1 5 5 1 5 5
2280 2290 2300 2310 2320 2330 2340 2350 2360 2370 (cm1)
Figure 2. The n1 parallel fundamental band of cyanogen photographically recorded in 15.3 h at a gas pressure of 2.67 104 Pa (200 torr).
The Q-branches of four hot bands are identified. The line labeled G is a grating ghost. (Reproduced by permission from I-Yan Wang
and A. Weber, J. Chem Phys., 67, 3084 (1977).)
pure vibrational transition. Each of the molecular con- (see below); and (4) very long recording times (photograph-
stants, B and DJ , are specific functions of the vibra- ically or photoelectrically), well in excess of 100 h, are
tional state and are in general so labeled by an additional required to obtain high-resolution vibrationrotation spec-
subscript showing the value of the vibrational quantum tra. To obtain high-resolution Raman spectroscopic data of
number v. a quality comparable to that obtained routinely for IR spec-
The spectra shown in Figure 1 are the pure rotation spectra with Fourier transform (FT) spectrometers, recourse is
tra of cyanogen. In Figure 1(a) the S-branch appears as a now had to nonlinear Raman spectroscopic techniques (see
series of sharp lines with apparently no structure, situated Section 3).
on a fairly flat background. Figure 1(b) shows the details
of the high-J region after a longer exposure. Here the pure
rotational hot band of molecules in the degenerate n5 state 2.2 Use of interferometers
are clearly visible as resolved l-type doublets. Also seen are
the weaker signatures of the pure rotational spectrum in the Various attempts were made to overcome, or at least min-
degenerate overtone state 2n5 (here their l-type doubling is imize, some of the above limitations by using interfero-
not resolved). Figure 2 shows the parallel n1 fundamental metric methods to record Raman spectra. These have been
band. It is severely overlapped on the low-wavenumber side summarized and reviewed in Weber.8 The two-beam
of its Q-branch by the features of the vibrationrotational Michelson-type FT interferometric spectrometer holds the
hot bands n1 C n5 n5 , n1 C 2n5 2n5 , n1 C n4 n4 , and greatest promise in this regard (for FT techniques see also
n1 C 3n5 3n5 . (The symbols for degenerate vibrational Part 4 of this volume and Part 1 of Volume 2). Initial exper-
states, here n5 , 2n5 , 3n5 , and n4 , are also specified by the iments in which a Raman excitation unit designed for use
appropriate l-value, which appears as a superscript. These with a slit spectrograph was coupled to the FT instru-
are omitted here for brevity.) ment located at the National Solar Observatory on Kitt
These examples demonstrate the limitations of spon- Peak mountain in Arizona were promising and demon-
taneous Raman spectroscopy by conventional methods. strated, contrary to predictions, that the FT spectrometer
The most important factors determining these limitations operating in the visible region of the spectrum is indeed a
are: (1) the observed pure rotational spectra of polyatomic very suitable instrument for high-resolution Raman spec-
molecules are invariably unresolved superpositions of spectroscopy of gases. Even though there was an unfavorable
tra from the ground as well as the lower-lying vibrationally mismatch between the optical systems of the Raman source
excited states; (2) the spectra of symmetric top molecules and the FT instrument, and the experiments were performed
are resolved only into their J-structure; the K-splittings of with low laser power in an uncontrolled laboratory envi-
the rotational lines due to the DJK term are not resolved (the ronment that did not permit running the argon laser in
exception is the case of ammonia);25 (3) the Raman lines single mode, this work demonstrated the superiority of
are significantly broadened by Doppler and pressure effects high-resolution FT spectrometers versus the best grating
spectrographs. A large body of data on the pure rotation

and rotationvibration spectra of H2 and D2 ,26,27 as well 80
as of N2 , O2 , CO2 , C2 H2 , and C2 N2 ,8 was obtained. The

70
precision of the Raman data for hydrogen was very nearly 60
that of those determined by Fourier transform infrared (FT-
50
IR) quadrupole absorption spectroscopy but the Raman line
wavenumbers showed an offset by about 0.01 cm1 from 40
the IR wavenumbers. In these experiments the 90 scatter- 30
ing geometry was used. The Raman cell was inside the laser
20
cavity, the laser was run at ca. 2 W, with a multimode line
width of ca. 0.16 cm1 and commensurate gas pressures 10
were between 33 and 80 kPa (250 and 600 torr). Encourag- 0
2326.5
2327.0
2327.5
2328.0
2328.5
2329.0
2329.5
2330.0
2330.5
ing as these results were, they in turn showed that, in order
to fully exploit the capabilities of the FT technique, it is
necessary to employ long-term stabilized single-mode lasers Wavenumber /cm1
together with the forward-scattering geometry to reduce the
Figure 3. Part of the fully resolved Q-branch of 14 N2 recorded
Doppler width of the Raman lines. Moreover, perfect align-
with a FT spectrometer. (Reproduced by permission from Bendt-
ment between the optical systems of the Raman source unit sen and Rasmussen.29 )
and the FT instrument is required to ensure accurate deter-
mination of the Raman wavenumbers (e.g. avoidance of
the above-mentioned offset). So far only existing general- 7
purpose FT instruments have been employed (see their
6
descriptions in Weber8 ) but an FT instrument specifically

designed for Raman spectroscopy of gases has been con- 5
structed at the University of Aarhus and very encouraging
initial results have already been published.28 30 4 0.018 cm1
Figures 3 and 4 show, respectively, the fully resolved 3
Q-branch and the S(6) line of the fundamental rota-
tionvibration band of 14 N2 . The complete fundamental 2
band, including the O-, Q-, and S-branches, was recorded
1
several times at a resolution of 0.015 cm1 and a gas pres-
sure of 4.04.7 kPa (3035 torr). The recording time for 0
each interferogram was 100 h. A separate Raman spec- 2387.8 2388.0 2388.2 2388.4 2388.6 2388.8 2389.0
trum of the fundamental band of CO revealed that the Wavenumber /cm1
measured wavenumbers of its Q-branch lines were sev- Figure 4. The S(6) line of the rotationvibrational Raman spec-
eral thousandths of a reciprocal centimeter smaller than trum of 14 N2 . (Reproduced by permission from Bendtsen and
those calculated from the parameters obtained from high- Rasmussen.30 )
resolution IR spectroscopy.31 This difference was used as
a calibration correction to the Raman data. Table 1 com- obtained by other workers using different techniques.32 34
pares the molecular constants obtained from fits of the These results should be of comfort to those who feel
uncalibrated and the calibrated Raman data with the results discouraged by the statement, No one in their right mind
14
Table 1. Molecular constants for N2 (in cm1 ).
Reference n0 B0 B 102 D0 106 D 109
32 2329.9116 (2) 1.73697 (10) 10.4 (5)

33 2329.9124 (7) 1.989622 (10) 1.73777 (54) 5.763 (40)
34 2329.9168 (3) 1.989574 (54) 1.7384 (3) 5.76 (3)
This work, uncalibrated 2329.90926 (6) 1.9895924 (30) 1.736970 (65) 5.7606 (76) 11.1 (1)
This work, calibrated and including 2329.91173 (6) 1.9895924 (38) 1.736973 (36) 5.761 (11) 10.43 (20)
wavenumbers of Maki and Wells31
Data taken from Bendtsen and Rasmussen18 (2000).

would voluntarily record Raman spectra of gases on a FT

machine,35 res ipsa loquitur. 60
0 = 5145.308
1
0 = 19429.77 cm1

Relative line width R /IR
cm
50
0
50
2.3 Raman line widths
40
1
Line widths are usually given as the full width at half- 0 cm
100
30
maximum (FWHM) of the line intensity. Pressure broad-
ening of rotational Raman lines is a function of the lower 20 1
rotational state. For diatomic molecules (N2 , O2 , CO) they 2000 cm
1
decrease from 0.15 105 cm1 Pa1 (0.15 cm1 atm1 ) 10 3000 cm
to nearly 0.05 105 cm1 Pa1 (0.05 cm1 atm1 ) with
increasing quantum number J, while for CO2 the change 0
0 20 40 60 80 100 120 140 160 180
is from nearly 0.3 105 cm1 Pa1 (0.3 cm1 atm1 ) to
Scattering angle ()
0.22 105 cm1 Pa1 (0.22 cm1 atm1 ). The values for
rotationvibration lines are less well established but are Figure 5. Relative Doppler linewidths for Stokes Raman and IR
transitions as function of the scattering angle for the argon laser
expected to be not too different from these. A working
l0 D 514.5 nm exciting radiation. (Reproduced by permission
gas pressure of ca. 1.33 103 Pa (10 torr) for nonpolar from Weber.8 )
molecules would therefore reduce the effect of pressure
broadening to a realm that permits the combined use of Raman and IR transitions are thus seen to be equal only
equally weighted high-resolution Raman data and IR data in for q D 0 , i.e. for forward scattering. The use of the
a comprehensive band analysis of vibrationrotation spec- conventional 90 scattering geometry will always result
tra. More difficult, however, is the minimization of Doppler in Raman lines that are significantly broader than the
broadening. The Doppler width (FWHM) of a Raman line corresponding IR lines. When Raman data obtained with
is a function not only of the frequency of the exciting scattering geometries for which q > 0 are to be included
radiation and the molecular rotation or rotationvibrational in a comprehensive determination of molecular parameters,
transition, but also of the scattering geometry. This Doppler they should be weighted, for equal signal to noise ratio, by
Doppler
width, nQ Raman , is given by6,36 at least the inverse square of their relative Doppler widths.

Doppler 2p T This weight would have to be augmented by additional
nQ Raman D 2R ln 2 factors that account for other causes of line broadening (e.g.
c M
pressure effects, line overlaps, broadening due to nonlinear
q effects, etc.).
4nQ 20 C nQ 0 nQ R sin2 C nQ R 2 11
2
where R is the gas constant, T the absolute temperature, M
the molecular weight, nQ 0 the wavenumber of the exciting 2.4 Intensity alternation
line, nQ R the wavenumber of the Raman shift (positive
for anti-Stokes and negative for Stokes lines), and q the In Figures 1, 2 and 3 an alternation in intensity is seen
scattering angle, measured from the direction of the incident between neighboring lines. This effect is characteristic of all
(laser) beam. When q D 0o , this expression reduces to molecules that have equivalent nuclei and is determined by
the nuclear spin statistical weights of the rotational levels.
Doppler Doppler 2p T Such weights have been determined for several diatomic
nQ Raman D nQ IR D 2R ln 2 nQ IR 12
c M and simple polyatomic molecules and can be found in many
since jnQ R j nQ IR , which is the IR molecular rotation or books on molecular spectroscopy (see Herzberg23 ). General
rotationvibration wavenumber. The ratio of the Raman to but simple-to-use formulas have also been derived that
IR Doppler line widths is then given by allow the computation of nuclear spin statistical weights

for any symmetric top molecule.37
2
Doppler
nQ Raman nQ 0 nQ 0 q

D 1C4 C sin2 13
Doppler nQ R nQ R 2
nQ IR
2.5 Supersonic jet expansions
This relationship is shown in Figure 5 as a function of the
scattering angle for Stokes Raman shifts of 500, 1000, Supersonic jet expansions of gases are used for sev-
2000, and 3000 cm1 , respectively, excited by the l D eral purposes: (1) to suppress the population of higher-
514.5 nm argon laser line. The Doppler widths of the lying rotational and rotationvibrational levels by adiabatic
cooling of a gas to temperatures well below its equilibrium For the determination of anharmonic coefficients of the
thermodynamic condensation point and thereby simplify the vibrational potential function it is desirable to observe
observed spectrum for measurement and analysis; (2) to not only overtones but also vibrational hot bands. The
produce a high density of weakly bound (hydrogen or van very crowded and overlapping vibrational bands observed
der Waals bonded) molecular complexes for spectroscopic with Raman and IR spectroscopy of stagnant gases at
study; and (3) to afford a method for the study of supersonic ambient temperatures prevent accurate determinations of
gas dynamics with high spatial resolution. Raman spec- the wavenumbers of hot bands54 56 (Figure 2). Supersonic
troscopy with supersonic gas expansions was first reported expansions are very efficient in lowering the rotational tem-
in 1976.38 Here the focus was on a direct determination of peratures of expanding gases into the range of 20 K or less.
the local dynamical properties of an expanding jet beam by Vibrational cooling is, however, very inefficient. The result
means of Raman spectroscopy. The pure rotational Raman is that transitions from excited vibrational states persist
spectrum of expanding CO2 gas was recorded and the rota- even in supersonically cooled gases but are nearly devoid
tional temperature and absolute density of CO2 monomers of any rotational structure. Table-top Raman spectrometers
were determined as functions of axial distance from the noz- can therefore be used to good advantage in the accurate
zle. Further studies showed the advantage of beam cooling determination of vibrational wavenumbers. An example
through the observation of pure rotational Raman scattering is the Raman spectrum of the totally symmetric n1 band
of CO2 in the vibrational ground (000 0) and lowest excited of CCl4 shown in Figure 6.57 Even though the rotational
(011 0) states,39 as well as the first experimental observa- branches that would flank the Q-branches are forbidden for
tion and study of van der Waals molecules (argon dimers) this mode by the selection rules, the rotational structure of
by Raman spectroscopy.40,41 Further work on several other
the Q-branches themselves give them sufficient width and
molecules was done by Luiks et al.42,43 While these inves-
underlying background to cause significant loss of resolu-
tigations were performed with low-resolution, table-top
tion in the room temperature spectrum (Figure 6a) and thus
spectrometers, supersonic jet sources were also used in con-
lower accuracy of the vibrational wavenumbers. In the jet-
junction with high-resolution FT spectrometers.44
cooled spectrum (Figure 6b) most of the rotational structure
of the Q-branches is absent. The Q-branches are narrow
2.6 Vibrational Raman spectra and better defined, the underlying background is absent,
Very significant information on the molecular structure and

conformation is also available from rotationally unresolved
0.60
vibrational Raman bands. This is especially true for heavy
spherical top molecules where only the rotational branch
0.40
Intensity (count s1 pixel1)
maxima nOP and nRS are resolved from the central

Q-branch.45 Information from the band contours and the
separation of the OP and RS rotational branch maxima 0.20
has yielded valuable data on the Coriolis coupling con- (a)
stants for the triply degenerate vibrational states of MF6 0.00
(M D S, Se, Te, Mo, W, or U), M(CH3 )4 (M D C, Si, C35Cl237Cl2
C35Cl337Cl
0.04 C35Cl37Cl3
Ge, Sn, or Pb), P4 , As4 , and OsO4 molecules. Asymmet- C35Cl4
ric top molecules with hindered or partially free internal
rotors present another rich field of investigation. Most of
the work on such molecules has been done on liquid-phase 0.02
systems with often ambiguous results. Since most of the
C37Cl4
modes that affect the conformation are of low energy, the
Raman spectroscopic technique has considerable advan- (b) 10
0.00
tages over IR spectroscopy. Whereas the spectra of liq- 430 440 450 460 470 480
uids are often affected by the presence of intermolecular
Wavenumber /cm1
forces, giving rise to broad overlapping bands, the spec-
tra of such molecules in the vapor or gaseous state have Figure 6. Raman spectrum of the n1 band of gas phase CCl4
recorded with a CCD detector. (a) Stagnant gas at room tempera-
sharper features, allowing vibrational assignments, deter-
ture (23 C) and at a pressure of 130 hPa (100 torr). (b) The same
mination of torsional potential functions, conformational spectrum recorded from a CCl4 N2 gas mixture cooled to 75 K
energy differences, etc. with less ambiguity and greater in a supersonic jet expansion. (Reproduced by permission from
reliability.46 53 Tejeda et al.57 )
and weak isotopic components are now well developed and Raman spectroscopy, where only one excitation line suffices
measurable (the natural abundance of the C37 Cl4 isotopic to generate a fully informative and complete Raman spec-
species is 0.019%). trum, the resonance technique employs as many UV and
VUV excitation lines as possible in order to map out the
upper electronic state by means of the Raman effect.70 74
2.7 Scattering cross-sections There are definite advantages to be had with resonance
Raman spectroscopy in the UV and VUV regions: fluores-
Raman scattering cross-sections have been reported for cence that mostly originates from the lower excited states
many molecules.21,22 Since then, additional results have is absent, the spectra are simple and highly selective, they
been obtained, e.g. for the methyl radical58 and for tri- allow the study of reaction processes in excited electronic
methylgallium and trimethylaluminum.59 There is further states, they permit the appearance of symmetry-forbidden
interest in scattering cross-sections at ultraviolet (UV) electronic transitions,75,76 they enable a quantitative assess-
wavelengths at which resonance effects enhance the inten- ment of large geometry changes, and they generate transi-
sities of Raman bands to a level useful for practical tions to highly excited vibrational states of the electronic
applications.60,61 ground state that are generally not observed with nonres-
onant Raman spectroscopy.70
2.8 Resonance Raman spectra 2.9 Raman spectroscopy under extreme

conditions
Raman spectroscopy is generally done in the visible region
of the spectrum for which most molecules have their first Though spectra are generally recorded at ambient or at
excited electronic state at energies well above the energy of low temperatures there are situations where it is advan-
the exciting photons (see Part 4 of this volume). Resonance tageous to heat the samples to high temperatures. Thus,
effects manifest themselves when the energies of the excit- hydrogen and deuterium were burned in flames in order to
ing photons approach the energy of an (usually the lowest populate higher-lying rotational levels and thereby deter-
excited) electronic state. This resonance enhances not only mine additional molecular constants.27 The pure rotational
the intensity of the Raman scattering but also reveals fea- spectrum of diatomic sulfur was investigated at elevated
tures often not observed with nonresonant excitation. These temperatures77 as was the vibrationrotation spectrum of
features generally have their origins in the antisymmetric P4 .78 (Note, however, that a recent high-resolution FT-IR
components of the Raman scattering tensor (see the treat- study reported significantly different results.79 ) An intrigu-
ment by Koningstein2 ). In some cases the resonance, or ing study of the vibrational Raman spectrum of sulfur at
pre-resonance, is already evident with photons in the vis- temperatures up to 700 C and at pressures of 0.1 and
ible region, as is the case with measured depolarization 0.2 MPa (1 atm and 2 atm, respectively) allowed the inves-
factors that differ significantly from expected values,1,62 tigation of tetratomic sulfur, S4 . Two structural isomers,
with the halogen and interhalogen compounds,63,64 or the chain-like and cyclic, were found. This was facilitated with
(NO2 )2 dimer,65 while in other cases UV photons are used the aid of resonance Raman enhancement (different laser
to map out the electronic properties of a molecule,66 to do exciting lines) together with the 32 S/34 S isotopic effect
photochemical studies by means of Raman spectroscopy,67 thereby discriminating against the modes belonging to S3
or to simply enhance otherwise weak Raman features.68 and other cyclic S6 , S7 , and S8 molecules.80 Also, the vibra-
While on the one hand there are specific benefits to be had, tional Raman spectrum of sodium iodide was studied at
there can be, on the other hand, severe problems, for the 1084 K. Bands of both the monomer and the dimer were
accompanying fluorescence overpowers the weaker Raman identified and the thermodynamic properties of the dimer
features. A clean way to avoid the resonance fluorescence molecule were calculated.81
is to shift the Raman excitation wavelength from the visi- Shortly after the discovery of the Raman effect the spec-
ble into the IR region of the spectrum, as was successfully tra of gases were recorded under conditions of elevated
done in the case of the NO2 ! N2 O4 system65 (also see pressures in order to increase the intensity of the Raman
the section on Applications of FT-Raman Spectroscopy scattered light. The pressure dependence of Raman intensi-
in Part 5 of Volume 2), or all the way into the UV. ties was first investigated by Weiler.82 In that study the pure
More specific, and focused on the exploration of molec- rotational Raman spectrum of gaseous CO2 was recorded
ular structure and behavior in excited electronic states, is under pressures ranging up to 8 MPa (80 atm). It was not
resonance Raman spectroscopy in the UV and vacuum- until the 1950s and thereafter that quantitative studies were
ultraviolet (VUV) regions.69 In contrast to nonresonant undertaken that related the intensities, shapes, widths, and
positions of the Raman lines to the gas density. (The gas physical parameter of concern. This polarization, being a
density is a physically more meaningful quantity than the function of the impressed electric field, is expanded in a
pressure that controls these effects. The unit of density is power series of the electric field strength E,
the amagat, which is the ratio of the density of a gas to
that at standard conditions.) This aspect of Raman spec- P D c1 E C c2 EE C c3 EEE C 14
troscopy thus relates to the study of intermolecular forces.
where the cm are tensors of rank m C 1. These bulk elec-
Notable accomplishments were not only the quantitative
tric susceptibility tensors can ultimately be expressed in
determination of line-broadening parameters,18 20 but also
terms of the molecular polarizability tensors. The first term
the discoveries of free rotation of diatomic hydrogen in the
(m D 1) gives rise to linear, spontaneous Raman scatter-
condensed (liquid or solid) state,83,84 and the existence of
ing discussed in the previous section. The second term
free, or quasi-free, rotations of diatomic oxygen and nitro- gives rise to hyper-Rayleigh and hyper-Raman scatter-
gen as well as methane in the liquid state.85 In addition, ing. The third term, cubic in the electric field strength,
narrowing and shifts of the Q-branches of pressurized nitro- is responsible for the appearance of a variety of coher-
gen and carbon monoxide were observed.86 These effects ent Raman processes the most prominent of which are the
are due to the broadening and ultimate overlapping, i.e. CARS, stimulated Raman scattering (SRS) with its two spe-
interactions, of the rotational energy levels with increasing cial variantsstimulated Raman gain (SRG) and stimulated
density. They thus provide important data for the deter- Raman loss (SRL) processes and photoacoustic Raman
mination of intermolecular potentials and must be allowed scattering (PARS). Hyper-Rayleigh and hyper-Raman scat-
for in high-temperature thermometry and the modeling of tering by molecules have been studied earlier but these
coherent anti-Stokes Raman scattering (CARS) spectra (see processes now play a more prominent role in the field
below). These few examples are only suggestive of the of solid-state/condensed-matter physics and will not be
large amount of work that is being done using Raman spec- discussed here91 100 (see Hyper-Raman Spectroscopy).
troscopy in the study of compressed gases and liquids.87 (Hyper-Rayleigh scattering of incident light with wavenum-
ber nQ 0 produces scattered light with wavenumber 2nQ 0 , while
in hyper-Raman scattering the scattered light has wavenum-
3 NONLINEAR RAMAN bers 2nQ 0 nQ i where the nQ i are the molecular rotation
SPECTROSCOPY or rotationvibrational wavenumbers.2,4,5 Hyper-Rayleigh
scattering and its extension to hyper-Raman scattering was
The previous section dealt with experiments and con- predicted on the basis of rigorous quantum mechanics
cepts capable of being understood within the framework before the arrival of the laser.91,92 )
of Placzeks polarizability theory. Within this conceptual
framework any attempts to modify the Raman scattering
process by externally impressed electric fields were not 3.1 Coherent anti-Stokes Raman scattering
successful88,89 and potential modifications of the molecular
polarizability by such fields were shown to be ineffective.90 CARS spectroscopy has been most prominently pursued,
Even the use of high-power continuous-wave lasers with both theoretically and experimentally (see Part 4 of this
their high-energy density in a collimated beam is in princi- volume). This intense engagement stems from the practical
ple not different from the earlier use of mercury arc lamp applications of CARS to high-temperature thermometry and
excitation of Raman spectra. The intrinsic coherence of the species determination in flames, combustors, jet exhausts,
laser light is not utilized and individual molecules scatter etc. The literature is now very extensive and cannot be
light independently and with arbitrary phase relationships considered fully here.101 111 The following examples will
to one another. The resulting scattered light is incoherent. have to suffice to illustrate the capabilities and also the
Use of high-power pulsed lasers or of several laser beams limitations of this technique to molecular structure studies.
that interact with the scattering medium generates electric Highly accurate values of the molecular parameters for
fields of sufficient strength to give rise to new effects that nitrogen, oxygen and other diatomic and linear molecules
can only be explained on the basis of a nonlinear scatter- have been obtained at Doppler limited resolutions (ca.
ing formalism (see Part 4 of this volume). In such cases 0.003 cm1 ). Figure 7 shows the fully resolved Q-branch
the scattering process is viewed as the collective, phase- of 14 N2 obtained via an intracavity CARS technique.112
correlated behavior of a molecular ensemble and, in analogy Though the wavenumber scales are reversed and the inten-
to Placzeks polarizability theory for the single molecule, sity ratios of the lines are different on account of the
the induced dielectric polarization of the bulk medium P quadratic dependence of the CARS intensity on the sus-
(dipole moment per unit volume) is the (phenomenological) ceptibility c3 , this spectrum is in full agreement with the
Q(0)Q(2) Q(4) Q(6) Q(8) Q(10) Q(12) Q(14) 6=N

K = 02 4 6
N2 in air 5
213 C.P.S.
20 mbar, 297 K 7 4
CARS intensity
P1: 116W, 19429.77 cm1

3
P2: 180 mW
RC = 0.25s 8 2
1
Obs. 0
CARS intensity
2330.0 2329.0 2328.0 2327.0 2326.0
Wavenumber /cm1 Calc.
Figure 7. CARS spectrum of the Q-branch of nitrogen in air at
a total pressure of 2 kPa. P1 is the argon laser power, P2 is the
dye laser power, and RC is the time constant. (Reproduced by
permission from Frunder et al.112 .) The spectrum was recorded
intracavity with the argon and dye laser stabilized spectrometer
described in Schrotter et al.108 (Compare this CARS spectrum Calc.: TROT = 540 K
with the FT spectrum shown in Figure 3.)
more recent spontaneous Raman spectrum obtained with an

FT spectrometer (see Figure 3). Examples of many other
spectra are found in references101 112 and elsewhere in the
literature. Of considerable interest are the results obtained 3001 3002 3003 3004
for molecular radicals and ions,112 114 as well as for Wavenumber /cm1
gaseous van der Waals and hydrogen-bonded complexes, Figure 8. High-resolution CARS spectrum of CH3 produced by
and their solid condensates and phase transitions.116 119 266 nm photolysis of methylodide in a molecular beam. (Repro-
duced by permission from Triggs et al.112 )
Figure 8 shows the high-resolution CARS spectrum of the
CH3 radical produced from CH3 I by photolysis with 266 nm SRG and SRL techniques do not suffer from all of these
UV radiation.113,114 This radical plays an important role shortcomings. For these processes the gain (or loss) depends
in reaction chemistry and its spectroscopic study was, for only linearly on the third-order susceptibility and the inter-
many years, elusive. ferences due to the nonresonant parts are absent from the
There are, however, several shortcomings that afflict signal. SRG and SRL spectra thus allow a direct compari-
CARS spectra. Foremost among these is the dependence of son with spontaneous Raman spectra and their interpretation
the CARS signal intensity on the square of the third-order in terms of conventional spectroscopic models is straight-
nonlinear susceptibility c3 . This susceptibility, which is forward. The discrimination of strong Raman transitions
a complex quantity, is composed of a resonant part that against weaker ones remains, however.
involves the molecular rotation and rotationvibration fre-
quencies, and a nonresonant part. The contribution of the
resonant part to the CARS line intensity depends, among 3.2 Stimulated Raman gain or loss spectroscopy
other factors, critically on the Raman line width, while the
presence of the nonresonant part causes line distortions and The techniques of SRG and SRL spectroscopy have been
line shifts. These effects make a direct interpretation of described in a number of publications120 124 (see Part 4
the observed CARS spectra in terms of the molecular con- of this volume). An example of recent results is given in
stants difficult and require extensive computer modeling. In Figure 9, which shows the SRL spectrum of the n4 -band
addition, most often only the strong Raman features (large of diborane, 11 B2 H6 . Only the Q-branch lines of this mode
Raman scattering cross-sections) are sufficiently amplified were observed, even though the O-, P-, R-, and S-branch
to be recorded. The quadratic dependence of the CARS transitions are allowed by the selection rules. Because of
intensity on the third-order susceptibility causes a fur- the lack of these transitions, in this as well as in most
ther discrimination against the weaker Raman transitions. other cases where only the Q-branches are observed, there is
The overall result is that CARS spectra show most often often an underlying uncertainty about the correctness of the
only the Q-branches of the stronger Raman modes. The assignments, a problem that is usually solved by checking
n3 -band of ethane,128 the n1 -band of ketene,129 jet-cooled

ethylene,130 oxygen,131 and benzene(water)n clusters.132
3.3 Photoacoustic Raman scattering

(b)
SRL processes populate excited rotation or rotationvib-
rational states. Relaxation from these levels is mainly by
collisional energy transfer to other, unexcited molecules.
If pulsed lasers are used for the pumping step, then the
energy transfer is also pulsed. A pressure wave propagates
through the gas and is detected by a suitably placed micro-
phone. Tuning the lasers through the rotational or rota-
791.0 790.5 790.0 789.5 789.0 788.5 788.0 tionvibrational resonances then maps out the photoacous-
(a) Wavenumber /cm1 tic Raman spectrum. Since for elastic scattering no energy is
Figure 9. SRG spectrum of the n4 -band of diborane, 11 B2 H6 .
injected into the gaseous medium, there is no Rayleigh line
(a) Experimental spectrum; (b) spectrum computed with the con- in the PARS spectrum. This absence can be used to great
stants derived from the n7 n4 IR spectrum. The agreement advantage in pure rotational Raman spectroscopy where the
betwen the experimental and simulated spectra is excellent. very intense Rayleigh scattered radiation has been the tra-
(Reproduced by permission from Flaud et al.126 ) ditional bane of Raman spectroscopists.133 135 (See also
Part 8 of Volume 2.)
the consistency of combination differences. Furthermore,
line overlaps introduced uncertainties in the wavenumbers
and the analysis ultimately depended, therefore, on a band 3.4 Double resonance techniques
contour approach applied to overlapping lines to obtain
the spectroscopic constants for the upper state.125 Of these The great increase in the strength of a Raman signal due
constants only the band origins and the three rotation con- to any of the nonlinear processes (CARS, SRG, SRL,
stants for the n14 states of 11 B2 H6 and 10 B11 BH6 are listed PARS, etc.) is often still judged to be inadequate for high-
in Table 2. A more recent study of the n17 n4 difference resolution spectroscopy (note, for example, that no O-, P-,
bands by means of high-resolution IR spectroscopy126 of R-, or S-branch transitions were observed for the SRL
10
B2 H6 and 11 B2 H6 at a resolution of 0.003 cm1 provided diborane spectrum owing to their low intensity) and sev-
a slightly different, and more trustworthy, set of constants eral means have been proposed and explored to increase
for the n14 state of 11 B2 H6 . These are also listed in Table 2 the intensity of the Raman signal. A simple increase in the
for comparison. The different results obtained from the laser pump power is contraindicated owing to the alternat-
SRL and IR spectra demonstrate the extreme care that must ing current Stark broadening and shifts of the transitions
be exercised in relying exclusively on Raman data, espe- that result from the intense electric fields at the focus of
cially those from only the Q-branches, no matter how high the laser beams, for this brings forth an undesirable loss
the instrumental resolution. Similar work, based on stim- of resolution. To enhance the strength of the observed
ulated Raman spectroscopy only, or on a combined SRL Raman signals while maintaining high resolution, two
and IR study, resulted in new information for CH3 D,127 the double-resonance methods have been developed. In the first,
Table 2. Molecular constants for the n14 states of 10

B2 H6 and 11
B2 H6 (in cm1 ).
Constant IRa SRLb

11
10
B2 H6 11
B2 H6 B2 H6
En 819.292954 (16) 790.991212 (60) 790.98366 (84)

A 2.654704146 (430) 2.65449530 (340) 2.6534216 (680)
B 0.636984207 (510) 0.602152079 (770) 0.60227751 (800)
C 0.581801691 (560) 0.553052251 (750) 0.5529128 (170)
a Data taken from Flaud et al.126
b Data taken from Domenech et al.125
Virtual state
J =1
K = 1 Q-branch
C6H6+
Virtual state
Relative ion signal
J =3
529 nm 591 nm
1B 0
2u
532 nm 594 nm
J =5
3
J =7 V2 = 2 2
J
1
0
3
30 MHz 1
V2 = 1 2 J
1A
1g 0 1
1609.5 1609.51 1609.52 1609.53 0
v
Raman shift /cm1
Figure 10. RamanREMPI spectrum of the O Q1 J D 0, K D 3
2, K D 1 transition of the n16 e2g mode of benzene in a Ground 2
J
molecular beam. The right-hand side shows the double-resonance state 1
transition leading to the benzene ion. (Reproduced by permission 0
from Leuchs et al.137 ) Pumping Spectroscopy
Figure 11. Schema of RamanRaman double resonance spec-
stimulated Raman pumping of rotationvibration levels is troscopy. (Reproduced by permission from Bermejo et al.142 )
followed by ionization detection. First proposed in 1983136
this method has since been developed into a powerful tech-
nique especially with the resonance multiphoton ionization
3.5 Line widths and line shapes
(REMPI) process. Figure 10 shows the RamanREMPI
spectrum and the associated energy level diagram for the
O
Q1 J D 0, K D 2, K D 1 transition of the n16 e2g In CARS spectroscopy the probe laser is tuned through
mode in benzene in a molecular beam as an example of the molecular resonances and the line widths of the tran-
the capability of this technique.137 In a similar double sitions thus play a critical role in determining the size and
resonance technique, excitation of rotationvibration levels shape of the CARS signal. While this aspect may be benign
of glyoxal by stimulated Raman pumping was followed by when considering high-resolution CARS spectroscopy of
laser-induced fluorescence.138 140 low-pressure gases for the purpose of molecular struc-
While the goal of these double resonance techniques is to ture studies, it is of major importance for thermometry
enhance the strength of the observed Raman signal, a dif- and species determination of gases and gas mixtures at
ferent RamanRaman double resonance technique has been moderate to high temperatures and pressures. Since the
developed to probe selectively excited rotationvibrational extraction of accurate temperatures from observed CARS
states,141 143 emphasis being given to those states that are spectra requires computer modeling, the results depend
difficult to access even by IR spectroscopy. The Q-branches on the quality of the line width and line shape data that
of the 2n2 n2 bands of 12 C2 H2 and 13 C2 H2 have been are contained in the modeling procedure. Great efforts
recorded and analyzed using the pumping scheme shown in have therefore been expended to measure Raman line
Figure 11,142 while Figure 12 shows the Q-branch spectra widths and line shapes,144 154 and to theoretically account
of the n1 C n2 n2 bands of 13 C2 H2 and of 12 C2 H2 as well for them in terms of molecular interactions as func-
as of the 2n2 C 2n5 n2 band of 13 C2 H2 .143 These examples tions of temperature and gas density.155 161 The accuracy
demonstrate the enormous advances that have been made of temperature determination with CARS using inaccu-
in recent years in high-resolution Raman spectroscopy of rate input parameters has received attention from several
gases. quarters.162
wavenumber) domain; wavenumbers were determined by

12C H
2 2
1 + 2 2
means of one of several kinds of length-based
Raman intensity (a.u.)
260 Pa, 195 K measurements, be they done with classical spectrographs

or tunable laser sources. An ensemble of molecules pre-
2 pared to oscillate coherently by means of pulsed laser beams

is not interrogated as to its dynamical behavior by means
of these techniques. In contrast, pico- and femtosecond
spectroscopies were developed for the purpose of study-

ing the dynamical behavior of molecular ensembles, and

0 these nonlinear spectroscopies in the temporal domain offer
3358 3360 significant opportunities for high-resolution Raman spec-
troscopy. In pulsed excitation the phase coherence of the
(a) Wavenumber /cm1
ensemble of vibrating, rotating molecules does not last
indefinitely and is subject to a free exponential decay, with a
3 13
C2H2 characteristic phase relaxation time T2 , after the termination
1 + 2 2
260 Pa, 195 K
of the excitation. The measurement of this phase relaxation,
2
and the determination of T2 , is accomplished by perform-
ing a time-resolved CARS experiment. After excitation by
the pump and Stokes pulses, the probe pulse is applied at a
1 time t D td after the excitation reaches its maximum at time
t D t0 . By varying the time delay t D td t0 while keep-

ing the frequency difference between the laser and Stokes
0
pulses constant, the relaxation of the coherent excitation
3340 3341 3342 is mapped out as a function of time. A Fourier transfor-
(b) Wavenumber /cm1 mation of this temporal interferogram yields a spectrum in

frequency units (Hz) from which the wavenumbers, nQ , are
obtained via the relation nQ D nc , where n is the frequency
13C H
2 2 and c is the speed of light in vacuum. No spectroscopy
21 + 25 5
260 Pa, 195 K
in the conventional sense is employed. Qualitative descrip-
2 tions of this nonlinear spectroscopy in the temporal domain
have been given in several places8,106 but for a complete
quantitative exposition see Manz and Woste163 and Felker
1 and Zewail.164
Using this technique the Q-branch of the n1 -band of
methane was investigated with picosecond lasers at a higher
0 resolution than was possible with stimulated Raman spec-
3331.0 3331.5 3332.0 troscopy and the remaining tensor coefficient was deter-
(c) Wavenumber /cm1 mined. Also, the higher resolution available with these
methods resulted in the observation of splittings in the pure
Figure 12. Q-branch spectra of 12 C2 H2 and 13 C2 H2 obtained with
the RamanRaman double resonance technique. (a) Lines marked rotational spectrum of oxygen, not capable of observation
with an asterisk correspond to transitions of the n1 Q-branch of the with frequency domain techniques. More recently time-
H12 C13 CH molecule present in the sample in natural abundance. resolved CARS spectroscopy with femtosecond lasers was
(b) The line marked with an asterisk corresponds to O(2) of n1 . applied to the study of the relaxation of hydrogen in its first
(c) The line marked with an asterisk corresponds to O(4) of n1 . excited vibrational state.165,166 The greater spectral band
(Reproduced by permission from Bermejo et al.143 )
width of femtosecond laser pulses enabled the excitation
of the full manifold of rotational levels of H2 populated at
room temperature and the resulting temporal interferogram
4 TIME-DOMAIN NONLINEAR
is shown in Figure 13. The rotational coherence is seen to
SPECTROSCOPY persist up to long time delays, but progressive loss of coher-
ence is easily noticeable. The analysis of this interferogram
In all of the foregoing versions of Raman spectroscopy, yielded shift and broadening coefficients of greater preci-
experiments were performed in the frequency (really, sion and accuracy than those available heretofore.
20
10
0
0 20 40 60 80 100
10
0
100 120 140 160 180 200
10
CARS signal (a.u.)
5
0
200 220 240 260 280 300
0
300 320 340 360 380 400
0
400 420 440 460 480 500
5
0
500 520 540 560 580 600
5
0
600 620 640 660 680 700
Delay/ps
Figure 13. Femtosecond generated CARS temporal interferogram of H2 at a pressure of 3 bar. Curve, theory; squares, experiment.
(Reproduced by permission from Lang et al.165 )
The method of time-delayed CARS spectroscopy in the separately measured, if at all. The problem is actually worse
temporal domain has several shortcomings, the principal when single longitudinal mode lasers are employed, for
one of which is that only the wavenumber differences then the actual value of the laser wavelength is not known
between individual rotational or rotationvibrational lines a priori since the laser may oscillate on any one of the
can be determined. For the determination of their abso- modes under the gain profile. In view of this the preci-
lute values, the absolute wavenumber of at least one of sion of Raman wavenumber determinations may be very
these transitions must be known. Adding the differences to high (close to 0.001 cm1 ) but their accuracy can still be
such absolute wavenumber will then yield the wavenumbers wanting. Incorporation of such wavenumbers into a set of
of the other transitions. The accuracies with which the IR determined wavenumbers then can lead to serious sys-
wavenumber differences can be determined with femtosec- tematic errors in the evaluation of the molecular constants.
ond time delayed spectroscopy is of the order 0.0001 cm1 . In the case of pure rotational (spontaneous) Raman spectra
this problem is easily solved by averaging the wavenum-
bers of the Stokes and anti-Stokes lines. An example of
5 WAVENUMBER CALIBRATION OF this problem is the earlier work on the FT-Raman spec-
RAMAN SPECTRA trum of H2 recorded with the FT spectrometer located on
Kitt Peak mountain. In this work the Raman wavenum-
The precision of wavenumbers of the spectra of molecules bers of the Q-branch lines exhibit a systematic difference
in the gaseous state determined by high-resolution Raman of 0.01 cm1 with respect to the values determined from
spectroscopy now rivals that determined by high-resolution the IR quadrupole absorption spectrum.26 Similarly, the
IR spectroscopy. Special attention must, however, also be wavenumbers of the recent FT-Raman spectrum of nitrogen
given to their accuracy, since Raman determined wavenum- had to be corrected on the basis of a systematic difference
bers require not only knowledge of the value of the between the values for the measured Raman lines of CO
measured wavenumber of the Raman line peak but also and those calculated on the basis of molecular constants
the wavenumber of the exciting line. In earlier work the determined by IR heterodyne frequency (in MHz, not in
wavenumber of the exciting line was either assumed to cm1 ) measurements.30 While in this latter case these cor-
be known (multimode HeNe laser, ArC laser) or was rections were small, they were of sufficient magnitude
to merit a comment in the published work. A similar A metallic phase of hydrogen under conditions of very
problem arises in nonlinear Raman spectroscopy where high pressure was predicted as early as 1935 and many
the wavenumber of the excitation pulse often depends on attempts have since been made to experimentally verify
the experimental parameters. Careful attention is therefore its existence. Besides being of intrinsic interest, this quest
now given to this problem in recent work.124,167 Indepen- for metallic hydrogen is of importance for substantiating
dent calibration of the spectrometer, be they classical FT models of the constitution and properties of the cores of
instruments or those based on nonlinear techniques, is there- the giant planets and Raman spectroscopy is one of the
fore essential. The heterodyne determined IR wavenumbers diagnostic tools used to study the behavior of hydrogen
and constants derived from them31 are judged to be the gas at very high pressures.183,184 Fraunhofer lines in the
most reliable and suitable for calibrations of spectra at the solar spectrum of the Earths atmosphere show depths and
Doppler limit of resolution at ambient temperatures. (See widths as functions of the zenith angle. In the literature
also Part 7 of this volume.) this filling in is known as the Ring effect.185 188 Its
physical cause is attributed to Raman scattering of the suns
radiation by gases of the Earths atmosphere, primarily the
6 SOME APPLICATIONS diatomics N2 and O2 . Similarly, Fraunhofer lines in the
ocean also are affected by the Ring effect.189,190 Raman
Applications of Raman spectroscopy are now legion. Be-
scattering even plays a role in the study of planetary
sides molecular structure studies the best known application
atmospheres,191 196 and the possibility of observing Raman
is the determination of temperatures and species in com-
scattering of Lyman-a radiation by molecular hydrogen in
bustion systems by CARS spectroscopy.144 154 Similarly,
interstellar space has been suggested.197,198 These are, of
the Lidar technique is used to probe the temperature and
course, natural phenomena, beyond mans intervention and
species concentration of the Earths upper atmosphere.
not applications of Raman spectroscopy to these objects
(These applications are described in Parts 8 of Volumes 2
in the conventional sense.
and 4.) Somewhat less well known are the uses of Raman
spectroscopy in the determination of gaseous species con-
tained in bubbles entrained in solids,168 171 in liquids,172 7 SUMMARY
and molecules entrapped in clathrates.173 176 In these latter
examples the goal is the determination of the paleocli- The major advances in Raman spectroscopy of gases have
matic composition of the Earths atmosphere over the last been traced from the mid-1950s to the present time. Spon-
100 000250 000 years by assaying the gaseous species taneous Raman spectroscopy and various schemes of non-
concentrations in the cavities of clathrate hydrates and gas linear Raman spectroscopy have been briefly described and
bubbles in ice cores taken from different locations and illustrated. High-resolution Raman spectroscopy of gases
at different depths. Inconsistent N2 /O2 ratios have been has now advanced to the stage of Doppler limited res-
reported.174 176 Recent measurements on the Vostok ice olution. Special attention is given to the problems of
core taken at a depth of 3316 m showed shifts of the Raman precision and accuracy of wavenumbers determined by
Q-branch line in the clathrate hydrate of 5 cm1 relative to Raman spectroscopy. A few examples are cited where
the value for the free molecule in the gas phase, while Raman spectroscopy is of special advantage for a variety
for oxygen similar results, including line splittings, were of applications.
obtained.176 These results depend on which polarization
configuration is employed, VV or HV. Samples taken from
a depth of 860 m revealed both bubbles and clathrates. Here ACKNOWLEDGMENTS
the nitrogen line exhibited a narrowing relative to the width
of the line for the free molecule in the gas phase. The Thanks are due to Drs W.J. Lafferty and D. Bermejo, and
position of the gas phase line in the bubble was nearly Profs J.W. Nibler and L.D. Ziegler for helpful discussions
coincident with that for the free molecule, as distinct from and advice, to Prof. Ziegler for providing a preprint of his
the N2 line in the clathrate.176 article, and to the various authors for permission to use
In a different application anti-Stokes rotational Raman illustrations from their publications.
scattering from hydrogen gas is used to calibrate the
intensity of the Thomson scattered signal from tokamak
plasmas.177 180 In another application Raman spectroscopy ABBREVIATIONS AND ACRONYMS
is used very effectively for mapping out temperature and
density profiles of supersonic jets in order to better under- CARS Coherent Anti-Stokes Raman Scattering
stand theIR gas dynamics.181,182 PARS Photoacoustic Raman Scattering
REMPI Resonance Multi Photon Ionization 23. G. Herzberg, Molecular Spectra and Molecular Structure
SRG Stimulated Raman Gain II. Infrared and Raman Spectra of Polyatomic Molecules,
SRL Stimulated Raman Loss D. van Nostrand Co., New York (1945).
SRS Stimulated Raman Scattering 24. D. Papousek and M.R. Aliev, Molecular VibrationRotation
VUV Vacuum-ultraviolet Spectra: Theory and Application of High Resolution
Infrared, Microwave and Raman Spectroscopy of Poly-
atomic Molecules, Elsevier, Amsterdam (1982).
REFERENCES 25. B. Cloppenburg, K. Manczak, H. Prockl, H.W. Schrotter and
G. Strey, Z. Naturforsch., 34a, 1160 (1979).
1. G. Placzek, Rayleigh Streuung und Raman Effect, in 26. D.E. Jennings, A. Weber and J.W. Brault, Appl. Opt., 25,
Handbuch der Radiologie, 2nd edition, ed. E. Marx, 284 (1986). (For earlier reports, see citations 14 given in
Akademische Verlagsgesellschaft, Leipzig, 205374, Vol. 6 reference 8.)
(1934). [English translation by A. Palm, Rayleigh Scatter- 27. D.E. Jennings, A. Weber and J.W. Brault, J. Mol. Spec-
ing and Raman Effect, U.C.R.L. Translation No. 526(L).] trosc., 126, 19 (1987).
2. J.A. Koningstein, Introduction to the Theory of the Raman 28. J. Bendtsen, Appl. Opt., 34, 3377 (1995).
Effect, D. Reidel Dordrecht (1972).
29. J. Bendtsen, F. Rasmussen and S. Brodersen, Appl. Opt., 36,
3. G. Placzek and E. Teller, Z. Physik, 81, 209 (1933).
5526 (1997).
4. D.A. Long, Raman Spectroscopy, McGraw-Hill, New York
(1977). 30. J. Bendtsen and F. Rasmussen, J. Raman Spectrosc., 31, 433
(2000).
5. D.A. Long, The Raman Effect, John Wiley & Sons, Chich-
ester (2001). 31. A.G. Maki and J.S. Wells, Wavenumber Calibration Tables
from Heterodyne Frequency Measurements, National Insti-
6. A. Weber, High Resolution Raman Studies of Gases, in
tute of Standards and Technology, Special Publication 821,
The Raman Effect, ed. A. Anderson, M. Decker, Inc.,
US Government Printing Office, Washington, DC (1991).
New York, 543757, Vol. 2 (1973).
7. A. Weber (ed.), Raman Spectroscopy of Gases and Liq- 32. A. Tabyaoui, B. Lavorel, G. Millot, R. Saint-Loup, R. Chaux
uids, Springer Verlag, Berlin (1979). and H. Berger, J. Raman Spectrosc., 21, 809 (1990).
8. A. Weber, Interferometric Raman Spectroscopy of Gases, 33. D. Reuter, D.E. Jennings and J.W. Brault, J. Mol. Spec-
in Fourier Transform Raman Spectroscopy: From Con- trosc., 115, 294 (1986).
cept to Experiment, eds B. Chase and J. Rabolt, Academic 34. J. Bendtsen, J. Raman Spectrosc., 2, 133 (1974).
Press, New York, 209270 (1994).
35. P.J. Hendra, Fourier Transform Raman Spectroscopy, in
9. W. Jeremy Jones, Can. J. Phys., 78, 327 (2000). Modern Techniques in Raman Spectroscopy, ed. J.J.
10. D.W. Lepard, D.E.M. Sweeney and H.L. Welsh, Can. J. Laserna, John Wiley & Sons, Chichester, 80 (1996).
Phys., 40, 1567 (1962). 36. A. Weber, Doppler Broadening of Raman Lines, in Pro-
11. D.W. Lepard, D.E. Shaw and H.L Welsh, Can. J. Phys., 44, ceedings of the XIth International Conference on Raman
2353 (1966). Spectroscopy, London, eds R.J.H. Clark and D.A. Long,
12. D. Lepard, PhD thesis, University of Toronto (1966). John Wiley & Sons, New York, 7980 (1988).
13. D.E. Shaw and H.L. Welsh, Can. J. Phys., 45, 3823 (1967). 37. A. Weber, J. Chem. Phys., 73, 3952 (1980); 74, 4754 (1981);
76, 3964 (1982); 88, 3428 (1988).
14. B.P. Stoicheff, Can J. Phys., 32, 339 (1954).
15. A. Langseth and B.P. Stoicheff, Can. J. Phys., 34, 350 38. I.F. Silvera and F. Tommasini, Phys. Rev. Lett., 37, 236
(1956). (1976).
16. E.G. Cox, D.W.J. Cruickshank and J.A.S. Smith, Nature, 39. H.P. Godfried and I.F. Silvera, J. Chem. Phys., 78, 121
175, 766 (1955). (1983).
17. E.G. Cox, Revs. Modern Phys., 30, 159 (1958). 40. H.P. Godfried and I.F. Silvera, Phys. Rev. Lett., 48, 1337
(1982).
18. K.S. Jammu, G.E. St. John and H.L. Welsh, Can. J. Phys.,
44, 797 (1966). 41. H.P. Godfried and I.F. Silvera, Phys. Rev. A, 27, 3008
19. A.D. May, G. Varghese, J.C. Stryland and H.L. Welsh, Can. (1983).
J. Phys., 42, 1058 (1964). 42. G. Luijks, S. Stolte and J. Reuss, Chem. Phys., 62, 217
20. L. Rosenmann, J.M. Hartmann, M.Y. Perrin and J. Taine, (1981); Chem. Phys. Lett., 94, 48 (1983).
Appl. Opt., 27, 3902 (1988). 43. W. Knippers, G. Luijks, S. Stolte and J. Reuss, Chem. Phys.
21. W.F. Murphy, W. Holzer and H.J. Bernstein, Appl. Spec- Lett., 109, 532 (1984).
trosc., 23, 211 (1969). 44. B.A. Palmer and R.S. McDowell, The Los Alamos
22. H.W. Schrotter and H.W. Klockner, Raman Scattering Fourier Transform Spectrometer: Applications to Molecular
Cross Sections in Gases and Liquids, in Raman Spec- Spectroscopy, in Proceedings of the 7th International
troscopy of Gases and Liquids, ed. A. Weber, Springer Conference on Fourier Transform Spectroscopy, ed.
Verlag, Berlin, 123166 (1979). D. Cameron, SPIE, Bellingham, WA, 104109, Vol. 1145
(1989). (An abbreviated version of this word is given in 70. B. Hudson, Spectroscopy, 1, 22 (1986).
reference 8. pp. 236237.) 71. P.B. Kelly, S. Li and B. Hudson, Applications of Ultra-
45. Y.M. Bosworth, R.J.H. Clark and D.M. Rippon, J. Mol. violet Resonance Raman Scattering in Molecular Elec-
Spectrosc., 46, 240 (1973). tronic Spectroscopy, in Advances in Laser Science-II, eds
46. J.R. Durig, J.F. Sullivan and T.S. Little, Utility of Gas M. Lapp, W.C. Stwalley and G.A. Kenney-Wallace, Amer-
Phase Raman Data for Structural and Potential Function ican Institute of Physics Conference Proceedings No. 160,
Determinations, in Raman Spectroscopy: Sixty Years On; American Institute of Physics, New York, 433435 (1987).
Vibrational Spectra and Structure, eds H.D. Bist, J.R. Durig 72. B.S. Hudson, Ultraviolet Resonance Raman Studies of
and J.F. Sullivan, Elsevier Science Publishers BV, Amster- Electronic Excitations, in Time-Resolved Vibrational Spec-
dam, 569597, Vol. 17B (1989). troscopy, eds A. Laubereau and M. Stockburger, Springer
47. J.R. Durig, J.F. Davis and G.A. Guirgis, J. Mol. Struct., 346, Verlag, Berlin, 170174 (1985).
197 (1995). 73. B.S. Hudson, Vacuum Ultraviolet Resonance Raman Spec-
48. J.R. Durig, G.A. Guirgis, I. Badejo and J.F. Davis, J. Mol. troscopy, in Recent Trends in Raman Spectroscopy, eds
Struct., 354, 15 (1995). S.B. Banerjee and S.S. Jha, World Scientific, Singapore,
49. V.M. Grosev, H.-W. Schrotter and J. Januscheit, J. Raman 368385 (1989).
Spectrosc., 26, 137 (1995). 74. L.D. Ziegler, Y.C. Chung, P. Wang and Y.P. Zhang, Res-
50. T.S. Bican, H.-W. Schrotter and V.M. Grosev, J. Raman onance Rotational Raman and Hyper-Raman Scattering: A
Spectrosc., 26, 787 (1995). Probe of Picosecond Dynamics, in Time Resolved Spec-
troscopy, R.J.H. Clark and R.E. Hester, John Wiley &
51. M.J. Lee, F. Fusheng, S.W. Hur, J. Liu, T.K. Gounev and Sons, Chichester, 55112 (1989).
J.R. Durig, J. Raman Spectrosc., 31, 157 (2000).
75. R.J. Sension and B.S. Hudson, J. Chem. Phys., 90, 1377
52. J.R. Durig, L. Zhou, T. Schwartz and T. Gounev, J. Raman (1989).
Spectrosc., 31, 193 (2000).
76. J.K.G. Watson, W. Siebrand, M. Pawlikowski and M.Z.
53. J.R. Durig, D.T. Durig, J.B. Robb, II, G.A. Guirgis, M. Zhen
Zgierski, J. Chem. Phys., 105, 1348 (1996).
and H.V. Phan, J. Raman Spectrosc., 31, 203 (2000).
77. P.A. Freedman, W.J. Jones and A. Rogstad, J. Chem. Soc.
54. I-Yan Wang and A. Weber, Ind. J. Pure Appl. Phys., 16, 358
(1978). Faraday Trans. II, 71, 286 (1975).
55. E. Kostyk and H.L. Welsh, Can. J. Phys., 58, 534 (1980). 78. N.J. Brassington, H.G.M. Edwards and D.A. Long, J. Raman
Spectrosc., 11, 346 (1981).
56. I-Yan Wang and A. Weber, J. Chem. Phys., 67, 3084 (1977).
79. V. Boudon, E.B. Mkadmi, H. Burger and G. Pierre, Chem.
57. G. Tejeda, J.M. Fernandez-Sanchez and S. Montero, Appl. Phys. Lett., 305, 21 (1999).
Spectrosc., 51, 265 (1997).
80. M.S. Boumedien, J. Corset and E. Picquenard, J. Raman
58. S. Hadrich, S. Hefter, B. Pfelzer, T. Doerk, P. Jauernik and Spectrosc., 30, 463 (1999).
J. Uhlenbusch, Chem. Phys. Lett., 256, 83 (1996).
81. M. Fields, R. Devonshire, H.G.M. Edwards and V. Fawcett,
59. H. Sano, G. Mizutani and S. Ushioda, Chem. Phys. Lett.,
Chem. Phys. Lett., 240, 334 (1995).
194, 398 (1992).
60. J. Burris, T.J. McGee and W. Heaps, Appl. Spectrosc., 46, 82. J. Weiler, Ann. Physik, 23, 493 (1935).
1076 (1992). 83. J.C. McLennan and J.H. McLeod, Trans. R. Soc. Can. III,
61. G.W. Faris and R.A. Copeland, Appl. Opt., 36, 2684 (1997). 23, 19 (1929).
62. L. Li, T. He, X. Wang, J. Zuo, C. Xu and F.-C Liu, Chem. 84. J.C. McLennan and J.H. McLeod, Nature, 123, 60 (1929).
Phys. Lett., 262, 52 (1996). 85. M.F. Crawford, H.L. Welsh and J.H. Harrold, Can. J. Phys.,
63. P. Vogt, M. Schmitt and W. Kiefer, J. Raman Spectrosc., 26, 30, 81 (1952).
861 (1995). 86. A.D. May, J.C. Stryland and G. Varghese, Can. J. Phys., 48,
64. P. Vogt, M. Ganz and W. Kiefer, J. Raman Spectrosc., 27, 2331 (1970).
69 (1996). 87. A.I. Burstein and S.I. Temkin, Spectroscopy of Molecular
65. C. Dyer and P.J. Hendra, Chem. Phys. Lett., 233, 461 Rotation in Gases and Liquids, Cambridge University Press,
(1995). Cambridge (1994).
66. P.B. Kelly and B.S. Hudson, Chem. Phys. Lett., 114, 451 88. W. Buchheim, Phys. Z., 36, 694 (1935).
(1985). 89. A.E. Douglas, J. Chem Phys., 16, 849 (1949).
67. P.B. Kelly and S.G. Westre, Chem. Phys. Lett., 151, 253
90. C.A. Coulson, A. Maccoll and L.E. Sutton, Trans. Faraday
(1988).
Soc., 48, 106 (1952).
68. R.E. Barletta and J.T. Veligdan, Appl. Spectrosc., 49, 532
91. T. Neugebauer, Z. Phys., 155, 380 (1959).
(1995).
69. G. Herzberg, Molecular Spectra and Molecular Structure 92. T. Neugebauer, Acta Phys. Acad. Sci. Hung., 10, 221 (1959).
III. Electronic Spectra and Electronic Structure of Poly- 93. M.J. French, Hyper Rayleigh and Hyper Raman Spec-
atomic Molecules, D. van Nostrand Co., Princeton (1966). troscopy, in Chemical Applications of Nonlinear Raman
Spectroscopy, ed. A.B. Harvey, Academic Press, New 112. H. Frunder, D. Illig, H. Finsterholzl, A. Beckmann, H.W.
York, 239279 (1981). Schrotter, B. Lavorel and G. Roussel, High-resolution
94. D.A. Long, Theory of Hyper Rayleigh and Hyper Raman CARS spectroscopy of Gases in the Cavity of a CW Argon
Scattering, in Introduction to Nonlinear Laser Spectro- Ion Ring Laser, in Laser Spectroscopy VI, Springer Series
scopy, revised edition, eds M.D. Levenson and S.S. Kano, in Optical Sciences, eds H.P. Weber and W. Luthy, Springer
Academic Press, New York, 165179 (1988). Verlag, Berlin, 334335, Vol. 40 (1983).
113. N.E. Triggs, M. Zahedi, J.W. Nibler, P. DeBarber and
95. L.D. Ziegler and J.L. Roebber, Chem. Phys. Lett., 136, 377
J.J. Valentini, J. Chem. Phys., 96, 1822 (1992).
(1987).
114. K. Ravichandran, Yin Bai and T.R. Fletcher, Chem. Phys.
96. L.D. Ziegler, Y.C. Chung and Y.P. Zhang, J. Chem. Phys.,
Lett., 261, 261 (1996).
87, 4498 (1987).
115. G.R. Mohlmann, Chem. Phys. Lett., 115, 226 (1985).
97. Y.C. Chung and L.D. Ziegler, J. Chem. Phys., 88, 7287
(1988). 116. R. Beck and J.W. Nibler, Chem. Phys. Lett., 148, 271
(1988); 159, 79 (1989).
98. L.D. Ziegler, J. Raman Spectrosc., 21, 769 (1990).
117. R.D. Beck, M. Hineman and J.W. Nibler, J. Chem. Phys.,
99. C.C. Bonang and S.M. Cameron, Chem. Phys. Lett., 192, 92, 7068 (1990).
303 (1992).
118. K.H. Lee, K.W. Brown, N.E. Triggs, A.D. Richardson,
100. D.J. Campbell and L.D. Ziegler, Chem. Phys. Lett., 201, 159 N.H. Rich and J.W. Nibler, J. Chem. Phys., 98, 10100
(1993). (1993).
101. G.L. Eesley, Coherent Raman Spectroscopy, Pergamon 119. K.H. Lee, N.E. Triggs and J.W. Nibler, J. Phys. Chem., 98,
Press, Oxford (1981). 4382 (1994).
102. J.W. Nibler and G.V. Knighten, Coherent Anti-Stokes 120. P. Escherick and A. Owyoung, High Resolution Stimulated
Raman Spectroscopy in Raman Spectroscopy of Gases and Raman Spectroscopy, in Advances in Infrared and Raman
Liquids, ed. A. Weber, Springer Verlag, Berlin, 253299 Spectroscopy, eds R.J.H. Clark and R.E. Hester, Heyden,
(1979). London, 130187 (1982).
103. S. Yu Volkov, D.N. Kozlov, A.N. Prokhorov, V.V. Smirnov 121. J.S. Santos Gomez, Coherent Raman Spectroscopy, in
and V.I. Fabelinskiy, Coherent Anti-Stokes Spectroscopy Modern Trends in Raman Spectroscopy, ed. J.J. Laserna,
of Molecular Gases, in Proceedings of the Institute of John Wiley & Sons, Chichester, 305342 (1996).
General Physics, ed. P.P. Pashinin, Nova Science Publ.,
Commack, NY, Vol. 2 (1987). 122. H. Berger, B. Lavorel and G. Millot, High Resolution Stim-
ulated Raman Spectroscopy of Gases, in Recent Trends in
104. M.D. Levenson and S.S. Kano, Introduction to Nonlinear Raman Spectroscopy, eds S.B. Banejee and S.S. Jha, World
Laser Spectroscopy, revised edition, Academic Press, New Scientific, Singapore, 6577 (1989).
York (1988).
123. D. Bermejo, J. Santos, P. Cancio, J.L. Domenech, C.
105. G. Marowsky and V.V. Smirnov (eds), Coherent Raman Domingo, J. Ortigoso and R. Escribano, J. Raman Spec-
Spectroscopy, Springer Verlag, Berlin (1992). trosc., 21, 197 (1990).
106. R.J.H. Clark and R.E. Hester, Advances in Non-linear 124. J. Santos, P. Cancio, J.L. Domenech, J. Rodriguez and
Spectroscopy, John Wiley & Sons, Chichester, Vol. 15 D. Bermejo, Laser Chem., 12, 53 (1992).
(1988).
125. J. Domenech, D. Bermejo, J.-M. Flaud and W.J. Lafferty, J.
107. J.W. Nibler and G.A. Pubanz, Coherent Raman Spectro- Mol. Spectrosc., 194, 185 (1999).
scopy of Gases, in Advances in Non-linear Spectroscopy,
126. J.-M. Flaud, W.J. Lafferty, H. Burger, G. Pawelke, J.-L.
eds R.J.H. Clark and R.E. Hester, John Wiley & Sons,
Domenech and D. Bermejo, J. Mol. Spectrosc., 203, 339
Chichester, 150, Vol. 15 (1988).
(2000).
108. H.W. Schrotter, H. Frunder, H. Berger, J.P. Boqillon, B. 127. D. Bermejo, J. Santos, P. Cancio, J.L. Domenech, C.
Lavorel and G. Millot, High Resolution CARS and Inverse Domingo, J.M. Orza, J. Ortigoso and R. Escribano, J. Raman
Raman Spectroscopy, in Advances in Non-linear Spec- Spectrosc., 21, 197 (1990).
troscopy, eds R.J.H. Clark and R.E. Hester, John Wiley &
Sons, Chichester, 97147, Vol. 15 (1988). 128. D. Bermejo, J. Santos, P. Cancio, J.M. Fernandez-Sanchez
and S. Montero, J. Chem. Phys., 97, 7055 (1992).
109. H.W. Schrotter and J.P. Boquillon, State-of-the-Art on High
Resolution CARS, in Raman Spectroscopy: Sixty Years 129. R. Escribano, J.L. Domenech, P. Cancio, J. Ortigoso,
On, eds H.D. Bist, J.R. Durig and J.F. Sullivan, Elsevier, J. Santos and D. Bermejo, J. Chem. Phys., 101, 937 (1994).
Amsterdam, 114, Vol. 17B (1989). 130. L. Abad, D. Bermejo, R. Escribano, V.J. Herrero, J. Santos,
110. D.A. Greenhalgh, Quantitative CARS Spectroscopy, in I. Tanarro, G.D. Nivellini and L. Ramonat, Chem. Phys.
Advances in Non-linear Spectroscopy, eds R.J.H. Clark Lett., 227, 248 (1994).
and R.E. Hester, John Wiley & Sons, Chichester, 193251, 131. G. Rouille, G. Millot, R. Saint-Loup and H. Berger, J. Mol.
Vol. 15 (1988). Spectrosc., 154, 372 (1992).
111. T.S. Bican, J. Jonuscheit, U. Lehner and H.W. Schrotter, J. 132. P.M. Maxton, M.W. Schaeffer and P.M. Felker, Chem. Phys.
Raman Spectrosc., 26, 699 (1995). Lett., 241, 603 (1995).
133. G.A. West, D.R. Siebert and J.J. Barrett, J. Appl. Phys., 51, 157. X. Michaut, J.P. Berger, P.M. Sinclair and H. Berger, J.
2823 (1980). Quant. Spectrosc. Radiat. Transfer, 60, 585 (1998).
134. J.J. Barrett, Photoacoustic Raman Spectroscopy of 158. J. Buldyrova, J. Bonamy and D. Robert, J. Quant. Spec-
Gases, in Chemical Applications of Nonlinear Raman trosc. Radiat. Transfer, 62, 321 (1999).
Spectroscopy, ed. A.B. Harvey, Academic Press, New York 159. M.-L. Dubernet and P.A. Tuckey, Chem. Phys. Lett., 300,
89169 (1981). 275 (1999).
135. J.J. Barrett and D.F. Heller, J. Opt. Soc. Am., 71, 1299 160. D. Pieroni, J.M. Hartmann, F. Chaussard, X. Michaut,
(1981). T. Gabard, R. Saint-Loup, H. Berger and J. Champion, J.
136. P. Escherick and A. Owyoung, Chem. Phys. Lett., 103, 235 Chem. Phys., 112, 13351343 (2000).
(1983). 161. F. Chaussard, X. Michaut, R. Saint-Loup, H. Berger, P. Jou-
137. M. Leuchs, M. Crew, J. Harrison, M. Hineman and J.W. bert, B. Lance, J. Bonamy and D. Robert, J. Chem. Phys.,
Nibler, J. Chem. Phys., 105, 4885 (1996). 112, 158 (2000).
138. D.A. King, R. Haines, N.R. Isenor and B.J. Orr, Opt. Lett., 162. S. Kroll, P.-E. Bengtsson, M. Alden and D. Nilsson, Appl.
8, 629 (1983). Phys. B, 51, 25 (1990).
139. A.B. Duval, D.A. King, R. Haines, N.R. Isenor and B.J. Orr, 163. J. Manz and L. Woste (eds), Femtosecond Chemistry,
J. Opt. Soc. Am. B, 2, 1570 (1985). VCH, Weinheim, Vols 1 and 2 (1995).
140. A.B. Duval, D.A. King, R. Haines, N.R. Isenor and B.J. Orr, 164. P.M. Felker and A.H. Zewail Molecular Structures
J. Raman Spectrosc., 17, 177 (1986). from Ultrafast Coherence Spectroscopy, in Femtosecond
141. R.Z. Martinez, D. Bermejo, J. Santos, J.P. Champion and Chemistry, eds J. Manz and L. Woste, VCH, Weinheim,
J.C. Hilico, J. Chem. Phys., 107, 4864 (1997). 193260, Vol. 1 (1995).
142. D. Bermejo, P. Cancio, G. Di Lonardo and L. Fusina, J. 165. T. Lang, K.-L. Kompa and M. Motzkus, Chem. Phys. Lett.,
Chem. Phys., 108, 7224 (1998). 310, 65 (1999).
143. D. Bermejo, R.Z. Martinez, G. Di Lonardo and L. Fusina, 166. T. Lang and M. Motzkus, J. Raman Spectrosc., 31, 65
J. Chem. Phys., 111, 519 (1999). (2000).
144. Laser Diagnostics in Combustion, Appl. Phys. B, 51(1) 167. M.L. Orlov, J.F. Ogilvie and J.W. Nibler, J. Mol. Spectrosc.,
(1990). 185, 128 (1997).
145. A.C. Eckbreth, Laser Diagnostics for Combustion Temper- 168. J.J. Blaha, Raman Microprobe Spectroscopic Analysis, in
ature and Species, 2nd edition, Gordon & Breach, Amster- Vibrational Spectra and Structure, ed. J.R. Durig, Elsevier,
dam (1996). Amsterdam, Vol. 10 (1981).
146. For CO2 , H2 O, N2 and O2 see L. Rosenmann, J.M. Hart- 169. G.J. Rosasco and J.H. Simmons, Am. Ceram. Soc. Bull., 53,
mann, M.Y. Perrin and J. Taine, Appl. Opt., 27, 3902 (1988). 626 (1974).
147. For H2 see R.L. Farrow, L.A. Rahn, G.O. Sitz and G.J. 170. G.J. Rosaco and E. Roedder, Geochim. Cosmochim. Acta,
Rosasco, Phys. Rev. Lett., 63, 746 (1989). 43, 1907 (1979).
148. For H2 rare gas mixtures see J.P. Berger, R. Saint-Loup, H. 171. M.C. Drake, G.J. Rosasco, R. Schneggenburger and R.L.
Berger, J. Bonamy and D. Robert, Phys. Rev. A, 49, 3396 Nolen, Jr, J. Appl. Phys., 50, 7894 (1979).
(1994). 172. M. Lankers, J. Popp, G. Rossling and W. Kiefer, Chem.
149. For H2 perturbed by N2 see P.M. Sinclair, J.P. Berger, X. Phys. Lett., 277, 331 (1997).
Michaut, R. Saint-Loup, R. Chaux, H. Berger, J. Bonamy 173. S.L. Miller, in Physics and Chemistry of Ice, eds E. Whal-
and D. Robert, Phys. Rev. A, 54, 1 (1996). ley, S.J. Jones and I.W. Gold, Royal Society of Canada,
150. For O2 see G. Millot, B. Lavorel and G. Fanjoux, J. Mol. Otawa, 4250 (1973).
Spectrosc., 176, 211 (1996). 174. J. Nakahara, Y. Shigesato, A. Higashi, T. Hondoh and
151. For N2 see G. Fanjoux, G. Millot and B. Lavorel, J. Raman C.C. Langway, Jr, Phil. Mag. B, 57, 421 (1988).
Spectrosc., 27, 475 (1996). 175. F. Pauer, J. Kipfstuhl and W.F. Kuhs, Geophys. Res. Lett.,
152. For the Fermi diad in CO2 see G. Fanjoux, B. Lavorel and 22, 969 (1995).
G. Millot, J. Raman Spectrosc., 29, 391 (1998). 176. B. Champagnon, G. Panczer, B. Chazallon, L. 1Arnaud,
153. For H2 and H2 argon mixtures see T. Lang and M. Motzkus, P. Duval and V. Lipenkov, J. Raman Spectrosc., 28, 711
J. Raman Spectrosc., 31, 65 (2000). (1997).
154. For H2 in supercritical H2 O and CO2 see S.F. Rice and 177. T. Yamauchi and I. Yanagisawa, Appl. Opt., 24, 700 (1985).
J.J. Wickham, J. Raman Spectrosc., 31, 619 (2000). 178. F. Flora and L. Giudicotti, Appl. Opt., 26, 4001 (1987).
155. L. Bonamy, J. Bonamy, D. Robert, S.I. Temkin, G. Millot 179. M. Bassan, L. Giudicotti and R. Pasqualotto, Appl. Opt., 32,
and B. Lavorel, J. Chem. Phys., 101, 7350 (1994). 5313 (1993).
156. J. Forsman, J. Bonamy, D. Robert, J.P. Berger, R. Saint- 180. F. Orsitto, A. Brusadin, E. Giovannozzi, D. Santi, R. Bar-
Loup and H. Berger, Phys. Rev. A, 52, 2652 (1995). tiromo and P. Pizzolati, Appl. Opt., 34, 2712 (1995).
181. G. Tejeda, B. Mate, J.M. Fernandez-Sanchez and S. Mon- 189. G.W. Kattawar and X. Xu, Appl. Opt., 31, 6491 (1992).
tero, Phys. Rev. Lett., 76, 34 (1996). 190. X. Xu and G.W. Kattawar, Appl. Opt., 33, 4835 (1994).
182. B. Mate, G. Tejeda and S. Montero, J. Chem. Phys., 108, 191. R.T. Brinkman, Ap. J., 154, 1087 (1968).
2676 (1998).
192. G. Kattawar, A.T. Young and T.J. Humphreys, Ap. J., 243,
183. I.F. Silvera, Metallic Hydrogen, in MetalInsulator Tran- 1049 (1981).
sitions Revisited, eds P.P. Edwards and C.N.R. Rao, Taylor
and Francis, London, 2142 (1995). 193. L. Wallace, Ap. J., 176, 249 (1972).
184. R.J. Hemley and N.W. Ashcroft, Phys. Today, August, 26 194. M.J.S. Belton, L. Wallace and M.J. Price, Ap. J., 184, L143
(1998). (1973).
185. J. Joiner, P.K. Bhartia, R.P. Cebula, E. Hilsenrath, R.D. 195. H. Fast, R. Poeckert and J.R. Auman, Ap. J., 187, 403
McPeters and H. Park, Appl. Opt., 34, 4513 (1995). (1974).
186. K.V. Chance and R.J.D. Spurr, Appl. Opt., 36, 5224 (1997). 196. W.D. Cochran and L.M. Trafton, Ap. J., 219, 756 (1978).
187. M. Vountas, V.V. Rozanov and J.P. Burrows, J. Quant. 197. A. Dalgarno and D.A. Williams, Mon. Not. R. Astron. Soc.,
Spectrosc. Radiat. Transfer, 60, 943 (1998). 124, 313 (1962).
188. C.E. Sioris and W.F.J. Evans, Appl. Opt., 38, 2706 (1999). 198. J.C. Brandt, Pub. Astron. Soc. Pac., 77, 212 (1965).
Near-infrared Spectroscopy of Synthetic and
Industrial Samples
Charles E. Miller
DuPont Engineering Technologies, Houston, TX, USA
1 INTRODUCTION different materials types is encountered, representative sam-

pling issues are often a concern, and speed and ease of
During the past several years, the great utility of the near- analysis are often top priorities.
infrared (NIR) region of the spectrum has been increasingly With regard to information content, it could be said
exploited to enable rapid and nondestructive analyses of that NIR spectroscopy enjoys an additional simplification
a wide array of synthetic and industrial materials.1 7 The advantage over mid-infrared spectroscopy, in that only a
attractiveness of NIR spectroscopy for such analyses is limited subset of functional groups can produce significant
mainly attributed to its great sampling ease, which enables vibrational bands in the NIR. For the most part, only those
the analyses of materials in gas, liquid, solid, slurry, pow- functional groups containing CH, NH, and OH bonds
der, fiber, film, plaque, and flake form with little or no can produce any vibrational overtone bands of appreciable
preparation. intensity in the NIR region. Vibrations from other bonds,
The absorptivities of the vibrational overtone and combi- such as CDC and CDO, can influence the intensity and
nation bands that are found in the NIR spectral region are, position of combination bands in the NIR, especially those
in general, much lower than the absorptivities of bands in at longer NIR wavelengths (2000 to 2500 nm). Although
the mid-infrared and visible spectral regions. As a result, many functional groups can influence the position and
NIR light can penetrate much deeper into a sample than shape of NIR absorption bands, the groups listed above
infrared or visible light, while still being capable of con- are most often the main contributors to the NIR spectrum
veying compositional information through weak vibrational of a material.
absorptions. The beneficial consequences of NIRs deeper Unfortunately, there are two additional properties of NIR
penetration depth are numerous: overtone and combination bands that are rather unappeal-
1. NIR absorption spectra of thick condensed-phase ing: a single functional group in a molecule can produce
samples (from millimeters to centimeters thick) can be many different combination and overtone bands (the com-
readily obtained. plication effect); and all of these bands tend to be rather
2. NIR transmission and reflection spectroscopy of broad and highly overlapped with one another (the overlap
highly scattering materials (i.e. materials with many effect), especially for condensed-phase materials.
refractive index interfaces) can be done. The particularly strong overlap effect for condensed-
3. Relatively large volumes of materials can be sampled phase materials is rather damaging for industrial analysts,
with NIR radiation, thus reducing concerns regarding who most often wish to utilize the sampling advantages of
representative sampling. NIR to analyze such materials.
Curiously, despite these difficulties, a large number of
These advantages of NIR analysis are of particular impor- successful NIR analyses of a wide array of materials have
tance in the chemical industry, where a wide array of been demonstrated over the past several years, and many
are undoubtedly being implemented in industrial research
John Wiley & Sons Ltd, 2002. laboratories, quality laboratories, and manufacturing plants.
The purpose of this article is to provide an overview of 2.4

the specific NIR spectral regions that are of use for the anal- 2.2
yses of industrially relevant materials. Due to the arguments 2.0
discussed above, it is rather difficult to provide defini- 1.8
1.6
tive assignments of individual NIR bands. Instead, attempts
1.4
Intensity
will be made to provide approximate assignments of 1.2
absorbances that are of interest for several common indus- 1.0
trial analyses. The sections in this article are divided accord- 0.8
ing to material class. Within each section, NIR spectra are 0.6
used to illustrate some of the more common absorbances. 0.4
0.2
Where appropriate, a separate discussion of the effects of
0.0
sample state (such as crystallinity, morphology, phase) on
the NIR spectrum is presented within each section. 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500
(a)
2 HYDROCARBONS
2.1 Saturated hydrocarbons
Intensity
Although NIR analyses of industrially relevant materi-
als were done as early as the 1930s and 1940s,8 11
one of the most highly visible applications that launched
NIR back into the industrial world in the 1980s was
the octane-number-in-gasoline application.12 Success in
this application depended heavily on the ability of NIR
to distinguish between the different saturated hydrocar-
700 800 900 1000 1100 1200 1300 1400 1500 1600
bon functional groups: methyl (CH3 ), methylene (CH2 )
(b) Wavelength / nm
and methine (>CH). Figure 1 shows the NIR spectra
of cyclohexane, which contains only methylene groups, Figure 1. NIR transmission spectra of n-heptane (. . . .) and
cyclohexane ( ) in the range (a) 16002500 nm and
and n-heptane, which contains both methylene and methyl (b) 7001600 nm. The discontinuity in the spectra at 1100 nm
groups. The characteristic methylene doublets at approx- is an instrumental artifact.
imately 17001800 nm and 23002350 nm are the most
prominent features in the cyclohexane spectrum. Other double bonds are not directly visible in the NIR region.
bands in the cyclohexane spectrum are caused by resonance However, bands from CH bonds that are adjacent to these
effects involving both CH stretching and bending vibra- double bonds are not only observable in the NIR spectrum,
tions, which are explained in detail elsewhere.13,14 The n- but also tend to be relatively strong, sharp, and well-
heptane spectrum looks much the same as the cyclohexane resolved.21 24 In general, CH bands from bonds adjacent
spectrum, except for some prominent shoulders on the to unsaturation appear at significantly shorter wavelengths
low-wavelength sides of the methylene doublets (denoted than those obtained from the saturated hydrocarbon groups.
by arrows in the figure). These absorbances are character- Figure 2, which compares the NIR spectra of ethane and
istic of the methyl group. Such subtle, yet highly repro- ethylene, illustrates the unique lower-wavelength position
ducible NIR absorbances can be utilized to discriminate and sharpness of such CH bands.
between different saturated hydrocarbons, thus enabling Not only can NIR be used to detect unsaturation, but
a wide range of quantitative and qualitative hydrocarbon it can also be used to discriminate between different
analyses.1,3,15 20 types of unsaturation in molecules, provided that there
is at least one CH group adjacent to the double bond.
Figure 3 shows the estimated spectra of cis-, trans- and
2.2 Unsaturated hydrocarbons 1,2-unsaturation in polybutadienes, in the first overtone
NIR region (15501850 nm). These spectra, which were
Unsaturated hydrocarbons also play an important role in estimated from actual spectra of different polybutadiene
the chemical industry, most notably in polymerization reac- polymers,25 indicate a strong contrast between the NIR
tions. Unfortunately, vibrational bands from carboncarbon spectroscopy of the 1,2-unsaturation group relative to the
Near-infrared Spectroscopy of Synthetic and Industrial Samples 3
Intensity
Intensity
0 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500
7500 6500 5500 4500
Wavenumber /cm 1 (a)
Figure 2. NIR transmission spectra of ethane ( ) and ethylene

(. . . .).
Intensity
Intensity
700 800 900 1000 1100 1200 1300 1400 1500 1600
(b) Wavelength / nm
1550 1600 1650 1700 1750 1800 1850 Figure 4. NIR transmission spectrum of benzene ( )
Wavelength / nm compared to the spectrum of cyclohexane (. . . .) in the range
(a) 16002500 nm, and (b) 7001600 nm. The discontinuity in
Figure 3. NIR spectra of cis- ( ), trans- (. . . .) and 1,2- the spectra at 1100 nm is an instrumental artifact.
(. . ) unsaturation in polybutadiene, estimated by applying
classical least squares analysis to the NIR spectra of various at approximately 24502500 nm. In addition, the CH
polybutadienes.25
stretching first and second overtone bands for aromatic
cis- and trans-unsaturated groups, and a more subtle con- CH groups (observed in the regions 16001800 nm
trast between the spectroscopy of the cis- and trans- groups. and 11001250 nm, respectively) are at distinctly lower
NIR bands from CH groups that are adjacent to car- wavelengths than the same bands for aliphatic CH groups.
boncarbon triple bonds are also rather sharp and distinct. The bands at 21002250 nm arise from the combination
The location of these bands depends on the nature of the of CH stretching and aromatic ring CC stretching modes.
adjacent functional group, but they can generally be found The CC ring stretching contribution to these bands makes
at approximately 1070, 1530, and 1880 nm. them rather sensitive to substitution patterns on the aromatic
ring. The dramatic effect of ring substitution patterns on
these bands is illustrated in Figure 5, in which the spectra
2.3 Aromatic hydrocarbons of benzene, toluene, and o-dichlorobenzene are compared.
The strong contrast in NIR absorbance bands from aliphatic

and aromatic hydrocarbons is illustrated in Figure 4, 2.4 Hydrocarbon polymers
which compares the NIR spectra of cyclohexane and
benzene. Perhaps the most distinct NIR absorbance bands Probably the most common hydrocarbon polymers of indus-
for aromatic hydrocarbons are the series of combination trial interest are polyethylene and polypropylene (Fig-
bands at 21002250 nm and the combination band ure 6).26 31 For polyethylene, which contains predominantly
Intensity
Intensity
2100 2150 2200 2250 2200 2300 2400 2500

Wavelength /nm Wavelength /nm
Figure 5. NIR transmission spectra of benzene ( ), toluene Figure 7. NIR spectra of polyethylene film before ( ) and
(. . .), and o-dichlorobenzene (. . ) in the range 21002250 nm. after (. . . .) uniaxial drawing and difference (. . ) between these
two spectra (magnified for clarity).
approximately 2260 nm. However, distinctive differences

in the polyethylene and polypropylene absorbance profiles
throughout the NIR spectrum are also observed.
Intensity
2.5 Sample state effects hydrocarbons
Although the CH group band assignments discussed above

provide a good general guide, it is important to note that
these bands are affected by the intermolecular environment
of the groups. Therefore, many of these bands are affected
by changes in sample phase, morphology, degree and type of
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
crystallinity, and molecular orientation (for polymers). This

Wavelength /nm is especially the case for the methylene combination bands
Figure 6. NIR spectra of polyethylene ( ) and polypropylene in the range 22502500 nm. For example, the NIR spectrum
(. . . .) from 1500 to 2500 nm. of polyethylene film before and after uniaxial stretching
(Figure 7) shows the subtle, yet reproducible, shifting and
methylene groups, the NIR spectrum is dominated by the changes in these bands as a function of molecular orienta-
characteristic methylene doublets at approximately 1700 tion of the polyethylene polymer chains.26,30 It is important
1800 nm and 23002350 nm. In addition, several bands to note that randomly polarized light was used to obtain the
are observed in the region 23502500 nm, which can be spectra in Figure 7, and therefore the effects observed indi-
attributed to resonance effects between the different CH cate changes in the methylene bands due to changes in the
stretching and bending modes in the polymer.13,14 local molecular environment, rather than the macroscopic
Although polyethylene contains predominantly methy- polymer chain environment! Very similar shifts in NIR CH
lene functional groups, there is also a small but significant bands in this region have been observed upon changes in
number of methyl and methine groups that can be present. the degree of crystallinity of polyethylene polymers.30
These groups reflect either ends or branches in the
polymer chains, which can greatly influence the physical
properties of the polymer. It has been shown that NIR can 3 KETONES, ESTERS, ETHERS, AND
be used to detect low levels of these functional groups in PEROXIDES
polyethylene, to enable discrimination between high- and
low-density polyethylenes, for example.30 With the exception of some weak overtone stretching bands
The main distinguishing feature of the polypropy- of CDO and carboxylate groups, carbonoxygen bonds
lene spectrum (Figure 6) is the methyl CH band at do not directly produce bands in the NIR spectral region.
Intensity
Intensity
1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500
(a) Wavelength /nm
Figure 9. NIR spectrum of di-isopropyl ether ( ) compared to
the spectrum of polypropylene (. . . .) in the range 15002500 nm.
bands to distinguish the ether functionality. However, the

presence of the ether group is conveyed through slight shifts
of the CH bands from where they would be without the
Intensity
ether functionality. Similarly, the OO bond vibration in

peroxides cannot be monitored directly in the NIR spec-
trum, but the peroxide functionality causes shifts in the band
positions of neighboring CH groups in the peroxide, thus
enabling one to distinguish a peroxide from its hydrocarbon
equivalent.
700 800 900 1000 1100 1200 1300 1400 1500 1600
(b) Wavelength /nm 3.1 Polymers with ketone, ester, and ether
Figure 8. NIR transmission spectra of acetone (. . ) and ethyl groups
acetate (. . . .) compared to the spectrum of cyclohexane ( )
in the range (a) 16002500 nm and (b) 7001600 nm. The dis- There are many different polymers containing ketone,
continuity in the spectra at 1100 nm is an instrumental artifact.
ester, and ether groups that are of great industrial inter-
est. These include polyether-polyols,32 37 polyesters,33,38,39
More often, they affect the NIR spectrum by shifting the and ethylene copolymers.40 Figure 10 shows the spectra
bands of neighboring CH, NH, and OH functional groups. of poly(ethylene glycol) (PEG) and ethylene vinyl acetate
Figure 8 shows the spectra of acetone and ethyl acetate, (EVA) copolymer, compared to the spectrum of cyclohex-
compared to the spectrum of cyclohexane. The distinct ane. When the spectrum of PEG is compared to the cyclo-
carbonyl overtone bands between 2100 and 2200 nm are hexane spectrum, it becomes clear that the ether linkages in
clearly visible in the acetone and ethyl acetate spectra. In the polymer chain repeat unit of PEG influence the position
addition, it appears that the methylene and methyl CH bands and shape of the peaks of the adjacent ethylene CH groups
are at slightly lower wavelength in these spectra than the (in the ranges 1650 to 1850 nm, and 2250 to 2500 nm). The
equivalent bands for a hydrocarbon material (see Figure 1, spectrum of EVA contains CH overtone peaks (in the range
the spectrum of n-heptane, for example), thus indicating that 1650 to 1800 nm) that are very similar, but not identical,
the ketone and acetate functionalities cause a distinct shift to those of cyclohexane. However, the EVA spectrum also
in the position of CH bands. In some special cases, much- contains the unique overtone and combination bands from
weaker third overtone CDO stretching bands in the range the acetate (COO) group in the range 20002200 nm.
of 1530 to 1600 nm can also be observed, although they It is important to note that the ether and acetate groups
are often difficult to find in the sea of CH combination in PEG and EVA, being good hydrogen-bond acceptors,
bands in that region. are rather effective at attracting moisture. The NIR region
Figure 9 shows the spectrum of di-isopropyl ether, com- 19002050 nm is often a very useful range for detecting
pared to the spectrum of polypropylene. Unlike the ketone moisture in such materials. Although the absorption bands
and ester spectra discussed above, there are no distinct CO in this range for PEG and EVA in Figure 10 are not very
small but significant shifts in the methylene bands at around

1720 nm and 22002500 nm, as well as subtle shifts in
the characteristic aromatic combination bands in the range
2100 to 2200 nm (although these are difficult to observe
in the plot). These shifts are caused by changes in the
Intensity
molecular environments of the ethylene and terephthalic

acid (aromatic) segments in the polymer, which alter the
vibrations of the CH and CDC groups in these segments.
The significant decrease in the absorption band at around
1910 nm is caused by a decrease in the absorbed water con-
tent of the polymer, which is coincident with the annealing
process.39,41
1100 1300 1500 1700 1900 2100 2300 2500
Wavelength /nm
Figure 10. NIR spectra of PEG ( ) and EVA (. . ) copoly-
4 ALCOHOLS AND ORGANIC ACIDS
mer compared to the spectrum of cyclohexane (. . . .) in the range
11002500 nm. These compound classes serve to introduce the OH group,
which, along with the NH group (discussed later), are the
strong to the eye, it is likely that there are some weak only NIR-active functional groups capable of being both
absorptions from the small amount of absorbed moisture hydrogen-bond donors and acceptors. This fact is important
that is in the polymers. for several reasons:
Hydrogen bonding has a profound influence on
3.2 Sample state effects ketones, esters, ethers, the vibrational spectrum, particularly in the NIR
and peroxides region.42 44
Changes in molecular environment (due to changes in
Like the case for hydrocarbon materials discussed above, sample state, temperature) are usually accompanied by
the spectra of ketones, esters, ethers, and peroxides can be changes in the hydrogen bonding of OH and NH groups
subtly, yet reproducibly, altered by changes in sample state. in a material.
A good example of such effects is observed by comparing Therefore, the NIR spectra of materials containing OH
the NIR diffuse reflection spectra of a polycrystalline poly- and NH groups are usually affected by changes in
mer, such as poly(ethylene terephthalate) (PET), before and sample state and temperature.
after annealing at 230 C (Figure 11).39 The annealing pro-
cess causes disorganized polymer chains to form ordered, Figure 12 shows the spectra of neat methanol, iso-
crystalline regions, thereby increasing the crystallinity of propanol, and cyclohexane. By far the most distinguishing
the material. In this case, changes in crystallinity cause features in the NIR spectra of alcohols and organic acids,
which can actually dominate the spectrum, are the combi-
nation and overtone OH stretching bands observed around
19002200 nm and 14001650 nm, respectively. Further-
more, it is important to note that the intensity, shape, and
location of these bands are greatly influenced by the hydro-
gen bonding state of the OH groups in the molecule, thus
Intensity
causing these distinguishing OH bands to be very sensi-

tive to sample state effects, such as dilution, phase, and
temperature. In the case of neat materials, the OH groups
are mostly hydrogen-bonded, and therefore produce rather
broad bands. In the case of neat methanol and isopropanol
(Figure 12), the OH bands are rather strong and broad,
which is indicative of the high degree of intermolecular
1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 hydrogen bonding that occurs in these neat alcohols.
Wavelength /nm For longer-chain hydrocarbon alcohols and fatty acids,
Figure 11. NIR reflectance spectra of PET before ( ) and which contain a lower fraction of OH groups, the spectra
after (- - - - ) annealing at 230 C.39 are not badly overwhelmed by the OH bands. Figure 13
two spectra are very similar. The only differences that

can be seen are very weak bands in the stearic acid
spectrum at approximately 14001550 nm, 2010 nm, and
21502250 nm, as well as what appears to be an upward
baseline shift in the stearic acid spectrum between 1800
Intensity
and 2250 nm. These weak features are most likely caused
by overtones and combinations of OH and COO vibrations
in the COOH acid group.
4.1 Sample state effects alcohols and organic

acids
1300 1500 1700 1900 2100 2300 2500
(a) The dramatic effects of hydrogen bonding and sample
state on NIR OH bands are illustrated in Figures 14 and
15. Figure 14 shows the NIR spectra of methanol at two
different temperatures. A small increase in temperature
Intensity
Intensity
700 800 900 1000 1100 1200 1300

(b) Wavelength /nm
Figure 12. NIR transmission spectra of methanol ( ) and
2-propanol (. . ) compared to the spectrum of cyclohexane
(. . . .) in the range (a) 13002500 nm and (b) 7001300 nm. The 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200
discontinuity in the spectra at 1100 nm is an instrumental artifact. Wavelength /nm
Figure 14. NIR transmission spectra of methanol at room tem-
perature ( ) and at approximately 30 C (. . . .) in the range
13002200 nm.
Intensity
Intensity
1100 1300 1500 1700 1900 2100 2300 2500

Wavelength /nm
Figure 13. NIR spectrum of stearic acid ( ) compared to the
spectrum of n-heptane (. . . .) in the range 11002500 nm. 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200
Wavelength /nm
shows the spectrum of stearic acid, which is a long-chain Figure 15. NIR transmission spectra of neat 2-propanol ( )
C-11 fatty acid, compared to the spectrum of a non- and 2-propanol diluted in carbon tetrachloride (. . . .), scaled for
acid equivalent (n-heptane). To a first approximation, these clarity.
causes small but significant shifts in the combination and 1440 nm

first overtone stretching OH bands in the methanol spec-
trum (at 20002100 nm and 14001600 nm, respectively)
to lower wavelengths. This shift in the band position is
indicative of the shift in the distribution of hydrogen-
bonded states of the methanol molecules towards more
Intensity
weakly hydrogen-bonded states. It is important to note that
these shifts were caused by only a few degrees Celsius
change in temperature, which underscores the extreme tem-
perature sensitivity of NIR OH bands. Furthermore, it is a
interesting to note that there were no obvious shifts in the
CH bands (16501800 nm) as a result of the change in
temperature. b
Figure 15 shows the spectra of isopropyl alcohol neat
and diluted to approximately 10% in carbon tetrachloride 1300 1500 1700
solvent. Because carbon tetrachloride is not expected to Wavelength /nm
contribute any significant bands in the NIR spectrum, all of
Figure 16. NIR spectra of crystalline sucrose (a) and amorphous
the bands in the spectrum of the diluted alcohol are from sucrose (b) in the range 13001700 nm.45
the alcohol. The most dramatic difference between these
two spectra is observed at approximately 1420 nm, where a alcohols are much lower than for shorter-chain analogs, the
sharp peak appears upon dilution of the alcohol. This peak sample state effects of the OH bands discussed above still
is most likely attributed to a free (non-hydrogen-bonded) apply. In the case of stearic acid (see Figure 13), the effect
OH group in the alcohol. As the alcohol is diluted, there are of the acid group is somewhat concealed in the spectrum as
fewer opportunities for intermolecular hydrogen bonding, a localized positive baseline offset between the CH overtone
and therefore the number of free OH groups increases. stretching bands at 17001800 nm and the CH combination
Other differences in these spectra in the OH stretching bands at 23002500 nm. Organic acid groups form partic-
regions (14001600 nm and 19502150 nm) are also the ularly strong hydrogen bonds with one another, and it is
result of the dramatic change in the distribution of the generally true that the width of OH stretching combination
hydrogen-bonded states of the alcohol molecules to weaker and overtone bands in the NIR spectrum increase with the
hydrogen-bonded and free states. In addition, as in the strength of the hydrogen bond of the OH group. Therefore,
case for the temperature effect discussed above, dilution this feature, which might look like a baseline offset at first,
does not appear to affect the CH bands of the alcohol (from is more likely the result of a very broad OH stretching peak
1650 to 1850 nm) to a large degree. from a very strong acidacid hydrogen bond that is present
For condensed phase materials, hydroxyl bands can also in the solid form of the organic acid.
be greatly affected by the crystal structure or phase of Interpretation of NIR OH bands can be further compli-
the material. This can be illustrated using the spectra of cated by the presence of moisture in the material. Any
crystalline and amorphous sucrose in the first combina- materials that contain OH groups can also attract ambient
tion OH stretching region (see Figure 16). The crystalline moisture, which is readily absorbed by the material through
material contains an ordered structure of molecules joined hydrogen bonding. Because the bands from any hydrogen-
together by specific types of hydrogen bonds, whereas bonded OH group are rather broad in general, it can be
the amorphous material contains a more random OH difficult to distinguish between bands from OH groups in
hydrogen-bonding pattern.45 As a result, the spectrum of the material and OH groups from absorbed moisture. In the
the crystalline material contains discrete OH bands from the OH combination band region (19002200 nm), bands from
different types of hydrogen bonds that are observed in the absorbed moisture generally occur at the lower wavelength
crystalline spectrum (including a particularly strong band at end of this range (around 19002000 nm). Therefore, the
1440 nm, which corresponds to a weakly hydrogen-bonded chance of being able to determine the OH content in the
OH group that is frozen into the crystalline lattice). In material in the presence of moisture interference increases
contrast, the spectrum of the amorphous material contains if the OH band of the material is close to the upper end
only a single broad OH band, indicative of the more random of this wavelength range (or 20002200 nm). Use of the
distribution of hydrogen bonding states in the material. first overtone OH stretching region (14001600 nm) for
Although the relative contributions of OH and COOH such analyses is somewhat more challenging, due to the
bands in the NIR spectra of longer-chain fatty acids and additional presence of CH stretching combination bands.
4.2 OH-containing polymers 5 NITROGEN-CONTAINING

COMPOUNDS
There is a wide variety of OH-containing polymers that
are of industrial interest: from low-OH-content PEGs to For nitrogen-containing compounds, the nature of the NIR
ethylenevinyl alcohol (EVOH) copolymers to higher-OH- spectrum greatly depends on whether an NH bond exists
containing celluloses, gelatins, and poly(vinyl alcohol)s on the molecule. Primary and secondary amines, which
(PVAs).9,32 36,40,46,47 As in all cases, the OH bands in contain NH bonds, produce distinctive NH stretch/bend
such materials are greatly affected by the hydrogen-bonding combination bands in the range of 1900 to 2100 nm and
states of the OH groups. Furthermore, it is generally the NH stretching overtone bands in the range of 1500 to
case that the OH combination bands from the polymer are 1650 nm. The NH2 group in primary amines typically pro-
at significantly longer wavelength (20002200 nm) than the duces a pair of strong combination bands at approximately
OH bands from absorbed moisture (19002000 nm), thus 19502050 nm and 20502200 nm, along with a strong NH
enabling NIR to distinguish between these two types of OH stretching overtone band at approximately 15001600 nm.
group. In contrast, the NH-group in secondary amines typically
Figure 17 shows the spectrum of an EVOH copoly- produces a somewhat weaker series of combination bands
mer, compared to the spectrum of polyethylene. Note the in the range of 20502200 nm, a strong NH stretching over-
unique band for the EVOH copolymer at 20002200 nm, tone band at 15001600 nm, and a weaker band in the
which can be assigned to the OH groups in the polymer. region of 16001650 nm.3,4,48 51
In this case, the polymer OH band appears to domi- Primary and secondary amide functionalities also pro-
nate the spectrum in the OH region, and it is difficult duce unique NIR bands, which are a result of over-
to detect any moisture-based OH absorption bands in the tones and combinations of unique amide NH and CDO
bond vibrations.4,11,52,53 Primary amides are characterized
19002000 nm region. In the case of PEG (see Figure 10),
by a strong combination band in the region of 1930 to
relatively fewer OH groups are present on the poly-
2000 nm, some weaker combination bands between 2000
mer chains, and the amount of these groups (commonly
and 2250 nm, and a band in the range of 14501550 nm,
referred to as the hydroxyl number) greatly impacts the
which is most likely an NH stretching overtone band.
physical properties of the material.32 36 It is quite possi-
Similarly, secondary amides are distinguished by several
ble that the two small but significant bands in the PEG
combination bands in the range of 19902250 nm and over-
spectrum that are observed in the OH combination band
tone bands in the range of 13501550 nm. The spectrum
region (at approximately 1920 and 2060 nm, see Figure 10)
of polyamide-6 (or Nylon-6), which contains secondary
are from absorbed moisture and polymer OH groups,
amide repeat units, is shown in Figure 18.54 Note the series
respectively. of bands in the region of 1950 to 2250 nm, which are
most likely combination bands involving NH stretching,
Intensity
Intensity
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500
Figure 17. NIR spectrum of EVOH (. . . .) compared to the Figure 18. NIR spectrum of polyamide-6 (Nylon-6) ( )
spectrum of polyethylene ( ) in the range 15002500 nm, compared to the spectrum of polyethylene (. . . .) in the range
scaled for clarity. 15002500 nm, scaled for clarity.54
NH bending, and CDO stretching vibrations in the CONH

secondary amide group. The weaker band at approximately
1950 nm is most likely due to the presence of absorbed
moisture in the polymer.
Unlike their primary and secondary analogs, tertiary
Intensity
amines and amides do not produce any unique bands in
the NIR spectrum. However, the CH bands from the groups
adjacent to the amine or amide functionality are most likely
shifted from where they would be located for the pure
hydrocarbon analog.
Amino acids, which are of great interest for bioprocess-
ing, produce unique NIR absorptions from both the amine
(NH2 ) group and acid (COOH) group in the regions 1100 1300 1500 1700 1900 2100 2300
14001650 nm and 19502200 nm.55,56 However, the spe- Wavelength /nm
cific nature of these absorptions depends greatly on the Figure 20. NIR spectra of a polyurethane polymer ( ) and
sample state and whether the molecule is in the neutral urethane model compound (- - - - ) compared to the spectrum of
or zwitterionic form. In the case of crystalline glycine cyclohexane (. . . .) scaled for clarity.37
(Figure 19), the molecules are in the zwitterionic form (i.e.
NH3 C and COO groups are present) and strong hydro- urethane NIR bands in the range of 19002200 nm and
gen bonds between amine NH groups and carboxylate oxy- 14001550 nm for the model urethane compound, which
contains only the urethane functional group and some
gen atoms are present.55 As a result, broad NIR absorbances
CH functional groups. Similar, but not identical, bands
in the NH stretching overtone and combination regions
in these regions are observed in the polymer spectrum. It
(12501300 nm, 15501650 nm, and 20002200 nm) are
should be noted that the polyurethane polymer represented
observed.
in Figure 20 contains both urea and urethane functional
Urea and urethane functional groups, although distinct,
groups. As a result, some bands that are observed in the NH
produce NIR bands in the same regions as the primary
stretching combination and overtone regions of the polymer
and secondary amide groups. These groups are of particular
spectrum that are not seen in the urethane model compound
interest in the polyurethanes industry, because polyurethane
spectrum could be assigned to the urea groups. However,
polymer chains are typically linked together with both ure-
it is also likely that sample state effects play a role in these
thane and urea groups. These groups produce combination differences as well, and these will be discussed later.
bands in the 19002200 nm region and NH overtone bands Nitriles, isocyanates, and azo compounds are very
in the 14001550 nm region. Figure 20 shows the spectra common materials that can produce rather distinct bands
of a polyurethane polymer, along with a low-molecular- in the NIR spectrum. The nitrile group produces a
weight model urethane compound.37,57 60 Note the unique unique first overtone CN stretching band at approximately
18801950 nm (see the spectrum of acetonitrile in Fig-
ure 21). Likewise, the isocyanate group produces a weak
2147 but distinct NDCDO stretching overtone band in the range
18501950 nm.37 Azo compounds can also produce unique
NIR bands in the ranges 14001550 nm and 19002200 if
there are NH bonds adjacent to the NDN double bond.
Intensity
2050
5.1 Sample state effects nitrogen-containing

a
compounds
1600
b 1270 Like OH groups, NH groups can interact with one another
through hydrogen bonding. As a result, NH bands in the
NIR spectrum can also be highly influenced by sample
1100 1800 2500 state and temperature.4,11,61 65 One particularly illustra-
Wavelength /nm tive example of such sample states effects on NH bands
Figure 19. NIR reflectance spectra of (a) a-glycine and is the comparison of the NIR spectra of two different
(b) g-glycine in the range 11002500 nm.55 crystalline phases of glycine (Figure 19).55 These two
Intensity
Intensity
1100 1300 1500 1700 1900 2100 2300 2500 1500 1600 1700 1800 1900
Figure 21. NIR spectrum of acetonitrile ( ) compared to the Figure 22. NIR spectrum of o-dichlorobenzene ( ) compared
spectrum of cyclohexane (. . . .) in the range 11002500 nm. to the spectrum of benzene (. . . .) in the range 15001900 nm.
crystalline forms involve a different set of discrete hydrogen in the NIR spectrum. However, the high electronegativity
bonds between the NH groups and the carboxylate (COO) of the halogen atoms can create strong secondary effects
groups in the zwitterionic form of the glycine molecule. on the NIR spectroscopy of neighboring CH, NH, or
The differences between these two hydrogen-bonding envi- OH groups. Figure 22 shows the NIR spectra of benzene
ronments primarily affects the spectra in the combination and o-dichlorobenzene in the first overtone CH stretching
band region (19002250 nm), although there are also sig- region. Note that the aromatic CH stretching overtone
nificant changes in the inflection of the spectrum in the bands for the chlorinated compound are, in general, at
first and second overtone NH stretching regions (around significantly lower wavelengths than the same bands for the
15501650 nm and 12501350 nm, respectively). These non-chlorinated compound. Similar effects can be observed
differences are due to the presence of different sets of rela- in the NIR spectra of other industrially relevant materials,
tively broad NH bands from strongly hydrogen-bonded NH such as poly(vinyl chloride) (PVC) versus polyethylene.
groups, which are somewhat concealed as subtle changes In addition, the low NIR-activity of CX bonds makes
in inflection. In fact, the only NH band that is distin- some halogenated materials very useful for fundamental
guishable in either of the spectra is the combination band NIR materials studies. For example, materials such as car-
at 2147 nm for the a-glycine phase, which is most likely bon tetrachloride and chloroform are very useful NIR sol-
attributed to a weak NH hydrogen bond that is present only vents, because they either have no appreciable bands in the
in the a-phase. NIR region or many useful NIR transmission windows.66
Sample state effects can also be observed in the spectra Similarly, per-halogenated hydrocarbons, such as teflon
of the polyurethane polymer and urethane model compound polymer and freon gases, can be used as solid or gas
(see Figure 20). Although the urethane groups in the model phase diluents for NIR studies.
compounds are chemically identical to those in the polymer,
the urethane bands in the NH/COO stretching combination
band region (19002200 nm) are quite different in terms 7 AQUEOUS SYSTEMS
of both shape and location. This difference is most likely
caused by the different hydrogen-bonding environments of From an NIR spectroscopists perspective, aqueous solvents
the urethane NH groups in the two states. In fact, it has can be considered an even stronger case of alcohols,
been shown that NIR bands in the 19002200 nm range because their spectra are even more dominated by OH
can be used to detect subtle phase-segregation effects in bands. In fact, because the OH groups in aqueous sys-
polyurethane polymers.60 tems are in the bulk (highly hydrogen-bonded) state,
these OH bands are also very broad. The spectrum of
water (see Figure 23) contains several broad, yet distinctive
6 HALOGENATED COMPOUNDS bands at 19002100 nm, 14001550 nm, 11001250 nm,
and 9001050 nm.
Because the fundamental vibrational frequencies of CX Spectral differences between different aqueous solvents
(X D halogen) bonds are rather low compared to CH, NH, typically comprise shifts in the shapes and positions of these
and OH bonds, they do not produce any strong bands broad OH bands. The spectra of water, concentrated sodium
Hydrogen-bonding patterns in aqueous solvents can also

be altered through temperature changes. In fact, it was
mentioned earlier that OH bands can be very sensitive to
temperature (see the spectra of methanol in Figure 14). As
a result, these broad OH bands for aqueous systems are not
Absorbance
only altered by changes in solute and solute concentration,

but also by changes in temperature.69,71,72
All of the factors discussed above present several chal-
lenges to those doing NIR analyses of solutes in aqueous
media:
the extreme dominance of the NIR OH bands from
water;
the sensitivity of these OH bands to solute and temper-
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
ature changes;
(a) the relatively low concentration of the analyte.
Despite these difficulties, the NIR region has been shown

to be useful for the analyses of several aqueous systems,
including bioprocesses56,73 75 and bodily fluids.76 78
An interesting exception to the NIR water/OH dominance
rule discussed above is the analysis of some highly conju-
Absorbance
gated hydrocarbons, such as hemoglobin.76 78 In this case,

hemoglobin produces bands from electronic transitions,
rather than overtone and combination vibrational transi-
tions, in the NIR spectral region. Because these electronic
bands are rather strong, and are altered by the oxidation
state of the iron atom in the hemoglobin, it is possible to use
NIR spectroscopy to determine the degree of oxygenation
700 800 900 1000 1100 1200 1300 of hemoglobin in blood.
(b) Wavelength /nm
Figure 23. NIR spectra of water ( ), phosphoric acid
(86%) (. . . .) and sodium hydroxide (50% solution) (. . ) 8 INORGANIC MATERIALS
at room temperature in the range (a) 13002500 nm and
(b) 7001300 nm. The discontinuity in the spectra at 1100 nm At first thought, the NIR region should not provide much
is an instrumental artifact.
useful information for inorganic materials, because they
hydroxide, and concentrated phosphoric acid at ambient usually do not contain any CH, NH, or OH bonds. However,
temperature are compared in Figure 23. In this case, the there are several factors that could result in useful NIR
different hydrogen-bonding environments of the OH groups analyses for inorganic materials:
in the three media are reflected by the dramatic differences weak overtone bands from CDO groups and bonds of
in the positions and shapes of the OH bands in the NIR hydrogen with low-atomic-weight atoms;
spectrum. It is interesting to note that these OH band the presence of electronic, instead of vibrational, tran-
effects are not confined to the typical OH overtone and sitions in the NIR region;
combination regions (14001600 nm and 19002150 nm, secondary effects of solutes on the NIR spectra of
respectively), but are somewhat smeared over the entire solvents containing NH or OH bands;
spectral range from 1400 to 2500 nm. This probably reflects the presence of OH groups in the crystal structures of
the fact that the bands that correspond to strongly hydrogen- some minerals;
bonded OH groups in these media are extremely broad, and the presence of surface OH, NH or CH groups on many
appear in the spectrum as baseline offsets and inflections. common inorganic substrates.
Similar changes in OH band patterns can be observed by
adding various organic or inorganic solutes,67 70 which Inorganic materials that contain CDO groups, such as
change the hydrogen-bonding patterns of the OH groups carbonates, and NDO groups (such as nitrates) can pro-
in the solution. duce weak overtone bands at the longer NIR wavelengths
(19002500 nm). However, these bands are often masked

by much stronger CH, NH, or OH bands that are in the same
part of the spectrum. In contrast, inorganic compounds with
light-atom hydrogen bonds, such as HF and LiH, produce
rather strong combination and overtone bands in the NIR
Intensity
spectrum.
Probably the most common example of inorganic mate-
rials with electronic transitions in the NIR region are the
rare earth oxides of holmium, dysprosium, and hafnium,
which are used in the NIST (National Institute of Stan-
dards and Technology) NIR wavelength standard SRM1920
(Figure 24).79 The relative sharpness and temperature sta-
bility of these electronic NIR bands make these materials 1600 1700 1800 1900 2000 2100 2200
rather attractive as NIR wavelength standards. Wavelength /nm
The effects of inorganic solutes on the NIR spectrum of Figure 25. NIR reflectance spectra of aluminum oxide powder
water have been exploited for many years to characterize before ( ) and after (. . . .) coating with POMA.81
the hydration of inorganic ions in aqueous solutions.67,68 As
discussed in an earlier section (see Figure 23), the NIR OH chemistry. Figure 25 shows the NIR reflectance spectra of
bands from water are very sensitive to the hydration state, aluminum oxide powder before and after treatment with a
or structure of ion/OH hydrogen bonds in the solution. As solution of poly(octadecyl methacrylate) (POMA).81 This
a result, it is possible to use NIR spectroscopy to assess the treatment causes a thin film of POMA to adsorb to the sur-
thermodynamics of hydration for inorganic ions. faces of the alumina particles. In this case, the uncoated
Some naturally occurring minerals, such as those based alumina powder produces significant OH bands in the
on metal hydrates, actually contain OH groups in their crys- 19002200 nm region, which can be assigned to surface
talline structure. In such cases, it is possible to use NIR OH hydroxyl groups and adsorbed moisture. More specifically,
stretching combination and overtone bands in the regions the sharp band at approximately 1882 nm, due to its sharp-
20802500 nm and 13501600 nm, respectively, to dis- ness, location and intensity, is most likely caused by free
criminate between different minerals. Such an application (non-hydrogen-bonded) OH groups on the alumina surface.
was recently demonstrated for several different hydrates of Once the organic coating is added, characteristic hydro-
aluminum: gibbsite, boehmite, and diaspore.80 carbon CH bands in the range of 17001800 nm appear
Even if bulk inorganics do not have vibrational or elec- and the free OH band at 1882 nm virtually disappears,
tronic bands in the NIR region, many will have hydroxyl thus indicating that the formerly free OH surface groups
groups and adsorbed moisture on the surface, both of which are now hydrogen-bonded to the molecules of the organic
can be strong NIR absorbers. As a result, it is possible coating.
in some cases to use the NIR spectrum to study surface
9 GASES
NIR spectroscopy can also be used to analyze materials

in the gas phase, provided that the material contains CH,
NH, OH, or CDO functional groups. In this case, the lower
Intensity
absorptivities of NIR bands versus mid-infrared bands are

not necessarily an advantage, because molecular densities
in the gas phase are much less than molecular densities in
condensed phases. Nonetheless, NIR can be very useful
for the analyses of some gases,82,83 particularly in the
case of compressed gases, where the molecular density is
sufficiently high.
1000 1200 1400 1600 1800 2000 2200 2400 2600 Due to the ability of gas molecules to rotate, the vibra-
Wavelength /nm tional spectra of gases can be very different from those of
Figure 24. NIR reflectance spectrum of NIST SRM1920, which condensed phases. This is true not only for the fundamen-
is a mixture of rare earth oxides, in the range 11002500 nm.79 tal vibrational bands observed in the mid-infrared spectrum,
ACKNOWLEDGMENTS
The author gratefully acknowledges Ann Brearley and

Randy Hernandez (DuPont Central Research and Devel-
opment, Wilmington, DE), for providing access to an NIR
Intensity
spectrometer that was used to collect some of the spectra

used in this article.
EVA Ethylene Vinyl Acetate

1700 1800 1900 2000 2100 2200
EVOH EthyleneVinyl Alcohol
Wavelength /nm NIST National Institute of Standards and Technology
Figure 26. NIR spectra of water vapor ( ) compared to the PEG Poly(ethylene glycol)
spectrum of liquid water (. . . .) in the range 17002200 nm, PET Poly(ethylene terephthalate)
scaled for clarity.
POMA Poly(octadecyl methacrylate)
PVA Poly(vinyl alcohol)
but also for the overtone and combination bands observed
PVC Poly(vinyl chloride)
in the NIR spectrum. The effect of molecular rotation is par-
ticularly strong in the case of small molecules, for which
the separation between rotational energy states is relatively REFERENCES
large.
Perhaps the most commonly encountered gas of industrial
1. O.H. Wheeler, Chem. Rev., 59, 629 (1959).
interest that is also NIR-active is water vapor.66 Figure 26
2. R.F. Goddu, Adv. Anal. Chem. Instrum., 1, 347 (1960).
shows the NIR spectrum of water vapor and liquid water
in the OH combination band region. The numerous sharp 3. K.B. Whetzel, Appl. Spectrosc. Rev., 2, 1 (1968).
bands that are unique to the water vapor spectrum are 4. L.G. Weyer, Appl. Spectrosc. Rev., 21, 1 (1983).
caused by transitions between rotational energy states. In 5. E. Stark, K. Luchter and M. Margoshes, Appl. Spectrosc.
fact, these rotational lines are often too closely spaced to Rev., 22, 335 (1986).
be resolved by most commercial NIR instruments. Further- 6. C.E. Miller, Appl. Spectrosc. Rev., 26(4), 275 (1991).
more, the entire envelope of these rotational OH bands 7. J.J. Workman, Appl. Spectrosc. Rev., 34, 1 (1999).
for water vapor is shifted to significantly lower wavelength
8. U. Liddell and C. Kasper, J. Res. Natl. Bur. Stand., 11, 599
than the broad OH band for liquid water. This effect is (1933).
due to the fact that the gas phase water molecules cannot
9. J.W. Ellis and J. Bath, J. Am. Chem. Soc., 62, 2859 (1940).
hydrogen bond to one another, which causes the energy of
10. L. Glatt and J.W. Ellis, J. Chem. Phys., 15, 884 (1947).
the OH combination vibration to increase (corresponding
to a decrease in the wavelength of the combination band). 11. L. Glatt and J.W. Ellis, J. Chem. Phys., 16, 551 (1948).
NIR bands with similar rotational structure are observed 12. J.J. Kelly, C.H. Barlow, T.M. Jinguji and J.B. Callis, Anal.
for other industrially common small-molecule gases, such Chem., 61, 313 (1989).
as HCl66 and HF, although the positions of the bands for 13. J.L. Duncan, Spectrochim. Acta., 47A, 1 (1991).
these gases are different. 14. A.S. Bonanno, J.M. Ollinger and P.R. Griffiths, in Near-
As the size of the gas molecule increases, these sharp Infrared Spectroscopy, Bridging the Gap Between Data Anal-
rotational bands start to blend into one another, producing ysis and NIR Applications, ed. K.I. Hildrum, Ellis-Horwood,
New York, 1928 (1992).
the more common single-band profile. As a result, the
NIR spectra of large-molecule gases (such as most volatile 15. F.W. Rose, J. Res. Natl. Bur. Stand., 19, 143 (1937).
organic compounds) have bands that appear more like the 16. F.W. Rose, J. Res. Natl. Bur. Stand., 20, 129 (1938).
bands of condensed phase materials. However, these bands 17. R.R. Hibbard and A.P. Cleaves, Anal. Chem., 21(4), 486
are often significantly shifted from the bands observed for (1949).
the condensed phases of such materials, due to the different 18. A. Evans, R.R. Hibbard and A.S. Powell, Anal. Chem., 23,
molecular interaction (or hydrogen-bonding) environments 1604 (1951).
in the gas and condensed phases. 19. C. Tosi and A. Pinto, Spectrochim. Acta, 28A, 585 (1972).
20. O.C. Mullins, N.B. Joshi, H. Groenzin, T. Daigle, C. Crowell, 49. R.A. Russell and H.W. Thompson, Proc. R. Soc. London,
M.T. Joseph and A. Jamaluddin, Appl. Spectrosc., 54(4), 624 A234, 318 (1956).
(2000). 50. K. Whetsel, W.E. Roberson and M.W. Krell, Anal. Chem.,
21. M. Buback, Z. Naturforsch., 39A, 399 (1984). 30(10), 1594 (1958).
22. R.F. Goddu, Anal. Chem., 29(12), 1790 (1957). 51. K.B. Whetsel, W.E. Roberson and M.W. Krell, Anal. Chem.,
23. R.T. Holman and P.R. Edmondson, Anal. Chem., 28(10), 30(10), 1598 (1958).
1533 (1956). 52. S.E. Krikorian and M. Mahpour, Spectrochim. Acta, 29A,
24. A.J. Fenton, Jr and R.O. Crisler, Anal. Chem., 36(12), 620 1233 (1973).
(1959). 53. K.T. Hecht and D.L. Wood, Proc. R. Soc. (London), 235, 174
25. C.E. Miller, B.E. Eichinger, T.W. Gurley and J.G. Hermiller, (1956).
Anal. Chem., 62, 1778 (1990). 54. C.E. Miller, S.A. Svendsen and T. Ns, Appl. Spectrosc.,
26. L. Glatt and J.W. Ellis, J. Chem. Phys., 15, 884 (1947). 47(3), 346 (1993).
27. A. Evans, R.R. Hibbard and A.S. Powell, Anal. Chem., 23, 55. C.E. Miller and D.E. Honigs, Spectroscopy, 4, 44 (1989).
1604 (1951). 56. S. Schmidt, M. Kircher, J. Kasala and J. Locaj, Bioprocess
28. L. Glatt and J.W. Ellis, J. Chem. Phys., 19, 449 (1951). Eng., 19(1), 67 (1998).
29. C.E. Miller, Appl. Spectrosc., 43, 1435 (1989). 57. E.W. Crandall and A.N. Jagtap, J. Appl. Polym. Sci., 21, 449
(1977).
30. C.E. Miller, Appl. Spectrosc., 47(2), 222 (1993).
58. E.W. Crandall, E.L. Johnson and C.H. Smith, J. Appl. Polym.
31. C. Tosi, Makromol. Chem., 112, 303 (1968). Sci., 19, 897 (1975).
32. C. Jones and J.A. Brown, Adv. Instrum., 38, 429 (1983). 59. C.E. Miller, P.G. Edelman and B.D. Ratner, Appl. Spectrosc.,
33. C.L. Hilton, Anal. Chem., 31, 1610 (1959). 44, 576 (1990).
34. P.A. Turley and A. Pietrantonio, J. Cell. Plast., 20, 274 60. C.E. Miller, P.G. Edelman, B.D. Ratner and B.E. Eichinger,
(1984). Appl. Spectrosc., 44, 581 (1990).
35. F. DeThomas, S. Monfre and P. Brimmer, NIR News, 1, 14 61. J.E. Sinsheimer and A.M. Keuhnelian, Anal. Chem., 46(1), 89
(1990). (1974).
36. J.M. Chalmers and W.C. Campbell, in Proceedings of Intl. 62. L.A. Strait and M.K. Hrenoff, Spectrosc. Lett., 8(4), 165
Conference on Analytical 27. Applied Spectroscopy, ed. (1975).
C.S. Creaser, Royal Society of Chemistry, London, 167 63. K.B. Whetzel, W.E. Roberson and M.W. Krell, Anal. Chem.,
(1988). 32(10), 1281 (1960).
37. C.E. Miller and B.E. Eichinger, J. Appl. Polym. Sci., 42, 2169 64. S.E. Krikorian, Spectrochim. Acta, 37A(9), 745 (1981).
(1991).
65. J.W. Ellis and J. Bath, J. Chem. Phys., 6, 723 (1938).
38. R.B. Roy and C. Kradjel, J. Polym. Sci., Polym. Chem. Ed.,
66. I. Murray and P.C. Williams, Chemical Principles of Near-
26, 1733 (1988).
infrared Technology, in Near-infrared Technology in
39. C.E. Miller and B.E. Eichinger, Appl. Spectrosc., 44, 496 the Agricultural and Food Industries, eds P. Williams
(1990). and K. Norris, American Association of Cereal Chemists,
40. A. Khettry and M.G. Hansen, Polym. Eng. Sci., 36(9), 1232 St. Paul, MN, 1734 (1987).
(1996). 67. K. Buijs and G.R. Choppin, J. Chem. Phys., 39(8), 2035
41. M. Fukuda, H. Kawai, N. Yagi, O. Kimura and T. Ohta, (1963).
Polymer, 31, 295 (1990). 68. G.R. Choppin and K. Buijs, J. Chem. Phys., 39(8), 2042
42. G. Pimental and A.C. McClellan, The Hydrogen Bond, (1963).
W.H. Freeman, San Francisco (1960). 69. O.D. Bonner and G.B. Woolsey, J. Chem. Phys., 72, 899
43. P. Schuster, G. Zundel and C. Sandorfy, The Hydrogen (1968).
Bond, II. Structure and Spectroscopy, North-Holland, Ams- 70. C. Jolicoeur, N.D. The and A. Cabana, Can. J. Chem., 49(12),
terdam (1976). 2008 (1971).
44. H.W. Siesler and K. Holland-Moritz, Infrared and Raman 71. W.C. McCabe, S. Subramanian and H.F. Fisher, J. Phys.
Spectroscopy of Polymers, Dekker, New York (1980). Chem., 74(25), 4360 (1970).
45. A.M.C. Davies and C.E. Miller, Appl. Spectrosc., 42(4), 703 72. V. Fornes and J. Chaussidon, J. Chem. Phys., 68(10), 4667
(1988). (1978).
46. J.A. Mitchell, C.D. Bockman, Jr and A.V. Lee, Anal. Chem., 73. K.S.Y. Yeung, M. Hoare, N.F. Thornhill, T. Williams and
29, 499 (1957). J.D. Vaghiani, Biotechnol. Bioeng., 63(6), 684 (1999).
47. K.H. Bassett, C.Y. Liang and R.H. Marchessault, J. Polym. 74. S. Vaidyanathan, B. McNeil and G. Macaloney, Analyst,
Sci. A, 1, 1687 (1963). 124(2), 157 (1999).
48. F.H. Lohman and W.E. Norteman, Anal. Chem., 35(6), 707 75. A. Hagman and P. Sivertsson, Process Control Qual., 11(2),
(1963). 125 (1998).
76. J.W. Hall and A. Pollard, in Leaping Ahead with Near- 79. V.R. Weidner, P.Y. Barnes and K.L. Eckerle, NBS J. Res.
infrared Spectroscopy, ed. G.D. Batten, Royal Australian Natl Bureau of Standards, 243253 (1986).
Chemical Institute, Melbourne, 421430 (1995). 80. H. Ruan, R.L. Frost and J.T. Kloprogge, Appl. Spectrosc.,
77. K. Norris and J.T. Kuenstner, in Leaping Ahead with Near- 55(2), 190 (2001).
infrared Spectroscopy, ed. G.D. Batten, Royal Australian
81. C.E. Miller and T.-K. Yin, J. Mater. Sci. Lett., 8, 467 (1989).
Chemical Institute, Melbourne, 431436 (1995).
82. P. Kaspersen and I. Linnerud, Proc. Annu. ISA Anal. Div.
78. I. Oda, Y. Ito, H. Eda, T. Tamura, M. Takada, R. Abumi,
Symp., 33, 73 (2000).
K. Nagai, H. Nakagawa and M. Tamura, in Proceedings
SPIE- Int. Soc. Opt. Eng., 1431, Proc. Time-Resolved Spec- 83. P. Werle, R. Muecke, F. DAmato and T. Lancia, Appl. Phys.
trosc. Imaging Tissues (1991). B: Lasers Opt., B67(3), 307 (1998).
Near-infrared Spectroscopy of Biomaterials
W. Fred McClure and Donald L. Stanfield
North Carolina State University, Raleigh, NC, USA
1 INTRODUCTION OH stretching modes.6 8 Those were the dark ages for

NIR. Only a faithful few hung on to the vision that the
1.1 Historical perspective NIR spectrum contained useful analytical information.9 13
These workers provided a sturdy foundation on which Karl
Although Sir Isaac Newton was the first to use a prism in Norris and other pioneers would later build modern NIR
1666 to demonstrate that white light was made up of var- technology.
ious colors,1 it was not until 1800 that William Herschel In the late 1950s and early 1960s, Karl Norris of the
first observed the heating effect of radiation beyond the United States Department of Agriculture (USDA) Labora-
red portion of the visible spectrum.2 4 It seems plausible tories in Beltsville, MD, began an in-depth study of the
that Herschels experiments were, in fact, the discovery of optical properties of dense light-scattering materials. It was
the near-infrared (NIR) portion of the electromagnetic spec- a natural transition from his earlier work with biological
trum. Langley invented the bolometer in 1881, a device that samples.14,15 By 1975, research centers with NIR scan-
extended spectral measurements deep into the infrared (IR) ning technology were established in the USA at USDA,
region. Yet, the potential of the NIR region for analytical Beltsville, MD, North Carolina State University, Raleigh,
measurements has been recognized for fewer than 80 years.5 NC, Russel USDA Research Center, Athens, GA and Penn-
Modern chemometrics (or statistics in chemistry), the one sylvania State University , College Station, PA. In the early
tool that has enhanced NIR appeal, has grown to maturity 1980s, the popularity of NIR was drawing everyone into
only since 1980. this exciting field. Today, there are more than 30 manu-
World War II stimulated the commercial development of facturers supporting the NIR community with offerings of
mid-infrared (MIR) technology for the purpose of inves- hardware (even precalibrated instruments) and software.16
tigating the properties and origin of rubber, fuels and
lubricants. The first single-beam MIR spectrometer was
introduced in the early 1940s, and 1947 saw the commercial
1.2 Biological samples
introduction of the first double-beam instrument. Because
Generally, a biological sample is any sample that has
of the difficulties of interpreting NIR spectra, analytical
its source in living organisms or may be constituted from
spectroscopists avoided the NIR region until the 1950s,
plants or animals (including humans). The term is used here
when Perkin-Elmer combined visible and NIR spectrome-
in its broadest sense to include agricultural, food and animal
try into one commercial instrument. Thus, NIR technology
applications of NIR spectrometry.
was further isolated from the MIR community, despite the The spectral characteristics of plant and animal materials
common origin of absorption bands in both spectral regions, are dependent on their chemical makeup. Organic, by def-
with the MIR region containing bands due to fundamental inition, these samples embody atoms of carbon, hydrogen,
vibrational transitions and the NIR region containing largely oxygen, sulfur, phosphorus and nitrogen along with less
overtones and combination bands involving CH, NH and pronounced amounts of other elements. Macroscopically,
biological samples are fashioned from proteins, fibrous car-
John Wiley & Sons Ltd, 2002. bohydrates (cellulose, moisture, minerals and vitamins) and
non-fibrous carbohydrates (sugars, starches and fats or oils). making band assignments with derivative spectra tenuous.20
All are complicated molecules that are composed of much In addition, the wavelength accuracy of dispersion-based
simpler compounds such as amino acids, monosaccharide instruments from one instrument to another could account
and disaccharide sugars, fatty acids, alcohols and glycerol. for 12 nm differences in peak positions when compared
The non-fibrous portions may be arranged in an intricate with the literature. All band assignments are given in units
pattern within a fibrous matrix with different levels of of nanometers.
hydration to create a wide range of properties properties The discussion is organized according to biological sam-
that are often elusive and unpredictable.17 ple type. Water, oil, protein and starch, major components
found in plant materials (components which occupied the
interest of NIR researchers in the early years of devel-
1.3 Qualitative NIR spectroscopy opment) are discussed first. A general discussion of these
Qualitative analysis of the spectra of biological samples components is followed by a discussion of plant materials
allows one to do at least four things: (1) identify or classify (grain, tea, coffee, grass, tobacco and forages), i.e. materials
unknown substances, (2) compare the relative intensities of already studied by numerous researchers. Spectral charac-
known absorption bands, (3) detect the presence of impuri- teristics of food/drink (cereals, meats, additives and sweets),
ties and (4) determine presence or absence of constituents fiber (cotton, flax, wool and wood) fruits and human sam-
in a mixture. All four are chemometric intensive and require ples (skin, hair, blood and urine) are then considered.
a computer to perform the necessary computations.18 Obviously, there is neither time nor space to discuss all
This article will discuss the qualitative spectral charac- biological samples. The samples selected appeared to be
teristics of biomaterials in terms of the major chemical the most common of the biomaterials discussed in the NIR
components that produce those characteristics. It is not the literature. Some spectra, such as water, are duplicated from
singular purpose of this article to make band assignments. time to time to enhance an understanding of certain spectral
Nonetheless, assignment of absorption bands to functional characteristics.
groups will be made where possible. One should recognize,
however, that assigning bands in the NIR region is a diffi-
2 FUNDAMENTAL COMPONENTS
cult task. The molecular structure of biomaterials is never as
simple or as predictable as the models used in textbooks.19 2.1 Water, oil, protein and starch
Anomalous spectral information (related to light scattering,
diffuse reflection, specular reflection, impurities, refractive Water, protein, oil and starch, four major biochemical
index, etc.) is always superimposed on the vibrational data, components in plant materials, will vary in concentration
further confounding efforts to relate absorption bands to depending upon the species. Levels (percent dry basis)
specific transitions. of these components in living tissue are approximately
Spectra of selected biomaterials, shown in this article, as follows: (1) water concentration may vary from 10%
were acquired to form a basis for the discussion of spectral (walnut shells on trees) to 95% (algae), (2) protein usually
characteristics and band assignments. Recorded at a reso- falls within the range 140%, (3) lipids (or oils) can vary
lution of 10 nm on a Foss NIRsystems Model 6500 spec- from 0.1 to 45% and starch can vary from 3 to 60%,
trometer at room temperature (23 C), these spectra contain all depending on the species and stage of maturity. It
digitized absorbance values at 2-nm intervals covering the is generally agreed that there is little starch in animal
range from 400 to 2500 nm. Some spectra were shortened tissue (K. Norris and D. Hopkins, personal communication,
to 7002500 nm if the visible region contained little infor- 2000). Figure 1 shows the absorption spectra of soy protein
mation of interest. Liquid samples, each 1-mm thick unless (85%), distilled water, vegetable oil and starch. The bands
stated otherwise, were run in the transmission mode and most commonly associated with water, oil, protein and
plotted as absorbance [log1/T]. Solid samples were run starch are summarized in Table 1.
by diffuse reflection in a static sample cell and plotted as
log1/R. Diffuse reflection spectra plotted in this way will
2.1.1 Water
also be referred to as absorption spectra in this article.
For each biosample considered, the newly scanned spec- Measuring the water content of biomaterials was one of the
tra are presented first, followed by references to the pub- first (if not the first) application of near-infrared reflection
lished literature. Second-derivative spectra are presented to spectroscopy (NIRS). Today, it remains a popular appli-
enhance unresolved absorption bands. However, it should cation because of the sheer ease and simplicity of the
be kept in mind that asymmetric bands will cause peaks technique. Water is a strong absorber in the NIR region
in a derivative spectrum to shift from their true position, and its presence must always be reckoned with when trying
Near-infrared Spectroscopy of Biomaterials 3
3.50 because it established the approximate position of the major

3.00
absorption bands of water. Moreover, Collins was one
of the first to observe band shifts as the water temper-
2.50 2180
ature changed. He also noted a marked increase in the
Absorbance
Water 1932 2058

2.00
2308
magnitude of the maximum absorption for bands at 770,
2346
1.50 1734
980 and 1210 nm with an increase in temperature. These
Protein Oil
changes were predictable by Roetgens hypothesis, which
1.00 1718
1760 2100 implies that water is a mixture of at least two kinds
1208 1454
0.50
1406
of molecules whose relative amounts change as the tem-
1204 2144 Starch
0.00 perature changes.24 Later workers produced evidence that
600 1000 1400 1800 2200 2600 liquid water is composed of three components, namely the
Wavelength / nm monomer (the simplest form), the dimer and the trimer,
Figure 1. Absorption and log1/R spectra of protein, water, oil although higher multimers are almost certainly present.25
and starch. Oil and water were run in the transmission mode and Collins determined that water exhibited maxima at 740,
protein and starch in the reflection mode.
850, 970, 1170, 1430 and 1940 nm. Other researchers have
published band assignments for liquid water that differ
Table 1. Band peak wavelengths of water, oil, protein and starch
(nm). slightly from the above findings.26 Of course, the 1430-
and 1940-nm bands correspond to the first overtone of
Water Oil Protein Starch OH stretch and combination of the OH bending and
stretching modes, respectively.27 The 970-nm band is the
1210 1208 1204 second overtone band and the 740-nm band is the third
1454 1406 1465 1464 overtone of the OH stretching mode. There is a weak
1718 1734
combination band near 1200 nm.
1932 1760 1932 1932
2114 Curcio and Petty28 conducted a similar study in 1951
2058 of liquid water at 20 C. Band maxima were observed at
2100 760, 970, 1190, 1450 and 1940 nm. In addition, Curcio and
2180 Petty gave a table of absorption coefficients at the individual
2308 2302 2290
band maxima. Curcio and Petty and Collins observed a
2346 2342 2324
shoulder at 850 nm which was not observed in Figure 1.
Collins found this band to be enhanced when the water
to determine the concentration of other constituents. Even temperature was raised to 95 C.
when dried, biological samples are very hygroscopic and, It is interesting that the absorption coefficient at 1940 nm
if not sealed in a waterless environment immediately after is more than four times that at 1450 nm and the absorption
drying, will quickly reabsorb moisture. Determining the coefficient at 1450 nm is almost 25 times larger than that at
salinity of foodstuffs, forage or other biomaterials is an 1190 nm. This information becomes very important when
exciting prospect (even though salt has no definitive absorp- deciding which spectral range to use for determining the
tion bands) because researchers have found the concentra- water content of biological samples. For example, when
tion of salt in water influences the position of the water working with living tissue (tissue with a moisture content
absorption band.21 Temperature variations cause changes in in excess of 30% by weight) it would be advisable to
the relative amounts of free and bonded OH groups, which consider wavelengths shorter than 1940 nm in order to
can also be caused by changes in salt levels. Cognizance of ensure linearity of the measurements. Likewise, if the
interfering effects (such as salt and temperature) must be moisture content is 30% or lower, measurements at 1940 nm
taken, making it important to control one while studying would be a likely candidate.
the effect of the other.22 Maeda et al. looked at variations of water spectra at
Two prominent bands, at 1454 and 1932 nm, show up in different temperatures. Their study (using multiple linear
the absorption spectrum of water (see Figure 1). However, regression, principal components regression and partial
bands at 762, 960 and 1152 nm became definable in the least-squares regression) led them to the conclusion that
second-derivative spectrum of water. a feature close to 1449 nm (due to the combination of
One of the earliest studies of the NIR absorption spec- antisymmetric and symmetric stretching modes of water)
tra of liquid water at temperatures of 0 and 90 C was is actually made up of five components ascribed to the
conducted by Collins.23 This work was not concerned water species with no, one, two, three and four hydrogen
with biological samples but it is germane to this article bonds. Bands due to water species with no hydrogen bonds
increase considerably with temperature, whereas those with 1.40

more than two hydrogen bonds decrease.29 Tanaka et al. 1.20
1934 2108
Wheat
demonstrated the power of NIR technology to classify min- 1468 1774
1.00 Gluten
eral waters using principal component analysis. Using three
Absorbance
Starch
wavelengths (1184, 1611 and 1709 nm) from the original 0.80
Bran
absorption spectra and the Mahalanobis distance method, 0.60 1202
they found that discrimination was 100%.30 0.40 994

Germ
0.20
2.1.2 Oil
0.00
600 1000 1400 1800 2200 2600
Biomaterials contain lipid fractions commonly referred to
as oils. Lipids are a class of compounds which contain long- Wavelength / nm
chain aliphatic hydrocarbons and their derivatives, such as Figure 2. Log1/R spectra of wheat, gluten, starch, bran and
fatty acids, alcohols, amines, amino alcohols and aldehy- wheat germ. All spectra were recorded in the reflection mode.
des, including waxes, fats and other similar compounds.31
Kamishikiryo et al.32 identified five prominent absorption 2.1.4 Starch
bands in soybean, monoolein and triolein oils with peaks
Animal tissue contains little or no starch. However, plant
at 1208, 1720, 1766, 2306 and 2348 nm. Peak absorptions
tissue may contain as much as 60% by weight (dry basis).
appear in the oil spectrum in Figure 1 at 1208, 1406, 1718,
Figure 1 shows the spectrum of starch having maxima at
1760, 2144, 2308 and 2346 nm. Of the seven bands identi-
1204, 1464, 1932, 2100, 2290 and 2324 nm. The broad
fied, only the 2306-nm band was found to be quantitative
combination band at 2100 nm, commonly called the car-
for oil in oilwater emulsions.
bohydrate band, is always present in the spectrum of plant
Spectra of various oils exhibit a sharp CH stretch (first
tissue. In the presence of a sizeable amount of protein
overtone) absorption at 1722 and 1760 nm. The stronger
bands peaking at 2310 and 2346 nm are combination bands the analytical quantification of either starch or protein is
involving CH stretching along with other vibrational always more difficult. We shall see later (Figure 2) that
modes in the molecule.33 While the weaker absorption band the starch fraction is largely responsible for the char-
near 1700 nm is thought to be the first overtone of the CH acteristics of the dry-wheat spectrum.35 In the second-
stretch, the bands near 1208 nm are the second overtones. derivative spectrum of starch (not shown), the first overtone
Takamura et al. at Nara Womens college in Japan stud- of the CH stretch in the 1700-nm region is clearly
ied second-derivative spectra of edible oils undergoing lipid present.
oxidation. Although they found several peaks in the dif-
ference spectra of oil and oxidized oil, it was the peak
at 2084 nm that was highly correlated with peroxide val- 3 PLANT MATERIALS
ues. In addition, they discovered that the peak shifted
and weakened during the reduction of hydroperoxide to 3.1 Grains
hydroxide.34
Originally developed for measuring liquids,9 13 it was
2.1.3 Protein not long before applications of NIR spectroscopy were
Figure 1 reveals protein absorption bands at 1208, 1734, extended to include solids.15,36 NIR analysis of grain and
2058, 2180, 2302 and 2342 nm. Protein, critical to biopro- seeds was first introduced by Norris and Hart in 1965.37
cesses in both plant and animal tissue, has three major NH These research programs formed a strong foundation upon
combination bands at 1980, 2060 and 2180 nm. The first which to build modern NIR analyses of biomaterials. It
overtone of the NH stretch can be seen at 1510 nm. is generally recognized that absorption in the NIR region
Kamishikiryo et al.32 conducted a study of the spectral is due to the CH, NH and OH stretching vibrations
properties of bovine serum albumin (BSA) in which they plus combination bands, with the fundamental absorptions
found eight absorption bands at 1380, 1516, 1692, 1738, occurring in the MIR region. Specific band assignments
1856, 2056 and 2170 nm. Of these eight bands, only the are always difficult at best. Even when assignments are
band at 2170 nm was quantitative with respect to BSA made, the bands tend to shift from one product to another.
concentration. Measurements of absorbance [log1/T] at In the following sections analyses of newly recorded
2170 and 2306 nm provided a good estimate of protein in spectra will be made in the light of work reflected by the
the emulsions. literature.
3.1.1 Wheat and wheat components 2.00

1
2
Law and Tkachuk38 were the first to make extensive assign-
3
Absorbance
ments of NIR bands in the spectra of wheat and wheat com- 1.50
ponents (gluten, starch, pentosans, lipid and water). Their Sesame 4
work concentrated on assignments to various overtones 1.00 Sunflower
Barley 5
6
and combination vibrations of CH, NH, OH and CDO Wheat Grass
bonds. They claimed to be the first to publish spectra of 0.50 Oats
wheat and wheat components except for the low-resolution

spectra between 1000 and 2000 nm for whole grains.39 The 0.00
600 1000 1400 1800 2200 2600
paper by Law and Tkachuk provided an extensive list of Wavelength / nm
papers that support their band assignments.9,13,33,40 52 This
Figure 3. Log1/R spectra of six grains: (1) sesame, (2) sun-
list of references not only provides an excellent source of flower, (3) wheat, (4) barley, (5) grass and (6) oats. All samples
support for making band assignments, but also it names were scanned in the reflection mode.
many people who were working with NIR as early as
1938.40 Osborne and Douglas55 studied the degree of starch damage
The spectra for ground wheat, gluten, starch, bran and in flour, making extensive band assignments.
germ are shown in Figure 2. The wheat spectrum shows
the dominant water (1940 nm) and carbohydrate (starch
3.1.2 Other grains
band at 2108 nm) bands while the prominent protein peaks
at 2054 and 2184 nm (seen in the gluten spectrum) serve Figure 3 shows the absorption spectra of six grains includ-
only to broaden the starch band in the wheat spectrum. The ing ground wheat (the latter given for comparison). The
1980-nm band, due to the primary amide group,42,43,48,53 spectra of sesame (spectrum 1) and sunflower (2) seeds,
was not observed by Law and Tkachuk because of the noted for their high oil content, exhibit very prominent oil
diminishing effect of water and starch absorptions at 1930 bands that peak at 1720, 1760 (the first overtone of the CH
and 2090 nm, respectively. stretching mode), 2308 and 2346 nm (CH stretching and
The type of wheat seems to have no effect on the bending combination bands). Wheat, barley and oats seem
number or position of the absorption bands. Major peaks in to have far more starch than sesame and sunflower seed.
dried gluten spectra occurred at 1190, 1500, 1730, 1980, However, the high protein in the latter may be masking
2050, 2180, 2290 and 2470 nm. The characteristic peak starch absorption.
of gluten spectra occurs at 1980 nm and is caused by the Kaffka and Balogh made attempts to determine oil,
high concentration of primary amide groups in gluten.38 protein, water and fiber in sunflower seeds using NIR
The spectrum of dried gluten was found to be similar spectroscopy.56 A multiple-linear regression program was
to the transmission spectrum of porcupine quill published used to select the most important wavelengths for quantita-
by Hecht and Wood44 in 1956, except for the intense tive determinations. These wavelengths were closely allied
peak at 1980 nm produced by the primary amide group. with the findings of other researchers.57,58
Although Fraser54 observed an absorption at 2010 nm in all
peptide-containing polymers except silk, Law and Tkachuk
did not find it in the gluten spectrum. In addition to the 3.2 Tea
possibility that it is not present in all gluten samples was
the possibility that the band was overshadowed by the Figure 4 gives typical spectra of red and green tea leaves.
primary amide band and/or inadequate resolution of the The samples were scanned as is from commercial pack-
spectrometer used in the study. This work was conducted ages. Major bands in these spectra appear at 990, 1194,
with a Cary 17I spectrometer that automatically varied the 1498, 1518s, 1726, 1934, 2064s, 2158, 2306 and 2350 nm
slit from 2.6 to 0.2 mm over the scanning range from 2500 (s D shoulder). The bands at 1518 and 2164 nm appear
to 1000 nm. as shoulders on the water (1450 nm) and carbohydrate
Pentosans (spectrum not shown) exhibited similar peaks (2100 nm) bands, respectively. Spectra of instant and herb
to starch, with the major peaks occurring at 1200, 1450, teas are given for comparison. The instant tea mix had
1540, 1930, 2090, 2320 and 2490 nm. Wheat bran and added sucrose that dominates the protein bands, producing
germ have spectral characteristics similar to those of ground a peak at 2102 nm. There is evidence of lipid absorptions
whole wheat. It is noted, in this case, that the particle size at 2306 and 2350 nm for all four teas.
of bran and germ, noted by the downward shift of these Caffeine is present in tea in concentrations ranging from
two spectra, was much smaller than that for ground wheat. 3 to 5%. In addition, theobromine and theophylline are
1.20 2306 R
in both spectra at 482, 554, 650, 756, 1194, 1448, 1576,
R = red tea 2064
2350 G 1726, 1762, 1934, 2138, 2310 and 2348 nm. The spectra
1.00 G = green tea 2158
I = instant tea 1934 look very much alike with the exception that the two peaks
Absorbance
H = herb tea
0.80 I at 2306 and 2346 nm are much weaker in the spectrum of
1518 1726
0.60 2102 2330
H
decaffeinated coffee. These bands appear in the spectrum
1498
of ground decaffeinated (not instant) coffee. This seems
0.40 1472 to indicate that the process of producing instant coffee
1194
0.20 990 removes the CH bonds that cause the absorption bands at
0.00
2306 and 2346 nm.
600 1000 1400 1800 2200 2600 Evidence points to bands at 1726 and 1764 nm being the
Wavelength / nm first overtone of the CH stretching vibration. Furthermore,
Figure 4. Log1/R spectra of four dried tea samples: (1) red tea, it is apparent that the bands at 2138, 2306 and 2346 nm are
(2) green tea, (3) instant tea and (4) herb tea. All spectra were CH combination bands.
recorded in the reflection mode.
present, but in much smaller quantities for theobromine

3.4 Tobacco
(0.17%) and for theophylline (0.013%) in black tea.59
Ikegaya et al.60 made a study of Japanese tea and made Five absorption spectra of flue-cured tobacco are given in
several band assignments. They indicated that the strong Figure 6: high nicotine (7.1%), low nicotine (1.3%), high
band at 1690 nm (we find this band to be at 1660 nm) and a sugar (24.2%), low sugar (2.3%) and an unanalyzed sample.
band near 2240 nm were associated with CH3 in the caffeine Absorption bands in all spectra peak at 454, 548, 670, 756,
molecule. In a related study, implied in the same paper, they 1194, 1494, 1586, 1724, 1938, 2124, 2310 and 2354 nm.
found that the intensity of both these bands increased with As expected, all spectra show evidence of chlorophyll
increasing CH3 concentration in the structure. It seems safe by the presence of the chlorophyll absorption band at
to say that the bands at 2306 and 2350 nm are combination 670 nm. High chlorophyll content is usually indicative of
bands (CH stretching and bending modes) and that the improper processing (or a reduced yellowing period). Since
peaks in the neighborhood of 1730 and 1760 nm are the the major biochemical change during yellowing is the
first overtone of the CH stretching mode. conversion of starches to sugars, a curtailed yellowing time
produces tobacco with low sugar content.61 It is interesting
to note in this case that the carbohydrate band at 2124 nm
3.3 Coffee complements the chlorophyll information at 670 nm. That
is, the high-sugar sample had a low chlorophyll content and
The spectra of ground regular and instant decaffeinated cof- the low-sugar sample had a high chlorophyll content. Both
fee are given in Figure 5 along with the second-derivative the low-nicotine and high-sugar samples had high sugar
spectrum of regular coffee. Major absorption bands appear levels as noted by the presence of well defined carbohydrate
bands at 2124 nm. The shift of this band from 2100 to
1.5 0.04
1.3 0.02 0.8 454

Second derivative
548 HiNic
0.00 0.7 670 2310
1.1
Absorbance
1938 2124
Decaf 2nd deriv. 2306 0.02 0.6 756 1724 2354
Absorbance
0.9 2346 LoNic 1494

0.04 0.5 1586 HiSug
0.7
1934
2124 0.06 0.4 1194
0.5 1726
1760 0.08 0.3 LoSug
1464 Log (1/R) HiSug LoSug
Reg
0.3 1202 0.10 0.2
0.1 0.12 0.1 Unk
600 1100 1600 2100 2600
0.0
Wavelength / nm 300 800 1300 1800 2300
Figure 5. Log1/R and second-derivative spectra of coffee. The Wavelength /nm
log1/R spectrum was recorded in the reflection mode. There Figure 6. Log1/R spectra of five flue-cured tobacco samples:
are eight major absorption bands in the ground-coffee spectra: High nicotine (HiNic 7.1%), low nicotine (LoNic 1.3%),
1202 nm (H2 O), 1726 and 1760 nm (lipids), 1934 nm (H2 O), high sugar (HiSug 24.2%), low sugar (LoSug 2.3%) and
2124 nm (carbohydrates) and 2310/2436 nm (lipids). Scanning an unanalyzed sample (Unk). All samples were scanned in the
was performed in the reflection mode. reflection mode.
2124 nm is probably due to the presence of a sizeable 1

2
amount of protein in both samples. 2.50
3
Measuring nicotine is relatively easy and robust in spite 4
2.00
Absorbance
of the fact that nicotine does not exhibit a noticeable 5
absorption band.62,63 However, a close look at the second- 1.50
derivative spectra would reveal a unique nicotine absorption
at 2146 nm that is not influenced by usual sample adulter- 1.00
ation. Looking at the carbohydrate band (2124 nm), it is
0.50
predicted that the unknown sample has about the same
amount of sugar as the low-sugar sample. 0.00
600 1000 1400 1800 2200 2600
Wavelength / nm
4 FOOD AND DRINK Figure 8. Log1/R spectra of meat and fish: (1) low-fat ham-
burger (7% fat), (2) high-fat hamburger (20% fat), (3) herring,
4.1 Food (4) tuna and (5) turkey. The spectra have been shifted with respect
to each other for clarity. All spectra were scanned in the reflection
mode.
4.1.1 Cereals
Absorption spectra of table cereals are given in Figure 7: The combination OH bands (around 1934 nm) for all cereals
(1) Honey Corn Flakes, (2) Corn Flakes, (3) Special-K and have a normal shape whereas the first overtone of OH
(4) Froot Loops. The presence of crystalline sucrose is stretch at 1450 nm is influenced by overlapping absorbers
evident by the sharp peak at 1440 nm in the spectra of both to such an extent that it is difficult to define the water peak.
Honey Corn Flakes and Froot Loops. Davies and Miller Other than these features, the spectra of table cereals are
suggested that this sucrose band is due to an OH bond much alike.
which is held in the crystalline structure so that it cannot
take part in the inter- or intramolecular hydrogen bonds.64 4.1.2 Meats, poultry and fish
Davies and Miller used a shatterbox to continue particle
size reduction while periodically recording the spectrum, The spectra of beef [low-fat (7%) and high-fat (20%) ham-
resulting in an enhanced reflection (decreased absorption) of burger], poultry (turkey) and fish (herring and tuna) are
the NIR energy. This reduction in particle size sharpened all given in Figure 8. Containing more than 70% water, these
features except the 1440-nm band; the intensity of the 1440- samples produce spectra that are dominated by water bands
nm band decreased as grinding continued. Hence Davies at 1450 nm (first overtone of the OH stretching mode) and
and Miller concluded that additional hydrogen bonding took 1934 nm (the combination band of OH vibrations). High-
place as the unbound OH was exposed by crystal fractures. fat hamburger and herring have prominent lipid peaks at
Comparing the carbohydrate peak (2100 nm) in both 1730, 1764 (first overtone of CH stretch), 2308 and 2350 nm
Corn Flakes and Special-K confirms that Special-K has (combination bands of CH symmetric stretching and bend-
added protein by the fact that the carbohydrate band of ing modes). Significant evidence of the low-fat content of
Special-K is broader than the same band for Corn Flakes. poultry is provided by the absence of lipid peaks in the
turkey spectrum.
1.80 Chen and co-workers at the USDA laboratories, Belts-
1
1.60
2
ville, MD were among the first to study the spectral char-
1.40 acteristics of poultry carcasses. In spite of the fact that
Absorbance
1.20 3 the spectra of healthy and abnormal carcasses look very

4
1.00 much alike, Chen and co-workers were able to classify
0.80 carcasses with a 92% accuracy (8% errors) using neural
0.60 networks.65 67
0.40
0.20
0.00 4.1.3 Fresh fruit
300 700 1100 1500 1900 2300
Wavelength / nm Kawano and co-workers have conducted experiments to
Figure 7. Log1/R spectra of four cereals: (1) Honey Corn measure the sugar content of peaches,68 the sugar content
Flakes, (2) Corn Flakes, (3) Special-K and (4) Froot Loops. Note of satsuma oranges69 and other agricultural products.70 74
the appearance of the crystalline sucrose band at 1450 nm. Samples were not available in order to include original
3.50
1934 1 1.60 1
2
3.00 2
2096
Absorbance
2.50 2276 1.20 3
Absorbance
3
2306
2.00 1464 4
2272
2394
4 0.80
1.50
5
1.00 0.40
0.50
1182 0.00
0.00 600 1000 1400 1800 2200 2600
600 1000 1400 1800 2200 2600
Wavelength / nm
Wavelength / nm
Figure 10. Log1/R spectra of five popular sweets: (1) sac-
Figure 9. Absorption spectra of (1) molasses, (2) honey, (3) corn charin, (2) aspertame, (3) white chocolate, (4) dark chocolate and
syrup and (4) lemon extract. Major peaks appear in all spectra at (5) sucrose. All spectra were recorded in the reflection mode.
1464, 1934, 2096, 2272, 2276, 2306 and 2394 nm. All spectra
were scanned in the transmission mode.
sucrose present was in the amorphous form, the 1440-nm
spectra in this article. Readers are encouraged to refer to band would not be present.81,82
the original publications referenced here.
4.2 Drink
4.1.4 Additives: molasses, honey, corn syrup and
lemon extract
4.2.1 Milk and water
The spectra of molasses, honey, corn syrup and lemon
Milk has been enjoyed by young and old alike around the
extract are shown in Figure 9. All spectra show promi-
world for centuries. It is a product that has a number of NIR
nent water absorptions at 1464 and 1934 nm. However,
spectral features related to milk quality, composition and
molasses, honey and corn syrup produce spectra that exhibit
physical properties. Milk has been subject of intensive NIR
a strong carbohydrate band peaking at 2096 nm. A similar
investigation over the past 40 years.83 99 More recently,
band appears in the spectrum of lemon extract but it peaks
Sasic and Ozaki100 made an in-depth study of milk in
at 2074 nm. It is suspected that the 2074-nm band is also
order to make band assignments. Although their study was
due to carbohydrates in the product but is shifted owing
conducted at a sample temperature of 40 C whereas our
to molecular interactions not encountered in molasses.75,76
samples were scanned at room temperature, their results
Note that the combination bands in the spectra of lemon
were invaluable reassurances in making band assignments
extract appear at 2272, 2306 and 2394 nm with the first
for the samples we scanned.
overtone transitions appearing at 1696, 1730 and 1764 nm.
Milk is made up of more than 80% water and has a spec-
Orange juice spectra (not shown) do not manifest these
trum that looks very similar to water (see Figure 11). The
absorptions.77 80
most obvious difference between the spectra of water (spec-
trum 5 in Figure 11) and milk is that the milk spectra have
4.1.5 Sweets: sucrose, saccharin, extra and higher absorbances than water at all wavelengths. A careful
chocolate
4.50
Nowhere in the world are sweets more important to peo- 4.00 0.08
2nd derivative
ple than in the United States. In recent years, attention of 0.05

3.50 0.02
0% 0% 1
2
Absorbance
0.01
NIR spectroscopy has been focused on sucrose, chocolate 3.00 0.04 3
0.07 4
and sweet substitutes. Spectra of these products are given in 2.50 0.10 H&H
H&H Half & half
5
0.13
Figure 10. The water bands (at 1450 and 1934 nm) are min- 2.00 2232 2261 2290 2319 2348 2377
Wavelength / nm
2406
imal in the spectra of all these products. Note also that the 1.50
1.00
1440-nm band due to the presence of non-hydrogen-bonded
0.50
OH in the crystal structure64 is present in all spectra shown. 0%
0.00
Furthermore, it is interesting that the sugar substitutes pos- 300 700 1100 1500 1900 2300
sess this same band. The presence of the 1440-nm band Wavelength / nm
in chocolate suggests that sucrose in chocolate is present Figure 11. Absorption spectra of five milk samples with varying
largely in the crystalline, not the amorphous form. If all fat content. Spectra were collected in the transmission mode.
look at milk spectra reveals an interesting correlation. As 4.50

2nd derivative 0% Ethanol
the concentration of fat globules and protein micelles in the 4.00
0.28
0%
milk samples decreases, the absorption spectra shift down- 3.50 0.00
Absorbance
0.28
ward. This shift is a light scattering effect, an expected 3.00 1
0.55
100% 2
occurrence in turbid liquid samples. Light scattering causes 2.50 3
2210 2244 2278 2312 2346 2380 2414
2.00 Wavelength /nm 4
the light to travel further than it would in a sample (like 5
1.50 6
water) free of this peculiarity. This idiosyncrasy is always
1.00
present in turbid (milk and muddy water) and solid (plant
0.50
tissue) samples. 0.00
100% Ethanol
The second obvious difference between the spectra of 300 700 1100 1500 1900 2300
milk and water is the presence of a dominant linear term that Wavelength / nm
tilts the milk spectra upward in the 4001300-nm region, an Figure 12. Absorption spectra of six ethanolwater mixtures.
effect caused by color transitions associated with the milk Spectra were obtained in the transmission mode.
complex. This is another anomaly associated with turbid
samples, and it becomes more prominent as the sample 4.2.3 Spirits
becomes darker in color. For example, if one were to add
brown soluble dye to the milk, the color change would be The term spirits is used here as an alcoholic bever-
exhibited as an increase in the slope of the spectra between age category that includes everything except beers and
400 and 1300 nm. wines. This category has been divide into two subgroups:
A less obvious, but still noticeable difference is the pres- (1) liquors (including scotch, gin, vodka and bourbon)
ence of absorption bands due to fat in the region of 1700 and (2) miscellaneous (including tequila, liqueur, rum and
and 2300 nm. These peaks can be clearly observed in the cognac). Beers and wines are treated separately.
spectra of half-and-half (a product containing half cream Absorption spectra of Chivas Regal scotch, Beefeater gin,
and half milk). Half-and-half certainly has the highest con- Smirnoff vodka and Wild Turkey bourbon are shown in
centration of fat, a fact that is substantiated by the obvious Figure 13. There are four regions where these biomaterials
fat bands at 1730, 1778 and 2312 and 2354 nm. The two differ from one another: (1) the first overtone of the OH
bands at 1730 and 1778 nm have been assigned to first over- stretch in the region of 1456 nm, (2) the first overtone of
tone CH2 stretching vibrations of fat and the two bands at the CH stretch (1694, 1732 and 1764 nm) in the region of
2312 and 2354 nm to combination stretching and bending of 1700 nm, (3) the combination band of OH transitions in
CH2 and CH3 in the fat molecules. The second-derivative the region of 1940 nm and (4) the combination bands of CH
spectra (inset in Figure 11) clearly indicate the quantitative vibration modes in the 2300-nm region. It is interesting that
relationship between the magnitude of the bands and fat the water bands are inversely correlated with the alcohol
concentration. content whereas the magnitude of the intensity of the
second-derivative bands is clearly proportional to alcohol
content the same phenomenon that was demonstrated in
4.2.2 Ethanolwater mixtures
Figure 12.
Figure 12 gives the absorbance spectra of six samples of
water with an ethanol content ranging from 0 to 100% 3.50 1-Scotch (40%)
2nd derivative
by volume. Ethanol absorption bands have peaks at 1694, 3.00 0.05
2-Gin (47%)
3-Vodka (50%)
1732, 1764, 2270, 2304 and 2352 nm. These are confirmed 0.03
4-Bourbon (55%)
2.50 0.00
1
Absorbance
as ethanol bands by the absence of these bands in the water 0.03 1

4
spectrum. The bands at 2270, 2304 and 2352 nm originate 2.00 0.05 4
1670 1687 2278 1721 1738 1755 1772

in the combination stretching and bending of CH2 and CH3 1.50 Wavelength / nm
in ethanol and the three bands at 1694, 1732 and 1764 nm 1.00
are the first overtone bands of the CH2 stretching modes.
0.50
All six bands clearly correlate with ethanol concentration
in water. It is interesting that the water bands at 1450 (the 0.00
300 700 1100 1500 1900 2300
first overtone OH stretch) and 1940 nm (the OH combina- Wavelength / nm
tion band) are inversely related to ethanol concentrations in
Figure 13. Absorption spectra of four commercial spirits:
the sample. It can be shown that an ethanol calibration can (1) scotch (with 40% ethanol), (2) gin (47%), (3) vodka (50%)
be used to determine the ethanol content of many alcoholic and (4) bourbon (55%). Scanning was performed in the
drinks. transmission mode.
4.0 2nd derivative

6.00 2nd derivative
1-Tequila (17%) 0.13
3.5 0.04 2-Liqueur (22%) 5.00 0.16
1
3-Rum (34%) 0.00
3.0 0.01 4
Absorbance
4-Cognac (40%)
17%
4.00 0.06
Absorbance
2
0.01
2.5 0.13 3
0.03
2.0 3.00 2214 2241 2268 2295 2322 2349 2376
40% Wavelength/nm
0.06
1634 1666 1698 1730 1762 1794
White wine 4
1.5 Wavelength /nm
2.00 Sake 3
Red wine 2
1.0 1.00 Blush wine 1
0.5
0.00
0.0 300 700 1100 1500 1900 2300
300 700 1100 1500 1900 2300
Wavelength / nm
Wavelength / nm
Figure 16. Absorption spectra of four different wines: (1) white,
Figure 14. Absorption spectra of four commercial spirits: (2) sake, (3) red and (4) blush. Spectra were obtained in the
(1) tequila (with 17% ethanol), (2) a liqueur (22%), (3) rum transmission mode.
(34%) and (4) cognac (40%). Scanning was performed in the
transmission mode. Figure 16 shows the absorbance and truncated second-
derivative spectra for four wines: (1) Sutter Home blush,
Figure 14 shows the absorption and truncated second-
with 9% alcohol, (2) Sutter Home cabernet sauvignon,
derivative spectra of Tequila Rose liqueur, with 17%
13%, (3) Hakutsuru sake, 14.5%, and (4) Sutter Home
alcohol, Amaretto liqueur, 22%, Bacardi rum, 34%, and
white, 12.5%. Like the beer spectra, these spectra are
Hennessy cognac, 40%. The same thesis as given above
amazingly similar. The combination bands of the CH
for liquors holds true for these miscellaneous drinks, that is,
stretching and bending modes are evident in the 2300-nm
both water bands and alcohol bands are correlatable with the
region. These bands, like those shown for other alcoholic
alcohol concentration. The spectra of Tequila Rose liqueur
drinks, are quantitative, as are the first overtone (of the CH
and Hennessy cognac have shoulders at 2096 nm due to the
stretch) bands in the 1700-nm zone. The spectrum of the red
addition of carbohydrates in these two products. The fact
wine shows only a slight color effect in the 4001300-nm
that the cognac drink was darker than the other three is
region.
clearly demonstrated by its strong absorption in the region
4001300 nm.
The absorbance and second-derivative spectra of dark
5 FIBER
and light beers, along with an absorption spectrum of
6% alcoholwater mixture, are given in Figure 15. With 5.1 Cotton, flax and wool
the exception of the color-transition effects in the region
from 4001300 nm, all three spectra look remarkably alike. Absorption spectra of wool, flax and cotton are given in
However, enhancement of the alcohol combination bands Figure 17. A strong cellulose band appears in the spectra
in the region of 2300 nm using a second-derivative math of cotton and flax at 2276 nm while the 2294 and 2340 nm
treatment of the absorbance spectra clearly indicates the bands in wool are attributed to lipids. Both lipid bands
quantitative nature of the alcohol bands despite the fact arise from combination CH stretch and bending modes.
that the three spectra are quite similar. The spectrum of wool (wool is almost 100% protein) has
4.50
Wool
4.00
2nd derivative 1.20 2294 2340
0.05
Lite 1940 2174
3.50 0.02
1.00 2054
0.01
Absorbance
1
3.00 0.03
Absorbance
3 Drk 6% Flax
2.50 0.06 1 0.80 1500 1736
2228 2262 2296 2330 2364
2 1698 2276
Wavelength /nm
2.00 3
0.60 1184
Cotton
1.50
1.00 0.40 2106 2348
0.50 0.20
0.00
300 700 1100 1500 1900 2300 0.00
Wavelength / nm 600 1000 1400 1800 2200 2600
Figure 15. Absorption spectra of (1) light-colored and (2) dark Wavelength / nm
beer along with the spectrum of a 6% ethanolwater mixture. Figure 17. Log1/R spectra of wool, flax and cotton. The reflec-
Spectra were obtained in the transmission mode. tion mode was used to obtain these spectra.
two prominent protein peaks at 2054 and 2174 nm. These the absorption spectra where the wood and knot materials
protein bands (combination bands from CH vibrations) also differ, one in the region of 1700 nm and the other in
have first overtone bands in the 1700 nm region (1698 and the region of 2300 nm (the first overtone and combination
1736 nm). Cotton and flax each have strong carbohydrate regions of the CH modes of vibration, respectively). The
bands at 2106 nm. second-derivative inset enhances the absorption spectra in
The spectra of rough-ground soft-pine wood and pine the 1700-nm region. Pine knot has bands at 1632, 1696
knot are shown in Figure 18. Wood, like cotton and flax, and 1718 nm, which do not appear in the spectrum of
has a cellulose peak near 2270 nm. Two regions show up in pine wood. These three bands are probably due to the
presence of pine rosin in the knot.
1.40
2nd derivative
1.20 0.01
Pine knot
1.00
0.01
Pine wood 1788 6 HUMAN SAMPLES
Absorbance
0.00
1632 1788 1828 Wood
0.01 1590
0.80 1696 1718
1566 1622 1678 1734 1790 1846 6.1 Skin
Wavelength (nm)
0.60 Knot
0.40 Figure 19 shows the absorbance spectra of the thumb-pads
0.20 of the left and right hands of three subjects, called Don, Fred
0.00
and Katie. These three subjects definitely were of different
300 700 1100 1500 1900 2300 body types: Don was short and weighed 125 kg, Fred
Wavelength /nm was 1.88 m in height and weighed 100 kg and Katie was
Figure 18. Log1/R spectra of pine wood and pine knot. The slim and weighed only 55 kg. Absorbance measurements
spectra were obtained in the reflection mode. were taken with an NR-6645ANO3P Interactance Probe on
1.60
1.40 Fred
1.20 Don
Katie
1.00
Absorbance
0.80
0.60
0.40
0.20
0.00
1000 1200 1400 1600 1800 2000 2200
(a) Wavelength /nm
2nd derivative 2nd derivative
0.005
0.00
LH RH
0.006 0.01
LA
LA
RH
0.008 LH 0.02 UA
UA
0.009 0.02 Water
Fat
Fat
0.011 0.03
1172 1182 1192 1202 1212 1222 1122 1154 1186 1218 1250
(b) Wavelength / nm (c) Wavelength / nm
Figure 19. Spectra of human skin: (a) log1/R spectra of the left and right hands of three subjects (Fred, Don and Katie); (b) abridged
second-derivative spectrum of the left and right hands of Fred; and (c) abridged second-derivative spectrum of the upper and lower
right arm of Katie. Spectra were obtained in the interactance mode.
a Foss NIRsystems Model 6500 spectrometer. Transmission a person is right- or left-handed. If the person is right-
characteristics of the fiber optics limited the useful scan handed the fat band is lower for their right hand than for
range to 4002200 nm. The spectra in Figure 19(a) appear their left. Massie used bands in this region to develop
to contain little information other than the 1200- and 1450- a hand-held NIR fat tester for hamburger.101 One neat
nm water bands. demonstration, in an effort to excite grade-school students
Two truncated second derivatives, however, demonstrate about NIR technology, was to go into a group with the
once again that information is present for both qualitative fat meter, measure the right and left palms and tell
and quantitative work. For example, Figure 19(b) shows them whether they were right or left-handed (K.H. Norris,
that the fat band at 1210 nm can be used to predict whether personal communication, 2000).
1.20
1.10
1.00 2054
2180
6.2 Hair
0.90
Absorbance
1940
0.80 2298
2352
0.70 A spectrum of black human hair is given in Figure 20.
0.60 Major absorption bands appear at 1186, 1502, 1578s, 1698s,
1502 1736 1810
0.50 1736, 1810s, 1940, 2054, 2180, 2298 and 2352 nm (s D
0.40 shoulder). Hair, like wool, is largely protein. Therefore, the
0.30 1578 1698
1186 two bands at 2054 and 2180 nm are combination CH bands.
0.20
600 1000 1400 1800 2200 2600 The two bands at 2298 and 2352 nm are combination CH
Wavelength / nm bands for lipids. The first overtone bands of the CH stretch
Figure 20. Log1/R spectrum of black human hair. The reflec- and deformation modes are found in the region around
tion mode was used to obtain this spectrum. 1700 nm.
4.00 0.01
566 8.6% HmG 1932

Second derivative
3.00 0.01
462
Absorbance
1108
720
2.00 145 0.00
794 836 1190
1.00 0.01 774 968

16.5% HmG 926
755 1154
0.00 0.01
400 800 1200 1600 2000 2400 700 800 900 1000 1100 1200
(a) Wavelength / nm (b) Wavelength / nm
0.03
0.02
0.02
0.01
Second derivative
8.6% HmG
Second derivative
0.01
0.01
0.00 2258 0.00 1624
16.5% HmG
0.01 0.01
1692 1740
2290
2170 1514 1784
0.02 2056 0.01
0.03 0.02
2000 2100 2200 2300 2400 1500 1600 1700 1800
(c) Wavelength / nm (d) Wavelength / nm
Figure 21. Triplicate spectra of human blood having 8.6% and 16.5% HmG: (a) absorption spectra; (b) second-derivative spectra in
the short-wavelength region; (c) second-derivative spectra in the region 20002400 nm; (d) second-derivative spectra in the region
15001800 nm. Absorption spectra were obtained in the transmission mode.
6.3 Blood 7 FUTURE TRENDS IN NIR

SPECTROSCOPY
The relatively good transparency of biological materials in
the NIR region of the spectrum permits sufficient photon
McClure et al.,5 in 1989, speculated that the future of
transmission through organs in situ for the monitoring of
NIR would evolve into four major directions: (1) Fourier
cellular events. This one attribute makes the study of blood
transform near-infrared (FT-NIR), (2) hyphenated systems,
using NIR technology very exciting.102 Triplicate spectra of
(3) NIR imaging spectrometry and (4) hand-held NIR spec-
aliquots of two samples of blood are given in Figure 21(a).
trometry. We are now seeing those projections unfolding on
One sample had a hemoglobin content of 8.6% and the
the spectroscopy playing field. In 1989, one FT-NIR com-
other 16.5%. There are four major absorption bands, at 462,
pany (perhaps two) marketed components that enabled them
566, 1450 and 1932 nm. Additionally, there is definitely a
to scan the NIR region. Today, the list of NIR instrument
light-scattering difference between the two sets of spectra
manufacturers includes numerous producers of FT-NIR
noted by the constant separation of the two spectral sets
spectrometers. Furthermore, at least two NIR instrument
throughout the 4001200-nm region. Major bands which
makers have produced dedicated FT-NIR systems.
show a strong quantitative relationship with hemoglobin
are at 1154, 1514, 1624, 1692, 1740, 2056, 2170, 2258,
2290 nm (see Figure 21bd). The bands agree with the
findings of Kuenstner and co-workers103,104 and others.105 7.1 Hyphenated systems
Two dimensional (2D) correlation spectroscopy has drawn

6.4 Urine attention to the importance of hyphenated systems. Bar-
ton and co-workers have demonstrated numerous enhance-
The second-derivative spectra of urine samples from six ments by taking advantage of multi-domain information on
patients are given in Figure 22. The ages of the patients the same sample set.107 110 Barton and Himmelsbach111
(four female and two male) ranged from 26 to 67 years. stated:
There appeared to be no gender or age correlations in
the recorded spectra. However, the two absorption bands Current research indicates that it is possible to correlate
at 2154 and 2200 nm, identified as urea absorptions, are one region of the spectrum against another, for exam-
ple, the MIR against the NIR to provide an interpretation
in keeping with the results of Shaw et al.106 Shaw et al.,
of the NIR spectrum. A map of the squared correlation
working with 177 urine samples, concluded that urea quan- coefficients can be plotted to show frequencies (wave-
titation is as accurate as the reference method (enzymatic lengths) of highest correlation. Since the MIR is more
conductivity). Creatinine analysis by NIR was slightly less easily interpreted because you can see the fundamental
accurate than the Jaffe rate method and the accuracy of the vibrations, this assists the spectroscopists with the diffi-
NIR method for measuring protein could not approach that cult task of assigning the observed absorbances in the
NIR where the combination and weaker overtones over-
of the reference method. It was, agreed however, that the lap and possess considerable intercorrelation. . . Progress
NIR method for protein could be used as a coarse screening in this area has been aided by the appearance of instru-
tool for protein levels above 0.3 g L1 . ments which cover multiple regions of the spectrum. The
sample remains in place while the visible, NIR and MIR
0.03 spectra are taken. Thus, sampling error and multiple instru-
ment error are removed. Additionally, software packages
are available which will combine data from multiple regions
Second derivative
0.02 and instruments. . . the analyst is challenged to consider the

2154
interpretable and the measureable; the qualitative and the
2200 quantitative aspects of an analysis. The possibilities for the
0.01
extraction of new information from the spectral and refer-
ence data are limited only by our ability to conceptualize
0.00 the models involved.
0.01
2000 2100 2200 2300 2400 Noda and co-workers112 114 proposed a generalized 2D
Wavelength / nm mapping scheme that relies on sample perturbations to
generate the correlation. They demonstrated the method
Figure 22. Derivative spectra of six patients ranging in age from
26 to 67 years. The major differences appear at 2154 and 2200 nm for many different applications.114 119 Both techniques by
due to urea absorption. The absorption spectra of these samples Bartons and Nodas groups will continue to aid in the
were obtained in the transmission mode. understanding of NIR spectra.
7.2 Imaging spectrometry and remote sensing Typical hand-held instruments incorporate narrow-band
interference filters and a single detector to acquire inten-
Imaging spectrometry, relatively new to the list of analyt- sity data. Such instruments are more compact and less
ical methods, is now technically feasible for both micro- expensive than laboratory instruments or process analyz-
scopic and earth sensing from aircraft and spacecraft. Micro ers. Unlike laboratory instruments and process analyzers,
images of cell structure can indicate composition distri- hand-held instruments are taken to the sample. Hand-held
bution within and among cells. Roberts et al.120 tested an instrument data are acquired from samples with little or no
NIR imaging spectroscopic system to identify bran, gluten sample preparation (grinding, homogenizing, etc.). Mori-
and starch in wheat. The system permitted the recording moto et al. developed a hand-held device for measuring
of images between 900 and 1900 nm in steps of 50 nm. protein in mixtures of protein and sugar and for mea-
Starch was correctly identified at 1550 nm. However, such suring nitrogen in grass tissue.124 McClure developed a
a direct study of images was not sufficient to characterize hand-held reflection spectrometer for measuring chlorophyll
all the constituents. The images, therefore, were linearly and moisture in plant tissue.125 McClure et al.126 designed,
combined by applying discriminant analyses. The more rel- constructed and tested a hand-held instrument for measur-
evant wavelengths (950, 1450 and 1500 nm) were detected ing moisture and vanillin in vanilla beans. McClure and
by using stepwise discriminant analysis. Mapping of the Stanfield127 developed a device for measuring moisture and
chemical constituents was obtained by applying canonical nicotine in tobacco leaves and tobacco products.
discriminant analysis techniques. In the segmented image, Hyvarinen and Niemela128 produced a hand-held mois-
the percentages of well-classified pixels were 92% for bran, ture analyzer that utilized an integrated two-channel detec-
95% for gluten and 99% for starch. tor and an electrically modulated source. The design, pistol-
Near-field imaging systems are becoming more and more shaped and battery operated, was intended for measuring
popular.121 Harris et al.122 pointed out that the technol- moisture of granular materials. They produced more than
ogy for conducting experiments with spatial resolutions 200 devices for the Finnish fuel peat production fields for
exceeding the diffraction limit is already available. Not yet determining optimum gathering time.
considered to be routine measurement techniques, super- The traditional deterrent to hand-held technology has
resolution measurements can be expected to expand rapidly. been the lack of capital built into the selling price to cover
Remote sensing, stimulated by improved low-noise the cost of maintenance of such devices. However, with
matrix sensors, finds application in earth sensing. For exam- the coming of Internet enhancements it may become trivial
ple, Goetz et al.123 wrote about using imaging spectrometry to maintain such low-cost devices. Both instrument checks
for sensing minerals in soil and for vegetation mapping. and calibration maintenance could be performed daily if
Even as early as 1985, technology was available for acquir- necessary with little input from the manufacturers.
ing laboratory-like spectra (images simultaneously for more
than 200 contiguous spectral bands) remotely from aircraft
and spacecraft. NASA, in the absence of military chal- DEDICATION
lenges, has turned its arsenal of fly-over spectrometers on
agricultural and ecological applications. The authors dedicate this article to their mentor and friend,
Karl H. Norris, whose leadership guidance and encourage-
ment over the years made this writing possible.
7.3 Hand-held spectrometry
For more than 50 years, NIR spectroscopy has evolved pri- ABBREVIATIONS AND ACRONYMS
marily into process and laboratory instruments. Process
analyzers (sometimes characterized as in-line or on-line) BSA Bovine Serum Albumin
operate in hostile environments whereas laboratory instru- NIRS Near-infrared Reflection Spectroscopy
ments are designed to function in controlled environments.
As such, laboratory instruments are neither portable nor
ruggedized for harsh environments. REFERENCES
Driven by the need to make quick checks on the com-
position and quality of samples at locations remote to the 1. G.W. King, Spectroscopy and Molecular Structure, Holt,
Rhinehart and Winston, New York (1965).
laboratory or process line, researchers have taken an interest
in developing constituent-specific instruments, much like 2. W. Herschel, Philos. Trans. R. Soc. London, 90, 255 (1800).
miniaturized pH probes, that can be implemented anywhere 3. W. Herschel, Philos. Trans. R. Soc. London, 90, 293 (1800).
with little effort. 4. W. Herschel, Philos. Trans. R. Soc. London, 90, 284 (1800).
5. W.F. McClure, R.E. Williamson and A.M.C. Davies, Near- and Near-infrared Reflectance: a Cost Study, Department of
infrared Instrumentation: Status, Trends and Future Con- Agriculture Federal Grain Inspection Service, Washington,
cepts, in Proceedings of the 2nd International NIRS Con- DC (1980).
ference, Tsukuba, Japan, eds M. Iwamoto and S. Kawano, 32. H. Kamishikiryo, K. Hasegawa, H. Takamura and
Korin Publishing, Tsukuba, 6579 (1989). T. Matoba, J. Food Sci., 57, 1239 (1992).
6. E.D. Archer, Lubrication, 55, 13 (1969).
33. R.T. Holman and P.R. Edmondson, Anal. Chem., 28, 1533
7. E.D. Archer, Lubrication, 55, 25 (1969). (1956).
8. B.G. Osborne and T. Fearn, Near Infrared Spectroscopy in 34. H. Takamura, N. Hyakumoto, N. Endo, T. Matoba and
Food Analysis, John Wiley & Sons, New York (1986). T. Nishiike, J. Near Infrared Spectrosc., 3, 219 (1995).
9. W. Kaye, Spectrochim. Acta, 6, 257 (1954). 35. D.P. Law and R. Tkachuk, Cereal Chem., 54, 256 (1977).
10. W.C.C. Kaye, C. Canon and R.G. Davaney, J. Opt. Soc. 36. K.H. Norris, Agric. Eng., 45, 370 (1964).
Am., 41, 658 (1951).
37. K.H. Norris and J.R. Hart, Direct Spectrophotometric
11. H.A. Willis and R.G.J. Miller, Spectrochim. Acta, 14, 119 Determination of Moisture Content of Grain and Seeds, in
(1959). Proceedings of the International Symposium on Humid-
12. K.B. Whetsel, Appl. Spectrosc. Rev., 2, 1 (1968). ity and Moisture: Principles and Methods of Measuring
13. O.H. Wheeler, Chem. Rev., 59, 629 (1959). Moisture in Liquids and Solids, Washington, DC 1963, ed.
J.P.N. Winn, Rheinhold, New York, 1925 (1965).
14. K.H. Norris and W.L. Butler, IRE Trans. Biomed. Electron.,
8, 153 (1961). 38. D.P. Law and R. Tkachuk, Cereal Chem., 54, 874 (1977).
15. K.H. Norris, Trans. ASAE, 7, 240 (1964). 39. D.R. Massie and K.H. Norris, Trans. ASAE, 8, 598 (1965).
16. W.F. McClure, Anal. Chem., 66, 43A (1994). 40. J.W. Ellis and E.L. Kinsey, Phys. Rev., 54, 599 (1938).
17. I. Murray and P.C. Williams, Chemical Principles of Near- 41. J.D. Bath and J.W. Ellis, J. Phys. Chem., 45, 204 (1941).
infrared Technology, in Near-infrared Technology in the 42. A. Elliott, Proc. R. Soc. London, Ser. A, 211, 490 (1952).
Agricultural and Food Industries, eds P. Williams and
K. Norris, American Association of Cereal Chemists, St. 43. R.D. Fraser, Nature (London), 176, 358 (1955).
Paul, MN, 1734 (1987). 44. K.T. Hecht and D.L. Wood, Proc. R. Soc. London, Ser. A,
18. P. Williams and K. Norris (eds), Near-infrared Technology 235, 174 (1956).
in the Agricultural and Food Industries, American Associ- 45. V.N. Nikitin, Vestnik Leningrad Univ., 5, 33 (1950).
ation of Cereal Chemists, St. Paul, MN (1987).
46. V.P. Korelova, Leningrad Lesotekh. Akad., 2, 167 (1950).
19. T. Bellamy, The Infra-red Spectra of Complex Molecules,
Chapman & Hall, London (1975). 47. J.A. Mitchell, C.D. Bockman and A.V. Lee, Anal.Chem.,
29, 499 (1957).
20. W.F. McClure, NIR News, 5, 14 (1995).
48. R.F. Goddu and D.A. Delker, Anal. Chem., 32, 140 (1960).
21. T. Hirschfeld, Appl. Spectrosc., 39, 740 (1985).
22. T. Begley, Proc. Annu. Users Conf. Pacific Scientific, Rock- 49. I. Glatt and J.W. Ellis, J. Chem. Phys., 19, 449 (1951).
ville, MD, 3 (1984). 50. I. Glatt, D.S. Webber, C. Seaman and J.W. Ellis, J. Chem.
23. J.R. Collins, Phys. Rev., 26, 771 (1925). Phys., 18, 413 (1951).
24. A. Roentgen, Wied. Ann., 45, 91 (1892). 51. P. Larose, Can. J. Chem., 39, 2394 (1961).
25. B. Bousfield and G. Lowry, Proc. Faraday Soc., 6, 85 52. R.G. Miller and H.A. Willis, J. Appl. Chem., 6, 385 (1956).
(1910). 53. S.E. Krikorian and M. Manpour, Spectrochim. Acta, Part A,
26. J.G. Bayly, V.B. Kartha and W.H. Stevens, Infrared Phys., 29, 1233 (1973).
3, 211 (1963). 54. R.D. Fraser, J. Chem. Phys., 24, 89 (1956).
27. R.E. Miller, Adv. Mol. Vib. Collision Dyn., 1A, 83 (1991). 55. B.G. Osborne and S. Douglas, J. Sci. Food Agric., 32, 328
28. J.A. Curcio and C.C. Petty, J. Opt. Soc. Am., 41, 302 (1981).
(1951). 56. K.J. Kaffka et al. Acta Aliment., 11, 253 (1982).
29. H. Maeda, Y. Wang and Y. Ozaki, Temperature-dependent 57. T. Hymowitz, J.W. Dudley, F.I. Collins and C.M. Brown,
Near-infrared Spectral Variations of Alcohols: Chemomet-
Crop Sci., 14, 713 (1974).
rics and Spectroscopic Analysis (Poster), in Proceedings
of the 8th International Conference on Near-infrared Spec- 58. B.S. Miller, Y. Pomeranz, W.O. Thompson, T.W. Nolan and
troscopy, Essen (Posters), eds H.W. Siesler and A. Koch, J.W. Hughes, Cereal Foods World, 23, 198 (1978).
Haus der Tecnik, Essen, 101 (1997). 59. F.A. Lee, Basic Food Chemistry, AVI, Westport, CT
30. M. Tanaka, A. Shibata, N. Hayashi, T. Kojima, H. Maeda (1975).
and Y. Ozaki, J. Near Infrared Spectrosc., 3, 203 (1995). 60. K. Ikegaya, M. Iwamoto, J. Uozumi and R.K. Cho, Deter-
31. P.E. Parker, F.F. Niernberger and United States Federal mination of Chemical Compositions of Japanese Green Tea
Grain Inspection Service, Oil Content Analysis of Sun- by Near Infrared Spectroscopy, in Proceedings of the
flower by Nuclear Magnetic Resonance, Solvent Extraction, 2nd International NIRS Conference, Tsukuba, Japan, eds
M. Iwamoto and S. Kawano, Korin Publishing, Tsukuba, 84. I. Ben-Gera and K.H. Norris, Isr. J. Agric. Res., 18, 117
353357 (1989). (1968).
61. T.C. Tso, Production, Physiology and Biochemistry of 85. N.S. Ahmed and F.R. Helal, Milchwissenschaft, 32, 272
Tobacco Plant, Ideals, Beltsville, MD (1990). (1977).
62. W.F. McClure, A. Hamid, F.G. Giesbrecht and W.W. 86. D.L. Biggs, J. Assoc. Off. Anal. Chem., 61, 1015 (1978).
Weeks, Appl. Spectrosc., 38, 322 (1984). 87. K.I. Ereifej, P. Markakis, G. Charalambous and G. Inglett,
63. F.G. Giesbrecht, W.F. McClure and A. Hamid, Appl. Spec- Instrumental Analysis of Foods: Recent Progress,
trosc., 35, 210 (1981). 237242, Vol. 2 (1983).
64. A.M.C. Davies and C.E. Miller, Appl. Spectrosc., 42, 703 88. R.G. Black, Analysis of Raw Milk by NIR, in Pro-
(1988). ceedings of the NIR84: International Symposium on Near-
infrared Reflectance Spectroscopy, Melbourne, Victoria,
65. Y.R. Chen, B. Park, M. Nguyen and R.W. Huffman, Proc. Australia, eds D. Miskelly, D.P. Law and T. Clucas, Royal
SPIE, 2786, 121 (1996). Australian Chemical Institute, Parkville, Victoria, 105113
66. Y.R. Chen, Proc. SPIE, 1836, 46 (1993). (1984).
67. Y.R. Chen, et al., J. Food Process Eng., 21, 33 (1998). 89. P. Robert, D. Bertrand, M.F. Devaux and R. Grappin, Anal.
68. S. Kawano, H. Watanabe and N. Iwamoto, Measurement of Chem., 59, 2187 (1987).
Sugar Contents in Intact Peaches by NIRS, in Proceedings 90. G.P. Downey, R.D. Bertrand and P.M. Kelly, Appl. Spec-
of the 2nd International Conference on Near Infrared Spec- trosc., 44, 150 (1990).
troscopy (ICNIRS), Tsukuba, Japan, eds M. Iwamoto and 91. R. Giangiacomo, F. Braga and C. Galliena, Use of NIR
S. Kawano, Korin Publishing, Tsukuba, 343352 (1990). Spectroscopy to Detect Whey Powder Mixed with Milk
69. S. Kawano, T. Sato and M. Iwamoto, Determination of Powder, in Making Light Work: Advances in Near-
Sugars in Satsuma Orange Using NIR Transmittance, in Infrared Spectroscopy (Proceedings of the 4th International
Making Light Work: Advances in Near-infrared Spec- Conference on Near Infrared Spectroscopy), Aberdeen,
troscopy (Proceedings of the 4th International Conference Scotland, eds I. Murray and I. Cowe, VCH, Weinheim,
on Near Infrared Spectroscopy), Aberdeen, Scotland, eds 399407 (1992).
I. Murray and I. Cowe, VCH, Weinheim, 387393, Vol. 1 92. Y.I. Posudin and V.I. Kostenko, Izv. Vyssh. Uchebn. Zaved.,
(1992). Pishch. Tekhnol., 3, 66 (1993).
70. S. Kawano, Bunseki, 1, 41 (1994). 93. R. Sueyasu, K. Kitamura and H. Kashiwa, Jpn. Kokai Tokkyo
71. S. Kawano, K. Takehara, T. Sato and M. Iwamoto, Deter- Koho, Patent No. H05-273124 (1993).
mination of Sucrose Content in Sugarcane Juice by NIR 94. S. Atanassova, Anal. Lab., 4, 232 (1995).
Spectroscopy, in Proceedings of the Third International 95. Y. Ren, B. Zhang, J. Lu, B. Bing and K. Wang, Yingyong
Conference on Near-infrared Spectroscopy (ICNIRS) Brus- Huaxue, 13, 109 (1996).
sels, Belgium, eds R. Biston and N. Bartiaux-Thill, Agri-
cultural Research Center, Gembloux, 510516 (1990). 96. R. Tsenkova, S. Atanassova, Y. Ozaki, K. Itoh and K. To-
yoda, Interrelation Between Composition and Near Infrared
72. S. Kawano, Gekkan Fudo Kemikaru, 14, 95 (1998). Spectra of Milk, Blood Plasma and Rumen Juice for
73. S. Kawano, Reito, 73, 770 (1998). Lactating Cows, in Proceedings of the 9th International
Conference on Near Infrared Spectroscopy, Verona, Italy,
74. J.Y. Chen, et al., J. Near Infrared Spectrosc., 7, 265 (1999).
eds A.M.C. Davies and R. Giangiacomo, NIR Publications,
75. W.I. Bell, Cereal Foods World, 27, 468 (1982). Chichester, 729733 (2000).
76. E.D. Dumoulin, B.P. Azais and J.T. Guerain, J. Food Sci., 97. R.T. Carl, Fresenius J. Anal. Chem., 339, 70 (1991).
52, 626 (1987). 98. R. Frankhuizen and N.G. VanderVeen, Neth. Milk Dairy J.,
77. M. Twomey, G. Downey and P.B. McNulty, J. Sci. Food 39, 191 (1985).
Agric., 67, 77 (1995). 99. J. Thomasow and M. Paschke, Milchwissenschaft, 36, 65
78. D.G. Evans, C.N.G. Scotter, L.Z. Day and M.N. Hall, J. (1981).
Near Infrared Spectrosc., 1, 3344 (1993). 100. S. Sasic and Y. Ozaki, Appl. Spectrosc., 54(9), 1327 (2000).
79. C.N.G. Scotter and L.Z. Day, The Authentication of Orange 101. D.R. Massie, Fat Measurement of Ground Beef with a
Juices Using Near Infrared Reflectance Spectroscopy, in Gallium Arsenide Infrared Emitter, in Quality Detection in
Making Light Work: Advances in Near-infrared Spec- Foods ed. J.J. Gaffney, American Society of Agricultural
troscopy (Proceedings of the 4th International Conference Engineers, St. Joseph, MI, 17, Vol. 1 (1976).
on Near Infrared Spectroscopy), Aberdeen, Scotland, eds
102. F.F. Jobsis, Science, 198, 1264 (1977).
I. Murray and I. Cowe, VCH, Weinheim, 394398 (1992).
103. J.T. Kuenstner and K.H. Norris, J. Near Infrared Spectrosc.,
80. C.N.G. Scotter and A. Legrand, Fruit Process., 5, 255
2, 59 (1994).
(1995).
104. J.T. Kuenstner, K.H. Norris and W.F. McCarthy, Appl.
81. D. Baker and K.H. Norris, Appl. Spectrosc., 39, 618 (1985). Spectrosc., 48, 484 (1994).
82. D. Baker, Cereal Foods World, 30, 389 (1985). 105. K.H. Norris and J.T. Kuenstner, Rapid Measurement of
83. J.D.S. Goulden, J. Dairy Res., 31, 273 (1964). Analytes in Whole Blood with NIR Transmittance, in
Leaping Ahead in Near-infrared Instrumentation, eds G. 113. I.A. Noda, E. Dowrey and C. Marcott, J. Mol. Struct., 224,
Batten, P.C. Flinn, L.A. Welch and A.B. Blakeney, Royal 265 (1990).
Australian Chemical Institute, North Melbourne, Victoria, 114. Y. Ozaki, Y. Liu and I. Noda, Appl. Spectrosc., 51, 526
431436 (1995). (1997).
106. R.A. Shaw, S. Kotowich, H.H. Mantsch and M. Leroux, 115. I. Noda, Y. Liu, Y. Ozaki and M.A. Czarnecki, J. Phys.
Clin. Biochem, 29, 11 (1996). Chem., 99, 3068 (1995).
107. F.E. Barton, II and D.S. Himmelsbach, Correlation Spec- 116. I. Noda and Y. Ozaki, Bunko Kenkyu, 44, 236 (1995).
troscopy Across Multiple Spectral Regions to Aid Cal-
ibration Model Interpretation, in Leaping Ahead in 117. I. Noda, Y. Liu and Y. Ozaki, Mikrochim. Acta, Suppl, 14,
with Near-infrared Spectroscopy, eds G. Batten, P.C. 609 (1997).
Flinn, L.A. Welch and A.B. Blakeney, Royal Australian 118. I. Noda, G.M. Story, A.E. Dowrey, R.C. Reeder and C. Mar-
Chemical Institute, North Melbourne, Victoria, 4852 cott, Macromol. Symp., 119, 1 (1997).
(1995).
119. Y. Ozaki, Y. Liu and I. Noda, Macromol. Symp., 119, 49
108. F.E. Barton, II and D.S. Himmelsbach, Correlation Spec- (1997).
troscopy: Near-infrared, Mid-infrared and Raman Spectral
120. P. Roberts, D. Bertrand, M.F. Devaux and A. Sire, Anal.
Regrions as Complementary Techniques for the Study of
Chem., 64, 664 (1992).
Wheat Properties, in Near Infrared Spectroscopy: The
Future Wave, eds A.M.C. Davies and P. Williams, NIR 121. B.D. Boudreau, J. Raja, R.J. Hocken, S.R. Patterson and
Publications, Chichester, 2631 (1996). J. Patten, Rev. Sci. Instrum., 68, 3096 (1997).
109. F.E. Barton, II, D.S. Himmelsbach and R.B. Russell, Appl. 122. T.D. Harris, R.D. Grober, J.K. Trautman and E. Betzig,
Spectrosc., 47, 1920 (1993). Appl Spectrosc., 48, 14A (1994).
110. F.E. Barton, II, D.S. Himmelsbach, J.H. Duckworth and 123. A.F.H. Goetz, G. Vane, J.E. Solomon and B.N. Rock, Sci-
M.J. Smith, Appl. Spectrosc., 46, 420 (1992). ence, 228, 1147 (1985).
111. F.E. Barton, II and D.S. Himmelsbach, Near Infrared 124. S. Morimoto, W.F. McClure and D.L. Stanfield, Appl. Spec-
Reflectance Spectroscopy and Other Spectral Analyses, in trosc., 55, 182 (2001).
Proceedings of the Second Conference on Spectroscopy 125. W.F. McClure, D. Moody, O. Kinoshita, S. Morimoto and
Across the Spectrum: Techniques and Applications of D.L. Stanfield Appl. Spectrosc., submitted (2001).
Analytical Spectroscopy, Hatfield, UK, eds A.M.C. Davies
and C.S. Creaser, Royal Society of Chemistry, Cambridge, 126. W.F. McClure, C.M. Hargrove, M. Zapf and D.L. Stanfield,
240247 (1991). unpublished results.
112. I. Noda, A.E. Dowrey, C. Marcott, G.M. Story and Y. Ozaki, 127. W.F. McClure and D.L. Stanfield, unpublished results.
Appl Spectrosc., 54, 236A (2000). 128. T. Hyvarinen and P. Niemela, Proc. SPIE, 1266, 99 (1990).
Far-infrared Spectroscopy
Peter R. Griffiths
1 THE FAR-INFRARED REGION is split into two paths allows the region from 4000 to
400 cm1 to be covered with fairly high efficiency with
Spectral regions are defined both by the nature of the no change of optical components. However, the region
instrumentation used for the measurement and the types between 400 and 200 cm1 is inaccessible with these
of transitions that are accessible. For example, not only are instruments because KBr cuts off completely at about
the sources of continuous radiation for near-infrared (NIR) 350 cm1 . Although beamsplitters fabricated with cesium
and mid-infrared (MIR) spectrometry different, but different iodide as the substrate and compensator plate allow spec-
types of instruments are frequently used to encode the tra to be measured to a little below 200 cm1 (since CsI
spectrum. Similarly, the strongest bands that absorb MIR cuts off at about 150 cm1 ), they have poor performance at
radiation are fundamental vibrational modes while all bands high wavenumber. Thus, because most MIR FT-IR spec-
that absorb in the NIR are overtones and combinations. trometers cover the region from 4000 to 400 cm1 , the FIR
It is less easy, however, to make such facile comparisons region is now generally thought of as starting at 400 cm1 .
between MIR and far-infrared (FIR) spectroscopy. Even The long wavelength end of the FIR region corresponds
the definition of the FIR region of the spectrum has been to the point that optical spectrometers no longer function
a subject of some controversy. In 1962, an international well and microwave technology is used to measure spectra.
commission for spectroscopy1 defined the MIR as the This wavelength is approximately 1 mm, and so the low-
region between 2.5 and 50 m (4000 and 200 cm1 ) since wavenumber limit of the FIR spectrum is 10 cm1 . Thus
that was the region covered by the better MIR grating the FIR was defined as the region ending at 10 cm1 .1
spectrometers at that time. By default, therefore, the start However, the long wavelength of the FIR is just as ill-
of the FIR region was generally thought of as 50 m, or defined as the short wavelength end, since FIR spectra
200 cm1 , and was so defined by this commission. Almost have been measured at longer wavelength than 1 mm using
40 years later, however, even this definition has changed. interferometers and to shorter wavelength using microwave
Although many grating spectrometers that were com- technology. Nonetheless, the FIR spectrum is generally
mercially available in the early 1960s covered the region understood to be the region between 400 and 10 cm1 (25
from 4000 to 200 cm1 , they required at least two grat- and 1000 m).
ing changes in the middle of the scan to do so. The
fact that Fourier transform infrared (FT-IR) spectrometers
have superseded grating monochromators as the instru- 2 TRANSITIONS THAT OCCUR IN THE
ment of choice for the measurement of MIR spectra has FAR-INFRARED
led to a redefinition (albeit unofficial) of the MIR spec-
trum. MIR Fourier transform spectrometers usually incor- Whereas the only transitions that are usually observed in
porate beamsplitters with a KBr substrate and compensator the MIR spectrum are intramolecular vibrational modes,
plate. The multilayer dielectric film at which the beam many different types of transitions are of sufficiently low
energy that they occur in the FIR spectrum. For organic
John Wiley & Sons Ltd, 2002. compounds, they include intramolecular stretching modes
involving heavy atoms, skeletal bending modes involving On the other hand, if X and Y are both heavy atoms, then
the entire molecule, torsional modes, ring puckering vibra- the reduced mass is much greater than one, and the effect is
tions of small-ring molecules, and intermolecular vibra- to reduce the frequency of the stretching mode significantly.
tions of hydrogen-bonded molecules and charge-transfer If we assume that the force constant of 127 I79 Br is the same
complexes. as that of HCl, the effect of the reduced mass (m D 48.704)
For inorganic and organic solids, as well as polymers, is to reduce its vibrational frequency compared to that
crystal lattice modes (phonon bands) are found in the FIR of 1 H35 Cl by a factor of (48.704/0.9722)1/2 , to 422 cm1 .
spectrum. Photon ionization spectroscopy of semiconduc- In fact, the vibrational fundamental is found in the FIR
tors is also investigated in this region. Some representative at 268.4 cm1 , showing that the force constant for IBr is
examples of these types of transitions will be shown in this significantly less than that of HCl.
article. For polyatomic molecules, the effective reduced mass
Finally, pure rotational transitions of light molecules is quite difficult to calculate, because no vibrational mode
extend into the FIR from the microwave region. The best can be described purely in terms of the stretching of two
known of these are the rotational lines of atmospheric water bonds. For example, some inorganic chemists have tried to
vapor. These lines are so strong that even a good purge assign particular bands in the FIR spectrum of complexes
is usually insufficient to prevent them from interfering to a pure metalligand stretch, but the motion is always far
with measurements in the FIR, so that almost all FIR more complicated that this. It is the opinion of this author
spectrometers are operated under vacuum. Not all rotational that the conclusions made by workers who have tried to
lines due to atmospheric species are a bad thing, however. assign bands in the FIR to a pure metalligand stretching
For example, some pure rotational transitions have allowed mode are frequently in error. In the last part of the 1990s,
further light to be shed on processes that are taking place programs that calculate the normal modes of molecules with
in the stratosphere. up to 20 or 30 atoms using density functional theory have
Brief descriptions of each of these transitions will be become commercially available.2 Thus it is now a relatively
given in this article. This is not intended to be a compre- easy matter to calculate all (3N 6) normal modes of any
hensive summary. A few representative examples of the inorganic species with fewer than about 12 atoms and see
types of transitions that occur in the FIR region will be for oneself whether the assignment of any band in the
given. FIR spectrum to the specific motion of just two atoms is
correct or not.
Vibrational frequencies can be divided into two broad
3 SMALL ORGANIC AND INORGANIC categories, group frequencies and fingerprint frequencies.
COMPOUNDS Group frequencies are characteristic of small groups of
atoms (e.g. OH, CN, CDO, CONH2 , etc.) and the
3.1 Intramolecular stretching modes involving
vibration is largely localized with that group. Finger-
heavy atoms
print frequencies are due to less localized motion and are
highly characteristic of the specific molecule. Since low-
Assuming simple harmonic motion, the fundamental stret-
frequency modes of polyatomic molecules usually involve
ching frequency, nQ e of a diatomic molecule, XY, is given
large amplitude motion of many or all of the atoms in the
by equation (1): molecule, it is highly unlikely that any stretching mode
k
nQ e D 2p 1 found in the FIR will be a group frequency.
m
where k is the stretching force constant and m is the reduced
mass of the molecule, i.e. mX mY /mX C mY , where mX 3.2 Skeletal bending modes
and mY are the masses of X and Y, respectively. If X is
hydrogen (1 H), and Y is significantly heavier (say 35 Cl), The highest frequency bending mode of any functional
then m is approximately equal to unity (actually 0.9722 group is usually no more than one-half the frequency of
for 1 H35 Cl). If X is deuterium (2 H),pthe fundamental
p fre- the corresponding stretching mode. For example, the sym-
quency is reduced by approximately 2 (actually 1.9459 metric and antisymmetric stretching modes of aliphatic CH2
for 2 H35 Cl) if k is unchanged. In practice, the fundamen- groups are found at 2926 and 2853 cm1 , whereas their
tal frequency of the 1 H35 Cl and 2 H35 Cl stretching modes highest-frequency bending mode (the in-plane deforma-
are 2990 and 2143 cm1 , respectively, so that the stretch- tion) absorbs at 1460 cm1 . Bond deformations involving
ing force constant is approximately the same for the two heavier atoms than hydrogen or deuterium often absorb in
isotopomers. the FIR. For example, the CCl stretching mode typically
Far-infrared Spectroscopy 3
absorbs near 700 cm1 and CBr and CI stretching modes

absorb at even lower wavenumber (longer wavelength).
Thus, it would be expected that CCl, CBr and CI bend- (a)
ing modes will absorb below 400 cm1 . These vibrations
cannot be used as group frequencies, however, because
they involve not only the motion of the carbon and halo-
gen atoms but also the motion of other lighter atoms in the
Transmittance
molecule (especially hydrogen). (b)
The more rigid the molecule, the more likely that vibra-
tions of its atoms will be coupled and the longer will be the
wavelength of its lowest frequency vibrational modes. For
example, all para-disubstituted aromatic compounds, p-X-
C6 H4 -Y, have a band that absorbs near 100 cm1 that is
caused by the in-phase motion of the substituents X and
Y above and below the plane of the aromatic ring.3 Since
all atoms in the ring move in the opposite direction, this
vibration involves all 12 atoms of the molecule (or more if
X and Y contain more than one atom).
The lowest frequency bending mode of a small molecule 50 40 30
known to this author is the skeletal bending mode of carbon Wavenumber /cm1
suboxide, ODCDCDCDO, which absorbs at 63 cm1 .4 Figure 1. FIR spectrum of trifluoroacetyl fluoride (b) with pure
rotational spectrum of water vapor above (a) to show poten-
tial interference. [Reproduced from J.R. Durig, A.R. Fanning,
3.3 Torsional and ring-puckering modes T.G. Sheenan and G.A. Giurgis, Spectrochim. Acta, 47A, 279
(1991)8 by permission of Elsevier; copyright 1991.]
There are two types of vibrational modes that absorb in the
FIR from which useful physico-chemical information can 1 cm1 . To determine the potential barrier to internal
be obtained: torsional modes and ring-puckering modes.
rotation accurately, the torsional mode should not be mixed
Thermodynamic data obtained in the 1930s showed that
with low frequency intramolecular bending modes.
rotation about single bonds is not free. The restricted motion
Analogous information can be obtained from ring puck-
around a single bond is quantized and gives rise to torsional
ering modes, which are also large-amplitude anharmonic
modes, sometimes called internal rotations. As a result,
vibrations. By analyzing the frequencies for these ring
small groups with both two-fold (C2n ) symmetry, e.g. OH
bending motions, it is possible to obtain the potential
and CHO, or three-fold (C3n ) symmetry, e.g. CH3 , bonded
to a large group such as an aromatic ring undergo librational surfaces for the different conformational isomers. For
motion with respect to the heavier anchor group. Polar example, the FIR spectrum of 1,3-disilacyclobutane is
groups, such as CHO, give rise to fairly strong bands in shown in Figure 2.9 While a one-dimensional model of the
the FIR spectrum, but the corresponding torsional modes
of nonpolar groups, such as CH3 are much weaker. The 95
spectrum yields information about the potential function of

the torsion.
Durigs group have extensively studied the potential bar- 85
riers for many small molecules by FIR spectroscopy.5 7 A
good example is the spectrum of trifluoroacetyl fluoride,
75
CF3 CFO, shown in Figure 1.8 Seven torsional transitions
can be seen in this spectrum, starting with the fundamental H2Si SiH2
(n D 1 0) at 45.65 cm1 and ending with the n D 7 6 70
transition at 33.40 cm1 . The transitions from one excited 40 60 80 100
state to another (the so-called hot bands) are surprisingly Wavenumber /cm1
intense because these modes are very anharmonic. From
Figure 2. FIR spectrum of 1,3-disilacyclobutane in the vapor-
this spectrum, it was determined that the three-fold barrier phase measured at 0.5 cm1 resolution. [Reproduced from Irwin
to internal rotation, V3 , for this molecule is 382 2 cm1 , et al. (1977)9 by permission of the American Chemical Society;
while the six-fold barrier, V6 , is much smaller at 8 copyright 1977.]
ring-puckering vibration yields reasonably accurate values 400 370

228
for the height of the potential barrier, more recent work has
316 (a)
involved the development of two- and three-dimensional 400 371
models that allow for the interaction of the ring-puckering 430 228
mode with other low-frequency vibrational modes. These (b)
401 313 262
interactions alter the levels of the potential function and, if 207
neglected, can result in poorly calculated barrier heights. (c)
400 200
3.4 Intermolecular vibrations /cm1
407
Even if the atoms involved in a vibration are light, a stretch- 199
291
ing vibration can still absorb in the FIR if the force constant, 345 (d)
k, is very low (see equation 1). Intermolecular interac-
282
tion can be very weak for the case of hydrogen-bonded
493 449 199
and charge-transfer complexes. Bands due to the stretch- (e)
ing of hydrogen bonds can absorb over a wide fre-
320
quency range, from as high as about 600 cm1 to below 459 272
249
10 cm1 . Usually hydrogen-bonded molecules in the liq- 322 (f)
uid state exist in so many conformations that the effect 385 311
406
of hydrogen bonding is to spread the absorption over a (g)
spectral region that is well over 200 cm1 wide. Thus lit-
400 300 200
tle detailed information of the nature of hydrogen bonds
/cm1
between individual molecules can be obtained from the
FIR spectrum of pure liquids and solutions. Conversely, Figure 3. FIR spectrum of the following monosaccharides:
the spectra of hydrogen-bonded solids can show relatively (a) D-arabinose; (b) L-arabinose; (c) xylopyranose (wood sugar);
(d) galactose; (e) mannose; (f) glucose; (g) levulose (D-fructo-
sharp bands that have been shown to be useful for dif- pyranose). [Reproduced from Yang et al. (2000)10 by permission
ferentiating between different monosaccharides, as shown of the editor.]
in Figure 3.10
Unlike the case for hydrogen-bonded molecules in the absorption. A polar molecule with a moment of inertia, I,
liquid state, which exist in many different conformations, within a potential well of depth, V, and semiangular aper-
the geometry of charge-transfer complexes is fairly well ture, x, due to its nearest neighbors has a harmonic libra-
defined. As a result, the stretching mode of the weak tional frequency given by equation (2):
bond between the two associated molecules can give
rise to a fairly narrow band. One early example of this p V 1/2
!0 D 2
type of vibration was reported by Lake and Thompson,11 x 2I
who investigated the interaction of iodine and substituted
For many liquids, these librational frequencies are app-
pyridines. Two bands are seen in the FIR spectrum, with
roximately 100 cm1 . These bands are also seen in the
the lower frequency band being the intermolecular stretch-
spectrum of nonpolar but polarizable liquids such as ben-
ing mode. Because the distribution of the electrons in zene, but are significantly weaker, since they depend on the
the II bond is distorted by the charge transfer, the II induced dipole.
stretch becomes allowed and is also observed in the FIR
spectrum.
4 INORGANIC SOLIDS
3.5 Liquid lattice bands 4.1 Phonon bands
Many polar and polarizable liquids show a broad absorp- At certain wavelengths, the atoms in most inorganic solids
tion band in the region below 150 cm1 . This band can move cooperatively, giving rise to so-called phonon modes,
be explained by a model of damped rotational oscillations or lattice vibrations, that generally absorb in the FIR.
(librations) of the molecules in the liquid state. These bands In ionic crystals such as NaCl or KBr, the cooperative
are sometimes known as liquid lattice bands and the motion of all the anions in one direction and the corre-
phenomenon is sometimes known as Poley-type librational sponding motion of the cations in the other can give rise
to exceptionally strong lattice modes, i.e. the absorption 100

index, k, of these bands can be very large. The reflectance
at normal incidence is given by equation (3):
n 12 C k 2 50
RD 3
n C 12 C k 2
where n is the refractive index. Since k n, the reflectance
Reflectance (%)
approaches 100% near the maximum of such lattice modes. (a)
0
The radiation reflected from ionic crystals was called rest-
strahlen (residual rays) by German scientists. This name 100
has since been applied to a strong band in the reflection
spectrum of any crystalline material. For KBr, the lat-
tice mode is centered at about 125 cm1 . For 4-mm thick
crystals, the high-wavenumber wing of this band starts to 50
absorb strongly at 400 cm1 , which accounts for the long-

wavelength cut-off of this material. Similarly the maximum
absorption of the lattice band of NaCl is at about 185 cm1 , (b)
0
and NaCl cannot be used as a MIR window material much
0 200 400 600
below 650 cm1 . Because of their very strong and localized
/cm1
reststrahlen bands in the FIR, alkali halides were com-
monly used as order-sorting filters in the earliest FIR grating Figure 4. FIR reflection spectra of a YBa2 Cu3 O6 single crystal at
spectrometers. 10 K(- - - - ) and 300 K( ) with the electric field (a) parallel
to and (b) perpendicular to the c axis of the crystal. [Reproduced
The optical properties of certain ceramic inorganic oxides from Genzel (1999)14 by permission of Springer Verlag; copyright
have become particularly important with respect to the 1999.]
phenomenon of high-temperature superconductivity, since
the optical properties have a strong influence on the elec- 150
trical properties. (As shown by Smith and Chugtai in (a)
Volume 4 of this Handbook,12 the dielectric function is 100
equal to the square of the refractive index.) Inorganic 50
solids, including these ceramic materials, often absorb FIR
2
0
radiation so strongly that it becomes necessary to mea- (b)
400
sure the reflection spectrum rather than the transmission
spectrum. The refractive index and absorption index are 200
then determined through a KramersKronig analysis of
the reflection spectrum, as discussed by Claybourn in Vol- 0
100 200 300 400 500 600 700
ume 2 of this Handbook.13 The reflectance spectra of a
/cm1
single crystal of YBa2 Cu3 O6 perpendicular and parallel to
Figure 5. Imaginary part, e2 (n), of the dielectric function of a
the c axis of the crystal at 300 K and 10 K are shown in
YBa2 Cu3 O6 single crystal calculated using a KramersKronig
Figure 4, while Figure 5 shows the imaginary part of the analysis of the reflectance spectra shown in Figure 4. [Reproduced
dielectric function of this material calculated from these from Genzel (1999)14 by permission of Springer Verlag; copyright
spectra. 1999.]
Interferometers for FIR spectrometry have been modi-
fied so that the crystal can be held in one arm, with the obtained. An interesting example of this type of measure-
sample replacing the fixed mirror. Because of the fact that ment has been reported by Genzel14 for indium antimonide.
the phase lag occurring on reflection depends on the opti- The phonon mode for InSb is found at 175 cm1 . In
cal constants, the interferogram measured in this way is addition, the onset of a plasma edge causes absorption
asymmetrical. This approach, known as dispersive Fourier below 100 cm1 that also leads to variation of both the
transform spectroscopy, allows the refractive index spec- refractive index and the absorption index. The variation
trum of the crystal to be measured directly for cases of very of reflectance and phase angle calculated from the dis-
strong absorbers. By calculating the sine and cosine Fourier persive FIR spectrum of InSb is shown in Figure 6(a)
transform of the interferogram measured in this way, both while the corresponding optical constants are shown in
the reflectance and the phase change on reflection can be Figure 6(b).
Reflectance Ir I2 techniques. Instead, if the temperature is slightly elevated

Ir I2 +
Phase angle
1.0 to between 10 and 20 K, an existing photon might lift
r =
0.8 r the electron from the first excited state to the conduc-
0.6 tion band, causing a photo-conducting signal. Even for
r Ir I2
0.4 extremely low impurity levels, this signal can now be
0.2
0.0
detected by the change in current through the sample, rather
0 50 100 150 200 250 300 than by the absorption of FIR radiation. The photother-
(a) Wavenumber / cm1 mal ionization spectrum of germanium at 10 K, which is
reproduced in Figure 7, shows some hydrogen-like series
10.0 of lines caused by shallow impurities of aluminum and
5.0
k boron.
k
2.0 n n
n, k
1.0 5 POLYMERS
0.5
Bands in the FIR spectra of polymers can be due to
0.2 both intra- and intermolecular modes. Intramolecular vibra-
0 50 100 150 200 250 300
tions can include both localized torsions of functional
(b) Wavenumber / cm1
groups and modes that are more delocalized along the
Figure 6. Dispersive Fourier transform spectra of InSb. (a) Ref- chain, while intermolecular modes include rotatory and
lectance and reflection phase. (b) The refractive index spectrum
translatory motion of the chains. In addition, the absorp-
nn and the absorption index kn derived from the reflection
spectrum. [Reproduced from Genzel (1999)14 by permission of tion of FIR radiation can arise from disorder-induced
Springer Verlag; copyright 1999.] absorption and absorption by relaxational or non-resonant
phenomena. Finally, absorption may be caused by impu-
rities accidentally or deliberately present in the polymeric
4.2 Photothermal ionization spectroscopy material.
The position of skeletal vibrational bands in the region
Photothermal ionization spectroscopy is a very sensitive between 150 and 300 cm1 is, of course, dependent on the
method for detecting shallow impurities in semiconductors chemical structure of the polymer but can also depend
such as silicon or germanium. At low temperatures, only on other parameters such as the dipole moment, the
the ground state is occupied. A photon can excite the elec- repeat mass unit and the cohesion energy, Ecoh . For
tron from the ground state to a conduction band, causing example, Bershtein and Rhyzhov15 found that the position
the onset of photconductivity. Since the density of impu- of the maximum of the low-frequency skeletal mode,
rities is only 109 to 1012 cm3 , however, this extremely nsk , was proportional to (Ecoh )1/2 for polystyrene (PS),
weak absorption cannot be detected by conventional optical poly(methyl methacrylate) (PMMA), poly(vinyl chloride)
(PVC), poly(acrylonitrile) (PAN), poly(chlorostyrene)
(PCS), poly(vinyl fluoride) (PVF) and poly(vinyl acetate)
Photoconductivity (arb. units)
(PVA), see Figure 8. This proportionality shows that the

force constant of the low-frequency skeletal motion of the
backbone chain depends on the non-bonded interactions
between the nearest units of both the chain whose skeletal
motion gives rise to this band and the neighboring chains.
Hydrogen bonding is usually considered the strongest
secondary force in the solid state of polymers and the speci-
ficity and magnitude of this type of interaction may strongly
influence the chain interaction and packing. The direct anal-
50 60 70 80 90 100 ysis of the low-frequency vibrations of hydrogen-bonded
Wavenumber / cm1 polymers is rare, but can be very useful. For example, Frank
and Fiedler16 showed that the FIR spectra of a series of
Figure 7. Photothermal ionization spectrum of germanium at
10 K showing some hydrogen-like line series caused by shal- polyamides were dependent on the number of CH2 groups
low aluminum and boron impurities. [Reproduced from Genzel between the amide groups and the way that the chains were
(1999)14 by permission of Springer Verlag; copyright 1999.] packed. Polyamides occur in two conformations forming
(a)
5
300
() / cm1
4 1 400
1 200
() / cm1
240
7
100
2
200
sk (cm1)
0 2
200
3 120
3 0
80
160 6
(b)
40
8 300
0
4 200
2 4 6
100
Ecoh1/2 (kJ mol1)1/2
Figure 8. Plot of nsk against (Ecoh )1/2 for: (1) PS; (2) PMMA; 100 5 0
(3) PVC; (4) PAN; (5) PCS; (6) PVF; and (7) PVA. The point
marked (8) is the calculated value for polyethylene (PE). [Re-
50
drawn from Bershtein and Rhyzhov (1984).15 ]
6 100
very different crystalline modifications named a and g. In 0
the a conformation, the molecules form an all-trans chain, 50
which is only possible if the chains are antiparallel. For
g-polyamides, the plane of the CONH group is perpen- 0
dicular to that of the CH2 groups and in this form the 100 200 300
neighboring chains are parallel. It was found that Nylons / cm1
2, 4 and 6 are usually in the a-conformation, while Nylons Figure 9. Absorption spectra of Nylons in the (a) a and (b) g
8, 10 and 12 occur in the g form. This leads to signifi- conformations: (1) Nylon 4; (2) Nylon 6; (3) Nylon 2; (4) Nylon
cant differences in their spectra, as shown in Figure 9. It is 8; (5) Nylon 10; and (6) Nylon 12. [Redrawn from Frank and
noteworthy that the MIR spectra of Nylons are so similar Fiedler (1979).14 ]
that it is often difficult to distinguish between them on the
basis of their MIR spectra. This is obviously not the case effect of varying the temperature of low-density and lin-
for their FIR spectra. ear high-density PE was reported by Frank and Leute15
The peak observed in the range of 100110 cm1 has and their results are summarized in Figure 10. The low-
been assigned to a translational vibration of the amide density PE had 40 branches per 1000 carbon atoms, which
group. The shift of this band to 116 cm1 in Nylon 2 resulted in the presence of side chains of different lengths.
(polyglycine) can be understood because only one CH2 The branches inhibit crystallization, so that the volume
group is centered between the two amide groups. When the fraction that is in the crystalline form is only 40%.
polymer is subjected to an elastic strain or after inelastic For the high-density PE, there were only 3 branches
prestraining, this band shifts to lower frequency. The other per 1000 carbon atoms, and 80% is in the crystalline
bands that absorb between 100 and 200 cm1 are assigned form.
to oscillations of the hydrogen bond. It is known that Goldstein et al.16 reported that even more information on
hydrogen bonds can form planar cyclic dimers and three of polymer crystallinity can be obtained from the FIR spec-
the six hydrogen-bond vibrational modes of these dimers trum of isotactic polypropylene (iPP), where four bands are
(the torsion, antisymmetric deformation and antisymmetric seen (Figure 11). With increasing crystallinity, some nar-
stretching) are infrared active. rowing of the bands is observed and the higher intermolecu-
Far-infrared spectroscopy allows the crystallinity of poly- lar interactions within the crystalline regions of the polymer
mers to be examined. Perhaps the best known lattice causes some of the bands to split. It is now well established
mode of any polymer is that of PE at 72 cm1 . The that the most general contribution in the FIR spectra of
() / cm1
5
() /cm1
4 4
3 3
2 2
1 1
100
100
200
)
T (K
)
200
T (K
300
300
50 70 90 50 70 90 110
(a) / cm1 (b) / cm1
Figure 10. Change in the FIR spectra of (a) low-density PE and (b) linear high-density PE with temperature. [Redrawn from Frank
and Leute (1983).17 ]
disordered polymers in the region below 150 cm1 orig-

inates from the broad-band absorption caused by liquid
lattice or Poley-type absorption (vide infra) and the sharp
lattice bands due to the crystalline polymer are often super-
(a) imposed on this broad absorption. Thus on decreasing the
degree of crystallinity of iPP, the intensity of the bands due
to crystalline iPP decreased and the background absorption
related to the disordered regions increased.
A review of FIR spectroscopy of polymers that is more
(b)
extensive than this brief summary has been written by
Bershtein and Rhyzhov.19
Absorbance
(c) 6 PURE ROTATION SPECTRA
Although the BornOppenheimer approximation is that the

electronic, vibrational and rotational states of molecules are
separated by a sufficient energy that they can be considered
(d) independent of one another,20 both vibrational and rota-
tional transitions can be observed in the FIR region. The
lighter the molecule, the higher is its rotational energy. Thus
whereas microwave spectroscopy is needed to observe the
transitions between rotational states of large molecules, the
(e) corresponding transitions of small molecules can usually
0 100 200 300 400 be observed in the FIR. In addition, the more symmetri-
/cm1 cal the molecule, the simpler is its pure rotation spectrum.
Figure 11. FIR spectra of iPP with different degrees of crys- Thus because the principal moments of inertia of water are
tallinity: (a) 52%; (b) 58%; (c) 66%; (d) 70% and (e) 75%. [Re- all different and because the molecular weight is only 18,
drawn from Goldstein et al. (1973).18 ] the rotational spectrum consists of a plethora of unequally
6 56
11
8
7 9 10
12
14 14 15 17 30
13 19
16 18 21
20

23 24 43
22 25 28
29
26 31
27
35 42
33 39 55
32
34 38
40
37 45 57
41 47
36 44 50 54
49
46 51 58 63
48
5253
62 6465
61
Emission increasing
10 11 12 13 14 15 16 17 18 19 20
(a)
96
109
110
118
116117 119
99 112 114
120
108 113 115

111b
105
106 111a
107
89 95 102
100b 104

103

98 101
77 100a
94 97

88 93
92
91
84 90
76 86
80
87
74 79 83 85
67 75 82
73
69
81
78 80
68 72
63
64 71
62 65 70
61
66
20 21 22 23 24 25 26 27 28 29 30
(b) Wavenumber /cm1
Figure 12. FIR emission spectrum of the stratosphere measured at a resolution of 0.067 cm1 with a phase-modulated NPL cube
interferometer and an InSb detector, mounted in a Concorde airliner. All the numbered lines have been assigned to species in the
stratosphere. The lines marked with asterisks are those used for determining the concentration of HNO3 , NO2 , N2 O and SO2 . [Reproduced
from Harries (1972);22 crown copyright is reserved.]
spaced lines across the entire FIR from 10 to 400 cm1 . This ABBREVIATIONS AND ACRONYMS
is most unfortunate, as the ubiquitous presence of atmo-
spheric water vapor means that FIR spectrometers must be iPP Isotactic Polypropylene
evacuated to avoid interference by the rotational lines of PAN Poly(acrylonitrile)
water vapor. PCS Poly(chlorostyrene)
As the molecular weight of a molecule is increased, its PE Polyethylene
rotational transitions shift towards the microwave region. PMMA Poly(methyl methacrylate)
Thus it is rare that the pure rotation spectrum of molecules PS Polystyrene
with many more than six atoms can be measured in the PVA Poly(vinyl acetate)
FIR. However, transitions between very high rotational PVC Poly(vinyl chloride)
states of large molecules can be observed without too much PVF Poly(vinyl fluoride)
difficulty due to their high Boltzmann population at ambient
temperature. Many years ago, this author measured the pure REFERENCES
rotation spectra of CDCl3 , CDBr3 and CDI3 .21 Because the
rotational constant is so small, the lines in the pure rotation 1. Triple Commission for Spectroscopy, J. Opt. Soc. Am., 52,
spectrum of CDI3 are separated by only about 0.4 cm1 476 (1962).
and yet transitions between rotational states as high as 2. H. Yoshida and H. Matsuura, Calculation of Vibrational Fre-
J D 80 ! 81 could be observed. quencies by HartreeFock-based and Density Functional
The fact that pure rotational energy levels are so small Theory, in Handbook of Vibrational Spectroscopy, eds
J.M. Chalmers and P.R. Griffiths, John Wiley & Sons, Chich-
means that the Boltzmann population of excited rotational ester, 20122028, Vol. 3 (2002).
states may be quite high even when the temperature is
3. P.R. Griffiths and H.W. Thompson, Proc. R. Soc. London,
well below 0 C. For example, when the first supersonic Ser. A, 298, 51 (1967).
transports were being built, there was considerable con- 4. F.A. Miller, D.H. Lemmon and R.E. Witkowski, Spectrochim.
cern about the gases emitted by these planes. One way that Acta, 21, 1709 (1965).
was successfully used to estimate the concentration of sev- 5. P. Groner, J.F. Sullivan and J.R. Durig, in Vibrational Spec-
eral molecules of concern was the measurement of their tra and Structure, ed. J.R. Durig, 405487, Vol. 9, Elsevier,
FIR emission spectrum.22 The emission spectrum of the Amsterdam (1981).
stratosphere measured with a phase-modulated interferome- 6. J.R. Durig and A.W. Cox, in Fourier Transform Spec-
ter equipped with a liquid helium cooled InSb detector, with troscopy: Applications to Chemical Systems, eds J.R. Fer-
the equipment mounted in a prototype Concorde airliner, raro and L.J. Basile, 215274, Vol. 1, Academic Press, New
York (1978).
is shown in Figure 12. The lines marked by asterisks are
those used for estimating the concentrations of HNO3 , NO2 , 7. J.R. Durig, Far IR Spectroscopy, Applications, in Ency-
clopedia of Spectroscopy and Spectrometry, ed. J. Lindon,
N2 O and SO2 . The resolution (0.067 cm1 ) and sensitivity Academic Press, New York, 498504 (1999).
of these measurements, which were taken over 30 years
8. J.R. Durig, A.R. Fanning, T.G. Sheenan and G.A. Giurgis,
ago, is quite remarkable even today. Spectrochim. Acta, 47A, 279 (1991).
9. R.M. Irwin, J.M. Cooke and J. Laane, J. Am. Chem. Soc., 99,
3273 (1977).
7 SUMMARY 10. L. Yang, S. Weng, L. Yang and J. Wu, Spectrosc. Spectral
Anal., 20, 189 (2000).
Far-infrared spectroscopy was probably in its heyday bet- 11. R.F. Lake and H.W. Thompson, Proc. R. Soc. London, Ser.
ween 30 and 40 years ago, when Gebbies group at the A, 297, 440 (1967).
National Physical Laboratory in Teddington, Middlesex, 12. D.M. Smith and A.R. Chughtai, Determination of Optical
UK demonstrated some quite remarkable measurements. and Dielectric Properties of Polymers by Vibrational Spec-
(The spectrum shown in Figure 12 exemplifies the type of troscopy, in Handbook of Vibrational Spectroscopy, eds
J.M. Chalmers and P.R. Griffiths, John Wiley & Sons, Chich-
spectra that were measured by this group.) Today, far fewer ester, 25232532, Vol. 4 (2002).
FIR spectrometers are being used but there are still several
13. M. Claybourn, External Reflection Spectroscopy, in Hand-
measurements that require measurements in this spectral book of Vibrational Spectroscopy, eds J.M. Chalmers and
region. Indeed, in Volumes 4 and 5 of this Handbook, P.R. Griffiths, John Wiley & Sons, Chichester, 969981,
few, if any, applications of FIR spectroscopy are discussed. Vol. 2 (2002).
Nonetheless, because of the wide variety of measurements 14. L. Genzel, Top. Appl. Phys., 74, 169 (1999).
that can be made in this spectral region, it is likely that there 15. V.A. Bershtein and V.A. Rhyzhov, J. Macromol. Sci.-Phys B,
will be a resurgence of FIR spectroscopy in the future. 23, 271 (1984).
16. W.F.X. Frank and H. Fiedler, Infrared Phys., 19, 481 (1979). 20. P. Groner, Normal Coordinate Analysis, in Handbook of
17. W.F.X. Frank and U. Leute, in Infrared and Millimeter Vibrational Spectroscopy, eds J.M. Chalmers and P.R. Grif-
Waves, ed. K.I. Button, Academic Press, New York, 51, fiths, John Wiley & Sons, Chichester, 19922011, Vol. 3
Vol. 8 (1983). (2002).
18. M. Goldstein, M.E. Seeley, H.A. Willis and V.J.I. Zichy, 21. P.R. Griffiths and H.W. Thompson, Spectrochim. Acta, 24A,
Polymer, 14, 530 (1973). 1325 (1968).
19. V.A. Bershtein and V.A. Rhyzhov, Adv. Polym. Sci., 114, 43 22. J.E. Harries, NPL Report DES 16, National Physical Labora-
(1994). tory, Teddington (1972).
Resolution and Instrument Line Shape Function
Peter R. Griffiths
1 DEFINITION OF RESOLUTION order of diffraction, and l is the wavelength. The angular

dispersion, Da , of a grating is found by differentiating this
Resolution can be defined as the minimum separation of formula with respect to wavelength. For a constant angle
two infinitely sharp lines of equal intensity that allows the of incidence, a, Da , is given by equation (2):
presence of two lines to be seen in the measured spec-
db jmj sin a C sin b
trum. An alternative, but not exactly equivalent, definition Da D D D 2
of resolution is the full width at half height (FWHH) of an dl d cos b l cos b
infinitely sharp isolated line measured by a spectrometer. If the grating is operated with nearly normal incidence
The first definition is the more accurate since the resolution (a 0 ) and small angles of diffraction, Da is only weakly
of two lines depends on their shape as well as the width and dependent on wavelength. The linear dispersion, DL , is
the instrument line shape (ILS) function of the spectrom- given by the product of Da and the focal length of the
eter. In this article, we shall consider the ILS function for monochromator, f, and tells how far apart in distance two
two types of spectrometer, scanning monochromators and wavelengths are separated in the monochromators focal
Fourier transform (FT) spectrometers. It should be noted plane. The reciprocal linear dispersion, Rd , equal to DL1 ,
that for monochromators, the ILS function is sometimes represents the number of wavelength intervals contained in
known as the spectral slit function. each interval of distance, x, along the focal plane. Thus
[equation (3)]:
2 MONOCHROMATORS dl
Rd D DL1 D fDa 1 D 3
dx
2.1 Non-diffraction-limited scanning
The geometric spectral bandpass, sg , of a grating mono-
monochromators
chromator is given by equation (4):
Let us first consider the case of a grating monochromator sg D Rd W 4
for which the widths of the entrance and exit slits are
equal.1 The spectral bandpass is the half-width of the where W is the slit width. If the entrance slit is illumi-
wavelength distribution passed by the exit slit. Thus the nated with monochromatic radiation, the grating acts like a
spectral bandpass is controlled by the linear dispersion of mirror and an image of the entrance slit is formed in the
the monochromator as well as the width of the slits. The plane of the exit slit, which is known as the focal plane of
well-known grating equation is equation (1): the monochromator. As the grating is rotated, the image of
the entrance slit is swept across the exit slit, as shown in
dsin a C sin b D ml 1
Figure 1. This physical operation is equivalent to the math-
where d is the groove spacing, a is the angle of incidence ematical operation of convolution, where the convolution of
to the grating normal, b is the angle of diffraction, m is the two functions, gl and f0 l is defined by equation (5):
1
John Wiley & Sons Ltd, 2002.
Gl D gl f0 l D gl f0 l li dl 5
0
2 Instrumentation for Mid- and Far-infrared Spectroscopy
W equal to WRd . Thus two very narrow lines will be separated

Slit width Fixed position
exit slit
to baseline if they are separated by a wavelength interval of
outline 2WRd and any visual sign that there are two lines present
H Slit height in the spectrum will be lost if they are separated by less
Moving than WRd . If the spectral feature is not a delta function,
entrance the measured spectrum will be the convolution of the ILS
slit image
function with the spectral feature. For bands that are much
wider than the ILS function, the measured spectrum will
No overlap be a fairly true representation of the band. Conversely, if
the ILS function is much wider than the line or band, the
shape of the measured feature will closely resemble that of
the ILS function.
25% overlap
Direction 2.2 Non-diffraction-limited polychromators with

of image
travel array detectors
50% overlap
If the entrance slit is much narrower than the exit slit,
or vice versa, it is fairly easy to recognize from Figure 1
that the ILS function will be rectangular. If the entrance
75% overlap
slit is somewhat wider than the exit slit, the ILS func-
tion is trapezoidal. Although this situation should rarely be
encountered for a correctly operated scanning monochro-
mator with a single detector, it is interesting to speculate
100% overlap
on the ILS of a polychromator with a wide entrance slit
(a) and an array of small detectors. Instruments of this type
have been used for Raman spectrometry with a charge-
coupled device (CCD) two-dimensional array detector and
100 near-infrared spectroscopy with a linear InGaAs array.
Half-width Since it is customary to operate these instruments with
Percentage of image of line
radiation emerging (sg) the image of the entrance slit filling at least two pixels
from exit slit
50
of the detector, it is very likely that the ILS would be
trapezoidal for both of these types of spectrometers. The
easiest way to demonstrate this type of behavior is the mea-
surement of the atomic emission line from a neon lamp
measured on an instrument that was designed for CCD-
0 50 100 200
Raman spectroscopy. We carried out such a measurement
Percentage of image overlap on a CCD-Raman spectrometer, with slits set at 30, 15,
(b) (slit function)
5, 2.5 and 1 m (Figure 2). It can be seen that for the
Figure 1. Generation of the ILS function of a scanning mono-
30-m slit, the response (which approximates the ILS func-
chromator with equal entrance and exit slit widths. In (a) the
image of the entrance slit is being moved across the exit slit, tion) is nearly rectangular. For the 15-m slit, the ILS is
with the shaded area representing the extent of overlap which somewhat trapezoidal, and for the narrower slits, the ILS
is proportional to the signal reaching the detector. In (b) the slit becomes more triangular. The resolution obtained with the
function is shown as a function of wavelength. 2.5- and 1-m slits is approximately equal. This result
implies that the image of a 2.5-m entrance slit fills one
When both functions are rectangular, as they are if pixel on this instrument.
f0 l li represents the profile of the entrance slit and From Figure 2 it can be seen that the actual ILS func-
gl represents the profile of the exit slit, the result of tion for a 2.5-m entrance slit is not exactly triangular but is
their convolution, Gl, is a triangular function. Thus a rounded at the top and bottom. Deviations from ideal behav-
monochromatic line (delta function) measured by a grating ior are caused by optical aberrations. The increased inten-
spectrometer operating in the absence of aberrations and sity of the spectrum shown in Figure 2(a) near 12 315 cm1
diffraction effects has a triangular shape, i.e. the ILS func- is the most obvious manifestation of the variation in the
tion is said to be triangular. The FWHH of this function is response of the CCD from pixel to pixel.
Resolution and Instrument Line Shape Function 3
4000
3500
3000
Counts
2500
2000 (a) (b) (c)

(d)
1500
1000 (e)
500
(cm1) + (cm1)
12 290 12 295 12 300 12 305 12 310 12 315 12 320
2
Figure 3. Sinc ILS function of a diffraction-limited grating
Wavenumber /cm1
monochromator. This is also the ILS function of a FT spectrom-
Figure 2. Atomic emission line measured with a polychromator eter with triangular apodization (vide infra). (Reproduced from
and CCD array detector with the following slit widths: (a) 15 m; Fourier Transform Infrared Spectrometry by P.R. Griffiths and
(b) 15 m; (c) 5 m; (d) 2.5 m; (e) 1 m. The rounding at the J.A. de Haseth, Wiley Interscience, 1986.)
discontinuities of the trapezoid is the result of optical aberrations.
2.3 Diffraction-limited case
At very small slit widths, the width of the image of a

line at a given wavelength is larger than the width of
the slits. In this case, the ILS function is no longer tri-
angular because the image of the entrance slit in the focal
plane of the monochromator is controlled by aberrations
and diffraction. When aberrations affect the ILS function
of a scanning monochromator with entrance and exit slits Figure 4. The Rayleigh criterion of resolution is found when the
of the same width, the result is that the triangular shape maximum of one infinitesimally narrow line with a sinc2 shape
lies at the same wavelength as the first zero of another line with
is somewhat rounded. This rounding can be seen in the
the same shape. (Reproduced from Fourier Transform Infrared
wings of the trapezoidal slit function shown in Figure 2. Spectrometry by P.R. Griffiths and J.A. de Haseth, Wiley Inter-
In the case of diffraction-limited resolution, the image of science, 1986.)
the monochromatic line in the focal plane is no longer
the same shape as the entrance slit (even in the absence The resolving power is an alternative way of expressing
of aberrations) but is rather controlled by the diffraction the capability to distinguish between two wavelengths. For
pattern. In this case the ILS function is described as a a monochromator, the theoretical resolving power is given
sinc2 (sinc squared) function, where sinc x D sin x/x, and by equation (6):
sinc2 x D sin x2 /x 2 (Figure 3). Two very narrow neigh-
WG jmj
boring lines measured with this function are just resolved Rth D D jmjN 6
when the central maximum of one line described by this d
function falls on the first minimum of the other. (Figure 4). where WG is the width of the grating and N is the total
This condition is sometimes known as the Rayleigh cri- number of grooves that are illuminated by the beam in the
terion of resolution. Although the Rayleigh criterion was monochromator. Because the diffraction order is given by
developed strictly for lines with a sinc2 shape, the term equation (7):
has come to be used when any two narrow lines of equal dsin a C sin b
mD 7
intensity are resolved with a 20% dip between them. The l
diffraction-limited spectral bandpass, which is the FWHH
the theoretical resolving power can also be written as
of the central maximum in the units of wavelengths, is
equation (8).
given by Rd Wd , where Wd is the diffraction limited slit
width, i.e. the slit width equal to the half-width of the cen- dsin a C sin bN WG sin a C sin b
tral maximum. Rth D D 8
l l
Since the maximum possible value of (sin a C sin b) is 2, of each function calculated individually.3 The FT of the
the maximum possible theoretical resolving power for a interferogram I is the true spectrum, InQ 0 , while the FT
grating of illuminated width WG is given by equation (9): of the boxcar function D is the sinc function, fnQ , given
2WG by equation (13):
Rth max D 9
l sin2pnQ
fnQ D 2 D 2sinc2pnQ 13
2pnQ
3 FOURIER TRANSFORM (FT) which is shown in Figure 5(a).
SPECTROMETERS Thus the spectrum of a monochromatic line would have
the appearance shown schematically in Figure 5(b). The
3.1 Effect of finite optical path difference function fnQ is, therefore, the ILS function of a FT spec-
trometer in which the interferogram has not been weighted
The ILS of a FT spectrometer is determined by two factors, in any way, i.e. unapodized.
the maximum optical path difference between the beams in The FWHH of this sinc function is 0.605/, but because
the two arms of the interferometer and the divergence of of the shape of the sinc function, two lines separated by
the beam passing through the interferometer.2 We will first 0.605/ would not be resolved. In practice, two monochro-
consider the case for a collimated beam passing through matic lines of equal intensity would have to be separated
by 0.73/ to achieve a dip of 20% of their intensity, i.e.
the interferometer. This is a theoretical construct, since
to achieve a resolution equivalent to the Rayleigh criterion
a perfectly collimated beam comes from a point source.
Because the area of a point source is zero, no photons can for a sinc2 function (Figure 6). The lines are resolved to
be emitted. In practice the ILS function caused by the finite baseline when they are separated by 1/, which is often
optical path difference must be convolved with the ILS known as the nominal resolution of a FT spectrometer.
caused by the finite beam divergence to obtain the true ILS Although this description has been given in terms of
function of a FT spectrometer. the ideal cosine interferogram shown in equation (10) and
Let us consider the case of an ideal interferogram for the cosine FT shown in equations (11) and (12), it is
which there is no phase delay between the true and the equally valid for the case of interferograms where there
measured interferogram. The interferogram of a collimated is a wavenumber-dependent phase lag qnQ between the true
beam of monochromatic radiation of intensity InQ 0 at and measured interferograms caused by a non-ideal beam-
wavenumber nQ 0 at an optical path difference is given by splitter or delays introduced by the electronics, shown in
equation (10): equation (14):
I D 0.5 InQ 0 cos2pnQ 0 10 I D 0.5 InQ 0 cos2pinQ 0 qnQ 14
To obtain the true spectrum, the cosine FT must be calcu- Since cosa C b can be written as (cos a cos b
lated from equation (11): sin a sin b), the effect of the phase lag is to turn
1 the ideal interferogram shown in equation (10) into a
complex function that can be written in the transcendental
BnQ D I cos2pnQ d 11
1 exponential notion as equation (15):
where the limits of integration are between C1 and 1. I D 0.5 InQ 0 exp2pinQ 0 15
In practice, of course, the optical path difference cannot be
varied by an infinite amount and the interferogram can only To obtain the spectrum, the complex FT must be calculated
be sampled over a finite optical path difference, say between from equation (16):
and C. In this case, we can think of the interferogram 1
as being multiplied by a boxcar function D which is BnQ D I D exp 2pinQ d 16
unity between and C and zero for 1 < < 1
and C < < C1. The FT of this function is given by Despite the difference in formalism, the ILS function for
equation (12): spectra calculated in this way is exactly the same as the
1 sinc function shown in equation (13).
BnQ D I D cos2pnQ 0 d 12 This function is far from ideal for examining spectra
1
containing many lines with greatly different intensities
Mathematically, the FT of the product of two functions because the side-lobes from a strong line can be mistaken
such as I and D, is the convolution of the FTs for the main lobe of a weak line. The amplitude of the
0.605/

(cm1) +
(cm1)
1
(a)

B()

(cm1)

1
1

1 1
1 + 1
(b) 2 2
Figure 5. (a) Sinc ILS function of a FT spectrometer with no apodization (boxcar truncation). (b) The result of convolving the sinc
ILS function shown in (a) with a very narrow line at wavenumber nQ 1 . (Reproduced from Fourier Transform Infrared Spectrometry by
P.R. Griffiths and J.A. de Haseth, Wiley Interscience, 1986.)
weighting the interferogram with a function, A, that

resembles the boxcar function, D, in that it has a value
of zero for 1 < < and C < < C1, i.e. where
no data were sampled. However, unlike D, which has a
constant value of unity for all values of between
and C, A D 1 only when D 0 and decays as jj
increases. The ILS function is no longer the sinc function
shown in equation (13), but instead is the FT of A.
Since the amplitude of the side lobes of all functions that
decay continuously as increases is smaller than that of
the sinc function, any such function is more appropriate
for high-resolution spectroscopy than the boxcar function,
0.73/ D. Since the side-lobes of a sinc function were once
Figure 6. Diagram showing that the minimum separation of two referred to as feet or podes (Greek poos), the process of
sinc functions required for the resultant to have a dip of 20% of removing these feet is called apodization, and the function
the maximum of either line is 0.73/, where is the maximum
A is called an apodization function.
optical path difference of the measurement. (Reproduced from
Fourier Transform Infrared Spectrometry by P.R. Griffiths and The simplest apodization function is a triangle, where
J.A. de Haseth, Wiley Interscience, 1986.) AT is given by equation (17):

most intense side-lobe is over 20% of that of the central AT D 1 for C
lobe and the amplitude of higher-order lobes decays very
slowly. This problem can be ameliorated by numerically AT D 0 for < and > C 17
The FT of AT is a sinc2 function given by equation (18): They calculated the absolute value of the largest secondary
lobe normalized to that of the central lobe and plotted it
sin2 pnQ against the FWHH of the ILS function normalized to the
fT nQ D D sinc2 pnQ 18 FWHH of the corresponding sinc function. Their results
pnQ 2
are shown in Figure 7. An ideal apodization function would
Thus a FT spectrometer in which interferograms are weigh- give rise to a point at the bottom left of this graph. However,
ted by a triangular apodization function has the same ILS for all the functions investigated by Norton and Beer,
function as a diffraction-limited grating monochromator and there is an empirical locus to the left of which there are
lines separated by the nominal resolution (1/ cm1 ) have no points.
a dip of 20% between their maxima. Norton and Beer selected three functions that gave rise
The sinc2 function is far from an optimal ILS function, to points on this locus where the FWHH was broadened to
however, as the amplitude of the most intense side-lobe is 1.2, 1.4 and 1.6 times that of the corresponding sinc func-
still 5% of that of the central lobe and the FWHH is one tion. These functions have been called the NortonBeer
and a half times greater than that of the corresponding sinc weak, medium and strong apodization functions, and are
function. The ideal ILS function would have minimal side- commonly used in Fourier transform infrared (FT-IR) spec-
lobe amplitude while retaining a FWHH close to that of troscopy. The coefficients Ci are shown in Table 1.
fnQ . To investigate the feasibility of achieving this goal, The ILS functions corresponding to these NortonBeer
Norton and Beer4 studied over 1000 polynomial functions functions are shown in Figure 8. Many other functions have
of the general form of equation (19): also been used in FT-IR spectroscopy. For example, the
HappGenzel function is favored by the Nicolet Instrument
2 i Corporation. In general, however, the three NortonBeer

n

A D Ci 1 19 functions have proved to be very successful for most aspects
iD1
of FT-IR spectroscopy.
0
10
3.2 Effect of beam divergence through the
Normalized height of largest secondary maximum |h /h0|
interferometer
Any beam that is incident on a grating or an interferometer

is usually assumed to be collimated. This implies that it
originates from a point source. In practice, in order for it
101 to emit any photons, the source must have a finite size. For
FT spectrometry, a simple calculation2 has been used to
show that the half-angle of divergence of the beam passing
through the interferometer cannot exceed a value given (in
radians) by equation (20):
1/2
nQ
amax D 20
102 nQ max
where nQ is the resolution (i.e. the maximum of the optical
path difference) and nQ max is the highest wavenumber in the
spectral range of interest. The effect of a divergent beam
is to change the shape and width of the ILS function from
Table 1. Coefficients Ci for the NortonBeer apodization

103 functions.
1.0 1.2 1.4 1.6 1.8 2.0 2.2
Normalized half-width [W /W0] C1 C2 C3 C4
Figure 7. Plot of the normalized height of the largest secondary
lobe for a series of apodization functions studied by Norton and Boxcar 1.000000 0 0 0
Beer as a function of their half-width relative to the half-width of Weak 0.384093 0.087577 0.703484 0
the corresponding sinc function. [Reproduced from R.H. Norton Medium 0.152442 0.136176 0.983734 0
and R. Beer, J. Opt. Soc. Am., 66, 259 (1976), with permission Strong 0.045335 0 0.554883 0.399782
from the Optical Society of America, 1976.]
1.0 increment (annulus) of solid angle d at a is given by

equation (21):
0.8
dI D InQ d cos2pnQ 0 cos a 21
0.6
Using the small angle approximation [equation (22)]:
0.4
I1
I0 a2
0.2 cos a 1 22
2
0.0 we have [equation (23)]:
0.2
a2
dI D InQ d cos 2pnQ 0 1 23
(a) 1/ 2/ 3/ 4/ 2
1.0 The full solid angle of the beam inside the annulus, ,
that we selected is approximately equal to 0.5pa2 , so
0.8
a2 /2 D /2p. Hence [equation (24)]:
0.6

Amplitude
0.4
dI D InQ d cos 2pnQ 0 1 24
2p
I2
0.2 If we use the full aperture of the interferometer [equa-
I0 tion (25)]:
0.0
max max
0.2

ID dI D InQ d cos 2pnQ 0 1
0 0 2p
(b) 1/ 2/ 3/ 4/
nQ 0 xmax
1.0 Q
D Inmax sinc Q
cos 2pn0 1 25
2p 4p
0.8
Thus the effect of the finite aperture is to modulate the
0.6 envelope of the interferogram with a sinc function. (Note
that it also produces a scale change, which can be accounted
0.4
I3 for by multiplying all wavenumbers in the spectrum by
0.2 (1 max /4p).) Provided that the solid angle is restricted
I0
so that the sinc function is positive, the sinc function acts
0.0 in the same manner as an apodization function. For this
0.2 reason, the broadening of the ILS function caused by the
effect of an extended source is sometimes called optical
1/ 2/ 3/ 4/
apodization. Thus a beam from a monochromatic source
(c) Frequency (cm1 if in cm) with a solid angle of max steradians measured using an
Figure 8. The ILS functions corresponding to the three Nor- interferometer with an infinite optical path difference will
tonBeer apodization functions: (a) weak; (b) medium; (c) strong. give rise to a spectrum the shape of which is the FT of the
On each plot, the sinc function corresponding to the ILS function
sinc function, i.e. a boxcar.
of an unapodized interferogram is also shown. [Reproduced from
R.H. Norton and R. Beer, J. Opt. Soc. Am., 66, 259 (1976), with Let us assume that the maximum resolution required
permission from the Optical Society of America, 1976.] at wavenumber nQ is nQ cm1 . The maximum optical path
difference max is, therefore, equal to 1/nQ cm and the
the ones given above. The discussion below follows the resolving power R D nQ /nQ D max nQ . If we define max as
treatment by Brault.5 equal to k 4p/R, the initial region where the sinc function is
Let us consider the effect of the path difference between positive reduces to k < 1. The fringe amplitude for various
an on-axis ray and a ray that is off-axis by an angle a for values of k is shown in Figure 9. It can be seen that when
a monochromatic beam of wavenumber nQ 0 and intensity k D 0.5, the fringe amplitude has halved by the end of
InQ 0 . When the optical path difference between the two the scan. This condition, which corresponds to the one
beams in the interferometer (optical retardation) is , the shown in equation (20), only has a very small effect on the
path difference between the on-axis and off-axis rays is spectrum and is generally used as the criterion for setting
cos a. Thus the response to the radiation in a small the maximum solid angle used for a given measurement.
Amplitude at x = 0 proportional to k ; max = k 4 being degraded. Similarly, the function of the J-stop is to
R
ensure that the beam passing through the interferometer is
1.5
sufficiently collimated that the resolution is determined by
the maximum optical path difference.
The Jacquinot stop is the limiting aperture of the optics
of all high-resolution FT-IR spectrometers. Some bench-top
o max
1.0 FT-IR spectrometers operating with a resolution of 1 cm1

2
or poorer are not equipped with a Jacquinot stop. In this

case, some other component (usually the detector) serves
Fringe amplitude max sinc
as the limiting aperture and a J-stop is not needed.
k = 0.5 3.3 Effect of poor mirror drive

Maximum amplitude for
this path difference
The ILS function of FT-IR spectrometers is largely gov-
0.25
erned by the maximum optical path difference and the
divergence of the beam. In the unlikely case that the drive
of the moving mirror tilts during the scan, an effect similar
0 x
to the one described in the previous section will be seen,
2 i.e. the resolution will be degraded. The design of all inter-
ferometers used in modern FT-IR spectrometers is such that
this effect is rarely seen and it will not be described here.
x=
Figure 9. Fringe amplitude as a function of path difference when REFERENCES
the beam passing through the interferometer diverges with a solid
angle equal to 4pk/R. The amplitude at zero path difference
1. J.D. Ingle, Jr and S.R. Crouch, Spectrochemical Analysis,
is proportional to k. [Reproduced by permission from Brault
Prentice Hall, Englewood Cliffs, NJ (1988).
(1985).5 ]
2. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
For FT spectrometers designed to measure spectra at Spectrometry, Wiley Interscience, New York (1986).
high resolution, the solid angle is set by the diameter of 3. C.T. Foskett, The Fourier Transform and Related Concepts:
an aperture mounted at a focus of the beam known as A First Look, in Transform Techniques in Chemistry,
ed. P.R. Griffiths, Plenum Publishing Co., New York, 1138
the Jacquinot stop (frequently abbreviated to J-stop). It
(1978).
is noteworthy that this aperture serves the same purpose
as the entrance slit of a grating monochromator, which is 4. R.H. Norton and R. Beer, J. Opt. Soc. Am., 66, 259 (1976);
Erratum J. Opt. Soc. Am., 67, 419 (1977).
to ensure that the beam is sufficiently collimated at the
grating that the resolution is determined by the width of 5. J.W. Brault, Fourier Transform Spectroscopy, in High Res-
olution in Astronomy, Proceedings of the 15th Advanced
the exit slit. When the width of the entrance and exit Course of the Swiss Society of Astronomy, eds A.O. Benz,
slits are equal, the maximum allowed amount of light M.C.E. Huber and M. Mayor, Saas Fee, Souverny, Observa-
passes through the monochromator without the resolution toire de Geneve, Switzerland (1985).
Monochromators for Mid-infrared Spectroscopy
Michael Ford
Consultant, Maidenhead, UK
1 BASIC REQUIREMENTS well separated; with the Littrow design this would involve
additional optics.)
The conditions which have to be met for a functional mid- Around 1890, Angstrom and Julius each used simple
infrared (MIR) monochromator are significantly different monochromators with rock salt prisms and lenses, pre-
from those for the visible region. Firstly, there are few sumably based on the conventional visible spectrometer,
convenient materials which transmit radiation over the full an exception to the use of reflecting optics referred to in
wavelength range of interest, leading to the almost universal Section 1 above. (It is worth noting that visible monochro-
use of reflective, rather than transmissive, optical elements. mators were used only for visual observations; visible spec-
Secondly, the lack of bright sources and, until relatively tra were recorded photographically using polychromators,
recently, of high sensitivity detectors, imposes a need for or spectrographs, covering a wide spectral range simultane-
high etendue or throughput. ously. Coblentz emphasizes the contrast with the necessary
Thirdly, the system needs to be sealed so that the sequential, point by point recording in the IR.)
strong absorptions due to atmospheric CO2 and water vapor Coblentz used two designs, both with a single pass
may be minimized by desiccation, purging or evacuation. through the rock salt prism. In the first (Figure 1), radiation
Fourthly, but less often considered, it has to be remembered from the entrance slit is collimated by a concave (presum-
that the instrument itself radiates very significantly in the ably spherical) mirror on to the prism and the emerging
MIR region, a situation which has been likened to mak- radiation focused by a similar mirror on to the exit slit.
ing visible measurements in a white hot spectrometer! How The position of the prism is fixed in relation to the sec-
these requirements have been met in actual commercial and ond mirror and wavelengths are selected by moving the
research monochromators is described in the following. entrance slit and first mirror assembly. In this system, the
prism can only be in the position of minimum deviation
for one wavelength, but this merely complicates the calcu-
2 EARLY MONOCHROMATORS lation of the wavelength from the measured angles and the
known refractive index data for rock salt. (With all these
In his classical book on infrared (IR) spectra1 published in early monochromators, the procedure was to set the angles
1905, Coblentz describes in some detail the monochroma- for a certain wavelength, make the necessary readings,
tors which he used for his work. In this section they, and then reset to the next wavelength and so on for the range
some used by earlier workers, are described together with of the scan). The second design was based on that of
some explanation of the reasons for the choice of those Wadsworth3 (Figure 2). In this design the prism is mounted
specific designs. In particular, it is interesting to note that on a turntable together with a plane mirror which reflects
the Littrow design,2 so popular later, was not used even the radiation onto the prism; the collimating and telescope
though it had been published in 1862. (It is suggested that arms remain fixed and wavelengths are selected by rotation
this was in order to keep the source and radiometer detector of the turntable. This system has the advantage that, once
the prism is set for minimum deviation at one wavelength,
John Wiley & Sons Ltd, 2002. it remains so for all wavelengths. In both designs, mirrors
G 3.1 Early and unusual designs
The first commercially available IR spectrometer, the model

D83, made by Adam Hilger of London in 1913, used the
Wadsworth type of monochromator, described earlier, with
spherical mirrors made of nickel and a choice of NaCl, KBr
or other prisms.
Although the double beam spectrometer designed by
Lehrer4 before the war is not believed to have been pro-
duced commercially, it is important as introducing the
R F1
P S optical null double beam concept incorporated later in
S the Perkin-Elmer model 21 and subsequently in many
C other instruments. (It is presumed that both Lehrer and
Wright (at Dow Chemicals) independently devised the
F2
null balance system at about the same time.) It used an
Figure 1. Coblentz monochromator. Wavelength setting is per- unusual double monochromator (Figure 3), the first part
formed by rotation of the source S, entrance slit F1 and collimating using an on-axis spherical mirror and the second a Lit-
mirror assembly about the prism P, while the prism, exit slit F2
and radiometer R remain stationary. C, cell holder; G, exhaust
trow system, both parts incorporating 30 , back reflecting
pump for radiometer. (Reproduced by permission of PerkinElmer prisms.
Instruments.) A monochromator design adopted in a number of com-
mercial instruments, especially those of relatively high
C resolution, is the Ebert5 system which utilizes a single large
spherical mirror for both collimator and telescope (or two
smaller mirrors on the same spherical surface). This config-
uration, illustrated in (Figure 4) as used in the Perkin-Elmer
model 180, gives good correction of spherical aberration but
does suffer from significant astigmatism. The effect of the
astigmatism may be minimized by the use of slits curved
about the optic axis so that the aberration is always tan-
gential to the slit. The 180 design was unique in the ability
to select any one of seven diffraction gratings mounted on
a turntable and scanned independently. At 0.15 cm1 , this
was probably the highest resolution MIR monochromator
ever offered commercially.
All grating monochromators require some means of elim-
Figure 2. Wadsworth system. Wavelength setting is performed ination of unwanted orders of diffraction from the grating.
by rotation of the prism and plane mirror assembly about a point In the case of the 180, and also the Hilger infragraph,
below the prism base, while the collimating mirror (illustrated this is achieved by the selection of the appropriate long
at C) and telescope mirrors remain stationary. (Reproduced by pass or band pass filter from several mounted on a filter
permission of PerkinElmer Instruments.)
wheel. The alternative to the use of filters is a sec-
of various focal lengths were used (from 35 to 100 cm). Slit ond, prism monochromator as employed in the Perkin-
sizes were about 15 to 20 mm long by 0.2 to 2 mm wide. Elmer models 125 and 225 and the Hilger GM3. In the
Perkin-Elmer examples the second monochromator was
of the Littrow design while the Hilger instruments used
3 COMMERCIAL MONOCHROMATORS the CzernyTurner arrangement (see Section 4) for both
monochromators.
Since the vast majority of the monochromators used in An additional or alternative method of effective elim-
commercial spectrometers have been variants of the Littrow ination of stray radiation at shorter wavelengths was
design, this section has been divided into two parts. The the use of a crystal chopper. Such a beam chopper has
first deals with non-Littrow, including early and unusual, alternate sectors of short wavelength transmitting crystal
designs and the second with the sequence of developments and apertures giving the same transmission as the crys-
and adaptations of the Littrow design. tal. When this rotates, radiation absorbed by the crystal
Monochromators for Mid-infrared Spectroscopy 3
Hilfs system Houpt system
Blende
Linsentenster
Austritts
spalt
Konsole fr
Beleuchtungs spiegel
Hilfsspalt
Eintrittsspalt
Bolo-
meter
30
Figure 3. Lehrer double monochromator. Wavelength scanning is performed by rotation of both prisms in opposite directions.
Figure 4. Optical schematic of the model 180. (Reproduced by permission of PerkinElmer Instruments.)
is modulated and detected while that transmitted by the scanning is performed by rotation of the Littrow mirror
crystal is not modulated and therefore not detected. This about a vertical axis. The design is very compact but, as a
system was mainly used in far-infrared instruments, such consequence, great care must be taken to avoid stray radi-
as the Perkin-Elmer model 301. ation due to reflections from the sides of the monochromator
In 1951 Walsh6,7 proposed a system for increasing the cover and prism base. The numerous products incorporating
resolution of a monochromator by double or multipassing this basic design included the Grubb Parsons S.3, the Uni-
of the radiation through or off the dispersing element cam SP200 and the Beckman IR5 and most were available
by reflection from mirrors placed beside the slits. In his with a choice of prism, usually NaCl and KBr, but including
original design detection of radiation not multipassed was CaF2 , CsBr and CsI. Some early instruments used spherical
avoided by chopping the radiation at a point between the mirrors but, unless long focal lengths were employed, aber-
returning mirrors. The double pass version was employed, rations were severe and greatly reduced resolving power.
in a Littrow configuration, in the Perkin-Elmer model 112. Later, improved methods for the production and replica-
A variation of this design for double pass, avoiding the need tion of aspheric mirrors resulted in the universal use of
for the chopper, used mirrors above or below the slits and parabolic mirrors for collimation and ellipsoidal mirrors for
was employed in an Ebert monochromator in the Perkin- producing a reduced image of the exit slit on the detec-
Elmer model E1 and in a Littrow configuration in the Grubb tor. Prior to the availability of order-sorting filters for the
Parsons model S.3. longer wavelengths, in a number of instruments, such as the
Perkin-Elmer model 221, the Littrow mirror was replaced
3.2 Developments and adaptations of the Littrow with a diffraction grating. This served as the main disper-
design sive element at shorter wavelengths, in conjunction with a
filter to remove second and higher orders, while at longer
A typical Littrow prism monochromator, the Perkin-Elmer wavelengths the grating was turned to zero order and served
model 137, is illustrated in Figure 5. Radiation from the as the Littrow mirror for a normal prism configuration.
entrance slit is collimated by the off-axis parabolic mirror Later instruments employed two, four or more gratings,
and passes through the prism to the Littrow mirror which each used in first, or first and second orders. Various con-
returns it along a similar path to the exit slit. Wavelength figurations were used for multiple gratings, including two
Littrow Prism
mirror
Paraboloid
mirror
Thermocouple Exit slit
Plane mirror Entrance slit
Ellipsoid mirror
Plane
mirrors Toroid
mirror
Toroid
mirror
Sampling area Source
Sector
mirror Plane Toroid
Plane
mirrors mirror
mirror
Figure 5. Optical layout of the Infracord spectrophotometer. (Reproduced by permission of PerkinElmer Instruments.)
Monochromators for Mid-infrared Spectroscopy 5
Golay
M14 detector
M13
P4
M11
Magnetic M8
P3 M12
M10
M7 Slits
P2
G1 M9
P1 G2
Monochromator bridge
Star wheel no.1 M4 M4

Transmittance
h M5
Reference itc
cell sw
well am M6
Be
Sample Photometer
cell bridge
well
M2 M1 Star wheel no.2
M3
M3 Equalizing
Source
Figure 6. Schematic diagram of optics (prism/grating double monochromator). (Reproduced by permission of Spectronic Unicam.)
back to back, two such pairs switched by additional mirrors the usual scale change at 2000 cm1 . Prior to the now
and four in an overlapping square on a single, stepper inevitable use of stepper motors, many monochromators
driven turntable. An exception to the use of only first and incorporated elaborate gearboxes to provide a range of scan
second orders was the Grubb Parsons G.S.2 which used speeds.
multiple orders of a single grating in conjunction with a In conjunction with scanning, slit widths in nearly all IR
fore monochromator with two switchable prisms of CaF2 monochromators have been driven so as to give approx-
and KBr. imately constant response from the detector at all wave-
Probably the most elaborate Littrow monochromator, a lengths, accepting the consequent change in resolution.
double monochromator, ever produced commercially was Again, prior to stepper motors, slit drives typically used
the Unicam SP100 (Figure 6). This could be used in either cams, in some cases with mechanical linkages to provide
a prism grating configuration, with a choice of two gratings different resolution slit programs, coupled to the scan drive.
and four prisms or, by insertion of a plane mirror in front of
the gratings, as a single, prism instrument. In addition, the
monochromators were evacuable to minimize atmospheric 4 RESEARCH MONOCHROMATORS
absorptions.
Unlike visible monochromators, many of which were Unlike commercial monochromators, which are designed
designed to make measurements at a single selected wave- to scan wide wavelength ranges, the vast majority of
length all, or virtually all, MIR monochromators were monochromators for research have been designed to scan
designed to be scanned. A range of different mechanisms narrow wavelength ranges at high resolution and ordinate
were employed from simple lead screws giving a nonlinear accuracy. The typical system used has been an Ebert or
relationship to wavelength to geometric sine or cosecant CzernyTurner8 (similar to Ebert but without the con-
linkages giving linear wavelength or wavenumber drive straint of a single spherical surface for the mirrors) grating
to gratings. However, by far the most commonly used monochromator plus a fore or post prism monochromator.
were cams for both prisms and gratings; they could read- An example, which was used by Mould9 to establish the
ily be designed to deal with multiple gratings as well as standard IUPAC ammonia spectral wavenumbers, is shown
D
M1
F
G
Figure 7. Optical layout of grating spectrometer. C and D, detectors; MI, switch mirror; F, vibrating chopper; G, source.
in Figure 7. Others, not illustrated, include those of Thomp- the typical wavenumber range covered in a single scan is
son et al., Fastie,10 and Plyler.11 very small.
The diffraction limited resolution, in wavenumbers, of a Although the monochromator illustrated relied on purg-
grating monochromator is the reciprocal of the difference, ing with dry, CO2 free air to minimize atmospheric absorp-
in centimeters, of the optical path off the two edges of the tions, many others were enclosed in vacuum casings but,
grating. It follows that, in order to achieve a resolution as the source and sample were usually external to this there
of 0.1 cm1 or better, typical of such instruments, there was still an air path requiring purging.
is a minimum requirement of a 10 cm wide grating at
30 . Since both ultimate resolution and dispersion increase
with grating angle it is advantageous to use high grating ACKNOWLEDGMENTS
angles, usually in the range 30 to 45 . However this has
the drawback that multiple interchangeable gratings and/or The author gratefully acknowledges the help and mate-
multiple orders must be used to cover the full spectral range. rial provided by the late Dr Henry Mould and Dr Francis
(Commercial monochromators generally use much lower Dunstan.
grating angles in order to cover at least one octave with
each grating.)
A general requirement of research monochromators is a REFERENCES
very precise and reproducible scan drive for the grating,
since a major use has been to establish accurate wavenum- 1. W.W. Coblentz, Investigations of Infra-red Spectra, Car-
ber values for the positions of gas spectral lines to determine negie Institution of Washington (1905). (Reprinted 1962,
molecular constants and provide calibration standards. The sponsored by the Coblentz Society and the Perkin-Elmer
usual procedure for performing such measurements has Corporation.)
been by comparison with wavelength, l, standards already 2. O. Littrow, Am. J. Sci., 35, 413 (1862).
established in the visible or near-infrared using higher 3. F.L.O. Wadsworth, Phil. Mag., 38, 137 (1894).
orders, n, of reflection from the grating and relying on the 4. E. Lehrer, Z. tech. Physik, 7, 169 (1942).
grating equation 5. H. Ebert, Wied. Ann., 38, 489 (1889).
nl D 2d sin q cos j 1 6. A. Walsh, Nature, 167, 810 (1951).
7. A. Walsh, J. Opt. Soc. Am., 42, 94 (1952).
where d is the grating line spacing, q is the angle of
the grating to the optical axis and j is the angle of 8. M. Czerny and A.F. Turner, Z. Physik, 61, 792 (1930).
incidence on the grating when q is zero. Selection of the 9. H.M. Mould, W.C. Price and G.R. Wilkinson, Spectrochim.
required order is achieved by appropriate setting of the Acta, 15, 313 (1959).
prism monochromator; unlike in commercial instruments, 10. W.G. Fastie, J. Opt. Soc. Am., 42, 641 (1952).
synchronous scanning of this is often unnecessary, since 11. E.K. Plyler, J. Res. Natl. Bur. Stand. (US), A64, 29 (1960).
Infrared Filtometers
Paul Wilks
Wilks Enterprise, Inc., South Norwalk, CT, USA
1 INTRODUCTION by comparing the infrared absorption in a sample cell

with that in a cell containing a standard mixture of gases.
1.1 What is a filtometer? Initially, no attempt was made to scan the spectrum or
isolate specific wavelengths. Hence the term nondispersive
The word filtometer is a contraction of the terms filter infrared or NDIR. Later on it was found that performance
and photometer and pertains to a device that makes and discrimination could be improved by using optical
use of band pass filters for wavelength discrimination as windows that would isolate spectral regions and eventually,
opposed to prisms, gratings or interferometers. Filtometers as filter technology evolved, specific absorption bands. True
have sometimes been referred to as NDIR (nondispersive NDIR instruments are rarely manufactured today, having
infrared) instruments. This is not a good description since been replaced by scanning spectrometers or some sort of
most NDIR instruments look at the entire infrared spectrum filter instrument such as those described below.
at all times, while the filtometer sees one or several specific The modern era of filter instruments began with the
wavelengths. A better description would be selective- development of the circular variable filter (CVF) by Optical
wavelength spectrometer. Coating Laboratory in cooperation with Beckman Instru-
A filtometer consists of a source of radiation, energy ments who were trying to produce a rapid scan gas
collecting optics, a beam chopper, sample compartment, chromatography/infrared (GC/IR) instrument. Although the
another set of collection optics and a radiation detector or GC/IR instrument was not successful, the CVF was taken
detectors. On some filter instruments a pulsed source is used over by Wilks Scientific Corporation and made the basis of
which eliminates the need for a beam chopper. a family of portable analyzers that became widely used for
The electronics include power supplies that operate the both gas and liquid analyses.
source, control temperatures and signal amplifiers and It soon became apparent that many filtometers were
related circuits that convert detector signals to concentration being used for a single application, i.e. oil-in-water analy-
values. sis or monitoring the concentration of a specific gas. Such
applications could be handled by infrared absorption mea-
surements at a specific wavelength. As a result, a relatively
inexpensive fixed wavelength filter replaced the expensive
1.2 The evolution of filtometry
CVF. The majority of infrared filtometers going into service
today are dedicated, fixed wavelength instruments.
During the early days of infrared spectroscopy, in the late
1940s and early 1950s, a class of gas analyzers evolved
in Germany (at I. G. Farben) and later on in the US (at
2 INTERFERENCE FILTERS
Baird Associates and Mine Safety Appliance). These were
generally referred to as Luft type NDIR analyzers, after The heart of a filtometer is the interference filter that defines
the inventor and the fact that the instruments operated the band of radiation falling on the detector.
Interference filters are formed on a substrate material
John Wiley & Sons Ltd, 2002. that transmits radiation in the range of the required band
pass of the filter. Thin films of different index materials 2.1 Wavelength vs frequency
are carefully laid down on the substrate in a vacuum
chamber. Because of the waveform nature of light, internal Spectroscopists tend to locate infrared absorption bands
reflections from the film layers will vary from constructive in terms of frequency (reciprocal centimeters or cm1 )
to destructive interference, depending on the wavelength, because of the relationship between the frequency of an
the indices of the films and their thickness. infrared absorption band and the vibrational frequency of
By using a number of layers of two or three different a chemical bond. Optical devices, on the other hand, relate
materials, filters that pass a relatively narrow band of radi- naturally to wavelength (nanometers or micrometers for-
ation with a specified center wavelength can be constructed merly microns). Interference filters are almost always spec-
(Figure 1). Typically, such filters have a bandwidth of about ified in terms of their wavelength transmission band or
2% of the wavelength, although 1% and 0.5% filters are wavelength cutoff point. This sometimes leads to confu-
sometimes produced.1 sion, especially in the 2 to 5 m region which is also the
Interference filters have the ability to pass large amounts 2000 to 5000 nm region and the 5000 to 2000 cm1 region.
of incident radiation because slits or apertures are not The conversion from reciprocal centimeters to micrometers
required to restrict their band pass. They can also accept is as follows:
sharply converging radiation although some shift in the 1
10 000 D 1 m D 1000 nm 1
center wavelength may occur. cm1
The technology of making narrow band pass filters has Wavelength shift in interference filters is directly propor-
been refined to the extent that such filters can be pro- tional to the change in thickness of the individual layers
duced for any specific wavelength in the infrared and that make it up. CVFs and linear variable filters (LVFs)
near-infrared regions of the spectrum. The coatings are which will be discussed in the later sections vary linearly
generally hard and durable. In many filtometers, the fil- with respect to wavelength. It would be almost impossible
ters are used as windows in the detector case. This has to produce CVFs or LVFs that are linear with respect to
the dual effect of improving performance by reducing frequency (wavenumbers).
energy losses and cutting costs as a result of the small size The following two paragraphs are reprinted, with permis-
required. sion, from the Spectrogon catalog on Optical Interference
Filters, 15th edition, Parsippany, NJ, USA.
1.0
2.2 Effects of varying angles of incident beams
The filter wavelengths are specified at normal angle of
incidence. If filters are intended to be used at other than
normal angle of incidence, the filters special characteristics
have to be considered in order to achieve the intended
spectral function. All interference filters will shift to shorter
wavelength as the angle of incidence of a collimated beam
deviates from the normal. This effect may, to some extent,
Transmission
be used to tune a filter to a desired shorter wavelength.

The angle effects can be approximately calculated by the
following formula:
lq D l0 ne 2 sin2 q1/2 ne 1 2
Typical values for the effective index (ne ) for an infrared

filter are in the region 1.52.8. At high angles of incidence,
30 or more, transmission will decrease and a distortion of
the filter band pass shape will eventually occur (Figure 2).
0.0
8.0 9.0 10.0 11.0 12.0 2.3 Effects of varying temperature
Typical filter band pass (m)
Interference filters will shift to longer wavelength with
Figure 1. Band pass of a typical narrow band-pass filter. increasing temperatures or to shorter wavelength with
Infrared Filtometers 3
5
wavelength (%)
Shift to shorter
0 10 20 30 40 50
Angle of incidence ()
Figure 2. Filter band shift with incident angle change. (Courtesy
of Spectrogon.)
2
Shift in wavelength (%)
0
250 150 50 50 150
1
3
Temperature (C)
4
Figure 3. Filter band shift with temperature. (Courtesy of Spec-
trogon.)
decreasing temperature (Figure 3). This wavelength shift Figure 4. A three-segment CVF. (Courtesy of Optical Coating
is caused by thermal expansion/contraction of the coat- Laboratory, Inc.)
ing materials. The shift is often expressed in nanome-
ters per degree Celsius and is normally extremely small
(0.010.2 nm C1 ). The temperature effect can be approx- 2.5 Linear variable filters
imately calculated by the following formula:
LVFs are formed in much the same way as CVFs except that
l0
l C D l0 C T 3 the substrate is moved linearly out from under the shield
1 C as each layer is laid down. The resulting variable filter is
rectangular rather than a segment of a circle.
The linear aspect of the LVF makes it useful as a window
2.4 Circular variable filters over a detector array. Typical spectral ranges for LVFs are
one octave, e.g. 2.5 to 5 m or 5.5 to 11 m. With a 50-
CVFs are multilayered interference filters that are formed element array this gives a theoretical resolution of 0.05 m
by rotating the filter substrate out from behind a screen as with the short wavelength segment and 0.1 m with the long
each layer is deposited. Thus each layer is wedge shaped wavelength segment.
and the band pass will gradually change in wavelength from
one end of the filter segment to the other. Typical mid-
infrared CVFs (Figure 4) cover the range from 2.5 to 15 m 2.6 Filter resolution/procurement
in three segments. Wavelength selection is made by rotating
the CVF in front of a slit until the desired wavelength is Users of interference filters acquire them in one of two ways:
passed through the slit. It is possible to produce a single
beam infrared spectrum by rotating the CVF past the slit 1. They have filters manufactured for them to their spec-
at a uniform speed while recording the output from the ifications.
detector on a strip chart recorder. The resulting spectrum 2. They consult catalogs listing standard or surplus filters
will be linear in wavelength. published by filter manufacturers.
The first approach is usually by far the more expensive varies as the square of the ratio of f numbers. Thus, an
since it costs the filter manufacturer just as much to make f/1 optical system has 100 times the energy throughput of
one filter disk as it does to make several dozen identical a f/10 system.
disks. This is because filters are generally made in large Most dispersive infrared spectrometers have f numbers
vacuum tanks capable of handling a large number of disks at in the 4.5 range. In Fourier transform infrared (FT-IR)
a time. On the other hand, the buyer can specify the required instruments, f numbers are generally higher, in the 6 to 10
band center, bandwidth and the transmission required for range. Lower f numbers are not feasible in part because
his specific application and the potential sales volume may of the constraints imposed by the optical requirements of
be able to justify the cost. Furthermore, he can reject a gratings and interferometers.
filter run if the resulting filter fails to meet the original Interference filters can accept energy converging on them
specifications. Many filters listed in the filter manufacturers at very steep angles without seriously affecting their reso-
catalogs are there for this very reason: they failed to meet lution, although there may be some shift in band center.
customer specifications! This means that very short focal length optics can be used
Users of filters will attempt to acquire filters by the sec- with them, leading to very low f numbers and hence high
ond approach whenever possible. Many absorption bands in energy throughputs.
the mid-infrared, especially those of liquids, are relatively
broad so that considerable latitude may exist in selecting a
filter that will provide the required absorption data. The cost 3 FILTOMETER TYPES
of 2.5-cm filter disks purchased from filter catalogs varies
from US $200 to US $700 depending on wavelength. The 3.1 Single wavelength
cost of a filter run to meet close specifications may reach
This is the simplest filtometer version. It consists of an
as high as US $20 000.
infrared source, a means of modulating the radiation from
Filter resolution or band pass is relatively controllable
the source, and a sample compartment (see Section 4)
by the filter manufacturer. Typically, the band pass will be
through which the radiation passes and is then focused on
2 to 5% of the wavelength, although, as noted above, it is
the detector.
possible to construct 1.0% and even 0.5% band pass filters
The filter may be a window somewhere in the optical
if necessary. However, from an instrument performance
path or it may serve as the window in the detector housing.
standpoint it is desirable to use as broad a band pass as
The size of the beam at the window location may require
possible to maximize energy transmission and hence the
a filter size of 1 to 2.5 cm in diameter while the detector
signal-to-noise ratio (S/N) (see following section).
windows are typically 4 mm square. Thus about 30 detector
windows can be cut from a 25 mm disk.
Mid-infrared detectors used in filtometers are usually
2.7 The importance of optical speed either lead salt (2 to 5 m) or pyroelectric (all wavelengths).
Because of the high energy throughput of filtometer opti-
Optical speed in an optical system defines its light gathering cal systems, it is rarely necessary to use cooled mercury
power or, in the case of a photometer, its energy throughput. cadmium telluride (HgCdTe) detectors.
A way of expressing optical speed is the f number which Pulsed sources are beginning to replace traditional hot
is essentially the ratio of the focal length of the mirrors or wire infrared sources. Electronic pulsing eliminates the
lenses divided by their diameter. need for mechanical beam choppers with their attendant
focal length service problems. Miniature incandescent bulbs such as
f number D 4 used in small flashlights can be pulsed at 16 to 24 Hz.
diameter
Low mass infrared emitters pulsable at 6 to 8 Hz are also
A photometer with mirrors having a focal length of 500 mm becoming available.
and a diameter of 50 mm is an f/10 system, while one with Incandescent sources emit from the visible through the
mirrors with 50 mm focal length and 50 mm diameter is an near- and mid-infrared to about 5 m with the long wave-
f/1 system. length cutoff being a function of the glass or quartz
If radiation is considered to be radiating from a point envelopes. Pulsable infrared emitters peak at around 4 m
source equally in all directions, then the f number defines and have sufficient output to be used out to 10 to 11 m.
that portion of the sphere of radiation that is intercepted by As is the case with other types of filtometers, single
the optical system. The larger the f number, the smaller wavelength models are generally dedicated to a specific
the area of the sphere intercepted and vice versa. Since we application with the sample handling system being an
are dealing with areas, the difference in energy throughput integral part.
3.2 Dual wavelength Filter Fuel cell

wheel
Chopper
wheel Beam
More recent versions of filtometers designed to measure splitter
a single component make use of two wavelengths the Detector
analytical wavelength and a reference wavelength either sample
free of absorption or at a compensating wavelength. The
use of a dual detector system has a significant effect on the Mid-
overall performance of a filtometer. infrared Lens
source
The amount of signal present in a photometric system
above the background noise level is the single most impor-
tant parameter affecting quantitative performance. The S/N Detector
(reference)
determines overall accuracy and minimum detectable con-
centrations; it is a major factor in reproducibility and Figure 6. Schematic of a filter wheel filtometer.
long-term stability.
The signal is a result of the radiation produced by the readings are stored in a microprocessor and concentration
source, the energy throughput (f number) and the efficiency computations performed at the end of each cycle.
of the sample handling optics. Noise comes principally Multi-element detectors with each element having its
from source fluctuations plus random electronic noise in own narrow band pass filter are available. At present,
the detector(s) and the amplifying circuits. because of size and optical considerations, the maximum
As shown in Figure 5, noise in the two channels, which number that can be installed in standard size cans is
is principally caused by minute thermal and schlieren four. Multi-element detectors have an advantage over filter
effects on and near the source, is roughly in synchronism. wheels in that their elements can be measured simultane-
When the two channels are simultaneously chopped and ously rather than sequentially as with the filter wheel with
the signals subtracted, most of the synchronous noise is the resulting signal-to-noise improvement.
eliminated resulting in as much as a 10-fold increase in
the S/N.
3.4 Variable wavelength filtometers
3.3 Multiwavelength filtometer The first variable wavelength filtometers were constructed
around the CVFs described in Section 2. Wavelength selec-
Multiwavelength filtometer systems are achieved either tion is accomplished by manually rotating the filter past a
with a filter wheel or a multi-element detector. They are slit. It is also possible to produce a low resolution, single
used when several components in a mixture are to be beam spectrum with a built-in motor drive.
quantified. A microprocessor version has come into widespread use
Filter wheel instruments (Figure 6) with as many as for process stream and ambient air monitoring. The instru-
15 individual filters are manufactured. Each filter is ment can be programmed to select specific wavelengths,
rotated past the detector sequentially; individual absorbance store absorbance data from each, and process the data to
Reference signal
display individual component concentrations. Because of
optical considerations mentioned earlier, wavelength rather
than frequency positions CVFs. However, the micropro-
cessor can be used to convert wavelength data into fre-
quency units.
LVFs are replacing CVFs for reasons of cost and flexi-
bility of design. Several instruments have been introduced
Analytical signal that move LVFs past a slit in front of a single detector to
achieve wavelength selection.
Another approach is to fabricate the LVF to cover a linear
detector array. The source image is either moved linearly
along the array to select individual wavelength/detectors
Combined signal 0.0001 or the image itself is distorted into a linear shape so that
Figure 5. Noise reduction possible with a dual-wavelength fil- the array is illuminated end-to-end. This latter procedure
tometer. enables an entire spectrum to be produced with each cycle
Aluminized Sample Internal

Aluminized
surface surface reflections
surface Sample Filters (IR /IA)
Linear
Reflector variable
filter
Detectors
Interface Source
connections Interface
Pyroelectric connections
array Figure 8. Liquid transmission filtometer optical diagram.
Embedded
Pulsable micro-electronics
infrared
source
Figure 7. LVF/array infrared spectrometer.
of the beam chopping frequency. It also gains S/N through

simultaneous wavelength measurement and multiscan sig-
nal averaging.
As a result of the availability of linear pyroelectric arrays
and LVFs, a simplified mid-infrared spectrometer has been
designed (Figure 7) that has virtually no air path or focusing
optics. The result is a very compact hand-held instrument
that can be used for material identification, quantitative
analysis and similar applications.
4 INTEGRATED SAMPLE HANDLING

Figure 9. A portable filtometer that accepts a variety of sample
Unlike infrared spectrometers, which are general-purpose cells. (Courtesy of Wilks Enterprise, Inc.)
instruments, filtometers are usually dedicated to specific
applications. For this reason, sample-handling equipment is Frequently, filtometers are used as process monitor-
ordinarily an integral part of the overall instrument. In some ing instruments and are equipped for these applications
models the source and detector optics are actually built with flow through cells, which can be based on ATR or
around the sample cell. It is important to note that nearly transmission.
all filtometers sample systems have few or no optical paths
exposed to ambient air, practically eliminating atmospheric
absorption effects. 4.2 Gas analysis
Perhaps the broadest use of infrared filtometers is in gas

4.1 Liquid sampling analysis. Here again the cell is designed into the filtometer
as an integral part. For high sensitivity, long cell path
The diagram in Figure 8 shows a typical layout for liquid lengths are required and the optical schematic of a folded
transmission sampling. (See Figure 7 for an attenuated total path cell is shown in Figure 10. Conventional 5 and 10 mm
reflection (ATR) sampling system.) Figure 9 is an infrared cells are used where concentrations greater than 1% are to
filtometer that accepts a variety of sample handling stages. be measured.
ATR is most often the method of choice for relatively high
concentration sampling because it provides the required
very short effective path lengths that are also very repro- 4.3 Optical fibers
ducible. It is best suited to viscous samples because of the
ease of covering and cleaning the ATR sampling surface. Infrared filtometers can be easily adapted to optical fibers
For trace and low concentration analyses, quartz cuvettes for remote sensing applications. The source and the detector
and conventional infrared transmission cells are used. can be coupled closely to the outgoing and returning fiber
Optical chopper
Reference
wavelength
Gas cell filter
Detector
Source Beam
splitter
Detector
Analytical
M2 M1 Sample Sample M4 M3 wavelength
in out
filter
Figure 10. Schematic of a dual-wavelength gas analyzer with a multipass gas cell.
bundles. Since fiber systems are relatively inefficient, the mid-infrared fibers resulting in greater sensitivity at lower
high signal-to-noise characteristic of the filtometer works cost. A typical design is shown in Figure 11.
to its advantage. A further advantage of filtometers over
FT-IR instruments is that the former can be located much
4.5 Reflectance measurements
closer to the point of measurement, thus greatly reducing
the required length of optical fibers. Because of their compactness and light weight, filtometers
are being adapted to surface measurements, either by specu-
lar or diffuse reflection. For diffuse reflection measurements
4.4 In-line infrared sensors of samples moving on conveyor belts, for example, fil-
tometers are mounted permanently above the conveyor line.
Infrared plugs, which consist of a pulsed source, a fil- Diffuse reflection measurements are generally made in the
ter/detector and a thermistor mounted adjacent to an ATR near-infrared where the inefficiency of collecting diffusely
optical element, can be inserted directly into a process reflected radiation is offset by the much higher intensity
stream to monitor a component concentration continu- of near-infrared radiation and infrared sensitivity of the
ously. Such infrared sensors can eliminate the need for detector.
Figure 11. An in-line infrared sensor with built-in source, filter detector and ATR sampling crystal. (Courtesy of Wilks Enterprise, Inc.)
Specular reflection measurements can be made in the measuring infrared absorbance at one or two wavelengths.
mid-infrared where specific wavelengths related to individ- When this is the case, an infrared filtometer can be used
ual coatings can be isolated by means of narrow band pass to supplement or replace an infrared spectrometer. Fur-
filters. Hand held versions of filtometers can be placed on thermore, filtometers can frequently be used in environ-
metal plates and the coating thickness determined. ments totally unsuited to spectrometers especially FT-IR
instruments.
Infrared filtometers have the following characteristics
5 CALIBRATION PROCEDURES, that will suggest where they may be applied:
SIGNAL PROCESSING They are not affected by vibration and will operate over
a broad ambient temperature range.
Present day infrared filtometers contain versatile micropro- They have few and in some cases no moving parts.
cessors that permit the installation of calibration data and Continuous monitoring models are designed to operate
various modes of operation. for long periods of time without attention.
Infrared spectrometers and early filtometer models typ- Portable models are small, compact and may operate
ically present data in terms of absorbance which is then from 12 V power supplies and thus can be used in the
converted to concentration by referring to previously pre- field.
pared calibration charts. The microprocessors contained in Filtometers, being dedicated to specific applications,
modern filtometers enable the absorbance vs concentration are simple, uncomplicated instruments that can be oper-
relationship to be programmed into them so that their read- ated by relatively untrained operators.
out is directly in the form of concentration in any desired Process monitoring instruments can be located close to
units. In some applications, simple go or no-go signals the process stream minimizing the required length for
are all that is required. sample lines and for optical fibers when used.
In some cases where the filtometer is to be dedi- Finally, there is the cost factor. Filtometers range in
cated to monitoring a single component, for example price from under US $5000 to as high as US $50 000
a refrigerant gas, calibration information can be factory for some process stream monitors. Typically, however,
installed. When a portable gas analyzer is used to check most filtometers are priced at under US $10 000, a
for the presence of a number of different gases, a range fraction of the cost of an FT-IR instrument.
of spectral profiles can be installed in the instruments Manufacturers of infrared filtometers will provide poten-
memory. tial users with assistance in adapting their instruments to
Field calibrations can be carried out in the usual fash- specific applications. This involves selecting the proper
ion by comparing instrument absorbance of concentra- optical filters and sample handling equipment, as well as
tions of known samples and inserting these relationships developing calibration procedures.
into the microprocessor memory. Calibration procedures
based on mid-infrared measurement are generally much
simpler than near-infrared procedures because the ana-
lytical absorption bands are usually sharper and more
7 FUTURE DEVELOPMENTS
isolated.
Like most analytical instruments, filtometers are bene-
fiting from technological developments in other fields.
The availability of pyroelectric detector arrays when com-
6 MID-INFRARED FILTOMETER
bined with mid-infrared LVFs and suitable software pro-
APPLICATIONS grams will result in wavelength programmable mid-infrared
filtometers of considerable versatility while still main-
As the size and scope of this handbook attests, the appli- taining the cost and portability advantage over FT-IR
cations of vibrational spectroscopy are very broad and spectrometers.
diverse. Infrared filtometers fill a niche that has been cre- Synthetic diamonds are becoming less expensive and
ated by the widespread use of infrared spectrometers, by larger in size to the extent that this material will soon
means of which, many practical applications of infrared be practical to use as ATR sampling elements in process
technology have been developed in environmental anal- filtometers. Its refractive index, mid-infrared transmission,
ysis, product quality, process control, materials identi- nontoxicity and chemical resistance make the diamond
fication, hazardous vapor monitoring and many others. ideally suited to analytical systems based on ATR. At
Frequently, the required information can be obtained by the same time, surface treatment techniques are improving
the surface resistance to erosion and coating of traditional oil monitoring, pharmaceutical quality control and chemical
ATR materials like zinc selenide and zinc sulfide. Moldable processing.
glasslike, mid-infrared transmitting materials, like AMTIR,
permit the fabrication of lower cost optical elements such as REFERENCE
hemispheres for in-line sensors. These and other advances
will result in the expanded use of mid-infrared filtometers 1. J.D. Rancourt, Optical Thin Films, McGraw-Hill, New
in such widely diverse fields as food analysis, lubricating York (1987).
Continuous Scanning Interferometers for
Mid-infrared Spectrometry
Richard S. Jackson
Bruker Optics Inc., Billerica, MA, USA
1 INTRODUCTION for an understanding of the performance limitations of inter-

ferometers, and many current design features. Modern inter-
The standard interferometer used in mid-infrared spectrom- ferometers can be broken down into a set of components,
etry has a long history. In 1887 a two-beam interferome- a large number of which are common to many interferom-
ter was used in the famous MichelsonMorley ether-drift eter designs. These components are dealt with in the next
experiment,1 designed to determine whether or not light section, and their properties are discussed in the context of
needed a medium in which to travel. Some years later, their effect on interferometer performance. Finally, several
the same interferometer was used in the first systematic examples of modern interferometers are given to illustrate
study of the fine structure of spectral lines.2,3 The com- the points made in earlier sections, and to show how inter-
putational requirements of the Fourier transform, however, ferometer design is affected by the intended end use of the
precluded the routine use of the interferometer for studies spectrometer.
of complex spectra until the 1960s. At that time, Fourier
transform spectrometers, based on the Michelson interfer- 2 PRINCIPLE OF OPERATION OF AN
ometer, became commercially available for mid-infrared
spectroscopy. Fourier transform spectrometers have now INTERFEROMETER
almost completely replaced dispersive instruments for mid-
Figure 1 shows a schematic diagram of a basic Michelson
infrared spectroscopy because of their inherent advantages.
interferometer, as used in a mid-infrared spectrometer, with
Although the interferometer in all modern Fourier transform
all the essential major components. Light from the infrared
infrared spectrometers is based on the design of Michelson,
source is focused onto the Jacquinot stop, A, an aperture
over the years a large number of variations have evolved
that limits the size of the source seen by the interferom-
to meet the requirements of a wide range of applications.
eter. Together with the mirror, B, which limits the solid
Fourier transform infrared spectrometers are now used in
angle seen by the interferometer, the Jacquinot stop con-
areas as diverse as the research laboratory, satellites, pro-
trols the etendue of the spectrometer, and thus the beam
cess monitoring systems, ship- and balloon-borne systems
divergence through the interferometer. The purpose of the
for the measurement of the stratosphere, and portable mili-
Jacquinot stop will be discussed in more depth later. After
tary units for the remote sensing of chemical warfare agents.
the Jacquinot stop the light is collimated and directed to the
This article initially discusses the basic principles of
beamsplitter assembly, C, where, in an ideal case, it is split
operation of interferometers used for mid-infrared spec-
into two beams of equal intensity. In a mid-infrared interfer-
troscopy, to lay the groundwork necessary for later sections.
ometer the beamsplitter is a thin film, D, which is deposited
Sources of error in interferometers are then discussed in on a substrate, E. Since the substrate changes the optical
some detail, since an understanding of these is necessary pathlength in one arm of the interferometer, a compensat-
ing plate, F, of equal thickness, is placed in the other arm
John Wiley & Sons Ltd, 2002. of the interferometer. After the beamsplitter, one of the two
destructive interference at different optical retardations. The

G intensity at the detector, I, is thus a function of the opti-
I
B
cal retardation , and is called the interference record. In
C
an ideal interferometer, when the optical retardation is zero,
which is often referred to as zero path difference, all the
wavelengths simultaneously constructively interfere. At this
point a very large signal, referred to as the centerburst, is
observed at the detector. For an ideal interferometer and
a polychromatic source with intensity Il the interference
E record is given by
H D
1

F I D 0.5Il 1 C cos 2p dl 1
J 1 l
Only the modulated component is generally of interest for
A
spectroscopy, and the direct current component is therefore
Figure 1. Schematic diagram of a standard Michelson interfer-
filtered out. The resulting (alternating current) signal is
ometer. The direction of travel of the light is indicated by the
arrows. Details are given in the text. referred to as the interferogram. In practice an efficiency
term, hl, must be included to account for factors such
beams passes to a fixed mirror, G, is reflected, and returns as beamsplitter efficiency, detector response, etc., and the
to the beamsplitter. The other beam passes to a mirror, H, interference pattern contains a wavelength dependent phase
that is mounted on a bearing, with a drive that permits it to term (see Section 3.5). All the intensity terms can be
be displaced along the optical axis. After being reflected, combined to give the overall instrument spectral response,
it also returns to the beamsplitter, where it combines with BnQ , which is a function of wavenumber, nQ , the reciprocal
the beam from the fixed mirror. After the beamsplitter, on of wavelength. The interferogram can then be expressed as
1
average half the light returns along the incident optical path
I D BnQ exp2pinQ dnQ 2
to the source and half is directed to the sample, usually by a 1
focusing mirror. The light is finally focused onto a detector Equation (2) is a Fourier transform, the solution of which
after transmission through, or reflection from, the sample. yields the single beam spectrum, BnQ .
In all modern spectrometers the beam from a HeNe laser, In practice the interferogram described by equation (2)
I, is also passed through the interferometer, and is detected must be digitized, which has two important consequences.
by one or more laser detectors, J. The purpose of the HeNe The first is that the range of the integral is restricted to some
laser is discussed in depth later. finite optical retardation, , which limits the resolution of
To envision how the interferometer works it is easier the resulting spectrum. This truncation of the interferogram
to initially consider monochromatic light of wavelength also leads to a sinc (i.e. sin x/x) instrument line shape,
l, such as that from the reference laser, and an ideal which is usually undesirable because it has large side lobes.
beamsplitter. The difference in the distances the light has These side lobes are therefore generally suppressed by
to travel in the two arms of the interferometer is referred to the process of apodization, in which the interferogram is
as the optical retardation. If the optical retardation is zero, multiplied by a weighting function that is equal to one at
constructive interference occurs between the two beams and the centerburst and decreases monotonically with increasing
all the light passes to the detector. If the moving mirror is retardation. If the interferogram is triangularly apodized, i.e.
displaced by an amount 14 l, then the optical retardation is it is multiplied by a linear ramp function that is equal to
1
2
l, and destructive interference occurs between the two one at the centerburst and zero at , then the instrument
beams so that no light is passed to the detector, and line shape is given by a sinc2 function:
all the light returns to the source. If the moving mirror
is displaced a distance 12 l then the optical retardation sin2 pnQ
AnQ D 3
is l, and constructive interference again occurs. Thus, if pnQ 2
the mirror is moved at constant velocity the intensity at which is the Fourier transform of the apodization function.
the detector varies sinusoidally as the interference cycles. For a sinc2 instrument line shape, two narrow spectral lines
Maxima occur when the optical retardation is an integer separated by 1/ cm1 will have a dip of 20% of the peak
number of wavelengths, and minima occur when the optical intensity between them. This satisfies the Rayleigh criterion
retardation is a half-integer number of wavelengths. for the resolution of two spectral lines,
A broadband source is composed of a continuous dis- 1
tribution of wavelengths, which undergo constructive and nQ D 4

Continuous Scanning Interferometers for Mid-infrared Spectrometry 3
and is therefore defined as the nominal resolution of the the HeNe laser are also recorded, then digital signal pro-
spectrometer.4 The second consequence of digitizing the cessing can be used to convert the data points from equal
interferogram is that the integral in equation (2) is replaced intervals of time to equal intervals of optical retardation. At
by a discrete sum. According to the Nyquist theorem, this the time of writing, three commercially available Fourier
limits the spectral bandwidth that can be measured to half transform spectrometers use this method of sampling the
the sampling frequency. infrared signal: the Spectrum 1 from Perkin-Elmer (Bea-
If the mirror is moved with a constant physical velocity, consfield, UK), the Matrix from Bruker Optics Inc. (Biller-
v, then the wavenumber dependent frequency of the detector ica, MA, USA), and the interferometers made by Manning
signal is given by Applied Technology (Troy, ID).
Regardless of the exact sampling interval, the fact that
fnQ D 2vnQ Hz 5 it is referenced to a HeNe laser gives a Fourier trans-
The frequencies, fnQ , are usually referred to as the Fourier form spectrometer its inherent wavelength stability. The
frequencies. The interferometer scan speed is usually cho- frequency precision of a Fourier transform spectrometer
sen in the range 0.1 to 10 cm s1 , so that the Fourier in the mid-infrared is often quoted as 0.01 cm1 , which
frequencies are on the order of audio frequencies and the is readily achievable on a short-term basis. The long-term
signal is therefore easily digitized. frequency precision is on the order of 0.04 cm1 , the error
In addition to the infrared beam, the beam from a HeNe being due to changes in the laser cavity length and in inter-
laser is passed through the interferometer. Since the laser ferometer alignment with temperature.6 It is possible to
is highly monochromatic a sinusoidal signal is generated achieve a long-term wavelength stability of approximately
by the laser detector. As in the case of the infrared sig- 1 part in 107 by using a frequency-stabilized HeNe laser,
nal, electronic filters are used to remove the direct current but because of cost this is usually only done in very high
component. In an infrared spectrometer this laser signal resolution instruments. It is worth noting at this point that
serves multiple purposes. In a continuous scanning Fourier the reason a laser is universally used as a reference in
transform spectrometer it is essential that the mirror move Fourier transform spectrometers is that it is inherently a
with a constant velocity to avoid spectral artifacts. This is better reference for velocity control and sampling interval
discussed in more detail later in this article. A constant than any physical, inductive, or capacitive method. This is
velocity is achieved by comparing the frequency of the because it directly measures optical retardation and optical
laser signal to that of a quartz oscillator, and using the velocity, whereas physical or capacitive methods assume a
resulting error signal in a feedback loop to correct veloc- constant relationship between optical retardation and phys-
ity fluctuations. The fast Fourier transform algorithm used ical retardation. There is error associated with the use of a
in all spectrometers also requires that the interferogram be HeNe laser as a reference, because fluctuations in the refrac-
sampled at equal intervals of optical retardation. This cri- tive index of the air cause slight changes in the optical
terion can readily be satisfied by triggering the sampling retardation at the HeNe wavelength relative to the opti-
of the infrared signal at the zero-crossings of the laser cal retardation at infrared wavelengths.6 This error is very
interference signal. A HeNe laser emits light with a wave- small, however, and is only non-negligible for specialized
length of 632.8 nm (15 798 cm1 ), so if the infrared signal measurements, such as those at very high resolution, or for
is sampled at every zero crossing, i.e. twice per HeNe step scan interferometry (see Instrumentation for Step-
wavelength, then according to the Nyquist theorem the scan FT-IR Modulation Spectrometry). In these cases it
maximum frequency measured corresponds to 15 798 cm1 . can be eliminated by the use of a vacuum spectrometer, or
It is sometimes desirable to digitize the infrared signal at careful thermal isolation and control.
every second, third, etc., zero crossing, giving a correspond- The final, and very important, use of the laser is to ensure
ing reduction in spectral range. For measurements in the that interferograms from consecutive scans are added coher-
visible or ultraviolet (UV), several modern spectrometers ently. This means that corresponding data points in all the
also incorporate electronics that double or quadruple the interferograms that are to be coadded must be acquired at
frequency of the laser signal, allowing a free spectral range exactly the same optical retardation. One method of achiev-
of more than 63 000 cm1 . ing this is to pass a beam of white light through an auxiliary
Recently, Fourier transform spectrometers have been interferometer attached to the same drive mechanism as the
built in which the sampling of the infrared signal is not primary (i.e. infrared) interferometer. The white light inter-
synchronized to the zero crossings of the HeNe laser. In a ferogram that is generated has a very sharp centerburst, and
scheme originally proposed by Brault,5 a fast sigma-delta the interferometer is constructed such that this centerburst
converter is used to sample the infrared interferograms at is displaced relative to the infrared centerburst. The start
equal intervals of time. If the times of the zero crossings of of scan is triggered by the white light centerburst, and the
infrared interferogram is then digitized at the zero-crossings travel is therefore always known. Laser quadrature also has
of the HeNe laser interferogram. Although this method the advantage that if one laser interferogram is close to a
ensures coherent addition of interferograms, it suffers from zero crossing at the mirror turnaround, then the other is not.
the limitations that only single-sided interferograms may be The possibility that noise in the laser signals can cause the
acquired, and only in one direction of mirror motion. This number of fringes to be miscounted is therefore eliminated.
leads to a very poor duty cycle (the ratio of the time during Laser quadrature does have the drawback that it is more
which data acquisition occurs to the total scan time). Fur- expensive to implement than interferogram correlation, but
ther, better photometric accuracy is obtained using double- interferometers employing this technique can scan indefi-
sided interferograms,7 and they are required for certain nitely, even under harsh conditions, without losing count
experiments (including many emission experiments such as of the zero-crossings.
Fourier transform Raman) where phase correction of single-
sided interferograms is not reliable.
All of the limitations imposed by using a white light
3 SOURCES OF ERROR IN
interferogram can be obviated by fringe counting, a tech- INTERFEROMETERS
nique used in almost all modern spectrometers. With suf-
ficiently fast electronics it is possible to count the zero- There are many sources of error in Fourier transform
crossings of the HeNe laser interferogram, and thereby keep spectroscopy,8 a number of which occur in the interferom-
track of the mirror position. This method is much cheaper eter. Sources of error in the interferometer may affect the
to implement than the white light interferogram, and allows resolution, wavenumber precision or accuracy, signal-to-
double-sided interferograms to be acquired in both direc- noise ratio, photometric precision or accuracy, or baseline
tions of mirror travel. The only potential problem occurs stability. An understanding of these effects is therefore
when the mirror reverses direction at each end of the scan. important. Many aspects of interferometer design are in
It is difficult to determine the direction of mirror travel fact aimed at reducing or eliminating these errors. Each of
when the velocity is close to zero at the turnaround points, the major sources of error is discussed below.
because the appearance of the laser interferogram is the
same for both directions. If the mirror reverses direction 3.1 Beam divergence
close to a zero-crossing then small amounts of noise in
the laser interferogram, for example from vibrations, may Equation (2) given above to describe the interferogram
cause the electronics to miscount the number of crossings. is only strictly valid if the beam passing through the
The counting error is never large though, and the problem interferometer is perfectly collimated. In practice this is not
can usually be alleviated by correlating the new interfero- possible, the amount of beam divergence depending on the
gram to the previously acquired average interferogram, and size of the source and the focal length of the collimating
if necessary shifting it to get coherent addition. The only mirror. The effective size of the source is therefore usually
case in which the combination of fringe counting and inter- limited by a stop, called the Jacquinot stop, which is shown
ferogram correlation is not reliable is when the noise level as A in Figure 1. In some interferometers, the detector or
in individual interferograms is very large. In this case it is source act as the Jacquinot stop, but the principle is the
possible for interferogram correlation to actually introduce same. For any finite-sized source, there are off-axis rays
large errors, and it must be disabled. that undergo a different optical retardation to on-axis rays.
A common method used to track the direction of the mov- Figure 2 shows an interferometer with both on-axis rays,
ing mirror, and thus eliminate the need for interferogram and off-axis rays that subtend an angle a to the optical
correlation, is called laser quadrature. There are many vari- axis. The on-axis ray is reflected from mirror M1, and from
ations of this, but they all work on the same basic principle. mirror M2. Mirror M2 has moved a distance L from the
Interferometers employing laser quadrature control incorpo- position, M20 , at which there is zero path difference. The
rate a second laser detector, usually in the other arm of the difference in the pathlengths between the two arms of the
interferometer. The laser beam that is incident on one of the interferometer, or the optical retardation, for the on-axis ray
detectors has an optical phase lag of a quarter wavelength is thus 2L.
relative to the laser beam at the other detector, so that there To determine the optical retardation for the off-axis ray
is a 90 phase difference between the two laser interfero- we can consider just the mirror surfaces M2 and M20 , in
grams. When the moving mirror is traveling in one direction one arm of the interferometer, since M20 is geometrically
the first laser interferogram leads the second by 90 , and equivalent to M1. The two rays follow an identical path to
when the direction of travel is reversed the second laser the point marked A and, assuming a plane wavefront, have
interferogram leads the first by 90 . The direction of mirror an equivalent path after the line BC. For the lower ray,
The optical retardation, 2L, at which this occurs must be no

L
greater than that required to achieve the desired resolution,
C
nQ . Since, from equation (4), the resolution is 1/2L, we

therefore have
AB M1 nQ
a 11
nQ max
In practice, for conventional low resolution (>4 cm1 )

measurements in the mid-infrared the etendue of the spec-
trometer is usually limited by either the size of the source
or the size of the detector, and the beam divergence in the
interferometer will be less than that given in equation (11).
M2 M2 For high resolution measurements, however, or measure-
ments at shorter wavelengths, the size of the Jaquinot stop
Figure 2. Schematic diagram showing the paths of on-axis and must be reduced, which limits the signal-to-noise ratio of
off-axis rays passing through a Michelson interferometer. The rays the spectrometer. There are also certain types of emis-
are shown for two positions of the moving mirror, M2. The path
sion measurements, such as those of airglow, in which
of the rays when the optical retardation is greater are shown as
dashed lines. the source can be large, but very weak. For these types of
measurements it would be desirable to have an interferom-
the pathlength AB is given by eter in which the optical path difference did not depend
as strongly on the angle subtended at the beamsplitter.
sin2 a
AB D 2L 6 Interferometers in which this is the case are termed field-
cos a widened, since for a given resolution the field-of-view is
For the upper ray, the pathlength AC is given by increased. An example of a field-widened interferometer is
2L shown in Figure 3. This design, which was suggested by
AC D 7 Bouchareine and Connes,9 uses back-silvered quartz prisms
cos a
instead of the mirrors in a conventional Michelson inter-
The optical retardation for the off axis ray is thus ferometer. Since the prisms have a refractive index that
is greater than that of air, the optical retardation can be
D AC AB
increased by moving one of the prisms perpendicular to
D 2L cos a 8
Thus, for a given displacement of the moving mirror, the

optical retardation for an off-axis ray is shorter than for an
on-axis ray. The mirror must therefore move a greater phys-
ical distance to pass through a complete interference fringe
for the off-axis ray, which means that it appears to have
a longer wavelength. Since off-axis rays are indistinguish-
able from on-axis rays with a slightly longer wavelength,
the effect of beam divergence in an interferometer is to
limit the achievable resolution. We can obtain a quantita-
tive measure of this effect by considering the optical path
difference between the on-axis and off-axis rays:
x D 2L1 cos a
La2 9
As the optical retardation is increased, the on-axis and off-

axis rays will be exactly out of phase when the optical path
difference, x, is equal to half a wavelength. If the highest
wavenumber of interest is nQ max , then the first time this will Figure 3. Schematic diagram of a refractively scanned, field-
occur is when widened interferometer, as proposed by Bouchareine and Connes.9
1 The large arrow indicates the movement of the back-silvered
L 10 quartz prism that is used to change the optical retardation.
2nQ a2
the beam direction, which inserts more material in one arm 3.3 Velocity errors
of the interferometer. This design is thus an example of a
refractively scanned interferometer. It can be shown that Velocity errors have already been briefly discussed. They
refraction of the light at the prism faces results in an opti- arise mainly from either poor bearing or drive quality, or
cal path difference that, to a first order approximation, is vibration. The simplest case to consider is one in which
independent of incident angle. This interferometer design the constant velocity has a sinusoidal perturbation, of fre-
is not suitable for high resolution measurements, because quency fe , superimposed upon it, which causes a secondary
very large prisms would be required, but for measure- modulation of the interferogram. If there is a phase lag
ments of extended sources the sensitivity can be up to 20 between the HeNe laser and the infrared detection chan-
times higher than that obtained with a standard Michelson nels, then the resulting sampling error leads to sidebands,
interferometer. so that energy at wavenumber nQ also appears at nQ fe /2v.
It should be clear from equation (9) that in addition In practice this situation can be caused by vibration from,
to its effect on resolution, changing the etendue of the for example, pumps or air conditioners, and manifests itself
interferometer will also change the wavenumber calibration. as ghost peaks that are identical to the real peak but of a
The measured position of a peak will be a weighted average different, and possibly negative, intensity. If the vibration
of the positions that would be found from the on-axis and is more complex, or the interferometer has a bad bearing,
off-axis rays, which undergo different optical retardations. then energy from all wavenumbers effectively appears at all
A change in the etendue will not only occur when the other wavenumbers, leading to serious performance degra-
Jaquinot stop is changed, but also any time the sample, or dation. While velocity errors due to poor bearing quality are
sampling accessory, vignettes the beam. A derivation of the easily remedied by using a better bearing, those caused by
magnitude of this effect is beyond the scope of this article, external vibrations are not so easily solved. It is sometimes
but it can be shown10,11 that the measured wavenumber, possible to provide vibration isolation for the whole spec-
nQ m , is given by trometer, but it is generally not possible to provide more
than a limited amount of vibration isolation between the
a2
nQ m D nQ c 1 12 interferometer and the spectrometer baseplate, because of
2
problems caused by the input and exit beams moving rel-
where nQ c is the wavenumber that would be found from the ative to other optical components. There are a number of
on-axis ray. viable solutions for velocity errors caused by vibration. One
method, which is employed by Midac Corporation (Irvine,
CA, USA) in interferometers designed for remote sens-
3.2 Sampling errors ing, is to use very high bandwidth servo loops to control
the mirror. This is possible as long as the interferometer
It was stated earlier in this article that the infrared signal has very high mechanical stiffness, so that it has no res-
must be digitized at equal intervals of optical retardation. onant frequencies within the servo bandwidth. In practice,
Errors of only a few percent in the sampling positions on the such a high stiffness is not easy to achieve, but very care-
interferogram can lead to significant errors in the measured ful design of the Midac interferometer has resulted in a
intensities. The effect of sampling error on the measured minimum resonant frequency of approximately 1000 Hz.13
spectrum depends on whether or not the errors are sys- The interferometer in the IFS 66 series spectrometers from
tematic or random. The most common type of systematic Bruker Optics Inc. also has a high minimum resonant fre-
error occurs when the infrared interferogram is sampled quency of 800 Hz. A second method is employed by Bruker
at every zero crossing of the HeNe laser interferogram. Optics Inc. and On-Line Technologies (East Hartford, CT,
Offsets in the comparators can then lead to slightly dif- USA) in interferometers designed for quality control and
ferent sampling points for the rising and falling parts of on-line monitoring. Both of these interferometers have piv-
the laser interferogram, and therefore to slightly different oting mirror mechanisms, with the center of gravity of the
retardations between these points. This leads to ghost peaks mechanism located such that external lateral forces do not
at multiples of half the laser wavenumber. Random errors tend to induce rotation. All four of these interferometers are
in sampling position lead to a reduction in the achievable described in more detail later in this article. Another solu-
signal-to-noise ratio of the spectrum. It can be shown12 that tion to the problem of velocity errors is to match the phase
for a positional error l, the maximum signal-to-noise ratio delays of the HeNe laser and infrared signal channels, in
is given by which case the sampling errors caused by velocity fluctua-
4 tions are minimized.14 This concept has been extended, and
S/Nmax D 13 significantly improved, in instruments in which the infrared
lnQ max
signal is sampled at equal increments of time, rather than automatically align the interferometer when the spectrom-
equal increments of optical retardation. In these spectrometer is first turned on. The major detrimental effect from
eters, which were briefly mentioned earlier, it is possible this type of misalignment is long-term baseline drifts due
to compensate for the filter delays during the digital signal to effects such as temperature changes in the laboratory.
processing that converts the interferogram from equal incre- Modern interferometers compensate for these long-term
ments of time to equal increments of optical retardation.5,15 alignment drifts by using one or more of the solutions dis-
cussed below for dynamic misalignment.
In the case of dynamic misalignment, the alignment
3.4 Alignment errors changes during the scan, and the effects are somewhat dif-
ferent than with a static misalignment. Tilt of the moving
Misalignment of an interferometer causes parallel rays mirror of a Michelson interferometer during the scan is in
passing through different parts of the interferometer to fact a frequently quoted source of error in interferometers,
undergo slightly different optical retardations. This is shown but it is really just a special case of the general problem of
schematically in Figure 4, for a Michelson interferometer in dynamic misalignment. In any interferometer there are six
which one of the mirrors is tilted. If the beam has a diameter possible motions for the moving mirror mechanism: three
D, then a small tilt of one mirror by an angle q leads to a rotational and three translational. In an ideal interferometer,
difference in optical retardation, , for the extreme rays of all but one of these motions either would be impossible or
approximately would have no effect on either interferometer alignment or
D 2D tan q 14 optical retardation. If an undesired motion affects the inter-
ferometer alignment then stringent requirements are placed
The interference records for the extreme rays are therefore on mechanical tolerances, and the interferometer tends to be
out of phase with each other by an amount susceptible to vibration. The effect of undesired motion on
4pD optical retardation is usually less severe, since this is largely
f D tan q 15 compensated for by the use of the reference laser. The sim-
l
plest case to consider is that of a Michelson interferometer,
This leads to a loss of modulation efficiency in the interfer- in which the principal undesired motion is tilt of the mov-
ometer, the effect being more pronounced at short wave- ing mirror. If the tilt is caused by nonstraightness of the
lengths. For a static misalignment, i.e. one in which the mirror travel, and it is severe enough, then it causes a loss
degree of misalignment is independent of the position of of resolution because the interferogram becomes steadily
the moving mirror, this leads to a loss of signal at the more attenuated with increasing optical retardation. This
detector. Over short time frames, this signal loss is con- problem tends to manifest itself in spectrometers designed
stant and does not pose a serious problem in interferometer for high resolution (nQ <0.05 cm1 ), especially if measure-
design. Almost all interferometers have either fine pitch ments are to be made in the visible or UV regions of the
screws for alignment of one or more optical components, spectrum. The more common problem in medium and low
or use the dynamic alignment mechanism (see below) to resolution interferometers is that the mirror tilt is not repro-
ducible from scan to scan, in which case it leads to baseline
instabilities. The degree to which the mirror tilt must be
controlled depends on both the shortest wavelength that is
to be measured, and the purpose of the measurement. If
2 the maximum permissible difference in optical retardation
D across the width of the beam is one-tenth of the short-
est wavelength, lmin , being measured (a commonly used
criterion), then for small q
lmin
q< 16
20D
The mid-infrared spectral region extends to 4000 cm1 ,
and therefore for a spectrometer with a beam diameter of
5 cm the maximum permissible tilt is only 2.5 rad. Many
Figure 4. Schematic diagram showing the effect of mirror mis- mid-infrared spectrometers are also designed with inter-
alignment on optical retardation in a Michelson interferometer. changeable optical components that allow measurements to
The direction of travel of the light is indicated by the arrows. be made in other spectral regions. In such spectrometers the
design of the interferometer to reduce or eliminate errors Examples of such interferometers (those in the Bruker IFS
such as mirror tilt is critical. For example, if the spectrom- 66 and Midac spectrometers) are given later in this article.
eter is designed to also be capable of measurements in the
near-infrared and visible regions of the spectrum, then the 3.4.2 Dynamic alignment
maximum permissible mirror tilt is reduced to only about
0.5 rad. Dynamic alignment was originally developed by Buijs,16
There are three methods of controlling the errors caused and subsequently commercialized by Bomem (Quebec,
by dynamic misalignment. The simplest of these in prin- Quebec, Canada), in the late 1970s in their DA series of
ciple, although not necessarily in practice, is to elimi- spectrometers. These spectrometers were designed for very
nate the misalignment mechanically by reducing all unde- high resolution (nQ 0.003 cm1 ) work, and it is not pos-
sired motions to a negligible level. The second method is sible to mechanically control mirror tilt over such large
dynamic alignment, in which the alignment of the interfer- retardations. Bomem have also shown that dynamic align-
ometer is actively controlled by a servo system throughout ment can control mirror tilt even when the instrument
the entire scan. The third method is that of optical compen- is subject to significant vibration,17,18 such as in systems
sation, in which the interferometer optics are designed such installed in aircraft or balloons. Dynamic alignment is
that undesired motions do not change the optical alignment. now also used in medium and low resolution laboratory
Each of these methods has its own advantages and disad- instruments from several manufacturers, including Nicolet,
vantages, and more than one is often used in any particular Digilab (Randolph, MA, USA), formerly Bio-Rad (Cam-
interferometer design. Which methods are incorporated in bridge, MA, USA), and Shimadzu (Tokyo, Japan). The
an interferometer design depends on the preferences of principle of dynamic alignment is shown in Figure 5. The
the designer, and the purpose for which the interferome- HeNe laser beam is either expanded to a diameter of a few
ter is being built. Each of these methods will be considered millimeters, or split into three paths, and then detected by
in turn. either three segments of a quadrant detector or three sep-
arate detectors. The three detectors yield three sinusoidal
interferograms, the relative phases of which depend on the
3.4.1 Mechanical design relative optical retardation of the three beams. Any tilt of
the moving mirror changes the relative phases of the three
At least some of the undesired motions in an interferometer
signals, and electronics then control a servo system that
are usually restricted mechanically so that they are within
realigns the interferometer. Although the diagram shows an
acceptable limits. In some cases, this is easy to accom-
interferometer in which the fixed mirror is realigned by the
plish, such as in the case of rotation of a plane mirror
servo system, interferometers have also been built in which
about the drive axis in a Michelson interferometer. This
the alignment of either the beamsplitter or the moving mir-
will lead to changes in interferometer alignment if the mir-
ror is adjusted. In Figure 5, for clarity an interferometer
ror is not mounted exactly perpendicular to the axis. Several
methods have been employed to control roll in Michel-
son interferometers that utilize cylindrical bearings. These
include springs that are attached to the rear of the drive, or
an extension that is attached to the bearing, with a sleeve
that slides over a bar coated with poly(tetrafluoroethylene)
(PTFE). An ingenious design is employed by Thermo Nico-
let (Madison, WI, USA) hereafter referred to simply as
Nicolet in their current spectrometers, one of which is
discussed in more detail in a later section. A small, but pow-
erful, magnet that is mounted on the bearing tracks a narrow
steel bar that is aligned parallel to the direction of mirror
motion. Other Michelson interferometers solve the problem
of roll about the drive axis by using noncylindrical bearings,
or occasionally even dual bearings. Although mechanical
control of bearing roll is simple, this is not necessarily
the case for other undesired motions in the interferome- Figure 5. Diagram showing the path of three parallel laser beams
through a Michelson interferometer. The three beams are detected
ter. To mechanically eliminate motions such as mirror tilt, by three separate detectors, and the signals from these detectors
the interferometer must either be built to very precise toler- are then fed to three actuators that control the alignment of one
ances, or the drive mechanism must be carefully designed. of the mirrors.
with three separate detectors is shown. This arrangement a change of refractive index with wavenumber), the optical
has two advantages over the use of a single quadrant detec- pathlength in the arm of the interferometer that contains the
tor. The first of these is that the laser beams, and thus substrate has a wavenumber dependence. A compensating
the laser detectors, are moved out of the infrared beam plate, made of the same material as the substrate, and of
path. In addition, the sensitivity of the system to mirror the same thickness, is therefore placed in the other arm of
tilt is maximized, because the distance between the detec- the interferometer. Thus, although the optical path length
tors is as large as possible. This consideration is especially in each arm of the interferometer exhibits a wavenumber
important if the interferometer is to be used at very short dependence, the optical path difference does not. In prac-
wavelengths. Compared to a quadrant detector, however, tice, it is not possible to make the substrate and compensat-
three detectors are expensive to implement, and the associ- ing plates exactly the same thickness, and there is therefore
ated optics require additional alignment. Both schemes are a small change in optical path difference with wavenum-
currently in use in commercial instruments: Nicolet uses ber. This results in a small, wavenumber dependent, phase
quadrant detectors in all their instruments, whilst Digilab shift in the interference record, referred to as phase errors,
uses separate detectors in all their spectrometers, including or chirping of the interferogram. As long as the resulting
the FTS 6000 spectrometer that has a spectral range that asymmetry in the interferogram is small, however, these
extends into the UV. errors can be corrected using software.4 It is worth noting
that additional phase errors can be introduced by the elec-
3.4.3 Optical compensation tronic filters in the spectrometer, but since these are not part
of the interferometer they will not be considered further.
In an interferometer that is optically compensated, small,
undesirable motions of the moving mirror may still occur,
but the optical design is such that some or all of the 3.6 Double modulation
more deleterious motions have a negligible effect on the
interferometer alignment. Clearly, in a standard Michel- There are many factors in an infrared spectrometer that limit
son interferometer, although mirror tilt is a problem, the the photometric accuracy or precision, but most of these
alignment of the moving mirror is immune to any shear are beyond the scope of this article. Readers are referred
motion in the plane of the mirror. If the plane mirrors to the book by Griffiths and de Haseth4 and the article by
in a Michelson interferometer are replaced by cube-corner Birch and Clarke8 for more information. One effect will
retroreflectors (see below), however, then the reverse is be considered, however, because it is significant, and can
true. In a cube-corner Michelson interferometer the effect be eliminated by suitable design of the interferometer. If
of mirror tilt is minimized, but shear of the moving mir- light that exits the interferometer strikes a plane surface
ror will cause interferometer misalignment. The effect of that is perpendicular to the beam, such as a sample or a
shear in a cube-corner interferometer, however, is much window, then some of it will be reflected back into the
smaller than the effect of tilt in a plane mirror interfer- interferometer. This light is modulated a second time, and
ometer, and undesired motion is therefore much easier to half of it again exits the interferometer in the direction
control mechanically. Cube-corner Michelson interferom- of the detector (the other half returning to the source).
eters are currently used in all Thermo Mattson (Madison, Because the light has been modulated twice the interference
WI, USA) and Jasco (Tokyo, Japan) interferometers, as well record contains components at double the frequency of the
as Brukers 120 HR spectrometer. Optical elements that are light that was modulated only once. The doubly modulated
used in modern interferometers for optical compensation are light therefore appears to be at twice its true wavenumber.
discussed in more detail later, in the section on interferome- Some idea of the magnitude of this effect can be gained
ter components. Several examples of optically compensated by considering a sample of refractive index 1.5, which will
interferometers are also given at the end of the article. reflect about 4% of the incident light. After being modulated
a second time half of this, or about 2%, will be returned
to the detector. Since 10% transmission corresponds to
3.5 Phase errors an absorbance of 1, and 1% transmission corresponds
to an absorbance of 2, this amount of stray light can
During the discussion on interferometer operation earlier obviously have a significant effect on measurements of
in this article, it was stated that the beamsplitter assem- strongly absorbing samples. The problem is compounded
bly consisted of a thin film deposited on a substrate, and by the fact that a spectrometer usually contains several
a compensating plate. The substrate is necessary because such surfaces, and some of them may be made of a high
a mid-infrared beamsplitter film is too thin to be self- refractive index material that reflects considerably more
supporting. Because the substrate exhibits dispersion (i.e. than 4% of the incident light.
In some (although not all) interferometers that contain for s-polarized light (i.e. light polarized perpendicular to
cube-corner retroreflectors, this problem can be eliminated the plane of incidence) is 0.49, resulting in an average effi-
by blocking half of the beam as it exits the interferometer. ciency that is greater than 0.86. Although the efficiency of
By referring to figures shown later in this article, it can be a thin germanium film is high in the mid-infrared, it should
seen that light that is reflected back into the interferometer be noted that it causes significant polarization of the beam
in one half of the beam will be returned to the detector exiting the interferometer, and this effect is even worse for
in the other half. The disadvantage of blocking half the materials with a lower refractive index. The usable high
beam is of course that the signal-to-noise ratio of the spec- wavenumber limit of a germanium thin film beamsplitter
trometer is reduced. Beam blockers are therefore usually is about 5500 cm1 , after which the germanium becomes
only used when photometric accuracy is important, such as opaque. Multi-layer thin films can be used to increase the
for quantitative analysis of strongly absorbing samples, or usable spectral range of the beamsplitter, whilst simultane-
measurements of the transmission of optical materials. ously reducing the polarization effects. These multi-layer
films make use of thin film interference effects, and a
detailed discussion of their design is beyond the scope of
4 COMPONENTS OF AN this article.
INTERFEROMETER The dielectric or semiconducting thin films that are
used for mid-infrared beamsplitters are too thin to be self-
The major components of interferometers used for mid- supporting, and, as previously mentioned, they are therefore
infrared spectroscopy can be broken down into the beam- deposited onto a substrate. This substrate is a plate of opti-
splitter, the fixed mirror, and the scanner mechanism. The cally transparent material with very flat surfaces. If the
scanner mechanism itself consists of a mirror, a drive sys- surfaces of the substrate plate (or the compensating plate)
tem and a bearing. Each of these components comes in are not flat, then the optical path difference in the interfer-
many different types, and the successful design of an inter- ometer varies across the width of the beam, which reduces
ferometer relies on a suitable combination of components. the modulation efficiency in a way exactly analogous to
The choice of components also depends on the end design the case of interferometer misalignment discussed earlier.
goal: a spectrometer designed for on-line analysis has very As with interferometer misalignment, the effect is more
different requirements than a spectrometer designed for a severe at short wavelengths, and the ability of the substrate
research laboratory. Later in this article, various examples material to take, and hold, a very good polish is therefore
of interferometers are given that illustrate different aspects important. The most common material used as a substrate
of design. in mid-infrared spectrometers is potassium bromide. KBr
transmits from 400 cm1 up through the visible, but its
softness puts an upper limit on the usable spectral range of
4.1 Mid-infrared beamsplitters about 11 000 cm1 . KBr is also hygroscopic, and the inter-
ferometer must be either purged or desiccated to protect the
An ideal beamsplitter is nonabsorbing, and will transmit optics. If better short-wavelength performance is required
50% of the incident light and reflect 50% of the inci- than can be obtained using KBr, calcium fluoride can be
dent light. For real beamsplitters this is never the case at used as a substrate material instead. CaF2 does not transmit
all wavelengths, and a beamsplitter efficiency is therefore below 1200 cm1 , but the usable upper limit extends to the
defined as vacuum UV. The long wavelength performance of a mid-
hnQ D 4RnQ TnQ 17 infrared beamsplitter can be improved by the use of cesium
iodide as a substrate, which transmits to below 200 cm1 ,
where RnQ and TnQ are the wavenumber depen- but the material is very soft and has a tendency to cold-
dent reflectance and transmittance, respectively. For a flow, so that it has a practical upper limit of only about
nonabsorbing beamsplitter, RnQ C TnQ D 1, and thus for 5000 cm1 . CsI is also extremely hygroscopic, and great
0.28 RnQ 0.72, the beamsplitter efficiency exceeds care must be taken to protect the beamsplitter from atmo-
80%. Efficiencies this high can be obtained from a sin- spheric moisture. One other material that is sometimes used
gle thin film of material if the material has a high refractive as a beamsplitter substrate for mid-infrared spectroscopy is
index, and therefore in many mid-infrared beamsplitters this zinc selenide. This has a usable spectral range that extends
film is made of germanium. Germanium has a refractive from about 550 cm1 through the near-infrared, and has
index of 4.0, and therefore, for a 45 incidence angle, the the advantage that it is nonhygroscopic. This beamsplit-
reflectivity for p-polarized light (i.e. light polarized paral- ter substrate is therefore useful in applications in which
lel to the plane of incidence) is 0.24, and the reflectivity the interferometer may be in a harsh environment, such
as process monitoring or control. Zinc selenide does have by the interferometer drive system. Shearing the retrore-
a high refractive index, however, which results in high flector displaces the output beam at the beamsplitter, and
reflection losses. This is true even if the beamsplitter is therefore causes misalignment of the interferometer, but the
anti-reflection coated, because of the very broad wavelength effect is smaller than that produced by tilting a flat mir-
range over which it must be used. The performance of an ror, and it is therefore easier to eliminate mechanically.
interferometer that uses a beamsplitter with a zinc selenide It should be noted that if a retroreflector is used in one
substrate is therefore lower than the performance of the arm of the interferometer it must be used in both, other-
same interferometer that uses a beamsplitter with a KBr wise the beams returning from the two arms would not
substrate. meet at the beamsplitter. It is possible to use a combi-
nation of a cube-corner retroreflector and a flat mirror to
eliminate the effects of both tilt and shear. An interferom-
4.2 Mirrors and mirror assemblies eter with this mirror configuration was originally used for
length measurements by Terrien,19 and its properties have
The concept of optical compensation was introduced earlier, been discussed in detail by Murty.20 In this configuration,
in which optical elements in the interferometer minimize the which is shown in Figure 6(c), the flat mirror is fixed and
effect of undesired motions of the moving mirror. Figure 6 the retroreflector moves to change the optical retardation.
shows several mirror assemblies used in mid-infrared inter- The combination of the two mirrors ensures that the beam
ferometers, as well as the effect of mirror tilt and mirror returns along its own path, regardless of any tilt or shear of
shear on the beam path. The simplest moving mirror is the retroreflector, and it is therefore termed fully compen-
a single flat, shown in (a). Tilt of the flat mirror seriously sated. This design does have the disadvantage, however,
misaligns the interferometer, because it displaces the return- that the clear aperture of the interferometer is only half the
ing beam at the beamsplitter. Shear of the mirror, on the size of the retroreflector. This means that either the signal-
other hand, has no effect on either the alignment or the to-noise ratio of the spectrometer is reduced, or a large
optical retardation. Figure 6(b) shows a cube-corner retro-
retroreflector must be used, which is both heavy and expen-
reflector, which, as the name implies, consists of three flat
sive. A variation of the combination of retroreflector and
mirrors that are mutually perpendicular. Light entering the
flat mirror is currently used in the spectrometers manufac-
cube-corner reflects from all three faces before exiting par-
tured by Oriel Instruments (Stratford, CT, USA), and as one
allel to the entrance beam. The parallelism of the entrance
component in the interferometer manufactured by On-Line
and exit beams is maintained even if the retroreflector is
Technologies that is described in detail later in this article.
tilted or sheared from its correct position, and for small
Figure 6(d) shows a mirror assembly that works on a very
tilts there is negligible displacement of the output beam
similar principle to the combination of the retroreflector and
at the beamsplitter. Tilting the retroreflector does cause a
flat mirror. In this design, the retroreflector is replaced by
change in optical pathlength, but this can be compensated
a parallel mirror pair. Shear of this mirror pair along any
of the three translational axes results in no change in either
the interferometer alignment or the optical retardation. This
is also true for rotation of the mirror pair about either of the
two axes that are in the plane of the page. Rotation about
an axis perpendicular to the plane of the page also does not
change the interferometer alignment, but does change the
optical retardation. This optical element is incorporated in
the interferometer designs used by both On-Line Technolo-
gies and Perkin-Elmer, which will be described in detail in
the next section. In the Perkin-Elmer interferometer, rota-
tion of the parallel mirror pair is in fact used to control the
optical retardation.
While optical compensation is an important consideration
(a) (b) (c) (d)
in the design of an interferometer, other factors must also
Figure 6. Mirror assemblies that are commonly used in interfer- be considered. The most obvious of these is cost; depending
ometers designed for mid-infrared spectroscopy. The top row of on the design goals, it may be cheaper to control the
figures shows the beam path in a perfectly aligned interferome-
ter. The second and third rows show the effect of tilt and shear, effects of undesired mirror motions using either mechanical
respectively, on the beam path. The four types of mirror assembly design or dynamic alignment. Another factor that must be
are described in detail in the text. considered is the mass of the moving mirror mechanism.
An interferometer that is to be used for fast time resolved pretensioned steel wire or cable. In addition to the primary
measurements, for example, must be capable of scanning cable drive, however, the Bruker IFS 120HR also has a
as rapidly as possible, with as high a signal-to-noise ratio secondary voice coil drive. This interferometer is described
as can be achieved in the limited scan times available. in more detail later in this article.
Therefore, although the designs shown in Figures 6(b) or In an interferometer that is to be used only at the very
6(c) can be useful for spectrometers to be used in harsh long wavelengths of the far infrared, the mirror is often
environments, they are clearly less well suited to rapid stepped, rather than driven at a continuous velocity, and the
scan kinetics measurements. With the exception of the infrared beam is modulated using a chopper. In this case, it
interferometer used by Manning Applied Technologies, is possible to use a lead-screw to drive the interferometer.
which will be described in detail later, the single flat mirror This has the advantage that the interferometer can achieve
shown in Figure 6(a) is in fact the best choice, since for a large optical retardations, and therefore high resolutions.
given clear aperture it has the smallest possible mass. Even Lead screws are currently used by both Specac (Orping-
for spectrometers that are to be used at only moderate scan ton, Kent, UK) and Sciencetech (London, Ontario, Canada)
speeds, in a given interferometer increasing the mass of the in their far-infrared spectrometers. More information about
moving mirror assembly will reduce the efficiency of the interferometers used for far-infrared spectrometry can be
duty cycle, and therefore degrade performance. found in another article in this volume (see Instrumenta-
Although most commercially available Fourier transform tion for Far-infrared Spectroscopy).
spectrometers are designed primarily for operation in the The principle of operation of voice coil and galvanometer
infrared, some are also capable of measurements at shorter drives is the same. They consist of one or more fixed mag-
wavelengths, even to the edge of the vacuum UV. In these nets and one or more coils through which current is passed.
spectrometers, the reflectivity of the mirror assemblies must The magnetic field induced by the current interacts with
be considered. In the near-infrared region of the spectrum the permanent magnetic field to produce a driving force,
there is an interband absorption at 12 000 cm1 , which or torque, that is proportional to the current. In general,
causes the reflectance of aluminum to fall to only 0.87, the coil is attached to the moving mirror assembly, since
and in the UV the reflectance of oxidized aluminum is less it is lighter than a permanent magnet, and the permanent
than 0.93. This poses a problem in interferometers in which magnet is fixed. In one interferometer design, however, this
there are many reflections, because the beam can become situation is reversed and the permanent magnet is attached
significantly attenuated. As an example, if we take a value to the moving mirror. This interferometer is described in
of 0.93 for the reflectance of a mirror surface, then the more detail later.
reflectivity of the mirror assembly shown in Figure 6(c), in
which there are seven reflections, is only 0.60.
4.4 Interferometer bearings
4.3 Interferometer drives The bearing is one of the most important components in an
interferometer, and the choice of bearing type depends on
Almost all commercially available Fourier transform spec- a large number of factors, including straightness (for inter-
trometers are driven by either a voice coil drive, if the ferometers with linear motion), friction, cost, rigidity, and
interferometer motion is linear, or a galvanometer drive, mass of the moving component. Of these factors, friction
if the interferometer motion is rotary. The exceptions to is one of the most important. The bearing need not be fric-
this are interferometers designed for very high resolution, tionless, but to avoid velocity errors it is important that the
and interferometers designed exclusively for use in the far travel be very smooth. It is also beneficial if the coefficients
infrared. of static and dynamic friction are similar, since this mini-
The natural line width for infrared spectra of low pressure mizes stick-slip motion (stiction) at the mirror turnarounds.
gases can be much less than 0.01 cm1 , and it is there- Finally, the effect of friction on bearing wear must be con-
fore sometimes desirable to make measurements at very sidered. Either bearing wear should be negligible over the
high resolution. From equation (4) it can be seen that to life of the interferometer, or the interferometer should be
achieve resolutions that are better than 0.01 cm1 , an opti- able to compensate for its effects.
cal retardation of more than 100 cm is required. It is not Historically many interferometers used for mid-infrared
possible to achieve retardations this high using voice coil or spectroscopy used an air bearing, which consists of a pis-
galvanometer drives, and an alternative drive system must ton and a cylinder, with a gap of 2 to 5 m between them.
be used. In current commercially available high-resolution Compressed air, typically at a pressure of about 2 atm, is
interferometers the moving mirror is usually driven by a forced into the gap to act as a lubricant. Either rectangular
bearings, or occasionally dual bearings, have been used to have demonstrated resolutions of better than 0.1 cm1 , and
avoid roll about the bearing axis. An air bearing can pro- operation into the visible region of the spectrum.
vide a long, very smooth mirror motion, with exceptional Flexure systems have been used to provide linear trans-
straightness, and negligible static or dynamic friction. Its lation in mid-infrared interferometers. Flexure systems typ-
drawbacks are that it is expensive, requires a supply of dry ically have limited travel, and it is therefore difficult to
air or nitrogen, and has lower stiffness than a contact bear- achieve high resolutions, but they do have several advan-
ing. Because of these limitations, air bearings are no longer tages over bearings that use sliding or rolling surfaces.
in common use in interferometers, but in a research spec- Flexures can provide high lateral stiffness, and exhibit no
trometer, where performance is the primary concern, an air wear. They also have negligible friction, and consequently
bearing is still a good option. Top-of the line research spec- provide very smooth motion with no stiction at the mirror
trometers manufactured by both Bruker Optics and Digilab turnarounds. Currently, Bruker Optics Inc. uses flexures as
use air bearings, and are capable of operation over very part of the hybrid bearing in the IFS 120 HR spectrom-
wide spectral ranges, at high resolution. These interferometer, and On-Line Technologies uses a complex flexure
eters are described in more detail later. system in their interferometer, which is designed for process
Several manufacturers of interferometers have used bear- applications.
ings in which a piston slides inside an outer sleeve. For For rotational motion, interferometers use either ball
obvious reasons, these bearings are often referred to as races, or flex-pivot bearings. Ball races are very cost effec-
syringe bearings. Beckman no longer manufactures Fourier tive, have very high rigidity, and can provide very smooth
transform infrared spectrometers, but in the 1980s they promotion. Their major drawback is that they are subject to
duced an instrument with a ground glass syringe bearing wear, and consequently interferometer performance may
and dynamic alignment. This bearing had a tendency to degrade after prolonged operation. An alternative to ball
bind, however, producing serious velocity errors. Digilab races is a flex-pivot bearing, which provides rotational
has also used syringe bearings in the past, in their FTS- motion with the advantages of flexures that were outlined
7 spectrometers. Currently, Nicolet uses syringe bearings above. Because of their advantages, flex-pivots have been
in all of their mid-infrared spectrometers. The details of used by many manufacturers in different interferometer
the design vary somewhat, depending on the instrument designs. Some of these interferometers are based on rotating
model, but in all cases the bearing consists of a piston with mirror assemblies, but flex-pivots have also been used to
graphite composite rings inside a glass tube. One of these provide translational motion. This can be achieved by using
interferometers is described in detail in Section 5. a porch swing design, such as that used in the Midac inter-
Another example of an interferometer bearing based on ferometer described below. Flex-pivots vary somewhat in
low friction sliding surfaces was designed by Gebbie at the details of their design, but they all operate on the same
the National Physical Laboratory, Teddington, UK.21 This principle as the one shown in Figure 7. They consist of two
interferometer uses a polished glass block of rectangular interlocking cylinders, A and B, which are not in contact
cross-section, which is oriented such that two faces form with each other, but are connected by flexure plates. The
an inverted V. The mirror carriage slides on this glass two cylinders can rotate relative to each other over a lim-
block on PTFE pads. The design of this bearing shares ited deflection angle, with very little center shift. For small
many characteristics of a syringe bearing, but it does have deflection angles, if the bearing is not loaded excessively
some potential advantages. First, gradual wear of the sliding then it can be cycled indefinitely without failure.
surfaces over the life of the interferometer will not result
in increased play in the bearing. In addition, any wear
products will fall out of the bearing, rather than being
trapped inside it, and therefore cannot interfere with the
smoothness of travel. Bearings similar to this are used in
the interferometers in the Digilab Excalibur and the Bruker
IFS 120HR spectrometers, both of which are described in
detail later.
Linear translation bearings for interferometers can also
be based on rolling, rather than sliding surfaces. Roller
bearings consist of either ball bearings or crossed rollers A B
trapped between highly polished hard-steel rails. Thermo Figure 7. Cutaway diagram showing the principle of operation of
Mattson has used these bearings, in combination with a flex-pivot bearing. The arrows indicate rotation of one section
cube-corner retroreflectors, in all their interferometers, and relative to the other.
5 EXAMPLES OF INTERFEROMETER B
H E L C
DESIGN D K
In this section some examples of interferometer design

are given. There are a very large number of variations N S
of the Michelson interferometer that have been designed
over the years for use in infrared spectrometers, and it is
not possible to list them all. Some of these designs have J
G A
been used in commercial spectrometers, but many have F
not. When choosing examples, preference has been given
to interferometer designs that have been commercialized,
because at least to some extent they have been proven.
Figure 8. Schematic diagram showing the construction of the
The specific examples given here have mainly been chosen interferometer used in the Avatar spectrometer manufactured by
because they illustrate certain aspects of interferometer Nicolet. The direction of travel of the light is indicated by the
design. The inclusion, or omission, of a particular design is arrows. Details are given in the text.
not a reflection on the performance of the interferometer,
or its suitability for a particular task. Detailed descriptions over previous syringe bearing designs is that the graphite
of many other interferometer designs can be found in the composite is formulated so that the coefficients of static and
patent literature, and in the references cited therein. dynamic friction are very close, and stick-slip motion is
therefore minimized at the mirror turnarounds.23 A unique
feature of this interferometer design is that the permanent
5.1 The standard Michelson interferometer magnet, H, is attached to the moving mirror, while the voice
coil, I, is fixed. The reason for this is that extra coils, J, can
The top-of-the-line research spectrometers currently made then be used to pivot the moving mirror about a point in the
by Digilab and Bruker can be considered classic interferom- center of the flexure ring. These coils, in conjunction with
eter designs. They are both Michelson interferometers that an expanded HeNe laser beam and a quadrant detector, K,
use air bearings, flat mirrors, and voice coil drives. The thus form a dynamic alignment system that can adjust the
Digilab FTS 6000 also has a dynamic alignment system, tilt of the moving mirror. This is in contrast to all other
with piezoelectric transducers on the fixed mirror, whereas commercially available dynamically aligned spectrometers,
the Bruker IFS 66 and IFS 66v rely on a precision inter- in which the tilt of the fixed mirror is adjusted. The
ferometer to achieve the required performance. The use advantage of dynamically aligning the moving mirror is
of lightweight flat mirrors, along with powerful voice coil that in some cases dynamic alignment of the fixed mirror
drives, means that the interferometers in these instruments can produce a large enough movement of the output beam
have efficient duty cycles, and can scan at very high speeds relative to the detector to impair system stability.23 The
(greater than 10 cm s1 ). In addition, the minimal number additional weight of the permanent magnet does increase
of reflections afforded by the use of flat mirrors, and the the mirror turnaround times, but the duty cycle is not
use of high performance bearings, permits these interferom- seriously affected except at either very low resolutions, or
eters to operate at the very short wavelengths of the UV. very high mirror velocities. Since this spectrometer is aimed
The Digilab FTS 6000 has a specified range of 50 000 cm1 at the routine market, moderate velocities and resolutions
to 10 cm1 , while the Bruker IFS 66v covers the range from are used, and this is therefore not a serious concern. The
5 cm1 to 55 000 cm1 . final novel feature of this interferometer is the inclusion
Nicolet also uses standard Michelson interferometers in of a small, powerful, permanent magnet, L, on the moving
their spectrometers, but they no longer use air bearings. mirror mechanism. It was mentioned earlier that roll about
The interferometer used in the Avatar spectrometer serves the bearing axis is undesirable, and this is prevented by the
as a good example, because it has several novel design small magnet tracking an iron bar that is external to the
concepts.22 Figure 8 shows a schematic diagram of this interferometer. Variations of this magnetic anti-roll system
interferometer. As in any Michelson interferometer, there is are in fact used in all of Nicolets current interferometer
a beamsplitter, A, a fixed mirror, B, and a moving mirror, designs. This magnet not only prevents rotation of the
C. In this design, the moving mirror is attached via a moving mirror assembly, but also lifts and carries most
flexure diaphragm, D, to a piston, E, that is mounted on of the weight. The magnet is located above the center-of-
graphite composite bushings, F. This assembly fits inside gravity of the mirror carriage, so that it does not produce
a glass tube, G, to very close tolerances. An improvement any torque, and carries approximately 85% of the weight.
The bearing utilized by Digilab in its Excalibur interfer- to the direction of the infrared beam, but the use of flat
ometer is similar to a syringe bearing in that it also utilizes mirrors in the interferometer means that this has no effect.
sliding surfaces that are in contact, but have a very low This design, which for obvious reasons is also termed a
coefficient of friction. As with the Nicolet design, the Dig- porch-swing interferometer, was first described by Walker
ilab interferometer is dynamically aligned, but in this case and Rex in 1979.24 Other porch-swing interferometers have
with the alignment mechanism on the fixed mirror. In the been incorporated in commercial instruments, but the Midac
Digilab interferometer the moving mirror is attached to an interferometer is significant because of its very careful
inverted V-shaped carriage, which rides on an optically flat mechanical design.13 One of the major considerations dur-
glass prism with a square cross-section. The glass prism ing the design of this interferometer was insensitivity to
is smooth enough that it does not cause significant veloc- vibration, since it was intended for field measurements. For
ity errors. The smoothness of the surfaces also means that portability reasons, the interferometer also had to be com-
the bearing will exhibit very little wear over the life of the pact, and capable of operating without any utilities other
interferometer, and any such wear is compensated for by than electrical power. To achieve the required insensitivity
the dynamic alignment system. to vibration, the structure of the interferometer was ana-
lyzed for resonant frequencies. All resonant frequencies
below 1 kHz were then removed by redesign. Once the low
5.2 An interferometer based on pure resonant frequencies had been removed from the interfer-
mechanical design ometer, it was possible to increase the bandwidth of the
voice coil drive to control the velocity errors caused by
An example of an interferometer that uses neither dynamic
vibration. The most important aspects of the interferome-
alignment nor optical compensation, but relies purely
ter for vibration control are the designs of the mirror and
on mechanical design, is the one produced by Midac
carriage, the swing arms, and the pivots. The mirror and
Corporation.13 The major components of this interferom-
mirror carriage are machined from a single piece of alu-
eter are shown in Figure 9. The moving mirror and mirror
minum to avoid resonances caused by the mirror mounting.
carriage, A, are suspended from the interferometer block
The mirror surface consists of two concentric areas, which
by two roughly H-shaped swing arms, B. The arms are
are individually polished. The larger, inner, area is used for
attached to the interferometer block and the mirror carriage
the infrared beam, while the outer area is used for the HeNe
at eight pivot points. The arms, mirror carriage, and inter-
laser and the white light interferogram. The outer area is set
ferometer block form the arms of a parallelogram, so that
back 75 m from the inner area so that the centerburst of the
when current is applied to the voice coil drive, C, the mirror
white light interferogram occurs before the centerburst of
swings backwards and forwards. The motion of the mov-
the infrared interferogram. Two identical H-shaped swing
ing mirror also involves a shear component perpendicular
arms are used to suspend the mirror carriage, rather than
four individual arms, because of the increased rigidity. It
was also found that flex-pivot bearings do not provide suf-
(a) ficient rigidity, and ball races were therefore used instead.
Each of the eight pivot points in fact consists of two ball
races that are separated by a sprung washer, which helps
damp any residual vibrations in the swing arms. This design
has proven itself over many years by successfully collecting
spectra while mounted in moving helicopters and tanks!
C B A
(b) 5.3 Double pendulum interferometers
An example of an optically compensated interferometer is

the double pendulum, or wishbone, interferometer25 devel-
oped by Kayser-Threde GmbH (Munich, Germany), and
currently used by Bomem in their MB series spectrom-
eters. The basic layout of this interferometer is shown
Figure 9. Schematic diagram showing the construction of the in Figure 10. Two cube-corner retroreflectors are rigidly
interferometer used in the pectrometer manufactured by Midac.
A plan view is shown in (a) and a side view in (b). The direction attached to an arm that rotates about a pivot point, so that
of travel of the light is indicated by the arrows. Details are given the optical retardation changes in both arms of the interfer-
in the text. ometer during a scan. In contrast to a standard Michelson
Figure 11. Schematic diagram showing the double pendulum

interferometer used in many spectrometers manufactured by
Figure 10. Schematic diagram showing the double pendulum Bruker. The direction of travel of the light is indicated by the
interferometer used in many spectrometers manufactured by arrows.
Bomem. The direction of travel of the light is indicated by the
arrows. the pivot axis. This means that velocity errors caused by
interferometer, because the mirrors in both arms of the external lateral forces, for example vibration, are minimized
interferometer move, the optical retardation is four times because they tend to act equally on both mirrors. In the
the physical retardation. Rotation of the arm about the pivot interferometer manufactured by Bruker, undesired rotation
axis causes shearing of the retroreflectors relative to the about the axis that passes through the pivot point and the
beamsplitter, but the motion of one compensates for the beamsplitter is controlled by using two spatially separated
motion of the other. The interferometer is optically com- flex-pivot bearings.
pensated for shear of the mirror assembly along all three
axes and rotation about the axis parallel to the line con-
necting the two mirrors. These motions will not change 5.4 Fully compensated interferometers
either the interferometer alignment or the optical retarda-
tion. Rotation about an axis parallel to the line connecting The interferometers produced by Perkin-Elmer may be
the pivot and the beamsplitter will cause misalignment of considered some of the classic examples of optical compen-
the interferometer, but this motion can be controlled by use sation. There are many variations, depending on the model,
of a suitably rigid pivot. Several variations of this design but all the interferometers use a parallel mirror pair as the
can be found in the patent by Kayser-Threde.25 moving element. As an example,27 the interferometer from
Bruker Optics Inc. uses a variation of the double pendu- the Spectrum 1 spectrometer is shown in Figure 12. One
lum interferometer,26 shown in Figure 11, in many of their arm of the interferometer contains a flat mirror A, and a
spectrometers. The principal difference between this design parallel mirror pair, B/C. The other arm contains a flat
and the interferometer described above is the inclusion of mirror, D, that can be tilted, using a motor, to align the
folding mirrors in each arm of the interferometer. These interferometer. Note that this interferometer is not dynami-
folding mirrors allow the interferometer to be aligned with- cally aligned: mirror D is adjusted and then remains in the
out moving the beamsplitter or one of the retroreflectors, same position throughout the scan. Mirror E is used only
and make the interferometer more compact. Inclusion of to make the design more compact, and does not affect the
these mirrors also allows a smaller, and therefore lighter, operation of the interferometer. Rotation of the mirror pair
arm to be used to hold the retroreflectors, permitting higher about a pivot point, F, changes the optical pathlength in one
scan speeds. A further advantage is that the center of of the arms of the interferometer. Since the mirrors B and C
mass of the mirror assembly can be made coincident with are parallel, the direction of the beam is unchanged when
A C C
D B
D
E
A
E

interferometer used in process spectrometers manufactured by On-
Line Technologies. The direction of travel of the light is indicated
by the arrows. Details are given in the text.
interferometer used in the Spectrum 1 spectrometer manufactured a cube-corner retroreflector, D, and the other side of the
by Perkin-Elmer. The direction of travel of the light is indicated double-sided flat mirror, C, so that the returning beam also
by the arrows. Details are given in the text. returns along its incident path. The interferometer built by
On-Line Technologies is scanned by moving the retrore-
they are rotated. It is therefore possible to set mirror A flector, D, although it could also be scanned by moving
perpendicular to the beam direction, so that all rays return the double-sided flat mirror, C. The beamsplitter, A, and
along their incident path. It is not too difficult to see that the flat mirror, B, are precision mounted in a casting, E,
translation of the mirror pair along any of three perpen- so that they are parallel. If this component is considered
dicular axes, or undesired rotation about two of the axes, in conjunction with the flat mirror, C, then it can be seen
causes neither a change in the optical path, nor a devia- that tilt or shear of the beamsplitter-mirror assembly has
tion of the beam returning to the beamsplitter. Although no effect on the interferometer alignment. Tilt or shear of
this interferometer has the advantage of being fully com- the retroreflector, D, clearly also has no effect on the opti-
pensated, it does suffer from some limitations if used in cal alignment. Finally, if the double-sided flat mirror, C, is
spectrometers designed for research applications. Unless the tilted, then its effect on the beam in one arm of the inter-
two mirrors in the parallel pair are well separated, and mir- ferometer is compensated for by its effect on the beam in
ror A is very large, it is only suitable for moderate to low the other arm. Shear of this mirror would obviously have
resolution measurements. The moving mirror mechanism no effect. The interferometer is very easy to manufacture,
can also be relatively heavy, especially if the mirror pair is because once the flat mirror, B, and the beamsplitter have
well separated. Nevertheless, for routine and quality control been mounted in the casting, the alignment of all the opti-
applications the interferometer is very stable and scan speed cal components is not critical. Although the interferometer
and resolution are rarely major considerations. This design is fully compensated, velocity errors caused by vibrations
has in fact also been used in Perkin-Elmers Model 1600, are still a potential problem. This is solved by mounting
Paragon 1000, and Spectrum 1000 spectrometers, and has the cube-corner retroreflector on a flexure system, which
been sold in thousands of instruments worldwide. has a center of gravity that is located such that external
On-Line Technologies also use a fully compensated forces do not tend to cause mirror movement. The flexure
interferometer in the spectrometers they manufacture for system is too complex to be described in detail here, but
process monitoring applications.28 Their design, shown in more information can be found in US Patent 5 486 917.29
Figure 13, is interesting because it combines the two opti-
cal elements shown in Figures 6(c) and 6(d) to produce an
interferometer that is both easy to manufacture and very 5.5 Very high resolution interferometers
insensitive to vibration. In one arm of the interferometer,
the incoming beam is reflected from the beamsplitter, A, Interferometers designed for very high resolution measure-
from a flat mirror, B, and from one side of a double-sided ments have some unique design requirements, both for
flat mirror, C, before returning to the beamsplitter via the the bearing and for the interferometer drive system. Cur-
same path. The other arm of the interferometer consists of rently, there are three very high resolution spectrometers
C A
H
C
D C E
F
A
B
E C H
D
B
G
Figure 14. Diagram showing the bearing assembly used in the
interferometer used in the ultra-rapid scanning interferometer
very high resolution 120 HR spectrometers manufactured by
designed by Manning Applied Technologies. The direction of
Bruker. Details are given in the text.
travel of the light is indicated by the arrows. Details are given
in the text.
commercially available: the DA8 from Bomem, and the
IFS 120 M and IFS 120 HR from Bruker Optics Inc. which discussion of the ultra-rapid scanning interferometer devel-
are capable of resolutions of 0.0026 cm1 , 0.0035 cm1 , oped by Manning Applied Technologies.30 This is a spe-
and 0.0012 cm1 , respectively. All of these interferometers cialized interferometer designed for spectroscopic measure-
are based on a standard Michelson design. The Bomem ments of fast transient processes, and has some unusual
interferometer uses flat mirrors and a dynamic alignment features. The most important of these is that, in contrast to
system to maintain the interferometer alignment over the most mid-infrared interferometers, the mirror drive does not
very high optical retardations, whereas both Bruker inter- use reciprocating motion, but instead rotates continuously
ferometers use cube-corner retroreflectors. Voice coil drives at high speed. The layout of this interferometer is shown
clearly cannot be used for such large optical retardations, schematically in Figure 15. One arm of the interferometer
and the moving mirror in the DA8 and IFS 120 M interfer- contains a rotating mirror, A, a cube-corner retroreflector,
ometers is driven by a pretensioned steel wire or cable. The B, and a flat mirror, C. This combination of mirrors ensures
drive used in the IFS 120 HR is more complex, and con- that the beam returns along its incident path, regardless of
sists of a hybrid bearing with a dual drive system, shown the tilt of the rotating mirror, A, relative to the axis of rota-
in Figure 14. The retroreflector is attached to a plate, A, tion. The cube-corner retroreflector, E, in the second arm
that is mounted to a second plate, B, via four flexures, C. of the interferometer, serves only to fold the beam, thereby
A voice coil, D, is mounted on plate A, which sits inside keeping the design compact. If mirror A is tilted relative to
a cylindrical magnet, E, that is mounted to the lower plate, the drive shaft and motor, D, and the two points at which
B. This part of the interferometer drive works in a manner the beam is incident on this mirror are not symmetrically
similar to the porch-swing interferometer described above. disposed about the rotation axis, then the optical retardation
When current is applied to the voice coil, the retroreflector will change during rotation of the mirror. The optical retar-
moves relative to plate B, changing the optical retardation. dation of this interferometer can be varied by changing the
The lower plate, B, is also attached to a main carriage, F,
positions at which the beams are incident on the rotating
which sits on polished steel rails, G. PTFE pads mounted
disk. It should be noted that during one complete rotation
between the carriage and the rails provide low friction slid-
of mirror A, the optical retardation changes through two
ing surfaces. A pretensioned wire, H, is used to move the
complete cycles. Because this interferometer uses continu-
main carriage along the steel rails. The voice coil drive
ous rotation, rather than reciprocating motion, it is capable
is used for low-resolution measurements, and to correct
of very high scan speeds. Time resolutions of 1.5 ms have
for velocity errors in the wire drive during high-resolution
in fact been demonstrated at 8 cm1 resolution.31
measurements.
5.6 An ultra-rapid scanning interferometer 6 SUMMARY
An article on rapid scanning interferometers for mid- This article has dealt only with interferometers that are used
infrared spectrometry would not be complete without a principally for mid-infrared spectroscopy, but even within
this limited context it should be clear that interferometer 7. D.B. Chase, Appl. Spectrosc., 36, 240 (1982).
design is an extensive field that includes aspects of mechan- 8. J.R. Birch and F.J.J. Clarke, Spectrosc. Eur., 7, 16 (1995).
ics, optics, and electronics. Interferometers are also used 9. P. Bouchareine and P. Connes, J. Phys. Radium, 24, 134
for many other applications, and aspects of interferome- (1963).
ters designed for one application are often applicable to 10. J. Chamberlain, Principles of Interferometric Spectroscopy,
those designed for other applications. The interested reader Wiley-Interscience, New York (1979).
can find a vast amount of additional information about the 11. A.E. Martin, Infrared Interferometric Spectrometers, in
design and applications of interferometers in the literature. Vibrational Spectra and Structure, ed. J.R. Durig, Elsevier,
Interferometry is also a field that has advanced enor- Amsterdam, 8994, Vol. 8 (1980).
mously over the years, and continues to do so. In the future 12. T. Hirschfeld, Quantitative FT-IR: A Detailed Look at the
we can expect many advances in all aspects of design, espe- Problems Involved, in Fourier Transform Infrared Spec-
cially electronics. Current literature should therefore always troscopy: Applications to Chemical Systems, eds J.R. Fer-
be consulted if knowledge of the state-of-the-art in inter- raro and L.J. Basile, Academic Press, New York, 193242,
Vol. 2 (1979).
ferometer design is important.
13. US Pat. 4 693 603 (Sep. 15, 1987) (to Midac Corp.).
14. L. Logan, Proc. Opt. Photo-Opt. Instrum. Eng., 191, 110
ACKNOWLEDGMENTS (1979).
15. US Pat. 5 923 422 (Jul. 13, 1999) (to Bruker Analytik GmbH).
The author would like to thank Richard Spragg of The 16. US Pat. 4 345 838 (Aug. 24, 1982).
Perkin-Elmer Corp., Gerald Auth and Terry Cavener of 17. H. Buijs, Proc. Opt. Photo-Opt. Instrum. Eng., 191, 116
Midac Corp., Tony Bonanno of On-Line Technologies, (1979).
Matthew Smith and John Coffin of Thermo Nicolet, Richard 18. H. Buijs, D.J.W. Kendall, G. Vail and J.-N. Berube, Proc.
Crocombe of Digilab, and Chris Manning of Manning Opt. Photo-Opt. Instrum. Eng., 289, 322 (1981).
Applied Technologies for supplying information about their 19. J. Terrien, Rev. Opt., 38, 29 (1959).
past and present interferometer designs. The author is also
20. M.V.R.K. Murty, J. Opt. Soc. Am., 50, 83 (1960).
very grateful to Tony Bonanno of On-Line Technologies,
Arno Simon of Bruker Optik, Chris Manning of Manning 21. S. Johnson, Fourier Transform Infrared. A Constantly Evolv-
ing Technology, Ellis Horwood, New York (1991).
Applied Technologies, and Peter Griffiths of the University
of Idaho for taking the time to review this manuscript, and 22. US Pat. 5 883 712 (Mar. 16, 1999) (to Nicolet Instrument
Corp.).
for their comments and suggestions.
23. US Pat. 5 896 197 (Apr. 20, 1999) (to Nicolet Instrument
Corp.).
ABBREVIATIONS AND ACRONYMS 24. R.P. Walker and J.D. Rex, Proc. Opt. Photo-Opt. Instrum.
Eng., 191, 88 (1979).
HeNe Helium Neon 25. US Pat. 4 383 762 (May 17, 1983) (to Kayser-Threde GmbH).
PTFE Poly(tetrafluoroethylene) 26. US Pat. 5 309 217 (May 3, 1994) (to Bruker Analytische
Messtechnik).
27. US Pat. 4 881 814 (Nov 21, 1989) (to The Perkin-Elmer
REFERENCES Corp.).
28. US Pat. 5 675 412 (Oct. 7, 1997) (to On-Line Technologies,
1. A.A. Michelson and E.W. Morley, Phil. Mag., 24, 449 Inc.).
(1887). 29. US Pat. 5 486 917 (Jan. 23, 1996) (to On-Line Technologies,
2. A.A. Michelson, Phil. Mag., 31, 338 (1891). Inc.).
3. A.A. Michelson, Phil. Mag., 34, 280 (1892). 30. US Pat. 5 898 495 (Apr. 27, 1999).
4. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared 31. C.J. Manning, Smoke and Mirrors: Ultra-rapid-scan FT-IR
Spectrometry, John Wiley & Sons, New York (1986). Spectrometry, in Fourier Transform Spectroscopy; Eleventh
International Conference. AIP Conference Proceedings 430,
5. J.W. Brault, Appl. Opt., 35, 2891 (1996). ed. J.A. de Haseth, American Institute of Physics, Woodbury,
6. D.D. Weis and G.E. Ewing, Anal. Chem., 70, 3175 (1998). 8495 (1998).
Instrumentation for Step-scan FT-IR Modulation
Spectrometry
Christopher J. Manning
Manning Applied Technology, Troy, ID, USA
1 INTRODUCTION was also important to the inception of the FT-IR spectrom-

eter industry.
Michelsons invention in 1880 of a new interferometer1 was Mertz invention2 of rapid-scanning interferometric spec-
made possible by a convergence of commercial technolo- trometry was the foundation for commercial Fourier trans-
gies, of which the most important were optical fabrication form spectrometry. From the time Mertz showed the
and precision machining. In the late 1800s, the United advantage of rapid-scan operation until the present day,
States had only a limited optical industry, which was mainly rapid-scan interferometers have dominated the marketplace.
concerned with the manufacture of cameras and eyeglasses. The experimental literature also has been largely dominated
Long traditions of both scientific research and commercial by rapid-scan interferometry for most of the last 35 years.
production had established a larger European manufactur- There are, however, certain types of measurements that are
ing base for optical components. Perhaps coincidentally, difficult to make with these instruments. Arguably, the most
Michelsons first interferometer was constructed in Berlin.1 important of these are fast, time-resolved measurements of
After he returned to the United States, Michelson relied kinetic processes. These reactions take place in much less
on a series of skilled craftsmen who could fabricate opti- time than it takes a typical interferometer to complete a
cal components, do very precise machine work, as well as single scan. Other measurements involve modulated pro-
assemble and align complex instruments.1 cesses where the period of the modulation is shorter than the
Modern Fourier transform infrared (FT-IR) spectrome- scan time of a rapid-scanning interferometer. For both these
try is also the result of a convergence of technologies. The types of measurements, it is presently advantageous to use
availability of components needed to assemble both rapid- step-scan interferometry. A number of other approaches are
and step-scan interferometers is driven by large, exter- possible and have been demonstrated. Much of the discus-
nal market forces, many orders of magnitude larger than sion that follows applies to these other approaches, which
the market for spectrometers. In particular, the availability include various rapid-scan3 7 and slow-scan8 instruments.
of computers was one key to the inception of the FT-IR Modulation measurements allow observation of any spec-
spectrometer industry, which today sells instruments val- troscopic feature that is varied by an external perturbation,
ued in the hundreds of millions of dollars per year (US$, while ignoring all of the spectral features that are invariant.
at the time of writing). In contrast, the computers them- In complex systems, the selective observation of particu-
selves involve a market well beyond ten billion dollars lar species or submolecular features is especially useful.
per year. The high performance of computers has been, In polymer systems, modulation measurements allow sub-
at times, an important force driving further advances in molecular groups to be observed according to their motion
FT-IR spectrometry. The availability of mass-produced gas during molecular strain. Possibly, the most important sys-
lasers, together with inexpensive and precise machining, tems to be studied by these methods will be biological.
The great complexity of biological systems often renders
John Wiley & Sons Ltd, 2002. their vibrational spectra intractable to analysis. The field
of infrared modulation measurements is only now about WI, USA) introduced the Magna 850 followed in subse-
20 years old. Dramatic improvements in speed and signal- quent years by the Magna-IR 860 and Nexus 870 models.
to-noise ratio (S/N) of measurements have been made in All of these instruments could be operated in either the
the past few years. rapid-scan or the step-scan mode.
Modulation measurements readily divide into two subsets. The current state-of-the-art in both step- and rapid-scan
In the first subset, the optical path difference, or retardation, FT-IR measurements has been virtually defined by digi-
is held as constant as possible, while a given datum tal signal processing. It was the commercial availability
is acquired; this is the approach generally used for fast of computers, analog-to-digital converters (ADCs) and the
time-resolved spectroscopy that is described in the articles CooleyTukey fast Fourier transform (FFT) algorithm16
by Rodig and Siebert (Instrumental Aspects of Time- that made commercial rapid-scan interferometry practi-
resolved Spectra Generated Using Step-scan Interfer- cal around 1970; these are the essential elements of
ometers) and Smith and Palmer (Fast Time-resolved digital signal processing. Computers, ADCs, and numer-
Mid-infrared Spectroscopy Using an Interferometer). In ical algorithms have advanced dramatically in the past
the second subset of step-scan measurements, the aver- 30 years and are still critical to state-of-the-art interfero-
age optical path difference is held constant while the path metric spectroscopy.
difference is modulated about the average value. This lat- Intel coined the term native signal processing for
ter technique is called path-difference modulation,9 phase digital signal processing calculations done on a personal
modulation,10,11 or sometimes, internal modulation, and is computer. Intel were surprised by the rapid emergence of
one focus of this article. A similar division of step-scan components optimized for digital signal processing. (In this
measurement types can be made using the criterion of how article and elsewhere, DSP is used to indicate both dig-
the data are processed and recorded. When the optical path ital signal processing and digital signal processors.) Intels
difference is held constant, data usually are recorded as an response to the emergence of DSP has been to build digital
explicit function of time, i.e. time-resolved. When the signal processing capabilities into their central processing
path difference or sample is modulated, data are usually units. Intel termed the additional capabilities multi-media-
demodulated and recorded as the real and imaginary parts extensions (MMX) because of their utility in DSP of both
of the response to modulation, i.e. phase-resolved. Essen- audio and video signals. The MMX capabilities are quite
tially the same distinctions about data treatment apply when impressive. For example, a state-of-the-art Texas Instru-
using rapid-scan or slow-scan instrumentation for phase- ments dedicated digital signal processor, the TMS320C67,
and sample-modulation measurements. operating at a 150 MHz clock speed is capable of approx-
Fellgetts 1951 thesis12 presented advantages of interfer- imately 1 billion floating point operations per second. The
ometry and inspired both the Connes and Mertz.13 In 1966, computing power of an Intel or Advanced Micro Devices
the Connes reported14 a truly modern step-scan spectrom- (AMD) processor operating at a 1 GHz clock rate is very
eter, which was applied to the study of planetary spectra. nearly the same. One major difference between the two
Since that time, especially in the last 10 years, an increasing is that dedicated DSPs generally have deterministic timing
number of measurements benefiting from step-scan opera- that is more suitable for real-time systems such as inter-
tion have been described. One of the important forces driv- ferometer servocontrol. Both PCs and DSPs are suitable
ing the availability of step-scan systems today is the ubiq- for rapid signal processing where the exact timing of the
uity of rapid-scan FT-IR systems a direct result of their results is not critical. Both types of hardware have been
commercial success. The widespread measurement and use- used for the control and data processing of step-scan FT-IR
fulness of static infrared spectra was partly responsible for spectrometers.
the interest in time-resolved measurements. The ubiquity The cost of computing has decreased by approximately
also created a supply of instruments suitable for modifica- a million-fold over the last 40 years. A typical step-
tion. In 1988, Palmer et al. reported the first modification scan FT-IR modulation measurement now requires about
of a commercial rapid-scanning instrument for step-scan 1 billion floating point operations over a 15 min measure-
operation.15 Shortly before, in 1987, Bruker (Karlsruhe, ment time, almost entirely for demodulation of the detector
Germany) introduced a step-scan option for their IFS 88 FT- signal. In 1960, such calculations would have been almost
IR spectrometer. This instrument has since been succeeded unimaginable; they would have taken 42 min on an IBM
by the IFS 66 and the Equinox 55. In 1990, the Digilab 704 and cost approximately US$ 280. This year (2000) the
division of Bio-Rad (Cambridge, MA, USA) (now Digilab, same calculations done on either a DSP or PC would take
Randolph, MA, USA) introduced the FTS-60A, which has about 1 s and would cost about US$ 0.0001, substantially
been succeeded by the similar FTS-6000. In 1994, Nicolet less than the depreciation of an infrared spectrometer over
Instrument Corporation (now Thermo-Nicolet) (Madison, the 15 min measurement time.
Instrumentation for Step-scan FT-IR Modulation Spectrometry 3
2 INSTRUMENTATION AND
PRINCIPLES OF OPERATION
2.1 Purpose of interferometer; scanning
The purpose of an interferometer in a Fourier transform

spectrometer is to modulate the infrared radiation intensity
(a)
or optical signal. The most common type of interferometer
is Michelsons design, shown schematically in Figure 1.
Interferometric modulation is a spectral multiplexing that
allows the intensity of all wavelengths of radiation to be
measured simultaneously. The intensity pattern resulting
from modulation is called an interferogram. Figure 2(a)
shows the centerburst portion of an interferogram signal,
which is composed of a sum of sinusoidal components.
Each modulation frequency represents a different wave-
length of radiation. Figure 2(b) shows that these sinusoidal
frequency components are superimposed to form the com-
posite interferogram signal. Under the Fourier transform,
sinusoids of different frequency are orthogonal, allowing
the interferogram signal to be decomposed into a spectrum
that represents the intensity for each wavelength as a sin-
(b)
gle datum.
The interferogram generally is modified by interaction of Figure 2. Plots showing summation of sinusoidal components
modulated radiation with a sample and then recorded from to form a composite interferogram signal: (a) composite signal;
(b) component sinusoids.
an infrared detector output. The interferogram that results
from spectral multiplexing can be decoded by Fourier
constant velocity. The motion of the mirror Doppler shifts
transformation to recover the spectrum. In a Michelson
interferometer, modulation is achieved by dividing the the radiation reflecting from it. Both mirrors direct the
radiation into two beams at a beamsplitter, as shown in beams back to the beamsplitter where they are recombined
Figure 1. One beam generally passes to a fixed mirror, to form two new beams.
which may be considered to be the reference arm of Figure 3 illustrates several methods for scanning a
the interferometer. In Figure 1, this is the mirror that is moving mirror. By far the most common approach is
supported by a piezoelectric transducer. The other beam constant-velocity scanning, or rapid-scanning (Figure 3a).
passes to a moving mirror, which may be translated at The optical path difference of the interferometer is varied
linearly in time. When the maximum path difference is
Amplifiers attained, the mirror may be rapidly translated back to the
starting point. In some systems, the mirror may be scanned
Filter sinusoidally to allow it to move rapidly, while reducing
Piezo the rate at which the mirror is accelerated and decelerated.
Sinusoidal scanning is shown in Figure 3(b). Figure 3(c)
Mirrors
Laser
depicts step-scan operation in which the mirror is stopped
Voice coil detector most of the time, and moved quickly from one retardation
drive
value to another. Figure 4(a) shows step-scan operation
where the mirror position is modulated about a constant
average position (see below).
Beamsplitter Because the beam in one arm has been Doppler shifted,
Laser IR source
after recombination both new beams exhibit a time-depen-
dent interference that causes the intensity of radiation of
each wavelength present to be modulated with a unique
IR detector
frequency. If the interferometer is rapid scanning, the mod-
Figure 1. Michelson interferometer. ulation can be conceptualized as the beat frequency between
as shown in Figure 1. For most measurements made with
Displacement
rapid-scanning interferometers, the beat frequencies, often
called Fourier frequencies, fall in or near the audio range.
A reference beam from a heliumneon (HeNe) laser usu-
ally is passed through the interferometer simultaneously
with the infrared radiation. Because the reference laser is
(a) Time
monochromatic and has a precisely known wavelength, its
interference signal provides a very accurate gauge of the
mirror position. The modulation frequency of the refer-
Displacement
ence laser is a very accurate gauge of the mirror velocity

and corresponds exactly to the Doppler frequency shift
generated by the moving mirror. Until quite recently, all
rapid-scan FT-IR spectrometers used the reference laser
interference signal as a sampling trigger for interferogram
recording, as well as an input to the velocity servo. In step-
(b) Time
scan FT-IR spectrometry, the points at which the infrared
signal is sampled are determined by the laser interference
Displacement
signal. Almost invariably, the laser signal also is used as a

gauge for feedback control of the mirror position.
2.2 Difficulty with time-dependent

measurements in rapid-scanning mode
(c) Time
Figure 3. Plots showing optical path difference vs time for If a sample or an optical component changes over time it
different scanning methods: (a) rapid scan; (b) sinusoidal scan; will produce a second modulation of the radiation inten-
(c) step scan. sity that may be confused with the interferometric spec-
tral multiplexing. Under favorable conditions the second
modulation can be cleanly separated from the spectral
multiplexing. Favorable conditions occur if the second
modulation frequency is lower than the frequency sepa-
ration between adjacent spectral elements or greater than
(a) the highest spectral modulation frequency. For example,
photoelastic modulators (PEMs) have been successfully
used to measure vibrational circular dichroism with rapid-
scanning interferometers; see, for example, the article by
Nafie (see Vibrational Circular Dichroism). PEMs gener-
(b)
ally operate at frequencies well beyond 10 kHz. Continuous
scanning slowly enough to keep the infrared modulation
frequencies below the polarization modulation frequency
is trivial. The most difficult circumstances occur when
the second modulation produces components that overlap
the Fourier frequencies. The multiplication of the dynamic
(c) sample absorption into the interferogram produces convo-
Figure 4. (a) Optical path difference vs time during step-scan lution of the modulations in the spectral domain. If the
with phase modulation. (b) Interference intensity vs optical path second modulation includes a range of unknown frequen-
difference. (c) Detector signal resulting from phase modulation of cies, deconvolution may not even be possible.
(a) applied to (b).
the Doppler-shifted beam and the unshifted beam. In a 2.3 Purpose of step-scan operation
two-beam interferometer one of the recombined beams is
directed back to the source and lost. The other recom- The primary purpose of step-scan operation is to remove the
bined beam generally passes to the sample and detector, time dependence of the interferometric spectral multiplexing
so that measurements of time-dependent systems can be produced by rapid scanning is virtually independent of
made conveniently. A variety of other modulations then may wavelength.
be applied without interference from the Fourier frequen- One advantage of phase modulation is that it is very
cies. By stopping the moving interferometer mirror while useful in rejecting multiplicative noise.18 There are other
each data point is measured, the Doppler shift drops to advantages specific to the type of measurement. For exam-
zero, and consequently the Fourier frequencies drop to zero. ple, in photoacoustic measurements, phase modulation pro-
There is still interferometric modulation, but it appears as vides a tool for a discriminating the depth of absorbers
a static phase difference between the components of the within a solid sample.19 The same depth discrimination
recombining beams. The static phase difference is a func- also can be achieved by amplitude modulation, for example
tion of interferometer mirror position. The interferometer by a rotating chopper.20 However, the effect of multiplica-
mirror position can be varied over a desired range by step- tive noise is significantly greater when using amplitude
ping through a series of positions, as shown in Figure 3(c). modulation than phase modulation,18 and almost always
Because stepping is never infinitely slow, there is still a substantially more than predicted by theory10 that considers
time dependence to the overall spectral multiplexing. How- only additive noise.
ever, the effective frequencies (or virtual frequencies17 ) Time-dependent measurements can be divided into two
in step-scan operation are generally shifted at least one categories. The first category is explicitly time-resolved
order of magnitude lower, and sometimes as much as three measurements as discussed in this Handbook by Rodig and
or four orders of magnitude lower, relative to conven- Siebert (Instrumental Aspects of Time-resolved Spec-
tional rapid-scan measurements. The effective modulation tra Generated Using Step-scan Interferometers) and
frequencies are proportional to the overall mirror scan or Smith and Palmer (Fast Time-resolved Mid-infrared
stepping rate. The static spectral multiplexing while the Spectroscopy Using an Interferometer). Time-resolved
mirror is stopped allows a wide range of time-dependent measurements usually employ a pulse excitation of the
measurements to be made conveniently, by providing clean sample. The second type of measurement employs phase
separation of the sample modulation from the spectral mul- modulation or sample modulation and has been called
tiplexing modulation. phase-resolved21 to distinguish it from time-resolved
In most step-scan modulation measurements, the mir- measurements. These modulations generally involve sinu-
ror is not held at a fixed position. Rather, the position soidal variation of the infrared radiation intensity. Sample
of one or both of the interferometer mirrors (and conse- modulation is generated by varying a physical parameter,
quently, the optical path difference) is modulated to produce such as electrochemical potential, strain or pH, that affects
a cyclic variation of the intensity of the interferometer out- the infrared absorption of the sample. In time-resolved
put. Figure 4(a) shows how the optical path varies as a measurements, a series of intensity values are recorded at
function of time. Phase modulation provides a constant discrete times after the excitation; the transient event and
modulation frequency for all wavelengths passing through intensity measurements are repeated at each of the retarda-
the interferometer. It is usually produced by sinusoidally tion steps. The complete data set is a series of discretely
modulating the position (sometimes called dithering) of sampled interferograms for each time after the pulse exci-
one or the other interferometer mirrors, such that the inter- tation. Usually, the time dependence being measured is
ference intensity is modulated according to the slope of the related to a sample that changes over time, but in some
interferogram over the range of mirror motion. Figure 4(b) cases, such as photoacoustic measurements, a time depen-
is a portion of the interferogram signal illustrating the inter- dence also arises from the detection method.
ference intensity as a function of optical path difference. The signals originating from phase modulation, sample
The result of interaction between the phase modulation modulation and PEMs or other optical modulators generally
and the interferogram is shown in Figure 4(c). This plot interact in a multiplicative fashion. Multiplication of sinu-
is intensity as a function of time, which is the detector soidal signals, or sinusoidal components of signals, result
signal resulting from phase modulation. For small phase in sum and difference frequencies. Consequently, just three
modulation amplitudes, the interferogram slope is essen- sinusoidal modulations and their interactions can produce
tially constant, and the amplitude of the modulated signal 13 distinct frequency components. The sum and differ-
is approximately equal to the first derivative of the inter- ence frequencies of two interacting modulations are called
ferogram. The constant modulation frequency is in contrast sidebands. Figure 5 illustrates the case of a low frequency
to rapid-scan operation, where each wavelength is modu- sample modulation interacting with a phase modulation and
lated with a unique Fourier frequency. It should be noted a polarization modulation generated by a PEM. Generally,
that the depth of modulation produced by phase modulation sample modulations occur at frequencies on the order of
varies with wavelength,10 while the depth of modulation tens of hertz, while phase modulations are often in the
f PEM f PM f S f PEM f PM + f S f PEM + f PM + f S
f PEM + f PM f S
f PM f s f PM + f s f PM f s f PEM + f s
fS f PM Break f PEM f PM f PEM f PEM + f PM

Frequency
Figure 5. Hypothetical frequency spectrum of triple-modulation experiment: fs , sample modulation frequency; fPM , phase modulation
frequency; fPEM , photoelastic modulation frequency.
range of hundreds of hertz. PEMs typically operate at 37 Thermo-Nicolet Corporation, as well as modifications of
or 74 kHz. Thus, the lowest frequency signal appearing in commercial systems22 and custom-built systems.14,23
Figure 5 is indicated as a sample modulation, fs . The phase
modulation frequency, fPM , with its sidebands at fPM fs
and fPM C fs , is next highest in frequency. A break in the 2.4 Demodulation
axis is indicated to account for the fact that the next cluster In both rapid-scan and step-scan operation, demodulation
of modulation frequencies is in the range of tens of kilo- operations are tailored to the modulations that are present
hertz. The cluster of nine modulation and sideband frequen- in the signal; various types of spectral information may be
cies, to higher frequency beyond the break in the axis, corre- recovered, depending on the types of modulation employed.
spond to the PEM frequency together with sidebands. Two In rapid-scan operation, where the sample is static and the
of the frequencies are the sidebands generated only by the sampling of the infrared signal is triggered by the laser,
interaction of the sample modulation with the polarization the appropriate demodulation operation is simply Fourier
modulation, at fPEM fs and fPEM C fs . The sidebands transformation of the interferogram signal, usually followed
generated by the phase modulation are at fPEM fPM and by phase correction.24 Certain step-scan measurements also
fPEM C fPM . There are also sidebands of the sidebands at can be demodulated by a single Fourier transform. In
fPEM fPM fs , fPEM fPM C fs , fPEM C fPM fs , particular, time-resolved measurements described in this
and fPEM C fPM fs . In a typical step-scan measurement, Handbook (see Instrumental Aspects of Time-resolved
these modulation signals would be present at each step of Spectra Generated Using Step-scan Interferometers and
the interferometer. Fast Time-resolved Mid-infrared Spectroscopy Using an
The intensities of the signals at each frequency as a Interferometer) fall into this category. Focal plane array
function of optical retardation is governed by the spectral imaging, as described in this Handbook (see Instrumenta-
information associated with each signal. Figure 5 ignores tion for FT-IR Imaging), and elsewhere,25 27 is another
the issue of each frequency component having a phase example of a step-scan measurement for which this simple
associated with it. In fact, there is a real and imagi- demodulation is appropriate. In these two examples, either
nary frequency spectrum; the intensities of the real and an image or a transient event is recorded at each step of
imaginary portions indicate the phase for each of the an interferometer mirror. The resulting set of intensity val-
signals. The interactions of the phases of the modula- ues over retardation for any single pixel or any single time
tions is quite intricate, but must be resolved properly delay is equivalent to an ordinary interferogram and may
to obtain accurate spectral data. A full description of be demodulated with a single Fourier transformation.
the process is beyond the scope of this article. In brief, In the case of phase modulation or sample modulation
proper resolution of the phase begins with knowledge measurements, the demodulation process is invariably more
of the relative phases of the modulation signals and the complicated than a single Fourier transformation and will
ADC sampling. The data from the ADC for each step is be treated more fully after the hardware and types of
Fourier transformed, and the phases and amplitudes are then modulations are described. In most step-scan modulation
recovered. measurements using DSP, after a preliminary demodulation
A variety of interferometer types have been operated in step, the real and imaginary intensity values resulting from
step-scan modes. These include the commercial step-scan each modulation can be Fourier transformed as if they were
interferometers available from Digilab, Bruker Optics and a static interferogram described above.
2.5 Mirror drive/support and creep, but this can be readily compensated by the
servoloop.
In every successful step-scan interferometer design to date, At least two custom-built step-scan interferometers30,31
the mirror position is measured by the use of a refer- have used a piezoelectric drive element on the shaft of
ence laser and the resulting signals are used to provide a micrometer driven by a stepper motor. The micrometer
feedback control of the mirror position. A variety of inter- provides slow, coarse positioning, while the piezoelectric
ferometer designs, mirror supports, drive mechanisms and element provides rapid, fine positioning. Typically stepper
servoloops have been employed in step-scan measurements. motors have a 200-step per turn ratio, but a microstepping
As of this writing, all commercial step-scan interferometers controller may be used to divide each step into arbitrarily
employ a voice coil drive (or the very similar galvanometer small increments. Stepper motor operation should not be
drive) of the type used in construction and operation of confused with stepping of an interferometer mirror, though
conventional rapid-scan FT-IR spectrometers. Some mirror the first can drive the second. Similarly small increments
carriages are suspended on air bearings (Bruker, Digi- of motion may also be obtained by the use of a direct cur-
lab) that offer excellent straightness and smoothness of rent motor with encoder and gearbox driving a micrometer
travel. Graphite-on-glass bearings are also used (Thermo- or leadscrew. For microstepping, the principle of operation
Nicolet). A ball-bearing slide has also been used in a is that the currents in the stepper motor coils are changed
moving cube corner interferometer (Oriel, Stratford, CT, incrementally to move the rotor infinitesimally. In the nor-
USA), which has been converted for step-scan operation mal operation of a stepper motor, the currents are simply
(Manning Applied Technology, Troy, ID, USA). Voice coil switched on and off to produce full steps. While the full
drives generally have a large free range of travel and allow resolution with which the current can be controlled is not
the interferometer to be seamlessly switched between step- realized in the motion, in fact, a microstepping drive allows
scan and rapid-scan operation. They are inexpensive, as mirror positioning with resolution of a few tens of nanome-
well as simple. The main disadvantage of voice coil drive ters or better. Likewise, a direct current motor with an
mechanisms is that they are relatively slow and generate encoder and gearbox can easily and repeatably realize very
only modest forces.28 small displacements. With a piezoelectric element to allow
For several reasons, some commercial step-scan interfer- fine and rapid position control to subnanometer resolution,
ometers (Digilab, Thermo-Nicolet) use a piezoelectric drive these step-scan interferometers are perhaps better in perfor-
mechanism on the fixed mirror.29 The second drive can mance than voice coil systems. It is possible to operate such
allow more rapid correction of retardation errors, as well systems in a rapid-scan mode, but the friction forces in the
as dynamic alignment. To be useful for dynamic alignment, screw are relatively large. Thus, they are not well-suited
the piezoelectric drive system must have multiple elements. for rapid-scan operation, which is a significant detriment to
The maximum displacement of piezoelectric drivers is usu- their market potential.
ally quite limited, on the order of micrometers, unless the
elements are quite large. The force required to drive a mir-
ror for phase modulation scales linearly in amplitude and 2.6 Mirror control
as the square of the frequency. The power scales as the
cube of the frequency. A voice coil drive is capable of In all step-scan FT-IR systems, the signal from the laser
very large phase modulation amplitudes, over thousands of must be resolved into a feedback signal used to govern the
fringes of the reference laser, at low frequencies. The ampli- mirror position. There are differences of opinion regarding
tude capability falls as the square of the frequency until the the relative merits of various approaches. One view holds
voice coil inductance becomes significant, beyond which that the laser intensity and laser detector photodiodes are
point the amplitude response decreases as the cube of the subject to drift and that the servo should be alternating
frequency.28 current-coupled to negate the drifts. The opposing opinion
Piezoelectric transducers face analogous limitations that is that the intensity of the laser and the drift of the pho-
stem from their capacitive nature. Even piezostacks require todiodes can be tracked closely. Further, the drift of the
100 or more volts for actuation, while some elements photodiodes is quite small and slow. In some sense, both
require 1000 or more volts. The piezoelement capaci- views are correct, and the best choice of servo system may
tance must be charged and discharged very rapidly to depend on the application. To date, all of the alternating
effect phase modulation. Thus, a relatively high-current, current-coupled servo systems have relied on modulation
high-voltage amplifier is required to drive a piezoelement, of the mirror position.22,29,32 The main disadvantage of this
while a relatively low-voltage amplifier is required to drive mirror motion is that it complicates time-resolved measure-
a voice coil. Piezoelectric transducers exhibit hysteresis ments. For modulation measurements, which are the focus
of this article, the additional modulation caused by the ser- All of these alternating current-coupled approaches rely on
vocontrol mirror motion is removed easily by filtering. modulation to modify the symmetry of the interferometric
The alternating current- and direct current-coupled ap- system.
proaches lead to different algorithms for conversion of It should be noted that there is some confusion in the ser-
the laser interference intensity signals to mirror position, voloop positioning performance specification in step-scan
thence to position error and feedback correction. In the case interferometers. Some manufacturers specify root mean
of alternating current-coupled systems, the fundamental square (RMS) position error, others specify the maximum
principle is to demodulate the laser signal and to use deviation in physical displacement of the mirror, while
the intensities of the various odd and even harmonics as others specify the displacement of the mirror in optical
feedback.32 If the mirror position is modulated about a path difference. It would be very helpful to establish a
point of even symmetry, i.e. an extremum of the laser standard. Further, the description should include the fre-
interference pattern, then the resulting signal at the detectors quency distribution of the errors. Often the errors are
will contain components of purely even symmetry; these are induced by environmental disturbances, specifically vibra-
the even harmonics. Any position error, i.e. deviation, from tion. Because of this, step-scan spectrometers often are
the point of even symmetry will appear as a component placed on vibration isolation tables to decouple ambient
of odd symmetry, which may be filtered and applied as noise. Ideally, spectrometers would be performance rated
feedback to correct the position to the point of exactly even for different environmental vibration levels.
symmetry. If the mirror position is modulated about a point Another serious problem with the specification of posi-
of odd symmetry, i.e. a zero-crossing of the laser, then tioning performance, which is rarely addressed, is that the
the resulting signal at the detectors will contain components reference laser is invariably used to measure the perfor-
of purely odd symmetry; these are the odd harmonics. mance of the servoloop of which it is an integral part. Any
The fundamental principle for direct current-coupled servo errors that affect the reference laser signal, such as inten-
systems is that the modulated intensity of the laser signal at sity fluctuations or photodiode drift, will not appear in the
any point in the sinusoidal waveform can be converted via position measurement, even when they produce real posi-
an arc sine or arc cosine function into mirror position. The tion errors. A more grievous oversight is that temperature
conversion is very accurate if the laser intensity is known variations of air or purge gas in the arms of the interfero-
to a high degree of accuracy. Thus, as in the case of the meter can easily produce much larger path difference errors
alternating current-coupled systems, the phase angle of the than the nominal positioning accuracy.18 These errors are
laser interference pattern can be resolved very accurately hidden in position stability measurements based on the ref-
within any one fringe. By fringe counting, it is possible to erence laser signals. One method to accurately estimate the
maintain nanometer-accuracy positioning over centimeters effect of temperature variations is to compare the fluctua-
of mirror travel. tions of the infrared signal to the fluctuations of the laser
Often, a quadrature detection system is used to resolve signal while the mirror is stopped at a particular optical
the ambiguity that arises in a system with a single laser path difference. The local slope of the infrared interference
detector. Because the mirror motion generates symmetrical record can be used to convert the magnitude of fluctuations
sinusoidal signals, it is not possible to discern the direction to position error. A more definitive method of resolving
of motion from a single laser detector. There are several these issues would be to use a separate laser, which is both
methods that may be used to resolve the ambiguity. By frequency- and intensity-stabilized, as a reference to mea-
far the most common is the use of a second laser detec- sure the servo performance. Better yet would be the use
tor sampling the second output of the interferometer. If the of an array detector with the stabilized laser illuminating
interferometer is symmetrical, this second laser output is the entire interferometer aperture such that the variation of
180 out of phase with the first. To alter the symmetry of the path difference across the aperture may be discerned. Sim-
interferometer and make the second output useful, a quarter- ilar information can be derived from a broad band infrared
wave plate is inserted in one arm. Together with a polarized source viewed near the centerburst with an array detector.
laser and a polarization filter in front of one of the detectors, In lieu of careful measurements, servo performance speci-
the second laser output can be adjusted to 90 out of phase fications and measurements should be viewed skeptically.
with the first, thereby offering direction information. One The modulation produced by an interferometer is a pro-
method described below is the use of a solid-state reference foundly accurate sinusoidal function of retardation. If the
laser, which can be frequency modulated. Analogously, incident intensity of a reference laser is constant, or known,
the moving mirror position may be dithered with a phase then the modulated intensity can be resolved to a very accu-
modulation to produce signals that reflect the absolute mir- rate determination of mirror position within any one fringe
ror position within a fringe, even in rapid-scan operation. via the inverse sine or inverse cosine functions. Again,
fringe counting is necessary for tracking the laser position In principle, the step-scan technique can be employed to
over more than one fringe. Some method must be used to observe the spectroscopic response of a system to any per-
resolve the ambiguity that each intensity value, except the turbation, so long as the effect is reproducible. If phase
extrema, corresponds to two positions within each cycle. modulation is used together with sample modulation, an
There are several other problems in determining the mir- extra step may be required in demodulation.35 Sample mod-
ror position in a direct current-coupled system. Perhaps the ulation can also be used without phase modulation, in which
best approach is to constantly monitor the intensity of the case demodulation requires only the same extra step as if
laser output with a small beamsplitter that taps a portion phase modulation were present.
of the laser output. These data may be used to normalize Usually, phase modulation is at a higher frequency than
the intensity of the signals observed during interferometric the sample modulation and acts as a carrier for the sample
modulation before they are used to calculate the mirror response signal. Many sample modulations are sharply
position. Other approaches are possible, such as calcu- limited in frequency. For example, the potential modulation
lating the incident intensity of the laser from the signals of an electrochemical interface is limited by the electrolyte
observed at two quadrature detectors described above. The resistance and the double layer capacitance. For polymer
laser intensity drifts relatively slowly after warm-up, so the stretching measurements, the required mechanical motion
updating need not be instantaneous. of the sample limits modulation frequencies to a few
An alternative alternating current-coupled servo could hundred hertz or lower, unless robust equipment is used.
exploit wavelength tuning of lasers. In particular, semicon- The overall demodulation scheme must be tailored to
ductor diode laser wavelengths usually shift when the drive the modulations present. In the case of a simple phase
current is changed. In general, diode lasers have a frequency modulation, the demodulation can be done conveniently
stability on the order of parts per thousand, while the HeNe in two steps. In the first step, the response of the sample
reference lasers invariably (as of this writing) found in com- or detector to the phase modulation frequency is recov-
mercial FT-IR spectrometers have a frequency stability on ered, and in the second step the spectrum is recovered. For
the order of parts per million. A variety of methods are many years, demodulation of step-scan interferometric sig-
available for stabilizing the wavelength of diode lasers, nals was accomplished by the use of lock-in amplifiers.
but have not yet appeared in FT-IR spectrometers. For A lock-in amplifier is conceptually quite simple,36 as indi-
high-resolution measurements, stabilized HeNe lasers with cated in Figure 6. An input waveform is multiplied by a
frequency drift on the order of parts per billion are read- reference sinusoid having the same frequency as the signal
ily available.33 To date, alternating current-coupled servo of interest, then low-pass filtered. The schematic diagram
mechanisms have relied on mirror motion to produce an in Figure 6 indicates a single-channel lock-in amplifier. If
alternating current signal for mirror position determination. two channels of this type are used with reference sinu-
It is possible to produce very similar alternating current
soids 90 out of phase, both the in-phase and quadrature
signals by frequency modulation of the reference laser. A
components of the signal of interest can be recorded. The
variety of methods, including the use of a Zeeman-split
consequence of multiplication is that sum and difference
laser,34 elasto-optic, electro-optic and acousto-optic modu-
frequencies are generated. The sum frequency is twice the
lators, can produce phase or frequency modulations suitable
reference and signal frequency, while the difference fre-
for accurately determining the mirror position in an alter-
quency is zero. Thus, the lock-in amplifier heterodyne shifts
nating current-coupled manner without any mirror motion.
the phase modulation frequency to zero, thereby convert-
ing the appropriate filtering from bandpass to lowpass. The
2.7 Modulation and demodulation lowpass filtering step will reject the higher frequency sum
signal, while passing the zero-frequency or direct current
A variety of modulation methods have been used in step- component. The lowpass filter can be set to an arbitrar-
scan FT-IR spectrometry. Perhaps the most prevelant has ily small passband at the expense of filter settling time.
been photoacoustic spectroscopy, where the infrared beam
is modulated by the phase modulation itself. This technique Gain Multiplier
Signal in RC integrator V out
is covered in more detail later in this article. Another type
of modulation may occur when the physical parameters of
a sample are perturbed. Three classic examples are electro- Filter
Ref in
chemical potential modulation (see Infrared Spectroelec-
trochemistry), strain modulation (see Dynamic Infrared
Linear Dichroism Spectroscopy) and electric field modu-
lation (see Vibrational Spectroscopy of Liquid Crystals). Figure 6. Schematic diagram of lock-in amplifier.
However, any frequencies close to the original signal, somewhat greater, but all of the frequency elements present
especially those caused by low-frequency multiplicative in the signal can be observed and recorded. One great
noise, are also shifted to near zero frequency and can advantage of this approach is that multiple modulations can
be very resistant to discrimination by lowpass filtering. be simultaneously recovered without the expense of addi-
While lock-in amplifiers are quite simple conceptually, they tional lock-in amplifiers.35 No filtering step is required to
usually have many useful controls that are quite compli- reject the 2f frequency component. A further advantage
cated. Controls often include input gain stages that can of DSP is that the very large dynamic ranges of broad-
be used to amplify weak signals, and a range of output band interferometric signals can readily be accommodated,
options such as filter bandwidth and frequency response, while preserving the very small signals resulting from sam-
which can be used to optimize the response for each ple modulation.
application. The use of DSP demodulation to recover the phase mod-
Analog lock-in amplifiers generally do not have sufficient ulation signal is quite straightforward. The detector signal
dynamic range to process a step-scan interferogram without should be digitized by an ADC of sufficient resolution to
compromising the S/N.31 The dynamic range of a rapid- represent all of the information present. This requirement
scan interferogram can exceed 1.5 million to 1, so that a true has become much easier to achieve in recent years as the
21-bit ADC would be required to preserve all of the spectral performance of ADCs has improved dramatically. The tim-
information through digitization. The dynamic range of a ing relationship between the sampling and the modulations
step-scan interferogram can be substantially larger than that present also must be accurately measured. This condition
of a rapid-scan interferogram because of the decreased can be difficult to achieve, particularly in the case of a PEM
mirror velocity. As the mirror velocity is decreased, the or other devices that must be controlled externally to the
bandwidth of the infrared spectral multiplexing is decreased spectrometer and ADC sampling.37 In the case of phase
proportionately. For every factor of p two decrease in the modulation, which is controlled internally by the spectro-
modulation frequency bandwidth, 2 times less detector meter, the condition can be easily met. The clock signal
white noise falls into the range of Fourier frequencies. used to drive ADC sampling can be derived from the same
It should be noted that detector noise is generally not quartz clock used to drive the phase modulation control.
the limiting noise source in step-scan measurements; this There are four main reasons why custom DSP soft-
analysis overlooks the effect of low frequency noise sources ware and hardware within the spectrometer have displaced
mentioned above. Nevertheless, the dynamic range of a the use of external lock-in amplifiers. The first reason is
step-scan interferogram can require more than 24 bits of the dynamic range issue raised above, which has been
ADC range, particularly to resolve the subtle responses largely answered by the emergence of DSP-based lock-
to sample modulations such as polymer stretching. DSP- in amplifiers. The second reason is that, to the end-user,
based lock-in amplifiers are now available with excellent the complexity of experiments is greatly reduced by build-
dynamic range. ing the signal processing hardware and software into the
More recently, lock-in amplifiers have been largely dis- spectrometer. For example, a step-scan dynamic dichro-
placed by custom DSP software and hardware built into ism measurement made with the use of a PEM requires
commercial step-scan spectrometers. These DSP capabili- at least three lock-in amplifiers, and preferably five. Ignor-
ties have been built into step-scan spectrometers to avoid ing for the moment the expense, each lock-in amplifier
the need for lock-in amplifiers. Digital demodulation can requires four cables a signal input, a reference input, and
proceed by either of two related routes. In a sense, the two outputs. A rats nest of 20 cables can be taxing to
product and summation steps, used in lock-in amplifiers as even the most patient experimentalist. Further, the input
described above, are an embodiment of the Fourier inte- and output stages of each lock-in amplifier must be cor-
gral. In the same sense, a lock-in amplifier produces the rectly adjusted for each measurement. In contrast, the DSP
real and imaginary parts of a single-point Fourier trans- approach requires only that the single ADC digitizing the
form of the signal of interest. The real and imaginary points detector signal have a correctly scaled input. A third rea-
are available for only the frequency of interest. Except for son for the displacement is that the cost of one additional
the narrow region of the frequency spectrum that is passed DSP chip in the spectrometer is much less than a complete
by the filtering step, the rest of the frequency spectrum is DSP-based lock-in amplifier, and in fact, can replace five
ignored by the lock-in amplifier. The functionality of a lock- or more dedicated lock-in amplifiers. The fourth reason is
in amplifier can be directly implemented in DSP algorithms. that by building the DSP hardware and algorithms into the
Another DSP method of producing the real and imaginary spectrometer, the fundamental tradeoff between filter time-
frequency components of a signal of interest is the Fourier constant and performance, which is detrimental to step-scan
transform itself. In this case, the computational burden is measurements,17 is avoided.
If lock-in amplifiers are used, it is helpful for the phase of layer thicknesses and absorption spectra is a formidable
modulation to be substantially higher in frequency than the mathematical problem.43 In many cases the problem can be
sample modulation so that two can be readily discrimi- simplified, such as when there is a thin layer of absorbing
nated by output filters. Frequency separation requirements material on top of a thick layer of weakly absorbing
are greatly eased by the use of DSP because the Fourier material.19
transform and some other digital signal processing algo- The next level of complexity is the use of multiple, simul-
rithms provide more precise frequency discrimination than taneous sinusoidal modulations. Because the modulations of
analog filters. The discrimination problem arises because different frequency are orthogonal under Fourier transfor-
the lock-in amplifier generates a component at twice the mation, or can be readily separated by lock-in amplifiers, a
phase modulation frequency that must be removed by the large number of phase modulations can be used simultane-
output filter. The output filter must allow the component ously. If lock-in amplifiers are used, there must be one for
due to the sample modulation to pass. If the 2f component each modulation frequency. Early reports31,35 used sinu-
is too low in frequency, then the filter settings appropriate soidal modulations, but the harmonics of square wave or
for removing it will reduce the amplitude of the sample other modulations can also be used for this purpose. The
modulation signal; some phase shift is unavoidable, but response to stepping can be conveniently extracted and
becomes larger as the filter time constant increases. Thus, removed,44 thereby simplifying spectral interpretation.
experiments performed with lock-in amplifiers are further The optical modulation applied to photoacoustic samples
complicated by constraints on modulation frequencies and can also be a pulse containing many frequency components.
relationships between modulation frequencies. Such pulses can be generated either by sudden stepping
It is possible to recover unique and useful spectral infor- of the mirror, or by the use of a shutter.45 Pulse-response
mation from additional modulations. The best known exam- data can be recorded as time-resolved measurements, or
ple of a third modulation is polarization, usually imposed the resulting signals can be demodulated to recover the
by the use of a PEM.38,39 An interesting alternative to the frequency response of the sample to each component of
use of a PEM is a rotating polarizer which achieves a sim- the excitation pulse.
ilar purpose more slowly.40 One elegant measurement used
double-modulation to simultaneously measure the back-
ground spectrum and the velocity-modulated spectrum of 3.2 Polymer stretching
ArHC .41,42
The polymer stretching measurements pioneered by Noda
et al.46 have attracted considerable attention from the poly-
3 APPLICATIONS AND BRIEF mer physics and spectroscopy communities and are also the
LITERATURE REVIEW subject of another article of the Handbook (see Dynamic
Infrared Linear Dichroism Spectroscopy). The premise
Applications of step-scan FT-IR spectrometers have multi- of these measurements is that during stretching of a polymer
plied as the commercially available systems have become sample, the molecular reorientation rates within the polymer
less expensive and easier to use. can be observed spectroscopically. Molecular reorientation
is the microscopic origin that gives rise to and governs the
macroscopic rheological properties of materials. Because
3.1 Photoacoustic spectroscopy infrared spectroscopic measurements are sensitive to molec-
ular orientation, stretching measurements offer a powerful
Currently, the most important industrial application of step- tool for probing molecular motions.
scan FT-IR spectrometry is photoacoustic measurements. Most polymer stretching measurements employ a sin-
Such measurements are particularly useful because layers gle sinusoidal modulation component, together with phase
on opaque substrates can be readily measured with minimal modulation.47 It is also possible to make the measurement
sample preparation. Two articles in this Handbook (see with multiple frequency components,35 or many frequency
Depth Profiling by Microspectroscopy and Photoacoustic components.48 In one report, a time-resolved measurement
Spectroscopy) describe these systems in some detail. was made using pulse excitation.49 The response of the
The simplest step-scan photoacoustic measurements use sample to the pulse excitation could have been demodulated
a single phase-modulation frequency to excite the sample. to recover the response of the sample to each frequency in
The time dependence of heat flow from various regions the excitation. If the phases of the many frequency com-
within the sample appears as a wavenumber-dependent ponents are aligned, then the sample is excited by a pulse.
phase shift. Deconvolving the sample response into a series The amplitude of the pulse is limited by the linear range of
the polymer sample, and by the transducer used to provide 4 LIKELY FUTURE INNOVATIONS
excitation. If the many frequency components have random
phase, the dynamic range of the sample and transducer may 4.1 Correction of position and modulation errors
be used more efficiently.48
The data acquired in dynamic infrared stretching mea- In recent years, a new approach for recording and
surements are often presented using Nodas methods of signal processing of infrared interferometric data has
two-dimensional correlation50 and generalized correlation.51 appeared.7,67 69 The new approach starts by recording both
The premise of the two-dimensional presentation is that the interference records of the reference laser and the
it reveals correlation between the motions of submolec- infrared signal. In conventional rapid-scan spectrometers
ular groups in the sample. These efforts have also been the laser interference pattern is used to trigger data sam-
expanded to pressure modulation,52 with limited success, pling from the infrared channel, so that only one set of data
and to shear modulation. is recorded per scan.70 In conventional spectrometers, once
the infrared data have been recorded, any errors of sample
timing or variations of velocity are permanently embedded.
3.3 Electrochemical modulation The new approach opens the possibility of powerful post-
processing steps that remove the effects of velocity error in
The use of electrochemical modulation for infrared spectro-
rapid-scan measurements. Various permutations of the new
scopic measurements was pioneered by Bewick and Pons
approach have appeared in the literature.
using a very slow-scanning monochromator. In the spectro-
A key processing step can be the convolution of inverse
electrochemical cell, a thin layer (5 m) of the electrolyte
transfer functions with the data sets.71 The transfer func-
was located between a platinum working electrode and a
tions of the infrared and laser signal channels, including
calcium fluoride window. The potential applied to the work-
detectors, preamplifiers, and even ADCs, are measured and
ing electrode was modulated and the detector signal was
inverted. The purpose of convolving the inverse transfer
demodulated with a lock-in amplifier. It took well over
functions is to remove the effects of delay and ampli-
a decade before the analogous experiment was performed
tude response in the signal channels, thereby restoring the
successfully using a step-scan interferometer by Budevska,
recorded data to a state corresponding, as closely as pos-
Pharr and Griffiths.53 56 This technique is described in the
sible, to the original optical signals. This step has the
article on infrared spectroelectrochemistry in this Handbook
effect of nulling any differences in delay between the
by Korzeniewski (see Infrared Spectroelectrochemistry).
two channels, as well as flattening the amplitude response.
Consequently, the measured signals become independent
3.4 Other applications of sample-modulation of velocity, thereby making the measurements immune to
measurements effects of velocity errors.
Interferograms measured with step-scan interferometers
Many other applications of sample modulation measure- have errors analogous to those encountered in rapid-scan
ments using phase-modulated step-scan interferometers operation, but rather than small variations in optical veloc-
have been reported. These applications include studies of ity, the errors for step-scan interferometers are caused by
the orientation dynamics of liquid crystals in a modu- variations in mirror position. Velocity errors in conven-
lated electric field.57 59 This work is also summarized in tional rapid-scan measurements are compensated to first
the article in this Handbook by Czarnecki (see Vibra- order by the use of triggering derived from the laser
tional Spectroscopy of Liquid Crystals). Guelachvili has signal. For step-scan interferometers, the infrared signal
used Zeeman modulation60,61 and Zeeman splitting with observed at the detector is very sensitive to mirror posi-
polarization modulation62 to observe radical species, con- tion, according to the derivative of the infrared interfero-
centration modulation to observe transient species63 and gram. In measurements described in the literature to the
velocity modulation to separate the spectral signature of present time, it has been assumed that the mirror has
charged species from neutral plasma emission.64 These been held accurately at the correct position, and only the
measurements have employed a high-resolution step-scan infrared interference signal has been recorded. The result
interferometer. Guenter and Alibert have used a modulation is that position errors are compensated to first order by
technique to extract the weak luminescence of mid-infrared signal averaging. To the extent that the position errors
light-emitting diodes from the thermal background with are zero-mean and that the infrared signal is not sharply
a step-scan interferometer.65 Keiderling has shown that curved, this solves the problem entirely for some measure-
impressive vibrational circular dichroism spectra can be ments. In the new approach for rapid-scanning interferom-
measured using a step-scan FT-IR spectrometer.66 eters, the reference laser interference record is measured
and the data are post-processed to remove the effects of the HeNe lasers used as retardation references. In addition
error. An analogous approach may also be appropriate to their considerable size advantages, semiconductor lasers
for application to step-scan measurements. The effects of are often 50% efficient, in comparison to the 0.01% effi-
position errors in subtle measurements may then be com- ciency typical of HeNe lasers. For most applications, the
pensated and removed. The presence of phase modulation loss of 10 or 15 W of electric power is an insignificant cost,
has been ignored in the logical development described especially because, for most applications, electricity must
here. A similar line of reasoning applies in the presence be continuously available for operating an infrared source
of phase modulation, with the added complication that that consumes at least 20 W. The real problem with HeNe
the infrared signal observed at the detector is propor- reference lasers is that the waste heat is often dissipated in
tional to the phase modulation amplitude. Perfect control the interferometer compartment where it can lead to false
of the phase modulation waveform and amplitude is not path difference errors. A variety of methods for stabilizing
possible, but the error compensation methods described diode lasers to the requisite part per million range can be
above can be extrapolated to correct for these issues contemplated.
as well.
4.2 Phase- and sample-modulation imaging 5 CONCLUSIONS

measurements
The utility of step-scan and other interferometric methods
As ADCs continue to evolve to higher speeds and reso- for extracting time-dependent spectral information from a
lutions, and the cost of processing larger data sets con- variety of systems is well established. DSP, coupled with
tinues to fall sharply, DSP approaches will become more increasingly powerful computers, will continue to open new
important. One particularly intensive signal-processing possibilities for measuring and processing massive data
burden is spectral imaging data sets, which can eas- sets. The use of data processing to account for nonideal
ily contain 100 megabytes of data per image. In recent behaviors of hardware is a very useful direction because
years, such infrared spectral imaging measurements have the cost of an additional unit of software is very low. By
become widespread.25 27 The various imaging methods are shifting complexity from hardware to software the unit cost
extremely powerful tools for extracting data from hetero- of new instruments can be reduced, while at the same time
geneous real-world samples. To date, these measurements realizing improvements in performance. These and other
have not involved either phase modulation or sample mod- enhancements will increase the utility of interferometric
ulation. Many of the low-frequency disturbances that are modulation measurements, whilst simultaneously decreas-
largely rejected by phase modulation, such as source inten- ing the cost of instrumentation. The application of modula-
sity drift, adversely affect imaging measurements. Because tion measurements to the extraction of relevant information
most real-world samples are heterogeneous, it is reason- from complex systems will continue to increase, perhaps
able to expect interest to continue to shift in the direction dramatically.
of imaging.
The use of phase modulation greatly improved the per-
formance step-scan measurements18 and can be anticipated ACKNOWLEDGMENTS
to improve the performance of imaging measurements. As
array detectors become faster, the useful phase modulation The helpful comments provided by Peter Griffiths of
frequencies will increase and continuous-scan measure- the University of Idaho, Tim Johnson of Pacific North-
ments will displace step-scan for static measurements.72 west National Laboratories, Richard Jackson of Bruker
Another consequence of faster array detectors will be time- Optics, Richard A. Palmer of Duke University, Guy
resolved spectral imaging and phase-resolved spectral imag- Guelachvili of Universite de Paris-Sud, Larry Mertz of
ing using sample modulation. These new experiments will Digiphase and Eric Jiang of Thermo-Nicolet are grate-
drastically increase the number of data points that must be fully acknowledged. Technical data provided by Richard
processed. Jackson, Robert Badeau of Thermo-Nicolet Instrument Cor-
poration and Raul Curbelo of BioRad/Digilab are appre-
ciated. The past, present and future customers of MAT
4.3 Solid-state reference lasers are warmly thanked for their contributions to this arti-
cle. The comments on servoloop performance are the
Semiconductor lasers will soon provide a source of result of a discussion with Richard Jackson, ongoing since
frequency- and intensity-stable coherent radiation to replace 1993.
ABBREVIATIONS AND ACRONYMS International Conference on Fourier Spectroscopy, Aspen,

CO, 301 (1970).
ADC Analog-to-digital Converter 24. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
AMD Advanced Micro Devices Spectrometry, John Wiley & Sons, New York (1986).
DSP Digital Signal Processing and Digital Signal 25. D.A. Simons, C.C. Clark, S.S. Smith, J. Kerr, S. Massey and
Processors J.-P. Maillard, SPIE Proc., 2198, 185 (1994).
MMX Multi-media-extensions 26. E.N. Lewis, I.W. Levin and P.J. Treado, US Patent 5 377 003
(1994).
27. E.N. Lewis, P.J. Treado, R.C. Reeder, G.M. Story, A.E.
REFERENCES Dowrey, C. Marcott and I.W. Levin, Anal. Chem., 67, 3377
(1995).
1. D.M. Livingston, Master of Light, Scribner, New York 28. C.J. Manning, AIP Conf. Proc., 430, 84 (1997).
(1973). 29. R. Curbelo and D.B. Johnson, US Patent 5 166 749 (1992).
2. L. Mertz, Astron. J., 70, 548 (1965). 30. J. Connes, H. Delouis, P. Connes, G. Guelachvili, J.-P.
3. W.M. Grim, III, J.A. Graham, R.M. Hammaker and W.G. Maillard and G. Michel, Nouv. Rev. DOptique Applique, 1,
Fateley, Am. Lab., 16, 22 (1984). 3 (1970).
4. S.I. Yaniger and D.W. Vidrine, Appl. Spectrosc., 40, 174 31. C.J. Manning and P.R. Griffiths, Appl. Spectrosc., 47, 1345
(1986). (1993).
5. K. Masutani, H. Sugisawa, A. Yokota, Y. Furukawa and 32. D. Debarre, A.C. Boccara and D. Fournier, Appl. Opt., 20,
M. Tasumi, Appl. Spectrosc., 46, 560 (1992). 4281 (1981).
6. A.J. Turner, US Patent 5 450 196 (1995). 33. Model 05 STP 901, Melles Griot, Carlsbad, CA, USA.
7. H. Weidner and R.E. Peale, Appl. Spectrosc., 51, 1106 (1997). 34. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
Spectrometry, John Wiley & Sons, New York, 131 (1986).
8. F. Ozanam and J-N. Chazalviel, Rev. Sci. Instrum., 59, 242
(1988). 35. C.J. Manning, G.L. Pariente, B.D. Lerner, J.H. Perkins,
R.S. Jackson and P.R. Griffiths, Multiple Modulation Step-
9. L. Mertz, J. Phys. (Radium), 19, 233 (1958). scanning Fourier Transform Infrared Spectrometry, in Com-
10. J. Chamberlain, Infrared Phys., 11, 25 (1971). puter Enhanced Analytical Spectroscopy, ed. S.D. Brown,
Elsevier, Chichester, UK, 121, Vol. 6 (1996).
11. J. Chamberlain and H.A. Gebbie, Infrared Phys., 11, 57
(1971). 36. P. Horowitz and W. Hill, The Art of Electronics, Cambridge
University Press, Cambridge, 628631 (1980).
12. P. Fellgett, PhD Thesis, Cambridge (1951).
37. R. Curbelo, US Patent 6 025 913 (2000).
13. L. Mertz, Personal communication (2000).
38. L.A. Nafie and D.W. Vidrine, Double-modulation Fourier
14. J. Connes and P. Connes, J. Opt. Soc. Am., 56, 896 (1966). Transform Spectroscopy, in Fourier Transform Infrared
15. R.A. Palmer, M.J. Smith, C.J. Manning, J.L. Chao, A.C. Spectroscopy, eds J.R. Ferraro and L.J. Basile, Academic
Boccara and D. Fournier, Step-and-Integrate Interferometry Press, New York, 83, Vol. 3 (1982).
in the Mid-infrared with Photothermal Beam Deflection 39. I. Noda, A.E. Dowrey and C. Marcott, Appl. Spectrosc., 42,
and Sample-Gas-Microphone Detection, in Photoacoustic 203 (1988).
and Photothermal Phenomena V, eds P. Hess and J. Pelzl,
Springer Series in Optical Sciences, Springer, Berlin, 50, 40. P. Malon and T.A. Keiderling, Appl. Spectrosc., 50, 669
Vol. 58 (1988). (1996).
16. J.W. Cooley and J.W. Tukey, Math. Computation, 19, 297 41. N. Picque and G. Guelachvili, Appl. Opt., 38, 1224 (1999).
(1965). 42. N. Picque and G. Guelachvili, Vibr. Spectrosc., 19, 295
17. B.O. Budevska, C.J. Manning, P.R. Griffiths and R.T. Rogin- (1999).
ski, Appl. Spectrosc., 47, 1843 (1993). 43. A. Harata and T. Sawada, J. Appl. Phys., 65, 959 (1989).
18. C.J. Manning and P.R. Griffiths, Appl. Spectrosc., 51, 1092 44. R. Curbelo, US Patent 5 612 784 (1997).
(1997). 45. B.O. Budevska and C.J. Manning, Appl. Spectrosc., 50, 939
19. E.Y. Jiang, R.A. Palmer and J.L. Chao, J. Appl. Phys., 78, (1996).
460 (1995). 46. I. Noda, A.E. Dowrey and C. Marcott, J. Polym. Sci., Polym.
20. K.H. Michaelian, Appl. Spectrosc., 43, 185 (1989). Lett. Ed., 21, 99 (1983).
21. R.A. Palmer, Spectroscopy, 8, 26 (1993). 47. R.A. Palmer, C.J. Manning, J.L. Chao, I. Noda, A.E. Dowrey
22. C.J. Manning, J.L. Chao and R.A. Palmer, Rev. Sci. Instrum., and C. Marcott, Appl. Spectrosc., 45, 12 (1991).
62, 1219 (1991). 48. C.J. Manning, US Patent 6 178 822 (2001).
23. R.E. Murphy and H. Sakai, Application of Fourier Spec- 49. H. Wang, R.A. Palmer and C.J. Manning, Appl. Spectrosc.,
troscopy to the Study of Relaxation Phenomena, Aspen 51, 1245 (1997).
50. I. Noda, Appl. Spectrosc., 44, 550 (1990). 61. M. Elhanine, R. Farrenq and G. Guelachvili, Mikrochim.
51. A.E. Dowrey, C. Marcott and I. Noda, SPIE Proc., 2089, 424 Acta., III, 265 (1988).
(1993). 62. M. Elhanine, R. Farrenq and G. Guelachvili, Appl. Opt., 26,
52. C. Marcott, G.M. Story, I. Noda, A. Bibby and C.J. Manning, 4024 (1989).
AIP Conf. Proc., 430, 379 (1997). 63. A. Benidar, G. Guelachvili and P.A. Martin, Chem. Phys.
Lett., 177, 563 (1991).
53. B.O. Budevska and P.R. Griffiths, Anal. Chem., 65, 2963
(1993). 64. P.A. Martin and G. Guelachvili, Phys. Rev. Lett., 65, 2535
(1990).
54. B.O. Budevska, C.M. Pharr and P.R. Griffiths, Proc. Soc.
Photo-Opt. Instrum. Eng., 2089, 420 (1993). 65. T. Johnson, Personal communication (2001).
55. C.M. Pharr and P.R. Griffiths, Anal. Chem., 69, 4665 (1997). 66. B. Wang and T.A. Keiderling, Appl. Spectrosc., 49, 1347
(1995).
56. C.M. Pharr and P.R. Griffiths, Anal. Chem., 69, 4673 (1997).
67. J.W. Brault, Appl. Opt., 35, 2891 (1996).
57. T. Nakano, T. Yokoyama and H. Toriumi, Appl. Spectrosc.,
47, 1354 (1993). 68. J.C. Brasunas and G. Mark Cushman, Appl. Opt., 36, 2206
(1997).
58. V.G. Gregoriou, J.L. Chao, H. Toriumi and R.A. Palmer,
Chem. Phys. Lett., 179, 491 (1991). 69. A. Turner, R.A. Hoult and M.D. Forster, US Patent 5 914 780
(1999).
59. R.A. Palmer, V.G. Gregoriou, A. Fuji, E.Y. Jiang, S.E.
Plunkett, L.M. Connors, S. Boccara and J.L. Chao, ACS 70. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
Symposium Series 598, Multidimensional Spectroscopy Spectrometry, John Wiley & Sons, New York, 124 (1986).
of Polymers, eds M.W. Urban and T. Provder, 99116 71. C. Manning and R.J. Combs, SPIE Proc., 3537, 181 (1999).
(1995). 72. C.M. Snively, S. Katzenberger, G. Oskarsdottir and J. Lauter-
60. G. Guelachvili, J. Opt. Sec. Am. B, 3, 1718 (1986). bach, Opt. Lett., 24, 1841 (1999).
Polarization Interferometers for Infrared
Spectrometry
Prasad L. Polavarapu
Vanderbilt University, Nashville, TN, USA
1 INTRODUCTION and millimeter wavelength regions until recently. With the

realization of the advantages9,10 that these interferome-
A majority of the Fourier transform infrared (FT-IR) spec- ters offer for polarization difference measurements, these
trometers used in research, industrial and teaching labora- interferometers have recently been developed for the mid-
tories are built around a Michelson interferometer,1 where infrared (MIR) region using flat mirrors as well as cube cor-
the unpolarized (or randomly polarized) incoming light is ner mirrors. New applications using these instruments are
divided into two components of equal amplitude by an being investigated in both the MIR and FIR regions.3,8 24
amplitude division beamsplitter, and these light components The commercialization of these interferometers has been
are recombined at the beamsplitter for interference. Thus a much slower than desired. Nevertheless, the advantages and
Michelson interferometer can be referred to as an amplitude new applications available with PDIs make it worthy to
division interferometer (ADI). A different concept of inter- expedite the development of these interferometers.
ferometry utilizes polarization as the basis for the division
and recombination of incoming light. An interferometer
based on this principle will be referred to as a polariza- 2 POLARIZATION DIVISION
tion division interferometer (PDI). For certain applications INTERFEROMETERS BASED ON A
polarization division interferometry is superior, at least in WOLLASTON PRISM
theory, to amplitude division interferometry. Different types
of polarization division interferometers are known.2,3 In a A schematic diagram for this interferometer2,4 6 is shown
broader sense they can be classified into two categories: in Figure 1. A Wollaston prism (designated BS in Figure 1)
those based on the division of polarized light by a Wollas- divides the incoming light (from an appropriate source or
ton prism,4 7 and those based on the division of polarized target) into two components with orthogonal linear polar-
light by wiregrid beamsplitters.3,8 24 The latter category izations. A Wollaston prism, made from either quartz or
can be divided into the following subgroups as those using: calcite crystals, contains two wedges held together such that
roof-top mirrors,8 flat mirrors,12 and cube corner mirrors.24 it forms a plane parallel plate. The optic axes of the two
The concepts of PDIs based on a Wollaston prism have been wedges are perpendicular to each other, and are parallel to
known2 for a long time, but actual spectroscopic applica- the respective external faces. If the incident light is linearly
tions have not emerged until recently.4 7 The spectroscopic polarized (using a polarizer P), with the polarization axis at
applications of PDIs based on wiregrid beamsplitters and 45 to either of these optic axes, then the two waves emerg-
roof-top mirrors appeared soon after the concepts were ing from the prism would have orthogonal polarizations and
presented by Martin and Puplett.8 The thrust of these equal amplitudes. The two polarization components of light
applications3 remained mostly in the far-infrared (FIR) exiting the prism travel with an angle a between their direc-
tions. This angle can be determined from the birefringence
John Wiley & Sons Ltd, 2002. of the prism material as a D 2ne no tan q, where ne and
P BS A L D P C BS A L D
Figure 1. PDI based on a Wollaston prism. P, linear polarizer; Figure 2. PDI based on a refractively scanned Wollaston prism.
BS, beamsplitter based on a Wollaston prism; A, linear polariza- P, linear polarizer; C, compensator; BS, refractively scanned
tion analyzer; L, lens; D, multi-element detector.4 6 Wollaston prism beamsplitter; A, linear polarization analyzer; L,
lens; D, single-element detector.7
no are the refractive indices for extraordinary and ordinary
detector, no mechanical scanning mechanism (such as that
rays and q is the angle of the wedge. When a linear polariza-
needed in the conventional techniques) is required. Since
tion analyzer with its axis parallel to that of the input polar-
there are no moving parts in this interferometer and it can
izer (designated A in Figure 1) is placed in their path, the
be miniaturized, this instrument is expected to have better
two polarizations can be projected onto the analyzers axis;
stability than the conventional interferometric systems, and
thus two light components emerge from the analyzer with
to be suitable for rugged imaging applications in the field.
the same polarization. An imaging lens (L) focuses these
Analytical and quality control applications in an indus-
two rays on to a focal plane array detector (D), where they
trial setting can also be anticipated for this interferometer.
produce an interferogram. Each pixel of the array detector
Limitations of this interferometer include low spectral res-
represents a different point of the interferogram in spatial
olution and difficulties in extending the measurements to
domain. This interference results from the optical path dif-
the MIR and FIR regions.
ference D da between the two rays, which increases with
A variant of this instrument (Figure 2) using refractive
the displacement d from the center of the prism along the
scanning and a single detector was also introduced by
face of the prism that receives incoming light. The Fourier
Buhler.7 In this instrument, the beamsplitter is similar to
transform of the interferogram recorded by the array detec-
that in Figure 1, except that the two wedges of the Wollas-
tor yields the spectral distribution of the source (that is,
ton prism are physically separated, with one wedge fixed
emission intensity of the source as a function of wave-
in position and the other mounted on a movable carriage.
length). The shortest wavelength that can be detected is
As the moving wedge translates, the optical path difference
given as lmin D 2aw/M, where w is the pixel width of the
experienced by the two polarization components of light
array detector (usually 25 m) and M is the magnification
changes. The scanning mechanism of the wedge is simi-
of the optical system (determined by the focal length of the
imaging lens). The spectral resolution, in wavenumbers, is lar to that used in refractively scanned interferometers.25
given by the relation nN D M/aw. The operating range is The detector signal as a function of the position of mov-
determined by angle q and transmission properties of the ing wedge represents the interferogram. A commercial
Wollaston prism and the sensitivity of the detector. Using instrument with this design using quartz as the Wollaston
different diverging angles a and a magnification factor of 8 prism beamsplitter has been available,7 but is not currently
it can be seen that this system can perform anywhere from available.
the visible region (above 300 nm) to the near-infrared (NIR)
region (2 m). Emission, absorption and Raman spec- 3 POLARIZATION DIVISION
tral measurements4 6 using this interferometer have been
demonstrated in the visible region. Similar measurements INTERFEROMETERS BASED ON
can be undertaken in the NIR region using a calcite prism. WIREGRID BEAMSPLITTERS
This interferometer was introduced commercially in the
early 1990s by Photonex Ltd (Glasgow, UK) and distributed An interferometer based on the principle of polarization
in the USA by Boston Electronics Corp. (Brookline, MA). division using a wiregrid beamsplitter was introduced by
Recently, however, this product was discontinued. Martin and Puplett8 for the FIR region. In this interferome-
Since spectral information is obtained from the Fourier ter the linearly polarized input light is divided by a wiregrid
transform of the interferogram registered at the array beamsplitter into two polarization components and brought
Polarization Interferometers for Infrared Spectrometry 3
back to the beamsplitter using two roof-top mirrors. The Table 1. Polarization division beamsplitters for the MIR and FIR
PDIs utilizing wiregrid beamsplitters and roof-top mirrors regions.
are referred to as MartinPuplett interferometers.
BaF2 a KBra KRS-5a W wiresb Mylarc
Support substrate BaF2 KBr KRS-5 None Mylar

3.1 MartinPuplett interferometers Blank substrate BaF2 KBr KRS-5 None None
Substrate thickness 3 mm 5 mm 4 mm 6 m
The principles of PDI can be illustrated using a collimated Spacing between 0.25 0.25 0.25 12 4
ray from a light source (SO), as shown in Figure 3. A wires (m)
linear polarizer, referred to as the input polarizer (P), Lower frequency 800 400 300
limit (cm1 )
polarizes the collimated light. The linearly polarized light
Residual (%) 13 5 40 25 12
falls on a polarization division beamsplitter (BS). For the
a Wiregrids
MIR region, wiregrid polarizers (metal wires on a suitable deposited on BaF2 , KBr and KRS-5 substrates can be pur-
substrate, such as BaF2 , KBr or KRS-5) with a matching chased from Specac Ltd (UK) or Molectron (USA). A beamsplitter can
be constructed from these substrates as described in Polavarapu et al.15
blank substrate have been used as polarization division b Freestanding W-wiregrids can be purchased from Specac Limited (UK).
c Wiregrids deposited on mylar substrate are available from Sciencetech
beamsplitters. For the FIR region, beamsplitters made of
Inc. (Canada).
freestanding tungsten wires (at 12 m spacing) as well as
aluminum wires (at 4 m spacing) deposited on a 6 m thick
in a perpendicular polarization component (S1 ) and this
mylar substrate have been used.20 In the case of mylar, the
component is now transmitted by the beamsplitter. Simi-
film thickness is small enough that a matching substrate
larly the polarization component S2 is rotated by the mirror
is not required. Table 1 summarizes the characteristics of
M2 to result in a parallel polarization component (P2 ) and
different polarization division wiregrid beamsplitters.
this component is now reflected by the beamsplitter. The
Polarization of the light coming through the input polar-
components S1 and P2 exit at 90 to the direction of the
izer can be resolved (see Figure 3) into two components,
input beam. When the mirrors M1 and M2 are at equal dis-
one parallel and the other perpendicular to the direction
tances, there is no optical path difference between S1 and
of wires in the wiregrid beamsplitter. The parallel compo-
P2 , so this position is referred to as the zero path difference
nent (P1 ) is reflected to a fixed roof-top mirror (M1 ) while
(ZPD) point. At the ZPD point, polarization of the output
the perpendicular component (S2 ) is transmitted to another
beam is identical to that of the incoming beam, which is set
roof-top mirror (M2 ) that is movable. The fixed mirror is
by the input polarizer P. As the mirror M2 travels away from
mounted on translational stages to give three degrees of ori-
the ZPD point, the polarization of the output beam changes
entational adjustments: tilt in the axis perpendicular to roof
depending on the length of the travel by the mirror M2 .
edge; lateral shear; and rotary shear. The polarization com-
For monochromatic input light of wavelength l, the
ponent P1 is rotated 90 upon reflection by M1 to result
mirror travel distances of l/8, 2l/8, 3l/8 and 4l/8 cor-
M1
respond, respectively, to phase differences between S1 and
MartinPuplett P2 of l/4, l/2, 3l/4 and l, or equivalently to polarization
interferometer P1 states of left circular, vertical, right circular and horizon-
S1
SO tal polarization. This modulation cycle is repeated as the
S2
M2 mirror travels to longer distances. This polarization modula-
P2 BS P tion is shown in Figure 4. For polychromatic light, different
wavelengths go through the above-mentioned polarization
states at different mirror travel positions, since for a longer
wavelength correspondingly longer mirror travel is needed
L1 A A to complete one cycle of modulation. Since the mirror
L1 L1 L1
travel needed for spectra obtained at nN cm1 resolution is
1/2 nN cm1 ), the number of full polarization modu-
SA SA
lation cycles each wavelength component undergoes at this
resolution is 1/lnN . For an interferometer scan to yield
L2 L2 L2 L2
4 cm1 resolution spectra in the 325 m region, this varies
(a) (b) (c) (d) between 800 and 100 modulation cycles. Thus this PDI
Figure 3. Schematic of a PDI based on a MartinPuplett design. serves to modulate the polarization of the input beam among
The interferometer and configurations (a)(d) for using the output all possible polarizations states and in the entire region
light in different measurements are shown. See text for details. appropriate for a given polarization division beamsplitter.
Figure 3 shows configurations appropriate for different

12 measurements. Configuration (a) is used to optimize the
H LCP V instrument. Since there is no linear polarization analyzer
in the beam, the detector should not see any interferogram
signal. The interferogram signal obtained in this configura-
(m)
8
H LCP V RCP H
tion is referred to as the residual interferogram. The smallest
peak-to-peak value of the residual interferogram obtained
to date is 13% (see Table 1) of that obtained in configura-
4 tion (b). In configuration (b), a linear polarization analyzer
H
LCP V RCP H LCP V RCP H is introduced in the output beam. A large interferogram is
0 1/2 2/2 3/2 4/2 5/2 6/2 7/2 8/2
generated in this configuration and the Fourier transform of
Mirror travel from ZPD (m) this interferogram gives the spectral intensity distribution
of the source. The transmission or absorbance of a desired
Figure 4. Polarization modulation incorporated by PDIs based
sample (SA) can be determined from the Fourier transforms
on wiregrid beamsplitters: H, horizontal polarization; LCP, left
circular polarization; V, vertical polarization; RCP, right circular of the interferograms measured in configurations (b) and
polarization. (c). Configuration (d) is used for measuring linear dichro-
ism and circular dichroism. When the analyzer is removed
For this reason, it is also appropriate to call this interfer- from configuration (c), the signal at the detector is pro-
ometer a polarization modulation interferometer. portional to linear and circular dichroism of the sample,
While the longer wavelength limit of a MIR Mar- which can be extracted respectively from the cosine and
tinPuplett interferometer is determined by the transmission sine Fourier transforms of the measured interferogram.
properties of the beamsplitter substrate, the shorter wave- MartinPuplett interferometers for the FIR region are
length limit is determined by the spacing of wires in the commercially available from Specac Ltd (Orpington, UK)
wiregrid beamsplitter and roof angle accuracy. The spac- and Sciencetech Inc (London, Ontario, Canada). Off-the-
ing of wires in the wiregrid beamsplitter should preferably shelf commercial instruments for the MIR region are
be four times smaller9 than the shortest wavelength to be not available, but custom-made instruments have been
measured. This condition is satisfactorily met for the MIR supplied13,14,24 in the past by Bomem Inc. (Canada).
region by the beamsplitters with 0.25 m grid-spacing. Nev-
ertheless, the extinction coefficient achievable with com-
3.2 Polarizing Michelson interferometers
mercially available wiregrid polarizers is of the order of
200 : 1. Using newer methodologies in photolithography, PDIs with flat mirrors are referred to as polarizing Michel-
wiregrid polarizers with extinction coefficients of the order son interferometers. When flat mirrors are used12 14 in
of 10 000 : 1 (at 10 m) have been made.26 Although they place of the roof-top mirrors shown in Figure 3, the
are not yet routinely available, the use of such polarizers components P1 and S2 return to the beamsplitter (Figure 5)
for constructing the beamsplitters will be essential for some with the same polarization. As a result, the exit beam traces
applications. The roof angle of the roof-top mirrors should the path of the input beam in the reverse direction. To
be9 within (l/4D)(360/2p) degrees of 90 , where D is the separate these two beams two different concepts have been
diameter of the beam. Commercially available roof-top mir- used. In the Bomem design12 (not shown here), the input
rors, with 2 arcsec tolerance for the roof angle, may be used beam is made slightly off-axis and focused before entering
for wavelengths as short as 2 m. For the wiregrid beam- the interferometer. The exit beam is made to focus at the
splitters with BaF2 , KBr and KRS-5 substrates, the longer same place as the input beam, but since they are slightly off-
wavelength cut-off is at 12, 25 and 30 m, respectively. axis a small prism placed at the focal point could be used
The collimated light exiting the PDI is better handled to divert the exit beam away from the input beam. Except
with lenses to avoid polarization distortions that are com- for these differences in the optical design, the principles of
monly introduced by the tight focusing mirrors. This is par- operation would be the same for the two interferometers
ticularly relevant for polarization sensitive measurements employing roof-top vs flat mirrors. A different version of
because light exiting PDI has polarization states that vary this interferometer is commercially available from Specac
with the distance of the moving mirror and these polariza- Ltd (Orpington, UK). In the Specac design (see Figure 5),
tion states are to be preserved to a high degree. For the the input polarizer is oriented at 45 to the incoming light
MIR region lenses made from KBr, ZnSe or BaF2 may be direction and the same polarizer serves the function of input
used. For the FIR region polyinfrared lenses22 have better and output polarizers. When the output light falls on this
transmission than the Si lenses. polarizer, the light reflected by polarizer is directed at 90
Polarization Interferometers for Infrared Spectrometry 5
M1 components P1 and S2 coming from the beamsplitter BS1

are reflected by the cube corner mirrors with the same
polarization. However, the light reflected by the cube corner
P1
mirror is displaced from the light incident on it. So, a
second beamsplitter BS2, with its grids oriented at 90 to
those of the first beamsplitter BS1, is used for processing
the polarization components reflected by the cube corner
S2
SO mirrors. Except for these differences, the operating principle
M2 is the same as that for the MartinPuplett interferometer.
P
P
BS 4 CONCLUSIONS
Figure 5. Schematic of a polarizing Michelson interferometer:
SO, light source; P, linear polarizer (oriented at 45 to the incoming Polarization interferometry for visible, NIR, MIR and FIR
light direction); BS, wiregrid beamsplitter; M1 , M2 , plane mirrors; regions is a well-developed subject. Attempts to commer-
P1 , polarization component parallel to the grid direction; S2 ,
cialize these instruments have appeared, but such instru-
polarization component perpendicular to the grid direction. The
output light exits at 90 to the direction of incoming light. In the ments have not yet found a place in routine laboratory
Bomem design,12 a small prism (not shown here) between P and measurements. Distinct advantages are offered,2,3 however,
BS diverts the output beam away from the input beam. by polarization interferometry. Hopefully these advantages
will be broadly recognized to generate renewed inter-
to the axis of incoming light. For dichroism measurements, est in making commercial products based on polarization
the output polarizer should not be present in the beam (see interferometry.
Figure 3d), so the Specac design is not suitable for dichro-
ism measurements, unless additional polarization modula-
tion accessories are introduced in the output beam. ACKNOWLEDGMENTS
Funding for PDI projects from NIH and NSF is gratefully
3.3 Polarization division interferometers with acknowledged.
cube corner mirrors
A PDI with cube corner mirrors has recently been custom ABBREVIATIONS AND ACRONYMS
built for the author by Bomem Inc. (Canada). A schematic
of this design is shown in Figure 6. When cube corner ADI Amplitude Division Interferometer
mirrors are used (in place of the roof-top mirrors), the PDI Polarization Division Interferometer
M1
REFERENCES
P1
1. A.A. Michelson, Light Waves and Their Uses, The Univer-
sity of Chicago Press, Chicago (1907).
2. M. Francon and S. Mallick, Polarization Interferometers,
M2 S2 Wiley Interscience, New York (1971).
BS1 SO
3. P.L. Polavarapu (ed.), Principles and Applications of Polar-
ization Division Interferometry, John Wiley & Sons, Chich-
BS2 P ester (1997).
4. T. Okamoto, S. Kawata and S. Minami, Appl. Spectrosc., 40,
691 (1986).
5. S. Takahashi, J.S. Ahn, S. Asaka and T. Kitagawa, Appl.
Spectrosc., 47, 863 (1993).
Figure 6. Design of a PDI with cube corner mirrors: SO, light 6. M.J. Padgett, A.R. Harvey, A.J. Duncan and W. Sibbett, Appl.
source; P, linear polarizer; BS1, BS2, wiregrid beamsplitters with Opt., 33, 6035 (1994).
grids perpendicular to each other and at 45 to the polarization axis
of P; M1 , M2 , cube corner mirrors; P1 , polarization component 7. NIRVIS: Computer Aided Training, Buhler Ltd. (Information
parallel to grids of BS1; S2 , polarization component perpendicular supplied by Professor P.R. Griffiths, University of Idaho.)
to grids of BS1. 8. D.H. Martin and E. Puplett, Infrared Phys., 10, 105 (1969).
9. D.H. Martin, Polarizing (MartinPuplett) Interferometric 17. P.L. Polavarapu, G.-C. Chen and Z. Deng, Appl. Spectrosc.,
Spectrometers for the Near- and Submillimeter Spectra, in 48, 1403 (1994).
Infrared and Millimeter Waves, ed. K.J. Button, Academic
18. P.L. Polavarapu, Z. Deng and G.-C. Chen, Appl. Spectrosc.,
Press, New York, 65148 (1982).
49, 229 (1995).
10. M.J. Dignam and M.D. Baker, Appl. Spectrosc., 35, 186
19. P.L. Polavarapu, Z. Deng and S. Weibel, Appl. Spectrosc., 50,
(1981).
98 (1996).
11. J.A. Bardwell and M.J. Dignam, Fourier Transform Polari-
20. P.L. Polavarapu and Z. Deng, Appl. Spectrosc., 50, 91
metry, in Fourier Transform Characterization of Polymers,
(1996).
ed. H. Ishida, Plenum, New York, 415443 (1987).
12. The Bomem DA3 Based Polarizing Interferometer Accessory,
(1996).
Technical Bulletin, Bomem Inc., Quebec, Canada (1987).
22. P.L. Polavarapu and G.-C. Chen, Appl. Spectrosc., 48, 1410
13. H. Ishida, Y. Ishino, H. Buijs, C. Tripp and M.J. Dignam,
(1994).
Appl. Spectrosc., 41, 1288 (1987).
23. P.L. Polavarapu, Appl. Spectrosc., 51, 770 (1997).
14. N. Raghunathan, N.S. Lee, T.B. Freedman, L.A. Nafie,
C. Tripp and H. Buijs, Appl. Spectrosc., 44, 5 (1990). 24. P.L. Polavarapu, P.K. Bose, A. Rilling and H. Buijs (in
preparation).
15. P.L. Polavarapu, G.-C. Chen and S. Weibel, Appl. Spectrosc.,
48, 1224 (1994). 25. W.M. Doyle, B.C. McIntosh and W.L. Clarke, Appl. Spec-
trosc., 34, 599 (1980).
(1994). 26. T. Leonard, personal communication (1995).
Diode Laser Spectrometers for Mid-infrared
Spectroscopy
Arlan W. Mantz
Connecticut College, New London, CT, USA
1 INTRODUCTION in High-resolution Fourier Transform Spectrometry of

Gases.
Semiconductor diode laser operation was first reported High-sensitivity detectors, discussed in Detectors for
about 1963 by Butler1 3 for lead salts (IIVI materi- Mid- and Far-infrared Spectrometry: Selection and Use,
als) and by Hall et al.,4 Nathan et al.,5 Holonyak et al.,6 were also developed during this period with background
and Quist et al.,7 for gallium arsenide (IIIV materials). limited performance in the case of longer wavelength
Semiconductor lasers form a unique class of laser devices detectors, and near quantum noise limited performance
because their output is tunable, a result of the temperature in the case of detectors operating closer to 3 m. The
dependence of the index of refraction of the material. Addi- availability of tunable semiconductor lasers, even with
tionally, the inherent line width of a semiconductor laser 10 to 100 W of cw power, when combined with these
was shown in 1969 by Hinkley and Freed8 to be about detectors provided real-time spectra with resolution and
50 kHz (0.0000017 cm1 at 2000 cm1 ), while the emis- signal-to-noise ratios (S/Ns) unheard of when using grating
sion line width from a free running laser using a standard spectrometers.
current supply (0.1 109 A(Hz)1/2 current noise) and a IIIV laser materials were developed primarily for near-
standard temperature controller (0.3 103 K thermal sta- infrared (NIR) fiber optic communication applications, so
bility) is typically 5 to 50 MHz (0.00017 to 0.0017 cm1 ), there are many spectroscopically interesting regions where
depending on both the temperature and current tuning these lasers are not currently available. IIVI materials,
rates of the laser. In 1983 Freed et al.9,10 measured line however, were developed primarily for spectroscopic appli-
widths of 20 kHz. This tunability and narrow emission cations including environmental monitoring of trace gases,
line width combine to make an interesting source for spectroscopy of hyperfine structures or unresolved Q branch
molecular spectroscopy, while the best resolution realized structures and measurement of molecular absorption line
in practice with state-of-the-art 10 m focal length echelle intensities.
grating spectrometers was 0.019 cm1 . Grating spectrome- Research applications of gas monitors, for ozone in the
ters are discussed in more detail in Monochromators for stratosphere for example, continue to this day, but most
Mid-infrared Spectroscopy. High-resolution Fourier trans- routine laser-based gas monitors now use NIR lasers
form spectrometers as developed primarily by Janine and because they operate at room temperature, and lead salt
Pierre Connes and colleagues at Laboratoire Aime Cotton, lasers require cryogenic cooling.
the only viable competing technology for tunable high In recent years, most spectroscopic laboratory research
resolving power instrumentation, were not available until applications in which mid-infrared (MIR) diode lasers are
about 1966.11 16 High-resolution Fourier transform spec- employed are concerned with measurement of absorption
trometers (resolution of 0.005 to 0.001 cm1 ) are discussed line shapes and line strengths followed by studies of van
der Waals molecules, molecular ions and collisional cooling
John Wiley & Sons Ltd, 2002. experiments.
2 THE SOURCE
Detector
In 1999, all MIR diode laser spectrometers used lead salt

lasers. Mantz17 describes semiconductor lasers in some
detail. In particular laser structures are described, and many Monochromator
references are provided for those interested in pursuing this Pulsed jet
topic in more detail. Diode laser

When comparing the state-of-the-art in lead salt (IIVI)
technology with gallium arsenide (IIIV) technology, much
of the available gallium arsenide technology is superior
in performance for spectroscopic applications. The supe-
rior features are continuous tuning in one mode, in many
cases single mode operation, and power per mode. The
problem with classical (IIIV) technology is that it is avail- Figure 1. An optical schematic of a TDL (tunable diode laser)
able only for wavelengths shorter than approximately 3 m spectrometer showing the diode laser, the experimental area (in
this case a pulsed supersonic molecular jet), the mode separation
where molecular absorption is generally weaker by about
monochromator and the detector. Note the absence of a permanent
three orders of magnitude than molecular absorptions in the calibration etalon. [Reprinted from Spectrochim. Acta Part A,
320 m spectral region, where vibrational fundamentals 52, V.M. Krivtsun, Yu.A. Kuritsyn, V.R. Mironenko, I. Pak and
are normally found. In spite of the experimental difficul- E.P. Snegirev, Spectroscopic Investigations with Pulsed Pb-salt
ties associated with weak absorptions, some groups have Diode Lasers, 925953, Copyright (1996), with permission from
Elsevier Science.]
expanded their research of line shapes into the NIR using
IIIV devices.
It must be mentioned that new, sophisticated laser 3 THE OPTICAL SYSTEM
structures utilizing IIIV materials have been reported18 in
which the laser emission region is engineered by designing Until the late 1980s most diode laser spectrometers con-
a quantum cascade structure instead of designing a material sisted of the laser with its current supply and temperature
with a band gap matched to the emission region, as has controller, a monochromator, a detector and electronics to
been done previously. Several groups around the world process the laser signal as shown in Figure 1. However,
are performing research on quantum cascade lasers and as mentioned earlier, intrinsic laser line widths as low as
reporting intriguing results. Some of these devices are being 20 kHz were measured using heterodyne techniques9,10 but
evaluated in spectroscopic applications,19 22 but the devices the line widths realized in a normally operating free running
are not yet available commercially. laser are more like 10 to as much as 50 MHz due to current
To summarize, the technology at this time should be and temperature tuning rates and the broadband noise level
characterized as in development, and we can expect inherent to the control electronics. Heterodyne measure-
significant improvements in the future, provided funding ments involve locking the diode laser emission wavelength
for research and development of these sources contin- to the emission wavelength of a highly stable (typically bet-
ues. Future improvements will be found in the ability ter than 1 part in 1012 ) CO2 laser which is operating only
to operate devices at temperatures high enough to per- on one CO2 molecular transition. The source of line width
mit thermoelectric cooling or room temperature opera- improvement is the stabilization of the diode laser emission
tion, as well as higher single mode power and improved wavelength by a highly stable reference, thus electronically
tunability. compensating for electronic drift and noise.
Detectors used in these spectrometers are usually cryo- The motivation for having narrow diode laser emission
genically cooled to achieve maximum sensitivity, and width is, of course, the realization of higher resolving power
they are described in Detectors for Mid- and Far- and the resulting ability to perform more precise inves-
infrared Spectrometry: Selection and Use. While most tigations of absorption line shapes, a topic of significant
research instruments use cryogenically cooled detectors, contemporary interest.
room temperature and thermoelectrically cooled (see Detec- In the mid 1980s, several groups began investigating the
tors for Mid- and Far-infrared Spectrometry: Selec- feasibility of stabilizing the emission wavelength of a diode
tion and Use) detectors are available with adequate sen- laser to an optical interference fringe produced by a highly
sitivity, but they are usually employed only in dedi- stable interferometer.
cated applications requiring long periods of unattended In order for this technique to be useful spectroscopically,
operation. the stabilized laser must be tuned while being stabilized. In
Diode Laser Spectrometers for Mid-infrared Spectroscopy 3
Evacuation valve spectrometer with emission wavelength stability better than

1 part in 108 by locking to an infrared (IR) fringe in
an etalon where the basic length is actually controlled
Beam splitter by locking the etalon to a visible fringe produced by
the etalon using a stabilized HeNe laser. Iodine-stabilized
HeNe lasers can achieve stabilities of about 1 part in 1011 ,
while most commercially available stabilized HeNe lasers
quote a stability of 2 parts in 108 . A stability of 1 part in
108 is the wavenumber scale precision realized in a step-
scan Fourier transform spectrometer (see High-resolution
Fourier Transform Spectrometry of Gases).
Winnewisser et al.23 25 developed a modified spherical
FabryPerot etalon, originally developed by Connes and
Superinvar rod spacers Scanner plate Connes15 by introducing a coupling plate inside the spheri-
Figure 2. Internally coupled FabryPerot interferometer. The cal FabryPerot; they refer to this as an internally coupled
entrance and exit beams (not shown) are perpendicular to the FabryPerot interferometer (icFPI). Tunability of the etalon
figure plane at the beamsplitter level. The rotation axis of the was accomplished by incorporating a KBr plate in the
tuning scanner plate is also perpendicular to the figure plane.
[Reproduced from C. Nicholas and A.W. Mantz, Applied Optics, interferometer cavity, and the plate could be rotated to
28, 4525 (1989), copyright Optical Society of America.] change the effective length of the interferometer (Figure 2).
Nicholas and Mantz26 scanned a fixed length icFPI by
the case of stabilization of the diode laser to an interference developing a diode laser step-scan system in which the
fringe this translates into the requirement that the fringe laser jumped from one fringe to the next; at each fringe
may be tuned in some manner. Tunable interferometric the laser was actively stabilized and data were collected.
fringes require a tunable interferometer. Schurman et al.27 29 further modified the icFPI by mount-
Two groups carried these investigations through to the ing one of the spherical mirrors on a piezoceramic stack
realization of an interferometrically controlled diode laser whose length could be varied; this combination replaced the
BS
Input
M2
Output
M1
Magnet
M1 Movable M3
Coil Piezoelectric
cats eye holder
(a)
+ Input
M2
+
Output
M1
TDL beam HeNe beam
(b) (c)
Figure 3. Michelson interferometer design. Note that cats eyes are utilized instead of plane mirrors or corner cube reflectors. This is
a stepping interferometer and the magnet is used to step from one position to the next. The piezoelectric holder supports a secondary
mirror in the cats eye; the secondary mirror position is modulated in order to provide an error signal to a servo loop which maintains a
fixed path difference between the two arms of the interferometer. [Reprinted from Spectrochim. Acta Part A, 52, A. Valentin, Tunable
Diode Laser Wavelength Control by a Michelsen Interferometer, 823833, Copyright (1996), with permission from Elsevier Science.]
Laser control Mirror

module Diode laser
Beam splitter
Feedback loop HeNe laser
icFPI
Trigger
Reference cell Beam splitter

Reference
Signal cell
PC Signal
DI.1 VCO Mirror
Detector
Figure 4. Block diagram of the wavelength-stabilized TDL spectrometer. The single mode TDL is locked to the 80 cm tunable internally
coupled icFPI by means of a feedback loop. The optical path length is varied by rotating a scanner plate which is also located inside
the cavity. By counting HeNe fringes the frequency scan becomes exactly reproducible, permitting averaging of scans. The second
part of the beam traverses the reference cell, and the third traverses a multiple reflection cell of Herriott design with a total optical
pathlength of 250 m. DLI, digital lock-in amplifier; VCO, voltage-controlled oscillator. [Reprinted from Spectrochim. Acta Part A, 55,
G. Winnewisser, T. Drascher, T. Giesen, I. Pak, F. Schmuling and R. Schieder, The Tunable Diode Laser: A Versatile Spectroscopic
Tool, 21212142, Copyright (1999), with permission from Elsevier Science.]
rotating scanning plate. The Winnewisser icFPI is available scanning icFPI changes the effective cavity length, defocus-
commercially (Laser Analytics Inc., Wilmington, MA). The ing the cavity and resulting ultimately in a loss of fringes;
wavenumber scale produced by these interferometers is the stepping Michelson interferometer does not suffer from
slightly nonlinear, because the fringe spacing changes as the this limitation. In the second case, the icFPI, the continu-
icFPI cavity length changes. The technique developed by ous scanning range of an icFPI is limited to approximately
Nicholas and Mantz26 maintains a constant fringe spacing, 1 wavenumber (cm1 ), while the Michelson imposes no
and the wavenumber scale utilizing this technique is linear. practical limit on the tuning range, which may, in prin-
Henry et al.30 32 utilized a high-resolution stepping ciple, be tens or even hundreds of wavenumbers (cm1 ).
Michelson interferometer (also called a Fourier transform Lead salt lasers, at the present state of the art, tune less
spectrometer) (Figure 3) derived from the original designs than 2 wavenumbers (cm1 ), so for lead salts this discussion
of Connes et al.16 The optical path difference between the may appear to be somewhat academic. However, lead salt
two arms of the stepping Michelson interferometer is con- distributed feedback (DFB) lasers may have continuous tun-
trolled by actively stabilizing the interferometer fringes ability of 20 wavenumbers (cm1 ), and in this case the
utilizing a stabilized HeNe laser typically stabilized to about limitations discussed above are real. Some lead salt DFB
2 parts in 108 , as was mentioned before. The interferom- lasers have been fabricated, but practically speaking, all
eter is scanned by jumping from one HeNe fringe to the DFB lasers operate in the NIR spectral region. Quantum
next fringe. While this technique is only available com- cascade lasers have also been developed with longer con-
mercially for research grade interferometers, it is quite tinuous mode tuning characteristics.
reliable and it has the possibility of possessing several Increased stability in the laser emission wavenumber
advantages when compared to the scanning icFPI. First, the (cm1 ) provides several other advantages. First, increased
wavenumber scale generated by the step-scanning Michel- laser line width results in a type of noise referred to
son is linear and highly precise, limited only by the stability as FM (frequency modulation) noise. Laser FM noise is
of the stabilized HeNe and the servo system. Second, the transformed into AM (amplitude modulation) noise in a
TDL
Removable Removable
Alignment HeNe
Monochromator
Removable
icFPI
Det.4
Stabilized
HeNe Det.1
Sample
call
icFPI
Det.3
Det.2
Figure 5. An optical schematic of a typical wavelength-stabilized TDL system. The laser is situated in a liquid nitrogen Dewar (labeled
TDL). A monochromator may be used to isolate a laser mode or to identify the precise spectral region in which the laser is operating;
if the removable mirrors are removed, the monochromator is bypassed. The spherical FabryPerot closest to the laser is a fixed-length
etalon with well-measured fringe spacings; the purpose of this etalon is to provide relative markers (in cm1 ) for the laser mode. The
stabilized HeNe laser and the second spherical etalon are used for the purpose of simultaneous controlled scanning and wavelength
stabilization of the TDL. In this system, the etalon length may be changed by means of a piezoceramic stack-mounted mirror. Other
elements in the system are self-explanatory. [Reproduced from J.S. Balent Jr and A.W. Mantz, Applied Optics, 31, 1907 (1992),
copyright Optical Society of America.]
spectrum; this results in increased noise in the spectrum, in the improved wavenumber scale (cm1 ) and through
or decreased S/N. Active frequency stabilization of the improved S/N.
diode laser emission results in reduced FM noise and The optical schematic, or block diagram, for a wave-
significant increases in spectral S/N. Active diode laser length-stabilized TDL spectrometer utilizing an icFPI is
stabilization results in a narrower laser line width when shown in Figure 4. Frequently a monochromator is inserted
the interferometer fringe is modulated to create a lock (Figure 5) between the diode laser and the icFPI; this is
signal in the system. The alternative to modulating the not shown in Figure 4. The monochromator is typically
fringe is modulation of the diode emission wavenumber used to isolate a particular laser mode in which case the
(cm1 ). Modulation of the emission of a diode can be monochromator remains in the system during measure-
carried out simply by modulating the current flowing ments. Sometimes the laser operates in a single mode
through the diode, but this results in increased effective and, in this case, the monochromator is used to determine
line width while still reducing AM noise. Finally, active (to about 0.5 cm1 ) where the laser is emitting. When a
stabilization provides improved photometric accuracy, both monochromator is used only to determine where the diode
Reference gas cell InSb

Lock-in amplifier
temperature
control 2 2 (I0)d 2
Tuning d (I0)d
fp BS fork Studied gas cell InSb
fp current Lock-in amplifier
control TBL d 3
3 3 (I t)d
fp 1 BS (I t)d
It
Lock-in amplifier I0 It
1 InSb TBL beam I0 d Integration
Clock (f2) d (16 bits)
Interferometer
f2 PC computer
D and D HeNe beam
optical control
n
a
D, D d(R)
Stabilized
D and D HeNe laser
electronic control
Sampling control
Figure 6. The general design of a Michelson interferometer stabilized laser. The interferometer block is shown schematically in Figure 3.
A portion of the TDL beam is directed to the interferometer which optically controls the emission wavelength of the laser. A stabilized
HeNe laser is used to control the optical path difference between the two arms of the interferometer. In this figure, the reference gas
cell is used to provide the background signal in order to present ratioed spectra in the output. Wavelength calibration for the diode laser
output is provided by the interferometer control system. [Reprinted from Spectrochim. Acta Part A, 52, A. Valentin, Tunable Diode
Laser Wavelength Control by a Michelsen Interferometer, 823833, Copyright (1996), with permission from Elsevier Science.]
laser is emitting, the monochromator is normally removed shape is Lorentzian, and at pressures intermediate between
from the optical system during the measurement process. these two extremes the classical model is a Voigt, which
Since monochromators typically transmit only 10 to 20% is the result of convolving a Gaussian, or Doppler, line
of the incident radiation, it is desirable to remove the shape with a Lorentzian line shape. Classical line shapes
monochromator from an optical system whenever possible. and broadening parameters are discussed in detail by Smith
The general design of a Michelson interferometer stabi- et al.34 It is also necessary to include pressure shifts of
lized laser is shown in Figure 6. A monochromator is not spectral lines if a complete description of experimental
generally used in this system, because the dispersion intro- effects is to be attempted; Smith et al.34 also deal with this
duced by the grating may produce alignment problems in effect, and it will not be discussed further here.
the Michelson interferometer. Thus the applications of this However, as the photometric precision of the measure-
system have so far been limited to lasers with single mode ments increased, it became necessary to include the shape
operation. and width of the laser emission profile in analyzing the
The reader is referred to several excellent articles describ- experimental data. It also became evident that classical
ing the stabilized diode laser systems. The icFPI is further models were frequently inadequate when describing these
described by Winnewisser et al.33 The laser actively sta- detailed line shapes under different pressure conditions.
bilized by a stepping Michelson interferometer is further
described by Valentin.32
5 TUNABLE DIODE LASER LINE SHAPE
4 LINE INTENSITY AND LINE SHAPE The absorption line shape which is measured experimen-
MEASUREMENTS tally is the result of a convolution of the line shape, in this
case a TDL, and the actual shape of the absorption line. As
Historically, classical models have been used to describe mentioned earlier, the diode laser line width is usually less
detailed studies of absorption line intensities and line shapes. than 0.0002 cm1 , while a calculated Doppler width for a
Classical line shapes include the Doppler profile which is very low pressure absorption line in a small molecule, like
realized at very low pressures, typically less than 0.2 Torr. N2 O, is usually one or two orders of magnitude larger than
At pressures of one atmosphere and higher the classical line this value. Surprisingly, the exact shape of the diode laser
1.0
0.9 37
H Cl: (10),R 2 D = 2.94 103 cm1
L = 2.6 103 cm1
0.8
P(HCl + N2) = 24.3 Torr
It /I0 P(HCl) = 0.038 Torr
0.7
0 and 0 (cm1 atm1)
0.6 calcobs
V 0 = 0.0776 0.005 V 0 = 0.0804
0.5
S 0 = 0.0810 0 = 0.0276 0.005 S 0 = 0.0853 0 = 0.0405

0.5
H 0 = 0.0805 0 = 0.0211 0.005 H 0 = 0.0847 0 = 0.0321

0.5
2942.68 2942.72 2942.76 cm1 2942.72
Voigt TDL intensity distribution Gauss TDL intensity distribution

Figure 7. Observed R(2) HCl line perturbed by N2 ( ) and least squares fit line including in the convolution product a Voigt
TDL intensity distribution and a Voigt function for the HCl line profile ( . . . . ). The calc-obs residuals are plotted as dotted curves for
different HCl line profiles: V, Voigt profile; S (or H), soft (or hard) collision model, using both a Voigt TDL intensity distribution and
a Gauss TDL intensity distribution. [Reprinted from Spectrochim. Acta Part A, 55, A. Henry, D. Hurtmans, Collision Narrowing and
Speed Dependent Effect on Broadening in Vibration Rotation Line Profiles Perturbed by Different Buffer Gases up to One Atmosphere
Difficulties in These Analyses Coming from the Quality of the Tunable Diode Laser Emission, 19671986, Copyright (1999), with
permission from Elsevier Science.]
emission profile is not well understood at this time. Early Position of

heterodyne beat frequency measurement results strongly CH4R (0)
suggested the emission profile is Lorentzian.8,9 Contem-
porary line shape studies assume the laser line shape to be KrCH4
Absorption (2nd derivative)
Doppler35 for a good diode. Recent work suggests that

some diodes are better characterized by assuming a Voigt
profile,35 although this may suggest a high current tuning
rate, and associated wavenumber jitter of the diode. Cur- ArCH4
rent work is in progress to more accurately characterize the
laser emission profile by investigating molecular absorp- NeCH4
tion at low pressure (<0.1 Torr) and at low temperatures
where the Doppler profile is of the same order of magni-
tude as the laser emission width. Molecules like CO may
be studied at temperatures as low as 30 K, while collisional p -H2CH4
cooling experiments permit the study of absorption features
at temperatures of 10 K and lower.
1310.2 1310.4 1310.6 1310.8 1311.0 1311.2 1311.4
cm1
6 ABSORPTION LINE PROFILE
Figure 8. Montage on the same frequency scale of the absorption
As mentioned earlier, at room temperature and at very bands of RCH4 (R D Kr, Ar, Ne) and p-H2 CH4 associated
low pressure the spectral line profile depends only on with the R(0) line of the n4 fundamental band of pure methane.
[Reprinted from Spectrochim. Acta Part A, 55, G. Winnewisser,
the molecular velocity distribution, and it is precisely
T. Drascher, T. Giesen, I. Pak, F. Schmuling and R. Schieder,
described by a Gaussian function, commonly referred to The Tunable Diode Laser: A Versatile Spectroscopic Tool,
as a Doppler profile,34 with a position, strength and a half 21212142, Copyright (1999), with permission from Elsevier
width dependent on the square root of the gas temperature, Science.]
J=1
J=2 Q
P
J=0
R
1310.2 1310.3 1310.4 1310.5 1310.6 1310.7 1310.8 1310.9

(a)
J=0
R
J=2
P
J=1
Q
1310.6 1310.7 1310.8 1310.9 1311.0

(b) cm1
Figure 9. The assignment of the (a) KrCH4 and (b) ArCH4 bands associated with the R(0) transition of methane. [Reprinted from
Spectrochim. Acta Part A, 55, G. Winnewisser, T. Drascher, T. Giesen, I. Pak, F. Schmuling and R. Schieder, The Tunable Diode
Laser: A Versatile Spectroscopic Tool, 21212142, Copyright (1999), with permission from Elsevier Science.]
(in K). At high pressure classical line shape theory predicts instruments described here. Two effects are realized with
that the line shape is approximately described by a Lorentz Dicke narrowing, first, the Doppler component of the broad-
profile,34 with a strength, position and half width the value ening parameter changes progressively from Gaussian to
of which depends on the collision gas partners. The general Lorentzian and its width tends toward zero when the pres-
line shape is described by a Voigt profile34 with a strength, sure increases.
position and both a Gaussian and a Lorentzian broadening The progressive change depends on the nature of the
parameter. collisions which are referred to as soft or hard collisions;
As pressure increases from a very low pressure the these models require a collisional narrowing parameter in
mean free path of the molecules actually decreases, with addition to the parameters associated with the Voigt profile.
each molecule confined to a small volume for a char- Two collision models have been described by Galatry36 and
acteristic time. This molecular confinement is referred to by Rautian and Sobelman.37 Nadezhdinskii38 performed
as Dicke narrowing. And, while Dicke narrowing is a systematic analyses of soft and hard collisions, and his
small effect, it is observable with high precision TDL studies suggest that soft collisions describe less efficient
spectral narrowing than spectral narrowing due to hard Q

collisions. As the pressure increases both models tend to
give the same narrowing results, and this complicates the
J=0
choice of models. J =1 J=2 R
P
7 A BRIEF SUMMARY OF SOME NOVEL
APPLICATIONS OF TUNABLE DIODE
LASERS

Precise absorption line strength measurements are important
applications of actively stabilized tunable lasers because of (a) 1311.30 1311.35
the high resolving power associated with these sources.
Figure 7, taken from Henry and Hurtmans,35 illustrates
J =0 6
the complexity of this analysis. First, the laser emission J =1 7 R
line shape appears to depend in a complicated way on P J =1 7
the quality of the laser. Henry and Hurtmans35 find some Q
lasers have Doppler emission profiles and some have Voigt
emission profiles. Furthermore, for the observed R2 HCl
(1-0) line shown in Figure 7, the soft and hard collision
models yield better calculated-observed residuals than a
Voigt model, indicating that a Voigt model is inadequate
for this particular line.
The present situation is that it is so far not possible
to predict which absorption lines for individual molecules
1311.04 1311.14 1311.24 1311.34
might be best described by confinement narrowing or
(b) cm1
by soft or hard collision models. These effects are only
discovered through the analysis of the experimental data. Figure 10. The assignment of the (a) NeCH4 and (b) p-H2 CH4
bands associated with the R(0) transition of methane. For p-
van der Waals molecules and molecular clusters are
H2 CH4 , the P(7) and Q(7) lines are broadened due to predis-
molecular systems with very shallow potential minima; sociation. The dissociation energy of the p-H2 CH4 complex
therefore, one can only study these molecular systems with was estimated to be 28 cm1 . [Reprinted from Spectrochim.
the aid of supersonic nozzles, or cold multipass cells, or Acta Part A, 55, G. Winnewisser, T. Drascher, T. Giesen, I. Pak,
collisional cooling cells. Figure 8 taken from Winnewisser F. Schmuling and R. Schieder, The Tunable Diode Laser: A Ver-
et al.,33 shows a montage of spectra for van der Waals satile Spectroscopic Tool, 21212142, Copyright (1999), with
permission from Elsevier Science.]
molecules with the structure RCH4 where R D Kr, Ar,
Ne and p-H2 in the region of the R0 line of the
n4 fundamental band of pure methane. Figure 9 shows by their velocity-modulated frequency shifts away from the
KrCH4 and ArCH4 with the appropriate quantum number absorption due to the neutral (sometimes parent) molecules.
assignments, and Figure 10 shows the same information This technique is a very important aid in assignment of
for NeCH4 and p-H2 CH4 . These studies are leading to quantum numbers to the spectral features as well.
a significantly improved understanding of these molecules Fluorocarbon plasmas are widely used for the processing
through the analysis of the spectra for the first time. of semiconductor materials, and, in an attempt to better
Linear carbon clusters have also been studied. It turns understand some chemistry associated with these plasma
out that odd numbers of carbon appear to produce clusters etching techniques, the spatial distribution of molecular
with linear molecular structure (Figure 11), while even ions, such as CFC , have been studied in He/C2 F6 plasmas.
numbered carbon clusters have proven to be quite elusive. Figure 12 is a portion of the velocity-modulated spectrum
Winnewisser et al.33 provide an excellent description of of CFC first reported by Kawaguchi and Hirota.41
the experimental configurations which are utilized in these Many different techniques,42,43 including high-resolution
investigations. Fourier transform infrared (FT-IR) spectroscopy,44,45 have
Molecular ion spectroscopy is largely possible due to the been developed to study this important class of molecule.
development by Gruebele et al.40 of a velocity-modulation Collisional cooling techniques as developed by Willey
technique whereby the molecular ion spectra are detected et al.46 involve the use of a low temperature cell, normally
14
10
Intensity (mV)
5
10 RQ (J,0)
14 J= 1 2 3 4 5 6 7
1 200 400 600 800 1000 1200 1400 1600 1800 2048
C2 2012.9892 cm1
Index
J = 50 48 46 44 42 40 38 36 34 32 30 28 26 24
Figure 13. The low temperature first derivative absorption spec-
trum of CH3 F at 1475 cm1 which is associated with RQJ, 0
2012.7 2013.0 2013.2 2013.5 transitions in the n5 band of this molecule. The cell temperature
(cm1) was maintained at 7.5 K during this experiment. One should note
the relative line intensities and the line widths of lines in this
Figure 11. P-branch spectrum of linear C9 at 2013 cm1 mea- spectrum compared to the room-temperature spectrum shown in
sured with the Cologne carbon cluster TDL spectrometer. Figure 14. Features which are the width of a pen trace are noise
[Reprinted from Spectrochim. Acta Part A, 55, G. Winnewisser, spikes. [Reproduced from D.R. Willey, K.A. Ross, V. Dunjko and
T. Drascher, T. Giesen, I. Pak, F. Schmuling and R. Schieder, A.W. Mantz, J. Mol. Spectrosc., 168, 301312 (1994), copyright
The Tunable Diode Laser: A Versatile Spectroscopic Tool, Academic Press.]
21212142, Copyright (1999), with permission from Elsevier Sci-
ence.]
12.5
10
Absorption
Intensity (mV)
0
R (2)
4
R (0) 5
0.01 cm 1
10
RQ (J,0)
Figure 12. Observed R(2) and R(0) transitions of CFC in the
X1 C state by magnetic field modulation of the discharge plasma. 12.5 J=1 2 3 4 5 6 7 8
The line center frequencies of R(2) and R(0) are 1776.3998 and
1 200 400 500 800 1000 1200 1400 1600 1800 2048
1769.7430 cm1 , respectively. The ordinate shows the absorption
intensity with an arbitrary scale. The time constant of the lock-in Index
amplifier was 1 s. [Reproduced from K. Kawaguchi and E. Hirota, Figure 14. The room temperature first derivative absorption spec-
J. Chem. Phys., 83, 14371439 (1985), copyright American trum of CH3 F at 1475 cm1 which is associated with RQJ, 0
Institute of Physics.] transitions in the n5 band of this molecule. The path length was
60 cm and the pressure was less than 1 Torr. Laser modulation
maintained at temperatures above 1 K with absorption path conditions for this scan were identical to those used to collect the
data shown in Figure 13. Line intensities peak at J values corre-
lengths (single pass) of approximately 4 cm, although we
sponding to T D 300 K, and the lines are wider than those shown
have recently completed testing of a 40-cm path length in Figure 13. Features which are the width of a pen trace are noise
collisional cooling cell. The cells are filled with helium spikes. [Reproduced from D.R. Willey, K.A. Ross, V. Dunjko and
gas to a pressure of 2050 mTorr for the microwave to A.W. Mantz, J. Mol. Spectrosc., 168, 301312 (1994), copyright
several hundred millitorr as we have used the technique. Academic Press.]
The helium gas equilibrates to the cell wall temperature,
and the analyte gas to be studied is allowed to flow very The first IR collisional cooling results were reported47
slowly into the cell. Analyte molecules cool to the gaseous for CH3 F and they are reproduced in Figures 13 and 14.
helium temperature in less than 100 collisions, and the The RQ(J,0) branch for J D 1 to 7 is plotted in Figure 13.
analyte molecules encounter tens of thousands of collisions These data were recorded with a cell temperature of 7.5 K.
before encountering a cell wall where they condense. Figure 14 shows exactly the same region recorded with a
0.009 4. R.N. Hall, G.E. Fenner, J.D. Kingsley, J.J. Soltys and R.O.
Obs-calc (%)
Carlson, Phys. Rev. Lett., 9, 366 (1962).

0.000 5. M.I. Nathan, W.P. Dumke, G. Burns, F.H. Dill, Jr and
G. Lasher, Appl. Phys. Lett., 1, 62 (1962).
0.009 6. N. Holonyak and S.F. Bevacqua, Appl. Phys. Lett., 1, 82
1.02 (1962).
7. T.M. Quist, R.H. Rediker, R.J. Keyes, W.E. Krag, B. Lax,
1.00 A.L. McWhorter and H.J. Zeiger, Appl. Phys. Lett., 1, 91
(1962).
Signal (mV)
0.98
8. E.D. Hinkley and C. Freed, Phys. Rev. Lett., 23, 277 (1969).
0.96
9. C. Freed, J.W. Bielinski and W. Lo, Appl. Phys. Lett., 43, 629
(1983).
0.94
10. C. Freed, J.W. Bielinski and W. Lo, Proc. SPIE, 438, 119
0.92 (1984).
0.90 11. J. Connes, Rev. Opt., 40, 45 (1961).
2147.0751 2147.0811 2147.0871
12. J. Connes, Rev. Opt., 40, 116 (1961).
Wavenumber /cm 1
13. J. Connes, Rev. Opt., 40, 171 (1961).
Figure 15. A collisionally cooled spectrum of the R(0) line in the
CO fundamental band recorded at 11 K with a helium pressure of 14. J. Connes, Rev. Opt., 40, 231 (1961).
0.27 Torr and a CO/He mixing ratio of 0.000237. 15. J. Connes and P. Connes, J. Opt. Soc. Am., 56, 896 (1966).
16. J. Connes, H. Delouis, P. Connes, G. Guelachvili, J.-P. Mail-
cell at 300 K. The difference in the intensity distribution as lard and G. Michel, Nouv. Rev. Opt. Appl., 1, 3 (1970).
a function of the rotational quantum number, J, is striking,
17. A.W. Mantz, Spectrochim. Acta, Part A, 51, 2211 (1995).
as is the reduced width of the absorption lines.
Recently, we recorded absorption data for the carbon 18. J. Faist, F. Capasso, D.L. Sivco, C. Sirtori, A.L. Hutchinson
and A.Y. Cho, Science, 264, 553 (1994).
monoxide molecule with a cell temperature of 11 K.48
The observed and the observed-calculated data are plotted 19. S.W. Sharpe, J.F. Kelly, J.S. Hartman, C. Gmachl, F. Cap-
asso, D.L. Sivco, J.N. Baillargeon and A.Y. Cho, Opt. Lett.,
in Figure 15. This data set was recorded at a corrected
23, 1396 (1998).
pressure of 0.27 Torr helium pressure and CO/He mixing
20. B.A. Paldus, T.G. Spence, R.N. Zare, J. Oomens, F.J.M. Har-
ratio of 0.000237. We have performed a simultaneous fit of
ren, D.H. Parker, C. Gmachl, F. Capasso, D.L. Sivco, J.N.
CO data between 296 K and 11 K, and these results49 are Baillargeon, A.L. Hutchinson and A.Y. Cho, Opt. Lett., 24,
being prepared for publication. 178 (1999).
21. A.A. Kosterev, R.F. Curl, F.K. Tittel, C. Gmachl, F. Cap-
asso, D.L. Sivco, J.N. Baillargeon, A.L. Hutchinson and
ABBREVIATIONS AND ACRONYMS A.Y. Cho, Opt. Lett., 24, 1762 (1999).
22. R.M. Williams, J.F. Kelly, J.S. Hartman, S.W. Sharpe, M.S.
AM Amplitude Modulation Taubman, J.L. Hall, F. Capasso, C. Gmachl, D.L. Sivco, J.N.
DFB Distributed Feedback Baillargeon and A.Y. Cho, Opt. Lett., 24, 1844 (1999).
DLI Digital Lock-in Amplifier 23. M. Reich, R. Schieder, H.J. Clar and G. Winnewisser, Appl.
FM Frequency Modulation Opt., 25, 130 (1986).
icFPI Internally Coupled FabryPerot Interferometer 24. T. Giesen, M. Harter, R. Schieder, G. Winnewisser and
S/N Signal-to-noise Ratio K.M.T. Yamada, Z. Naturforsch., 43a, 402 (1988).
TDL Tunable Diode Laser 25. H.J. Clar, R. Schieder, M. Reich and G. Winnewisser, Appl.
VCO Voltage-controlled Oscillator Opt., 28, 1648 (1989).
26. C. Nicholas and A.W. Mantz, Appl. Opt., 28, 4525 (1989).
REFERENCES 27. M.J. Schurman and A.W. Mantz, Proc. SPIE, 2205, 494
(1993).
1. J.F. Butler, A.R. Calawa, R.J. Phelan, Jr, T.C. Harman, A.J. 28. J.S. Balent, Jr and A.W. Mantz, Appl. Opt., 31, 1907 (1992).
Strauss and R.H. Rediker, Appl. Phys., 5, 75 (1964). 29. M.J. Schurman, V.J. Dunjko, S. Goldstein, M. Baron and
2. J.F. Butler, A.R. Calawa, R.J. Phelan, Jr, A.J. Strauss and A.W. Mantz, J. Quant. Spectrosc. Radiat. Transfer, 52, 379
R.H. Rediker, Solid State Commun., 2, 213 (1964). (1994).
3. J.F. Butler and A.R. Calawa, J. Electrochem. Soc., 112, 1056 30. A. Valentin, C. Nicholas, L. Henry and A.W. Mantz, Appl.
(1965). Opt., 26, 41 (1987).
31. A. Henry, A. Valentin, M. Margottin-Maclow and F. Rachet, 41. K. Kawaguchi and E. Hirota, J. Chem. Phys., 83, 1437
J. Mol. Spectrosc., 166, 41 (1994). (1985).
32. A. Valentin, Spectrochim. Acta, Part A, 52, 823 (1996). 42. T. Oka, Phys. Rev. Lett., 45, 531 (1980).
33. G. Winnewisser, T. Drascher, T. Giesen, I. Pak, F. Sch- 43. T. Oka and T.R. Geballe, Astrophys. J., 351, L53
mulling and R. Schieder, Spectrochim. Acta, Part A, 55, 2121 (1990).
(1999).
44. N. Picque, These de Docteur en Sciences, Universite de Paris-
34. M.A.H. Smith, C.P. Rinsland, B. Fridovich and K.N Rao, Sud, Centre dOrsay (1998).
Intensities and Collision Broadening Parameters from Infra-
45. N. Picque and G. Guelachvili, Appl. Opt., 38, 1224
red Spectra, in Molecular Spectroscopy: Modern Research,
(1999).
Vol. III, ed. K. Narahari Rao, Academic Press, Inc., Orlando,
FL, 111248 (1985). 46. D.R. Willey, R.L. Crownover, D.N. Bittner and F.C. De
35. A. Henry and D. Hurtmans, Spectrochim. Acta, Part A, 55, Lucia, J. Chem. Phys., 89, 1923 (1988).
1967 (1999). 47. D.R. Willey, K.A. Ross, V. Dunjko and A.W. Mantz, J. Mol.
36. L. Galatry, Phys. Rev., 122, 1218 (1961). Spectrosc., 168, 301 (1994).
37. S.G. Rautian and I.C. Sobelman, Sov. Phys. Uspekkii, 9, 701 48. B. Aoaeh, Y. Abebe, A.W. Mantz, C.D. Ball, F.C. De Lucia,
(1967). D.C. Benner and M.A.H. Smith, The 55th Ohio State Uni-
versity International Symposium on Molecular Spectroscopy,
38. A.I. Nadezhdinskii, Spectrochim. Acta, Part A, 52, 1041 Columbus, OH, 1216 June 2000, Paper ME06.
(1996).
49. B. Aoaeh, N. Kolodziejski, A.W. Mantz, D.C. Benner,
39. V.M. Krivtsun, Yu.A. Kuritsyn, V.R. Mironenko, I. Pak and V. Malathy Devi, M.A.H. Smith, C.D. Ball and F.C.
E.P. Snegirev, Spectrochim. Acta, Part A, 52, 925 (1996). De Lucia, The 55th Ohio State University International
40. M. Gruebele, M. Polak and R.J. Saykally, Chem. Phys. Lett., Symposium on Molecular Spectroscopy, Columbus, OH,
125, 165 (1986). 1216 June 2000, Paper ME07.
FabryPerot Interferometers for Mid-infrared
Spectroscopy
Tetsuo Iwata
University of Tokushima, Tokushima, Japan
1 FUNDAMENTALS OF THE is incident on the system, interference fringes are formed

FABRYPEROT INTERFEROMETER as the result of superposition of electric fields of many
transmitted light beams which are generated by multiple
reflections at the two inner surfaces. The two plates are
The FabryPerot spectrometer (FPS), originally developed
made slightly wedge-angled with respect to each other in
by C. Fabry and A. Perot in 1896, has not only extremely
order to avoid disturbing effects due to reflections at the
high spectral resolution but also a large luminosity-resolu-
outer uncoated surfaces. The plates are frequently held
tion product.1 3 However, its applicable spectral range is
with a fixed spacer and this arrangement is known as a
limited to a very narrow interval. Consequently, the FPS has
FabryPerot (FP) etalon. For use in the visible and UV
not been used very much in the mid-infrared region for mea-
spectral region, Ag is usually coated on a glass or a fused
suring continuum spectra. The main instrument to measure
silica plate, and for use in the IR region, dielectric mate-
the mid-infrared spectra is the Fourier transform spectrome-
rials such as Ge, Te, Te-NaCl, Zn-Na3 AlF6 , Ge-Na3 AlF6 ,
ter (FTS). The FPS is mainly used for measuring super-fine
etc., are coated on a KBr, NaCl, or ZnSe plate. In the far-
spectral structures of laser lines and Brillouin scattering in
infrared spectral region, a metal mesh is used in place of
the visible and UV region. Much effort has been made to
the plate.4 6 The use of a dual liquid crystal FP etalon
extend the applicable spectral range of the FPS, includ-
has been reported for measuring fluorescence and Raman
ing into the mid-infrared region. The FPS has potential in
images.7
the mid-infrared range, for example, as a remote sensing
instrument because of its compactness. In this article, we
will review the fundamentals of the FabryPerot interfer-
2 INTENSITY TRANSMITTANCE
ometer (FPI) and describe some methods for extending its
spectral range as a spectrometer. Figure 1 shows the FP etalon surrounded by air whose
The FPI is one of the multiple-beam, division-of-ampli- refractive index is unity. Suppose that the refractive index
tude interferometers. It consists of two flat, partially trans- of the FP etalon medium is n and that the etalon spacing
parent plates separated by a fixed distance. The inner is d. For simplicity, we neglect absorption in the etalon
surface of each plate is coated with metal or high refractive- medium. Also, we neglect the phase shift of the electric
index materials such as dielectric multilayers. The distance field of the light wave at each reflection on the two-
between the two plates is from 10 m to several centime- plate surfaces. Then, let us assume that a monochromatic,
ters, the flatness or the parallelism of the two surfaces parallel, p- (or s-)polarized light beam of a wavelength l
is better than 1/50 of a wavelength, the diameter of the (or wavenumber s) is incident on the FP etalon with an
plate is 210 cm, and the reflectivity of the coatings is in angle q to the optical axis. The complex amplitudes of the
the range of 8595%. When a quasi-monochromatic light incident and transmitted light amplitude are denoted by i
and a, respectively. As the result of multiple reflections, the
John Wiley & Sons Ltd, 2002. transmitted light is expressed by the following equation:
1 n 1
d I =1
I0
R = 0.1
itr 4t exp(i)
a
itr 3t
I = aa R = 0.3
itr 2t exp(i)
itr t itr 3
R = 0.8 R =0.95
itr 2 itt R = 0.6
itr
ir =2(m 1) =2m =2(m+1)
it

Figure 2. Airy distribution function as a function of for various
reflectivities (R D 0.95, 0.8, 0.6, 0.3 and 0.1), where q D 0, n D 1,
d D 1/6 cm, s D 9999 cm1 and m D 3333.
i
()
function with a period of 2p. The sharpness of the fringe
I0 = ii
pattern depends on R. When the value of R approaches
Figure 1. The amplitude of successive transmissions and reflec- unity, the fringe becomes sharp. In order to obtain the high
tions in a plane parallel plate. t and t0 are Fresnel amplitude
reflectivity, we have to coat the etalon plates with a metal
coefficients of transmission at the individual plane surfaces, and
r and r 0 are those of reflection at the individual plane surfaces. or dielectric material. In a realistic situation, however, we
have to take into account the absorption of the incident
a D itt0 C itt0 r 02 expi C itt0 r 04 expi2 C light, which we have assumed to be zero. In addition, we
cannot neglect the phase shift in reflections at the etalon sur-
itt0 faces any more; the value of the phase shift f is between
D 1
1 expi r 02 0 and p. Equation (3) can be rewritten as follows:
where t and t0 are Fresnel transmission coefficients and D 4pnds cos q 2f 5
r and r 0 are Fresnel reflection coefficients, respectively.3
From the Stokes relation,3 r 2 C tt0 D 1 and r D r 0 . Let If we introduce the energy absorption coefficient A into
the energy transmittance be T D tt0 and let the energy equation (2), it can be rewritten as:
reflectance be R D r 2 D r 02 . Then, R C T D 1. If we put the 1 R A 2
intensity of the incident light on the FP etalon as I0 D ii , ID I0 6
1 R 2 C 4R sin2 /2
that of transmitted light I is given by
where R C T C A D 1. Accordingly, the maximum value
T2 Imax is decreased by a factor of t D 1 R A 2 /1 R 2 .
I D aa D I0 2
1 R 2 C 4R sin2 /2 However, the shape of the interference fringe pattern for a
given reflectivity R is not changed. The value of t is called
where denotes the complex conjugate. The quantity is
a transmittance of the FP etalon. For instance, t 36%
a phase difference introduced by a double passage of the
for R D 0.9 and A D 0.04 (Ag coating), and t 92% for
light between the FP etalon plates, and is given by
R D 0.9 and A D 0.004 (multilayer coating). However, it
D 4pnd cos q/l D 4pnds cos q 3 should be borne in mind that such coatings are effective
only for a particular spectral range. It should be noted that
The intensity transmittance I/I0 is known as the Airy the peak transmittance of the FP etalon is 100% if A D 0.
distribution which represents the instrument function of the This means that the peak transmission intensity is equal to
FP etalon. The maximum value Imax is obtained when the incident one when A D 0.
D 4pnds cos q D 2pm 4
and then Imax D T2 I0 /1 R 2 , where m is an integer. The
3 ANGULAR DISPERSION AND THE
minimum value Imin D T2 I0 /1 C R 2 is obtained when D INTERVAL OF THE INTERFERENCE
2pm C 1/2 . The contrast C of the interference fringe pat- ORDER
tern is, therefore, given by C D Imax /Imin D 1 C R 2 /1
R 2 . Figure 2 shows the Airy distribution as a function of Equation (4) gives a condition for forming a bright inter-
with R as a parameter. The distribution is a comb-like ference fringe pattern. When an extended light source as
FabryPerot Interferometers for Mid-infrared Spectroscopy 3
Diffuse source Collimating lens d Focusing lens Screen
1 n 1
L1 L2
Figure 3. FP etalon and equal-inclination interference fringes.
shown in Figure 3 is used, the pattern consists of many For constant values of l, d, and n, we can obtain the
concentric circles. For example, many rays originating from following relation:
individual points on a circle including a point P on a diffuse
source are incident on the FP etalon at the same angle, q, dq 1
D 8
on the image plane. Other concentric fringes correspond to dm 2f
mC tan q
different incident angles, namely circles of different diam- p
eter on the diffuse source. Because the interference pattern
is originally localized at infinity, a focusing lens L2 placed This equation indicates that the distribution density of the
after the etalon brings back the pattern on to its focal plane. circular fringe pattern is high for outer circles and that
This kind of interference pattern is called equal-inclination the tendency is independent of the wavelength. Although
interference fringes or Haidinger fringes. the interference order m is largest at the center of the circle
If the light source emits two wavelength components and a large m makes the value of equation (8) small, a
of l and l C l with the order m and the spacing d small change in m has almost no effect on equation (8) as
of the FP etalon fixed, two fringe positions correspond- the value of m is very large.
ing to the individual wavelength components on the screen
should be different with respect to each other. Further-
more, we can understand from equation (4) that the two 4 FREE SPECTRAL RANGE AND
wavelength components must be emitted from different
positions on the extended light source. This implies that FINESSE
the wavelength dispersion of the FPI is introduced by
the variation of the spatial position of the light source Figure 4 shows a plot of equation (2) as a function of
or by the change of the incident angle q. This disper- wavenumber s with d D 1/6 cm, n D 1, and R D 0.8. The
sion mechanism of the FPI is quite different from that of distribution is a periodic cone-like function where each
the diffraction grating and that of the prism. The angu- interference order corresponds to the specified wavenum-
lar dispersion D can be calculated from equation (4) as ber. The period is shown as FSR and the full width at half
follows: maximum (FWHM) at the center wavenumber as W. The
2f FSR represents one order of interference. By differentiating
dq mC
D D D p D 1 7 equation (4) by m, we can obtain 4pnd cos qds D 2pdm. By
dl 2nd sin q l tan q substituting dm D 1,
The dependence of D on l, d, and q is shown in equa- 1

tion (7). FSR D ds D 9
2nd cos q
FSR the light loss in the FPI; which are called surface finesse
1.0 FD , aperture finesse FA , and absorption finesse, respec-
tively. Because the absorption loss can be made relatively
small, we consider only FD and FA . The FD and FA are
defined as:
M 4l
0.5 W FD D and FA D 13
2 da2
where M represents a wavefront error when the parallelism
of the two plate surfaces of the etalon is given by l/M and a
is a solid angle subtended by the aperture of the FPI at the
9999.0 10002.0 collimating lens. When distributions of individual finesse
(cm1) are Gaussian, the total finesse F can be expressed as:
Figure 4. FP fringes as a function of s. FSR, free spectral range. 1 1 1 1
2
D 2 C 2 C 2 14
F FR FD FA
W can be derived from equation (2) as follows:
Usually, we can make the value of FA relatively large
1 R
W D p FSR 10 regardless of the structure of FPI by using a long focal
p R length lens and a large aperture. However, the values of
A ratio of FSR to W gives a measure of fine-ness FR and FD have some upper limits. Because making a
of interference fringes and therefore it is called finesse FR ; high precision surface is usually more difficult than making
which can be interpreted as a measure of resolvable ele- a high reflective coating, FR > FD . Since the value of FD
ments in one FSR. The subscript R denotes the reflectivity is 50 at most, this value determines the total finesse F. For
finesse. p this reason, FD is called the limiting finesse. The limiting
FSR p R finesse can be relaxed to some extent in the IR region rather
FR D 11
W 1R than in the visible region because of the longer wavelength.
In this expression, it is pertinent to consider that W
is approximately equal to the resolution limit of the FP
etalon in wavenumber. In fact, W is nearly equal to 5 OPTICAL THROUGHPUT AND
the resolution criterion of Rayleigh which is defined for THEORETICAL RESOLUTION
a diffraction-limited instrument function. As a matter of
fact, while the Rayleigh criterion gives a 19% central Generally, the performance of the spectrometer should be
dip, the W gives a 17% dip.2 The resolving power < represented by not only the spectral resolution but also
in wavenumber (or wavelength) of FPI, therefore, can be brightness or optical throughput. The optical throughput L
defined as follows: of the spectrometer is given by a product of the intensity
s s FSR transmittance t and etendue U, i.e. L D tU. The etendue
< D D mFR 12
W FSR W is an invariant quantity in geometric optics and is defined
From this definition, we can understand that the finesse FR as a product of the effective area S of the collimating
corresponds to the total number of grooves in a diffrac- lens and the solid angle subtended by the FPI aperture
tion grating and that it can be interpreted as an effective at the collimating lens: U D S. Here, there is a trade-
number of light beams contributing to interference. For off between and < : < D 2p. From these relations,
example, the resolving power of the typical diffraction- L D 2ptS/<. In the case of the plane diffraction grating
grating monochromator is in the range of 104 105 for monochromator, the solid angle is given by D wh/f2 ,
Ntotal number of grooves D 104 105 and m D 1. It is where w is the slit width, h is the slit height, and f is
not very difficult for the FPI to attain < D 5 106 with the focal length of the collimator lens in the monochro-
m D 105 and FR D 50. The relation between R and < can mator. Because w D dq/ds fs/<, D dq/ds hs/< and
be derived from equations (11) and (12). If R approaches L D tS D dq/ds tShs/f<. Comparing the L value of
unity, FR and < become large. For instance, FR D 61.2 the FPI (LFPI ) and that of the typical grating monochro-
when R D 0.95 and FR D 312.6 when R D 0.99. mator (LG ) under the same spectral resolution, we can
In a practical situation, however, the value of the reflec- obtain LFPI > 100pLG when sdq/ds D 1.0 and h/f D
tive finesse FR is limited by other factors such as imper- 1/50. This implies that the FPI is a superior spectrometer
fections on the plate surfaces or a deviation of plates from for spectral resolution and optical throughput compared to
parallelism, a finite solid angle of the light collection, and the conventional diffraction grating monochromator.
Here, the general relation < D 2p can be derived deformation of the piezoelectric crystal when applying
as follows: When wavelength l has a maximum at the a stepwise voltage, wavelength scanning is performed.
center (q D 0) satisfying ml D 2d, wavelength l l The use of the piezoelectric device is promising for the
forms a ring of an angular radius q satisfying ml following reasons: simplicity and ease of operation; high
l D 2d cos q 2d1 q2 /2 . Therefore, ml D dq2 . In precision; compactness in instrumentation; and possibility
order to resolve the wavelength l from l l, we have in high speed scanning.
to exclude this ring from the aperture. This limits the
angular aperture. Its solid angle is given by D pq2 D
pml/d D 2pl/l D 2p/<. Therefore, < D 2p. 7 EXTENSION OF THE SPECTRAL
RANGE
6 WAVELENGTH SCANNING As mentioned earlier, the FPS has a major problem in

that the spectral range is limited to within the FSR in
In order to scan the wavelength of the FPS, it is necessary spite of the extremely high spectral resolution and high
to vary one of three variables in equation (4), incident angle optical throughput. For this reason, we have to insert an
q, refractive index n of the FPI medium, or spacing d of order sorter before or after the FPI so that the undesired
the FPI plates, while keeping the other two fixed. spectra outside the FSR are excluded. As the order sorter,
In order to vary the incident angle, we have to rotate the dispersion elements such as a narrow bandpass interference
FPI itself and it is somewhat difficult. We, therefore, have to filter, a prism, a diffraction grating, and the FP etalon can
insert a ring aperture on the focal plane of the focusing lens be employed. The effective use of the order sorter brings
in such a manner that the partial light flux corresponding about the wide spectral range.
to the angular difference between q and q C q only passes However, there is another problem with the FPI being
through the aperture. However, this method is rarely used in used as a spectrometer; it is the nonzero minimum inten-
a practical instrument for the following reasons: complexity sity problem. In the Airy distribution that was shown in
in operation; the largest angular-dispersion area around Figure 3, the light intensity between individual transmitted
q D 0 cannot be utilized; a light detection position on peaks cannot become completely zero. This phenomenon
the detector should be changed when the ring aperture is tends to mask weak lines with strong neighbors. Absorption
moved; and the optical throughput advantage of FPI cannot lines in a continuum background are to some extent filled
be utilized completely because only part of the interference in by the minima from the surrounding continuum. For this
pattern is detected. Although a multichannel ring detector reason, the single FPS is not always ideal for absorption
has been designed8 in order to cope with such problems, it spectroscopy. However, some methods for alleviating the
is not widely used because of cost. problem and for extending the spectral range have been
In order to vary the refractive index of the FPI medium, developed.
all of the FPI should be put into a closed container with win-
dows and the pressure of the inner gas should be changed
continuously or discretely. Then, a circular aperture should
8 SYNCHRONIZED SCANNING OF THE
be inserted on the focal plane of the focusing lens so that GRATING MONOCHROMATOR AND
unnecessary interference orders except the center one are THE FPS
screened. In order to record a spectrum, the output light
intensity of the FPI is plotted as a function of the gas pres- Helbert et al.9 have developed a synchronous wavelength
sure. This second method gives high precision because there scanning method between the grating monochromator and
is no mechanically moving part. However, the gas pressure the FPS. In this arrangement, the output light from the FPI
has to be changed from 0 to 30 atm for a wavelength range is focused onto the entrance slit of the grating monochro-
of 0.110 m when using air as the inner gas. mator. A schematic drawing of the principle of operation
The third method, varying the spacing d of the FPI plates, is shown in Figure 5. Here, WR denotes a spectral band-
is accomplished by mechanically moving one or both FPI width of the grating monochromator and W represents
planes while keeping the parallelism of the plates. The the Airy function. First, the wavenumber of the grating
appropriate size of the aperture should be inserted on the monochromator is set at sq , which corresponds to the
focal plane of the focusing lens in a manner similar to the sq1 th order of the FPS (Figure 5a). From this position,
second method. Although various mechanisms for changing the wavenumber of the grating monochromator is moved
the distance d have been devised, the use of a piezoelectric synchronously with that of the FPS (Figure 5b). When
crystal as an actuator is popular. By making use of the small the wavenumber arrives at the position of the (sq1 C x)th
WR
(a)
W
pq 1 pq+1x t q 1
t
(b) q 1
pq1x pq 1

(c)
t q 1 + Tq
pq1x pq 1

(d)
pq-1 t

(e)
tq
pqx pq
Figure 5. Principle of synchronized scanning of the FPI and the grating monochromator: WR , monochromator function; W, FP function.
[Reproduced by permission of Optical Society of America from J.M. Helbert, P. Laforie and P. Miche, Appl. Opt., 16, 21192126
(1977).]
order of the FPS in Figure 5(a), both scannings are stopped used in place of the grating monochromator in the previous
(Figure 5c). Then, while keeping the wavenumber of the synchronization method. The combined use of both instru-
grating monochromator at that position, the scanning of the ments alleviates the difficulty of the previous synchronized
FPS is returned to its original position (Figure 5d). From operation.10,11 In this optical configuration, the wavelength
these new wavelength positions of the FPS and the grat- dispersion along the slit height direction is determined by
ing monochromator, the synchronized scanning is started the FPI and that perpendicular to the slit height direction
again (Figure 5e). By a repetitive process of such oper- is determined by the diffraction grating of the spectro-
ations, a wide spectral range is covered maintaining the graph. Although this arrangement is simple and convenient,
high resolution. Helbert et al. have obtained a spectral res- there is a problem in identifying the overlapped spectral
olution of 0.05 cm1 for the whole visible spectral range, orders.
and have measured line spectra from a low-pressure mer-
cury lamp.
10 MULTIPLE FP ETALON
ARRANGEMENT
9 A COMBINATION OF THE
SPECTROGRAPH AND THE FPS Two or even more FP etalons are aligned in series in
order to obtain the wide spectral range.12 14 There are
In this combination, a spectrograph (or the polychroma- two configurations to arrange FSRs of the two FP etalons:
tor attached to a two-dimensional multichannel detector) is a multiple-integer tandem arrangement (Figure 6a) and a
FSR1 spacing etalon gives a wider FSR than the longer one.
Therefore, the transmitted maxima of the two etalons
coincide except that the longer etalon has many maxima
for every one of the shorter. In this way, a wide spectral
FSR2
range is obtained.
In the vernier tandem arrangement, two spacings of
the individual etalons are made slightly different with
respect to each other. Based on the vernier principle, this
arrangement gives a relatively wider spectral range than
(a) the multiple-integer tandem arrangement. The FSR of the
tandem system is given by the lowest common multiple
FSR1 (LCM) of the two FSRs (FSR1 and FSR2). The use of
multiple etalons in series has the following advantages: both
high optical throughput and high spectral resolution can be
obtained; and the nonzero minimum intensity problem in
FSR2 the Airy distribution function is alleviated to some extent.
This is because the instrument function of the tandem
system is given by the multiplication of the individual Airy
distribution functions.
Figure 7 shows the FPS system with the vernier tan-
dem arrangement originally developed by Sandercock and
(b) LCM of FSR1 and FSR2
colleagues.15,16 By using a six multipassing optical system,
Figure 6. Multiple etalon arrangement: (a) multiple-integer tan- spectral resolution of 103 cm1 was obtained for a range
dem arrangement; (b) vernier tandem arrangement. of 50 cm1 . They have measured Brillouin scattering of the
surface wave of silicon and the shear wave of polystyrene.
vernier tandem arrangement (Figure 6b).15,16 In the multi- In order to use this system, however, accurate adjustment of
ple-integer tandem arrangement, one spacing of the FP the optical system and high precision in optical components
etalon is set as an exact multiple of the other. The shorter are required.
Sandercock tandem
Fabry-Perot 6 pass Telescope
Oscilloscope Pinhole
Cryostat
OXF temperature
controller
Beam
splitter
Tandem ND
controller filter
Beam
splitter
Pinhole
Discriminator PMT
Argon-ion
laser
AT Shutter
computer controller
Figure 7. Schematic illustration of a six-pass tandem-FPI Brillouin-scattering apparatus. ND, neutral density; PMT, photomultiplier tube.
[Reproduced by permission of American Physical Society from G. Li, N. Tao, L. Van Hong, H.Z. Cummins, C. Dreyfus, M. Hebbache,
R.M. Pick and J. Vagner, Phys. Rev. B, 42, 44064425 (1990).]
11 USING THE FPI AS A center wavenumber n0 : FSR0 D kDI, where k does not
MULTICHANNEL WAVELENGTH need to be an integer. In such an optical configuration, FSR0
is divided by N>k equally, and the etalon spacing of the
SAMPLER FPI is changed N times with a step interval of FSR0 /N
in wavenumber. The procedures for i D 0 and i D 10
An effective and practical method has been proposed
i N 1 steps are schematically shown in Figure 8(ac)
recently for extending the spectral range.17 In this method,
the FPI and a second spectrometer are aligned in tandem and 8(df). Thereby we gather N spectra from the second
and the first FPI plays the role of a multichannel wave- spectrometer. Finally, a single high-resolution spectrum is
length sampler. As the second spectrometer, the following synthesized from the N spectra (Figure 8g). In principle,
three instruments can be used: a grating dispersion-type the spectral resolution is determined by the first FPI and
monochromator attached to a multichannel detector; a the spectral range is determined by the second spectrometer.
conventional FTS; or a multichannel FTS. Unlike the The advantages of the method are summarized as follows:
previous method using a combination of a FPS and a capability of high spectral resolution, which is not easily
spectrograph, the continuity of the spectrum is taken attained by the second conventional spectrometer alone;
into consideration. The principle of operation is shown applicability over a wide spectral range (it is impossible
in Figure 8. First, we adjust the FSR of the FPI to be to cover such a wide range by the FPI alone); capability
equal to k>1 times the sampling data interval DI (or of high signal-to-noise measurements (this benefit is due to
resolved wavenumber) of the second spectrometer at a the high optical throughput of the combined system); and
i=0 FSR0
(a)
DI S0
(b)
S0
(c)
0 FSR0 0 0 + FSR0

i=1
FSR1

(g) 01
1 = 0 +
(d)
DI = FSR0 /N
S1
(e)
S1
(f)
1 FSR1 1 1 + FSR1
Figure 8. Principle of operation of the combined system of the first FPI and the second spectrometer. (a) Intensity transmittance curve
of the FPI with FSR D FSR0 , where i D 0; (b) an output spectrum obtained from the second spectrometer, where DI represents the
sampling data interval and S0 indicates the area shown in gray; (c) a spectrum constructed by using wavenumber information in (a)
and intensity information in (b); (d) intensity transmittance curve of the FPI with FSR D FSR1 , where i D 1, whose peak positions are
shifted by n from those in (a); (e) an output spectrum obtained from the second spectrometer, where S1 indicates the area shown in
gray; (f) a spectrum constructed by using wavenumber information in (d) and intensity information in (e); (g) a spectrum synthesized
from that in (c) and that in (f). [Reproduced by permission of Society for Applied Spectroscopy from T. Iwata and J. Koshoubu, Appl.
Spectrosc., 52, 10081013 (1998).]
ease ofs construction involving moderate cost and simple qualitatively as follows: the FP interferogram includes OPD
operation. information up to infinity because of the multiple reflec-
tions occurring in the etalon medium. Although the actual
instrument does not give an infinite OPD because of the
12 MULTIPLEX FP SPECTROMETER attenuation of the intensity of the reflected light, the OPD
A method has been developed to obtain a high-resolution obtainable is still longer than that of the conventional FTS.
spectrum for a wide range by using a single FPI.18 22 In By Fourier transformation of the FP interferogram, many
this method, an FP interferogram is measured first and then harmonics of the fundamental wavenumber are obtained
it is inverse-transformed. The FP interferogram is obtained as shown in Figure 9(b).21 The higher harmonics are pro-
as a function of the spacing gap of the FP etalon as shown duced by interfering beams with the longer OPDs propor-
in Figure 9(a).21 The procedure is, therefore, very simi- tional to the number of multiple reflections. For the higher
lar to the conventional FTS. In comparison with the FTS, harmonics, the effect of convolution with the instrument
however, the optical path difference (OPD) required for function decreases. In conventional FTS, the convolution
attaining a given spectral resolution can be reduced by a effect (or the truncation of the interferogram with a finite
factor of 10. The reason why the FP interferogram inverse OPD) determines the spectral resolution. This is why such a
method gives such a high resolution can be explained high spectral resolution is obtained. This kind of instrument
is called a multiplex FPS.
Figure 10 shows a solar absorption mid-infrared
spectrum measured by the multiplex FPS,22 with spectral
Detected signal
0.000 0.003 0.006 0.009 0.012 0.015 0.018

(a) Optical path length (cm)
Signal intensity (a.u.)
n =1
2550 2575 2600 2625 2650 2675
(a)
Relative intensity
n =2
n =3
n =4
n =5
n =6 n =7
0 4000 8000 12 000 16 000 20 000

(b) Wavenumber /cm 1
Figure 9. (a) Calculated multiplex FPI interferogram produced
by a Gaussian-shaped input spectrum with center wavenumber
2600 cm1 and HWHM D 30 cm1 . The plate reflectivity is 0.9. 2550 2575 2600 2625 2650 2675
(b) Portion of calculated spectrum derived from Fourier transform (b) Wavenumber /cm1
of the multiplex FPI interferogram shown in (a). The interfero-
gram is truncated at x D 0.0075 cm. Note that the two lowest Figure 10. Solar absorption spectra: (a) first and (b) fifth har-
harmonics are not well resolved, but the higher harmonics are monics after all necessary corrections have been applied. The
clearly Gaussian. [Reproduced by permission of Optical Society resolution is (a) 0.226 and (b) 0.0045 cm1 . [Reproduced by
of America from W.B. Cook, H.E. Snell and P.B. Hays, Appl. permission of Optical Society of America from H.E. Snell,
Opt., 34, 52635267 (1995).] W.B. Cook and P.B. Hays, Appl. Opt., 34, 52685277 (1995).]
resolution 0.045 cm1 . The multiplex FPS, however, still REFERENCES

has some problems to be overcome. Requirements in pre-
cision for the driving mechanics and for optics are very 1. G. Hernandez, FabryPerot Interferometers, Cambridge
severe because the OPD should be varied from zero to University Press, Cambridge (1986).
some lengths while keeping the parallelism of the FP plates. 2. A.P. Thorne, Spectrophysics, Chapman & Hall, London
In addition, to realize zero OPD is physically impossible, (1988).
which implies that an elegant data processing technique 3. M. Born and E. Wolf, Principle of Optics, Pergmon Press,
should be developed for recovering the original spectrum Oxford (1975).
and for correcting the phase error in the FP interferogram. 4. H. Blancher, G. Bachet, R. Conlon and D. Aubert, Int. J.
However, a multiplex FPS can be made much smaller than Infrared Millimeter Waves, 6, 53 (1985).
a conventional FTS producing the same spectral resolution 5. D.A. Naylor and I. Furniss, Infrared Phys., 29, 159 (1989).
and covering a much broader spectral range. Because it is 6. W. Schatz, P.T. Lang, T. Kass, M.A. Heusinger and K.F.
light and compact, it is suitable for use in a satellite or for Renk, Int. J. Infrared Millimeter Waves, 13, 853 (1992).
real-time pollution monitoring in the mid-infrared region. 7. K.A. Christensen, N.L. Bradley, M.D. Morris and R.V.
Morrison, Appl. Spectrosc., 49, 1120 (1995).
8. P.B. Hays and J. Wang, Appl. Opt., 30, 31003107 (1991).
13 CONFOCAL FPI 9. J.M. Helbert, P. Laforie and P. Miche, Appl. Opt., 16,
2119 (1977).
Today, a confocal FPI is more popular than a plane FPI in 10. D.T. Speer, S. Von Laven, A.H. Karp and M. Stockton, Appl.
the visible region. The confocal FPI gives higher spectral Opt., 19, 2757 (1980).
resolution than the plane FPI, usually < 108 . It consists 11. G. Poulter and R.E. Jennings, Infrared Phys., 23, 4350
of two concave spherical mirrors separated by a distance (1983).
nearly equal to their radii of curvature. The focal points 12. A. Poglitsch, J.W. Beeman, N. Geis, R. Genzel, M. Haggerty,
of the two mirrors coincide on the optical axis. Because E.E. Haller, J. Jackson, M. Rumitz, G.J. Stacey and C.H.
concave mirrors are used, the surface finesse is not defined, Townes, Int. J. Infrared Millimeter Waves, 12, 859 (1991).
and so high resolution is achieved. Although the confocal 13. S. Gehrsitz, H. Sigg, H. Siegwart, M. Krieger, C. Heine,
design is easy to adjust, it has difficulty in wavelength scan- R. Morf, F.K. Reinhart, W. Martin and H. Rudigier, Appl.
ning and is therefore rarely used for spectral measurements. Opt., 36, 5355 (1997).
In order to change the spacing of the confocal mirrors, the 14. P.B. Hays, V.J. Abreu, M.E. Dobbs, D.A. Gell, H.J. Grassl
two mirrors must be exchanged with mirrors of different and W.R. Skinner, J. Geophys. Res., 98, 10 713 (1993).
curvatures. 15. S.M. Lindsay, M.W. Anderson and J.R. Sandercock, Rev.
Sci. Instrum., 52, 1478 (1981).
16. G. Li, N. Tao, L. Van Hong, H.Z. Cummins, C. Dreyfus,
ABBREVIATIONS AND ACRONYMS M. Hebbache, R.M. Pick and J. Vagner, Phys. Rev. B, 42,
4406 (1990).
FP FabryPerot 17. T. Iwata and J. Koshoubu, Appl. Spectrosc., 52, 1008
FPI FabryPerot Interferometer (1998).
FPS FabryPerot Spectrometer 18. Y. Yoshihara, A. Kitade and K. Okada, Jpn. J. Appl. Phys.,
FSR Free Spectral Range 19, 2523 (1980).
FTS Fourier Transform Spectrometer 19. T. Aoki, Appl. Opt., 29, 2364 (1990).
FWHM Full Width at Half Maximum 20. P.B. Hays and H.E. Snell, Appl. Opt., 30, 3108 (1991).
LCM Lowest Common Multiple 21. W.B. Cook, H.E. Snell and P.B. Hays, Appl. Opt., 34,
ND Neutral Density 5263 (1995).
OPD Optical Path Difference 22. H.E. Snell, W.B. Cook and P.B. Hays, Appl. Opt., 34,
PMT Photomultiplier Tube 5268 (1995).
Instrumentation for Far-infrared Spectroscopy
Peter R. Griffiths1 and Christopher C. Homes2

1
2
Brookhaven National Laboratory, Upton, NY, USA
1 SOURCES FOR FAR-INFRARED noted that microwave sources have been developed that
SPECTROMETRY emit highly monochromatic radiation at shorter wavelengths
than 1 mm (10 cm1 ) but they will not be covered in this
article. Because wavelengths between 1 mm and 100 m
Measurements of far-infrared (FIR) spectra are made far require a different measurement technology than the MIR,
more difficult than the corresponding measurement of mid- measurements made in this region are sometimes called
infrared (MIR) or near-infrared (NIR) spectra because of submillimeter wave spectroscopy.
the weakness of the source. The spectral energy density The radiance of synchrotron sources in the FIR can be
of a blackbody at 1400 K is almost 5800 times weaker at as much as three orders of magnitude greater than any
10 cm1 , and about 60 times weaker at 100 cm1 , than it of the sources mentioned above.1 From the standpoint of
is at 1000 cm1 . Since the sensitivity of any measurement obtaining the highest photon flux, therefore, synchrotrons
of a MIR or FIR spectrum is directly proportional to the are easily the best sources of FIR radiation. They have
spectral energy density of the source, it is apparent that FIR two important disadvantages, however, in that beam time
sources should be as hot as possible and have an emittance must be obtained and the experiment must be transported to
of close to unity. the synchrotron and not vice versa. If these factors do not
For measurements between about 400 and 100 cm1 , provide an excessive limitation, the synchrotron is certainly
the radiant power emitted by a silicon carbide (Globar ) the ideal source of FIR radiation, particularly for etendue-
source is as high as any conventional infrared source. limited applications such as reflectionabsorption infrared
Although Nernst glowers can be operated at higher tem- spectroscopy of thin films.
perature than Globars, they become quite transparent below
about 200 cm1 , so that their emissivity drops to the point
that they are of little use as FIR sources, despite their 2 DETECTORS
high temperature. Below 100 cm1 , the emissivity of a
Globar also becomes low and it is customary to use
2.1 Pneumatic amplifiers
a high-pressure mercury lamp for measurements between
50 cm1 and the onset of the microwave region of the
The first type of detector to be routinely used for FIR
spectrum. The reason why mercury lamps have proved to
spectroscopy using both grating monochromators and slow-
be so successful for FIR spectrometry is because emission
or step-scanning interferometers was the Golay pneumatic
from the plasma reinforces the emission from the hot quartz
amplifier detector.2 A schematic of one of these detectors
envelope of the lamp. However, the mercury lamp has sig-
is shown in Figures 1 and 2. A chamber containing a gas
nificant flicker noise, so even if the energy is better, the
of low thermal conductivity is sealed at one end with an
signal-to-noise ratio (S/N) may not be. It should also be
infrared transparent window (A) through which radiation
reaches a thin absorbing film (B). This film has a low
John Wiley & Sons Ltd, 2002. thermal capacity, so that its temperature rises rapidly, in
Pneumatic Ballasting The temperature of the absorbing film follows the fre-
chamber reservoir
Leak quency at which the incident radiation is modulated. Visible
light from a source (E) passes through a condensing lens
Window (F) to a line grid (E) and is focused on the flexible mirror.
D
A A meniscus lens (H) between the line grid and the mir-
ror focuses the beam so that, in the absence of any other
deformation, an image of one part of the line grid is super-
imposed on another part of the same grid. If the image of
a gap between lines coincides with a gap in the grid, light
Incident will be transmitted by the grid and hence the intensity of
radiation C
Flexible light reaching the sensor is measured.
mirror Golay detectors are remarkably sensitive and were com-
monly used with slow- and step-scanning interferometers
and lamellar grating interferometers (vide infra). These
instruments do not modulate the infrared beam at a fre-
B
quency above about 0.1 Hz, so that the beam is usually
Absorbing further modulated by a rotating chopper typically operating
film
at 13 Hz. When the older interferometers were superseded
Figure 1. Front section of a Golay detector. Chopped incident by the type of rapid-scanning interferometers incorporated
radiation passes through the window onto a blackened film,
causing the pressure of the gas in the pneumatic chamber to
in bench-top Fourier transform infrared (FT-IR) spectrome-
fluctuate and a flexible mirror to deform with the same modulation ters, the slow response speed of Golay detectors precluded
frequency as the chopper. A pinhole leak minimizes the effect of their use and they were, in turn, superseded by pyroelectric
long-term temperature variation. (Reproduced by permission of bolometers.
Cathodeon Ltd.)
turn warming the gas with which it is in contact. A rise 2.2 Pyroelectric bolometers
in the temperature of the gas in the chamber leads to
a concomitant increase in its pressure and, therefore, a Pyroelectric bolometers incorporate as their heat-sensing
distortion of the mirror membrane (C) with which the other element ferroelectric materials that exhibit a large spon-
end of the chamber is sealed. A fine leak (D) connects the taneous electrical polarization at temperatures below their
detection chamber with a ballasting reservoir of gas on the Curie point. If the temperature of these materials is changed,
other side of the mirror membrane to prevent slow changes the degree of polarization is changed. The change in polar-
in ambient temperature from affecting the detector. ization may be observed as an electrical signal if electrodes
Head Mirror and lens block

Grating
Light
source
Mirror L
H
F
G K
P-FET
light
To amplifier sensor
Figure 2. Rear section of a Golay detector. Visible radiation is passed through a grating and is focused onto the flexible mirror shown
in Figure 1. The reflected radiation passes back through the grating and is measured by a light sensor. When the flexible mirror is
undistorted, the grating prevents any light from reaching the detector. The greater the distortion of the mirror, the greater is the measured
signal. (Reproduced by permission of Cathodeon Ltd.)
Instrumentation for Far-infrared Spectroscopy 3
are placed on opposite faces of a thin slice of material is the operating temperature. The noise equivalent power
to form a capacitor. When the polarization changes, the (NEP) of these bolometers is 1.5 1013 W Hz1/2 . To
charge induced on the electrodes can either flow as a cur- minimize the effect of the MIR photon flux from ambient
rent through a relatively low external impedance or produce grey bodies, LHe-cooled filters must be placed in front of
a voltage across the slice if the external impedance is rel- these detectors. For increased sensitivity, these detectors
atively high. The detector will only give a signal when the may be operated at 1.5 K by pumping the helium bath
temperature of the element changes. through the lambda point. In this case, the NEP may
The most commonly used material for pyroelectric detec- be reduced as low as 8 1014 W Hz1/2 . The lambda
tors is deuterated triglycine sulfate (DTGS). The DTGS point of liquid 4 He is reached by pumping on the LHe
element is usually mounted so that the thermal resistance reservoir. At a pressure of 37.8 mm Hg, the LHe undergoes
between the element and its surroundings is large, so that a superfluid transition at 2.172 K. Continued pumping of
the thermal time constant is long. The voltage responsivity the superfluid down to a pressure of about 1 mm Hg lowers
of pyroelectric detectors is approximately proportional to the temperature of the fluid to 1.27 K.
the reciprocal of the frequency at which the incident radi- A further increase in sensitivity can be achieved using
ation is modulated. The frequency at which radiation of pumped liquid 3 He, for which the lowest operating tem-
wavenumber nQ is modulated by a rapid-scanning interfer- perature is 0.32 K. A pumped liquid 3 He bolometer also
ometer is equal to VnQ Hz, where V is the optical velocity has a lower bandwidth (80 cm1 as opposed to 700 cm1
(in cm s1 ). For an optical velocity of 0.316 cm s1 (5 kHz for a 4.2 K detector). When these two factors are taken
laser frequency), the highest modulation frequency of FIR in tandem, the reduction in Vn is an order of magnitude,
radiation (corresponding to 400 cm1 ) is 125 Hz, and the and a further restriction of bandwidth can improve this
lowest (corresponding to 10 cm1 ) is 3 Hz. In principle, figure even more. The 3 He bolometer has a typical NEP of
therefore, the performance of DTGS detectors fitted with 1.4 1015 W Hz1/2 , and has been used down to 1 cm1
a polyethylene window should be quite good for measure- using a mercury arc lamp source and a lamellar grating
ments of FIR spectra using a rapid-scanning interferometer. interferometer.
However, it should be noted that this frequency range The NEP of these bolometers may be compared to the
includes electrical line frequency (60 Hz in the USA) and its typical value for a DTGS detector of 5 109 W Hz1/2 .
first harmonic, and so spikes are often seen in FIR spectra Thus the typical increase in sensitivity of bolometers cooled
measured with a rapid-scanning interferometer. In addi- to 4.2 K over that of the DTGS detector at ambient temper-
tion, low-frequency building vibrations can lead to spikes ature is more than four orders of magnitude. This improve-
(glitches) in the spectra at other frequencies, so that good ment factor may be increased to six orders of magnitude
vibration isolation is often necessary. by cooling with pumped 3 He.
The second type of FIR detector is the InSb hot-electron
2.3 Liquid helium cooled bolometers detector. Absorption of radiation by free-carrier electrons
causes their mean temperature Te to rise above that of
Because the intensity of FIR radiation emitted by most the host lattice. Since electron mobility is proportional to
sources is very low, spectra measured using pyroelectric T3/2
e , this temperature rise can be sensed as a change in
or Golay detectors are often quite noisy. To improve the conductivity. The very low thermal mass of free electrons,
S/N, it may be necessary to use a detector with much higher with short energy relaxation times of 107 s, leads to very
performance than either of these two detectors. In this case, fast submillimeter and millimeter wave detectors. In one
the use of liquid helium (LHe) to cool the element is often modification of this type of detector, a specially-shaped
called for. n-type InSb crystal is coupled directly to a low-noise
Two types of LHe-cooled detectors have been used for preamplifier without sacrificing bandwidth or introducing
the measurement of FIR radiation. The first are either noise. The typical wavenumber response of these detectors
boron-doped silicon or germanium doped with copper, gal- is from 2 to 50 cm1 .
lium or antimony. When cooled with liquid 4 He, these
detectors have a low frequency cut-off of 20 cm1 . The
fundamental factor limiting the sensitivity of these detectors 3 INTERFEROMETERS
is Johnson noise, the power of which is given by equa-
3.1 The early days
tion (1):
Vn D 4kB nQ RT 1
In the late 1950s and early 1960s, two manufacturers
where kB is Boltzmanns constant, nQ is the bandwidth of introduced double-beam grating spectrometers for FIR
the detector, R is the resistance of the bolometer and T spectrometry. These were the Perkin-Elmer (Norwalk, CT,
USA) Model 301 and the Beckman (Irvine, CA, USA) equipped with a Moire fringe reference device that actuated
IR-11. Although these instruments allowed spectra to be a punched paper tape at equal intervals of path difference
measured down to 33 cm1 , the spectra were noisy, espe- and the spectrum was computed off-line. As crude as this
cially when high resolution was required. At about the instrument sounds, its sensitivity was remarkably high, in
same time, John Strong, at Johns Hopkins University in part because of its very high optical throughput.
Baltimore showed that it was possible to measure signif- By about 1965, the NPL group had switched to a much
icantly better FIR spectra with an interferometer provided more compact and easily fabricated design that became
that the long computing times required could be tolerated. the basis of the Grubb-Parsons Mark II interferometer, a
(Remember that the fast Fourier transform algorithm was photograph of which can be seen in the first article of
not reported until 1964 and that the power of the best this volume. This interferometer, which was an almost
computers at that time was not much better than that of exact replica of the NPL cube, operated in the step-and-
a small chip that played the Christmas carol God Rest integrate (step-scan) mode in which the moving mirror was
Ye Merry, Gentlemen in Christmas cards sold in the mid driven by a precise stepping motor so that there was no
1990s.) need for any fringe referencing. It allowed spectra to be
After Strongs work, the key developments of Michel- measured at resolutions up to 0.1 cm1 . It may be noted
son interferometers for FIR spectroscopy was carried out that efficient vibration isolation was needed to obtain the
in the 1960s by Alastair Gebbies group at the National best performance of this instrument.
Physical Laboratory (NPL) in Teddington, Middlesex, UK. These two instruments were so successful that they
Their instruments were quite small and, at least by todays spawned a number of other slow-scan or step-scan inter-
standards, crude, but allowed FIR spectra of remarkably ferometers for FIR spectrometry made by Beckman-RIIC
high quality to be measured to below 10 cm1 . Fortu- in the UK, Coderg in France, and Polytec in Germany.
nately, because the tolerances required for interferometry By the mid-1970s, however, they started to be superseded
are proportional to the wavelength, it was a relatively sim- by rapid-scanning interferometers that had started to gain
ple matter to make interferometers for FIR measurements popularity for MIR measurements. Since these instruments
where the wavelengths could be as long as 1 mm. The are identical to most rapid-scanning interferometers used
first commercial FIR Fourier transform spectrometers were for MIR spectrometry, and these instruments are described
based on the interferometers developed at the NPL. These in the article by Jackson3 in this Handbook, we will not
instruments were made by Research and Industrial Instru- consider their design here.
ments Co. (RIIC) (Strathclyde, UK) and Grubb-Parsons
(Newcastle-upon-Tyne, UK). RIIC was taken over by Beck-
man in the early 1970s. 3.2 Beamsplitters for far-infrared
The first of these FIR interferometers, the RIIC FS-520, interferometers
was introduced in 1964. As with most FIR spectrome-
ters, the entire optical path was evacuated. The FS-520 The biggest drawback to the use of Michelson interferome-
incorporated a continuous drive, where the optical path dif- ters for FIR spectrometry was the low efficiency of Mylar
ference could be varied at a rate between about 0.5 and beamsplitters. The modulation efficiency of the interferom-
500 m s1 , with 5 m s1 being a typical value for mea- eter, hnQ , is given by 4 R0 (nQ ) T0 nQ , where R0 nQ and T0 nQ ,
surements between 100 and 400 cm1 . At this slow speed, are the reflectance and transmittance of the beamsplitter at
the Fourier frequencies were well below 0.1 Hz so that the wavenumber nQ , respectively. The maximum efficiency is
radiation from the mercury lamp source had to be mod- found when R0 nQ D T0 nQ , D 0.5, in which case hnQ D 1.
ulated by a 13-Hz chopper. The throughput allowed for The sinusoidal variation of modulation efficiency with
this spectral region was very large and the FS-520 had wavenumber is known variously as interference fringes,
a poly(ethylene terephthalate) (PET) beamsplitter that was channel spectra, dielectric resonances and Fabry-Perot
over 20 cm in diameter. (The most common commercial fringes. If one considers the waves on a non-absorbing,
sources of PET are du Pont, who market it as Mylar , and parallel-sided sheet of a dielectric, the reflectance and
ICI, who marketed it as Melinex . Beamsplitters fabricated transmittance are given by equations (2) and (3):
from PET are commonly known in the USA as Mylar
beamsplitters and will be referred to as such throughout 2R2 1 cos
R0 D 2
the rest of this article.) After the beam passed through 1 C R2 2R cos
the sample compartment, the interference record was mea- and
sured with a Golay detector, and the detector signal was 1 R2
demodulated with a lock-in amplifier. The instrument was T0 D 3
1 C R2 2R cos
where D 4p nQ nt d cos qt is the relative phase shift (in 1.0

radians) between two adjacent emerging rays, d is the
0.8
thickness of the film, nt is its refractive index, and qt is
the angle of the beam inside the film to the surface normal. 0.6
4R 0T0
The beamsplitter efficiency has maxima at D 2m 1p
and minima at D 2mp, for m D 1, 2, 3, . . . . 0.4
Radiation polarized with its electric field polarized paral-
0.2
lel to the plane of incidence is denoted by p, while radiation 25 m 12.5 m 6 m 3 m
polarized with its electric field polarized perpendicular to
the plane of incidence is denoted by s. For any non-zero 0 200 400 600 800 1000
angle of incidence at the beamsplitter, the reflectances for Frequency / cm1
p- and s-polarized radiation have the different forms of Figure 3. Variation in the efficiency of 3-m, 6-m, 12.5-m and
equations (4) and (5): 25-m Mylar beamsplitters as a function of wavenumber for
s-polarized radiation.
tan2 qi qt
Rp D 4
tan2 qi C qt Table 2. Spectral range covered by Mylar
beamsplitters of different thicknesses.
and
sin2 qi qt
Rs D 5 Thickness (m) Spectral range (cm1 )
sin2 qi C qt
where qi and qt are the angles of incidence and transmission, 3 900250
6 450120
respectively, and are related by n D sin qt /sin qt . In a con- 12.5 22560
ventional Michelson interferometer, qi D 45 . The refrac- 25 10030
tive index of PET is about 1.6 and the index of absorp- 50 5015
tion k D 0 at most wavenumbers in the FIR. This yields 100 258
Rp D 0.013 and Rs D 0.115. At a maximum in the beam-
splitter response [equation (6)]:
long wavelengths (>200 m), where the efficiency of these

1R 2 beamsplitters is poor and blackbody sources are weak, mea-
T0 D 6 surements become quite difficult. Measurement of a FIR
1CR
spectrum over the full range from 400 to 10 cm1 using
The relative beamsplitter efficiencies of a PET beamsplit- Mylar beamsplitters requires five different beamsplitters,
ter for the two polarizations are summarized in Table 1. as shown in Table 2.
It can be seen that, because of the large difference in The beamsplitter efficiency may be improved if a mate-
efficiencies for the p and s polarizations, a Michelson inter- rial with a larger refractive index than Mylar is used. The
ferometer equipped with a PET beamsplitter is far more data in Figure 4 show the variation of R0 and T0 for both
effective for s-polarized than p-polarized radiation. s- and p-polarized radiation as a function of the refrac-
Figure 3 shows the variation in the calculated efficiency tive index of the material from which the beamsplitter is
of four common Mylar beamsplitters as a function of fabricated for incidence angles of 45 and 15 . The differ-
wavenumber for s-polarized radiation. To cover the entire ence between the values of R0 for s and p polarization (R0s
FIR region from 400 to 10 cm1 with a beamsplitter effi- and R0p , respectively) may be reduced not only by choos-
ciency greater than 0.5, it can be seen that several beam- ing a material with higher refractive index such as silicon
splitters must be used. The spectral range typically cov- (n 3.4) or germanium (n 4.0), but also by reducing
ered by Mylar beamsplitters is shown in Table 2. At the angle of incidence at the beamsplitter. The values of
Table 1. The beamsplitter efficiencies for p and
R0s and R0p are much more similar at 15 incidence, but to
s polarizations for a PET beamsplitter (nt D 1.6) design an interferometer for which the incidence angle is
for a conventional 45 Michelson interferometer. so small is not trivial.
Several manufacturers of FT-IR spectrometers have app-
Polarization R0 T0 4 R0 T 0 lied variants of this technology in their FIR spectrome-
ters. For example, ThermoNicolet Instrument Corporation
p 0.052 0.95 0.20
(Madison, WI, USA)4 have used solid silicon plates for
s 0.37 0.63 0.93
None 0.21 0.79 0.57 their FIR beamsplitters. (Thick Ge plates cannot be used as
FIR beamsplitters because of the strong phonon absorption
1.0
1.5 Silicon (2 mm)
i = 45
R0s 3 m Mylar
0.8 i = 15
12 m Mylar
Intensity (arb. units)

= 2 cm1
1.0
0.6
(ideal)
R0
0.4 R0p 0.5
0.2
0 200 400 600
Frequency / cm1
0.0 Figure 5. Comparison of single-beam spectra measured with a
1 2 3 4
3-m and 12-m Mylar film and a 2-mm thick silicon beam-
Mylar n Si splitter. Note that the silicon beamsplitter covers a much wider
Figure 4. Variation of R0 for s- and p-polarized radiation as a spectral range than either of the Mylar beamsplitters.
function of the refractive index of the material from which the
beamsplitter is fabricated at 45 and 15 incidence. Note the 6
2.0
much smaller difference between the values of R0s and R0p at
an incidence angle of 15 compared to 45 . 1.5
1.0
of germanium.) If the Si plate is several millimeters thick,
Intensity (arb. units)
0.5
then the beamsplitter response consists of many closely 4
spaced cycles (spacing 0.7 cm1 for a 2-mm piece of sil- 0.0
0 20 40 60 80 100
icon) that gives a single-beam background spectrum that is
quite unsuitable for spectroscopic applications. If the reso- Ge/Mylar
Mylar
lution required is less than this spacing, e.g. nQ 2 cm1 , = 2 cm1

2
then only the average beamsplitter response of 4R0 T0
0.73 is observed over essentially the entire wavenum-
ber range. This beamsplitter has excellent long-wavelength
response and has been used to obtain spectral information
down to 1 cm1 . Provided that the sides are highly par-
allel, this beamsplitter should operate continuously into 0 200 400 600
the MIR, with the exception of a strong phonon band at Frequency / cm1
620 cm1 . Except in those cases where very high resolu- Figure 6. Single-beam spectra measured with an uncoated 6-m
tion is required, a thick Si beamsplitter can replace all of Mylar beamsplitter and the same material coated with a thin layer
the Mylar beamsplitters, as demonstrated by the data in of germanium. (Inset) The single-beam spectrum measured with
the same source/detector combination, only with a filter cooled to
Figure 5. 4.2 K with a 100-cm1 cut-off in place.
If high resolution is required, however, the Si beamsplit-
ter should be only a few micrometers in thickness and it
is difficult to produce a free-standing Si film with highly a 6- m Mylar beamsplitter with a thin film of germanium
parallel surfaces that is so thin. To circumvent this prob- is seen by the data in Figure 6. The inset in this figure is
lem, Bruker Optics (Karlsruhe, Germany) have deposited for the same source/detector combination, only with a filter
a thin film of germanium onto a PET substrate. The result cooled to 4.2 K with a 100-cm1 cut-off in place. Cutting
is a multilayer dielectric germanium beamsplitter that has off the high frequency region of the spectrum shows the
all the advantages of the thick beamsplitter discussed pre- advantage of the coated beamsplitter at low frequency even
viously, but it can be used for measurements at higher more effectively.
resolution and, because of the very thin germanium coating, Using a similar rationale, Perkin-Elmer (Beaconsfield,
the absorption caused by phonons is minimal. The improved UK) have successfully vapor-deposited thin irregular metal
efficiency at long wavelength that is achieved after coating layers on a stretched polypropylene (PP) substrate for use as
a beamsplitter with wide wavelength coverage. Both metal Some of these problems were easily overcome. The turn-
mesh and metal layers deposited in a polka-dot pattern on around of contemporary interferometers is far smoother and
a dielectric substrate can make very efficient beamsplitters severe drum-heading is less commonly seen in contempo-
at long wavelengths. Metal layers deposited in a polka- rary interferometers. To a greater or lesser effect, however,
dot pattern are essentially the inverse of a metal mesh. drum-heading is still frequently seen even in contempo-
Either a metal mesh or the polka-dot require an optimized rary interferometers. The effect of vibrations and electrical
pattern for high efficiency. interference has been overcome by better mechanical and
electronic design. Evacuated instruments were designed and
built to preclude interference by water vapor. The perfor-
3.3 Rapid-scanning interferometers for mance was still rather poor, however, because of the low
far-infrared spectrometry efficiency of PET beamsplitters. If measurements are only
required down to 180 cm1 , beamsplitters made with an
Even though excellent spectra were measured using the analogous optical design to MIR beamsplitters can be used,
interferometers described in the previous section, the out- with CsI replacing the KBr substrate and compensator plate
standing performance of laser-referenced rapid scanning and a thicker layer of germanium being used. For mea-
interferometers for mid- and NIR spectrometry led instru- surements below 180 cm1 , however, either Mylar or
ment manufacturers to modify these instruments for FIR some of the special beamsplitters described in the previous
operation. At first, the only change made in the interfer- section, are required.
ometer when switching from MIR to FIR operation was
the substitution of a Mylar beamsplitter for the Ge/KBr
beamsplitter. At first, however, the performance of these 3.4 The Genzel interferometer
instruments was quite poor because of several effects. First,
as noted in the previous section, the efficiency of the Another innovative concept for FIR spectroscopy is being
Mylar beamsplitter is poor. Secondly, the Mylar film used by Bruker in their IFS113v spectrometer which incor-
(which is stretched taut but is still quite flexible) can be porates a Genzel interferometer. The design of this instru-
easily set into vibration when the moving mirror reaches ment is shown in Figure 7. Unlike a standard Michelson
the end of its scan and is rapidly decelerated, causing the interferometer where the light is collimated at the beam-
interferogram to be extremely unreproducible. (This effect splitter and is never focused until it emerges, in the Genzel
has been called drum-heading.) interferometer the light is focused onto the beamsplitter.
In principle, the optical bench of a FIR instrument can The beams that are transmitted and reflected by the beam-
be either evacuated or purged to eliminate water vapor. splitter pass to two collimating mirrors and hence to a
For many reasons, however, evacuated instruments are far double-sided moving mirror. As the mirror moves, the path
preferable to purged spectrometers for FIR operation. For in one arm increases and the path in the other arm decreases.
example, it is extremely difficult to remove all traces of The small size of the beamsplitter in the Genzel inter-
water vapor from the single-beam spectrum and the effect of ferometer allows several beamsplitters to be mounted on
residual rotational lines of water vapor are commonly seen a wheel and interchanged without breaking the vacuum.
in ratioed spectra, especially when they are measured at The reduction in size significantly reduces the amplitude
high resolution. Furthermore, in a purged system vibrations of the drum-head vibrations commonly associated with
caused by the incoming gas can create a real problem, large-diameter Mylar beamsplitters. Another feature of
leading to artifacts in the spectrum. Even low-frequency the Genzel interferometer is that the angle of incidence at
vibrations caused by the effect of people walking near the the beamsplitter (15 ) is much smaller than a standard
instrument and building vibrations can lead to interferences Michelson interferometer. This has the advantage of reduc-
below 200 cm1 , so that careful vibration and electrical ing the difference between the s and p polarizations (as
isolation is called for. Finally, another factor that degrades shown in Figure 4), allowing optical polarization studies
the performance of FIR spectrometers based on rapid- without having to reorient the sample.
scanning interferometers is interference by harmonics of For a displacement, x, of the moving mirror, an optical
the line frequency (60 Hz in the USA, 50 Hz in the UK). path difference of 4x is generated by the Genzel interfer-
When the velocity of the moving mirror is set to give a ometer. Although this has been claimed as an advantage
laser frequency of 5 kHz (the most common velocity for for the Genzel over the Michelson interferometer, any tilt
operation with a pyroelectric detector), the glitch caused introduced on moving the mirror leads to a doubling of the
by the 60-Hz interference is at 190 cm1 , which is right error in comparison to the error introduced in a standard
in the middle of the FIR region. Michelson interferometer with the same tilt. Thus we do
g
f
Figure 7. Genzel interferometer of the type incorporated in the Bruker IFS-113 FT-IR spectrometers: c, filter wheel; d, carousel on
which up to six beamsplitters may be mounted; e, moving two-sided mirror; f, reference interferometer; g, HeNe laser; h, spherical
collimating mirror.
not consider this to be a major advantage. Nonetheless, the type of interferometer for very high resolution FIR spec-
Genzel interferometer does have several significant advan- trometry is the cube-corner. The highest resolution mea-
tages for FIR spectroscopy. surements achieved with such an instrument were reported
by Kauppinen and Horneman5 using the interferometer
shown schematically in Figure 8. The cube corner retrore-
3.5 Interferometers for very high resolution flectors were fabricated from 15 15 cm flat mirrors and
spectrometry in the far-infrared the diameter of the beamsplitter was 31 cm. The maximum
optical path difference, max , obtained with this inter-
To obtain very high resolution measurements, some form of ferometer was 6.2 m. Since the full-width at half height
tilt compensation is required. To date, the most successful of a sinc instrument line shape function is 0.605/max ,
W2 M11
M5
M8 M7
W3I
G M6
W4 L1
M1
FP
M02 D L2
GB M01 SH
M13
W1 M12
B0 M10
B
M4 S
M2
M3
M9
Figure 8. Optical layout of the high-resolution cube-corner interferometer reported by Kauppinen and Horneman. [Reproduced from
Kauppinen and Horneman (1991) by permission of the Optical Society of America; copyright 1991.]
0.30 With a 125 m PET beamsplitter and a LHe-cooled

bolometer, the instrument can be used down to 5 cm1
0.25
(although measurements at the limit of the wavelength
0.20 range are not necessarily practical at the highest resolu-
tions). A detailed description of the design of the interfer-
Absorbance
0.15 ometer used in this instrument can be found in the article by

0.00095 cm1
Jackson3 in this Handbook. It is a Michelson type that uti-
0.10
lizes cube-corner retroreflectors, and has a clear aperture of
0.05 more than 6 cm. To reduce polarization effects, and increase
the effective area of the beamsplitter, the angle of incidence,
0.00 qi , is 30 . Because of the very long optical paths, to avoid
0.05
problems with water vapor in high resolution measurements
61.425 61.420 61.415 the instrument can be evacuated to 2 103 Torr.
Wavenumber / cm1
Figure 9. Line in the pure rotation spectrum of CO measured
without apodization at the maximum resolution of the Bruker 3.6 The lamellar grating interferometer
HR-002 spectrometer; the full-width at half height of this line is
0.00095 cm1 . (Data courtesy of Dr R.S. Jackson, Bruker Optics.) Because of the low reflectance of Mylar beamsplitters with
p-polarized radiation, the modulation efficiency of Mylar
the resolution obtained with this instrument should be
beamsplitters to unpolarized radiation never exceeds 50%,
approximately 0.001 cm1 . Spectra could be measured at
even at their optimum wavelength. For measurements
a resolution of 0.002 cm1 at 666 cm1 using a Golay
detector. When equal or higher resolution at wavenumbers below 50 cm1 , where the source intensity is extremely low,
below 100 cm1 was desired, a LHe-cooled bolometer was such low efficiency cannot be tolerated. One way around
needed. this problem is to use a lamellar grating interferometer, for
Bruker Optics now market a cube-corner interferometer which a beamsplitter is not required.
based on similar principles to Kauppinens interferometer. A schematic of a lamellar grating interferometer is shown
The IFS 120 HR has a maximum optical retardation of in Figure 11. Unlike the Michelson interferometer, which
greater than 7.1 m, giving an achievable resolution of better operates on the principle of amplitude division, lamellar
than 0.001 cm1 . One line in the rotation spectrum of CO grating interferometers operate on the principle of wave-
at 61.42 cm1 is shown in Figure 9, demonstrating that the front division. It consists of two sets of parallel interleaved
theoretical resolution is being achieved. To give an idea
of the high information content of spectra measured on this
instrument, the spectrum of H2 S2 over a wavenumber range
of 0.75 cm1 measured at a resolution of 0.0025 cm1 with
a mercury arc source and LHe cooled bolometer is shown
in Figure 10.
0.90
0.80
0.70
Transmittance
0.60
0.50
0.40
v4 = 1
0.30 Ground state
0.20
70 60 50 40 30
0.10 rQ branch
8 J
0.00
78.50 78.60 78.70 78.80 78.90 79.00 79.10 79.20 d
Wavenumber / cm1 Figure 11. Principle of the lamellar grating interferometer. Radi-
Figure 10. FIR spectrum of H2 S2 over a wavenumber range of ation reflected off the moving facets travels a distance of 2d/cos a
0.75 cm1 measured at a resolution of 0.0025 cm1 . further than the radiation reflected off the stationary facets.
mirrors, or facets. One set is fixed, while the other set can wavelength. Diamond is transparent throughout the entire
move in a direction perpendicular to the plane of the fixed FIR. However, its small size and high cost prevents it from
facets. When the moving facets have traveled a distance being a general-purpose window. The main use of bulk
d, the path difference between the rays that strike the fixed diamonds for FIR spectrometry has been as a window of
facets and the rays that strike the moving facets is 2d/cos a, a Golay detector. On the other hand, thin diamond films
where a is the angle of incidence of the median ray of may be fabricated by chemical vapor deposition, have high
the beam. Because almost all the radiation from the source strength and are not extortionately expensive.
reaches the detector, the efficiency of a lamellar grating can The best general purpose window material for FIR
be significantly higher than that of a Michelson interfero- spectroscopy is polyethylene. High-density polyethylene
meter with a PET beamsplitter. (HDPE) is often used as a window for gas and liquid cells;
The low-wavenumber cut-off nQ L of a lamellar interfero- however, this material has a fairly weak crystal lattice mode
meter is determined by the cavity effect, by which the at 72 cm1 . When windows are at least 5 mm thick, the
modulation of waves whose electric vector is parallel to the transmittance of this band can be less than 50%. Low-
sides of the cavity starts to decrease. This occurs when nQ L < density polyethylene (LDPE) has lower crystallinity than
0.3a1 , where a is the grating constant. Thus if the facets HDPE, so that this band is far weaker in LDPE than HDPE.
are 1 cm across, nQ L 3 cm1 . The lowest frequency that As a result, however, LDPE is also less rigid and not as
has been measured with a lamellar grating spectrometer is suitable as HDPE as a window material. The melting point
1.5 cm1 . The high-wavenumber limit results from the fact of LDPE is about 140 C and so this material is easily
that the grating can be considered as a series of long rect- melted. One useful way of preparing solid samples for FIR
angular slits. Diffraction of the beam at this grating results spectroscopy is to grind them with powdered LDPE and
in the interference being cancelled at the exit aperture. The compress the mixture while heating it to about 140 C. In
high-wavenumber limit nQ H is given by F/aS, where F is the view of the fact that blackbody sources emit much higher
focal length of the collimator and s is the diameter of the power in the MIR than the FIR region of the spectrum,
exit aperture. In practice, nQ H rarely exceeds about 150 cm1 . it is often preferable to remove as much of the short-
The lamellar grating interferometer is the instrument of wavelength radiation as possible if very long wavelengths
choice for measurements made below about 20 cm1 , and are to be measured. In this case, black polyethylene, of the
Beckman-RIIC once sold one of these instruments (the FS- type commonly used in garbage collection bags, comes in
820) that was optimized for this spectral region. However, useful as an optical filter since it absorbs most of the energy
there now is little interest in the measurement of spectra at above about 600 cm1 while having a transmittance greater
such long wavelengths, and there are no instruments of this than 50% below 200 cm1 .
type currently available commercially. Two other polyolefins that have been proposed for use as
One other instrument that was developed in the late windows for FIR spectrometry are PP and poly-(4-methyl-
1960s and early 1970 was the MartinPuplett polarizing pentene-1), or TPX . These materials are structurally
interferometer.6 This instrument is described in some detail related as 4-methyl-pentene-1 is a propylene dimer. These
by Polavarapu7 and will not be described here. Suffice it to materials have one great advantage over polyethylene for
say that, like the lamellar grating spectrometer, the perfor- use as a window. Whereas polyethylene is typically barely
mance of the polarizing interferometer in the far infrared is translucent, PP and TPX are transparent to visible light.
also not limited by the poor efficiency of the PET beamsplit- TPX is significantly more susceptible to air oxidation
ter and excellent results were obtained on this instrument. when illuminated with ultraviolet light, so that if a mercury
lamp is used as the source, optical filtering, e.g. with
black polyethylene, is necessary. Many FIR spectroscopists
4 WINDOW MATERIALS FOR tend to use PP thin films in preference to TPX when
FAR-INFRARED SPECTROMETRY used in a vacuum, since PP has a fairly high burst
pressure.
Since NaCl and KBr are opaque below 650 and
380 cm1 , respectively, they cannot be used as window
materials for FIR spectroscopy. Cesium iodide is transpar- 5 SUMMARY
ent to about 150 cm1 and hence can be used, not only as
a window for measurements down to this wavenumber, but Fourier transform spectrometers developed in three distinct
also for the preparation of alkali halide disks. spectral regions in the early 1960s. Pierre Connes and his
Silicon cuts on below 400 cm1 and is often useful, co-workers in France developed remarkably sophisticated
although its high refractive index does lead to some loss step-scan interferometers that permitted NIR spectra to
of energy. Both quartz and sapphire also transmit at long be measured with a resolution of better than 0.01 cm1 .
These instruments may be considered the forerunners of HDPE High-density Polyethylene

the step-scan interferometers made by Bruker, Bio-Rad LDPE Low-density Polyethylene
(Cambridge, MA, USA) and ThermoNicolet although their LHe Liquid Helium
principal application was in the field of astronomy. Low- NEP Noise Equivalent Power
resolution rapid-scanning interferometers were developed NPL National Physical Laboratory
by Larry Mertz and his colleagues at Block Engineering PET Poly(ethylene terephthalate)
(Cambridge, MA, USA) for remote sensing. Nonetheless, PP Polypropylene
the FT-IR spectrometers that are so prevalent in chemical RIIC Research and Industrial
laboratories today are direct descendents of these instru- Instruments Co.
ments. The interferometers that were developed for FIR
spectrometry in Gebbies laboratory have had no commer-
cial counterparts for at least 15 years. However, it could be REFERENCES
argued that these instruments did as much to demonstrate
the power of Fourier transform spectroscopy to the chem- 1. G.P. Williams, Synchrotron and Free Electron Laser Sources
ical community as any of the instruments developed for of Infrared Radiation, in Handbook of Vibrational Spec-
troscopy, John Wiley and Sons, Chichester, Vol. 1 (2002).
MIR and NIR spectrometry. Their performance was every
bit as good as todays rapid-scanning interferometers. How- 2. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
Spectrometry, Wiley Interscience, New York, 209212 (1986).
ever, the market for these instruments is so small today
that it has proved more lucrative to modify rapid-scanning 3. R.S. Jackson, Continuous-scanning Interferometers for Mid-
infrared Spectrometry, in Handbook of Vibrational Spec-
interferometers that were originally designed for MIR spectroscopy, John Wiley and Sons, Chichester, Vol. 1 (2002).
trometry than to compete with these instruments with slow
4. Nicolet Instrument Corporation, US Patent 4632553 (1986).
continuous scan or step-scan interferometers.
5. J. Kauppinen and V.-M. Horneman, Appl. Opt., 30, 2575
(1991).
ABBREVIATIONS AND ACRONYMS 6. D.H. Martin and E. Puplett, Infrared Phys., 10, 105 (1969).
7. P. Polavarapu, Polarization Interferometers for Infrared Spec-
DTGS Deuterated Triglycine Sulfate trometry, in Handbook of Vibrational Spectroscopy, John
FT-IR Fourier Transform Infrared Wiley and Sons, Chichester, Vol. 1 (2002).
Incandescent Sources for Mid- and Far-infrared
Spectrometry
Henry Buijs
ABB Bomem Inc., Quebec, Quebec, Canada
1 BACKGROUND and q is expressed in radians. A usable IR radiation source

should have maximum incident power with an etendue of
Infrared (IR) absorption spectroscopy is the study of IR 0.05 cm2 sr.
radiation absorbed by samples of a given thickness. The
distribution of absorption as a function of wavelength is
3 POWER EMITTED BY AN IDEAL
determined by means of a spectrophotometer. A spec-
trophotometer requires a source of intense multi-wavelength INCANDESCENT SOURCE
radiation to provide adequate intensity incident on the sam-
ple. The intensity of IR radiation is commonly expressed Plancks radiation law determines the power emitted by
in watts of incident power. The geometric distribution of a small aperture in a cavity that is at some equilibrium
intensity plays an important role in spectrometer design and temperature. This is called black body radiation and cor-
performance. Typical sample presentation techniques dic- responds to an ideal incandescent source. It describes the
tate that the illumination area on the sample be less than power density Pl per unit wavelength interval and per unit
0.25 cm2 , 0.6 cm diameter, while for micro-sampling the cm2 sr as shown in equation (2).
area is much smaller than this. To ensure that deviations C1
Pl .dl D dl 2
from Beers law are minimal, the illumination should be l5 fexp[C 2 /lT] 1g
approximately perpendicular to the sample surface with Since equation (2) gives the density function per unit wave-
a small range of angles. A useful source of IR radiation length interval, the equation is different when expressed per
should have high intensity distributed over a small area unit wavenumber interval, as shown in equation (3).
and a small range of angles.
C1 n3
Pn .dn D dn 3
[expC2 .n/T 1]
2 ETENDUE Expressing wavelength l in centimeters and wavenum-
ber n in cm1 , C1 D 1.19 1012 W cm2 sr1 and C2 D
The product of area of the beam, A cm2 , and its solid angle, 1.4388 cm K. C1 and C2 are the Planck first and sec-
steradians (sr), is known as etendue, cm2 sr. A circular ond radiation constants. The temperature T is expressed
symmetric beam illuminating a sample with total cone angle in Kelvin. Figure 1 shows the distribution of power as a
of q degrees has a corresponding solid angle where function of frequency for an ideal black body source at
1200 K.
p.q2 There are several useful relationships for sources that fol-
D 1
4 low Plancks distribution. Wiens displacement law states
that the maximum of the curve occurs at a wavenumber that
John Wiley & Sons Ltd, 2002. increases linearly with temperature. The StefanBoltzmann
0.0012 and far-infrared radiation and are installed in well over

W cm2 sr 1 cm 1
0.0010 95% of all contemporary commercial mid- and far-infrared

0.0008
spectrophotometers.
0.0006
0.0004 5 OTHER MID- AND FAR-INFRARED
0.0002 SOURCES
0 2000 4000 6000 8000 10 000 12 000
A source called a pulsIR manufactured by Ion Optics
Wavenumber / cm1 (Waltham, MA, USA) is quite useful for several applica-
Figure 1. Spectral radiance distribution for 1200 K ideal black tions. It consists of a thin metal foil heated via an electric
body source. current to high temperature. The surface is deliberately tex-
tured by means of ion beam etching such that the emissivity
law states that the distribution integrated over all wavenum- is greatly increased compared with bare metal. The source,
bers increases as the fourth power of temperature. Finally, being made from a thin metal foil, consumes little power
approximating the equation for low wavenumbers in the and has a very low heat capacity. This permits chopping
far-infrared shows that the intensity increases linearly with the IR intensity simply by turning the source on and off at
temperature. frequencies up to 10 Hz. The surface area of the foil is suf-
ficiently large to provide an adequate etendue for mid- and
far-infrared spectroscopy as described above. This type of
4 A PRACTICAL MID- AND source is principally intended for dedicated analyzers that
FAR-INFRARED SOURCE can sense, for example, toxic gases in an indoor environ-
ment via nondispersive IR spectroscopic techniques. Daly
A surface heated to elevated temperature will emit radiant et al.1 have described a novel extension of this type of
power following Plancks law but modified by its emissivity source. By texturing the metal foil surface with repeated
which may vary between 0 and 1 and may vary with patterns, resembling photonic bandgap materials, using fine-
wavelength. A surface with constant emissivity near 1 and line lithography, they have produced IR sources where the
being able to operate at elevated temperature is a good emissivity is tuned to relatively narrow spectral regions.
candidate for a mid- and far-infrared source. For dedicated analyzers, such a source includes the chop-
Sintered silicon carbide can be heated to a temperature ping and wavelength selection functions all in one small
over 1800 K by passing electrical current through it. A device with low power consumption.
silicon carbide rod (frequently known by the trade name
Globar ) can be maintained in air at 1400 K for a number
of years without degradation. The temperature of a silicon 6 VERY-FAR-INFRARED
carbide Globar can be maintained stable so that the radi- SPECTROSCOPY
ated power remains constant with time. Its emissivity is
quite constant with wavelength and is typically >0.8. In most cases where the Globar is not used for mid- and
Other materials, such as tungsten and sintered zirco- far-infrared spectroscopy, the reason is that greater power
niumthoriumcerium oxide (Nernst glower) have also density is needed. Below 100 cm1 , the intensity of the
been used as sources of mid-infrared radiation. These can be Planck curve falls off quickly, as shown in Figure 1. Spec-
heated to higher temperature than the Globar but exhibit tral measurements in this region require long observation
significantly lower emissivity at longer wavelength com- time to achieve sensitivity comparable to the mid-infrared.
pared with silicon carbide. Hence they are less suitable as After optimization of detector sensitivity and etendue, it
IR sources in the mid- and far-infrared. is desirable to optimize source radiation intensity as well.
Early versions of silicon carbide sources consisted of This requires significantly higher temperature of the emit-
a rod with silvered tips clamped between two electrodes. ting material. The plasma of a gas discharge can reach
The silvering had a tendency to become oxidized so that temperatures over 5000 K. However most gas discharge
contact with the electrodes was lost; thus older Globar plasmas have widely varying emissivity as a function of
sources had a rather limited lifetime. Today various forms wavelength. To achieve uniformly high emissivity requires
of silicon carbide igniters are manufactured. These are a plasma with high gas density. High pressure mercury and
designed to reliably ignite natural-gas-fired appliances. xenon discharge lamps are frequently used for spectroscopy
These igniters are most frequently used as sources of mid- in the very-far-infrared. So-called short arc high-pressure
Incandescent Sources for Mid- and Far-infrared Spectrometry 3
lamps are popular for bright visible illumination, but the gas is used to concentrate the radiation at a point where it can be
volume at high temperature is small which leads to limited used for spectroscopy and other applications. A synchrotron
etendue and reduced emissivity. The Phillips HPK 125 W source has an equivalent black body temperature in excess
high pressure mercury lamp has a large volume discharge of 10 000 K. It is particularly attractive for spectroscopy
using high pressure mercury vapor. This source has high and special illumination applications in the far UV and
etendue and high emissivity and an effective temperature of soft X-ray regions of the spectrum. But it has also been
approximately 5000 K. The plasma is enclosed in a quartz employed for vibrational spectroscopy in the mid- and far-
envelope that transmits radiation below 60 cm1 . Here its infrared. Synchrotron source radiation has been used effec-
intensity is four times the intensity of a Globar as expected tively for difficult microscopy analysis problems2 as well
from the ratio of temperatures. Above 100 cm1 the enve- as high resolution far-infrared spectroscopy. Synchrotron
lope is opaque and acts as an IR source with high emissivity sources are described in greater detail in an article by
at the temperature of the envelope which is in the range of Williams in this section of the Handbook (see Synchrotron
1500 K. and Free Electron Laser Sources of Infrared Radiation).
7 HIGH RESOLUTION INFRARED

9 TERAHERTZ SPECTROSCOPY
SPECTROSCOPY
In 1987, Grichkowsky et al.3 described a new method for
The slit width of a grating or the diameter of the Jacquinot far-infrared vibrational spectroscopy using subpicosecond
stop of a Fourier transform infrared spectrometer must be pulses of electric current to generate short bursts of electro-
set small in order to achieve high spectral resolution. As a magnetic radiation. The spectral distribution of the radiation
result the etendue is severely limited. At the same time is inversely proportional to the pulse duration. For subpico-
noise density increases with resolution, other conditions second pulse duration, the spectral bandwidth is in excess
being constant. Measurement time increases rapidly to of 30 cm1 in the far-infrared. With the propagation of short
many hours as the resolution is increased. Much work duration pulses of electromagnetic radiation through a sam-
has been done to optimize detectors for high resolution ple, it is possible to not only determine the absorption of
spectroscopy so as to reduce observation time. Despite this, the radiation but also the propagation delay. For a sam-
some work at high resolution has been performed using a ple with absorption bands distributed along the spectrum,
source more intense than the Globar . A carbon rod with the propagation of a short pulse will result in a disper-
a cavity drilled transversely can be heated to over 3500 K. sion of the pulse both due to the change in its spectral
To avoid deterioration it is enclosed in a housing with an shape and the differential changes in the index of refraction
IR transmitting window and purged with a buffer gas, such (dispersion) of the sample. Grichkowsky et al. showed an
as argon at several bars pressure. Such a source has high arrangement for spectroscopy where radiation due to sub-
emissivity and high temperature. Used in the mid-infrared it picosecond pulses was propagated through a sample and
has permitted reduction of observation time of between five subsequently detected using a triggered detection system
and ten times compared with using a Globar . The carbon
such that the detection is synchronized with the emitted
rod source is inconvenient, however, because it requires a
pulses. By varying the delay between emission of the radi-
significant amount of argon gas for purging and a power
ation pulse and detection of the transmitted radiation, the
supply that can furnish several hundred amperes of current.
pulse dispersion can be mapped. The graph thus generated
is the complex Fourier transform of the spectral distribu-
8 IR SPECTROSCOPY WITH A tion. By performing a complex inverse Fourier transform,
both the index of refraction and absorption spectra are
SYNCHROTRON SOURCE
obtained. The technique of terahertz spectroscopy is consid-
A number of synchrotron accelerator facilities exist specif- erably more sensitive than conventional spectroscopy but is
ically to be used to provide an intense source of radiation. currently limited to the far-infrared only.
A particle, such as an electron, when accelerated emits syn-
chrotron radiation. By angularly accelerating a stable beam
REFERENCES
of particles around a circular track, a continuous source
of intense radiation can be tapped from the beam. The
1. J.T. Daly, A.C. Greenwald, E.A. Johnson, W.A. Stevenson,
radiation is emitted along the direction of motion of the J.A. Wollam, T. George and E.W. Jones, Mater. Res. Soc.
particles. For a circular track it is emitted in a radial out- Symp. Proc. 607 (Infrared Applications of Semiconductors III),
ward direction all along the track. Special collection optics 175180 (2000).
2. D.L. Wetzel, J.A. Sweat and D.D. Panzer, Synchrotron Pow- 3. D. Grischkowsky, C. Chi, I.N. Duling, W.J. Gallager, J.M.
ered FT-IR Microspectroscopy Enhances Spatial Resolution Halbout and M.B. Ketchen, Photoconductive Gen-
for Probing and Mapping of Plant Materials, in Fourier eration of Sub Picosecond Electrical Pulses and their
Transform Spectroscopy: 11th International Conference, ed. Measurement Applications, in Proceedings of Topical Meet-
J.A. De Haseth, AIP Conference Proceedings 430, Am. Inst. ing on Picosecond Electronics and Opto-electronics
Physics, Woodbury, NY, 354357 (1997). (1987).
Synchrotron and Free Electron Laser Sources of
Infrared Radiation
G.P. Williams
Thomas Jefferson National Accelerator Facility, Newport News, VA, USA
1 INTRODUCTION all the sources have an important role to play in scientific

investigations, and it is important to match the source to
Light of a given wavelength can be characterized by its the experiment.
brightness, power, bandwidth, and time structure. In many
cases these definitions are evident from the units. For 2 FT-IR SPECTROSCOPY AND
example, power could be expressed in W cm, or it could
also be expressed in photons s1 per 0.1% bandwidth. BRIGHTNESS
Brightness, also called brilliance, or spectral radiance, is
Brightness is an important concept for many applications,
defined as power/(area solid angle) and will therefore
particularly those in which the samples are small. In the
be in units such as W cm mm2 sr1 . Brightness has the
infrared (IR) region where the signal-to-noise ratio (S/N) is
unique property of being at best conserved in an optical
often limited by detector noise, for a Fourier transform in-
system. As an example, we consider a 1200 K thermal
frared (FT-IR) instrument, the S/N per scan can be written
source of area 10 1 mm, which emits into an angle of
as equation (1):
2p sr. For this source the power could be of the order
of 20 W into a bandwidth of approximately 10 000 cm1 BnN nN xD
S/N D 1
and the area solid angle is 63 mm2 sr. Thus the bright- f1/2 A1/2
ness is 20/63 104 , or 30 W cm mm2 sr1 . In contrast,
where BnN is the source brightness, nN the resolution, the
we consider a synchrotron radiation source which emits
limiting etendue of the experiment, x the optical efficiency,
100 mW into the same bandwidth, namely 10 000 cm1 , but
D the detectivity, f the electronic bandwidth and A the
from a source size of 330 330 m, and into a solid angle
area of the detector. Here it is assumed that the etendue,
of 1 msr, giving a brightness of 0.1/0.032 103 104 , , the area solid angle product describing the sample, is
or 1 mW cm mm2 sr1 , a factor of 300 higher than the ther- less than that of the source. Note that equation (1) is based
mal source. Free electron lasers (FELs), with a total power on, but not identical with the equation used by Griffiths
of 100 W cm, emitting into a diffraction-limited phase space and de Haseth,1 and in particular here we have used f
of l2 , will, at a wavelength of 3 m, have a brightness of instead of 1/t where t is the total measurement time. The
100/3 103 2 or 107 W cm mm2 sr1 , a factor of 1010 equivalence of these quantities depends on the electronics.
higher again! For thermal sources one has little control over the bright-
These comparisons serve to introduce the terminology ness apart from raising the temperature or changing the
and a sense of scale, but miss many important points such emissivity, so that this term is usually ignored. However,
as temporal structure and total available bandwidth. In fact, when other, alternative sources are considered, this is the
appropriate term to optimize for use with samples mounted
This is a US Government Work and is in the public domain in in geometries where the etendue is limited, such as in
the United States of America. microscopy, in cryostats, or in high-pressure cells.
3 LIGHT PRODUCED BY reference to Figure 1. The passing of an electron produces

ACCELERATORS an electric field varying in time. If the charge is moving
close to the speed of light, the pulse is sharp because rel-
3.1 Qualitative description ativistic corrections have to be applied to the field. The
effect of these corrections is to fold the field forward as
Electron particle accelerators can produce very bright illustrated in Figure 1. We obtain the spectrum by using
beams of IR radiation. We can see how they work with the Fourier transform theorem to give the intensity as a
function of 1/time or frequency. The intensity can be cal-
culated by squaring the modulus of the electric field, and
the result is a useful source of light as will be seen from
e the calculations.
1 1 mrad There are two main classes of such sources. In a syn-

chrotron radiation source, such as is shown in Figure 1,
(a)
the electron is circulating in a storage ring, the pulse is
sharp and the resulting Fourier transform yields a broad-
band spectrum. In fact, owing to relativistic corrections, the
electric field is pointed in a forward direction into an angle
of 1/g, where g is the ratio of the mass of the electron to the
|E ()|2
E (t )
rest mass. This has a value of 1000 for a 500-MeV electron.

We find that if the electron describes a circle of radius r,
the pulse width t D 1/g2 r/cg, the 1/g2 being an addi-
tional relativistic correction. Putting in numbers we find that
(b) t (c) t D 2/109 3 108 D 1017 s. Thus the wavelength
Figure 1. Schematic of synchrotron radiation production by a indicated by the vertical dashed line in Figure 1(c) is given
relativistic particle. In (a), an electron is traveling relativistically by ct D 30 A 1 A D 1010 m. Note that the spectrum
in a magnetic field, its motion describing a circle. The influence extends to longer wavelengths with only a slow falloff.
of the electric field, modified owing to the relativistic motion, is
shown. g is the ratio of the mass of the particle to its rest mass. The second class of device, called an undulator, is
(b) Electric field as a function of time at the observer. (c) Fourier shown in Figure 2. Here, the electron is traveling through
transform of (b) which yields the broadband spectrum. a varying magnetic field yielding a longer sinusoidal
Undulator
N periods
E (t )
Radiation from an insertion device
N S
N S Radiation
S t
N Electron beam
S N
N
N S
S
Magnet poles
|E ()| 2
= 1
N

Figure 2. In an undulator or wiggler, electrons travel in alternating magnetic fields, yielding a truncated sinusoidal electric field as a
function of time. The Fourier transform of this is a sharp spectral peak.
Synchrotron and Free Electron Laser Sources of Infrared Radiation 3
pulse, the Fourier transform of which is a sharp, narrow Compared with thermal sources at 2000 K, which are
spectral band. This long pulse alone results in the syn- available, but are much hotter than conventional globar
chrotron spectrum effectively being compressed by about or Nernst sources, synchrotron radiation is 1000 times
four orders of magnitude with a corresponding increase brighter, undulator radiation about 107 times brighter, and
in the brightness. The undulator can be on a linear FEL radiation about 1011 times higher.
accelerator or a storage ring, although in the latter case
the electron beam energy and typical device characteris-
tics are such that the emission is not usually in the IR 3.2 Quantitative description
region.
In practice, the electrons in linear accelerators and stor- 3.2.1 Synchrotron radiation
age rings are accelerated by the radio-frequency cavities,
and are in bunches. The light which is emitted is given The complete calculations of synchrotron radiation power
by an incoherent superposition of the emission from a sin- and brightness are discussed in detail elsewhere.11 In the IR
gle electron, this output being proportional to the number region, however, simple equations can be used to calculate
of electrons in the bunch, which is as high as 1012 in the the spectra from storage rings. For reference we compare
National Synchrotron Light Source (NSLS), Brookhaven the numbers with those from a thermal source.
National Laboratory vacuum ultraviolet (VUV) storage It can be shown that in the IR spectral region, where the
ring, for example. Readers interested in a deeper introduc- synchrotron radiation source is typically diffraction limited,
tion will find several papers useful.2 7 the brightness is given12 by equation (2):
In passing, we mention the topic of edge radiation. Bsr nN 108 InN 2 W mm2 sr1 cm 2
This is radiation emitted by the longitudinal, rather than
transverse, velocity change of the electrons which occurs where I is the stored beam current in amperes and nN the
as they enter or exit dipoles on storage rings.8 10 In theory, frequency in wavenumbers (cm1 ). Actual storage rings
it is almost as bright as synchrotron radiation but with yield lower brightness toward higher wavenumbers owing
characteristics similar to the radiation which is emitted to the finite size of the electron beam.
when an electron hits a foil and called transition radiation. The well-known expression for the flux from a blackbody
The radiation is easier to extract since for a single edge it radiator is equation (3):
is emitted into angles close to g (defined above), and which h!3 d!
are much smaller than is the case for synchrotron radiation. Bbb ! D 2 2 h !/kT
3
2p c e 1
However, usually two edges interfere, opening up the angle,
and also the source is not a HermiteGaussian 0,0 beam Assuming emission into 2p sr (plus ignoring any obliquity
as shown in Figure 1(a), but has a hollow profile, and so factors), substituting T D 2000 K (the temperature of a
the brightness never exactly matches synchrotron radiation. typical thermal source), and converting to wavenumber
Owing to the smaller opening angle, diffraction plays a yields equation (4):
major role in increasing the source size and may limit the dnN
far-IR emission. Bbb nN D 6 1015 nN 3 W mm2 sr1 cm (4)
enN /1400 1
Finally, we discuss the free elctron laser (FEL). This
Calculations of the synchrotron radiation and blackbody
device is similar to what is shown in Figure 2, but the
brightness, using equations (2) and (4), are shown in
undulator is placed in an optical cavity or resonator, which
Figure 3. It is obvious that the synchrotron offers approxi-
reflects the emitted light back and forth. The electrons
mately three orders of magnitude advantage over the ther-
then become tightly bunched owing to interactions with an
mal source.
electromagnetic wave of light which is also traversing the
An easy way to calculate the power from both sources
undulator. This light can be spontaneously generated by the
is to work backwards, multiplying the brightness by the
electrons in the magnetic field, can come from an external
(area solid angle). For the synchrotron source, this requi-
seed laser, or can be the light generated from a previous
res knowledge of the emission angle and source size, both
bunch of electrons, reflected from mirrors which form an
of which are dependent on the particular storage ring. In
optical cavity outside the undulator. If this tight bunching is
the absence of specific parameters, an approximation is to
of the order of the wavelength of the light being emitted, the
use the diffraction limit of l2 . However, for synchrotron
electric field that has to be considered is now that of many
radiation the natural opening angle is given by equation (5):
electrons, so that the modulus squared of the electric field
1/3
is enhanced by the number of electrons squared, which, as l
qnat rad D 1.66 5
mentioned above, is many orders of magnitude. r
1019
1018 Infrared brightness
Photons s1 per 0.1% bandwidth mm2 sr 1

1017
1016 Synchrotron radiation
1015
Eq. (2)
1014 Practical limits
1013
2000K black body
1012
1011
1010
109
108
10 000 1000 100 10 1
Wavelength /m
Figure 3. Calculated brightness for synchrotron radiation (1 A) and a thermal source.
in Figure 4 for three wavelengths for a ring of radius 2 m.

In addition, one needs to know the electron beam diameter.
Parallel Full calculations of power for both synchrotron radi-
to orbit plane
ation and a thermal source are shown in Figure 5. The
figure illustrates one of the less exploited features of IR
synchrotron radiation, namely the enhanced flux at very
Power per unit vertical angle
long wavelengths. At wavelengths beyond 100 m, the

synchrotron flux can exceed that from a mercury arc lamp.
Perpendicular In a typical measurement at 10 cm1 , up to 20 times more
to orbit plane power is expected from the synchrotron.13
Another less used feature of synchrotron radiation is
the time structure. The pulse structure varies consider-
ably among the synchrotron radiation facilities, but pulse
durations ranging from tens of picoseconds to 1 ns are
common, with pulse repetition frequencies from hundreds
down to a few megahertz. This has been utilized in several
experiments, as will be discussed later.
One additional characteristic of the synchrotron source is
that it is absolute, since the power depends on the electron
20 0 20
beam current stored in the ring, which can be measured
Vertical angle (mrad)
to 1 part in 106 . In principle, it should also be a highly
stable source since the beam current is a slowly decaying
Figure 4. Vertical angular distributions above and below the orbit
function, although problems with unstable electron beam
plane for synchrotron radiation of various wavelengths for a
storage ring of 2-m radius. Dashed line, 1 m; solid line, 10 m; orbits can generate noise problems. The absolute nature of
dotted line, 100 m. the synchrotron source has been used in many experiments
and led to the discovery of the role of electrons in metals
where l is the wavelength, r the radius (in the same units in adsorbate dynamics and surface friction.14
as l), and qnat is the angle required to transmit 90% of the Despite these advantages, the full realization of IR syn-
emitted light at a given wavelength. A detailed calculation chrotron radiation as a productive source was limited until
of the vertical distribution of synchrotron radiation is shown the issue of electron beam orbit stability was addressed.
1018
1017 Infrared power
Photons s1 per 0.1% bandwidth 1016

2000 K black body
1015
1014
Synchrotron radiation
1013
Practical limit
1012
1011
1010
109
10 000 1000 100 10 1
Wavelength /m
Figure 5. Calculated power for synchrotron radiation (1 A) and a thermal source.
The VUV ring at the NSLS is a facility that has regularly 4 IMPLEMENTATION OF
improved orbit stability and allowed, in part, the successful SYNCHROTRON RADIATION
utilization of the IR programs.15,16
Owing to the large emission angles of IR synchrotron
3.2.2 Undulators and free electron lasers radiation, mechanical modifications of the storage rings are
usually required, involving an enlargement of the vertical
Finally, for completeness, we present a practical equation
opening angles subtended by the extraction ports. The
for the wavelength harmonics, ln , emitted by an undulator/
horizontal angles are usually not a problem. The largest
FEL. On-axis, these are given by equation (6):
vertical angle obtained on a storage ring is 90 mrad at

13.056lu K2 the NSLS, Brookhaven.18 The next major problem is to
ln D 1C 6
nE2 2 deal with the thermal load on the first mirror, which
requires a finite element analysis calculation. Solutions to
where ln is in angstroms, E GeV is the electron beam the thermal problem may be either water cooling of the
energy, lu cm is the undulator period and n is the harmonic. first mirror, X-ray beam interception via a water-cooled
K is called the deflection parameter, given in practical finger, or the use of a slotted mirror. Normally for rings
units by equation (7): of 2 m radius, water cooling is adequate, whereas for
K D 0.934lu B0 7 rings of 5-m radius, the power density is such that too
much distortion would occur if the entire beam struck the
where B0 is the peak magnetic field in tesla. As an example mirror.
we cite the FEL at Jefferson Laboratory17 in which the Once the first-mirror heating problem has been solved,
undulator period is 2.7 cm and B D 0.55, making K D 1.39. typical beamlines try to produce a focused beam on a
The first harmonic for a 40-MeV electron beam is then 1 wedged diamond window which then ends the storage
43 403 A, or 4.3 m. ring ultra-high vacuum requirements. The beam may then
Compared with storage rings, FELs have much shorter be collimated and transported to an FT-IR instrument and
pulse widths; the Jefferson Laboratory FEL pulses are the remainder of the instrumentation is standard laboratory
400 fs long and repeat at frequencies up to 75 MHz. based.
1E-5
NSLS
1E-6
1E-7
W cm mm 2 sr 1
1E-8
1200 K black body
1E-9
1E-10
Detector noise
1E-11
1E-12
1000 2000 3000 4000 5000
Wavenumbers /cm1
Figure 6. Power falling on to a 10-m sample at f/1 optics for a synchrotron source (NSLS, Brookhaven National Laboratory) and a
1200 K thermal source.
Figure 7. (a) Photograph of triangular fluid inclusion in fluorite; (b) IR spectrum of fluid inclusion and maps of bands indicated by
arrows. The inclusion is mainly composed of oil with a CO2 bubble probably surrounded by OH groups.
5 EXAMPLES OF EXPERIMENTS WITH The technique works well for samples which recover in the
SYNCHROTRON RADIATION time between pulses. Typically, spectra are recorded for one
fixed time difference between the pump and the probe, then
Perhaps the best example of the use of high-brightness, repeated for different time differences. The technique has
broadband IR light is microspectroscopy. The advantage is been applied to semiconductor materials, GaAs,23 and to a
immediately evident if one calculates the power through a superconducting film of Pb in which the dynamics of the
10-m pinhole illuminated with f/1 optics. We have done quasi-particles was studied.24
this in Figure 6 and it can be seen that for these conditions,
a 1200 K thermal source provides about 1 nW cm on to
the sample compared with 1 W cm from the synchrotron 6 CONCLUSIONS
source. Given typical HgCdTe detector noise characteristics
of 1011 W, the S/N advantages are clear. Spectroscopy The successful utilization of synchrotron radiation as a
can readily be performed at the diffraction limit with bright, broadband source for spectroscopy is evident from
synchrotron radiation.
Table 1. List of some of the IR synchrotron
It turns out to be relatively straightforward to couple a radiation facilities either operating or planned.
synchrotron source to an FT-IR/microscope combination,
and it was first done at Brookhaven National Laboratory in Laboratory Country
September 1993.19,20 With this combination of tools, it is
possible to make maps of samples using the IR vibrational NSLS, Brookhaven USA
bands as a contrast mechanism. To do this, the samples are ALS, Berkeley USA
raster-scanned, with a full spectrum being taken at each Aladdin, Wisconsin USA
NIST, Gaithersburg USA
point. Values of the strengths of chosen vibrational modes
SRC, Daresbury UK
are extracted at each point and the data plotted. As an LURE, Paris France
example we show in Figure 7 maps of the functional groups Max-Lab, Lund Sweden
for a triangular fluid inclusion in a piece of fluorite.21 The UVSOR, Okasaki Japan
measurements have a spatial resolution of 0.7l.22 Dane, Rome Italy
Another important application concerns pumpprobe SRRC, Hsinchu Taiwan
CLS, Saskatoon Canada
experiments. In these experiments a laser or FEL is syn- ANKA, Karlsruhe Germany
chronized to the pulses from a synchrotron. Synchrotron NSRL, Hefei China
storage rings typically run at 50 or 500 MHz and have bunch SSRF, Shanghai China
lengths of 50 or 500 ps repeating every 2 or 20 ns, respec- CAMD, Baton Rouge USA
tively. Samples are pumped by the laser and probed by BESSY-2, Berlin Germany
SpRing-8, Nishi Harima Japan
the broadband synchrotron source as shown schematically
LNLS-1, Campinas Brazil
in Figure 8. Thus no time resolution is demanded of the DELTA, Dortmund Germany
detector other than that required for the spectroscopy itself.
Far-infrared
(probe) pulse Spectrometer
Sample
Detector
Synchrotron source
NSLS VUV ring,

Brookhaven Natl. Lab.
Fiber-optic cable
transports pulses
to sample
Synchrotron
signal
Mode-locked (pulsed) laser
Figure 8. Schematic of pumpprobe spectroscopy using a synchrotron radiation source synchronized to a laser (G.L. Carr, personal
communication).
the number of facilities operating or being planned. In 6. S. Krinsky, M.L. Perlman and R.E. Watson, in Handbook
Table 1 we list all that we know about at the time of of Synchrotron Radiation, ed. E.E. Koch, North-Holland,
Amsterdam, Chapter 2 (1983).
writing.
Hence it is likely that the utilization of these synchrotron 7. C.J. Hirschmugl, M. Sagurton and G.P. Williams, Phys. Rev.
A, 44, 1316 (1991).
radiation sources will continue to increase as more and more
of their capabilities are required for research and analytical 8. R.A. Bosch and O.V. Chubar, Am. Inst. Phys. Conf. Proc.,
417, 35 (1997).
work. In addition, during the next decade we are likely
to see a continued development of accelerator-produced 9. R.A. Bosch, Nucl. Instrum. Methods A, 386, 525 (1997).
IR including improvements in existing facilities, as well 10. P. Roy, M. Guidi Cestelli, A. Nucara, O. Marcouille, P. Cal-
as the introduction of new facilities such as those using vani, P. Giura, A. Paolone, Y.-L. Mathis and A. Gerschel,
Phys. Rev. Lett., 84, 483 (2000).
multiparticle coherence.
11. S.L. Hulbert and G.P. Williams, Synchrotron Radiation
Sources, in Vacuum Ultraviolet Spectroscopy I, eds J.A.R.
Samson and D.L. Ederer, Experimental Methods in the Phys-
ACKNOWLEDGMENT ical Sciences, Vol. 31, Academic Press, San Diego, 1 (1998).
12. J.B. Murphy, personal communication.
This work was performed under the auspices of
13. R.P.S.M. Lobo, J.D. Laveigne, D.H. Reitze, D.B. Tanner and
the US DOE under contract numbers DE-AC02- G.L. Carr, Rev. Sci. Instrum., 70, 2899 (1999).
98CH10886, Brookhaven National Laboratory and DE-
14. P. Dumas, M.K. Weldon, Y.J. Chabal and G.P. Williams,
AC05-84ER40150, Thomas Jefferson National Accelerator Surf. Rev. Lett., 6, 225 (1999).
Facility. Thanks are due to BNL staff and to Paul Dumas,
15. L.H. Yu, E. Bozoki, J. Galayda, S. Krinsky and G. Vignola,
Fritz Hoffmann, Yves Chabal, Carol Hirschmugl, Dieter Nucl. Instrum. Methods A, 246, 165 (1986).
Moeller, Larry Carr, and Lisa Miller.
16. R. Biscardi, G. Ramirez and G.P. Williams, Rev. Sci. Ins-
trum., 66, 1856 (1995).
17. S.V. Benson, G. Biallas, C.L. Bohn, D. Douglas, H.F. Dylla,
ABBREVIATIONS AND ACRONYMS R. Evans, J. Fugitt, A. Grippo, J. Gubeli, R. Hill, K. Jordan,
G. Krafft, R. Li, L. Merminga, G.R. Neil, P. Piot, J. Preble,
NSLS National Synchrotron Light Source M. Shinn, T. Siggins, R. Walker and B. Yunn, Phys. Rev.
VUV Vacuum Ultraviolet Lett., 84, 662 (2000).
18. G.P. Williams, P.Z. Takacs, R.W. Klaffky and M. Shleifer,
Nucl. Instrum. Methods A, 246, 165 (1986).
REFERENCES 19. G.L. Carr, J. Reffner and G.P. Williams, Rev. Sci. Instrum.,
66, 1490 (1995).
1. P.R. Griffiths and J.A. de Haseth, Fourier Transform IR 20. G.L. Carr, P. Dumas, C.J. Hirschmugl and G.P. Williams,
Spectroscopy, Wiley-Interscience, New York (1994). Nuovo Cimento D, 20, 375 (1998).
2. J. Schwinger, Phys. Rev., 75, 1912 (1949). 21. N. Guilhaumou, P. Dumas, G.L. Carr and G.P. Williams,
3. J.D. Jackson, Classical Electrodynamics, John Wiley & Appl. Spectrosc., 52, 1029 (1998).
Sons, New York (1975). 22. G.L. Carr, personal communication.
4. H. Winick, Synchrotron Radiation Research, Plenum Press, 23. G.L. Carr, Appl. Phys. Lett., submitted.
New York, Chapter 2 (1980).
24. G.L. Carr, R.P.S.M. Lobo, J. LaVeigne, D.H. Reitze and D.B.
5. A. Hofmann, Phys. Rep., 64, 253 (1980). Tanner, Phys. Rev. Lett., 85, 3001 (2000).
Detectors for Mid- and Far-infrared Spectrometry:
Selection and Use
E. Theocharous1 and J.R. Birch2

1
National Physical Laboratory, Teddington, UK
2
Cliftonville, Dorking, UK
1 INTRODUCTION Equivalent Power (NEP) while there are a host of other

parameters that can be critical in an infrared measurement
An infrared detector is a transducer that converts infrared which are being ignored. Unfortunately these parameters
energy into some other measurable form, usually into an can vary from fabrication batch to fabrication batch, result-
electrical signal. Herschel used a liquid in glass thermome- ing in some products being well outside their specification.
ter when he discovered that there was a significant amount Due to this variability in the performance of infrared detec-
of energy in the part of the solar spectrum lying beyond tors the authors have found it necessary to buy more than
the red.1 In 1833 the thermopile was invented2 but it was one detector and evaluate all of them before selecting the
not until the 1930s, when the military realized the ben- best from that batch.
efits offered by infrared in the battleground that infrared Section 4 summarizes the most important characteristics
detector developments accelerated. Improvements in the of the major infrared detector families. Only detectors with
detection of infrared radiation have, of course, benefited significant response below a fairly arbitrarily chosen value
civilian applications. of 7000 cm1 (wavelength longer than 1.4 m) are included
This article is aimed at those who want to identify the in this article. In general, such detectors fall into two cate-
most appropriate infrared detector for their particular appli- gories: those covering the near-millimeter wavelength (far-
cation or who wish to use their chosen detector under infrared) region from ca. 3 to 700 cm1 , and those used in
its optimum operating conditions. To accomplish this, the the mid-infrared region from about 400 to 7000 cm1 . Fur-
most important parameters governing the performance of thermore, the discussion will be limited to single element
infrared detectors will be described in Section 2 along with detectors as the treatment of detector arrays and matrices
explanations as to how these parameters affect detector is deemed to be outside the scope of this article.
performance. Some of these are the same parameters that
are tabulated by infrared detector manufacturers in their
brochures so the reader can identify the parameters that are 2 PARAMETERS DETERMINING THE
critical to his application. Unfortunately, it is our experience PERFORMANCE OF AN INFRARED
that at least some infrared detector vendors exaggerate the DETECTOR
performance of their products. They can usually get away
with this by relying on the inability of the purchaser to Hudson3 gives one of the best treatments of infrared
verify their claims. Furthermore, infrared detector suppliers science. Despite its age, this book still presents a thor-
typically only check a couple of parameters of each detec- ough insight into the science of infrared detection. Hudson
tor they despatch usually its responsivity and the Noise recognizes that the term detector sensitivity is being
used ambiguously because it sometimes refers to the
John Wiley & Sons Ltd, 2002. signal-to-noise ratio (S/N) and sometimes to the detector
responsivity. This term is not being used in this arti- a 1 W radiant power is incident on the detector. Units of
cle. Instead, other detector parameters, which describe the D* are cm Hz1/2 W1 . Note the use of cm as the unit of
performance of infrared detectors unambiguously, will be length. This is used because a unit of measurement has
utilized. been adopted for D* called the Jones in honor of the man
Responsivity is the first of these parameters. It defines who contributed enormously in this area. The Jones is equal
the ability of an infrared detector to convert infrared radia- to 1 cm Hz1/2 W1 .
tion into an electrical signal. It is the ratio of the electrical The definition of D* makes it a very useful parameter,
output, expressed in volts or amperes, divided by the inci- as pointed out earlier, because it can be used as the first
dent infrared power in watts. However, the selection of an criterion in the comparison of the performance of different
infrared detector on the basis of the highest responsivity infrared detector types. It should be the starting point in
value should be avoided. Although the detector response is every detector selection process. Once a particular detector
an important parameter, it provides no information about type has been chosen, then the NEP of a particular detector
the ability of the detector to measure small signals. This is of this type with a specific active area can be calculated
because there is always some noise present in the output of using equation (1), in order to confirm that this detector
every detector and the presence of this noise can obscure can meet the signal-to-noise requirements of the particular
small signals. What determines the repeatability of a mea- application for which it is aimed.
surement using an infrared detector is not the magnitude of Some care is needed in the comparison of detectors
the signal since this can be set to any level by introducing in terms of D*. The value of D* is dependent on the
the appropriate amplification. Indeed electrical amplifica- measurement conditions. So, D* is written followed by
tion is used with every infrared detector but its main pur- parentheses which define these measurement conditions.
pose is to condition the magnitude of the signal generated The first is always the wavelength in order to take into
so that it can be read or displayed by another device. The account the dependence of the detector responsivity on
presence of the amplifier can only deteriorate the detection wavelength, while the second is the modulation frequency
capability of a detector or, at best, leave it unchanged. What to take into account the dependence of responsivity and
is more important is the S/N, because it determines the type noise power density on the modulation frequency. If the
A4 uncertainty contribution in a measurement. This ratio wavelength is omitted, then the value of D* given refers to
can be increased by a factor of two by doubling the signal or the wavelength of peak response of the detector. The last
by reducing the noise power within the detection bandwidth condition which must be specified with every D* value is
by the same factor. This leads to the introduction of another the field of view of the detector and this is discussed below.
detector parameter far more suitable for describing the abil- The ideal detector should generate no noise. The only
ity of a detector to measure small signals. NEP is the ratio of noise component at the output of an ideal infrared detector
the noise power density (noise power per unit bandwidth) should be the shot noise due to the photons reaching the
at the detector output divided by the detector responsiv-
detector from objects which are at ambient temperature and
ity, both measured at the appropriate modulation frequency.
which are in the field of view of the detector. Although
NEP is the necessary radiant power incident on the detector,
no infrared detector is ideal, in practice it is the shot
which generates a signal equal to the noise power present in
noise due to the thermal background which should be the
a unit bandwidth (units of NEP are W Hz1/2 ). The presence
dominant noise component in a well-designed detector.
of the frequency units confirms the importance of detection
When a detector is operated in this regime it is said to be a
bandwidth in the detection of small signals.
Background Limited Infrared Photodetector (BLIP). This is
Although NEP is a good parameter for the comparison
a condition that some detectors have almost reached after
of particular infrared detectors, it suffers from a drawback.
some years of development effort, and the proximity of the
It has been found both experimentally and theoretically that
operation of a detector to the BLIP condition is a measure
the NEP of most infrared detectors is proportional to the
of how good that detector type is in detecting small signals
square root of their active area.3,5 For this reason Jones6
under the specified operating conditions.
specified another detector parameter, known as the specific
Figure 1 summarizes the D* values of some of the
detectivity, D*, which is independent of detector active area
commercially available infrared detectors when they are
and for this reason refers to detector type rather than to a
specific detector as is the case for NEP. D* is given by: operated at the temperatures indicated. The same figure
also shows the background limited (BLIP) D* for the
A1/2 ideal photovoltaic and photoconductive detectors. All the
D D 1
NEP detector types as well as the calculated background limited
where A is the detector active area. D* is equivalent to the D* values are assumed to view a hemispherical surround,
S/N of a detector of unit area in a unit bandwidth when which is at a temperature of 300 K. The background limited
Detectors for Mid- and Far-infrared Spectrometry: Selection and Use 3
1013
PbS (selected) 300 K
(Hg 0.61 Cd 0.39) Te
x = 0.39 80 K
1012 PbS 300 K Theoretical peak D * for background limited condition

(average) of 300 K, 180 field of view
Photovoltaic Photoconductive
PbTe (selected)
77 K
Hg CdTe (Hg 0.73 Cd 0.27) Te
1011 77 K Ge: Cu (60)
80 K Ge: Au (P-type)
~15 K
D*(,,1)
PbS 77 K 77 K Ge: Cd
In Sb (selected) 25 K Ge: Zn (1)
Ge: Au 77 K 4K
(n-type)
80 K (Hg 0.80 Cd 0.20) Te
1010 80 K
Pb1x SnxTe
0.78 < x < 0.85 77 K
PbSe
InAs 77 K
300 K Pb1x Snx Te Immersed
thermistor
109 PV 0.78 < x < 0.85 77 K
28 C
PbTe 77 K Pb1x SnxTe
(average) 0.78 < x < 0.85 77 K Unimmersed thermistor
28 C
InSb PM E 300 K Ge: Hg 35 K
108
1.5
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
Wavelength / m
Figure 1. D* of a number of commercially available detectors as a function of wavelength. [Reproduced by permission from Wolfe
and Zissis (1989).]
D* has a minimum just beyond 10 m because this is the allow the radiation being measured to pass through without
wavelength at which the spectral radiance of a blackbody vignetting. As an example for the benefits involved, the D*
operating at 300 K has a maximum. Since the temperature of a detector will increase by a factor of 2 when the field
of the thermal background viewed by the detector is around of view of a detector is restricted to 60 (full angle).
300 K, the shot noise due to the background also has a Another approach, which can increase the D* of an
maximum, so the background limited D* has a minimum infrared detector but one which may not be always feasible,
around this wavelength. As the cut-off wavelength of a is to use a cold band-pass filter so that only the part
detector moves to shorter wavelengths, less of the thermal of the thermal background spectrum which falls within
background generated photons lie in the response curve the bandwidth of the cold filter reaches the detector and
of the detector so the shot noise they generate is reduced contributes to noise. The cold filters used for this purpose
and D* increases. For wavelengths longer than 14 m, the are nearly always mounted on the cold shield so they are
number of photons due to the thermal background (and thus also at the detector temperature. However, this method is
the noise they generate) can increase or decrease depending only useful when the wavelength range of the infrared
on the short wavelength cut off of the detector. Furthermore, radiation being detected is limited. The benefits provided
the signal photons in the same amount of incident power by the introduction of cold filters depend critically on the
increase with wavelength so D* values increase again. bandwidth and wavelength of peak transmission of the filter
It is important to stress that the D* values of all the detec- being used, as well as the detector being used. A good
tors shown in Figure 1 refer to a field of view of 180 (2p example of the use of cold filters is the use of short pass
solid angle). It is comparatively easy (and wise) to increase cold filters in front of InSb detectors. While InSb has the
the D* values shown in Figure 1 by reducing the solid highest D* values in the 4 m to 5 m region, its response
angle subtended by the ambient temperature background.7 extends to 1 m, albeit with poorer D* values. The addition
This is achieved by utilizing a cold shield, usually cooled of a short-pass cold filter transmitting below 2.7 m can
to the same temperature as the detector, so that the field easily increase the detector D* by an order of magnitude
of view of the detector is no more than is necessary to and extends the benefits of using an InSb detector in the
1.6 m to 2.5 m region where the availability of large area the smallest permissible active areas should be chosen.
and spatially uniform detectors is limited.8 It is always advantageous to demagnify the image
The response time t of an infrared detector defines how formed on the detector so the radiation collection is
quickly the output of the detector can follow a rapidly maximized for a particular detector active area. This
changing incident signal. Mathematically it specifies how may not always be possible if the detector cannot
quickly the output of the detector rises in response to a support the higher f/number of the converging beam,
step change in the incident signal. More frequently the which results from the demagnification. Benefits also
detector frequency bandwidth is used to convey similar result by optically immersing the detector9 and by
information since the response time in the time domain utilizing a field lens. Optical immersion is related to
is related to the bandwidth in the frequency domain via a the immersion of lenses in microscopy and allows the
Fourier transformation. The response of the detector R(f) effective collection area of a detector to increase. It
at a frequency f is related to the DC response R(0) by: is particularly effective with thin film photoconductor
detectors such as PbS and HgCdTe which can be
Rf D R0 [1 C 2pft2 ]1/2 2 deposited on the plane surface of the lens.
2. Linearity. A detector is said to be linear if its output
The detector bandwidth isp therefore the frequency at which is proportional to the infrared signal being measured.
the response falls to 1/ 2 of the DC response, and for
If this condition is valid then the detector responsiv-
detectors responding exponentially to step changes in the
ity is constant over a range of incident powers, which
incident signal, the bandwidth f is given by:
must accompany the definition of linearity along with
1 the specified tolerance in nonlinearity. It is important
f D 3
2pt to remember that a detector may be inherently lin-
ear but the detection system may be nonlinear due
To complicate matters, some detectors exhibit two different
to the unsuitable choice of support electronics. A
time constants and some others have time constants that
classic example is the use of a photovoltaic detector
are dependent on the operating temperature as well as
with an amplifier or read-out system with high input
being wavelength dependent. The contribution of support
impedance. This makes the output a logarithmic func-
electronics to the speed of response of detection systems
tion of the incident signal being measured. Operating
should never be ignored. Some detectors can be very
them in the reverse bias mode can increase the linear
fast, but effects like stray capacitance in the leads and
range of a photovoltaic detector. Radiometrists have
the amplifier input characteristics can reduce the detection
devised a number of methods for measuring detector
bandwidths by orders of magnitude.
nonlinearity, details of which can be found elsewhere.9
Thermal detectors have slower response times which are
They have also devised methods of correcting for the
rarely faster than below 1 ms. Photon detectors are inher-
ently faster, offering response times of 109 s. In infrared nonlinearity of detectors provided the linearity charac-
radiometry, relatively modest frequency bandwidths are teristics of a detector are fully known.
required, typically less than 1 kHz. The main driving force 3. Spatial uniformity of response. This parameter, along
for increasing the operating frequency is to operate at suffi- with the relatively poor NEP values associated with
ciently high frequencies where the contribution of the 1/f infrared detectors, provides the most serious uncer-
noise diminishes. Furthermore the detector must be able tainty contributions in infrared radiometry.8 Spatial
to respond to the modulator (mechanical chopper), which nonuniformity of response arises because the respon-
forms part of the noise suppression mechanism being uti- sivity of a detector varies from one point to the next
lized. It is important, however, to operate the detector in on its active area. It is defined as the largest devia-
a frequency region in which the detector response is flat tion of the detector response at different points on the
with respect to changing frequency to ensure that drifts in active area expressed as a percentage of the maximum
modulation frequency are not interpreted as changes in the response. Causes of spatial nonuniformity include poor
detector response. Alternatively, the operator can invest in fabrication, imperfections present on the active area,
a chopper whose operating frequency is actively controlled variable thickness of the detector element, relatively
by an external oscillator. high series resistance compared to shunt resistance
Other parameters that can contribute to or even determine and the proximity/positioning of the electrodes (tem-
the selection of an infrared detector include the following: perature measuring devices in thermal detectors) to
different parts of the active area. Some detectors even
1. Active area. Since the NEP of a detector is proportional exhibit different relative spectral response profiles at
to the square root of the active area,3,5 detectors with different points on their active areas.8 When a changing
distribution of irradiance illuminates a spatially nonuni-

Photon
form detector, some variations in the detector output detector
can easily (but wrongly) be interpreted as changes in
the total power incident on the detector.
4. Ageing effects. Ageing is the apparent change in the Thermal detector
detector responsivity over a period of time or after it
Response
has been exposed to high and /or short wavelength radi-
ation. Ageing in photomultipliers is well documented
but ageing in infrared detectors is less widely publi-
cized. Detectors should never be operated outside the
specified conditions but our experience indicates that
photoconductive thin film detectors exhibit the high-
est ageing characteristics and this is likely to originate
from ohmic heating generated by the large biasing cur- Wavelength
rents passing through their thin film structures. Figure 2. Spectral response of ideal thermal and photon detectors.
There are a number of further detector parameters which
This effect is mainly due to the fact that the absorptivity of
have to be taken into consideration when selecting the
a material depends partly on the apparent roughness relative
most appropriate detector for a particular application. These
include the operating temperature, the temperature coeffi- to the incident wavelength. As the wavelength increases the
cient of response, the shunt resistance or dark resistance, the apparent roughness decreases, resulting in the characteris-
series resistance (for large area detectors) and the capaci- tic deterioration of the absorption of black coatings with
tance. Most of these parameters are revisited in Sections 3 wavelengths.10 The presence of the window in front of
and 4. a thermal detector will also define an upper and a lower
wavelength transmission limit, but the window transmis-
sion effects will be ignored in this article as they receive
3 MAJOR DETECTOR CATEGORIES full treatment elsewhere.11 We shall concentrate briefly on
the parameters which determine the response of thermal
Infrared detectors are divided into two main groups: ther- detectors. Apart from the temperature coefficient of the
mal detectors and photon detectors. Thermal detectors rely physical process on which the thermal detector is based, the
on temperature changes generated by the incident radiation. responsivity of such a detector can be increased by increas-
They therefore utilize elements which have some property ing the thermal resistance (degree of isolation) between the
that is strongly dependent on temperature. The incident transducer and its mount.12 However this will also result in
infrared radiation increases the temperature of the element the deterioration of the temporal response of the detector.
and this is detected by the change in the temperature- Here, as on many other occasions, a compromise has to be
dependent property. The operation of the photon detectors reached by the design engineer. Although thermal detectors
relies on a change in the mobility of charge carriers, which have inherently slow response compared to photon detec-
is generated by the incident radiation. Brief summaries tors, they can still be sourced with modulation bandwidths
of the main characteristics of thermal and photon detec- extending to 1 kHz but this is at the expense of lower D*.
tors are given below along with a comparison of their Another parameter which the design engineer has at his
characteristics. disposal is the heat capacity of the transducer. Decreasing
the heat capacity increases the detector responsivity and
reduces the temporal response of the detector. Thermal
3.1 Thermal detectors capacity is directly related to physical size so it is obvious
that there are limits on how low the thermal capacity of a
Thermal detectors respond to temperature changes, which transducer can be made.
are directly proportional to the incident radiant power, so Finally, to conclude this section, we have a short dis-
the response of an ideal thermal detector per incident watt cussion on black coatings. Apart from the requirement for
should be constant at all wavelengths (Figure 2). However, a high (and spectrally flat) absorptivity, the black coating
in practice this is never the case because the absorptivity must have a low thermal mass in order to keep the heat
of black (absorptive) coatings with which the transducer capacity of the transducer low (see previous paragraph).
elements are coated in order to enhance their responsivity How low the thermal mass of the black coating must be
always deteriorates as the incident wavelength increases.10 depends on the thermal mass of the transducer itself since
the thermal mass of the black coating must be small relative aimed at relatively low-technology instruments such as gas
to that of the transducer. analyzers, fire detectors and intruder alarms. Finally some
Commercially available thermal detectors rely on two thermal detectors operate without any black coatings relying
main types of black coatings. The first is in the form of on the natural absorptivity of the transducer itself. The main
paint and is applied like paint. These types of coating benefit resulting from this is to keep the thermal capacity
are cheap and easy to produce but their thermal mass is of the transducer low.
relatively high and their absorptivity relatively low (total
reflectance values of a few percent) so they are address-
ing less demanding applications where their low cost and 3.2 Photon detectors
robustness are major advantages. The second type of black
Photon detectors operate by increasing the mobility of
coating is characterized by very high porosity, which means
charge carriers. In a semiconductor lattice, electrons will
low thermal mass, high absorptivities and relatively flat
populate the valence band whereas the conduction band,
spectral response. Harris13 gives an excellent review of the
which is separated by an energy gap Eg , is relatively empty.
metal-blacks, which come under this category. They include
A photon with energy higher than Eg can excite an electron
the gold-black and silver-black coatings, which are made in
from the valence to the conduction band where it is free to
evaporation chambers but not in vacuum. Instead, a small
move under the influence of an electric field. If the electric
pressure of nitrogen gas is retained so that the mean free
field is provided externally then the detector is said to
path of the metal atoms is small. This results in a gradual
operate in a photoconductive mode. When the electric field
loss of their kinetic energy so by the time they reach the
is provided internally then the detector is said to operate in
target they have no kinetic energy and just deposit on the
a photovoltaic mode.
substrate.13 This results in fibrous coatings that have high
The spectral response of photon detectors is character-
and relatively flat absorptivity over very wide wavelength
ized by a gradual increase in responsivity with increasing
regions. They represent the best black coatings on offer
wavelength, followed by a sharp drop above a certain wave-
for thermal detectors. Yet another little used black coating
length, as shown in Figure 2. The increase in the photon
is the platinum black,10 which is made by a wet fabrica-
detector response is due to the increasing number of photons
tion method. This black differs from gold black in that its
in a given amount of energy as the wavelength increases.
structure is not fibrous but dendritic, but our experience
Since each photon (assuming a quantum yield of unity)
has shown it to be robust, being able to survive cryo-
generates a free electron, the charge generated for a Watt
genic temperatures, and to have a high and spectrally flat
absorptivity.10 Figure 3 shows the relative spectral response of power incident on the detector will increase with wave-
of a pyroelectric detector coated with a platinum-black coat- length. For wavelengths above a certain threshold, photons
ing showing that its absorptivity changes by less than 5% do not have sufficient energy to excite electrons to the
in the 1 m to 13 m wavelength range. conduction band so no free charge carriers will be gen-
Although the discussion has concentrated on high absorp- erated. The cut-off wavelength in micrometers of a photon
tivity coatings with flat response, thermal detectors are detector is related to the band gap energy, expressed in
also available with multi-layer dielectric coatings, which electronvolts, by:
have good absorbance over certain wavelength regions such 1.24
lcutoff D 4
as the 8 m to 12 m band. However, these detectors are Eg
1.02
1.01
Relative response
1.00
0.99
0.98
0.97
0.96
0.95
0.94
1 2 3 4 5 6 7 8 9 10 11 12 13
Wavelength / m
Figure 3. Relative spectral response of a Pt-black coated pyroelectric detector in the 1 m to 13 m region.
Photon detectors can be divided into two main groups. do not have to be cooled to very low temperatures to sup-
These are the photoconductive and the photovoltaic detec- press thermally excited electrons. In fact they have the
tors and the main characteristics of each group are summa- highest D* values for a particular operating temperature.
rized below. Extrinsic photoconductors suffer from limited solubility of
the dopants in the host lattice (usually germanium and sil-
3.2.1 Photoconductive photon detectors icon). The low dopant concentrations result in absorption
coefficients which are much lower than those of intrinsic
Photoconductive detectors rely on the increase in their photoconductors. To overcome this, extrinsic photoconduc-
electrical conductivity when they are being illuminated tors are made to be very thick (many millimeters) in order to
by infrared radiation. Two types of photoconductors are ensure adequate absorption of the radiation being detected.
available. Intrinsic photoconductors are materials that have Extrinsic photoconductors require cooling down to 4 K to
relatively low energy band gaps which are smaller than have D* values comparable to those of intrinsic photo-
the energy of the infrared photons being detected. The conductors such as HgCdTe operating at 77 K. The over-
incident photon energy can then excite an electron from the whelming advantages of intrinsic photoconductors mean
valence band into the conduction band so the creation of the that extrinsic photoconductors are used exclusively for the
electron and the hole, which is also created in the valence detection of infrared radiation of wavelengths longer than
band, increase the electrical conductivity of the lattice and 20 m where intrinsic photoconductors are not available.
this is measured with the aid of an external electric field. However, extrinsic photoconductors have lower dielectric
However, there are few materials with energy band gaps
constants, which means that they are considerably faster
of less than 0.1 eV so the detection of infrared photons of
than intrinsic photoconductors and are also finding applica-
energies much lower than 0.1 eV cannot be accomplished
tions where speed of response is important.
by intrinsic photoconductors. Extrinsic photoconductors, on
Extrinsic photoconductors have benefited from some
the other hand, have energy band gaps which are relatively
recent improvements in their performance. Their dopant
large, but the addition of an impurity into the lattice
concentration has been increased to prevent the build-up
introduces extra energy levels in the energy gap between the
of positive charges that are left behind by the excited elec-
valence and conduction bands where the impurity dopants
trons. These extrinsic photoconductors are called Blocked
reside. If the impurity has more valence electrons than
Impurity Band (BIB) detectors and are addressing demand-
are needed to bond it to the semiconductor lattice, it can
ing applications in far-infrared astronomy.
donate one or more of the spare electrons to donor
Intrinsic and extrinsic photoconductive detectors have
bands which lie just below the conduction band. If the
square or rectangular active areas because they require
impurity has fewer valence electrons than are needed for
biasing and these shapes aid the generation of uniform
bonding to the host lattice, it can accept electrons, which
electric fields across the active areas of these detectors.
are provided from the valence band so holes are created.
These can be considered as positively charged entities that
can move under the influence of an external electric field, 3.2.2 Noise in photoconductor detectors
thus contributing to the conductivity of the lattice. The holes
reside in an acceptor band, which lies just above the For photoconductors operating at low frequencies, the 1/f
valence band. The gaps between the donor and acceptor noise component dominates.5,14 The effects of the 1/f
bands and the conduction and valence bands, respectively, noise can be avoided by operating at frequencies typically
are typically less than a few tenths of an electronvolt so higher than 1 kHz. Furthermore, the noise power density
at room temperature the thermal energy of the lattice is due to the 1/f noise is proportional to the square of the
sufficient to ionize the impurity bands. It is for this reason current flowing through the detector, so reducing the biasing
that the temperature of extrinsic photoconductors has to current can reduce the effects of this noise component. At
be lowered sufficiently, usually down to 4 K, to prevent intermediate frequencies a noise contribution referred to as
5,14
the ionization of the impurity bands. At low temperatures the Generation-Recombination (G-R) noise takes p over.
the impurity bands can be excited by the incident infrared This noise component contributes a factor of 2 to the
photons so the material exhibits extrinsic photoconduction. noise power of a photoconductor compared to that of a
The position of the donor and acceptor impurity bands can photovoltaic detector and is the origin of the difference
be tailored by the choice of impurity and this has resulted in between the theoretical BLIP curves of the two detector
the development of a number of extrinsic photoconductors, types shown in Figure 1. The power density of the G-R
which respond to infrared radiation longer than 100 m. noise decreases at higher frequencies5,14 and the thermal
Intrinsic photoconductors have very large absorption or Johnson noise becomes the dominant noise component.
coefficients and low intrinsic carrier concentrations so they Its noise power density is constant at all frequencies so it
is an example of white noise. The thermal noise current absorbed in the depletion region, the electronhole pairs
Inoise due to a conductor of resistance R operated at absolute which are created are separated by the electric field of the
temperature T in a bandwidth f is given by: p-n junction itself so current can be generated.
Figure 4 highlights the different modes under which a
4kTf 1/2 photovoltaic detector can be operated. The photovoltaic
Inoise D 5
R detector can be operated without any biasing by either mea-
where k is Boltzmanns constant. In calculating the total suring the short circuit current across the detector terminals
noise due to a photoconductor, one must not forget the or by measuring the open circuit voltage. These two con-
thermal noise due to the load resistor, which can be cal- ditions are indicated in Figure 4 by load lines 1 and 2,
culated from the same equation by substituting the load respectively. The open circuit voltage mode indicated by
resistance and the absolute temperature of the load resistor, load line 2 should be avoided because the generated voltage
as this is likely to be different from the temperature of the changes logarithmically with the incident infrared radia-
detector. The total noise at the output of a photoconduction strength. The short circuit current measurement is far
tor can be calculated by summing all noise components in more attractive because it provides a current output that
quadrature. The reader is referred to the book by Dereniak varies linearly with the incident radiation signal. Operating
and Crowe14 which provides a full treatment of the noise a photovoltaic detector under short circuit conditions can
sources associated with photoconductive infrared detectors be accomplished relatively easily with the aid of a trans-
and discusses the conditions which must be satisfied for the impedance amplifier. The amplifier will maintain its two
BLIP operation of an infrared photoconductive detector. input terminals at the same voltage but will not sink any cur-
rent so the photocurrent generated will be diverted through
the feedback resistor while the detector will experience a
3.2.3 Photovoltaic detectors short circuit across its terminals.
Photovoltaic detectors are formed at the boundary when two Photovoltaic detectors can also be operated in a reversed
semiconductors, one with excess donors (n-type) and one bias mode as shown by load line 3 in Figure 4. This mode
with excess acceptors (p-type), are brought into electrical of operation is referred to as the photoconductive mode,
contact. Because of the migration of free electrons to the although this is the wrong nomenclature, as was pointed out
p-type and free holes to the n-type material, there is a by Geist,15 because it differs significantly from the opera-
neutralization of the charges over a region at the boundary tion of photoconductive detectors. This mode of operation is
of the two materials called the depletion region. When attractive because it can extend the linear range of operation
radiation of photon energy higher than the band gap is of a photovoltaic detector and it can improve its temporal
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.05 0.10 0.15 0.20 0.25
Dark current 2
D
R RL = 10 k
20
E1
Photocurrent (A)
Irradiance
E2
40
E3
60
E4
80
E5
100
RL = 500
3 1
Reverse-biased mode Unbiased mode
V S /R L
Voltage (V)
Figure 4. Current to voltage characteristics for a photovoltaic detector. E0 to E5 represent IV profiles for increasing levels of
irradiance.9
response because it reduces the junction capacitance. How- the noise components in a photovoltaic detector/amplifier
ever, infrared photovoltaic detectors are rarely operated combination and identify the conditions which must be
under biased conditions because they have rather low shunt satisfied for BLIP operation.
resistance values, ranging from a few k for a 5 mm dia-
meter InSb detector operating at 77 K to about 10 for a 3.2.5 Other infrared photon detectors
1 mm diameter HgCdTe detector operating at 77 K whose
cut off wavelength is around 12 m, so biasing results in Two further categories of infrared detectors exist which
very large dark currents. differ in their operation from photoconductive and photo-
voltaic detectors. These are the photoemissive or Schottky-
barrier detectors, the best example of which is the PtSi
3.2.4 Noise in photovoltaic detectors detector. The second is the Quantum Well Infrared Photode-
The main sources of noise in a photovoltaic detector5,14 are tector (QWIP) whose operation resembles that of extrinsic
the shot noise due to the thermal background, the shot noise photoconductors, although no impurities are added to gener-
due to the photocurrent generated (signal), the shot noise ate the infrared absorption. The operation of these infrared
due to the reverse saturation current, the 1/f noise that is detector families is summarized in Section 4 in association
present at low frequencies and the thermal noise due to the with PtSi and quantum well detectors, respectively.
shunt resistance of the detector. The last contribution can
be calculated using equation (5) by substituting the shunt
resistance of the detector and the appropriate temperature.
4 SUMMARY OF INFRARED DETECTOR
The shunt resistance is the slope of the detector VI curve CHARACTERISTICS
at V D 0 V under dark conditions. Photovoltaic detectors
are operated in association with transimpedance amplifiers The characteristics of the most important infrared detector
as discussed earlier, so the amplifier noise contributions technologies are summarized below. Further information
must be included in the noise budget. There will therefore can be sourced from manufacturers brochures and reviews
be a noise component due to the thermal noise of the of the mid- and far-infrared detectors.5,16,17
feedback resistor (it can be calculated with the aid of
equation 5) and the amplifier noise, which consists of two
noise contributions the voltage noise density and current 4.1 Thermal detectors
noise density both of which are tabulated in the amplifier
data sheet. It is worth remembering that the amplifier input 4.1.1 The thermocouple and thermopile
offset voltage Voffset and input bias current Ibias determine A thermocouple relies on the thermoelectric effect to detect
the dark current via the relationship: the incident radiation. It is stable but rather fragile and
Voffset has a relatively slow response time which is longer than
Idark D Ibias C 6 10 ms. The specific detectivity of a thermocouple can reach
Rshunt
109 cm Hz1/2 W1 for time constants of 20 ms. Metal black
where Rshunt is the shunt resistance of the detector. This coatings are used to coat thermocouples to increase their
relationship is given to stress that the dark current values responsivity and extend the spectral response to above
tabulated by the manufacturers in their brochures are fairly 50 m. The signal generated by a thermocouple is small
meaningless because the true value of the dark current is so the performance of a detection system based on a ther-
a function of amplifier parameters as well as the detector mocouple relies heavily on the quality of the amplifier.
shunt resistance. Since the dark current also generates Although thermocouples can operate under DC condi-
shot noise, it is clear that the amplifier Voffset and Ibias tions, drifts associated with DC amplifiers recommend AC
contribute to the total noise indirectly. The choice of the operation.
amplifier is therefore important. When using a detector with Some benefits can be gained by connecting a number of
a high shunt resistance value, it is preferable to choose thermocouples in series, resulting in the well-known ther-
an amplifier with low input bias current (see equation 6). mopile. Early thermopiles used fine metallic wires and a
Infrared detectors have low shunt resistance values so an blackened gold foil about 0.5 m thick as the absorber.
amplifier with low input offset voltage should be selected. More recently evaporated thin film thermopiles have been
Finally, a rule of thumb for limiting the noise generated adopted which offer superior robustness. The time con-
by a photovoltaic detector/amplifier combination is that the stant of evaporated thermopiles ranges from 4 to 50 ms17
feedback resistance should not exceed the detector shunt depending on the type and thickness of the black coating
resistance. Dereniak and Crowe14 give a full treatment of and thermopile surface. Noise is dominated by the Johnson
Figure 5. Spatial uniformity of response of a thin film thermopile at 1.55 m.
noise contribution. Their D* can exceed 108 cm Hz1/2 W1 of a large area silver-black coated thermopile over its 6 mm
for a time constant of 20 ms. However, Stevens2 points out diameter active area is shown in Figure 5 and a character-
p
that, for some thermopiles, the A dependence of noise istic trough near its center is clearly visible. The response
does not apply. D* can therefore depend on active area. is higher near the edges because the thermometers moni-
Furthermore, the D* depends on the thermopile design2 as toring the temperature of the target are attached around the
well as the time constant. The thermopile design, active circumference of the active area. Heat due to the radiation
area and time constant should all be specified when spec- incident near the perimeter of the thermopile can be cou-
ifying the D* of a thermopile. D* values in the range pled to the thermometers much more efficiently compared
108 cm Hz1/2 W1 to 109 cm Hz1/2 W1 for an evaporated to the heat generated near the center.
thermopile operating at ambient temperature with a time
constant in the region of 4 to 30 ms are quite common,5
4.1.2 The bolometer
although D* values as low as 106 cm Hz1/2 W1 (nonevap-
orated thermopiles) and as high as 4 109 cm Hz1/2 W1 The bolometer utilizes a resistive element made of a mate-
have appeared in the literature.2 rial that has a high temperature coefficient of resistance.
Thermopiles are available in a variety of active area Infrared radiation incident on the transducer increases its
sizes ranging from circular shapes up to 6 mm in diame- temperature, resulting in a change of its resistance. Early
ter, to square and rectangular shapes, which accommodate bolometers used thin metal foils less than 100 m thick as
the shape of exit slits of monochromators. They are also their transducers. When coated with metal blacks they were
available in matched pairs to compensate for ambient tem- able to achieve D* values exceeding 108 cm Hz1/2 W1 and
perature changes. response times down to 102 s. They operate at ambient
One of the drawbacks of using thermopile detectors in temperature but they are fragile and have low resistance
infrared radiometry is their nonuniform spatial response. values. Thermistor bolometers quickly superseded metal
Some commercially available 3 mm 3 mm thermopiles foil bolometers because they utilize thermistor elements
have been found to exhibit spatial nonuniformities of made out of a mixture of nickel, manganese and cobalt
response of many tens of percent. The normalized response oxides which has a temperature coefficient of resistance of
4% C1 . They are rugged, resistant to vibration and can connection was made, a modulated voltage difference will
operate at ambient temperature. Their D* is proportional to appear across the crystal faces. The current or voltage
the square root of their time constant so faster response pulses can easily be measured using standard electrical
is accompanied by a reduction in D*. At low frequen- techniques.
cies, 1/f noise dominates so their maximum D* values One of the most widely used pyroelectric detectors is
occur around 20 Hz. They must be operated in a bridge based on triglycine sulfate (TGS) crystal. It is a colorless
circuit to overcome the changes of their resistance with crystal with a Curie point (the temperature at which the
ambient temperature. The D* of a thermistor bolometer crystal loses its pyroelectric characteristics) of only 49 C.
can be increased by nearly a factor of 4 by optically Furthermore, TGS crystals are hygroscopic so they have
immersing it using a germanium hemispherical lens. The to be operated in sealed containers behind windows, which
superconducting bolometer utilizes the large change in the can limit their wavelength range of response. Despite these
resistance, which occurs when a material becomes super- drawbacks, TGS crystals are used extensively because they
conducting. Although the temperature coefficient of resis- offer some of the highest D* values compared to other
tance is equivalent to 5000% C1 , the proper operation pyroelectric detectors, particularly at high frequencies. A
of the superconducting bolometer requires the transducer number of variant crystals, such as deuterated triglycine
to remain within the transition range of the material being sulfate (DTGS) and deuterated l-alanine doped triglycine
used. This means controlling its temperature to better than sulfate (DLATGS), have also been developed with higher
103 K. This is difficult to achieve in practice, restricting Curie points of around 60 C and 74 C.
the application of the superconducting bolometers to the Lithium tantalate (LiTaO3 ) has a lower pyroelectric
laboratory. Carbon bolometers rely on the large temper- coefficient18 compared to TGS but its Curie point is at
ature coefficient of resistance of carbon below 30 K. The 620 C. This means that LiTaO3 detectors can operate over
resistance of the carbon transducer increases exponentially a wide range of ambient temperatures, although measure-
as the inverse of the temperature so best responsivities ments at the UK National Physical Laboratory (NPL) indi-
are obtained for very low temperatures. Carbon bolome- cate that its temperature coefficient of response at 20 C is
ters usually operate at 2.2 K. However, their performance relatively high at around 0.2% C1 . LiTaO3 crystals are
is inferior to that of the germanium bolometer. The germa- not hygroscopic and can be operated without a window.
nium bolometer is a development of the far-infrared free Most commercially available LiTaO3 detectors have crys-
carrier absorption detector described later in this section. It tal thickness of around 50 m and are available with active
can be used with a black absorber or rely on the absorp- areas up to 3 mm 3 mm although detectors with a 10 mm
tion of a germanium lattice heavily doped with gallium. diameter active area can be fabricated as specials. Pyro-
The Ge(Ga) bolometer operates like a photoconductor and electric ceramic materials such as zirconate titanate (PTZ)
must operate at the liquid helium temperature. Under some have also found applications mainly because the material
conditions further benefits can arise by operating it at the is relatively easy to fabricate. The choice of material for a
lowest temperature possible, and operation at 2.2 K is rec- pyroelectric detector will depend on the size of the active
ommended. The D* of the Ge(Ga) bolometer can be as area required, the modulation frequency being detected and
high as 3 1013 cm Hz1/2 W1 at relatively low frequen- the range of ambient temperatures over which the detector
cies (20 Hz). It offers a very broad spectral response which must operate. Generally TGS pyroelectric detectors offer a
extends to beyond wavelengths of 1000 m. good compromise except when the ambient temperature is
likely to rise towards its Curie point or the presence of a
window is not acceptable.
4.1.3 Pyroelectric detectors
All pyroelectric crystals are also piezoelectric to some
Pyroelectric detectors rely on the property of certain crystals extent so they are very efficient microphones and accelero-
to generate electric charge on their surface when their meters. However, the authors experience indicates that
temperature is changed. This charge will be neutralized the way the pyroelectric crystal is mounted plays a crit-
soon after its formation due to charge leakage to the ical role in suppressing microphonics. Crystal mounting
surroundings. However, soon after the charge is generated, along with the uniformity of the thickness of the crystal
it can be detected as a voltage across the two pyroelectric and the quality of the black coating are the most important
crystal faces. Alternatively, if the two crystal faces are parameters that govern the degree of spatial nonuniformity
electrically connected, an electric current will flow briefly. of response encountered in pyroelectric detectors. Figure 6
By modulating the radiation incident on a pyroelectric shows the spatial uniformity of response of a LiTaO3 detec-
crystal, a series of electric current pulses will flow through tor exhibiting a 30% variation in its spatial response due
the external circuit or, in the case where no electrical entirely to the variation of the crystal thickness. Finally,
Figure 6. Spatial uniformity of response of a LiTaO3 pyroelectric detector at 1.55 mm. The nonuniformity arises due to the variation
of the crystal thickness.
the fact that these detectors only respond to changes in the infrared radiation is transferred to the membrane and then
power of the incident radiation need not be considered as a to the gas. The resulting increase in gas pressure distorts
drawback because modulation of the incident beam can be a diaphragm, the position/shape of which is monitored
coupled to noise suppression methods such as phase sensi- optically.16 The performance of the Golay cell as an infrared
tive detection. Unfortunately optimum D* values for most detector is surprisingly good considering its method of oper-
pyroelectric detectors occur for relatively low modulation ation. Its D* is within an order of magnitude of that of an
frequencies, typically between 10 and 20 Hz. ideal thermal detector.20,21 Its spectral response is governed
DTGS detectors are installed in most low-end Fourier by the window through which radiation enters the gas cell
transform infrared spectrometers. In these instruments, the and can extend to wavelengths of many millimeters. At
infrared beam is typically modulated at frequencies of long wavelengths the absorptivity of the absorbing mem-
between 120 Hz and 1.2 kHz well above the range in brane contributes to a lower responsivity. It is, however, a
which any pyroelectric detector has its highest response. rather delicate detector, sensitive to vibration, microphony
However, since DTGS is the only thermal detector oper- and air draughts. Further information on the operation of
ating at or slightly below ambient temperature that is Golay cells and the types of windows being used with them
sensitive enough to be used at modulation frequencies up can be found in the book by Kimmitt.16
to 1 or 2 kHz, it serves its purpose well. It may be noted
that, since the D* of DTGS detectors improves signifi-
cantly as the modulation frequency is decreased, reducing 4.2 Photon detectors
the optical velocity of the interferometer results in greatly
improved performance, provided that the dynamic range of The remit of this article is to deal with detectors respond-
the analog-to-digital converter is not exceeded. ing to wavenumbers lower than 7000 cm1 . This range
The Golay cell pneumatic detector is a broad band, room includes the long wavelength part of the response of ger-
temperature infrared detector which was developed in the manium detectors. Germanium photovoltaic detectors rely
mid 1940s.19 It consists of a small cavity that contains a on an indirect band gap transition, so their long wave-
low pressure inert gas, usually xenon. The cavity also con- length cut off (which is temperature dependent) decays
tains an absorbing membrane so the power of the incident to zero much more slowly than direct band gap detectors
1014 1.4
Absolute responsivity (A W1)

1.2
K)
D * (peak, 300 Hz,1 Hz) (cm Hz1/2 W1)
77
D( 1.0
J16
0.8
1013 0.6
0.4
0.2
0 C)
(3
16TE2 0.0
1.0
J 0.8 0.9 1.1 1.2 1.3 1.4 1.5 1.6 1.7
1012 Wavelength / m
Figure 8. Absolute responsivity of a 5 mm diameter InGaAs
detector.
C)
(25
J16 operating at room temperature. InGaAs detectors of this
type are widely available with diameters ranging from
1011
0.8 1.0 1.2 1.4 1.6 1.8 50 m to 5 mm. One supplier even offers a 10 mm 10 mm
Wavelength / m active area detector but our evaluation has shown it to have
Figure 7. D* of germanium detectors at three different temper- relatively poor performance8 despite a high price tag. How-
atures. [Reproduced by permission from EG&G Judson, Infrared ever, even with this well developed detector technology,
Detectors (1994).] the best detectors still have to be identified by selection.
One of the 5 mm diameter active area InGaAs detectors
such as indium gallium arsenide (InGaAs). The response evaluated even exhibited relative spectral response fluctu-
of germanium detectors covers the range from the visi- ations across its active area.8 Generally InGaAs detectors
ble to nearly 1.8 m. Germanium detectors have inherently have high quantum yields and their responsivities approach
low shunt resistance and a relatively large dark current at 1.2 A W1 at 1.55 m. They can be selected with spatial
300 K (see equation 6). The shunt resistance increases at nonuniformities of response as low as 0.5% even in the
lower temperatures, and at approximately 60 C the noise 5 mm diameter format and they have a low temperature
performance of germanium detectors becomes comparable coefficient of response below 1.6 m and shunt resistance
to that of InGaAs detectors. Figure 7 shows the improve- values well in excess of 1 M even for the 5 mm diameter
ment in D* which is gained as the operating temperature devices. InGaAs detectors exhibit nonlinearities of less than
of a germanium detector decreases. Note also the shift of 0.1% when operated in a short circuit mode for incident
the cut off wavelength toward longer wavelengths as the powers in excess of 1 mW at 1.55 m. The high power limit
temperature decreases. Germanium detectors are available of the linear range increases further by operating InGaAs
in very large sizes (over 10 mm in diameter). Their time detectors in a reverse biasing mode.
constant increases with active area but for small active area Extended InGaAs detectors are only available with active
devices the time constant for PIN photodiodes is below areas smaller than 2 mm in diameter. They have low
109 s. At 300 K the performance of germanium detectors shunt resistance and therefore relatively high dark currents.
is inferior to that of InGaAs in almost every respect except Thermoelectric cooling improves their performance and is
the maximum active areas available and cost. recommended since it provides temperature stabilization.
The band gap of InGaAs can be controlled by adjusting Extended InGaAs detectors with cut off wavelengths at
the relative concentration of indium and gallium, and this 1.9 m, 2.2 m and 2.6 m are readily available from a num-
has resulted in the development of InGaAs detectors whose ber of vendors.
cut off wavelength varies, in steps, from 1.6 to 2.6 m. They The lead sulfide (PbS) intrinsic photoconductor detector
are direct band gap detectors so their spectral response was one of the first infrared detectors to be used and still
is characterized by a sharp cut off. They can be conve- finds applications today in the 1 m to 3 m wavelength
niently divided into those responding below 1.6 m and range. The performance of PbS detectors was reviewed by
extended InGaAs detectors. The former have been exten- Johnson.22 It is manufactured by depositing very thin films,
sively developed to address the fiber-optic communications typically 1 m thick, on electrically insulating substrates
needs at 1.3 m and 1.55 m and provide excellent perfor- such as sapphire. The deposition can be by wet chemical
mance even at ambient temperature. Figure 8 shows the deposition or by evaporation. They are available with large
absolute responsivity of a 5 mm diameter InGaAs detector active areas (up to 10 mm by 10 mm) and are relatively
1012 detectors are less attractive in a nonmatrix format, because

the spectral response of PtSi has a maximum at short
wavelengths (1 m) and decreases monotonically to zero at
5 m. The response is higher at shorter wavelengths because
1011 they have higher photon energies and are therefore more
likely to excite the electrons sufficiently so the probability
D * (pk, 1000, 1)
(cm Hz1/2 W1)
of tunnelling through the potential barrier increases.24 PtSi

298 K 208 K 77 K
detectors suffer from inherently low quantum efficiencies,
1010 which can be lower than 1% above 4 m. Nevertheless
they had considerable success in two-dimensional array
form,17 mainly because of the ease of fabrication and
high manufacturing yields compared to alternative detector
109 matrix technologies. Single large area PtSi detectors were
evaluated at the NPL because they were supposed to have
good spatial uniformity of response. However, even at 77 K,
the series resistance was high enough compared to the shunt
108 resistance to generate a reduction in the detector response
1 2 3 4 5 near its center, while the maximum response was around
Wavelength / m the circumference, near the collecting electrode.
Figure 9. D* of a PbS photoconductive detector at three different Other Schottky barrier detectors such as the IrSi are under
temperatures for a 1 kHz modulation frequency. [Reproduced by development because they offer the possibility of operation
permission from Judson Technologies, LLC (1993).] at longer wavelengths.17
The lead selenide (PbSe) intrinsic photoconductor has
cheap. They can be operated at room temperature but they a band gap energy of 0.25 eV so its response extends
require temperature stabilization because they exhibit a to about 6.5 m. Its fabrication process is similar to that
high temperature coefficient of response, typically many of PbS, and active areas as large as 10 mm 10 mm are
percent per C. PbS detectors used to be regarded as readily available. Although D* increases from less than
unstable by pyrometer manufacturers.23 Their D* value 109 cm Hz1/2 W1 to 1010 cm Hz1/2 W1 on cooling from
improves considerably by cooling to 193 K. Cooling to 293 K to 193 K no further benefits are gained in D* by
77 K reduces the D* but extends the response range of cooling to 77 K. The time constant is faster than that of PbS
this detector to beyond 4 m (see Figure 9). They do suffer detectors, being about 5 s at 293 K but increases to 50 s at
from some disadvantages including a relatively slow time 193 K.22 PbSe detectors have a high temperature coefficient
constant, typically 1 ms, which deteriorates with cooling. of response (3% per K) so active control of the detector
Faster operation is possible at the expense of response. They temperature is required.23 Their output exhibits drift after
also suffer from poor spatial nonuniformity of response exposure to UV or visible radiation.
which is well over 10%, with some detectors exhibiting Indium antimonide (InSb) photovoltaic detectors benefit
nonuniformities in excess of 50%. from the location of their long wavelength cut-off, which
Indium arsenide (InAs) photovoltaic detectors are intrin- extends to approximately 5.5 m which is just beyond
sic direct band gap detectors whose response extends from the 3 m to 5 m atmospheric window. Photovoltaic InSb
1 m to about 3.2 m. When they are operated at 77 K their detectors operating at 77 K offer BLIP performance and
D* value toward the long wavelength edge is the best avail- the highest D* values, approaching 1011 cm Hz1/2 W1 in
able, approaching 1012 cm Hz1/2 W1 around 3 m. They the 4 m to 5 m range with a 180 field of view. Figure 10
have frequency bandwidths in excess of 1 MHz when they shows the D* of an InSb detector operating at 77 K as
are operated with load resistors of low resistance values. a function of wavelength for 180 and 60 field of view.
However, InAs is a relatively fragile material and InAs They are available in a variety of active area sizes from
detectors are only available with active area sizes of 2 mm below 0.1 mm diameter to an impressive 7 mm diameter.
in diameter or smaller. They offer good spatial uniformity of response even in the
Platinum silicide (PtSi) is an example of a Schottky 7 mm diameter format and good linearity characteristics.
barrier detector. PtSi offers advantages in the fabrication Their shunt resistance increases with decreasing active area.
of detector matrices because it can be interfaced directly At 77 K the shunt resistance of a 1 mm diameter InSb
to the readout electronics. PtSi detector matrices with as detector is typically 500 k but it decreases to 10 k at
many as 1024 1024 pixels are available.17 However, PtSi 60 C and to a few ohms at ambient temperature. Their
1012 detectors at shorter wavelengths because by changing the

er
filt
D* (,1 kHz, 1 Hz) (cm Hz1/2 W1)
ld
composition ratio of Hg and Cd the resulting compound
co has band gap energies which can vary from above 0.6 eV
8
P2 er
S filt to below 0.06 eV. HgCdTe detectors are grown using epi-
ith old
w
3 5c taxial methods on insulating substrates. HgCdTe detectors
SP 60
wi
th V= can be fabricated as a photoconductor (without a junction)
FO
1011 and as a photovoltaic detector. HgCdTe intrinsic photocon-
180
V=
FO ductive detectors have square or rectangular active areas to
ensure uniform bias current distribution over the active area.
Detector sizes vary from 50 m 50 m to 4 mm 4 mm
active areas. Because the dark resistance of these detectors
Detector temp. = 77 K
is low even at 77 K, they can be sourced in a split area or
1010 meander configuration. This increases the effective length
1 2 3 4 5 6
Wavelength / m and halves the width resulting in higher dark resistance
values. The manufacturing process allows the fabrication
Figure 10. D* of an InSb detector operating at 77 K as a function
of wavelength. [Reproduced by permission from EG&G Judson, of custom shapes and sizes relatively easily. Their respon-
Infrared Detectors (1994).] sivity and D* increase with increasing bias current but the
ohmic heating of the detector due to the bias current limits
good overall performance at 77 K recommends them for the responsivity and D* values attainable. The high bias
use for shorter wavelength detection with the aid of short- current also introduces a high 1/f noise component which
pass cold filters (see Section 2). The improvements in provides further restrictions to the maximum achievable D*
D* which result from the introduction of two cold filters values at low frequencies. Despite claims in manufactur-
are also shown in Figure 10. The authors are using a ers brochures the largest area they can be sourced in a
number of large area InSb detectors incorporating short photovoltaic guise is limited to 1 mm in diameter.
pass cold filters transmitting below 2.5 m for radiometric The response of Hg0.8 Cd0.2 Te covers the important 8 m
applications between 1.6 m (the cut-off wavelength of to 12 m atmospheric transmission window where this
InGaAs detectors) to about 2.5 m. InSb detectors from detector offers BLIP operation even at 77 K. With a dif-
certain suppliers are known to exhibit a condition known ferent HgCd composition ratio, the response of HgCdTe
as flushing, which arises when a cold detector is exposed photoconductive detectors extends to 26 m, where they
to UV or visible radiation. Preventing the exposure of these can operate adequately even at liquid nitrogen tempera-
detectors to short wavelength radiation, particularly when ture. They can be operated with a flat frequency response
they are cold, can prevent this condition. to beyond 100 kHz while the response of the smaller area
Photoelectromagnetic (PEM ) photodetectors form a vari- devices can be extended to beyond 1 MHz so they find
ation on the operation of photoconductive detectors because applications in Fourier transform spectroscopy. Figure 11
the photoconductive element is placed in a transverse mag- summarizes the D* values of a number of Hgx Cd1x Te
netic field so the electrons and holes generated by the inci- detectors as a function of wavelength. All except one
dent infrared radiation are deflected by the magnetic field are operating at 77 K. However, when photoconductive
in opposite directions as they drift through the semiconduc- HgCdTe detectors were evaluated at NPL, their perfor-
tor lattice. The electrons and holes reach different faces of mance was found to suffer from a number of drawbacks
the semiconductor and they can be detected as a voltage which prohibit their application in infrared radiometry.8
difference (using large input impedance instruments) or as Their spatial variation of response was very poor with
a current, if a low impedance instrument is utilized. The some detectors evaluated exhibiting nonuniformities higher
PEM detectors have few advantages compared to conven- than 50%. Furthermore their relative spectral response dif-
tional photon detectors and they require a magnet to operate fered from one point on their active area to the next. Their
so they have found very few applications. One exception nonlinearity increases as a function of irradiance rather than
is a PEM detector based on InSb that can detect radiation total power incident on the active area and their linearity
up to 7 m without the need for cooling. Operation at room range was very restricted. Some of these drawbacks were
temperature makes it attractive for some industrial applica- overcome by operating HgCdTe detectors behind integrat-
tions despite the relatively poor noise performance of this ing spheres25 although the responsivity of the detector was
detector and the additional requirement of the magnet. reduced by two orders of magnitude.
HgCdTe photovoltaic and photoconductive detectors. The Several alternative variable band gap alloys/semi-
structure of HgCdTe detectors resembles that of InGaAs conductors are being developed. These include HgZnTe,
1011
MCT-5-N
MCT-12.5
FT-IR-16
D *(, 10 K, 1,60)
(cm Hz1/2 W1)
1010 FT-IR-22
MCT-5-TE
FT-IR-24
109
1 4 8 12 16 20 24
Wavelength / m
Figure 11. D* of Hgx Cd1x Te detectors as a function of wavelength. All except the MCT-5-TE are operated at 77 K. [Reproduced by
permission from InfraRed Associates Inc HgCdTe Technical Data Sheet.]
HgMnTe and HgCdZnTe26 but none appear to be able to Also shown in Table 1 are the measured time constants
match the performance of HgCdTe detectors at present. for each impurity detector.5,17 Figure 12 shows the D* of
The principle of operation of extrinsic photoconductors some doped germanium detectors at different wavelengths.
was summarized in Section 3. Extrinsic photoconductors Extrinsic Ge detectors can operate in a current or voltage
require a host lattice of very high purity (before the mode and have relatively fast time constants, as can be
impurities are added). Historically, germanium and silicon seen from Table 1. The time constants of these detectors
have been the materials that could be fabricated with the can be reduced further by ensuring that the lifetime of the
highest purity so they were adopted as the host lattice for all generated charge carriers is kept low.17
extrinsic photoconductors. Table 1 shows a list of materials Copper doped germanium Ge : Cu has been one of the
which have been used as the impurity in Ge-doped extrinsic most popular extrinsic photoconductors with a response
photoconductors. Table 1 also shows the cut off wavelength extending to 30 m. The addition of copper impurity in
of the resulting detector, along with the temperature at the Ge lattice offers the advantages of introducing three
which the detector will operate under BLIB conditions. acceptor levels in the Ge forbidden gap. The impurity
band of most interest for long wavelength operation has
Table 1. Characteristics of germanium-doped extrinsic photo- a band gap of only 0.04 eV corresponding to the cut
conductors.5,17 off wavelength of 30 m. The Ge : Cu detector has to be
cooled to below 14 K for BLIP operation but is most
Impurity Cut off Operating Time frequently used at 4.2 K. The concentration of the copper
wavelength (m) temperature (K) constant
impurity is typically 1016 atoms cm3 and is partly limited
Au 8.3 77 <50 ns by the solubility of Cu in the Ge lattice. The detectors
Hg 14 35 100 ns are formed from cubes, of typically 3 mm or 5 mm side
Cd 22 25 50 ns before electrical contacts are added on two opposite faces.
Cu 30 14 100 ns Commercial products are offered with circular apertures
Zn 40 <10 50 ns up to 5 mm in diameter by incorporating limiting cold
In 120 4 <1 s
apertures. Operation at the liquid helium temperature is a
B 120 4 <1 s
Ga 120 4 <1 s serious drawback for extrinsic photoconductors and it has
restricted their applications to the laboratory.
1011 Table 2. Characteristics of silicon-doped extrinsic photo-

conductors.17
Impurity Cut off Operating

wavelength (m) temperature (K)
Tl 4.3 78
1010 S 6.8 78
In 7.4 78
D* (max, 780) (cm Hz1/2 W1)
Mg 12.1 5
Ga 17.8 20
Bi 18.7 29
Al 18.4 27
Ge : Zn
Ge : Hg As 23 13
109 B 30 15
Sb 32 10
Ge : Cu 1011
108
Ge : Au
D * (max, 780) (cm Hz1/2 W1)
107 1010
1 2 3 4 6 8 10 20 30 40 50
Wavelength / m
Figure 12. D* of various doped germanium detectors. All are
operating at 4.2 K except Ge : Au which is operating at 77 K. All
view a 295 K background with a 60 field of view. [Reproduced
by permission from Santa Barbara Research Centre, Infrared
Components Brochure, 17th edition.]
More recently, interest in silicon extrinsic photconductors 109

Si : Sb
has been resurrected17 because these detectors can be Si : Ga
fabricated in a monolithic construction which include the Si : As
detector and the support/readout electronics. The interest
lies mainly in the detection over the 3 m to 5 m and 8 m
to 14 m atmospheric windows.
Table 2 shows a list of materials which have been used
as the impurity in silicon-doped extrinsic photoconductors.
Table 2 also shows the cut off wavelength of the resulting 108
detector, along with the temperature at which the detector 0 5 10 15 20 25 30 35
has to operate for adequate suppression of thermal ioniza- Wavelength / m
tion of the impurity states. The cut off wavelength of silicon Figure 13. D* of various doped silicon detectors operating at
doped photoconductors depends on the concentration of the 4.2 K. All view a 295 K background with a 60 field of view.
impurity and shifts to longer wavelengths as the doping [Reproduced by permission from Santa Barbara Research Centre,
density increases.17 Infrared Components Brochure, 17th edition.]
Figure 13 shows the D* of some doped silicon detectors
at different wavelengths. width and depth, only two states can exist in the quantum
Quantum well detectors form a relatively new type of well, a ground state and one state very near the top of
infrared detectors which rely on engineering narrow band the well. Infrared photons can raise an electron to the
gap semiconductors.27 The energy states in a quantum well excited state where an applied electric field can generate
are discrete. By creating quantum wells of the appropriate a photocurrent. Because the two states are relatively well
defined, the absorption and therefore the spectral response 5 SUMMARY AND CONCLUSIONS
of these detectors is relatively narrow, unless quantum wells
of different characteristics are included, each absorbing in Photon detectors have two significant advantages over
a different spectral region. Their main advantage is that thermal detectors. Firstly, they respond over a narrower
detectors can be fabricated using well-established process- part of the spectrum compared to thermal detectors so the
ing technologies. Quantum well detectors lend themselves thermal background generated noise in photon detectors
well to detector matrix fabrication although they require will always be lower and their D* higher. Secondly, photon
slightly lower temperatures to match the D* values of detectors have much faster response times, typically 1 s or
HgCdTe detectors.28 Little or no effort has been directed faster compared to the response time of thermal detectors,
in the development of large area (diameter greater than which is restricted to slower than 1 ms.
1 mm) quantum well detectors because the advantages they The main advantages of thermal detectors are that
offer are of limited application to detectors of a nonmatrix they respond over much wider parts of the spectrum
format. (theoretically over the entire spectrum). Furthermore the
Photoconductors based on free carrier intra-band absorp- cooling requirements of photon detectors are much more
tion are used to detect radiation longer than 300 m.16 Their stringent than those of thermal detectors which can operate
main difference from other photoconductors is that the at ambient temperatures. Photon detectors have to be cooled
absorption of infrared radiation causes intra-band transitions sufficiently to ensure that the number of thermally excited
of electrons within the conduction band. This increases the electrons is small compared to those excited by the incident
mobility of the electrons resulting in an increase in the infrared radiation.
electrical conductivity of the material. This type of detector Detectors can be selected over a wide part of the infrared
requires semiconductors with high electron mobilities such spectrum which offer a performance which is close to the
as Ge and InSb. While free carrier absorption takes place BLIP condition. It was pointed out in Section 2 that the
at all temperatures, at room temperature it does not modify adoption of narrow band cold filters can improve the D*
the semiconductor electrical conductivity sufficiently to be value of a detector by orders of magnitude in some cases.
detected. At the liquid helium temperature, however, the The resulting narrow band spectral response of the detec-
electron population distribution in the conduction band can tor may not be acceptable. The active area of an infrared
be modified sufficiently by the absorption of the very long detector should be made as small as possible and this can
wavelength photons, resulting in a change in their mobility be achieved by demagnification, the adoption of a field
which is detected just like in other photoconductors. The lens or by the optical immersion method. The cold aper-
free carrier absorption increases linearly with wavelength, ture defining the detector field of view should be selected
while the number of photons in a unit energy of incident to minimize the noise contribution due to the thermal
power also increases with wavelength. The D* of free car- background.
rier absorption detectors shows a very rapid increase with However we also realize that minimizing the NEP of a
wavelength, making their operation attractive for wave- detection system is not always paramount. In these cases,
lengths longer than 1000 m. other detector selection criteria such as cooling require-
In the mid-1990s Grimes et al.29,30 showed that point ments, ruggedness, physical size and cost of purchase
contact Josephson Junctions will respond to both millimeter become more important. Then the task of selecting the most
and submillimeter wave radiation at wavelengths close to appropriate detector may become even harder.
their superconducting energy gap. Similar work showed
the spectral response of NbNb superconducting point
contact with an energy gap in the region of 2.2 K. Grimes ABBREVIATIONS AND ACRONYMS
et al.29 suggested that this and the associated structure
could be related to the Josephson Junction having a BIB Blocked Impurity Band
strong tendency to act as a source of such radiation. The BLIP Background Limited Infrared Photodetector
operating temperature of this device lay in the region of DLATGS Deuterated l-alanine Doped Triglycine
2.2 K and demonstrates the benefits of the macroscopic Sulfate
parameters of speed, high D and narrow bandwidths; such G-R Generation-Recombination
a device can be made to operate in a narrow band, voltage InGaAs Indium Gallium Arsenide
tunable manner to generate a continuous wave narrow NEP Noise Equivalent Power
band detector. This could lead to significant improvements NPL National Physical Laboratory
in detector effects with corresponding improvements in PEM Photoelectromagnetic
far-infrared spectrometers. QWIP Quantum Well Infrared Photodetector
REFERENCES 14. E.L. Dereniak and D. Crowe, Optical Radiation Detectors,

Wiley, New York (1984).
15. J. Geist, Appl. Opt., 25, 2033 (1986).
1. W. Herschel, Phil. Trans. R. Soc. London, 90, 284
(1800). 16. M.F. Kimmitt, Far Infrared Techniques, J.W. Arrowsmith,
Bristol, Ch. 4 (1970).
2. N.M. Stevens, Radiation Thermopiles, in Semiconductors
and Semimetals, eds R.K. Willardson and A. Beer, Aca- 17. D.C. Crowe, P.R. Norton, T. Limperis and J. Mudar, Detec-
demic Press, New York, 287317, Vol. 5 (1970). tors, in The Infrared & Electro-Optical Systems Handbook,
Electro-Optical Components, ed. W.D. Rogotto, Infrared
3. R.D. Hudson, Infrared System Engineering, John Wiley & Information Analysis Centre, Environmental Research Insti-
Sons, New York (1969). tute of Michigan (Ann Arbor, MI) and SPIE Optical Engineer-
4. BSI PD 6461 : 1995, Vocabulary of metrology, Part 3, Guide ing Press (Bellingham, Washington, DC), 175283, Volume 3
to the expression of uncertainty in measurement, BSI ISBN (1993).
0 580 23482 7. 18. E.H. Putley, The Pyroelectric Detector, in Semiconductors
5. W.L. Wolfe and G.J. Zissis (eds), The Infrared Handbook, and Semimetals, eds R.K. Willardson and A.C. Beer, Aca-
3rd edition, Office of Naval Research, Washington, DC demic Press, New York, 259285, Vol. 5 (1970).
(1989). 19. M.J.E. Golay, Rev. Sci. Instrum., 18, 357 (1947).
6. R.C. Jones, Proc. Inst. Radio Eng., 47, 1494 (1950). 20. R.W. Boyed, Radiometry, and the Detection of Optical
7. P.W. Kruse, Infrared Phys. Technol., 36, 869 (1995). Radiation, J. Wiley & Sons, New York (1983).
8. E. Theocharous, N.P. Fox and T.R. Prior, Proc. SPIE, 2815, 21. P.N. Dennis, Proc. SPIE, 234, 27 (1980).
56 (1996). 22. T.H. Johnson, Proc. SPIE, 443, 60 (1984).
9. W. Budde, Physical Detectors of Optical Radiation, in 23. R.J. Hutchinson and J.K. Moore, Proc. SPIE, 197, 287
Optical Radiation Measurements, Academic Press, New (1979).
York, Vol. 4 (1983).
24. J.M. Mooney, J. Appl. Phys., 65, 2869 (1989).
10. D.B. Betts, F.J.J. Clarke, L.J. Cox and J.A. Larkin, J. Phys. 25. E. Theocharous, T.P. Prior, P.R. Haycocks and N.P. Fox,
E: Sci. Instrum., 18, 689 (1985). Metrologia, 35, 543 (1998).
11. S.S. Ballard, K.A. McCarthey and W.L. Wolfe, Optical 26. J. Piotrowski, Laser Focus World, 135 (1997).
Materials for Infrared Instrumentation, Report No 2389-11-S,
University of Michigan (1959). 27. B.F. Levine, J. Appl. Phys., 74, R1 (1993).
12. R.H. Kingston, Detection of Optical and Infrared Radiation, 28. S. Gunapala, G. Sarusi, J. Park, T.L. Lin and B. Levine,
Springer-Verlag, BerlinHeidelbergNew York (1978). Physics World, 35 (1994).
13. L. Harris, The Optical Properties of Metal Blacks and 29. C.C. Grimes, P.L. Richards and S. Shapiro, Phys. Rev. Lett.,
Carbon Blacks, MIT (Cambridge, MA) and the Eppley 17, 431 (1966).
Foundation for Research (Newport, RI), Monograph Series 30. C.C. Grimes, P.L. Richards and S. Shapiro, J. Appl. Phys.,
No. 1 (1967). 39, 3905 (1968).
Optical Materials for Infrared Spectroscopy
D. Warren Vidrine
Vidrine Consulting, San Juan Capistrano, CA, USA
1 INTRODUCTION (ATR) elements. In both types of applications, the edges

of the optical material itself commonly function as the
Infrared (IR) spectroscopy is unique among the high- optical elements means of registration and mounting.
information-content spectroscopies in its ability to measure Because of the varied limitations of IR-transmitting mate-
easily all states of matter. Even better, it can do this rials, mid-IR design choices are often complex and dif-
with the sample segregated from the inner workings of ficult. The materials themselves cover a diverse gamut,
the instrument (unlike mass spectrometry and chromatogra- including some that are seriously toxic, frightfully expen-
phy). Nonetheless, the sample must be sampled somehow, sive, or amusingly exotic. The approach in this article
so the Achilles heel of IR spectroscopy is the requirement is first to introduce the general types of materials, then
for interfacial materials which transmit and/or reflect IR explore the optical, mechanical, thermal, and chemical
light over a broad spectral region. characteristics relevant to their applications in instrumen-
In near-IR, visible, and ultraviolet (UV) spectroscopy tation.
we take the abundant virtues of plastics, glass, and fused
quartz for granted, but in the mid-IR region no such
perfect materials exist and it is difficult to achieve 2 FAMILIES OF TRANSPARENT
even two qualities out of the good, cheap, easy tril- MATERIALS
ogy. This article is intended as a critique of the optical
materials commonly utilized in IRsample interface sys- Transparent materials group themselves rather neatly into
tems, along with some comments about their practical eight families of materials. The colligative characteristic
implementation. within each family may be chemical, mechanical, optical,
Several reference books have extensive tables of proper- or a combination of these. This section introduces these
ties for mid-IR optical materials, but there it is difficult to families, and attempts to give a broad-brush description of
find the salient properties assembled in a few concise tables. their salient features.
In addition, some of the properties (especially mechanical Table 1 is a list of optical materials with useful mid-
properties) have different values in different tables! The IR transparency. It is by no means complete, but rather
tables in this article began as an attempt to gather a work- represents materials I have used or about which I have
ing data set of materials properties for my own work and acquired information. However, this list does include all
contain data from a plethora of sources (including some the materials commonly used in mid-IR practice. In many
unpublished data and verbal communications). When these cases, listed materials are available from multiple manufac-
sources disagree significantly on a property value, an edu- turers and their actual properties may vary significantly. In
cated estimate has been made. particular, some of the materials are sold both as single-
The two basic uses for transparent materials in IR sam- crystalline materials and as polycrystalline or amorphous
pling are as windows and as attenuated total reflectance materials. I have tried to include some tradenames both
as a way of further identifying materials and as an aid to
John Wiley & Sons Ltd, 2002. locating the rarer materials.
Table 1. Families and names of transmissive optical materials.a
Family Material Other names Symbol
Alkali metal halides (salts) Sodium chloride Rocksalt, halite NaCl

Potassium chloride Sylvite KCl
Potassium bromide KBr
Cesium bromide Csium bromide CsBr
Cesium iodide Csium iodide CsI
Metal fluorides Lithium fluoride LiF
Magnesium fluoride Sellaite, Irtran-1 MgF2
Calcium fluoride Fluorspar, fluorite, Irtran-3 CaF2
Strontium fluoride SrF2
Barium fluoride BaF2
Heavy metal halides Silver chloride Cerargyrite, chlorargyrite, hornsilver AgCl
Silver bromide Bromyrite AgBr
Thallium bromochloride KRS-6, Tl(Br,Cl), TlBr0.4 Cl0.6 KRS-6
Thallium bromoiodide KRS-5, Tl(Br,I), TlBr0.42 I0.58 KRS-5
Metal oxides Magnesium oxide Irtran-5, crystalline magnesia MgO
Sapphire a-Alumina, sapphire, corundum aAl2 O3
Crystal quartz a-Quartz, rock crystal, silicon dioxide aSiO2
Zirconia Cubic zirconia (Hf-doped) ZrO2
IIVI chalcogenides Zinc sulfide b-ZnS, sphalerite, Irtran-2, Cleartran ZnS
Zinc selenide Irtran-4 ZnSe
Cadmium sulfide Greenockite CdS
Cadmium selenide Cadmoselite CdSe
Cadmium telluride Irtran-6 CdTe
Groups IV and IIIV Diamond (Type IIa) C
Silicon FZ-Si (undoped vacuum-grown float-zone) Si
Germanium Ge
Gallium arsenide GaAs
Glasses Silica glasses Fused quartz, vitreous quartz, Vycor vSiO2
Borosilicate glasses Pyrex, BK-7, LEBG BSCG
Fluorozirconate glass ZBLAN, HMFG HMFG
Arsenic trisulfide glass Orpiment glass As2 S3
AMTIR-1 Ge33 As12 Se55 glass AMTIR
Plastics Polyethylene (Usually, high-density) poly, polythene HDPE
Polymethylpentene TPX, PMP TPX
Polyester PET, PETG, Mylar, Melinex PET
Polytetrafluoroethylene Teflon TFE, PTFE PTFE
Fluorinated EPC Teflon FEP, TFE/HFP FEPC
a Names were selected on the basis of their descriptive value and common use; the names in parentheses include the commonest
tradenames. Owing to space limitations, many material names are vulgar: some proper names such as poly(4-methylpent-1-ene) and
poly(ethylene terephthalate) have been truncated, and lengthy designations such as fluorinated ethylenepropylene copolymer (EPC)
and tetrafluoroethylenehexafluoropropylene copolymer have been omitted. Symbols are for convenience; they are mostly chemical
formulas but also include some common designators for nonstoichiometric materials. Trademarks include Kodaks Irtran, Amorphous
Materials AMTIR, CVDs Cleartran, Cornings Pyrex and Vycor, Mitsuis TPX, and Duponts Teflon, Melinex, and Mylar.
2.1 Soluble halides 25 m) was originally defined by the transmission limit of

thick KBr optics. This family of materials has unrealized
Sodium chloride (table salt, NaCl) is the paragon of this potential: practitioners who only consider the commonly
family, composed of the IVII salts (except fluorides). The used ones (NaCl, KBr, and CsI) ignore the least hygroscopic
members of this family are of gratifyingly low refractive member, the versatile and inexpensive KCl. They neglect
index (RI), but are notably water-soluble, soft, and rela- the robust CsBr, which is as hard as KBr, but does not
tively weak. The pioneering work in the mid-IR region cleave and is notably resistant to shock and point stress.
was accomplished with NaCl optics. The modern low- Also neglected are the inexpensive (but difficult to find)
wavenumber limit of the mid-IR region (400 cm1 or polycrystalline NaCl and KCl, which combine superior
Optical Materials for Infrared Spectroscopy 3
strength (quadruple the practical strength of the crystalline made from it. After the preceding panegyric, it is important
materials!) with freedom from cleavage. CsI is a workhorse to note that KRS-5 has some severe faults. Thallium is
in the far-IR region, but in the mid-IR region it is inferior to frighteningly toxic; in the USA thallium salts are even
its cousins, being fingernail-soft and extremely hygroscopic. banned as rat poison because they are so dangerous! In
IR practice, the insolubility of KRS-5 offers considerable
protection against accidental poisoning, but polishing oper-
2.2 Fluorides ations are hazardous because of the toxicity of the dust.
Polish KBr, but leave KRS-5 to the professionals! Another
Calcium fluoride (fluorspar, CaF2 ) is the paragon of this limitation of KRS-5 is its tendency to cold flow, which
family, characterized by low water solubility and low RI. hampers its use in pressure windows.
Calcium fluoride is often used for aqueous transmission
cells because it resists corrosion by water. Barium fluoride
is slowly etched by water, but is sometimes used for aque- 2.4 Oxides
ous transmission measurements because its useful spectral
range includes the major low-absorptivity spectral regions These materials (MgO, Al2 O3 , SiO2 , and ZrO2 ) are in some
of liquid and frozen water. Solutions containing ammonium ways a mixed bag, but they are all hard, tough materials
ion rapidly corrode most fluorides, and fluorides tend to be with limited spectral range. Sapphire (clear, crystalline a-
sensitive to strong acids. This family (especially BaF2 ) has Al2 O3 ) is a commonly used (and fairly cheap) material for
the reputation of being sensitive to thermal shock, although mid-IR measurements. With its toughness, inertness, and
I have personally observed that some (not all!) BaF2 win- extreme hardness it is a well-deserved favorite for process
dows survive immersion in liquid N2 . A more probable applications. ZrO2 probably deserves more attention for
culprit is the deadly combination of thermal shock and ATR applications. Crystalline SiO2 is rarely used as it has
inferior properties (except in the far-IR region) to the ubiq-
transient point stress when windows are constrained in cell
uitous fused quartz. MgO in its pressed Irtran-5 form has
mounts.
suffered from incomplete transparency and low strength,
but new methods of producing its crystalline form may
allow the exploitation of MgOs superior spectral range.
2.3 Heavy metal halides
The members of this family are the opposite of brittle: they 2.5 IIVI (chalcogenides)
are actually malleable! The silver salts (primarily AgCl and
AgBr) are water resistant and transparent over the entire This peculiarly named family of materials includes two
mid-IR range, but they are weak materials prone to cold workhorses, ZnS and ZnSe. Their use was pioneered with
flow: their flexural apparent elastic limit (FAEL) strength their hot-pressed forms, Irtran-2 and Irtran-4. The advent
must be derated at least 25-fold in long-term pressure appli- of CVD (chemical vapor deposited) ZnS and ZnSe has
cations! AgCl is sometimes used as an optical solder, or extended their use to ATR applications. Although suscepti-
as a crushable index-matching gasket. Contact with met- ble to strong oxidizers and aggressive acids and bases, these
als can initiate rapid electrochemical corrosion; a tidy (but materials have proved both mechanically and chemically
rarely used) solution is to machine the cell body out of sil- robust in many applications and are favorites for process
ver! AgCl and AgBr photodegrade when exposed to actinic IR spectroscopy. Although zinc, sulfur, and selenium are all
light, preventing their use in monitoring photolytic reac- vital nutrients, both materials can be hazardous. Selenium is
tions. AgCl and AgBr are soluble in solutions containing toxic in remarkably small doses, and zinc-containing dusts
chelating agents, concentrated chlorides, or CN , S2 O3 2 , are legendarily allergenic. Cadmium telluride is occasion-
or Br3 ions. Strong oxidizers can strip the halide away ally used for similar applications, but it is fragile, toxic, and
from the silver salt, and reducing agents can convert AgCl chemically vulnerable. Note that the chalcogenide glasses
or AgBr into black deposits of Ag metal. (e.g. AMTIR and orpiment glass) are grouped with the
The thallium salts have outstanding virtues, and KRS-5 glasses in Section 2.7, not here.
has become a ubiquitous optical material in IR practice. Its
cousin KRS-6 is rarely used because it is neither harder
nor stronger than KRS-5 and has a more limited spectral 2.6 IV and IIIV (diamond family)
range. With its transparency and robustness KRS-5 has until
recently reigned as the premier material for routine labo- This group contains some of the most unique and useful
ratory ATR spectroscopy, and mid-IR optical fibers can be IR-transmitting materials which exist, all of them extremely
hard and brittle, but impervious to thermal shock because 2.8 Plastics
of their low expansion coefficients and high thermal con-
ductivities. Diamond is certainly the most expensive IR Several plastics have invaded the unlikely territory of
optical material in common use. Type IIa diamonds are gen- mid-IR optical materials. All of them have significant
erally used because of their superior IR transmission, which absorption bands within the mid-IR region but also
covers the whole mid-IR range except for a phonon band considerable regions of transparency. Their main uses
around 2000 cm1 (5 m). Diamond is the premier mate- are as stretched-film environmental windows, specialty-
rial for high-pressure cells, although great care has to be application cell windows, and molded Fresnel lenses. The
taken to avoid failure by crystal plane cleavage. Note that ease with which they are formed and molded is a major
although properly oriented and mounted diamonds achieve virtue, and their chemical resistances are unique and useful.
a strength of 8700 MPa, a more realistic strength figure for Although polyethylene cells have been used for far-IR
ordinary mounting is a mere 370 MPa! Inexpensive thin measurements for decades, there are as yet few laboratory
plates of synthetic diamond, produced by vapor deposition, mid-IR uses of plastic cells. Chief among these are several
have now become available. Silicon is the little cousin of schemes utilizing meshes or porous films surface tension
diamond, and a very inexpensive material because of the wicks a liquid sample into the pores to form an oil spot,
semiconductor industry. In practical window thicknesses, volatile solvents evaporate, and the measurement is made
Si transmits light over the entire mid-IR range except for in transmission.
a narrow two-phonon band at 610 cm1 (16.4 m). Note
that undoped vacuum-grown float-zone material should be
used; the more common Czochralski-grown Si has signifi- 3 COMPOSITE MATERIALS
cant absorptions from C and O impurities, including a broad
band at 1100 cm1 (9.1 m). Ge is a similar material, and 3.1 Coatings
even has useful transmission out to 700 cm1 (15 m)
through a 50-mm thickness of material. Si and Ge both
Optical coatings represent the earliest and most ubiquitous
have extremely high refractive indices, useful for ATR mea-
examples of tailored materials. MgF2 antireflection coatings
surement at small penetration depths but troublesome for
have a long history in visible optics, and they are applied
transmission measurements. One caveat: because of free
to some mid-IR optics as a way of passivating and
thermal electrons, both Si and Ge become opaque at mildly
hardening the surface. Although MgF2 has a reststrahlen
elevated temperatures. Ge has another undesirable charac-
band at 20 m, CaF2 (with its reststrahlen band at 46 m)
teristic: Ge coated with Pt or other metals can become
can be used instead if maximum performance at long
irreversibly opaque upon heating in an ordinary laboratory
wavelengths is needed. The soluble halides are often coated
oven. If the facile cleavage of Si and Ge becomes a serious
with proprietary organic substances which inhibit fogging.
detriment in a particular application, it may be possible to
Diamond coatings are not as impervious as the cachet
substitute polycrystalline GaAs (which is hard to find but
of their name suggests because the pinholes in vapor-
commercially available).
deposited diamond coatings can allow moisture to get under
the coating and cause it to flake off. Newer diamond
2.7 Glasses coatings perform better because underlayers have been
developed which passivate the underlying material and
The family of IR glasses includes several materials uti- prevent corrosional propagation of defects.
lized for IR optical fibers. Opticians may be upset by the
short shrift I give the silica glasses (lumping Pyrex and
BK-7 together, indeed!), but they are rarely used beyond 3.2 Tailored materials
the near-IR region. The first true mid-IR glass was orpi-
ment glass, now largely supplanted by ZnSe and AMTIR An early example of a tailored material is pressed KBr,
for discrete optics but being utilized in purer form for fiber originally utilized for sample pellets and molded lenses
optics. AMTIR is utilized for ATR elements when resis- but found useful because of its freedom from cleavage.
tance to acidic environments is needed. Fluorozirconate Polycrystalline (Polyclear) NaCl and KCl windows for
glasses were developed for optical-fiber applications, but strength applications are significant improvements over
are rarely used for discrete optics because CaF2 is generally the standard crystalline windows. CVD materials, although
superior. Not being stochiometric materials, glass compo- originally produced because of the difficulties of growing
sitions can be tailored, but such materials belong in the large crystals of some materials (including ZnS and ZnSe),
sections on composite and tailored materials. do exhibit superior properties compared with the natural
materials. Another class of tailored materials is a child of In addition to length limits and environmental fragility,
the CVD technique: provided that there is a reasonable a perennial problem for mid-IR optical fibers results from
match of thermal characteristics, several CVD materials their small diameters. It is simply impossible to cram
can be deposited in successive layers to form a composite much light into them with a thermal source, so fiber-
bulk material. With this technique, the superior surface based applications tend to be energy-starved. The mid-IR
qualities of ZnS can be achieved in an optical element region conspires against optical fibers: materials are brittle,
with the superior transmission properties of ZnSe. This kind absorbing, and environmentally fragile, sources are weak
of tailoring is in its infancy, but holds great promise for and detectors are noisy, and the long wavelengths require
creating superior mid-IR optical materials. thick cladding and take their toll with diffraction losses.
Hollow optical conduits transmit large energy throughputs
and offer other advantages, so they are a more popular
3.3 Optical fibers method of translocating mid-IR light.
Although most of the effort spent developing IR optical

fibers has been for near-IR communications fibers, signifi- 3.4 Matrix materials
cant progress is being made with mid-IR fibers. The good
news is that practical mid-IR fibers now exist; the bad A gray area exists with sample diluents. When there is
news is that none of them transmits light much beyond neither solution nor chemical interaction, these diluents are
1 m in that most important part of the mid-IR region below optical materials themselves, as surely as are microscope
2000 cm1 (>5 m). The premier optical fibers for the immersion oils. While the specialized index-matching oils
region above 2000 cm1 (<5 m) are fabricated from heavy such as are used for optical fibers are outside the scope
metal fluoride glasses (e.g. ZBLAN, HMFG). Their integral of this article, I should at least mention some common
cladding (an optically thick coating of lower-refractive- examples of these diluent matrices, which admirably fulfill
index composition) makes them good examples of tailored a variety of dispersion and index-matching roles. Many of
optical materials, and they are suitable for applications in these matrices are also used in bulk form as conventional
the 1020-m distance range. Unfortunately, these fluoride optical materials, and appear in the tables.
fibers are water-soluble and prey to serious environmen- One of the earliest sampling methods was the KBr pellet,
tal degradation. Presenting a table of mid-IR optical fiber consisting of the sample dispersed and finely ground within
specifications seems foolhardy, since improvement contin- a KBr matrix which was then cold-pressed into a transparent
ues and the optical parameters of individual batches still disk. A lesser-known version using powdered polyethylene
vary very widely. was tedious to make because of triboelectrically generated
There are currently a variety of kinds of optical fiber repulsion (unless manipulated in a radioactive environ-
for the <2000 cm1 (>5 m) region, but all of them have ment!). Another hoary but ubiquitous and effective sam-
severe limitations beyond the attenuation which determines pling method utilizes a mortar-and-pestle mull of the sample
their 1-m length limit. The chalcogenide fibers (e.g. in alkane oil (mineral oil, Nujol), pressed between KBr
AMTIR and orpiment glass) have reasonable environmental plates. A perfluoroalkane oil (perfluorokerosene, Fluorol-
resistance, but are frustratingly brittle and have limited ube) mull allows spectral coverage in the regions obscured
spectral range. Silver halide fibers have reasonable spectral by the alkane oils few absorption bands. For reference, the
range and may currently hold the length/attenuation record, spectra of these mulling oils are shown in Figure 1.
but are painfully sensitive to light and environmental Diffuse reflection spectroscopy commonly utilizes opti-
degradation. KRS-5 fibers are sensitive to environmental cally scattering diluent powders, especially KBr, diamond
degradation, susceptible to scattering losses upon bending, dust, or Si. These diluent powders are vital to the opti-
and prey to further losses because they are usually unclad. cal nature of the measurement, and reduce the confounding
Many of these fibers have limited numerical aperture (the effects of the samples anomalous dispersion. In a type of
cone of incident light effectively transmitted down the fiber) combination of gas chromatography with IR spectroscopy, a
because of the limited compositional tailoring developed so frozen argon matrix is utilized: the output of the chromato-
far for them. Some high-numerical-aperture fibers (mostly graphic column is mixed with argon and spray-painted on
orpiment glass) have been fabricated unclad or by utilizing to a moving cryocooled surface, producing a physical chro-
a polymer sheath in lieu of integral cladding, but these matogram consisting of the actual chromatographic eluate
are prey to ATR losses by surface contaminants or by the striped along a time axis of solid transparent argon. Argon
sheath. All of the <2000 cm1 (>5 m wavelength limit) (like the other noble gases) is fairly mid-IR transparent even
fibers suffer from serious entrance/exit losses because of in solid form, so it serves as a near-perfect matrix for the
their high refractive indices. eluate.
equations specifying the RI and dispersion of many optical

1.0 materials throughout their spectral ranges are available in
Nujol Fluorolube
several of the reference books cited at the end of this article.
0.8
Nevertheless, the RI value of a material is reasonably
Absorbance
0.6
constant across the mid-IR range, so for the purposes of
this article a single value of RI is given for each optical
0.4 material in Table 2.
RI affects three important elements of mid-IR measure-
0.2 ments. First, reflection within a window or a cell can
produce interference fringing which appears as a sinusoidal
0.0
addition to the spectrum. Organic liquid measurements
4000 3000 2000 1000 are particularly susceptible to these troublesome fringes
Wavenumber / cm1 because liquid transmission cells are often in the mil-
limeter to submillimeter thickness range. Because organic
Figure 1. Mid-IR spectra of common mulling agents.
liquids commonly have refractive indices around 1.5 (usu-
One could argue that IR solvents are also themselves ally 0.2 RI units), low-index optical materials are valued
optical materials, since they are commonly utilized as as trouble-free windows for liquid cells, since a near-
solvent matrices for samples. However, the line must be match in RI between the liquid and the window (a low
drawn somewhere, so the liquid chromatographic solvents, n) suppresses the intensity of interference fringing. Inter-
excellent supercritical solvents such as Xe and CO2 , and ference fringing can also be suppressed by wedging the
even the traditional (but dangerous) CCl4 and CS2 , will not cell volume, or by choosing resolutionthickness combi-
be covered in this article. nations which avoid fringing (e.g. 8 cm1 resolution with
a 1-mm cell, or second-derivative spectroscopy with thin
cells).
4 MATERIAL PROPERTIES OF Reflectance at a single surface (between two nonabsorb-
TRANSMISSIVE OPTICAL ing media with RIs, n1 and n2 ) is described by
MATERIALS
n2 n1 2
RD
n2 C n1
4.1 Optical characteristics
Note that n1 is often represented by air, n D 1.0.
Most mid-IR transmissive materials are transparent into Second, reflection losses are high whenever light tra-
the visible or near-IR region on their short-wavelength verses a large step change in RI. These are often termed
side, and have a long-wavelength cutoff which defines window losses, and a series of high-RI windows can eat
their useful spectral ranges. A few kinds of materials up much of a spectrometers beam intensity while spraying
(e.g. quartz, diamond, silicon, silicate glasses, and plas- reflected light into inconvenient places inside the instru-
tics) have strong absorption bands within their main ment. Although broadband mid-IR antireflection coatings
spectral range, which produce opaque gaps in their use- are available, they are expensive and sometimes of dubious
ful transmission range. In Table 2, two long-wavelength mechanical/chemical durability.
cutoff values are listed for each material: one value Third, materials of high RI are necessary for ATR spec-
for typical window thicknesses, and the other value for troscopy, a valuable technique for the measurement of many
the long internal pathlengths characteristic of ATR spec- difficult liquid and solid samples. The technique requires
troscopy. that n (the ATR elements RI minus the samples RI)
In addition to spectral range, the other salient optical be high enough to provide a conveniently wide range of
property of these materials is their refractive index (RI) (or supercritical angles, since the technique requires that the
n). The RI of a particular material is not constant across sample contact the surface where total internal reflection
the entire mid-IR spectral range; the value tends to decrease of the IR light is taking place. As will be described in
gradually towards longer wavelength, then decreases more other articles, the ATR interaction with the sample effec-
rapidly near the long-wavelength cutoff of the material. tively penetrates only a fraction of the lights wavelength.
This dependence of RI on wavelength is called dispersion. With very high-RI ATR elements, the reflection angle
Dispersion is definitely not a constant across the spectral can conveniently be far in excess of the critical angle,
range, but must be considered when prismatic elements or resulting in an even smaller effective sample penetration
lenses are used in optical systems. Extensive tables and depth. Since the penetration depth varies with wavelength
Table 2. Optical characteristics of transmissive optical materials.a
Material Symbol Refractive Long-wavelength Comments

index, limit (m)
n5 m T<6 mm T50 mm
Sodium chloride NaCl 1.52 18 15

Potassium chloride KCl 1.47 23 19
Potassium bromide KBr 1.54 30 25
Cesium bromide CsBr 1.66 42 38
Cesium iodide CsI 1.74 60 48 Easy to scratch
Lithium fluoride LiF 1.33 7 6
Magnesium fluoride MgF2 1.34 8 6.5 Birefringent
Calcium fluoride CaF2 1.40 10 8.5
Strontium fluoride SrF2 1.41 11.5 10
Barium fluoride BaF2 1.45 12 11
Silver chloride AgCl 2.00 27 21 Darkened by UV
Silver bromide AgBr 2.15 34 30 Darkened by UV
Thallium bromochloride KRS-6 2.19 25 21
Thallium bromoiodide KRS-5 2.38 45 40 Orangered
Magnesium oxide MgO 1.64 8 6 Birefringent
Sapphire a-Al2 O3 1.62 6 4 Birefringent
Crystal quartz a-SiO2 1.47/4 m 4 2.6 Birefringent
Zirconia ZrO2 2.13/0.5 m 5.5 5
Zinc sulfide ZnS 2.25 14 10.5
Zinc selenide ZnSe 2.43 20 15 Yellow
Cadmium sulfide CdS 2.24 15 12 Birefringent
Cadmium selenide CdSe 2.45 20 15 Visually opaque, birefringent
Cadmium telluride CdTe 2.70 28 20 Visually opaque
Diamond C 2.39 80 N/A
Silicon Si 3.42 50 8.5 Visually opaque
Germanium Ge 4.02 23 14 Visually opaque
Gallium arsenide GaAs 3.82 17 12 Visually opaque
Silica glasses vSiO2 1.404 m 4 2.6 OH absorption varies
Borosilicate glasses BSCG 1.502 m 2.5 2 Strong OH absorption
Fluorozirconate glass HMFG 1.50/0.5 m 8 6.5 Little OH absorption
Arsenic trisulfide glass As2 S3 2.41 13 8
AMTIR-1 AMTIR 2.51 16 11.5 Visually opaque
Polyethylene HDPE 1.54 d Visually translucent
Polymethylpentene TPX 1.43 d Visually transparent
Polyester PET 1.54/0.5 m d Visually transparent
Polytetrafluoroethylene PTFE 1.35/0.5 m d Visually translucent/opaque
Fluorinated EPC FEPC 1.34/0.5 m d Visually transparent/translucent
a Refractiveindex values are given for room temperature and 5 m wavelength, except as noted. Approximate wavelength limits are given
for two material thicknesses, <6 mm and 50 mm. Wavelength limits designated d (discontinuous) indicate interrupted regions of
transparency. Diamond has a narrow absorption band at 5 m. Silicon has a narrow 2-phonon absorption band at 16.4 m. Czochralski-
grown silicon has an additional absorption band due to oxygen at 9 m. Semiconductors (e.g. Si, Ge, CdTe, GaAs) display free-carrier
absorption at elevated temperatures, limiting their use at high temperatures.
and with the (nonconstant) n, the effective ATR sam- materials, since many of these crystalline materials are
pling depth varies severalfold across the mid-IR spectral cut at special angles in order to minimize birefringence
range. when used as windows. The RI of a material may also
Several other optical properties affect specific kinds vary with pressure and electrical charge, allowing the
of measurements. Some of the materials are birefringent, construction of photoelastic polarization modulators and
which means that the material has different RI values acousto-optically scanned monochromators. However, these
depending on polarization of the light beam through the are very specialized applications of mid-IR optical materials
crystal. Table 2 presents in-between RI values for such and beyond the scope of this article.
Table 3. Mechanical properties of transmissive optical materials.a
Material Symbol Strength, Fracture Hardness Density, Comments

Fa toughness (Knoop) D
(MPa) (MPa m1/2 ) (kgf mm2 ) (g cm3 )
Sodium chloride NaCl 2.4 (9.5) ? 18 2.16 Cleaves

Potassium chloride KCl 2.2 (11) ? 17 1.98 Least hygroscopic salt, cleaves
Potassium bromide KBr 1.1 ? 7 2.75 Cleaves
Cesium bromide CsBr 8.3 ? 20 4.44 Shock resistant, does not cleave
Cesium iodide CsI 2.7 (5.6) ? 20 4.53 Shock resistant, deforms (cold flow)
Lithium fluoride LiF 11 ? 99 2.64 Sensitive to thermal shock, cleaves
Magnesium fluoride MgF2 49 1.0 415 (575) 3.18 Rugged, cleaves
Calcium fluoride CaF2 37 (54) 0.5 158 3.18 Hard, fairly tough
Strontium fluoride SrF2 35 ? 154 4.28
Barium fluoride BaF2 27 ? 82 4.88 Sensitive to mechanical shock
Silver chloride AgCl 26 (<1) ? 10 5.56 Deforms (cold flow)
Silver bromide AgBr 30 (1.1) ? 7 6.47 Deforms (cold flow)
Thallium bromochloride KRS-6 21 ? 39 7.19 Deforms (cold flow), dust very toxic
Thallium bromoiodide KRS-5 26 ? 40 7.37 Deforms (cold flow), dust very toxic
Magnesium oxide MgO 27 ? 692 3.58 Tough
Sapphire a-Al2 O3 448 (689) 3.0 1370 3.98 Very hard, tough
Crystal quartz a-SiO2 41 ? 820 2.65 Brittle, inert, cleaves
Zirconia ZrO2 ? 2.0 1160 5.85
Zinc sulfide ZnS 60 (103) 0.5 (0.8) 160 (250) 4.09 Tough, dust allergenic
Zinc selenide ZnSe 55 0.33 120 5.27 Fairly tough, dust allergenic
Cadmium sulfide CdS 28 ? 108 4.82 Birefringent, dust toxic
Cadmium selenide CdSe ? ? 70 5.67 Brittle, birefringent, dust toxic
Cadmium telluride CdTe 6 ? 56 5.85 Very brittle, dust toxic, visually opaque
Diamond C 370 (8700) 2.0 8820 3.52 Very hard and inert, cleaves
Silicon Si 62 0.95 1150 2.33 Very hard and inert, visually opaque
Germanium Ge 49 0.66 780 5.33 Very brittle, visually opaque
Gallium arsenide GaAs 50 ? 721 5.32 Dust toxic
Silica glasses vSiO2 35 0.8 461 2.20
Borosilicate glasses BSCG 35 ? 500 2.35
Fluorozirconate glass HMFG 25 ? 225 4.33 Sensitive to thermal shock
Arsenic trisulfide glass As2 S3 8 ? 109 3.20 Brittle, dust toxic
AMTIR-1 AMTIR 18 ? 170 4.40 Brittle, visually opaque, dust toxic
Polyethylene HDPE 27 ? R60 0.95 LDPE deforms (hardness only R15)
Polymethylpentene TPX 24 ? R60 0.83 Fairly rigid
Polyester PET 98 ? R110 1.37 Strong, e.g. stretched-film windows
Polytetrafluoroethylene PTFE 22 ? R58 2.24 Deforms (cold flow), visually opaque
Fluorinated EPC FEPC 18 ? R25 2.15 Deforms (cold flow)
a Strength (Fa) is defined as the elastic limit or flexural apparent elastic limit. Crystalline strength is often limited by brittleness and cleavage;
higher effective strength can often be achieved by careful orientation and mounting. Parenthetical Fa values for CaF2 , NaCl, KCl, and ZnS are for
polycrystalline materials. Parenthetical Fa value for CsI is for work-hardened material. Parenthetical Fa values for AgCl and AgBr are empirical
strengths for long-term pressure applications where cold flow is the determining factor. The low Fa value for diamond reflects its brittleness and
easy cleavage; the parenthetical value is achieved by careful orientation and mounting. Fracture toughness measures the difficulty of propagating an
incipient fracture. Fracture toughness value for ZnS is for the crystalline material; the parenthetical value is for CVD material. Parenthetical hardness
values for MgF2 and ZnS are for hot-pressed materials (e.g Irtran). Hardness values for plastic resins are in Rockwell units.
4.2 Mechanical characteristics optical materials properties presented in this article. In addi-
tion to the inexactitude of the numbers, their application is
The published mechanical characteristics of optical mate- fraught with twists and turns. Note the two strength val-
rials are surprisingly fragmentary and contradictory. The ues for diamond: they differ by 24-fold, even though they
gleaning of the data for Table 3 was by far the most diffi- describe the same material! The subtleties of point stresses
cult part of my data-gathering effort, and these mechanical and incipient cracks mock attempts to design by the num-
data are still the most approximate and suspect of all the bers alone.
The first mechanical property goes by the somewhat mis- As indicated above, hardness is a liability in many appli-
leading title of strength. It is an appropriate designation cations of transmissive optical materials. Hard things shatter.
as long as the stories of Achilles, Icarus, and Varus are kept However, hardness also has its virtues. In applications where
firmly in mind! For materials subject to sudden failure, the contact with solid samples is necessary, harder materials last
load at the elastic limit is Fa, but for many materials incip- longer without dulling or scratching. Even where the prob-
ient failure is heralded by an increase in the load versus lem is only rough handling, hard windows such as ZnSe
deformation and Fa is then defined as the flexural appar- (which is hard compared with full-range materials such
ent elastic limit (FAEL), the point where deformation /load as KRS-5 or KBr) and sapphire (very hard!) are popu-
increases by 50%. No distinction between these definitions lar. The ascendancy of single-application instruments has
is made in the table. crystallized efforts to find durable optics for particular appli-
For well-behaved materials, circular pressure windows cations. A solution to this dilemma is slowly being found
can be designed according to the following equation: in tailored materials, which (like Damascus steel or a fine
nihonto) manage to be hard and tough at the same time.
SF K D2 p
Fa D
4t2
where Fa is the strength value for the material, D is the 4.3 Thermal characteristics
unsupported diameter of the window, t is the thickness
of the window, p is the load per unit area (pressure, in The thermal characteristics of common window materials
megapascals), SF is the safety factor, nominally SF D 4, are presented in Table 4. Thermal conductivity becomes
and K is a constant which depends on the mounting important when large temperature gradients or temperature
method: Kclamped D 0.75 or Kunclamped D 1.125. To illustrate shifts are part of the engineering lifetime of an optical
the design information obtained, this equation can be element. The premier family of high-thermal-conductivity
rewritten as windows is diamond and its relatives Si, Ge, and GaAs,

SF K 1/2 p 1/2
although MgO, sapphire, and ZnS also have notably high
t
D thermal conductivities.
D 4 Fa Thermal expansion, in addition to its potentiating effect
Windows can either be mounted so that the periphery is in translating T/t into destruction, is a parameter which
constrained flat (clamped), or unconstrained (unclamped). must be honored in any cell design for high or low temper-
Hard clamping, of course, imposes additional thermal expan- atures. Cell optics must survive repeated transitions to and
sion compensation and point stress engineering require- from room temperature. When low pressures are involved
ments. This all sounds fairly cut-and-dried, but the reality and unconstrained (or spring-mounted) optics are practical,
is that mid-IR optical materials are far from being a well- cell design can remain primitive and effective. However,
behaved bunch! The equations are included only to illustrate when pressure requirements necessitate that the optics be
the interrelationships among the mechanical parameters. constrained and temperature excursions are necessary, ther-
The flatness, scratch and dig, and deburring of contact mal compensation must be considered. KBrs expansion
surfaces are of paramount importance in maintaining win- coefficient is quadruple that of Hastelloy C! The diamond
dow strength. Much more than a few data tables is needed family (C, Si, Ge, and GaAs) have very low thermal expan-
to design a cell, unless it is for a low-pressure, room- sion coefficients, and CaF2 , BaF2 , and HMFG are blessed
temperature, low-toxicity application where neither life nor with thermal expansion coefficients similar to that of 316
property would be imperiled by catastrophic cell failure. stainless steel.
Fracture toughness is the second mechanical property The maximum-use temperatures are based partly on
listed. It could be the most predictive mechanical property, defined physical properties and partly on spectroscopic folk
since it addresses a very common (perhaps the most wisdom. Certainly KBr is not rated anywhere near its melt-
common) failure mechanism for mounted transmissive ing point of 734 C, but extended service at 300 C does
optical elements. Unfortunately, fracture toughness values etch its surface! Diamond does not self-destruct anywhere
are unavailable for the majority of transmissive optical near its rated temperature, but it does burn in air. An inter-
materials. The extremely high point stresses generated in esting group of optical materials are the semiconductors
very hard materials conspire with low fracture toughness (e.g. Si, Ge, GaAs, CdTe), where free carrier absorption
to create what we might characterize as brittleness. This prohibits their use as transmissive optical materials at high
combination of properties is a good explanation for the temperatures. In essence, at high temperatures these materi-
legendary brittleness of diamond, and the comparative als become too electrically conductive to transmit light. An
toughness of sapphire or fused silica. additional caveat for these materials (especially germanium)
Table 4. Thermal properties of transmissive optical materials.a
Material Symbol Thermal Thermal Maximum Comments

conductivity expansion use temperature
(mW cm1 K1 ) (106 K1 ) ( C)
Sodium chloride NaCl 65 44 400 Slightly sensitive to thermal shock

Potassium chloride KCl 65 36 400
Potassium bromide KBr 48 43 300 Sensitive to thermal shock
Cesium bromide CsBr 10 48 400
Cesium iodide CsI 11 50 200
Lithium fluoride LiF 40 37 400 Very sensitive to thermal shock
Magnesium fluoride MgF2 150 11 500
Calcium fluoride CaF2 97 19 600
Strontium fluoride SrF2 83 18 600
Barium fluoride BaF2 117 18 500 Sensitive to thermal shock
Silver chloride AgCl 11 30 200
Silver bromide AgBr 6 30 200
Thallium bromochloride KRS-6 7 51 200
Thallium bromoiodide KRS-5 5 60 200
Magnesium oxide MgO 435 12 >2000
Sapphire a-Al2 O3 240 8.4 1700
Crystal quartz a-SiO2 12 11 >1200
Zirconia ZrO2 19 9 >1000
Zinc sulfide ZnS 272 6.4 300
Zinc selenide ZnSe 180 7.3 300
Cadmium sulfide CdS 159 4.6 200
Cadmium selenide CdSe ? 4.9 >200
Cadmium telluride CdTe 63 4.5 300 Opaque at high temperatures
Diamond C 23 200 0.8 >700
Silicon Si 1490 4.2 300 Opaque at high temperatures
Germanium Ge 602 6 100 Opaque at modest temperatures
Gallium arsenide GaAs 523 4.5 200 Opaque at moderate temperatures
Silica glasses vSiO2 14 0.6 1070 Immune to thermal shock
Borosilicate glasses BSCG 11 3.2 500
Fluorozirconate glass HMFG 6.3 17 150 Sensitive to thermal shock
Arsenic trisulfide glass As2 S3 2 25 150
AMTIR-1 AMTIR 2.5 12 300
Polyethylene HDPE 5 120 120 Softens
Polymethylpentene TPX 1.7 117 160
Polyester PET 2.4 80 120 Phase change at 76 C
Polytetrafluoroethylene PTFE 2.5 100 250
Fluorinated EPC FEPC 2.5 95 250
a Maximum use temperatures for semiconductors (e.g. Si, Ge, CdTe, GaAs) are far below maximum survival temperatures. Their low maximum use
temperatures are due to thermally generated free-carrier absorption and so are thickness dependent. High-temperature chemical interactions with samples
and window mounting materials can be a serious materials problem. PET undergoes a phase change which can be utilized for tightening stretched-film
windows with a heat gun.
is that thin metal coatings may diffuse into the bulk mate- be irretrievably etched by momentary exposure to condens-
rial at ordinary laboratory-oven temperatures, rendering the ing moisture. The water solubilities listed in Table 5 show
optical element permanently opaque. why. Materials with low solubilities (e.g. BaF2 ) are gener-
ally resistant to this fogging effect, but are slowly etched
4.4 Chemical resistance by flowing liquid water. Such etching rapidly accelerates
if certain kinds of complexing agents are present (NH4 C is
Transmissive optical elements must be chosen for resis- a nemesis to BaF2 and CaF2 ). A list of common chemi-
tance to the chemical environment. Humidity is a traditional cal incompatibilities is included in Table 5 as a guide, but
enemy of mid-IR optics, and the alkali metal halides can be warned that this list is nowhere near inclusive. Also, in
Table 5. Chemical properties of transmissive optical materials.a
Material Symbol Solubility Incompatibilities Comments

in H2 O
(satd wt%)
Sodium chloride NaCl 26.5 Water, glycols, high humidity Inexpensive

Potassium chloride KCl 25.8 Water, alcohols, ether, wet solvents Inexpensive
Potassium bromide KBr 39.5 Water, alcohols, ether, humidity
Cesium bromide CsBr 55.4 Water, alcohols, humidity
Cesium iodide CsI 61.5 Water, alcohols, any humidity
Lithium fluoride LiF 0.27 Water, HF
Magnesium fluoride MgF2 0.0076 Conc. acids Rugged
Calcium fluoride CaF2 0.0015 Ammonium ion, some conc. acids Fairly tough
Strontium fluoride SrF2 0.012 Ammonium ion, hot HCl
Barium fluoride BaF2 0.16 Ammonium ion, acids
Silver chloride AgCl 0.00014 Oxidizers, chelators, conc. chloride Attacks base metals, UV darkens
Silver bromide AgBr 0.00001 Oxidizers, chelators, conc. chloride Attacks base metals, UV darkens
Thallium bromochloride KRS-6 0.3 Methanol, chelators, strong bases Dust very toxic
Thallium bromoiodide KRS-5 0.05 Methanol, chelators, strong bases Dust very toxic
Magnesium oxide MgO 0.00062 Conc. acids, ammonium ion
Sapphire a-Al2 O3 Insol. Conc. acids and bases Inert
Crystal quartz a-SiO2 Insol. HF, some hot conc. bases Inert
Zirconia ZrO2 Insol. HF, H2 SO4 Fairly inert
Zinc sulfide ZnS 0.00069 Strong oxidizers, some acids Dust allergenic
Zinc selenide ZnSe <0.001 Acids, strong conc. bases Dust allergenic
Cadmium sulfide CdS 0.00013 Acids Dust toxic
Cadmium selenide CdSe Insol. Acids Dust toxic
Cadmium telluride CdTe Insol. Oxidizers, some acids Dust toxic
Diamond C Insol. K2 Cr2 O7 C conc. H2 SO4 Expensive, very inert
Silicon Si Insol. HF C HNO3 Inexpensive, very inert
Germanium Ge Insol. Hot H2 SO4 , aqua regia Some interaction with redox
agents
Gallium arsenide GaAs Insol. Acids, strong oxidizers Dust toxic
Silica glasses vSiO2 Insol. HF, strong bases Inert
Borosilicate glasses BSCG Insol. HF, strong bases Inert
Fluorozirconate glass HMFG >1 Water, high humidity
Arsenic trisulfide glass As2 S3 0.00005 Alcohols, bases Dust toxic
AMTIR-1 AMTIR Insol. Bases Dust toxic
Polyethylene HDPE Insol. Hot organics, oxidizing acids Inexpensive
Polymethylpentene TPX Insol. Chlorinated aromatics Inexpensive
Polyester PET Insol. Halohydrocarbons Inexpensive
Polytetrafluoroethylene PTFE Insol. None Toxic fumes at high temperatures
Fluorinated EPC FEPC Insol. None Toxic fumes at high temperatures
a Materials
containing poisonous elements (e.g. Tl, As, Cd, Se) can be handled safely, but dust from grinding or polishing is very hazardous. Chemical
incompatibilities often become worse at elevated temperatures; the noted incompatibilities represent interactions at modest temperatures.
the liquid phase, incompatibilities worsen rapidly at higher streams. In many cases, hygroscopic particulates (e.g.
temperatures. A sometimes-overlooked source of cell cor- SO4 2 , salt) are present which form highly corrosive
rosion is chemical incompatibility between the optical ele- aerosols with even modest humidity. Once an aerosol is
ment and its metal mounting. Some transmissive optical present, acid gases which are otherwise benign to trans-
materials (e.g. AgCl and AgBr) are electrochemically active missive materials can become powerful corrosive agents by
themselves, and many others can provide sources of ions dissolving in the aerosol particles. For this reason, corrosion
for electrochemically mediated corrosion. Stainless steel is via particulates can sometimes be suppressed by elevat-
not always corrosion free! ing the temperature so that the particulates are no longer
Aerosols pose a serious fogging danger to mid-IR win- hydrous. Centrifugal or media filtration should normally be
dows exposed to the environment or to gaseous sample employed when particulates are a problem.
Among the comments in Table 5 are indications of although differential reflectance effects can produce mild
extraordinary toxicity. Many of the materials listed are polarization and beam coloration in off-axis optics. Metals
somewhat toxic and should be handled with care, but it are generally used for reflecting surfaces, and many metals
would not serve the expected readership simply to label have very good reflection in the mid-IR region. However,
everything toxic. Therefore, only materials easily capable metal reflection is inevitably lossy, unlike dielectric reflec-
of poisoning unwary spectroscopists are listed as toxic. tion. Dielectric-enhanced metal surfaces have limited use
Overall, most bulk optical materials are reasonably safe to in general mid-IR instrumentation because of the broad
handle (given minor prophylaxis such as hand-washing or wavelength range involved, and because it is difficult to
gloves), but several can become dangerous or deadly during produce adherent coatings with the thicknesses required at
ill-advised grinding and polishing operations. long wavelength. Broad-range dielectric reflectors do find
use as interferometric beamsplitters in Fourier transform
infrared (FT-IR) instrumentation, and FabryPerot dielec-
5 REFLECTIVE MATERIALS tric filters are extensively used in nondispersive infrared
(NDIR) instrumentation.
Reflective optics are an important, often essential, part The metals commonly utilized as reflecting surfaces
of mid-IR optical systems. Unlike transmissive optics, are shown in Table 6. Normal-angle reflectance at 5 m
reflective optics are usually composite materials: the reflect- (2000 cm1 ) gives a good representation of the mid-IR
ing surface is rarely the same composition as the bulk mate- reflectance of these materials. The normal-angle reflectance
rial. Typically, a reflective optical element consists of the at the red HeNe laser wavelength is also given, since red
actual reflecting surface bonded to a plate, block, or assem- lasers are commonly utilized for alignment of multielement
bly which provides a means of registration and mounting. optical systems. Some Knoop hardnesses of these metals
Reflective optical elements generally have a surface are listed; however, vacuum-coated metal surfaces are
which is either flat, spherical, elliptical, parabolic, or often much softer than bulk metals. Vapor deposition can
toroidal. Toroidal mirrors are now mostly historical arti- be similar to a gentle snowfall: porous, soft, and poorly
facts, a way to approximate an elliptical surface with a consolidated. Again, there is a great deal of technique and
non-computer-driven diamond lathe, but they are some- art in coating metals. In some cases, ion bombardment is
times still used when ellipses with equidistant focal points used to hammer down and harden the metal surface after
are needed. Most curved mirrors are concave (except for coating.
the convex mirror used as part of Cassegranian telescopes). In practice, optical coating is an art, rife with recipes
Although on-axis optics (such as Cassegranians and source and secret techniques. In many cases, there are several
spheres) are often utilized, most of the reflecting optics underlayers of different materials to assure bonding of the
in mid-IR instrumentation are off-axis. The design freedom reflective surface to the substrate. In addition, it is common
allowed by inexpensive diamond machining has revolution- practice to overcoat optics with a very thin dielectric coating
ized mid-IR optical design. (most commonly SiO) in order to provide a harder, more
Reflective optics have the advantage of intrinsic achro- durable surface. Silver is routinely overcoated with MgF2
maticity. There is no wavelength-dependent focus shift, to prevent tarnish.
Table 6. Properties of some reflective metals.a
Metal Symbol R(?) R(?) n k Knoop Chemical Comments

(5 m) (HeNe) (5 m) (5 m) hardness reactivities
Aluminum Al 0.98 0.90 8.5 46.0 25, 40 Hg, OH , HCl, H2 SO4 Easy to sputter
Beryllium Be 0.98 0.54 2.8 21.1 60, 125 HC , OH Stiff, lightweight, poisonous
Chromium Cr 0.96 0.64 3.2 16.8 935, 1800 HC , HCl Used to bond gold to glass
Copper Cu 0.99 0.95 3.0 29.6 40, 163 O2 , HNO3 , HC , OH Used for heat dissipation
Gold Au 0.99 0.94 2.7 30.2 30, 45 O2 C Cl , CN Highest mid-IR reflectance
Molybdenum Mo 0.99 0.57 3.2 29.1 600 O 2 , HC
Nickel Ni 0.95 0.64 4.7 18.0 200, 557 HNO3 , S, OH , HC Used for optical conduits
Platinum Pt 0.95 0.59 5.0 20.0 40, 200 Cl2 , CN , S Inert
Rhodium Rh 0.96 0.81 6.8 25.0 400 H2 SO4 , HC Very hard surface
Silver Ag 0.99 0.98 3.5 32.0 32, 60 HNO3 , CN , O3 , H2 S, S Overcoat needed to avoid tarnish
Tungsten W 0.98 0.51 2.4 22.5 820 HNO3 C HF, O2 , HC
a R(HeNe) refers to reflectivity at 0.633 m. Reflectivities, n, and k values are at room temperature.
The presence of a thin overcoat or surface contamination glass substrates if they are produced using float glass or
fortunately has a much smaller optical effect than might be traditional optical grinding/polishing techniques. Other sub-
imagined. Because of the nature of electromagnetic reflec- strates such as ceramics and graphite-filled polymers are not
tion from metals, there is a near-null of the lights electrical yet widely used, perhaps because aluminum and steel fulfill
field at the metal surface, so surface contamination within most requirements.
a small fraction of the lights wavelength does not result A description of optical figuring methods follows.
in light absorption, even if the surface contaminant has
high absorptivity. Where many reflections occur (e.g. White
cells; see Long Path Gas Cells), incremental contamina- 6.1 Traditional lap grinding/polishing
tion of the reflecting surfaces may have no effect for a
The ancient method of creating an optical figure is by
long time, but subsequent contamination can result in rapid
grinding and polishing on a wheel. Flats and spherical
degradation.
surfaces can be made easily, and the technique has even
Reflection at grazing angles is a more complicated mat-
been adapted to produce paraboloids. However, deeply
ter. Approximate optical constants are listed in Table 6.
curved optical elements are difficult to produce in this way,
Reflectance at any incidence angle can be calculated using
and the glass blanks commonly utilized must be accurately
the Fresnel equations with these constants. In general, most
bonded to a mounting plate, a difficult step.
metals become less reflective near the grazing angle, and
transition metals such as nickel are actually more reflective
than gold at certain incidence angles (see Hollow Optical 6.2 Float glass
Conduits for Vibrational Spectroscopy). Dielectric mate-
rials are even more reflective than metals at the grazing Plate glass is often made by floating molten glass on to
angle but have not become popular, perhaps because they a liquid tin surface. Glass made by the float technique is
may be more susceptible to performance degradation by usually fairly flat, and mirrors with quarter-wave (visible)
minor surface contaminants. See ReflectionAbsorption surface accuracy can be cut from low-stress float glass.
Spectroscopy of Thin Films on Metallic Substrates for This technique is generally limited to flat surface mirrors,
more information about grazing angle reflection. although float-glass segments are used to fabricate cube
It should be noted that visible colors do not denote light corners and rotational molding has been used to form
absorption in the mid-IR region. An infamous example concave optics.
is black anodized aluminum, which is fairly reflective in
the mid-IR region! In general, the mid-IR spectra of dyes
and pigments do not have the high absorptivities that are 6.3 Electroforming
characteristic of their visible spectra. The premier light-
absorptive material in the mid-IR region is amorphous One of the earliest ways to reproduce complex optical
carbon, and paints containing high loadings of carbon black surfaces was by electrolytic replication (electroforming). A
should be used where low reflectivity is needed. positive-curvature optical master element (usually either a
paraboloid or an ellipsoid) is prepared, then electrolytically
coated with nickel. Electroplating is continued until the
6 SURFACE FIGURING OF deposit is thick enough to hold its shape, then the nickel
dish is separated from the master and given a final
REFLECTIVE OPTICS electroplating of rhodium. The resulting parabola or ellipse
(usually with a characteristic hole at the axis) is used as
One way of classifying optical elements is by the method
a very-low f-number on-axis optical element, or more
used to create the optical surface figure. In some cases,
often cut into a number of off-axis mirrors. These mirrors
the optically figured substrate also functions as the mount-
are then bonded to mounting plates. The optical quality
ing surface, and in other cases, this substrate is rigidly
of electroformed mirrors varies widely, and the optical
bonded to a mounting surface or block. The most com-
registration of electroformed mirror elements suffers from
mon substrate for reflective optics is aluminum: it is
the uncertainties in the cutting and mounting process.
lightweight, easy-to-machine, and a well-known substrate
for coatings. Aluminum has few flaws as a substrate for
reflective optics; even its fairly high thermal expansion 6.4 Epoxy replication
coefficient is usually easy to deal with. Stainless steel
is often utilized when reasonable resistance to corrosion Once the appropriate optical master element exists, it is
is needed. Flat mirrors and spherical mirrors may have possible to produce large quantities of reflective elements
with the desired optical figure from simple materials 6.6 Molding
and crudely machined blanks. The master element is
prepared by coating with mold-release agent. A blank is Although many camera lenses are actually molded, the bulk
machined with a corresponding surface (e.g. its negative molding of reflecting elements has not yet become popular
curvature matching the positive curvature of the master), for mid-IR instrumentation. Although polymer resins have
then specialized epoxy is applied and the blank is pushed very high thermal expansion coefficients, moldable glass-
into close contact with the master. After the epoxy has filled and graphite-filled polymers are available with very
cured, the blank is released, having now acquired a thin low thermal expansion coefficients.
epoxy surface with a precise optical figure. The epoxy-
surfaced blank is coated with aluminum to produce the
mirror. With reasonably uniform epoxy thicknesses and ABBREVIATIONS AND ACRONYMS
low-shrink epoxy resins, excellent optical accuracy can be
obtained. Since the blank is generally machined complete CVD Chemical Vapor Deposited
with mounting surface, the optical and mounting surfaces FAEL Flexural Apparent Elastic Limit
of the resulting element are accurately registered. Mirrors
replicated in this fashion are limited to low-temperature
service, but can be fabricated with hard-to-make figures FURTHER READING
such as inside corners.
M. Bass et al., Handbook of Optics, McGraw-Hill, New York,
Vols 1 and 2 (1995).
6.5 Diamond turning D.E. Gray (ed.), American Institute of Physics Handbook,
McGraw-Hill, New York (1972) (especially Chapter 6).
Precision lathes with diamond bits are now capable of T. Katsuyama and H. Matsumura, Infrared Optical Fibers,
machining surfaces to optical precision. Several blanks Adam Hilger, Bristol (1989).
are typically mounted on the lathe spindle and machined P. Klocek (ed.), Handbook of Infrared Optical Materials, Marcel
simultaneously. Often, a final low-tech polishing oper- Dekker, New York (1991).
ation is still needed to polish out the residual stri- R.H. Perry and D.W. Green (eds), Perrys Chemical Engi-
ations, which otherwise produce minor but unsightly neers Handbook, McGraw-Hill, New York (1984) (especially
grating-like dispersion of light. These optics are exten- Sections 3 and 23).
sively used in mid-IR process instrumentation because L. Ward, The Optical Constants of Bulk Materials and Films,
the finished coated optical element contains nothing but Adam Hilger, Bristol (1988).
metal and is not susceptible to solvent attack or thermal W.L. Wolfe and G.J. Zissis (eds), The Infrared Handbook, US
degradation. Government Printing Office, Washington, DC (1978).
INTERNET REFERENCES
Advanced Materials http://www.cvdmaterials.com/index.htm
Alkor optical crystal components http://www.alkor.ru/00000700/crystall.htm
Almaz Optics, Inc: optical materials http://www.almazoptics.com/
Amorphous Materials http://www.amorphousmaterials.com/
AO glass, IKS-34 http://www.hi-ho.ne.jp/dhow/aoglass.htm
Argus International, Inc. http://www.argusinternational.com/amtir.html
CASIX http://www.u-oplaz.com/OO.htm
Cleveland Crystals, Inc. http://www.clevelandcrystals.com/
Crystals Land http://www.goldbridge.com/crystalsland/crystald.html
CRYSTRAN: Optics http://www.crystran.co.uk/
Database of optical indices http://www.sopra-sa.com/
David L. Sullivans FT-IR spectra library http://www.che.utexas.edu/dls/ir/ir dir.html
EKSMA: optical components http://www.microcontrolnt.com/prodotti/eksma/optical components/
optical materials.htm
Fiber types http://www.optical-fibers.com/Fiber Types.htm
Focus on: material data http://www.sspectra.com/focus materials.html
GRINM products: infrared http://www.grinm.com/ir.htm

High-pressure diamond optics http://www.hpdo.com/
Infrared fiber systems http://www.infraredfibersystems.com/
Infrared window materials http://infrared.als.lbl.gov/IRwindows.html
ir-windows.com http://ir-windows.com/
Janos Optical Material Selection Guide http://www.janostech.com/
Keystone Crystal Corp. http://www.zbzoom.net/kcc/index.html
KORTH http://www.korth.de/tra/compare.gif
Lattice Materials Corp. http://www.latticematerials.com/
LAYERTEC http://www.layertec.de/index.html
Luxpop Property Calculator http://www.luxpop.com/
MarkeTech International, single crystals http://www.mkt-intl.com/single.htm
MaTecK, Material-Tech. & Kristalle http://www.mateck.de/index.asp
Matweb Materials Property Data http://www.matweb.com/help.htm
Molecular Technology GmbH http://www.mt-berlin.com/
Multimode fluoride glass fibers http://thorlabs.com/thorcat/main/2725-000.HTML
NIST Property Data Summaries http://www.ceramics.nist.gov/srd/summary/advmatdb.htm
Optovac http://www.optovac.com/
Oriel Instruments Homepage http://www.newport.com/oriel/
Photonics Directory http://www.photonics.com/
Precision optics http://www.gsilumonics.com/c01optics frame/about.htm
Princeton Scientific Products http://www.princetonscientific.com/pages/crystal.html
Principal Metals http://www.principalmetals.com/utilities/hardness.htm
Remspec Corporation http://www.remspec.com/
Saphikon Sapphire Products http://www.saphikon.com/
SEDI Fibres Optiques http://www.sedi-fibres.com/index.php?th=68
Spectroscopy Central Ltd. http://www.spectroscopy.co.uk/irwcom/
SPIE http://www.spie.org/
TUFTRAN http://www.cvdmaterials.com/tuftran.htm
TWO-SIX Welcome to II-VI http://www.ii-vi.com/pages/prod-iivi-materials-znse.html
Vidrines tables http://www.vidrine.com/stuff.htm
VITRON Spezialwerkstoffe GmbH http://www.vitron.de/vhpe.shtml
Monochromators for Near-infrared Spectroscopy
Frank A. DeThomas1 and Paul J. Brimmer 2

1
ABB Automation, Woodstock, MD, USA
2
Foss Pacific, North Ryde, Australia
1 INTRODUCTION can have a larger number of grooves per millimeter than

ruled gratings (improving dispersion and resolution), have
Near-infrared spectroscopy (NIRS) is the measurement of lower levels of stray light and can be made in custom sizes.
the wavelength and intensity of absorption of near-infrared Holographic gratings allow greater design flexibility, and
(NIR) light by a sample. The NIR spectral region spans allow for improvement in performance over ruled gratings.
the range from 700 to 2500 nm (14 2854000 cm1 ) with Instrument design involves a series of compromises made to
absorptions representing overtone and combinations from achieve the desired speed, size, flexibility and performance,
the fundamental molecular vibrations found in the mid- and is the focus of this article.
infrared (MIR) region. These overtones and combinations This article focuses on aspects of the design of single-
mainly are associated with CH, NH, OH and SH beam NIR spectrophotometers based upon concave holo-
functional groups. NIRS uses chemometrics to analyze graphic gratings and single-element detectors. It is best
the functional groups present in a sample mixture. As to begin by discussing why monochromators have been
a quality and process control technique NIRS analysis historically used in the NIR region. Detector sensitivity,
methods are applied to raw material, intermediates and source intensity and sample absorptivity together deter-
finished products. NIRS historically was developed as a mine the optimum spectroscopic system. In this region
quantitative analysis technique, but during the past 10 years of the electromagnetic spectrum, the availability of high-
chemometrics methods have led to qualitative applications. intensity sources, sensitive detectors and holographic grat-
Detection limits depend on a number of factors, including ings enables the monochromator configuration to be effi-
resolution, the intensity of the absorption band, and the cient in the UV, visible and NIR regions.
complexity of the sample matrix. Typical detection limits Table 1 shows the sensitivity of the detectors, and the
are 100 to 500 ppm for transmission and 500 to 1000 ppm intensity of light sources for MIR, NIR and UVvisible
for reflection measurements. Applications include analysis (UVVIS) regions. It can be seen from Table 1 that the
in the agricultural, food, chemical, medical, pharmaceutical, optimum configuration is different in the different spectral
polymer and textile industries. regions. For both UVVIS and NIR regions the intense
Historically, there have been a number of different sources and sensitive detectors and the fact that all of
monochromator designs using one, two or even three grat- the wavelengths impinge onto the detector simultaneously
ings to achieve the desired scan range, resolution, dynamic make it difficult to reap the benefits of a multiplexing
range and stray light. Multiple gratings are used to improve approach such as Fourier transform spectrometry. In these
the dispersion achievable by single gratings, and to provide regions, dispersive systems match the energy delivered by
a means of reducing the amount of stray light in the sys- the source to the energy requirements on the detector very
tem. Today gratings are generally ruled holographically, so effectively. In the MIR, where the sources are less intense,
that they are robust and easy to manufacture. They also and the detectors are less sensitive, the energy require-
ments are better matched by having all of the wavelengths
John Wiley & Sons Ltd, 2002. impinge onto the detector simultaneously, which explains
2 Instrumentation for Near-infrared Spectroscopy
Table 1. Comparison of detector sensitivity, source intensity and sample absorptivity for different spectral
regions.
Component UVVIS NIR MIR
Detector sensitivity High (1013 ) Fairly high (1011 1013 ) Low (108 1010 )
Source intensity High (deuterium, tungsten) High (tungsten) Low (globar, ceramic)
Sample absorptivity Medium Low High
Common system Monochromator Monochromator Fourier transform
why Fourier transform techniques have replaced dispersive diffraction gratings combine the functions of optical
systems entirely in the MIR region. imaging and diffraction into a single element. This
In a Fourier transform near-infrared (FT-NIR) spectrom- simplifies the alignment and reduces the number of
eter, the signal-to-noise ratio (S/N) near the centerburst of optical elements and mounts required, thus reducing the
the interferogram can often exceed the dynamic range of manufacturing cost. Concave diffraction gratings were
the analog-to-digital converter. Also, the very high signal invented by Henry Rowland1 in 1882 and consist of finely
incident on the detector of FT-NIR spectrometers can lead spaced lines produced on the face of a spherical surface.
to detector saturation and/or nonlinear detector response, In the 1950s, Seya2 and Namioka3,4 proposed a basic
which results in poor photometry and hence to significant monochromator that consists of a single concave diffraction
deviations from Beers law. These obstacles can obviously grating with the locations of the entrance and exits slits
be overcome, as there are commercially available FT-NIR fixed and the grating rotating about the center of its face.
systems. Nevertheless, the ability to easily match the energy Noda et al. developed the modern theory for concave
requirements of NIR detectors to the energy output of holographic diffraction gratings in 1974.5
NIR sources by dispersive spectroscopic methods is why Holographic concave diffraction gratings result from the
monochromators have been historically used and are still application of laser technology, and the spatial modulation
used in the NIR region. capability of photo-resistive materials to form the hyper-
bolic lines described by Rowland. At the intersection of
two lasers, interference fringes are formed, which form the
1.1 Diffraction gratings hyperbolic lines in the photoresistive material deposited
onto a glass substrate. Since these gratings are formed
The fundamental grating equation for a plane grating (see
due to an optical, rather than a mechanical phenomenon,
Figure 1) is given by:
aberration-free images are formed, and the amount of stray
sin a C sin b D 106 m nl 1 light is much lower than in ruled gratings. Holographic
gratings can be generated on many other types of surfaces,
where a is the angle of incidence, b is the angle of diffrac- including plane, spherical and toroidal.
tion, m is the diffraction order (integer), n is the groove
density (grooves per millimeter), l is the wavelength (nm)
and f D a C b. 2 MONOCHROMATOR COMPONENTS
In NIR spectroscopy, the use of concave gratings has
superseded that of plane diffraction gratings. Concave This section will review the major components for a
monochromator and discuss some of the design considera-
tions that are made.
2.1 Concave holographic gratings

LB Exit
slit
For NIR measurements, a concave grating with a groove
density of 1200 grooves mm1 or more is generally used.
LA
Normal To specify a concave holographic grating for a monochro-
mator system, a number of instrumental factors must be
Entrance determined. The general design rules are to have highest
slit
groove density that will accommodate the required wave-
Figure 1. Fundamentals of the grating equation. length range, and provide the highest dispersion for the
Monochromators for Near-infrared Spectroscopy 3
required spectral resolution. The largest entrance and exit detection. The challenge is to develop controllable drive
slit widths are selected to maximize the energy throughput mechanisms that provide the necessary wavelength accu-
without degrading the desired spectral resolution. Concave racy and precision for the application.
gratings with longer focal lengths are selected to make A number of methods of scanning the grating have been
alignment easier and more reproducible (but with a loss investigated. Many are mechanical in nature, in which
in light-gathering ability). Large size gratings are selected the grating is rotated about a fixed axis with a motor.
to offset this loss. The spectrum is sampled over the portion of the grat-
Some of the main specifications for a concave holo- ing rotation where the radiation of interest passes through
graphic grating include: the exit slit. The motor provides direct velocity control
of the grating so that it can be scanned in the forward
Spectral region (nm). This is the region of interest and backward directions. In another variation two or three
that is specified by the minimum and maximum wave- gratings are mounted on the same shaft and rotated to
lengths. increase the data collection cycle through 360 . In another
Linear dispersion (nm mm1 ). This is the extent that design, electromagnetic drives are used to scan the grat-
the spectral interval is spread out across the focal field ing in one direction and a spring returns the grating to
or the instruments ability to resolve spectral detail. the starting wavelength of at the end of the scan. The
Wavelength of peak efficiency (nm). Since the effi- use of mechanical cam drives was once the method of
ciency of the concave holographic grating varies with choice and such drives are still available; however, they
wavelength, the wavelength of peak efficiency must are not as popular today because of the development of
be specified to show the spectral region of optimum small industrial drive motors that are simpler and more
performance. robust.
Slit height (mm). The slit dimension must be matched The combination of a motor and an optical encoder has
to the concave holographic grating dimension. proven to be the most rugged and cost-effective method of
Focal length (mm). This is distance between the slit
providing a reliable moving grating system. Because of the
and either the collimating mirror (if a plane grating is
trend towards higher resolution, optical encoders provide a
being used) or the concave grating.
needed increase in wavelength accuracy in the monochro-
Resolution (nm). This is the minimum separation of
mator. Small, high-precision industrial motors are able to
two adjacent absorption bands that can be distinguished
provide precise movement of the grating, increasing the
by the spectrometer.
wavelength precision achievable by the drive mechanism.
Aperture ratio (f number). The f number is given by
the focal length of the diffraction grating divided by
its diameter. The smaller the f number, the greater the
amount of energy from the source passed through the 2.3 Sources
monochromator and the greater the total energy flux.
Blank dimension (mm). This is the outer dimension The design objective is to focus as much energy as possible
of the substrate used to form the concave holographic from a source being operated at the lowest wattage into
grating the entrance slit. This is typically accomplished with a
Degrees of rotation. This is the number of degrees that combination of optics that are either independent of the
the concave holographic grating is rotated during the source or integrated with the source in the form of a
scan, which determines the scan range. reflector. In general there is an inverse relationship between
Groove density (g mm1 ). This is the number of source wattage and its lifetime. As the source power is
grooves per millimeter on the grating. The higher the increased to increase the available energy, its lifetime
number of grooves the greater the linear dispersion (and decreases. A balance must be developed that allows the user
therefore resolution). Also the groove density deter- to operate the monochromator for periods of from 6 to 12
mines which region of the electromagnetic spectrum months without having to change the source, while ensuring
will have the greatest efficiency. that enough photons are passed into the monochromator to
perform the desired analysis.
A number of lamps are available for the NIR region,
2.2 Grating drive including tungsten, tungstenhalogen, xenon and mer-
curyxenon. Each has advantages and disadvantages with
The design objective is to rotate the grating in a con- regard to cost, intensity, wavelength coverage, stability
trolled fashion so that different wavelengths of the dispersed and operating lifetime. Tungstenhalogen lamps are the
radiation are passed reproducibly through the exit slit for most commonly employed in the NIR region because
they produce a high energy, smooth spectral output that entrance slit. A source without a back reflector is typically
approximately follows the black-body curve corresponding used in combination with one or more mirrors to focus the
to the lamps color temperature. beam on the entrance slit.
The spectral range of the tungstenhalogen lamp is Historically, NIR monochromators have been prone to
from 300 to 3500 nm, where the longest wavelength is error in calibration transfer associated with changing the
determined by the opacity of the quartz envelope. Tung- source. This is not a simple problem because it is asso-
stenhalogen lamps feature a tungsten coil filament mount- ciated with the source, the sample and positioning errors
ed in a quartz glass envelope that is filled with an inert gas associated with the grating drive, and is greatly influenced
with a trace of a halogen (normally bromine). The presence by the complexity of the calibration methods used. In gen-
of the bromine gas creates a regenerative process known as eral, the more complicated the calibration with respect to
the halogen cycle. Once the filament is up to temperature, the number of wavelengths or factors used for the anal-
tungsten slowly evaporates from the filament and deposits ysis, or the larger the wavelength range or the lower the
on the relatively cool quartz envelope, where it reacts with constituent concentrations, the more sensitive the analytical
the bromine gas. The tungsten bromide has a relatively high performance becomes to changing the source.
vapor pressure and, at the temperature of the quartz walls, Small differences in the alignment of reflectors, mirrors
it will evaporate, so that the quartz walls are kept clean and and filaments cause changes in the emission curve of the
clear. Since the whole bulb is at the same high temperature, monochromator. The image of the filament is focused
the tungsten halide does not deposit anywhere else on the through the slit and carried through the monochromator
bulb. When the tungsten bromide reaches the filament, the system. Since no two sources are absolutely identical,
temperature is high enough to lead to decomposition, caus- a change in the source characteristics can affect some
ing tungsten metal to be deposited back on the filament, calibrations. Changing the source causes variations in both
releasing the bromine to go through another cycle and pick wavelength and absorbance scales.
up more tungsten from the quartz walls. The temperature Intensity differences between sources are commonly
of the bulb has to be kept high (250300 C) in order for compensated by taking a ratio (I/I0 ) of a reference or back-
the cycle to operate. The purpose of the inert gas in the ground scan (I0 ) to a scan of the sample (I). With strong
bulb is to reduce the rate of evaporation of the tungsten absorbers, the linearity of the system comes into play when
from the filament; in a vacuum, it would evaporate much the absorbance is relatively high, which is affected by the
faster at the normal operating temperature of the filament. amount of stray light. High absorbance can also lead to
If a hot spot forms on the filament, tungsten will evaporate changes in apparent band position and resolution, due to
from that spot more rapidly, leading to a process that even- this nonlinearity and spectral saturation. The most common
tually causes the filament to break. Also, since the hot spot approach of correcting for wavelength shifts is to use a sec-
is at a different temperature than the rest of the filament, ondary standard to correct for changes on the wavelength
it has a different emission characteristic, which causes the axis. Typical standards used for these purposes are inter-
emission curve of the bulb to change with time (resulting ference filters, polystyrene films or the National Institute
in wavelength shifts). of Standards and Technology (NIST) Standard Reference
Tungstenhalogen lamps have a color temperature bet- Materials (SRM) 1920, which is a mixture of rare earth
ween 2900 and 3400 K and are available from 10 to oxides which have known peak positions in the NIR region.
250 W with operating voltages from 6 to 24 V. Lamp The peak positions of these standards are measured with a
voltages are precisely regulated to maintain a constant given monochromator, and mathematically corrected with
envelope temperature (usually between 250 and 350 C) these known reference points.
and hence allow good instrumental precision. Cooler tem- Some vendors have switched to pre-aligned sources
peratures do not allow the halogen cycle to take place, where they are aligned in the x, y, and z direction to pro-
causing the bulb to blacken, while higher temperatures will vide constant energy and limit wavelength shifts caused by
cause excessive oxidation of the filament. When the correct changing the source. Another means of correcting wave-
operating conditions are met, the lamp lifetime can reach length shifts due to variations in the source is to add a dif-
10 000 h. fuser at the entrance slit in the form of screens, fiber optics
Tungstenhalogen sources are available for a number or etched glass to randomize the energy before it enters the
of non-scientific uses and are easily adapted to scientific monochromator. These approaches are not popular because
applications. These sources are available with and without they lose between 20 and 60% of the energy throughput,
back reflectors, which greatly affects the energy density and which can only be compensated by increasing the source
dispersion. Sources with metal-coated back reflectors are wattage (at the cost of source lifetime) or by increasing the
used to increase the energy density directed forward to the size of the entrance slit (at the cost of resolution).
2.4 Entrance and exit slits Table 2. Orders of NIR radiation.
Order (m) Wavelength (nm)

The entrance slit provides a narrow source of light for
dispersion and refocusing by the grating in the plane of the
1 1200 1400 1600 1800 2000 2200 2400
exit slit. The exit slit selects a narrow band of the dispersed 2 600 700 800 900 1000 1100 1200
spectrum for observation by the detector. In practice the 3 400 466.7 533.3 600 666.7 733.3 800
entrance and exit slits are of equal width allowing the image 4 300 350 400 450 500 550 600
of the entrance slit to pass to the exit slit. It should be noted
that the relationship between slit width and resolution is not
order from 1200 to 2400 nm will also have a number of
constant across the scan range and is usually measured (and
higher orders, as shown in Table 2. For a holographic grat-
specified) at the wavelength where the system has its peak
ing with a fixed groove density (n), a certain wavelength
response.
(l) and known angles of incidence and diffraction, a and
The design objective is to have the largest slit width for
b, equation (1) shows that the diffraction order (m) varies
the required resolution. The larger the slit width, the more
photons reach the detector. As the width of the entrance inversely with wavelength, so that if the diffraction order
slit is increased, the additional energy improves the S/N is doubled the wavelength is halved.
but decreases the resolution, which potentially decreases For example, at 2400 nm from the first order, there is
the selectivity of the analysis. If the slit width is reduced to a lesser but still considerable amount of 1200 nm energy
improve the resolution, the amount of energy reaching the from the second order, and a lesser amount again of 800 nm
detector, and hence the S/N, is reduced. This directly affects from the third order, etc. The detector does not know
the limits of detection and precision of the measurement. what wavelength the energy is, but only that energy strikes
The entrance slit is therefore a point of compromise in the it and is detected as a signal. If not somehow eliminated
design of the monochromator system, and will depend on from the system, this extra stray light will put an upper
whether S/N or resolution is more important for the desired limit on the maximum absorbance level measurable, which
analysis. will decrease the dynamic range of the spectrometer.
For diffuse reflection measurements, large exit slits are This is illustrated by the plot shown in Figure 2, which
desired to increase the sampling area (for inhomogeneous shows the relationship between absorbance and concen-
materials, for example) and to increase the amount of tration at several levels of stray light. In the absence of
light collected in these more light-limited measurements. In stray light, absorbance is assumed to increase until there
transmission measurements, which gather more light gener- is 0% transmitted light, or infinite absorbance. If there is
ally than reflection measurements, smaller slits are desired stray light present, the amount of transmitted light cannot
to minimize beam dispersion with increasing distance from decrease beyond this level to 0%, which limits how high the
the exit slit, producing a more collimated beam through the measured absorbance can increase, as shown in this figure.
sampling area. Since absorbance as defined by the BeerLambert law has a
logarithmic relationship to transmittance, stray light of 10%,
1% and 0.1% will result in upper absorbance limits of 1, 2,
2.5 Wavelength, order and interference filters and 3 absorbance units, respectively. Partway between zero
absorbance and this upper limit, the relationship between
The tungstenhalogen light source emits a continuum of
wavelengths across the visible and NIR regions. The pri- 4.0 0%
mary objective of the monochromator is to spatially sep- 3.5

arate these wavelengths with a known resolution and to 3.0 0.1%
Absorbance
Stray light
direct them sequentially to the detector. Concave holo- 2.5

graphic gratings, as discussed earlier, separate the con- 2.0 1%
tinuum from the source to discrete wavelengths, but have 1.5
several spatially overlapping orders that must be separated 1.0 10%
or eliminated. 0.5
Higher orders of the radiation may be thought of as impu-
rities to the first order, and act as stray light. The sample 0 10 20 30 40 50 60 70 80 90 100
being analyzed may or may not absorb this radiation, but Concentration
it does interfere with the monochromators detection sys- Figure 2. Relationship between absorbance and concentration at
tem. For example, an NIR monochromator that has its first several different levels of stray light.
absorbance and concentration will become nonlinear, which Table 3. Common detectors for NIRS.
adversely affects quantitative and qualitative analyses.
This stray light from higher orders of the grating must Type Detector Operating range (nm) D
therefore be filtered out so that only the first-order radiation
Photodiode Si 2001100 1013
reaches the sample and the detector. The first-order radia- Photodiode Ge 4001800 1011
tion diffracted from the grating is always the most intense, Photodiode InGaAs 9001700 1012
so that first-order radiation is generally preferred. However, Photovoltaic InAs 10003800 1010
some monochromators use the first and second orders of the Photovoltaic InSb 10007000 1010
grating to extend the scan range. Photoconductive PbS 9003300 1011
Photoconductive PbSe 10004800 109
The higher orders of radiation from a grating can be
largely eliminated through the use of bandpass interference
filters. Interference filters employ thin metallic or dielectric For diffuse reflection measurements in the 1100 to 2500 nm
layers to produce the desired interference at a specified region a matrix of large area (1 cm2 ) PbS detectors is
wavelength to reject the unwanted radiation by selective used to increase the area of detection and improve the
reflection. A cut-off wavelength filter can be used to filter S/N in these light-limited situations. A similar approach
out all absorption below a certain wavelength range. When is used for diffuse transmission in the 700 to 1100 nm
used to eliminate higher orders from a diffraction grating, region, in which case a matrix of large silicon detectors
these filters are called order-sorting filters. Order-sorting is used to increase the efficiency of measuring the diffusely
filters are either fixed (one filter) for a narrow scan range transmitted radiation without the use of additional focusing
(1200 to 2000 nm) or moving (multiple filter wedge or optics. Detector performance is optimal when the size and
wheel) for a large scan range (400 to 1100 nm, 1100 to shape of the beam are matched to the size of the beam at
2500 nm). Interference filters are generally mounted after the sample.
the grating, since in a beam of white light, heating of the In general, as the size of the detector increases, its
filter can occur and change the rejection wavelength range. response time constant increases leading to a slower res-
Moving or changing the interference filters (commonly ponse. The time constant of the detector, or matrix of
effected by a stepper motor) as the grating scans allows a detectors, needs to be much shorter than the measurement
larger scan range to be achieved without interference from time for a single resolution element if the band shape is
higher orders. These optical cut-off devices change from not to be deformed. Deformation of the band shape adds
their maximum transmission to their maximum rejection artificial features to the spectrum and limits the ability of
capability in about 100 to 400 nm, so the cut-off needs to calibrations to be transferred between spectrometers.
be at least this far away from the beginning of the higher
order wavelengths, or some residual stray light may not be
effectively rejected from the system. 3 DESIGN CONSIDERATIONS
There are many things to consider in the design of a

2.6 Detection monochromator, including the type of components selected,
and how they are configured. One need not only con-
There are three different types of single element detectors sider the performance specification on the absorbance and
used in the NIR region: photodiodes, photovoltaic and pho- wavelength axes, but also the requirements of the applica-
toconductive. The actual detector used is determined by a tion of the instrumentation as well. Thus such factors as
number of factors, which include sensitivity, noise, oper- the dynamic range of an analysis, the spectral resolution
ating wavelength range, response time, geometry (sizes) required and the measurement environment (e.g. ambi-
and dynamic range (linearity). Table 3 lists some of the ent light in a process measurement), which can determine
more common detectors for the NIR region along with their whether a sample is placed before or after the diffrac-
operating range and sensitivity (D ). Most of the detectors tion grating, must also be taken into account. In addition,
can operate at room temperature or can be thermoelectri- coupling to fiber optics, either externally or internally to
cally cooled. Thermoelectrical cooling generally shifts the the monochromator, can affects the overall monochromator
operation range to shorter wavelength and increases the sen- design. Finally, the users of NIR technology are demand-
sitivity with respect to room-temperature operation. ing the ability to transfer calibrations and libraries between
The number of detectors and their dynamic range, size, monochromators, which also affects the instrument design
shape and speed (time constant) all affect the resulting in terms of the ability to manufacture identical instruments.
spectrum. Detector choice is critical for NIR applications. In this section, these considerations are discussed.
3.1 Stray light This lower absorptivity is a distinct advantage for sam-
pling because it eliminates the requirement of making dilute
Stray light is light that is measured at the detector other than mixture of the analyte in a solvent or non-absorbing solid
what has passed through the monochromator, or light other prior to each analysis. For transmission measurements in
than the wavelength of interest. As shown earlier, it has the NIR, pathlengths typically start at 0.5 mm at 2500 nm
deleterious effects on the linearity in quantitative analyses, and can approach several centimeters at 700 nm. For NIR
and it deforms the shape of strong absorption bands, which diffuse reflection measurements, the depth of penetration
adversely affects qualitative analyses as well. Stray light depends on particle size and packing density but generally
can come from light scattered from internal instrument starts out at about 0.1 mm for most powders and increases
components, but can also come from higher orders from the to several millimeters as the visible region is approached.
diffraction grating, unless properly eliminated using order- This allows more of the sample to be analyzed, thereby pro-
sorting filters, as discussed earlier. viding a more representative analysis of the bulk properties
Internal reflections and scattering can be minimized by of the sample.
making all internal surfaces a matte black color, strategic There is a challenge to the instrument designer to achieve
placement of baffles or light shields inside the monochro- accurate measurements of both reflectance and transmit-
mator and careful finishing of the optical surfaces. External tance. For reflection spectroscopy, the beam size at the
reflections from windows or probes at the sample interface sample depends on the samples average particle size. Thus
are minimized by designing the optical geometry of these selecting the optimum detector area can present a problem.
interfaces to minimize the collection of specularly reflected If the sample is a fine powder of, say, 100 mesh (<0.2 mm)
radiation. For example, reflection measurements are made then the size of the beam at the sample can be small and
by illumination normal to the sample surface, with detectors a representative sample will still be observed. In this case,
mounted at 45 relative to the sample to minimize collection the size of the detector can be similar to that of the detec-
of specularly reflected light and to maximize the diffusely tors used for transmission measurements. As particle size
reflected component (which contains the information about increases and/or becomes more heterogeneous, sampling
the sample). errors increase. To reduce sampling error, the sample can
The most common source of stray light in a holographic be ground into a uniform powder, although there is a risk
system is randomly scattered light due to surface imperfec- of changing or destroying the constituents to be measured
tions on the optical surfaces of one or more components or and an involved preparation step is added. The alternative
interfaces in a spectrophotometer. These components can is to increase the sampling area to reduce the sampling
include radiation reflected from slits, the surfaces of order errors. Increasing the sampling area is achieved in a number
sorting filters, wavelength standards, entrance and exit win- of ways, including:
dows, fiber optic probes or sample cell widows. All these
Increasing the focal length of the mirror used to focus
interfaces can contribute to the overall stray light of a
the beam at the sample. This usually involves a cor-
system and limit the dynamic range of the measurement
responding increase in the area of the focal point (and
system. In order to be useful over a wide dynamic range,
the detector, since it is best to have the sizes of these
the stray light should be kept to less than 0.1%.
areas match). Often, multiple detectors are used.
Moving the sample by a rotatory or linear motion.
There are many devices used in NIR spectrometers
3.2 Sample absorptivity that rotate or transport the sample to increase the area
of interrogation. Although this approach increases the
Absorption bands in the NIR arise from molecular vibra- effective sampling area, it can also increase the noise
tions whose fundamental transitions appear in the MIR. NIR due to sample movement during the NIR scan. The
absorptions are caused by combinations of several funda- best of these devices actually steps the sample holder
mental modes (stretching and bending) or from the first so that data are only collected during the stationary
through third overtones of the fundamental XH stretch- period of the movement, thus eliminating a potential
ing vibrations. Combination bands are usually about 10 source of error.
times weaker than the weaker of the two fundamentals from
which they are derived. The intensity of the first overtone
may be one to two orders of magnitude weaker than the 3.3 Spectral resolution
corresponding fundamental, with concomitant reduction in
intensity as the overtones increase in order and approach NIR spectrophotometers are not perfect devices, produc-
the visible region (700 nm). ing an apparent broadening of narrow spectral bands.
The instrument resolution is a measure of the instruments Power

fundamental ability to separate adjacent spectral features. Electronic board supply
The bandpass is the spectral wavelength interval that may
be isolated by the monochromator. The bandpass is depen-
dent primarily on the width of the entrance and exit slits, Holographic
Entrance slit grating and drive
but can also depend on the width of the grating, system Source
aberrations, the time-constant of the detector and the rate
at which the grating is scanned. Order Wavelength
The instrumental bandpass can be determined by passing sorters standards
light from a reasonably monochromatic light source into the Exit slit
fixed entrance slit of the monochromator. The output from
Sample
the exit slit is detected and recorded as intensity versus
wavelength. The bandpass is then defined as the full width
at half the maximum height of the measured signal. (a) Detector electronics
The bandpass of holographic monochromators is limited
by the grating to 67 nm and is typically set above this Power
Electronic board supply
limit to about 10 nm. In the NIR region this corresponds
well with the absorption bandwidth of the typical analytes S
a
measured, although the width of some bands in the spectra m Holographic
of highly crystalline materials and liquid solvent are less p Entrance slit grating and drive
than this. A smaller spectral bandpass will not dramatically Source l
e
change the appearance of a measured spectrum of most
Wavelength
naturally occurring materials because the natural widths of Order
standards
sorters
the absorption bands are greater than the spectral bandpass Exit slit
of the monochromator, and generally only worsens the S/N
of the measurement.
(b) Detector electronics
Figure 3. (a) Pre-dispersive and (b) post-dispersive monochro-
mator designs.
3.4 Pre-post-dispersive monochromators
to the entrance slit of the monochromator, from which the
Monochromators can be applied in either a pre- or post- polychromatic light is focused upon the grating for disper-
dispersive design depending upon the requirements of sion. As the grating is rotated (scanned), the monochromatic
the application. In the pre-dispersive design shown in light is focused through an order-sorting filter to the exit
Figure 3(a), white (polychromatic) light from the source is slit and detector. The post-dispersive design is used to min-
focused through an entrance slit and passed to the grating imize sensitivity to stray light (room light) when sampling
for dispersion. The grating is rotated to scan the wave- in open areas.
length range of interest. In many designs, it takes less than
a second to scan the entire NIR spectrum. Radiation from
the grating is focused through an order-sorting filter and 3.5 Designing with fiber optics
the appropriate wavelength is passed through the exit slit.
The NIR beam may also be passed through a wavelength The largest and most common application of fiber optics is
standard to correct for changes in the wavelength axis. to transfer the light from the instrument to the sample and
After the exit slit, monochromatic radiation is focused back to a detector. These fiber-optic sample interfaces can
upon the sample via conventional or fiber optics, and the be either single fibers (30050 m in diameter) or multiple
diffusely reflected or transmitted radiation is collected for fibers (up to hundreds of individual fibers of 1100 m
detection. The pre-dispersive design minimizes the energy diameter). The fiber(s) are used to conduct the radiation to a
reaching the sample thus minimizing heating effects upon remote sample location for illumination typically through a
the sample. probe (interface) designed to suit a particular measurement.
In the post-dispersive design shown in Figure 3(b), poly- A detector may be mounted at the end of the probe or a
chromatic light from the source is focused upon the sam- second set of fiber(s) can collect the diffusely reflected or
ple either directly or through a fiber optic interface. The transmitted radiation for return to the detector, which is
reflected or transmitted light is then collected and passed mounted next to the monochromator.
Fiber-optic sample interfaces force a few design changes 3.6 Standardization of spectra
to be made in monochromator designs. The fibers simply
act as a means of transferring light, much like conventional The United States Pharmacopeia (USP) and other inter-
optics, but with greater flexibility. The major consideration national regulatory agencies are moving to set instrument
is to deliver as much of the NIR beam as possible into test requirements for both the wavelength accuracy and
the fiber-optic. For fiber-optic bundles, this is usually done repeatability and photometric response on NIR spectrom-
by matching the size of the fiber bundle to the size of eters, because these instruments are becoming commonly
the NIR beam. Coupling with single fibers usually requires used to generate quality data on final, released pharma-
some focusing of the NIR beam onto the fiber, as they are ceutical products. The USP guidelines on NIR mandates
very small in diameter, and would otherwise not deliver charting of the response against the results obtained during
sufficient NIR radiation to the sample. the initial instrument installation for wavelength accuracy
While the most familiar use of fiber optics has been as and repeatability (over the specific wavelength regions used
a sample interface, until recently the use of fiber optics in the calibration or library), as well as the photometric
as an internal component of a monochromator has been response over time. The problem is that there are not many
very unusual. Few texts discuss the design features that well-characterized standards in the NIR region. The data
fibers can bring to a spectrophotometer. Instead of treating that are available are not accurate or precise enough to
the monochromator as the smallest module of the system, meet the requirements for calibration transfer, which has
fibers allow the monochromator to be divided into sub- led individual vendors to develop their own solutions.
modules or individual components that can be separated. In addition, there is increased interest in obtaining NIR
This approach greatly increases the flexibility of design and spectrometers that give identical, or nearly identical spec-
the ability to configure a solution for a particular analytical tra so that quantitative calibrations and qualitative libraries
problem. can be transferred between systems. This would drasti-
In the design of a spectroscopic system for the NIR cally decrease the time and effort in method development,
region, the design flexibility afforded by fiber optics is which allows the NIR to provide useful data almost imme-
important. Replacement of conventional internal optics with diately, rather than after a lengthy calibration or library
fiber optics in the following components can reap significant development. It also makes the NIR more cost effective,
benefits in design of instrumentation: as the costs of analysis of calibration samples, and need of
chemometric expertise in developing methods is no longer
Slits. Fiber-optic bundles can be used to form the
entrance and exit slits of the monochromator system needed.
eliminating the need for mechanical slits. In order to make each monochromator provide nearly the
Entrance slit. A randomized fiber-optic bundle located same NIR spectra, the wavelength axis must provide similar
between the source and the entrance slit to the mono- resolution and wavelength accuracy across the scan range.
chromator prevents the filament image from being This initially would seem difficult, as the movement of the
carried through the optical system, resulting in a more grating is not linear but sinusoidal, and the wavelength scale
uniform illumination of the samples from instrument to of the spectrometer must be linearized. As part of this lin-
instrument. In addition, the use of fibers can also isolate earization process, the wavelength axis is compared and
the heat-generating lamp from the heat-sensitive optic corrected to internal and/or external wavelength standards.
components. Changing the number of the fibers and Wavelength standards vary between reflection and transmis-
their diameter controls the amount of energy transmit- sion as well as between vendors. Currently used standard
ted through the fibers. The larger the number of fibers, reference materials (SRMs) include National Institute of
the greater the amount of energy delivered, but at a Standards and Technology (NIST) SRM 1920 rare-earth
greater cost. oxides, polystyrene films and filters made from didymium
Exit slit. Fibers can be used to form the exit slit oxide, holmium oxide, dysprosium oxide, erbium oxide,
of the monochromator creating an effective coupling and others. The NIST SRM 1920 is used for reflectance
to the system. Fibers allow the beam to be shaped to measurements and the American Society for Testing and
the optimum illumination configuration for a particular Materials (ASTM) peak-finding algorithm is used to locate
sample, e.g. round, square, large, small. peaks within the maximum degree of precision and accu-
racy possible, given the instrumental and mathematical
Replacing components with fiber optics allows impro- parameters. Films and filters are typically used internally so
vements in the monochromator design, and increases the they are independent of the sampling mode. The stability
flexibility in optimizing the monochromator for difficult of the standards is always in question and a traceable path
applications. back to a master instrument should be used to demonstrate
that the wavelength and photometric response of a standard For quantitative applications the S/N is the most impor-
has not changed (within certain limits) over a stated period tant instrumental parameter. The S/N determines the overall
of time. accuracy and precision of a measurement, the minimum
A photometric repeatability test may be conducted in detectable concentration and even the number of sam-
the reflection mode using poly(tetrafluoroethylene) (PTFE) ples needed for a calibration. For qualitative applications,
samples doped with different levels of carbon black. A wavelength precision is the most important instrumental
series of these standards (2%, 10%, 20%, 40%, 60%, 80% parameter as calibration algorithms tend to be very sensitive
and 99%) with known photometric response is used to to band position and shape.
determine the instruments linearity. For transmission, a With the cost of chemometric calibrations exceeding,
series of neutral density screens can be used to achieve the doubling or even tripling the cost of the NIR monochroma-
same results. In either case, a sample reflecting or trans- tor, calibration transfer has developed into a major issue.
mitting 80% of the incident radiation is used to monitor Calibration transfer is mostly affected by the wavelength
the instrument response over time and verify that it has not accuracy and similar photometric response of the instru-
changed. ments. The overall performance of any monochromator
Where and how standardization is done is critical to results from the combination of source, filters, grating,
the performance and transferability of the data from any scanning mechanism, detector and sample interface; there-
monochromator. The number of reference points required fore, all the components should be carefully chosen for the
to fit a grating spectrum is dependent on the precision spectroscopic system to function in its optimum range of
of the scanning mechanism. Typically a standard is used performance for the desired application.
that provides multiple reference points across the spectrum
and the data are interpolated between the points. The
more data points, the better the fit and the better the ABBREVIATIONS AND ACRONYMS
transfer between instruments. Standards mounted inside the
instrument are very common because their presentation and ASTM American Society for Testing and
environment can be controlled, but it must be remembered Materials
that they are in a different optical location than the sample. NIRS Near-infrared Spectroscopy
Large differences can exist between wavelengths measured NIST National Institute of Standards and
from reference materials located inside and outside the Technology
spectrometer because of optical geometry differences. So PTFE Poly(tetrafluoroethylene)
for transferability issues, it is not only necessary to keep SRM Standard Reference Material
the monochromators as identical to each other as possible, USP The United States Pharmacopeia
but also to keep the instrumentsample interface as similar UVVIS UltravioletVisible
as possible.
REFERENCES
4 CONCLUSIONS
1. H.A. Rowland, Philos. Mag., XVI, 197 (1882).
Depending on the objective of the analysis, different instru- 2. M. Seya, Sci. Light (Tokyo), 2, 8 (1952).
mental and design parameters become important. It is 3. T. Namioka, Sci. Light (Tokyo), 3, 15 (1953).
extremely difficult, if not impossible, to design a single 4. T. Namioka, J. Opt. Soc. Am., 49 (1959).
NIR spectrometer optimized for all types of samples and 5. T. Namioka, H. Noda and M. Seya, J. Opt. Soc. Am., 64, 1031
analysis modes. (1974).
Near-infrared Array Spectrometers
Edward W. Stark
KES Analysis, Inc., New York, NY, USA
1 INTRODUCTION optics, analytical spectroscopy has gained from the immense

investment in fiber-optic communications technology.
The recent increased availability of array detectors has led This article primarily addresses photodiode array (PDA)
to the development of spectrometers based on diffraction spectrometers for measurement of the transmission through
grating polychromators, that is, grating spectrometers with and/or the reflection from samples for quantitative or qual-
multiple detectors that measure radiant power in a large itative analysis. The measurement of low radiant power
number of spectral detection bands simultaneously. Devel- emission, such as Raman scattering, is discussed in the
opment of aberration-corrected holographic diffraction grat- article Array Detectors for Raman Spectroscopy.
ings has allowed for the design of small polychromators
with good optical performance.
Array spectrometers have primarily been used in the visi- 2 ADVANTAGES OF ARRAY
ble and adjacent near-infrared region (NIR) of the spectrum SPECTROMETERS
due to the availability of detector arrays of reasonable cost.
The earliest systems used silicon detector arrays covering 2.1 Photometric performance
the visible and very-NIR regions from 400 to 1100 nm
wavelength. 38- and 76-element parallel-channel silicon The PDA combines high responsivity with low noise to
arrays were introduced for color measurement. These were provide highly sensitive measurements of spectral radi-
followed by self-scanned arrays which combined a multi- ant power. The resultant high signal-to-noise ratio (S/N)
plexer with the detector diodes on the same chip. The addi- is particularly important for NIR vibrational spectroscopy
tion of integrating amplifiers to the multiplexer improved utilizing the higher order overtone and combination absorp-
the performance. Germanium (Ge) and, later, indium gal- tions. Cryogenic cooling reduces the noise substantially
lium arsenide (InGaAs) detector arrays extended the range allowing detection of extremely weak signals, for example
of photodiodes to 1700 nm. Extended range InGaAs arrays in Raman spectroscopy and other scientific applications. For
and lead sulfide (PbS) film photoconductor arrays have absorption spectroscopy, however, arrays are more typically
recently become available increasing the range to and used at ambient temperature or cooled with thermoelectric
beyond the upper end of the NIR region, which is defined devices to as much as 70 C below ambient.
to be from 780 to 2500 nm.1 The fact that all the detector channels collect their
The development of InGaAs photodiode technology has energy simultaneously causes the array spectrometer to
been accelerated by the needs of fiber-optic communications, have increased performance due to the multichannel advan-
which is optimized in the regions of 1300 and 1550 nm. tage. The S/N is proportional to the square root of the
InGaAs array technology has also been applied to monitor measurement time. Each element of an n-channel array
power for dense wavelength division multiplexing (DWDM) uses the total measurement time, whereas a single detec-
in fiber-optic communications. As is the case with fiber tor time-shared among the same number of channels has at
best 1/n of the time for each channel. Therefore, an array
John Wiley & Sons Ltd, 2002. may have an increase in S/N proportional to the square
root of the number of channels compared to the equivalent . . .the wavelength interval, l, . . . of radiant energy leav-
time-shared detector. ing the exit slit of a monochromator measured at half the
The photocurrent from the photodiode is directly and lin- peak detected radiant power.3
early proportional to incident radiant power when operated
in the photovoltaic mode with zero bias voltage applied. For an array spectrometer, the detector element serves
Linear output signal response over eight to nine decades as the exit slit. Resolution may be measured by varying
of radiant input power is typical. The dynamic range of the wavelength of an essentially monochromatic source that
the photodetector generally far exceeds the dynamic range is irradiating the entrance slit and determining the output
of the associated electronics. Therefore, gain ranging may signal as a function of source wavelength.
be used to advantage if saturation of the readout amplifier The resolution of a typical array spectrometer is several
is avoided and the range of the analog to digital con- nanometers, although better resolution is possible over a
verter is not exceeded. The array spectrometer does not limited spectral range. In most NIR absorption spectroscopy
suffer from the multiplex disadvantage of Fourier trans- applications, a resolution of a few nanometers is quite
form (FT) instruments as the noise remains concentrated in sufficient. Commonly used flat-field grating polychromators
each channel rather than being spread across the spectrum. suffer from a varying linear dispersion so that the resolution
Also, the full dynamic range of the detector, amplifiers, changes over the wavelength range. To accommodate this,
and analog-to-digital converter (ADC) is available for the the spacing and width of the detector elements are usually
spectral signal in absorption measurements as compared to less than the resolution. Although charge-coupled device
the FT spectrometer, for which the interferogram dynamic (CCD) arrays of at least 2048 elements are available,
range is approximately equal to the spectral dynamic range multiplexed detector arrays commonly have 128, 256, or
times the square root of the number of spectral resolution 512 elements, each element providing a spectral data point
elements.2 within the spectral range. However, when the element
spacing is less than the resolution, the data points do not
represent independent resolution elements. Parallel pinout
2.2 Speed arrays of up to 76 elements are also in use with external
electronics.
The response time constant of a photodiode depends on As there are no moving parts involved in the spectral
three factors, the time constant of the circuit, the carrier scanning, array spectrometers have excellent wavelength
transit time, and the diffusion time of the carriers in detec- stability. The primary source of variation is thermal drift,
tors which are not fully depleted by bias voltage. All of which can be as low as 0.01 nm C1 . Wavelength accu-
these factors are improved by applying a bias voltage, racy depends on the alignment and reproducibility of the
unfortunately, at the cost of increased noise. Response to polychromator grating and the positioning of the array.
1010 Hz is possible when using this photoconductive mode Specifications of a few tenths of a nanometer absolute
with small detectors, low resistance, and sufficient bias volt- accuracy are possible and desirable. Due to the nonlinear
age. However, in spectroscopy sub-microsecond response relationship between data points and wavelength typical
times are generally not required and the slower photo- of array spectrometers, wavelength calibration and inter-
voltaic mode of operation is favored to reduce the noise. polation is commonly employed. However, for transfer of
Even in this mode, the detector response time constant does calibrations from one instrument to another, it is impor-
not usually limit the spectral data rate. The readout may tant that the native instrument wavelength scales be closely
reach a scan rate of 107 data points per second. This trans- matched.
lates to several thousand spectra per second (SPS) although
typical spectrometers output less than 100 SPS. No mechan-
ical scanning, which would limit the spectral data rate, is
2.4 Environmental ruggedness
required in a diode array spectrometer.
The array spectrometers simple combination of an entrance
2.3 Wavelength resolution, stability and slit, a fixed diffraction grating, and an array detector with its
accuracy readout electronics results in a rugged system with no crit-
ical moving parts. It is particularly useful in environments
Resolution, formerly known as spectral bandwidth or band- with high vibrations that would disable an interferome-
pass, of a dispersive spectrometer, in molecular spec- ter or scanning system. The array spectrometer can be
troscopy, is defined by the American Society for Testing more readily sealed and temperature controlled due to its
and Materials (ASTM) as small size.
Near-infrared Array Spectrometers 3
2.5 Cost 3.1 Spectral signal
The optical and mechanical simplicity of the array spec- Before considering the individual subsystems, the overall
trometer generally results in lower cost units than scanning relationships among them should be defined. In princi-
monochromator or FT instruments. Very low cost units ple, each parameter is a function of wavelength, l. The
have been produced that compete favorably with filter pho- spectrometer system begins with a source which emits
tometers. This cost advantage is particularly evident in the energy with a spectral radiance, LS l, measured in watts
visible and very-near-infrared where low cost silicon arrays per square centimeter-steradian-micrometer. The spectral
can be used. It is expected that InGaAs arrays will expe- radiance LlTl after the sample is LS leS lTl,
rience significant cost reduction as production techniques where eS l is the efficiency of the radiative transfer
improve and quantities increase. from the source through the sample including geometric
factors as well as other losses and Tl is the transmit-
tance/reflectance of the sample.
In the case of a scattering sample, the additional geomet-
3 THE ARRAY SPECTROMETER
ric losses caused by the samples redirection of radiation
SYSTEM in many directions may be considered as part of either
eS l or Tl. In diffuse reflection measurements, Tl
A spectrometer used for transmission or reflection spec- of a sample is usually established with respect to a unity
troscopy must be considered as a system composed of a reflectance diffuse reference sample so the geometric fac-
radiation source, the sample optics which transfer the source tors are largely incorporated in the efficiency eS l. The
energy to the sample and from the sample to the sensor, the practice varies among diffuse transmission practitioners.
wavelength selection optics, the detection device, and the In any case, one can define the spectral radiance at the
processing of the spectral signals. The array spectrometer entrance to the spectrometer as Ll Tl watts per square
systems considered here are those using an array of detec- centimeter-steradian-micrometer.
tors to measure radiant power at a number of wavelengths The optical geometry of the grating spectrometer defines
simultaneously. Two-dimensional arrays can add a spatial an effective area, Al (cm2 ) and an associated solid angle,
dimension to the spectral information and are also used
l (steradians), at each focus within the optical system.
together with spectral multiplexing, such as FT techniques, For maximum signal, the external source and sample optics
to provide multispectral imaging. Based on imaging usage, should be configured so that they do not limit these values.
the element of an array is often called a pixel, short for Optical absorption and reflection losses within the instru-
picture element. ment contribute an overall efficiency factor, el. Finally,
It is convenient, although not totally rigorous, to sep- the spectral responsivity of the detector, RD l (amperes per
arate the performance of the spectrometer system into watt or volts per watt), defines the conversion of the optical
photometric performance, which is related to its ability to power to a usable electrical signal. The spectral signal at
measure radiant power, and spectral performance, which is pixel p of the array may then be expressed as
related to the wavelength dimension of the spectral out-
put. The photometric performance of the overall system Sp l D LlTlAl
lelRD l dl 1
may be characterized by the S/N, photometric stability
(drift), spectral data rate, dynamic range, and photomet- The integration extends over the entire wavelength region
ric linearity. These parameters are primarily dependent on for which the spectral response is not zero.
the strength and stability of the source, the throughput In an array spectrometer, each detector element primarily
and efficiency of the polychromator optics, the detectiv- responds to a narrow band of wavelengths. Thus, to avoid
ity of the detector array, and the capabilities of the readout integration over wavelength, the values of Ll, Al,
l,
electronics. el, and RD l are often assumed to be constant for pixel
The spectral performance may be described in terms p at an effective wavelength of lp . Ignoring stray radiant
of spectral resolution, spectral range, wavelength accu- energy, the spectral response is assumed to be zero outside
racy and stability, and stray light. These parameters are the geometric instrument line shape (ILS) for pixel p. The
primarily controlled by the polychromator and the geom- integrated spectral response of pixel p is then assumed to
etry of the detector array. The three major subsystems be the spectral response at lp multiplied by the resolution,
of an array spectrometer, the source, the polychroma- lp , for that pixel, thus
tor, and the detector array and electronics, are discussed
below. Sp D Llp Alp
lp Tlp elp RD lp lp 2
Because the sample characteristics are application depen- Both conditions compromise the potential performance of
dent, the performance of the instrument itself is usually the array spectrometer.
described for a lossless sample, that is, for the case of
a sample having unity transmittance or diffuse reflectance
Tl D 1 such as might be used to determine a baseline 3.3 Noise
spectrum.
The sensitivity of the spectrometer is ultimately limited
by the noise. Noise in absorbance units is proportional to
3.2 Optical throughput the S/N of the transmittance or reflectance spectrum. Two
different types of photometric noise typically exist in a
Optical throughput, E, sometimes called the geometric spectrometer additive and multiplicative. Additive noise,
etendue of the system, is a measure of the ability of an as the name implies, includes various types of noise that
optical system to collect and transfer radiant energy. It is are added to the signal. With the exception of signal current
defined as the vector product of the area A and solid angle
shot noise, described below, the additive noise components
of a beam at its focus and the square of the refractive index are not a function of the signal strength or of wavelength.
of the medium in which the beam is focused.4 The optical The main sources of random additive noise are the detector
throughput is limited to its least value at any point in the and its associated electronics. External electrical sources
optical system. The relative magnitudes of A and
will be such as magnetic fields, portable telephones, and radio
different at different points within the system and their ratio stations can contribute significant random and coherent
may be modified by magnification due to a lens or mirror, noise unless the spectrometer is well shielded. Optical
but, once established, the throughput cannot be increased. sources, such as room lights, can also generate additive
Therefore, throughput matching throughout the system is coherent noise in the signals. Often, coherent noise may be
important to conserving the available radiant power. rejected by suitable signal processing.
Removing the terms in equation (1) that depend on Multiplicative noise, on the other hand, arises from mod-
the source radiance, sample transmittance/reflectance, opti- ulation of the signal energy from causes such as source
cal efficiency and detector responsivity leaves the optical instability, sample motion, vibration, and atmospheric tur-
throughput of the system, El, given by bulence within the instrument and it is, therefore, a function
of both the signal amplitude and the wavelength. The major
El D Al
l cm2 sr 3 multiplicative noise sources for array spectrometers are
typically within the source and/or sample optics. Both tung-
For example, in a spectrometer the throughput may be stenhalogen lamps and fiber optics may contribute signifi-
established by the area of the entrance slit, and the solid cant sensitivity to vibrations. Atmospheric water absorption
angle subtended by the grating as viewed from the slit. For in the optical path is often a significant variable.
a circular concave grating with a numerical aperture NAG
that is fully illuminated from a slit with area ASLIT at an
angle a from the grating axis, the throughput at the slit, 3.4 Resolution and spectral response
ESLIT , is
ESLIT D pASLIT NAG 2 cos a 4 The ILS, the shape of the spectral response of the polychro-
mator optics for each array element, is controlled by the
If a fiber optic is used as the entrance slit, ASLIT in width of the detector element or of the magnified entrance
equation (4) is the area of the fiber core. If the numerical slit image on the detector, the linear dispersion at the pixel
aperture of the fiber is less than NAG , the grating will not location, and the image quality of the optics, including
be fully irradiated and the lesser numerical aperture of the aberrations and defocusing due to field curvature. Most
fiber becomes the controlling factor. When throughput of vibrational spectroscopy applications of array spectrome-
an array spectrometer is established at the entrance slit, it ters do not require operation at the limit of the optical
is not a function of wavelength. resolution so wider slits and pixels are used to improve the
In many diode array spectrometers, the area of the photometric performance. Therefore, the influence of the
detector pixel is smaller than the entrance slit image so slit and pixel widths dominate the spectral response of each
the maximum throughput is defined by the projected area pixel and the ILS is triangular or trapezoidal. The full width
of the detector pixel and the solid angle subtended by the at half-height (FWHH) measure of resolution or bandpass
grating when viewed from the detector element. In this case, is appropriate for this line shape. However, the spectral
effective throughput is lower than the value at the entrance shape of Ll and RD l can significantly modify the over-
slit and it does vary with wavelength due to the geometry. all spectrometer response, thereby producing asymmetric
line shapes for the pixel signals, Sl. This effect causes a holographic pattern that exposes photosensitive material
a shift in the effective pixel wavelength lp that should be (photoresist) on the surface of the blank. The depth of the
considered in calibration of the wavelength scale. grooves is optimized based on the wavelength range to be
used. The greater the number of grooves the greater the
theoretical resolution of the grating. The closer the grooves
4 THE DIFFRACTION GRATING are spaced, the greater the dispersion of the grating, that is,
POLYCHROMATOR the larger the angle of diffraction for any given wavelength.
Gratings may be formed on a transparent blank, creating a
The basic optical system used in array spectrometers is transmission grating, or they may be coated after formation
the diffraction grating polychromator. If a plane diffrac- with aluminum or gold to produce a reflection grating.
tion grating is used, a lens or mirror collimates the energy The majority of gratings are replicas of a master grating.
from the entrance slit onto the grating which disperses the Although replica gratings may have slightly higher surface
energy as a function of wavelength. A second lens or mir- scattering than original masters, they have the advantage
ror focuses the collimated dispersed energy on the detector of providing essentially identical optical parameters for
array. An order sorting filter is needed to block unwanted many units.
energy from reaching the detector array. Using a reflection Figure 1 shows the geometry of a plane reflection grating
plane holographic grating, the crossed CzernyTurner con- polychromator. a is the angle of incidence of the collimated
figuration shown in Figure 1 can produce a flattened spec- energy, measured from the normal to the grating surface,
tral field and good coma aberration correction over a broad while b is the angle of diffraction of the energy for a specific
wavelength range. Rotation of the plane grating can be used wavelength. For a concave grating that focuses the energy,
to adjust the wavelength region imaged on the array. the angles are measured between the principal ray and the
A simpler optical configuration consists of an entrance axis of the grating. In either case, the fundamental grating
slit, a concave diffraction grating to disperse the light as equation may be written
a function of wavelength and to create the spectral image,
ml
an order sorting filter, and an array of detector elements D sin a C sin b 5
(see Section 4.7). Other geometries, including transmission s
gratings, are also possible, although less often used for array where m D 0, 1, 2, 3, . . . and the grating constant s is
spectrometers. the groove spacing in the same units as l.
Gratings are often specified in terms of groove density
(grooves per millimeter). The groove spacing is (in nano-
4.1 Diffraction grating principles meters) then found as the reciprocal of the groove density
multiplied by 106 . a and b have the same sign if they are
A diffraction grating consists of a large number of closely on the same side of the optical axis and opposite signs
spaced slits or diffracting lines. These lines are formed by if they are on different sides. By convention, angles are
generating grooves on the surface of a grating blank by considered positive if they are directed counterclockwise
ruling or, more often now, by using a laser to generate and the positive orders lie on the same side of the optical
axis as the slit. The opposite convention appears in some
r of the literature.

Mirror 1 Figure 2 plots the relationship between a and b with l/s
Detector as a parameter. For a particular grating the groove spacing s
array is a constant. Therefore, each contour represents a constant
value of l. The dashed diagonal line corresponds to l D 0.
It is also the locus of all wavelengths in diffraction order
Mirror 2 m D 0, which describes regular or mirror-like reflection.
r Order sorting
Values for m > 0 lie to the right and values for m < 0
filter
lie to the left of the m D 0 locus. Given any two values,
the third value may be determined. For example, the angle
Plane holographic of incidence a of a spectrometer might be 20 . For a
grating
selected wavelength that is 0.3 times the groove spacing,
the intersection of the a D 20 ordinate with the l/s D
Entrance slit
0.3 contour (marked with a circle in Figure 2) occurs at
Figure 1. Crossed CzernyTurner polychromator. b D 40 .
90
0.2
0.4
1.4
1.3
1.8
1.2
m=1
0.8
0.5
1.7
0.6
0.3
80
1.6
0.7
0.1
0
1.1
m = 1
1
70
0.9
1.9
1.5
0.1
Angle of incidence
60
1.8
50 1.7
0.2
1.
0.4
1.3
4
0.3
1.6
0.8
1.2
0.6
0.3
0.5
0.
40
1
0.
0.7
1.
0.
1
2 1.5
1
0.4
0.
0.5
30
9
0.6 1.4
1.3
20
0.7 1.2
0.
0.
3
0.
0. 1.1
4
10 0. 8
0.
0.8 5
0.
6
0.9 0. 1
0.
0.
5
0.
1 7 0.9
3
0.
0.
1
4
0
0
90 80 70 60 50 40 30 20 10 0 10 20 30 40 50 60 70 80 90
Angle of diffraction
Figure 2. Diffraction angles with l/s parameter.
Differentiating the grating equation with a constant 4.2 Resolution and polychromator spectral
yields the angular dispersion of the grating, db/dl. throughput
db
D ms cos b1 6 The classical definition of ILS as the shape of an infinitely
dl sharp isolated line source measured by the instrument is
For b near zero, i.e. close to the normal to the grating, not particularly applicable to array spectrometers due to
where cos b is slowly varying, the angular dispersion is the few data points within the resolution. When the resolu-
almost uniform. By the selection of a, any desired range of tion is defined by matched slit and detector widths or the
wavelengths can be placed near the normal. The angular detector width, only one or two detector elements will typi-
dispersion of the grating increases with the order so a cally respond to an isolated line source, that response being
grating operated in the second order will have twice the due to the spectral overlap discussed below and illustrated
dispersion as one used in the first order. In the case of in Figure 3(a) and (b) rather than the ILS. When there are
a plane grating, the diffracted energy is focused on the several pixels within the resolution, as in Figure 3(c), a
detector array by additional optics. Either lenses or mirrors crude approximation to the ILS may be obtained by inter-
may be used for this purpose. polation. An equivalent, and more measurable, definition
The linear dispersion, dx/dl, of the plane grating spec- of the ILS is the shape of the spectral output signal for
trometer is controlled by the focal distance, r 0 , of the exit each pixel as a sharp isolated line is scanned in wavelength
optics (see Figure 1). This is usually expressed as the recip- across the region of spectral response.
rocal linear dispersion The theoretical resolving power of a grating Rr is given
by the formula
dl l
D smr 0 1 cos b cos j 7 Rr D D Nm 8
dx l
where j is the angle of incidence of the principal ray at where N is the number of grooves which equals the grating
the detector. As can be seen later, the focal parameters of aperture divided by the grating constant s. For example,
a concave holographic grating may be modified to produce a 50 mm diameter grating with 600 lines per millimeter
a flat focal plane. In this case, the focal distance, r 0 , varies will have a resolving power of 30 000. At 1500 nm, this
with b, causing additional variation in the linear dispersion yields a theoretical resolving power of 0.05 nm. However,
across the spectrum. in vibrational spectroscopy, particularly in the visible and
Data point spectral positions

1 2 3 4 1 2 3 4
1.0
Relative throughput 1234
E ()/Em() 0.5
r
4
mbe
3 nu
el
2 Pix
1
Wavelength
= m = 2m = m
(a) (b) (c)
Pixel width = slit image Pixel width = 2 slit image Pixel width = 0.5 slit image
Figure 3. Spectral resolution and relative throughput.
NIR range where array spectrometers are commonly used, half the pixel is irradiated, the throughput area Al is half
the slit and detector element widths are usually chosen to the pixel area. Continuing to change the wavelength, the slit
provide a spectral resolution substantially larger than the image moves so that it is completely on the pixel. Al is
theoretical resolving power of the grating. Therefore, the now the full pixel area, and the maximum throughput area
Rayleigh criterion for resolution, which was created for Alp is realized. Continuing to change the wavelength, the
diffraction limited grating spectrometers with a sin x/x2 entrance slit image immediately starts to move off the pixel,
ILS, is not appropriate for array spectrometers. reaching the point of half overlap [Al D 0.5 Alp ], and
In small array spectrometers, the entrance slit width is finally, that of no overlap [Al D 0]. From equation (1),
typically fixed. In some cases, the entrance slit is simply the spectral response, S(p), for that pixel is proportional to
the end of an optical fiber. The entrance slit is imaged on the Al. The distance the slit image moves from the first to
array detector so the array elements now serve as the exit the second half overlap point is equal to the pixel width wp .
slits. Many array spectrometers are designed with relatively The spectral resolution for that pixel, lp in equation (2)
large spectral bandpass compared to the optical resolution is, therefore, given by the product of the linear dispersion
limit set by aberrations. In that case, the ILS is reasonably (nm mm1 ) in the spectral image plane and the pixel width.
approximated by the convolution of the entrance and exit For the matched optical throughput case of Figure 3(a), this
slit functions. is defined as lm . Each pixel produces one data point in
Imagine the following sequence. For monochromatic the spectral output so, for the usual case where the pixel
radiation filling an entrance slit of width wSLIT , an image pitch (the center-to-center spacing) is equal to the effective
of the entrance slit is formed at the plane of the detector pixel width, the data point spectral positions are spaced by
with a width wI given by lm , the resolution with equal pixel and slit image widths.
Figure 3(a) shows the overlapping sequence of spectral
wI D wSLIT r 0 r 1 cos f1 9
throughput, El/Em lp for several pixels where Em lp
The entrance slit image moves across the detector array as is the maximum throughput for the case of matched widths.
the monochromatic wavelength is scanned. In Figure 3(b) the pixel width is twice the width of the
For the case in Figure 3(a), assume that the width of entrance slit image. Now the entrance slit image controls
the entrance slit image, wI , is equal to the pixel width, wp . the area of the overlap so the greater pixel area does not
Initially the wavelength is adjusted so that the edge of the increase the maximum throughput area, Alp . Therefore,
entrance slit image is adjacent to the edge of a detector the maximum throughput is the same as for the matched
pixel but there is no overlap. For that wavelength and that width case in Figure 3(a). In the case of Figure 3(b), the
pixel, the spectrometer throughput area Al is zero and slit image is totally within the wider pixel for a wavelength
there is no response. As the wavelength is changed, the range equal to lm and then falls off as in the matched
entrance slit image moves onto the detector pixel. When case. As a result, the spectral resolution, l, is twice lm .
The data point spectral position spacing is also doubled due the spectrometer may be an additional significant source
to the increased pixel pitch forced by the wider pixels. The of stray radiant energy. Undispersed radiation from the
integrated radiant power on each pixel is doubled due to the entrance beam and from the zero order covers a broad
wider spectral range that is utilized. However, the radiant wavelength range as does the dispersed energy from unused
power on each pixel could be increased still further without diffraction orders. Reflection of energy from the order
additional loss of spectral resolution by increasing the sorting filter, detector window, and array surface is an
entrance slit width to obtain the matched optical throughput additional potential source of stray radiant energy. This may
condition. be particularly troublesome if it strikes the grating and is
In the case of Figure 3(c), the pixel width is one-half the redirected back to the array at a different location.
entrance slit image width. The throughput is now limited by The spectral sensitivity of a pixel to stray radiant energy
the pixel area to one-half of its matched value. However, the is not limited to the resolution of the spectrometer, lp , but
wavelength must still vary by lm to move the slit image it extends over the entire spectral response of the combined
between the one-half overlap positions so the resolution is detector and order sorting filter. Therefore, the effective
the same as the throughput matched case in Figure 3(a). stray radiant energy for a pixel is the integral over all
The integrated radiant power is, therefore, one-half that of wavelengths of the spectral stray radiant energy times the
the case in Figure 3(a). As can be seen on the projection spectral response of the detector and order sorting filter
of the relative spectral throughput, the fractional overlap combination. The effective spectral range for stray radiation
between pixels is also increased substantially, producing is therefore many times the resolution of the spectrometer,
more correlation between the response of adjacent pixels. placing stringent requirements on the blocking of stray radi-
There is sometimes confusion between resolution and data ant energy over a wide spectral range. A particularly diffi-
point spacing for array spectrometers. When there are sev- cult situation arises when there is a narrow region of high
eral pixels sharing the entrance slit image as in Figure 3(c), absorbance in an otherwise low absorbance spectrum. In
the data point spacing is reduced (i.e. there is more than this case, the integrated stray signal may be relatively large
one spectral data point within the spectral resolution), but and the spectral signal for the absorption region quite small.
the resolution is not improved. In this case, if there are Defining the signal due to the integrated spectral stray
sufficient data points within the resolution, the ILS may radiant energy on pixel p as S(p) the combined signal
be approximated by interpolation. However, the throughput S0 (p) is
for each pixel, and therefore, the S/N is reduced. S0 p D Sp C Sp 10
The measured absorbance is log[S0 p/Rp], where R(p)

4.3 Stray radiant energy, absorbance linearity is the signal for the reference, e.g. for Tl D 1 in equa-
and dynamic range tion (1). If S(p) equals zero, the measured absorbance
will equal the actual absorbance log[Sp/Rp]. As S(p)
approaches zero, the measured absorbance is limited to
Stray radiant energy, often called stray light, is defined
log[Sp/Rp]. The effect of low levels of stray light
by the ASTM as
on the linearity and dynamic range of the spectrometer
all radiant energy that reaches the detector at wavelengths measurement is surprisingly large, as shown in Figure 4.
that do not correspond to the spectral position under Significant absorbance nonlinearity occurs more than one
consideration.1 absorbance unit below the absorbance value produced by
the stray light alone.
In addition, external energy such as room light that is A typical measurement technique for stray radiant energy
within the spectral response of a pixel but is not a function utilizes a filter to block radiation in one spectral region
of the sample transmittance/reflectance Tl may reach the while allowing it to be detected over the remaining wave-
detector with similar effects. Stray radiant energy arises lengths. For example, a long-pass filter such as Schott
within the spectrometer from a number of sources, not all RG850, which has a sharp cut-on at 850 nm could be
of which are included in typical stray light specifications. introduced in the optical path, blocking short wavelength
This undesired energy adversely affects the linearity and radiation. Such a filter typically has an internal transmit-
dynamic range of the desired energy measurement. tance of <105 (actual absorbance >5) below 770 nm and
A common definition of stray light relates only to >0.95 above 900 nm.5 The measured absorbance in the
scattering from the grating and other optical surfaces within blocking region will be approximately log[Sp/Rp].
the optical path. However, each pixel typically accepts However, this value only considers stray energy transmit-
radiation over most of a hemisphere. Reflection of radiation ted by the filter, roughly the energy above 850 nm. For
onto the array from walls and other structures within a silicon detector system with a long wavelength cut-off
5.0
%
00
4.5 0.0
%
01
0.0
4.0
0.010%
3.5
Measured absorbance
0.100%
3.0
2.5
S (p)/R(p) = 1.000%
2.0
1.5
1.0
0.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Actual absorbance
Figure 4. Nonlinearity due to stray radiant energy.
of 1100 nm and a tungsten source, over 40% of the avail- 4.4 Multiple diffraction orders
able energy is at wavelengths below 850 nm. In addition,
spectral stray radiant energy is often higher at wavelengths The presence of the diffraction order, m, in the grating
close to the measurement wavelength due to scattering and equation (equation 5) indicates that several wavelengths,
multiple reflections between the detector and order sorting l, may be present at each angle of diffraction, b. m D 0
filter but such nearby stray energy is not included in the corresponds to regular reflection, which causes an unwanted
stray measurement using a long-pass filter. loss of energy and represents a potential source of stray
As is discussed next, shorter wavelength second- and light. The order m D 1 is usually used in vibrational
third-order energy may be a significant source of stray radi- spectroscopy, sometimes combined with m D 2 to obtain
ant energy, as it is defined above. In this case, significant an added shorter wavelength range. This is particularly true
energy is concentrated at a few specific wavelengths that for systems using array detectors because it is desirable to
are shorter than the measurement wavelength. Measurement have the array on the opposite side of the grating axis from
of this type of stray radiant intensity requires blocking the the entrance slit to avoid mechanical interference.
energy at the measurement wavelength to a level an order Figure 5 illustrates the spectral effects of multiple orders
of magnitude below the required stray energy level while and the concept of free spectral range. An idealized spec-
passing the shorter wavelengths. Adequate short-pass glass trometer is assumed with a detector array extending from
absorption filters are available only for certain regions, e.g. zero wavelength to 2.6 m in the first order containing 261
the Schott BG 39 filter for measurement wavelengths above pixels numbered from 0 to 260. Each pixel is exposed to
800 nm in silicon detector spectrometers. radiant power from several orders of diffraction, m. The
So, at best, the broadband filter stray energy mea- wavelength scale of the spectrometer for each order is also
surement technique only provides an order of magnitude shown in Figure 5. The increase in linear dispersion with
result but it does have the advantage of simplicity and order number causes a reduction in spectral bandwidth such
a rapid measurement. However, a complete description that the resolution for order m, lm, is equal to the reso-
of the response of an array spectrometer requires mea- lution for the first order, l1, divided by m. For example,
suring the response of each of the pixels as a function pixel 130 receives first order radiant energy in a spectral
of wavelength over the entire wavelength range of pos- bandwidth of 10 nm centered at 1.30 m. It also receives
sible sensitivity. This time consuming operation is justi- second order radiant energy in a spectral bandwidth of 5 nm
fied when the performance is critical, particularly when centered at 0.65 m and third order radiant energy in a spec-
attempting to match instrument characteristics for calibra- tral bandwidth of 3.3 nm centered at 0.325 m. Although
tion transfer. higher order energy may also be present, it is still lower
Wavelength (m)
0 0.5 1 1.5 2 2.5 1st order
0 0.25 0.5 0.75 1 1.25 2nd order
0 0.167 0.333 0.5 0.667 0.833 3rd order
120
Si (c = 1.1) Ext. InGaAs (c = 2.6)
1.10
1.30
100
0.85
Si (c < 0.65)
1.70
80
InGaAs (c = 1.7)
Relative signal
60
1.30
0.85
2nd order
2.60
40
0.55
1st order
0.65
3rd order
0.87
20
0.57
0.36 0.55
0.43
0.43
0.325
0
0 50 100 150 200 250
Spectral free range Pixel number (10 nm )
Figure 5. Spectral effects of multiple orders.
in intensity and shorter in wavelength than the first three and to 0.57 m third order radiation which would provide
orders. almost 15% of the first order signal. A long-pass order
The relative spectral signal produced by such a spectrom- sorting filter is required to reject the 0.85 m and shorter
eter, using a tungsten lamp source operated at 2800 K and wavelength energy. If a single filter is used for the entire
a unity quantum efficiency detector, is plotted in Figure 5 detector array, either at the array or elsewhere in the optical
for the first order, second order and third order. For pur- path, 0.85 m energy will not be available for detection
poses of illustration only, the grating efficiency variation at any pixel location. In practice, allowance must also be
with wavelength and order is not considered. made for the transition from blocking to transmitting of the
It is important to allow energy of only a single order filter.
to be detected by each pixel of the array. This is often In general, then, the shortest wavelength, l1 , usable in
accomplished by use of order sorting filters combined the first order is somewhat more than half the longest first
with the spectral response of the detectors. For first order order wavelength, l2 , being used because the second order
operation, higher-order shorter wavelengths are imaged on wavelength l2 /2 must be rejected and with it, the short
each pixel together with the first order but there is no wavelength energy in the first order. Extending this to
imaged energy from lower orders. In this case, a simple higher orders, free spectral range is defined as
long-pass absorbing glass filter often suffices to reject the
unwanted orders. l2 l1 D l2 jmj C 11 11
For example, in Figure 5, pixel 170 of our hypothetical
InGaAs detector detects first order energy at 1.70 m, the The first order free spectral range in this case is 0.85 m.
upper limit of the detector response. This pixel also is Assuming each type of detector is used to its long wave-
exposed to 0.85 m second order radiation, which would length limit, the free spectral ranges of Si, InGaAs and
generate somewhat more than half the first order signal, extended InGaAs (2.6 m) detectors are shown in Figure 5.
Each pixel of the array spectrometer is exposed to a 4.6 Woods anomalies

specific set of wavelengths of different orders. For example,
one pixel of a silicon detector array covering 0.4 to 1.0 m When the diffraction angle is large enough for a given order,
might be used to detect at 0.65 m in the first order. jbj for a higher order approaches 90 . Beyond this point, the
The second order energy at 0.325 m must be blocked energy in the higher order does not exist and more energy is
but the blocking at 0.50 m required for the pixel at the concentrated into the lower orders. These phenomena were
1.0 m spectral position is no longer required because that discovered by R.W. Wood in 1902 and the rapid variations
wavelength is not focused on the pixel at 0.65 m. By use in a lower order efficiency when jbj of a higher diffraction
of a segmented or linear variable filter, an array may be order is near 90 are called Woods anomalies. The second
used beyond its natural free spectral range as defined by order Woods anomalies in the first order spectrum occur
the detector cut-off wavelength. in the regions where the second order b2l is 90 . The
Second order operation requires rejection of both the efficiency in this region drops and the polarization increases
first order and the third order interference. The shorter so good spectrometer design avoids using it. The positions
wavelength third order energy may be rejected by a long- of Woods anomalies are a function of the grating groove
pass filter but first order rejection also requires a long period ratio, l/s, the angle of incidence, a, and diffraction
wavelength cut-off. This may be provided by the detector order. Solving the grating 1 for l/s with sin b D 1 yields
or by short-pass filters. Since the second order energy and l sin a 1
detector responsivity is much less than those of the first D for m < 0
s m
order, excellent long-wavelength blocking is required to 12
l sin a C 1
avoid excessive stray light in the shorter wavelength region D for m > 0
of the second order. Sometimes, the second order filter must s m
be divided into several regions due to the limitations of The spectral positions of Woods anomalies due to the
second order free spectral range, which equals the longest second and third orders are shown in Figure 6.
wavelength l2 /3 for each filter segment. When a D 0, the wavelength of the anomaly is the same
in the positive and negative orders. As a increases the
Woods anomaly splits and appears at two wavelengths
4.5 Grating efficiency because bl differs for the positive and negative orders.
For plane gratings, the second-order Woods anomaly can
The efficiency of the grating is a complex function of produce narrow but quite deep loss of energy, particularly
wavelength and polarization of the incident energy which for the polarization when the electric vector is perpendicular
depends on the groove density, the shape of the grooves, to the grating grooves. Concave gratings, on the other
the angle of incidence, and the grating coating reflectance hand, have a wider but shallower feature in this region
or grating blank transmission. The grating efficiency at any due to the curvature of the grating producing a range of
particular wavelength is reduced by the separation of the diffraction angles across the grating for each wavelength.
energy into several positive and negative orders. Standard In either case, Woods anomaly may produce substantial
holographic gratings have a sinusoidal groove profile which artifacts in the spectrum. Therefore, for any given first
provides a generally smooth and flat efficiency distribution. order wavelength range and a moderate angle of incidence,
there is a range of groove frequencies 1/s that should be
The groove depth is optimized for the wavelength region
of interest. Blazed gratings alter the groove shape so as to 1.0
Grating groove period ratio /
diffract more energy in a desired direction and, therefore, 0.9 1

they can increase the efficiency over a narrow wavelength m=
0.8 ly for
oma
region in either the positive or the negative order. Holo- 0.7 r an
o rde
ond =1
graphic gratings can be blazed by chemical and ion-etching 0.6 Sec aly f
or m
om
techniques. However, in many array spectrometer applica- 0.5 de r an
d or
Thir
tions, a broad wavelength region is desired and sinusoidal 0.4 Sec
ond
orde
groove holographic gratings are preferred. 0.3
Third
r an
oma
0.2 orde ly fo
Array spectrometers usually operate with low dispersion r ano rm
= 1
maly
in order to image the spectrum on a small array detector. 0.1 f or m
= 1
The maximum efficiency for such low dispersion sinusoidal
0 10 20 30 40 50 60 70 80 90
groove gratings is 34% in the first order. The efficiency in Angle of incidence
the second order is typically somewhat less than half this
value. Figure 6. Woods anomaly positions.
avoided to prevent a second order Woods anomaly from Array detector

A
appearing in the spectral range. B
Slit
1 sin a 1 1 sin a

Order sorting
< < m < 1 13 filter
2lmax s 2lmin

When present, the third-order anomaly is substantially
weaker than the second order and it often may be neglected.
The third order cannot exist for b3l 90 . The lack of the r r
third order may simplify the order sorting filter require-
ments by eliminating the third order for first order wave- P
lengths shorter than Concave holographic
1 sin a grating
lb3l D 90 D s m < 1 (a)
3
14
1 C sin a rD
lb3l D 90 D s m > 1
3 Grating
blank
Pin hole
4.7 The concave grating
One of the more important events in the history of spec-

troscopy occurred in 1882 when Henry Rowland invented Pin hole
rC Y
the concave diffraction grating. He showed that a spheri-
cal concave mirror, ruled with parallel lines equally spaced
along the chord of its arc, will produce a sharply focused
spectrum on a circle whose diameter is equal to the radius
of curvature of the grating surface. Fermats Principle states
that the distance from the object A to any point P on the (b) Laser 0
surface plus the distance from P to image point B must be
Figure 7. Concave holographic grating geometry: (a) usage;
a constant (Figure 7), (b) recording.
AP C PB D 2b D constant 15 The simplified geometry for usage and recording of
Using Fermats Principle, Rowland stated the concave holographic grating are shown in Figure 7(a)
and (b).
Should we increase b by equal quantities and draw the ellip- Using the nomenclature of Noda et al., and considering
soids or hyperboloids so indicated. . . The intersection of only the in-plane slit and image points z D 0, the follow-
these surfaces with any other surface form what are known ing simplified equations describe the first terms of the series
as Huyghens zones. By actually drawing these zones on expansion of the light-path function.
the surface we form a grating which will reflect or refract
the light of a certain wavelength to the given focal point.6 F D AP C PB C nml 16
This results in an aberration free image at a given wave- where n is the number of the groove and groove 0 passes
through the origin.
length. Such a hyperbolic fringe pattern is generated by the
The grating equations from F100 are:
interference pattern of two coherent point sources. A cen-
tury after Rowlands discovery, the combination of laser ml
sin a C sin b D
and photoresist technologies has allowed us to form the s
17
hyperbolic lines that he proposed. l0
sin sin g D
The generalized theory of the grating needed to describe s
the properties of various forms of gratings was developed The horizontal focal curve from F200 D 0 is
by Noda et al. in 1974.7 In the same issue of the Journal
cos2 a cos2 b cos a C cos b ml
of the Optical Society of America, they also published a ray C C
tracing procedure for holographic gratings.8 These papers r r0 R l0
2 2
provide the basis for the design of concave holographic cos g cos cos g cos
D0 18
grating polychromators. rC rD R
The astigmatic focus from F020 D 0 is film detectors, such as PbS, by deposition on a flat substrate.
Therefore, the polychromator must produce a reasonably
1 1 cos a C cos b
C 0 flat image plane at the detector surface to preserve the focus
r r R
and spectral resolution. Often, this is done by adjusting the
ml 1 1 cos g cos parameters of the horizontal focal curve.
C D0 19
l0 rC rD R Figure 8 illustrates a variety of positions and shapes of
The coma from F300 D 0 is the diffracted spectra obtainable by changing the coefficient
2 2 2 2
cos a cos a sin a cos b cos b sin b ml cos g cos g sin g cos cos sin
C C D0 20
r R r r0 R r0 l0 rC R rC rD R rD
The nonzero terms in g and resulting from holographic Cf . For this example, the spectral range is 950 to 1700 nm,
recording provide the flexibility for modification of the a is chosen to be 10 , the groove density is 250 grooves per
focal curve, astigmatism, and coma from the case of millimeter (s D 4 m) and the grating radius of curvature is
constant spacing and straight grooves where these terms 100 mm. The entrance slit is placed on the Rowland circle
are all zero. The coefficients of ml/l0 in equations (18), so Cf equal to zero produces the classic Rowland circle
(19), and (20) are often denoted as Cf , Ca , and Cc as they image. For this example, choosing Cf equal to 0.081 in the
are fixed by the recording geometry. 1 order optimized the flatness of the spectral (tangential)
Unlike the separate exit slits and detectors of the classical focus and its location relative to the astigmatic (sagittal)
recording spectrograph, which were arranged on a curve to focus. The astigmatic focal surface position is adjusted by
match the focal plane of the spectrograph, modern array Ca to coincide with the spectral focus and coma is optimized
detectors are fabricated as a single flat device, typically on with Cc . Rays are shown between the entrance slit, the
a semiconductor wafer or, in the case of photoconductive grating, and the optimized spectral focal point for 1700 nm.
Cf = 0.149
1700 nm
Cf = 0.125
1450 nm
1200 nm
Cf = 0.100
950 nm 1700 nm
1450 nm
1200 nm
Cf = 0.075 950 nm Cf = 0.100
Detector m < 0, < 0
m > 0, > 0 array
Cf = 0.050 Cf = 0.081 Astigmatic focus
Entrance slit
Cf = 0
Cf = 0.050
Cf = 0.025
Cf = 0.075 Cf = 0
Cf = 0.075
Rowland circle
Concave holographic grating
Figure 8. Concave holographic grating spectral foci.

Although the C1 order with Cf equal to 0.149 produces elements being unequally spaced in wavelength. This
a very flat image surface, there is interference between the is generally corrected by interpolation of the data. The
entrance slit position and the spectral image. The extreme variation in reciprocal linear dispersion computed from
angle of incidence on the detector array also adversely equation (7) with uniform detector pixel width results
impacts the design of order sorting filters. in variation of the spectral resolution as a function of
Even for Cf equal to 0.081 there is a significant angle of wavelength. The spectral resolution is also affected by the
incidence at the detector, and the focal distance, r 0 , mea- defocusing, the difference between the focal distance and
sured from the grating vertex to the spectral image plane, the distance to the flat array surface, as shown in Figure 10.
varies over the spectral range. This impacts the reciprocal Defocusing produces a blur ellipse with a length in the
linear dispersion and the linearity of the wavelength scale spectral direction approximately equal to the defocusing
as shown in Figure 9. distance divided by the f/# of the system and the cosine
The deviation from a linear wavelength scale results of the angle of incidence at the detector, f. For the
in the data points produced with equally spaced detector example here, this would produce a 100 m blur at the
Wavelength nonlinearity (nm)
20
10
10
20
30
0 5 10 15 20 25 30 35
(a)
Reciprocal dispersion (nm mm1)
30
25
20
15
0 5 10 15 20 25 30 35
(b) Image position (mm)
Figure 9. Wavelength scale deviations: (a) nonlinear wavelength scale; (b) reciprocal linear dispersion.
0.2
Defocusing distance (mm)
0.1
0.0
0.1
0.2
0 5 10 15 20 25 30 35
Image position (mm)
Figure 10. Defocusing distance.

long wavelength end of the spectrum, corresponding to integrating capacitor in a transimpedance amplifier. The
approximately 2 nm wavelength. Reducing the variations charges are then sequentially transferred to the output as
in wavelength scale and resolution become important when a video signal.
attempting to cover a wide wavelength range in a small Linear InGaAs and Ge detector arrays are used to extend
instrument. the responsivity beyond the 1100 nm limit of Si. Integrated
Stark and Landon9 added a field flattening lens to the readout electronics for these arrays are provided as a sep-
basic concave grating polychromator and adjusted the hor- arate complementary metal-oxide semiconductor (CMOS)
izontal focus so that the detector surface was perpendicular buffered multiplexer chip that is connected to the PDA
to the principal ray from the center of the grating to the by wire bonds. These readout electronics utilize a capaci-
center of the array. The more uniform spectral dispersion tive transimpedance amplifier for each pixel. The operation
of this configuration results in a relatively constant spectral alternates between photocurrent integration and readout.
bandpass for each array element.10 This configuration also To avoid the dynamic range limitations of charge storage
allowed insertion of a dichroic beam splitter to image the and transfer electronics, and to provide continuous readout
second order spectrum on a second array in addition to the of the photocurrent, a transimpedance amplifier using a
first order on a longer wavelength array.11 feedback resistor may be used. In this case, a parallel
pinout PDA may be used with a hybrid circuit preamplifier-
multiplexer assembly. Seventy-six-element InGaAs and Si
5 ARRAY DETECTORS FOR NIR PDAs, with pixel areas of approximately 1 mm2 (InGaAs)
SPECTROSCOPY to over 2.6 mm2 (Si), have been used to obtain about 20
to 200 times more radiant energy collecting area per pixel
Several different types of array detectors have been applied than that of the typical arrays using integrated electronics.
to NIR spectroscopy. CCDs and charge-injected devices
(CIDs) are discussed in Array Detectors for Raman
Spectroscopy. Within the wavelength limitations of these 5.1 Principles of photon detectors
silicon devices, these charge-transfer devices may pro-
vide extremely low noise for detection of weak signals, The class of detectors used to form arrays for vibrational
particularly when cooled to cryogenic temperatures. How- spectroscopy are known as photon detectors, as opposed
ever, the dynamic range is limited by noise generated in to thermal detectors which depend on an effect of heating
the readout and the maximum signal allowed due to fill- due to absorption of incident energy. When a photon with
ing the storage capacity of the pixel (the full well depth sufficient energy enters a semiconductor material, it may
limit). Low cost, uncooled, linear CCD detectors originally be absorbed to produce one free holeelectron pair. In the
designed for imaging applications have been applied to presence of an electric field within the material, the holes
spectroscopy; however, the dynamic range is quite limited. and electrons migrate in opposite directions, producing a
Charge-transfer array detectors are not readily available for voltage and/or current in the external circuit. Photon energy
longer wavelength operation due to the limitations of semi- is given by
conductor materials other than silicon. E D hn D hcl1 21
Array detectors that are used for much of visible and
where h is Planks constant and c is the speed of light. When
NIR spectroscopy have rather different requirements than
the photon has insufficient energy to free a holeelectron
those discussed for Raman spectroscopy. The energy being
pair, it is not absorbed and the quantum efficiency falls
measured is typically orders of magnitude greater and the
rapidly to zero. The response of a photon detector is,
measurement times allowed may be shorter. Linear dynamic
therefore, characterized by a long wavelength cut-off that
range is also a major consideration. Therefore, most array
is related to the band gap energy, Eg , of the material, and
spectrometers for visibleNIR spectroscopy utilize multi-
given by
plexed PDAs rather than charge-transfer devices. hc
Linear silicon detector arrays with integrated readout lc D 22
Eg
electronics generally incorporate 64, 128, 256, 512, or
1024 photodiodes with a pixel spacing (pitch) of 25 or Thermal excitation provides a distribution of band gap
50 m. Typical pixel heights of 0.5 to 2.5 mm provide aspect energies that tends to increase the long wavelength cut-off
ratios from 10 : 1 to 100 : 1 to accommodate the height somewhat. Therefore, changes in photodiode temperature
of the entrance slit image. These devices accumulate the cause significant responsivity variation at the wavelengths
photocurrent from each pixel as charge on a capacitor, near the cut-off. Higher temperatures also increase the
which may be either the photodiode capacitance or an probability of electrons in the valence band becoming
thermally excited into the conduction band thereby reducing Depletion region
the shunt resistance of the detector. Lower shunt resistance
p+ layer n layer
causes significant increases in the noise of the detector.
_
Consequently, cooling of a photon detector may signifi- Conduction band
_ _
cantly improve its performance; however, cooling tends to
_ _
reduce the cut-off wavelength.
The quantum efficiency, the ratio of the number of
free electrons generated per incident photon, is limited Band gap energy Eg
by surface reflection from the detector and the absorption + +
parameters of the detector material and geometry. Photon +
losses due to surface reflection from the detector may Valence band
be reduced by an antireflection coating. Photon detectors +
are often characterized by a relatively constant quantum
efficiency, independent of the photon energy, over a broad Figure 12. Photodiode pn junction energy. Redrawn from
spectral range below the cut-off wavelength. There are two Hamamatsu concept (http://usa.hamamatsu.com/opto-semi/photo-
major classes of photon detectors used to produce arrays for diodes/diodes.htm).
spectroscopic purposes. The more common is the planar
For example, to form a Si pn junction, boron is diffused
diffused photodiode, which consists of a semiconductor
into a wafer of n-type Si in individual regions defined by
pn junction formed in the substrate material by diffusion
a mask during the diffusion process. Each pC diffusion
of impurities at high temperatures. Silicon, germanium, and
InGaAs semiconductor wafers provide photodiode detectors region provides an individual active detector area sensitive
for the ultraviolet (UV) through the NIR regions of the to optical radiation. A neutral depletion layer is formed
spectrum. Photoconductive film detector arrays of lead between the pC layer and the n layer. An electric field
sulfide and lead selenide have also been developed to extend exists within this depletion layer which accelerates electrons
the long wavelength cut-off into the mid-infrared as far towards the n layer and holes towards the pC layer as
as 5 m. These arrays provide a lower cost alternative to illustrated in Figure 12. As a result, the depletion layer is
PDAs but the performance is also significantly poorer in emptied of free carriers.
detectivity, in dynamic range, and particularly, in speed When a photon with energy greater than the bandgap
of response. For that reason, PDAs will be covered in of the material enters the photodiode, it is absorbed and a
detail. holeelectron pair is created. The photon may be absorbed
in the pC layer, the depletion layer, or the n layer. In
each layer, the electrons are forced into the conduction
5.2 Photodiode theory of operation band leaving holes in the valence band. Holeelectron pairs
created in the depletion layer are rapidly accelerated in
The usual type of photon detector for array spectroscopy opposite directions by the electric field moving the holes
is the planar diffused photodiode. This structure consists into the pC layer valence band and the electrons into the
of a pn junction formed by diffusion of impurities into n layer conduction band. In the absence of an external
a semiconductor. Due to the impurities, the n region has voltage, holeelectron pairs generated outside the depletion
an excess of electrons and the pC region has an excess region diffuse more slowly until they reach the depletion
of holes, which are locations missing an electron. A cross- layer, where they are rapidly accelerated. The effectiveness
section of the photodiode is shown in Figure 11. The pC of this diffusion process is limited by the carrier lifetime.
and nC refer to more heavily doped regions compared to The thickness and doping concentration of the pC layer,
the doping levels of the n region. the substrate n layer, and the nC layer all establish the
spectral response and the response time of the detector.
Cap Active area Wire bond The PIN (p-intrinsic-n) photodiode is designed with a
Signal
nearly intrinsic n layer, with very low doping, between
p+ diffusion
a relatively thin highly doped pC layer and a relatively
Depletion region thick nC layer. When a reverse voltage is applied, it causes
RL
ntype the depletion region to extend throughout the intrinsic
n+ diffusion material. Photons are primarily absorbed in the depletion
layer where the holeelectron pairs are rapidly swept by
the electric field to the heavily doped p and n regions.
Figure 11. pn planar junction photodiode structure. The frequency response of a PIN diode may be as high
E E 103
Conduction Conduction
band band Germanium
Rshunt(T )/Rshunt(20 C)
102
Direct Eg = 0.75 eV Indirect Eg = 0.67 eV

101
Valence Valence InGaAs
band band Silicon
k k
In0.53Ga0.47As Ge 100
Figure 13. Energy-band diagram for InGaAs and Ge. (Redrawn

from Ban et al.12 with permission of Spectroscopy, Aster Publish- 101
ing, now Advanstar Communications.)
as 1010 Hz. Although such short time constants are not 102
20 10 0 10 20 30 40 50
required for multiplexed array detectors used in vibrational Temperature (C)
spectroscopy, the PIN structure is sometimes used anyway.
Figure 14. Relative shunt resistance vs temperature. (Redrawn
For spectral response beyond the 1100 nm cut-off of from Bertone and Webb.13 )
the silicon photodiode, Ge and InGaAs photodiodes have
been used. Germanium photodiodes are generally made VD across the detector in the absence of radiation.
with n type doping of a p substrate; therefore, they have
1
the opposite polarity from silicon photodiodes. Figure 13 ID D VD RSH 23
shows the energy-band diagram of Ge and InGaAs which
describes the energy E and momentum k necessary to create
holeelectron flow.12 5.3 InGaAs photodiodes
InGaAs has a direct band gap because the minimum
energy gap between the valence band and the conduction Several types of photodiodes based on InGaAs have been
band of 0.75 eV occurs at the same value of momentum k. developed with different energy gaps and, therefore, dif-
Therefore, when a photon is absorbed, an electron can trans- ferent long wavelength limits. The simplest InP (indium
fer directly from the valence band to the conduction band phosphide) based structure is produced by epitaxial growth
without a change in momentum. For Ge, on the other hand, of a layer of n InGaAs and a cap of n InP on an nC
the minimum energy gap of 0.67 eV occurs at different InP wafer as illustrated in Figure 15. In order to match
values of momentum k. Therefore, a phonon (a quantized the crystal lattice parameters of the InGaAs to the InP,
vibration of the lattice) must be created for an electron In0.53 Ga0.47 As is formed yielding an energy gap that estab-
to transfer from the valence to the conduction band. As lishes a cut-off wavelength of 1700 nm. p diffusion through
a result, the performance characteristics, such as quantum the InP cap produces the pn junction. The InP cap causes
efficiency, shunt resistance, dark current, capacitance, and some absorption below 950 nm reducing the responsivity in
temperature dependence of intrinsic Ge detectors, are not that region. SiN is added as an insulator and antireflection
as favorable as InGaAs, which explains why Ge has been coating. This structure has excellent performance even at
largely supplanted by the latter material. room temperature.
Detectors are not ideal diodes that allow no reverse
current, because reverse leakage current does occur, pri- Au/Zn
marily as a result of thermal excitation of electrons from
Signal
the valence to the conduction band. This leakage is char- SiN(AR) SiN(AR)
acterized as a parallel shunt resistance RSH . At similar SiN SiN
temperatures, detectors with lower energy gaps tend to have RL
p: lnP n: lnP
1 m
lower shunt resistance. The shunt resistance may be raised n: lnGaAs
3 m
significantly by cooling. The ratio of shunt resistance to 100 m
n+: lnP
that at 20 C is plotted versus temperature for Si, Ge, and
InGaAs photodiodes in Figure 14.13 It should be noted that
at 20 C, Ge photodiodes have substantially lower shunt Figure 15. Planar InGaAs/InP PIN photodetector. (Redrawn from
resistance than either Si or InGaAs. Shunt resistance affects Ban et al.12 with permission of Spectroscopy, Aster Publishing,
the dark current ID , the current that flows due to a voltage now Advanstar Communications.)
SiN
CO = 1.7 m n: InAs0.6P0.4 ~1 m
CO = 2.62 m
n: InAs0.82Ga0.18 ~3 m
aO = 5.985
CO = 1.7 m ~3 m
n: InAs0.6P0.4
a O = 5.985 Signal
n: InAs0.5P0.5
n: InAs0.4P0.6
RL
a /a" 0.33% n: InAs0.3P0.7 ~1 m
(each step) (each step)
n: InAs0.2P0.8
n: InAs0.1P0.9
aO = 5.985 n: InP
Figure 16. Compositionally graded InGaAs structure. (Redrawn from Ban et al.12 with permission of Spectroscopy, Aster Publishing,
now Advanstar Communications.)
The wavelength response of InGaAs can be extended by order to accommodate defective elements, the missing data
varying the composition. A wavelength limit of 2.62 m are often interpolated from the adjacent values. For this
is obtained with InAs0.82 Ga0.18 . The cap material is to be an acceptable compromise, the defective data points
InAs0.6 P0.4 which absorbs below 1.7 m causing a drop in should be separated by a minimum number of data points so
quantum efficiency below that wavelength. Other Inx Ga1x there is only one defect within the interpolation interval. A
alloys provide 1.9 m and 2.2 m wavelength limits. For requirement that there be no adjacent defects is minimally
these shorter wavelength compositions, the lattice matched acceptable; however, separation by five data points is a
cap material absorption wavelength limit is also lower. In more useful specification.
order to grow these materials starting with an InP wafer,
the composition is graded in steps, similar to the struc-
5.4 Responsivity
ture shown in Figure 16, to minimize the lattice mismatch.
Even so, due to lattice defects, low shunt resistance and Responsivity is defined as the detector output per unit input
high capacitance, the detector performance, while useful, is energy flux. For photon detectors it is usually expressed in
poorer at room temperature than might be expected. How- units of ampere-watt1 . Responsivity is a function of wave-
ever, cooling increases the shunt resistance as shown in length and, for modulated energy, the signal frequency f.
Figure 14.
Defects in the lattice may cause low shunt resistance Rl, f D Sl, fPl1 24
or low responsivity of the detector element. Although such where Sl, f is the output response signal and Pl
defects occur in silicon, they are more common in the more is the incident radiant power at wavelength l. Due to
complex InGaAs devices. Single element detectors showing the decreased energy per photon as wavelength increases
high dark current or low responsivity are simply rejected. and the generally constant quantum efficiency of photon
However, arrays of many elements have an increased detectors, responsivity typically increases as a function
probability of incorporating one or more bad elements of wavelength up to the long wavelength cut-off. For an
thereby decreasing the yield of perfect arrays. The density ideal detector with unity quantum efficiency, at low signal
of defects is relatively constant so the expected number frequency, and below the long cut-off wavelength, lc , the
of defects is proportional to the total array area. In order signal output current (in amperes) is
to maintain high yields, standard InGaAs arrays are usually
Sl D Plhc1 lq 25
specified allowing up to 2% inoperative pixels and extended
wavelength InGaAs arrays as much as 5% inoperative where q is the charge on an electron, 1.6 1019 C. For
pixels. Si arrays with defects are normally rejected. In l in micrometers, therefore, the ideal detector responsivity,
Rl, is the responsivity. The rms random noise, Nf, f, Ad , is

measured with an electrical bandwidth f centered at f.
Rl D SlPl1 D 0.806 l A W1 26 Detector noise tends to be a function of the square root
of the detector area, Ad . D is usually specified for the
If the signal current is integrated, the output signal may
optimum signal frequency fopt , which is a function of the
be expressed in terms of charge Q accumulated in the
detector time constant and the frequency dependence of the
integration time, T, or in the number of signal electrons,
noise.
Se , corresponding to Q, where
In order to compare detector types independent of
Ql D SlT C detector area, Ad , and the measuring bandwidth, f, Clark
27 Jones14 also defined the term specific detectivity, D
Se l D Qlq1 D SlTq1
(D-star), which is D normalized for f D 1 Hz and a
Detectivity, D, is defined14 as the S/N per unit quantity of detector area of 1 cm2 .
incident radiation, which is equal to the responsivity divided
D l, f0 D A0.5 0.5
d f Dl cm Hz0.5 W1 29
by the root mean square (rms) noise, N. The reciprocal of
D is the noise equivalent power (NEP), the radiant power The spectral specific detectivity D of common array
required for the signal to equal the rms noise of the detector. detector materials is shown in Figure 17.
Dl, f, f, Ad D Rl, fNf, f, Ad 1 The D values shown represent state-of-the-art single
element detectors and include only the noise contribution of
D Sl, fPl, f1 the detector and not of the associated electronics. However,
Nf, f, Ad 1 W1 28 the spectral characteristics of Figure 17 still apply as they
depend only on the responsivity Rl.
For the definition of D, the radiant power, Pl, f, is The random noise sources within the photodiode are shot
assumed to be modulated at a frequency, f. The signal noise due to the photocurrent and the dark current and
responsivity, Rl, f, which is a function of f due to Johnson noise due to the shunt resistance of the detector.
the detector time constant, produces a signal, Sl, f, also Shot noise is generated by random fluctuations of the
modulated at frequency f. The rms value of the signal current flowing through the detector, including both the
varies in accordance with the frequency dependence of photocurrent, IP , and the dark current, ID . Its mean square
1015
Si (300 K)
1014
Ext. InGaAs (233 K)1.83 m
1013
Ext. InGaAs (233 K) 2.11 m
D * (cm Hz1/2 W 1)
InGaAs (300 K)
1012 InAs (77 K)

Ext. InGaAs (233 K)2.48 m
InSb (77 K)
1011
Ge (300 K)
PbS (300 K) HgCdTe (77 K)

1010
Ext. InGaAs (300 K) 2.6 m
109
PbSe (300 K)
108
0 1 2 3 4 5 6
Wavelength /m
Figure 17. Specific detectivity D for photon detectors.

value is equal to the incident flux from the background that is within the field
of view of the detector.16 This background radiation adds a
hi2shot i D 2qIP C ID f 30 background photocurrent, IB , to the signal photocurrent, IS .
A convenient expression for rms shot noise is 5.7 Liquid nitrogen (LN2 ) cooled InGaAs arrays may contain
104 If0.5 pA when I is in picoamperes.15 a cold shield to reduce the background signal by limiting
Johnson or thermal noise is caused by the random the field of view of the detector, at least in the direction
motion of charge carriers in a resistive element at ther- perpendicular to the array length. At LN2 temperature
mal equilibrium causing a current corresponding to the (77 K) the shunt resistance of the detector elements is
instantaneous rate of change of charge. As the tempera- sufficiently high that the shunt resistance Johnson noise
ture is increased, the mean kinetic energy of the carriers and the dark current shot noise are both negligible. In one
increases, yielding an increased noise current. The Johnson example, the background signal for an LN2 cooled array
noise current in turn produces a noise voltage across the in a room-temperature spectrometer is quoted as 360 000
resistive element. Ideally pure reactances, such as a capac- electrons for a 60-s exposure.17 This corresponds to a
itor, are free from Johnson noise. The mean square value background photocurrent of approximately 1015 A. In this
of Johnson noise current from the shunt resistance of the particular case, the background shot noise is comparable to
detector is the noise of the electronics.
1
hi2Johnson i D 4kTRSH f 31
where k is Boltzmanns constant, 1.38 1023 Ws K1 . At 5.5 Photodiode and transimpedance amplifier
room temperature, rms Johnson noise current is approxi- circuits
mately 0.13 R1 f0.5 pA, when R is in megohms.
Uncorrelated noise currents add as the sum of the The performance of diode array spectrometers is often
mean square values so the total mean square detector limited by the electronics associated with the detector
noise is rather than the detector element itself. This is particularly
true when the detector is cooled to increase the shunt
hi2detector i D hi2shot i C hi2Johnson i resistance, thereby reducing the detector noise. For most
1
D 2qVD RSH 1
C IP f C 4kTRSH f 32 purposes, the photodiode is connected to the inverting input
of a transimpedance amplifier, that is, an amplifier that
The critical role of shunt resistance in detector noise and converts the input photocurrent to an output voltage signal.
the improvement potentially obtainable from cooling the It is the properties of this combined circuit that control
detector to increase shunt resistance is evident. However, the performance of the spectrometer. A generic equivalent
once the detector noise is below that of the associated circuit of this photodiodeamplifier combination is shown
electronics, there is little benefit in further cooling. The in Figure 18.
noise may also be reduced by limiting the electrical The polarity of the diode shown is consistent with Si and
bandwidth f to that required for the desired signal output. InGaAs detectors and would be reversed for Ge. The series
In the case of an integrating buffer amplifier, f D 1/2T resistance, RS , in combination with the amplifier circuit
where T is the integration time. input capacitance may limit very high frequency response
The theoretical limit of detectivity of a detector is the but both factors are usually very small, and therefore, are
noise that results directly from the photon fluctuations of insignificant in most vibrational spectroscopy applications.
CF
RF
i F2
RS

2
IA+ i A Filter
Ideal diode IP ID RSH CSH i D2 2
+ VO + OS

2
VR VOS + OS
+
+
Figure 18. Photodiode and transimpedance amplifier.

In the capacitive transimpedance amplifier, the feedback For example, when source is modulated by a chopper or
resistor is eliminated RF D 1, a reset switch is added to electrically, f2 would be chosen to pass the fundamental
discharge the feedback capacitor after integration, and no chopping frequency and, perhaps, some significant harmon-
filter is used. ics. It has been claimed that it is not possible to use a
The operational amplifier18 in the equivalent circuit is chopper and a lock-in amplifier with an array detector.19
an ideal amplifier. The difference potential between the However, if each pixel is sampled at a rate more than twice
actual amplifier inputs is VOS , the amplifier offset voltage, the highest significant modulation frequency, i.e. above
which may have either polarity. When no external reference the Nyquist rate, each pixel signal can be synchronously
voltage, VR , is applied to the detector, the photovoltaic demodulated and digitally filtered, gaining benefits equiv-
mode of operation is obtained. With reverse bias VR , the alent to a lock-in amplifier on each channel and more. In
detector is in the photoconductive mode. In either case, practice, this technique has been used to produce a time-
ignoring the random noise components, the output voltage shared dual-beam array spectrometer producing up to 30
for the resistor-capacitor feedback case is baseline and dark corrected spectra per second (SPS). The
sampling for that spectrometer is continuous at a rate of
VO D IP C ID C IA ZF VOS 1 C jGj 33 600 SPS for each pixel.9 11
The magnitude of the detector shunt impedance, jZSH j, is
where ZF is the feedback impedance and jGj is the
magnitude of the voltage gain. IP is the photocurrent, IA is jZSH j D RSH [1 C 2pfCSH RSH 2 ]1/2 38
the input bias current of the amplifier, and ID is the dark
current of the detector, given by and the magnitude of jGj becomes
1 1
jZF j
ID D VR C VOS RSH D VD RSH 34 jGj D
jZSH j
1
The primary advantage of photovoltaic operation with VR D RF RSH [1 C 2pfCF RF 2 ]1/2 [1 C 2pfCSH RSH 2 ]1/2
equal to zero is the reduction of the dark current which then 39
depends only on the amplifier offset voltage, VOS . where CSH is the combined shunt capacitance of the
The operational amplifier also has internal noise sources19 detector, the wiring, and the amplifier input. RSH is the
that are reflected as a mean square input current noise hi2A i shunt resistance of the photodiode. At f1 D 2pCSH RSH 1 ,
and a mean square input voltage noise hv2OS i, each propor- jZSH j D 0.707 RSH . Since CSH is proportional and RSH
tional to the effective noise bandwidth. When a feedback inversely proportional to pixel area, their product tends to
resistor, RF , is used, the added Johnson noise is be a constant for a given material. For room temperature
lattice-matched InGaAs at zero bias, CSH RSH is approx-
hi2F i D 4kTRF1 f 35 imately 1.6 103 s and the frequency breakpoint, f1 ,
occurs at about 100 Hz. Because of their very high shunt
The total rms noise voltage at the output of the amplifier
resistance, typical Si photodiodes at zero bias have CSH RSH
is, therefore,
on the order of 10 to 70 s and a frequency breakpoint, f1 ,
[hv2O i]0.5 D [hi2D i C hi2A i C hi2F i]0.5 jZF j C [hv2OS i]0.5 1 C jGj below 0.1 Hz.
36 At low frequencies, f < f1 , the gain magnitude, jGj,
The response to the input currents, including the desired approaches RF /RSH . Therefore, low RSH increases the
signal photocurrent and various offset and noise currents 1 C jGj multiplier for amplifier voltage offset and noise.
is controlled by the magnitude of the feedback impedance, If CF RF is less than CSH RSH , i.e. f2 > f1 , then the
jZF j. For the resistor-capacitor feedback circuit, gain, jGj, increases starting at f1 and then levels off
at f2 D 2pCF RF 1 . At high frequencies, f > f2 , jGj
jZF j D RF [1 C 2pfCF RF 2 ]1/2 37 approaches CSH /CF . Therefore, if the shunt capacitance
of the detector exceeds the feedback capacitor, the high
At f2 D 2pCF RF 1 , jZF j D 0.707 RF and the response to frequency multiplier 1 C jGj > 2. This is known as gain
input current has started to decline. At higher frequencies peaking. Unlike the current response, there is no high
the response is proportional to 1/f. The corner frequency, frequency rolloff and, in fact, the voltage gain may increase
f2 , is chosen to pass the desired signal frequencies and to and remain high to the limit of the amplifier open loop
reject current noise occurring at higher frequencies. For cur- bandwidth. If this broadband noise is sampled periodically
rent noise uniformly distributed in frequency, such as shot by a multiplexer, the noise is folded back in the frequency
and Johnson noise, the noise bandwidth is approximately domain so that all the noise power is contained below the
1.5 f2 . Nyquist frequency, i.e. one-half the sampling frequency.
Therefore, it is desirable to add a filter with a corner are read out simultaneously with, and subtracted from, the
frequency of about f2 at the output of the transimpedance corresponding integrated signals in an operation known as
amplifier before the multiplexer in order to limit the voltage correlated double sampling. This feature reduces the effects
noise bandwidth to the signal bandwidth. This filter also of reset offsets of the capacitor voltages at the beginning
produces a sharper 1/f2 2 rejection for the current noise, of each scan, of multiplexer related switching noise, and of
reducing the effective bandwidth for the current noise to some very low frequency noise.
approximately 1.18 f2 . In the case of the capacitive transimpedance amplifier,
the charge, QF , on the feedback capacitor, CFx , is
5.6 The capacitive transimpedance readout QF D CFx VF D IS C IB C ID C IA T 40

circuit
where the total photocurrent, IP , is equal to the sum of the
Integrated readout electronics utilize capacitive transimpe- desired signal current, IS , and the background photocurrent,
dance amplifiers and sampling readout circuits to eliminate IB . Ignoring random noise components, the output signal
the high value feedback resistor. A simplified diagram of voltage sampled at time T after integration begins may be
the circuit is shown in Figure 19. expressed as
The equivalent circuit of the photodiode and operational
amplifier, as shown in Figure 18 with RF and the filter VO D QF C1
F C VREF VOS TD0
removed, also applies to the photodiode and amplifier in D IP C ID C IA TC1
F C VREF VOS TD0 41
this circuit. The capacitive transimpedance circuit simul-
taneously integrates the detector current of each element where the currents are average values during the exposure
into a small capacitor, CF , for a controllable exposure time, and the offset voltage, VREF , is supplied to allow operation
T, after the reset is released. The value of the integrating of the circuit with a single power supply voltage.
capacitor may be selected between two values, CF1 and Ideally, for photovoltaic operation the bias voltage on
CF2 , for all pixels simultaneously. A smaller capacitor pro- the detector is zero and there is no average dark current.
duces a larger integrated signal for a given current input. Unfortunately, real amplifiers have an offset voltage VOS
There are usually two sample-and-hold circuits at the out- that appears as a bias across the detector. This bias
put of each buffer amplifier. One is activated to sample and voltage may have either polarity with a magnitude of up
hold the initial signal at the start of charge integration and to a few millivolts. As a result of the bias voltage, a
the second to sample and hold the signal due to the inte- dark current, ID , flows in the detector that is inversely
grated charge at the end of charge integration. These two proportional to the shunt resistance, RSH , of the detector.
signals for each pixel are then simultaneously transferred When the bias voltage at the detector anode is positive,
in sequence by the dual channel multiplexer to a differ- the dark current will have the opposite polarity from the
ence amplifier. Thus, the initial sample and hold signals photocurrent.
Readout circuits
Transimpedance amplifier
Reset Multiplexer
Charge reset sample-hold
CF1 gain1 CS1 Video difference

amplifier
CF2 gain2 +
Signal
sample-hold
Photodiode
+ CS2
Readn
+
VREF

Figure 19. Capacitive transimpedance circuit.

The amplifier input bias current, IA , might increase using a 10.5 pF capacitor has a transimpedance gain of
rapidly with temperature, depending on the type of amplifier 15 nV per e and an rms readout noise of 30 V.20 The
used. JFET amplifier input bias current may be on the order readout noise is, therefore, equivalent to 2000 electrons
of 1013 to 3 1012 A at 25 C but it increases by a factor input charge. When the 0.4 pF capacitor is used, the rms
of 2.3 for each 10 C rise in temperature. The JFET input readout noise increases to 80 V but the transimpedance
bias current is subtracted from the photocurrent but it is gain is 400 nV per e . Therefore, the read noise, although
added to the dark current when the amplifier bias voltage increased in voltage, is decreased to the equivalent of 200
is positive at the detector anode. The input bias current of electrons.
other types of amplifiers may have either polarity.
When the photocurrent is zero, the largest negative sum
of dark current and amplifier bias current produces the most 5.7 Dynamic range and signal amplitude
negative feedback capacitor voltage. resolution
T Dynamic range is a description of the usable range of input
VFmin D jID C IA jmax 42
CF radiant power for which the detector and the electronics
For example, at 20 C, a typical InGaAs array with 50 m will generate a detectable linear output signal. For detection
square pixels has a dark current of 5 pA, corresponding systems, it is often expressed as the maximum linear signal
to a shunt resistance of approximately 109
and a 5 mV divided by the rms noise at the output with no input
voltage offset. Assuming that the amplifier bias current radiant power. As noted earlier, in photovoltaic mode,
is negligible at 25 C and its temperature is not allowed the linear dynamic range of the detector itself may be as
to rise significantly, and that CF D 10 pF, then VFmin D high as 108 to 109 , far beyond the typical dynamic range
0.5 T. If the negative voltage saturation limit is 0.75 V, the of the associated electronics. When a signal is digitized,
maximum integration time allowable is about 1.5 s. Since the minimum detectable signal corresponds to the least
shunt resistance is approximately inversely proportional to significant bit (LSB) of the ADC. The dynamic range at
pixel area, a 50 1000 m pixel might have a maximum the converter is, therefore, limited to the range of the ADC,
integration time of 75 ms. Detector cooling to increase shunt for example, 65536 : 1 for a 16-bit ADC. Gain ranging
resistance becomes advantageous for large pixel arrays. can increase the overall dynamic range of the system by
The total mean square noise voltage at the output of the adjusting the signal amplitude to fall within the ADC range.
amplifier is The signal amplitude resolution describes the precision
within which a signal can be measured, i.e. the ability to
hv2O i D hi2D i C hi2A ijZF j2 C hv2OS i1 C jGj2 C hv2R i 43 differentiate small differences in the signal amplitude. This
precision is limited by the noise level accompanying the
where hv2R i is the added read noise from the reset and
signal which may vary with the signal strength. A measure
multiplexer. Often, the amplifier voltage noise is combined
of signal amplitude resolution is the S/N. It is important to
with the read noise in device specifications. The current
define the signal of interest for this to be a useful definition.
noise bandwidth f of the integrator jZF j is 1/2T.
When a signal is digitized, the smallest difference that can
During reset, the gain, jGj, of the circuit is zero. During
be detected is the LSB. Again, gain ranging can be used to
integration, however,
amplify the signal, thereby decreasing the effective size of
jZF j D 2pfCF 1 44 the LSB.
In vibrational spectroscopy, there are two very different
and the magnitude of jGj becomes situations to be considered. The first case, such as Raman
jZF j spectroscopy, involves the detection and measurement of
jGj D D 2pfCF RSH 1 [1 C 2pfCSH RSH 2 ]1/2 low radiant power in the presence of noise and, in many
jZSH j
45 situations, significant background radiation. The second
For frequencies f1 > 2pCSH RSH 1 , jGj approaches case, based on the absorption spectrum of the sample,
CSH /CF , which is constant with frequency up to the limit of measures sample transmittance or reflectance. This involves
the amplifier open loop bandwidth. Therefore, it is impor- measuring small changes in a relatively high radiant power
tant to minimize CSH to avoid excessive amplifier voltage baseline spectrum. For these two cases, different sources of
noise. This criterion is another factor leading to the use error and noise become important.
of small area pixels in arrays with the capacitive tran- The article Array Detectors for Raman Spectroscopy
simpedance amplifier circuit. Reduction of CF to increase considers the first case. Due to the weak signal, the sample
the transimpedance gain also increases the effect of ampli- and background photon shot noise may be the ultimate
fier voltage noise. For example, a typical InGaAs array limiting factors. Sufficient cooling of the detector can
increase the shunt resistance so that the detector noise is For the example array, if dark current is negligible due
insignificant. Signal integration increases the signal linearly to cooling and CF D 10.4 pF, Qmax D CF VCmax D 20.8
with integration time while the shot noise increases as 1012 C, equivalent to 130 106 electrons and the rms shot
the square root of time, improving the S/N. However, noise is 11.4 103 electrons or 175 V. The maximum S/N
integration time, and hence, the maximum signal are limited is then 2 V/175 V D 11 400 : 1. If dark current limits the
by the charge capacity of the integration capacitor and the integration time, the maximum photocurrent signal is 1.25 V
dark current plus photocurrent. Read noise from the array but the shot noise is unchanged. Therefore, the maximum
electronics, which is not affected by integration time, limits S/N is about 7100 : 1. In the high sensitivity mode with
the minimum detectable signal. Therefore, the maximum CF D 0.4 pF, Qmax is equivalent to 5 106 electrons and
dynamic range for weak signal detection is the maximum the shot noise is 2.2 103 electrons or 890 V. The
allowable signal divided by the read noise. maximum S/N is reduced by a factor of about 5.
In many cases, signal and noise parameters of the Ultimately, the multiplexed signal from the array is
integrating self-scanned array are specified in number of converted to a digital signal that is quantized by the
electrons, Ne . This is equivalent to specifying charge, Q, ADC. To avoid saturation, the largest signal amplitude
where Ne D Q/q. The rms noise of charge Q, in terms of must be adjusted to be within the range of the ADC.
electrons, is equal to the square root of Ne . Also, given a Then, the smallest signal from the array that can be
voltage, V, on a capacitor with capacitance C, the charge digitized equals the smallest step, or LSB, of the ADC.
Q equals CV. Thus, the signal in terms of electrons can State-of-the-art integrated high-speed ADCs suitable for
be computed. Conversely, if parameters are specified in multiplexed signals typically have a maximum resolution
electrons, they may be converted to charge or voltage. of approximately 16 bits meaning that the signal can be
A typical InGaAs array has a maximum output voltage divided up into a maximum of 216 (65 536) levels. The LSB
range Vmax of 2 V and a read noise of 30 V for the high is then 216 or 1 part in 65 536. The LSB of the ADC places
dynamic range mode. The usable signal is Vmax Vdark a limit on the amplitude resolution for a single conversion.
where Vdark is the dark voltage produced due to the The maximum sampling rate of 16-bit integrated ADCs is
dark current during the exposure time. For very short typically 100 to 500 kilo-samples per second (kHz). Faster
exposures, Vdark may be neglected and the dynamic range integrated converters are available but typically with 12 to
is 2 V/30 V D 6.7 104 . Under these circumstances a 16- 14 bit resolution.
bit ADC is required to maintain the dynamic range. Using It is common to co-add spectra to improve the S/N and
the high sensitivity mode, the maximum dynamic range dynamic range. For this to be possible, the noise level must
is 2 V/80 V D 2.5 104 . With longer exposures, Vdark be digitized by at least one or two levels of the ADC.
becomes significant. The maximum Vdark for the example Otherwise, the digital signal will remain unchanged for each
array is 0.75 V limiting the maximum signal to 1.25 V. At co-added spectrum and the noise level will not be averaged
saturation Vdark , the integrated dark current shot noise in the down. Therefore, it is good practice to have the rms noise
wide dynamic range mode is about 100 V and the dynamic level about equal to the LSB. The rms digitizing noise of
range is reduced to about 1.25 V/100 V D 1.25 104 . In an ideal converter is approximately 0.3 times the LSB so
the high sensitivity mode, the dark current shot noise is that will not contribute significant additional noise.
about 550 V and the dynamic range is about 2.3 103 . The resistorcapacitor transimpedance amplifier has
For weak signals, the S/N is maximized by using integra- advantages for use with large pixels at room temperature.
tion times sufficiently long to ensure that the photocurrent A typical InGaAs array with 300 3000 M pixels might
and the dark current shot noise are significantly larger than have a shunt resistance, RSH , of at least 10 M
and a CSH
the read noise. The maximum exposure or integration time of 90 pF. Therefore, f1 is about 177 Hz. In order to limit
Tmax is limited by the maximum charge, Qmax , that can be the low frequency gain jGj and to obtain tiny metal-on-
accumulated on the integrating capacitors. silicon low-noise feedback resistors, RF is set equal to
Tmax D Qmax I1 D CF VCmax I1 46 10 M
also. For comparison with the previous example
using 75 ms integration time, f2 is set at 6.7 Hz using a
The current, I, that is integrated in the positive direction 2400 pF capacitor and a filter with a corner at 6.7 Hz is used
is the sum of the positive detector dark current, positive at the output before the multiplexer. The effective current
transimpedance amplifier bias current, the background pho- noise bandwidth is, therefore, about 7.9 Hz and the voltage
tocurrent, and the desired signal photocurrent. In the neg- noise effective bandwidth is 10 Hz.
ative direction, photocurrent does not exist so the detector The amplifier has an input offset of 50 V yielding a
dark current and amplifier bias current may limit the inte- dark current of 5 pA, an input noise voltage from 0.1
gration time. to 10 Hz of 100 nV, an input bias current of 20 pA which
reduces with increased temperature, and an input current and sR for the associated random noises, yields
noise density of 20 fAHz1/2 . The maximum linear output sS
is 10 V. S 1 C
tD
S C sS
D S 48
Adding the Johnson noise of the feedback resistor to
R C sR R 1 C sR
equation (43) and removing the read noise term gives the R
mean square noise output voltage for the resistorcapacitor 0
A C sA D logt
configuration.
sS sR
hv2O i D hi2D i C hi2A i C hi2F ijZF j2 C hv2OS i1 C jGj2 D logt log 1 C C log 1 C 49
S R
47
In the absence of photocurrent, the output rms noise is and using the first term of the series expansion for log X,
approximately 1.74 V. Hence, the dynamic range before 1/2
the ADC is 5.7 106 . This is over 64 times the dynamic s S 2 s R 2
sA 0.434 C 50
range of the ADC so gain ranging of up to 64 : 1 can be S R
applied to fully utilize the signal dynamic range. If the gain
If the reference measurement time is significantly longer
is set such that the full scale of the ADC is 200 mV, the LSB
than the sample or the sample signal is substantially less
is 3.0 V, and the maximum photocurrent is 20 nA. The
than the reference signal, sA 0.434 sS /S. The minimum
photocurrent shot noise is 2.25 V and the combined noise
absorbance noise equals 0.434 divided by the maximum
is about 3 V. The maximum S/N is 6.7 103 . The noise
S/N of the measurement.
equivalent photocurrent under these conditions is 0.3 pA.
Therefore, for the self-scanned array cooled to obtain
The peak responsivity of InGaAs is about 1.0 AW1 so the
negligible dark current, and thus, for a maximum S/N of
minimum NEP of the array is 3 1013 W.
11 400, the minimum rms absorbance noise is
In absorption spectroscopy, the maximum signal arises at
the peak of the baseline energy spectrum. For example, for 0.434
sAmin D 38 Abs 51
a tungsten lamp at 2800 K, the peak signal typically occurs 11 400
at about 1200 nm (see Figure 5). The signal at 400 nm may
In the case, when the dark current limits the integration
be less than 2% of that value. Therefore, for a 16-bit ADC,
time,
the available dynamic range at 400 nm would be limited to 0.434
about 1200 : 1 and the rms absorbance noise on the baseline sAmin D 61 Abs 52
7100
would be about 360 AU. A band with an absorbance of
one would have a maximum S/N of 120 : 1 resulting in an For the resistorcapacitor configuration, the single scan
absorbance noise of 3 mAU. Similar, though less dramatic, S/N is 6.7 104 and sAmin is 7 Abs. In either configura-
reductions of dynamic range occur in the NIR. The signal tion, co-adding multiple scans would increase the S/N by
cannot be increased by simply increasing the source energy the square root of the number of scans at the cost of mea-
or the optical throughput due to saturation at the peak. surement time. The trade-off of measurement time versus
To increase the dynamic range capability of the ADC, S/N depends on the application. Although the time of a
gain ranging, similar to that employed in FT spectrometers, single scan is short, typically a fraction of a second so that
may be used. In this technique, as the pixels are scanned, the averaging multiple scans may not be a major drawback, an
gain is lowered for strong signals so the radiant power at the increasing number of applications in process control require
array can be increased. For this technique to be effective, rapid measurements that limit averaging.
the amplifier dynamic range must be greater than that of
the ADC, e.g. the buffer amplifiers must not saturate on
the strong signals and their S/N should exceed the ADC 6 SOURCES FOR VISIBLE-NIR ARRAY
dynamic range. This capability is uniquely provided by the SPECTROSCOPY
resistorcapacitor transimpedance amplifier.
In absorption spectroscopy, the transmittance, t, or The tungstenhalogen lamp is by far the most common
reflectance, r, to be determined is the ratio of the radiant source used in visible and NIR diode-array spectroscopy
power transmitted or reflected by the sample to that due to its low cost and general reliability as a radia-
irradiating it. Usually, the desired analytical measurement is tion source in these two spectral regions. A significant
the absorbance logt or, for reflectance, logr. Using problem does occur in some applications caused by the
transmittance, t, and absorbance, A, generically to represent tungstenhalogen lamps low intensity in the blue and UV
either a transmission or reflection measurement, S for the spectral regions. This low output in the blue region becomes
sample signal and R for the reference or baseline signal, sS even more dramatic and disconcerting when compared
to the tungstenhalogen lamps naturally large intensity where c1 and c2 are constants with values 5.8474
output in the adjacent NIR spectral region. When this 104 W cm2 sr1 m4 and 1.438 104 m K, respectively.
imbalance of radiation intensity is coupled with the high The emissivity, el, T of tungsten with a filament
NIR sensitivity of silicon photodiode detectors, applications temperature of 2800 K peaks at 0.4 m with a value of
which use data that should reasonably span the visible and 46.1% and falls to 22.4% as the wavelength increases
NIR regions are at a disadvantage. to 2.6 m (Figure 20). Thus both the blackbody radiance
To reduce this problem in the visible region, hybridized and the loss of emissivity reduce the NIR radiance of a
sources have been developed using blue light emitting tungsten filament. The rapid reduction in spectral radiance
diodes to augment the tungstenhalogen lamp. To further moving from the peak to the short wavelength region is
extend operation into the UV region, a deuterium arc also obvious. At a 2800 K filament temperature the spectral
lamp may be used either alone or in combination with a intensity at 0.400 m is less than 5% of the peak spectral
tungstenhalogen lamp if operation throughout the visible intensity in the NIR while at 0.250 m in the UV region,
is desired. Xenon flash and arc lamps have also been used the spectral intensity is a factor of about 4500 below the
when a particularly high radiant intensity source or short NIR peak. Increasing the filament temperature to 3100 K,
measurement time is needed. However, the higher cost and at a cost of a greater than ten fold reduction in lamp
lower stability of these arc sources has led to the use of life, still leaves the 0.25 m spectral energy at a factor
tungstenhalogen lamps whenever possible. of 1000 below the spectral energy at the NIR peak. As
a result, excellent blocking of the NIR energy is required
to make measurements in the blue and UV regions. Silicon
6.1 Tungstenhalogen lamp spectral radiance array detectors with reduced NIR sensitivity and enhanced
blue/UV response have also been developed.
The overall shape of the tungsten filament spectral radiance The spectral radiance of a tungsten filament varies
LS l, T is that of a blackbody at the filament temperature significantly with the filament temperature as shown in
modified by the spectral emissivity, el, T, of tungsten, Figure 21 for a 2800 K lamp.
which is a function of both wavelength and temperature. Below 1 m, the temperature sensitivity rises rapidly as
wavelength is decreased. Because the lamp is in equilibrium
LS l, T D el, Tc1 l5 ec2/lT 11
with its ambient surroundings, changes in the ambient
W cm2 sr1 m1 53 temperature are reflected in the lamp output resulting in
70
Black body (3100 K)

60
Spectral radiance (W sr1 cm2 m1)
50
40 Black body (2800 K)
30
Tungsten (3100 K)
20
Tungsten (2800 K)
10
0 0.5 1.0 1.5 2.0 2.5 3.0

Wavelength /m
Figure 20. Spectral radiance of tungsten filament.

3500
3000
2500
Fractional change per K
2000
1500
1000
Tungsten
500
Black body
0 0.5 1.0 1.5 2.0 2.5 3.0

Wavelength /m
Figure 21. Tungsten radiance temperature sensitivity.
baseline drift. Dual-beam spectrometer operation, optical filament. Hirschfeld23 extended this concept by enclosing
feedback for lamp control, and/or signal processing to the lamp in a polished reflective sphere with an opening just
reduce baseline offset effects are among the approaches large enough to accommodate the desired exit beam. With
used to deal with this problem. Due to the high temperature this geometry, the radiant energy that is not contained in
of the lamp envelope, there will be refractive index the exit beam is redirected to the filament. The reflectivity
changes and possibly turbulence in the surrounding air of the filament is 1 e so one-half to three-quarters of the
that can spatially and temporally modulate the energy. The energy striking the filament is reflected, much of it into
source optics may be designed to minimize these effects, the exit beam. The result is a substantial increase in the
which have been observed to produce short term energy apparent emissivity of the lamp filament and the usable
fluctuations on the order of 20 AU when illuminating a output radiation as well as a reduction in the electrical input
fiber optic.21 power required.
The filament of a typical low-voltage tungstenhalogen
lamp is coiled and often the coil is also coiled. The
output radiation is reduced by both the fill factor of the 6.2 The halogen cycle
coiled filament, which is the percentage of the source area
being utilized that is filament rather than the intervening The fill gas of a tungstenhalogen lamp is usually kryp-
gaseous space, and the transmission properties of the lamp ton or xenon with a trace of bromine or iodine added. The
envelope, which is usually quartz. A little appreciated effect halogen is added to remove evaporated tungsten from the
of the coiled filament is partial polarization of the lamp inner surface of the bulb, where it would absorb energy
radiant output. A typical lamp may have as much as 10% thereby reducing the lamp efficiency and heating the enve-
polarization.22 The combination of the grating and lamp lope, ultimately to the point of failure. The halogen reacts
polarization may result in a sample orientation sensitivity chemically with the tungsten deposit on the envelope to
that is significant for oriented samples. produce tungsten halides which evaporate at the relatively
It is relatively common to use a spherical reflector behind low temperature of the envelope. When the tungsten halide
a lamp to re-image the filament on itself thereby appearing reaches the higher temperature filament, it breaks down
to fill in the spaces and increasing the apparent fill factor. redepositing tungsten on the filament. Although filament
The energy reabsorbed by the filament is added to the evaporation and redeposition of tungsten are nonuniform,
electrical energy input such that the lamp voltage must be the tungstenhalogen cycle does increase filament life
reduced to maintain the desired operating temperature of the somewhat. The most important effect is the reduction of
bulb blackening with age. Operation of the halogen cycle 0.4

requires the envelope surface temperature to be above
(W cm2 nm1) at 30 cm
Fused silica window
200 C to ensure evaporation of the tungsten halide. There-
Radiant intensity
fore, small size bulbs are generally used, usually of quartz
0.1
due to its strength at high temperatures. Operation below
90% of the rated voltage is likely to reduce the bulb tem-
perature sufficiently to allow bulb blackening to proceed 0.04 UV glass
unabated. window
6.3 Tungstenhalogen lamp life and failure 0.01

modes 160 200 240 280 320 360 400
Wavelength / nm
Tungstenhalogen lamp life is a direct function of filament Figure 22. Deuterium lamp radiant intensity. (From ScienceTech
temperature. Lamps rated at 2800 K may have an average Incorporated, http://www.sciencetech-inc.com)
life of 10 000 h or more but raising the temperature to
3100 K will typically reduce the life to less than 1000 h. intensity drops steadily with increasing wavelength to about
Within a reasonable range around the rated voltage, lamp 5% of the maximum at 0.4 m (Figure 22). The lamp
life is approximately proportional to the 12th to 13th consists of an anode and cathode in a fused silica envelope
power of the applied voltage. Tungstenhalogen lamps filled with low pressure deuterium gas. A typical lamp is
operated within the voltage rating usually fail from melting operated at constant current of 300 mA with a voltage drop
or breakage of a thin spot in the filament. Sometimes of 80 V. A discharge starting voltage of about 400 V is
the thin spots are due to the uneven evaporation and required. A cathode warm-up heater is also used for starting.
redeposition of the tungsten. Another failure mode peculiar Depending on the specific lamp, the heater power may be
to halogen lamps is filament necking or notching. The ends reduced or eliminated after about 20 s.
of the filament are cooler due to thermal conduction of The arc instability results in a specified peak to peak out-
the lead wires, therefore, the halogen attacks the filament put intensity fluctuation of 0.03% to 0.05%, corresponding
at these points. As the filament gets thinner it heats to 130 to 217 AU. Based on peak to peak being about
more, stopping the erosion, but loss of tungsten continues five times the rms value, the rms absorbance noise is typ-
at cooler points closer to the lead wires. This process ically between 25 and 50 AU. The maximum output drift
results in a notch forming at each end of the filament. is typically 0.3% per hour corresponding to 1300 AU
Under continuous operation the notch may deepen until the per hour.
filament breaks. Deuterium arc lamps may be pulsed at frequencies of up
The electrical resistivity of the filament is much lower to 2 kHz to increase the maximum intensity. The average
at room temperature than when operating. Therefore, as current is maintained at 300 mA but a peak current may be
voltage is applied there is typically an initial current surge as high as 1.3 A. Pulsed operation generally uses a keep-
of about 14 times, and possibly up to 20 times, the normal alive current to avoid the need for re-ignition on each pulse.
operating current. The notched ends of the filament heat
more rapidly than the rest of the filament when the lamp is
turned on which increases the resistance and the need for 6.5 Xenon arc and flash lamps
power dissipation at this point over the rest of the filament.
Unless the lamp power supply limits the inrush current, Xenon lamps consist of a cathode and anode in an envelope
the thin areas of the filament may overheat sufficiently to filled with high pressure xenon gas. For spectroscopic use,
melt or break. The current remains above equilibrium for arc lengths of a few millimeters are defined by the cathode
some time after startup until the entire body of the lamp to anode spacing. The spectral distribution of radiant
reaches equilibrium temperature. This may take several intensity for a xenon lamp is composed of radiation from
minutes. the plasma, which is approximated by a 6200 K black body
distribution, and a series of strong lines in the 0.8 to 1.0 m
region as shown in Figure 23. The relative contribution of
6.4 Deuterium arc lamps the plasma and the line structure is a function of how hard
the lamp is driven.
Deuterium lamps provide intense UV radiation in the 0.185 Low power xenon lamps for continuous operation range
to 0.400 m region with a maximum at about 0.2 m. The from 75 to 150 W input power. A current regulated power
100
80
Spectral radiant intensity (rel. units)

Visible
region
60
Xenon lamp
40
Black body radiator
20
300 400 500 600 700 800 1000 1200 14001600 1800 2000 2200 2400 26002800
Reduced scale
Wavelength / nm
Figure 23. Spectral distribution of xenon lamp. (From ScienceTech Incorporated, http://www.sciencetech-inc.com)
supply providing 5 to 7.5 A direct current plus a high led to the development of excellent array detectors for
voltage trigger source are required. The arc length is these spectral regions. Array spectrometers have unique
typically 1.3 mm for a 75 W lamp up to 2.5 mm for a 150 W capabilities and features that enable many new applications
lamp. The highest radiance occurs in the plasma adjacent to be developed, from low light level scientific experiments
to the cathode tip. Xenon arc lamps require up to 20 min to on-line process control.
to stabilize. After stabilization, high stability lamps may
drift 0.5% and have short term fluctuations of up to 1%,
equivalent to 4300 AU. The average life of such a xenon ABBREVIATIONS AND ACRONYMS
continuous arc source is typically 1000 to 2500 h.
Xenon flash lamps provide short very high intensity ADC Analog-to-digital Converter
pulses of light. Pulse widths of a few microseconds and CIDs Charge-injected Devices
pulse repetition rates of 30 to 100 pulses per second are typ- CMOS Complementary Metal-oxide Semiconductor
ical. Operating voltages of up to 1000 V may be required DWDM Dense Wavelength Division
with trigger potentials of up to 10 kV. The peak to peak Multiplexing
output fluctuation of a typical xenon flash lamp is speci- LN2 Liquid Nitrogen
fied at 3% maximum, equivalent to 0.012 AU. This large LSB Least Significant Bit
fluctuation makes simultaneous dual beam detection imper- NEP Noise Equivalent Power
ative. Flash lamps have been used with dual beam scanning PDA Photodiode Array
grating spectrometers with a flash for each wavelength PIN p-intrinsic-n
position. However, dual-array spectrometers are particu- SPS Spectra Per Second
larly suited to pulse operation because all sample and
reference wavelengths are detected simultaneously. Xenon REFERENCES
flash lamps with a guaranteed life of 80 million flashes are
available. 1. American Society for Testing and Materials, Standard Termi-
nology Relating to Molecular Spectroscopy, ASTM Standard
E 131, ASTM, Philadelphia, PA (1999).
7 CONCLUSION 2. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
Spectrometry, Chemical Analysis, John Wiley & Sons, 67,
Vol. 83, New York, NY (1986).
Array spectrometers for the visible and NIR spectral
3. American Society for Testing and Materials, Standard Termi-
regions can be optimized for many different applications by nology Relating to Molecular Spectroscopy, ASTM Standard
selection of the components and the system design. There E 131, ASTM, Philadelphia, PA (2000).
is no single ideal configuration to fit all needs. Fortunately, 4. A. Thevenon, J. Flamand, J.P. Laude, B. Touzet and J.M.
the requirements of the telecommunications industry have Lerner, SPIE Proc., 815, 136 (1987).
5. Schott Glaswerke, Optical Glass Filters, Mainz, Germany. 15. L. Smith and D.H. Sheingold, Noise and Operational Ampli-
fier Circuits, Analog Devices Application Note AN-358,
6. H.A. Rowland, Philos. Mag., XVI, 197 (1883).
Analog Dialog 3-1 (1969).
7. H. Noda, T. Namioka and M. Seya, J. Opt. Soc. Am., 64, 1031
16. J.A. Jamieson, R.H. McFee, G.N. Plass, R.H. Grube and R.G.
(1974).
Richards, Infrared Physics and Engineering, McGraw-Hill,
8. H. Noda, T. Namioka and M. Seya, J. Opt. Soc. Am., 64, 1037 New York, 153 (1963).
(1974). 17. Understanding the Signal-to-noise Performance of InGaAs
9. E.W. Stark and D. Landon, High Speed. Large Dynamic Array Detectors in Spectroscopic Applications, JYHoriba
Range 5001800 nm Diode Array Spectrophotometer, Pre- Technical Note, Jobin Yvon, Inc. (http://www.jyhoriba.com).
sented at Pittsburgh Conference, New York, NY (1990). 18. R. Stata, Users Guide to Applying and Measuring Oper-
10. E.W. Stark, Methods for Evaluating the Performance of a ational Amplifier Specifications, Application Note, Analog
Diode Array Spectrophotometer, Presented at Third Interna- Devices, Norwood, MA (1969).
tional Conference on Near Infrared Spectroscopy, Brussels, 19. L. Smith and D.H. Shelngold, Noise and Operational Ampli-
Belgium (1990). fiers, Application Note, Analog Devices, Norwood, MA
11. E.W. Stark, US Patent 4 997 281 (1991). (1969).
12. V.S. Ban, G.H. Olsen and A.M. Joshi, Spectroscopy, 6(3), 49 20. Sensors Unlimited, Inc., Dynamic Range of the LX-series
(1991). InGaAs Linear Array, Technical Notes (http://www.sensors-
inc.com/980005a.htm).
13. N. Bertone and P. Webb, Noise and Stability in PIN Detec-
tors, Technical Note, Perkin-Elmer (http://opto.perkinelmer. 21. E. Stark, unpublished.
com/library/papers/tp7.htm). 22. T. Hirschfeld, personal communication (1979).
14. R. Clark Jones, Proc. IRE, 47, 1495 (1959). 23. T. Hirschfeld, US Patent 4 346 323 (1982).
Miniature Spectrometers
Roman V. Kruzelecky and Asoke K. Ghosh
MPB Technologies Inc., Pointe Claire, Quebec, Canada
1 INTRODUCTION integrated-optic spectrometers with no moving parts5 7 that

offer portability, robust optical alignment, spectral scan
Most molecules that comprise solid, liquid or gaseous rates exceeding 500 scans s1 and relatively high perfor-
samples have characteristic vibrational modes associated mance. The availability of these miniature spectrometers
with their chemical bonds that can interact with photons.1 provides new possibilities for extending the use of spec-
This results in optical absorption bands, mainly in the troscopy to on site process monitoring. As the overall
mid-infrared and near-infrared (NIR) spectral range, that cost of the miniature spectrometers decreases with produc-
are indicative not only of the chemical bond but also of tion volume, new application opportunities will develop for
the actual bonding configuration and the local chemical imbedding dedicated spectrometers permanently in equip-
environment.2 Non-contact infrared (IR) spectral reflection ment wherever spectral information can benefit the perfor-
or transmission measurements probe these characteristic mance and operating costs of the equipment. An example
molecular vibrational modes to provide valuable informa- is monitoring the combustion of an engine to maintain effi-
tion about the chemical bonding and composition in solids, cient fuel usage and predict maintenance requirements.
liquids and gases. IR spectroscopy, therefore, can yield more This article mainly considers integrated monolithic and
information than methods that merely provide data about the guided-wave spectrometers,5 7 focusing on the IOSPEC
relative concentration of the components of a sample. (integrated optical spectrometer) technology as developed
Large bulk-optic NIR dispersive3 and Fourier transform by MPB Technologies Inc. (Pointe Claire, Quebec, Canada),
infrared (FT-IR)4 spectrometers are routinely employed for in conjunction with INO (Sainte-Foy, Quebec, Canada).
product analysis under controlled laboratory conditions. Other fabrication technologies such as the LIGA pro-
However, IR spectroscopy has a broad range of applica- cess of molding microspectrometers using polymers,8 and
tions that extend outside the laboratory. Applications range the miniaturization of bulk-optic spectrometers,9 are also
from in situ material and process monitoring on a produc- discussed.
tion line, environmental monitoring, geological surveys and IOSPEC 5,6 employs a broad-band IR slab-waveguide
planetary exploration, to biomedical measurements on tis- structure to integrate an input IR fiber or slit, a con-
sue and blood samples. While the bulk-optic instruments are cave reflection grating, and a linear detector array at the
suitable for spectral analysis under controlled conditions, optical output, in a compact, monolithic structure (see
their operation and calibration can be sensitive to various Figure 1). Light is coupled into the spectrometer either
environmental operating conditions including mechanical directly through a miniature slit, or through a suitable
vibrations, thermal fluctuations, humidity and dust. More- IR fiber array. The pigtailed input fiber precisely defines
over, the overall cost, size and weight of the spectrometer the position of the diffracted signal at the output focal
systems are also important considerations. plane, providing very stable long-term optical alignment.
The evolution of linear detector technology has cre- The optical signal is guided within the slab waveguide
ated new possibilities for the realization of miniature onto a master blazed grating structure that also serves as a
concave reflector. The diffraction grating is formed in crys-
John Wiley & Sons Ltd, 2002. talline semiconductor material using proprietary multistep
The use of a relatively thin optical waveguide to provide

vertical optical confinement simplifies the optics and min-
imizes the amount of substrate material required. The
waveguide is formed by sandwiching a suitable, opti-
cally transmissive core material between two cladding
layers with a slightly lower refractive index than the core.
This facilitates optical confinement within the core layer
through total internal reflections at the core/cladding inter-
(a) (b) faces. A high-quality, single-crystal core layer should be
Figure 1. (a) Miniature slab waveguide spectrometer with pig- employed to minimize scattering at inhomogeneities and
tailed fluoride fiber, and (b) high-resolution 3 to 5 m Si-based grain boundaries.
slab waveguide spectrometer pigtailed to a ruggedized IR fluoride Design issues associated with the waveguide include
fiber array and coupled to a 256-channel PbSe array. the refractive index, the effective numerical aperture
(NA), spectral optical propagation loss, core size and
processing involving photolithography, reactive ion etching thermalmechanical characteristics. The NA of the wave-
(RIE) and additional chemical processing. This yields near- guide [see equation (1)] is controlled by the refractive index
ideal grating elements, with peak diffraction efficiencies of the core, n(core), and the difference in the refractive
typically exceeding 85%. Additional integrated optics help index between the core and cladding, n, such that10
to linearize the output focal plane, which is as wide as
20 mm, and to focus the dispersed signal onto the detector (NA)2 D 2ncore n 1
array. Depending on the resolution and bandwidth require- The corresponding maximum angle of propagation within
ments, the IOSPEC structure can be as small as 3.0 by the waveguide, qm , is given by equation (2):
4.5 cm. IOSPEC is currently commercially available in
NA
several broad-band models (1 to 5 m, 1 to 3 m, and 3 to sinqm D 2
5 m) coupled to 256-channel thermo-electrically cooled IR ncore
detector arrays. The packaging consists of a compact opti- For spectroscopy, it is desirable to have a waveguide with
cal spectrometer unit (OSU), as shown in Figure 1(b), that a large input acceptance angle to provide efficient input
is about 19.5 19.5 14 mm and weighs under 2 kg, and optical coupling. However, it is also desirable to have a
a separate electronics control unit. small vertical incidence angle on the grating to minimize
aberrations. By selecting a high-index core layer such as
Si, it is possible to have both a high input NA and a low
2 REVIEW OF FABRICATION propagation angle within the waveguide.
TECHNOLOGIES FOR MINIATURE Several core/cladding structures have been developed
SPECTROMETERS for our spectrometer based on either Si or ZnSe. One
type5 employs high-temperature chemical vapor deposition
This section reviews some of the fabrication technologies of group IV alloys. The NA of the waveguide can be
that are currently employed for the realization of miniature controlled from 0.3 to about 0.7 through the chemical
optical spectrometers. composition of the cladding layers. This structure offers
high-temperature stability to beyond 500 C and low optical
propagation loss, typically below 0.25 dB cm1 , even in thin
2.1 Slab waveguide spectrometers layers. The main loss mechanism in the various waveguides
is scattering and absorption at the core/cladding interface.
There are numerous advantages to the use of a solid As a result, increasing the core thickness, or reducing the
medium for the fabrication of IR spectrometers. Water input signal NA, reduces the propagation loss. In thicker
vapor exhibits significant optical absorption in the mid- waveguides with 1 to 2 mm core layers, the propagation
infrared, with broad peaks at 1.87, 2.7 and 6.2 m. Many loss can be below 0.1 dB cm1 .
IR spectrometers require special sealing and N2 purging Two types of slab waveguide spectrometers are feasible:
to reduce this background optical absorption. High-purity a fully integrated spectrometer in which all the optical
IR-transmissive materials are currently available, such as elements are microfabricated on the slab waveguide (see
Si and ZnSe, with optical absorption coefficients below Section 2.2); and a hybrid approach in which some of
0.005 cm1 in the 1.2 to 12 m range, facilitating the the optical elements, such as the grating, are fabricated
fabrication of solid-state spectrometers that are insensitive separately and subsequently affixed to the waveguide (see
to environmental background absorption. Section 3).
Miniature Spectrometers 3
2.2 Micromachining of spectrometer optics by scanning electron microscopy (SEM). Some roughness
resulted from the initial patterning of the NiCr mask by
Lithographic techniques facilitate the planar patterning of lift-off. This could be improved by patterning the NiCr
optical structures on a suitable slab waveguide to yield mask using a suitable RIE process. Some improvements
a fully integrated spectrometer suitable for mass pro- to the sidewall verticality should be attainable by cooling
duction. As shown in Figure 2(a), several spectrometers the substrate during the RIE and by using a low-pressure
can be patterned on a single 10 cm outer diameter sub- (<1 mTorr) plasma etch process such as electron cyclotron
strate. This microfabrication technology has matured con- resonance (ECR) RIE. More recently,13 chemically assisted
siderably for the fabrication of integrated circuits and is ion beam etching of InP-based waveguides and RIE etching
finding new opportunities for the fabrication of micro of SiO2 -based waveguides has yielded sidewall vertical
electromechanical systems (MEMS).11 The main techni- angles below 1 .
cal challenge is the patterning of the grating elements. The main drawback of this technology is that the
These involve submicrometer features and require optically waveguide core size is limited to about 10 to 15 m. This
smooth, perpendicular sidewalls to minimize optical signal does not provide sufficient optical coupling for general IR
scattering. The main techniques available are RIE12 and ion spectroscopy. However, one very large potential application
beam milling.13 is for wavelength-division multiplexing (WDM) in fiber-
The fabrication of planar gratings requires a deep (15 optic communications systems.13 Here, the coupling to
to 20 m), highly anisotropic etch through a suitable 10-m core communications fiber can be very efficient. By
mask. Some preliminary work was performed by MPB operating at higher diffraction orders, m 1, very high
Technologies in conjunction with Professor H. Craighead resolutions (<0.2 nm) are feasible over narrow bandwidths.
using the facilities at Cornells Nanofabrication Facility. The WDM spectrometer unit cost can be quite low in mass
A slab waveguide structure consisting of (5 m Si)/(10 m production quantities.
Si0.91 Ge0.09 ) was prepared on semi-insulating Si by
heteroepitaxy. The optical elements were patterned by
photolithography, using a 4 to 1 optical reduction scheme 2.3 Injection-molded spectrometers
to improve resolution. A Ni/Cr mask was employed for
the deep etch of the Si/SiGe/Si. An SF6 /O2 12 plasma The LIGA-technique has also been employed to fabricate
etch chemistry was employed in a radio frequency (rf) miniature spectrometers for the visible and NIR.8 (LIGA
magnetron system for the Si and SiGe layers. This provided is a German acronym for Lithography, Galvanoformung
a relatively high etch rate of about 1 m min1 at 10 mTorr and Abformung, with an English translation of lithogra-
that minimized degradation of the mask layer. Etch depths phy, electroforming and molding.) In the LIGA process, an
of 15 m were attained with a vertical to horizontal aspect X-ray-sensitive polymer resist sheet, of the desired thick-
ratio as high as 20. The sidewalls exhibited some pitting ness of the waveguide core, is bonded onto glass or Si.
on the scale of 0.1 m. A Ti mask is used to pattern the resist under exposure to
Figure 2(b) shows a planar view of one of the gratings X-ray synchrotron radiation. The exposed resist is removed
formed by the preliminary RIE at Cornell, as obtained by chemical stripping. This is employed to produce a
0001 15 KV X3,000 10 m WD24

(a) (b)
Figure 2. (a) Mask for integrated IR spectrometers, and (b) planar grating element formed in Si/SiGe/Si by RIE.
molding insert. The mold can then be used to produce 3 THE OPTICS OF MINIATURE
many copies by hot embossing. For the NIR, the spec- SPECTROMETERS
trometer waveguide consists of an air gap between two
plates containing vapor-deposited Au. This can provide This section discusses the unique features of slab waveguide
attenuation below 0.3 dB cm1 in waveguides exceeding spectrometers and the design considerations inherent in
0.5 mm in thickness for an input NA D 0.22. The loss is miniaturizing the spectrometer optics.
due to scattering and the finite reflectivity at the air/Au
interface.
A resolution of about 10 nm was obtained over a 3.1 Performance considerations
400-nm bandwidth in the NIR near 1.6 m.8 Extraneous
stray light and noise is about 5% of the peak transmit- The design of guided-wave and monolithic miniature spec-
tance of the LIGA spectrometer. The typical waveguide trometers presents new opportunities for the application of
core height is about 150 m. Core heights up to 500 m spectroscopy but also poses many technical challenges and
are possible, although there can be some degradation of trade-offs. Firstly, the requirements of the end application(s)
the accuracy, verticality and smoothness of the grating must be considered: the necessary spectral range, spectral
elements. resolution and overall signal-to-noise ratio (S/N). Although
an IR spectrometer operating from 1 to 12 m with 0.2-nm
resolution may be desirable for general laboratory measure-
2.4 Miniature bulk-optic spectrometers ments, the specifications for a more task-dedicated on site
spectrometer unit can often be focused on the necessary
The classic bulk-optic spectrometer can be reduced in size instrument parameters to meet the requirements of the field
by folding the optical path using mirrors. Various grating or on-line sample analysis. Due to the relatively small input
mountings are possible, but the Rowland geometry is the aperture and tight engineering tolerance of typical miniature
most common amongst the various commercial suppliers. spectrometers, optimal matching of the overall system com-
The use of a linear detector array at the spectrometer out- ponents is required to achieve high performance.
put eliminates the requirement for mechanical motion. The In the visible spectral range, the availability of low-cost,
input slit width is reduced relative to the full-size spec- Si-based CCD optical sensors with 512 to 1024 channels
trometers to be compatible with the typical pixel width of has provided the possibility of relatively high-resolution
linear detector arrays; about 25 m in the ultravioletvisible and inexpensive miniature spectrometers. In the NIR and
(UVvis) and 50 to 100 m in the IR. The important IR, 256-channel diode and photoconductive linear detector
arrays are becoming relatively standard. The main challenge
design considerations are the linearity of the output focal
to larger detector arrays is the multiplexing and the large
plane and the focal plane spot size. This requires careful
number of interconnects required. This can be simplified
design of the spectrometer optics and grating to minimize
using flip-chip interconnects in place of gold wire bonding
aberrations.
to improve the integration between the detector pixels and
Ocean Optics9 has pioneered the spectrometer on a
the switching electronics. As the development of integrated
printed-circuit card technology for the UVvis, vis and
IR quantum-well intrinsic photoconductor (QWIP) arrays
short-wave NIR spectral regions. By interchanging grat-
in IIIV compound semiconductors progresses, IR detector
ings, various spectral ranges and resolutions are feasi-
arrays with 1024 channels should become relatively stan-
ble in the 200 to 1100-nm range. These spectrometers
dard. The total operating spectral bandwidth (BW) avail-
benefit from the commercial availability of low-cost Si
able for a given alignment of the detector array [see equa-
charge-coupled device (CCD) detector arrays and drive
tion (3)] is determined by the product of the total number of
electronics, as originally developed for mass market appli-
available detector pixels, Np , and the bandwidth per pixel,
cations such as video cameras, providing as many as
lp :
1024 pixels. Due to the high detectivity of the Si CCDs,
BW D Np lp 3
detector cooling is not needed. Applications of the mini-
ature UVvisNIR spectrometers include color analysis, To minimize the instrument cost and maximize the useful
fluorescence spectroscopy and plasma diagnostics. Exten- information output, the detector pixels can be selectively
sion of this technology to the IR may be problematic positioned along the output plane of the spectrometer.
due to atmospheric absorption, requiring hermetic sealing, Although this tends to narrow the range of applications
and the high dispersion in air that significantly compli- of the instrument, it creates new opportunities for low-
cates the design of a wide linear focal plane in miniature cost, dedicated spectrometers that can be imbedded within
spectrometers. equipment.
y
where Ai is the input aperture, and i is the input solid
angle of acceptance. Simply maximizing Li to collect as
much input light as possible is not fruitful for miniature
Grating line spectrometers due to the inherent physical size constraints.
(blank)
Rowland circle A better approach is to maximize the overall S/N with
x rp
rj respect to the selected optical source. The responsivity of
ai ao most IR detectors, and hence the resulting S/N, generally
Image
depends on the incident power density (W cm2 ) on the
Source point detector. Therefore, a better design approach for miniature
z spectrometers is to maximize the resulting optical power
density at the detector pixels. A second factor to consider,
Figure 3. Concave reflection grating geometry.
which distinguishes most diffractive spectrometers from
FT-IR spectrometers, is that the optical signal is focused
3.2 Optical layout onto the input aperture of the spectrometer. Therefore, Ai
should not be larger than the input focal spot, Af .
The layout of most miniature spectrometers is based on Reflective optics employed in telescopes, such as ellip-
a modified version of the Rowland14 geometry, as shown soidal reflectors, provide a focal spot that is typically about
in Figure 3. This employs a concave grating of radius Rg 5 mm in diameter in the IR. In the case of refractive optics,
that minimizes the number of optical components required. the attainable effective focal diameter is generally deter-
In the classic Rowland geometry, the output focal plane mined by the quality of the collimation of the optical signal
lies along the Rowland circle of radius Rg /2. This is not and aberrations, rather than diffraction effects. For well col-
suitable for coupling to a linear detector array. In the limated light, the focal diameter (blur circle) is given by
IOSPEC geometry, the optical layout and grating are equation (5):
d3
modified using proprietary optical design methodology to df D K B2 5
provide a relatively wide linear focal plane (up to 20 mm) f
about 1.2 mm outside the output face of the waveguide. where dB is the beam diameter, f is the focal length, and
This facilitates the close coupling of a linear detector K is a factor which depends on the refractive index of the
array at the output plane over a relatively broad spectral lens (K D 0.067 for n D 1.5, K D 0.018 for n D 2.4 and
range (2000 to 4000 nm, typically). IOSPEC employs a K D 0.0087 for n D 4). Using high-index optical material
near-Littrow mounting (ai 0 ) to minimize aberrations can significantly reduce aberrations and provide a tighter
and facilitate high spectral resolution. The optical input focal spot. Arc sources can be focused into the smallest
is generally defined by a precision miniature slit that is focal spots and are therefore the most efficient illumination
aligned and positioned at the input face of the waveguide sources for miniature spectrometers.
spectrometer. An IR optical fiber can be pigtailed directly to The input aperture is given by wi hw , where wi is
the optical input of the spectrometer using a suitable epoxy, the width of the input slit and hw is the height of the
as shown in Figure 1(a), or the optical signal can be coupled waveguide. For IOSPEC , there is no intrinsic limit to
directly into the waveguide through the miniature input slit. hw . The thickness of the slab waveguide can therefore
The injected signal diverges in the plane of the waveguide be selected to optimize the net S/N. The input aperture
according to the effective NA of the input fiber or lens, as width, wi , determines the effective instrument resolution,
reduced by the refractive index of the waveguide core, but is or bandwidth per pixel. The Rowland geometry facilitates
confined in the vertical direction by the slab waveguide. The a one-to-one image of the input spot size at the output
concave reflection grating collects the incident radiation, plane in the absence of aberrations. A thin rectangular
disperses it into its spectral components, and focuses it at slit, fabricated using laser machining, is placed between
the desired focal plane. the input fiber and the waveguide to define the output
spot size according to resolution requirements. The detector
pixel width is generally matched to the input slit width. In
3.3 Input optics more recent models of IOSPEC , a miniature variable input
slit is employed to facilitate operational trade-off between
The effective optical collection of a spectrometer is gen- resolution and S/N.
erally described by its input luminosity, as given by equa- The optical throughput can be improved through the
tion (4): use of a circular to rectangular aperture converter. One
Li D Ai i 4 method is to employ a bundle of smaller-core optical fibers
10 000
1000
dB km1
Input aperture Output aperture
100
10
1
0 1 2 3 4 5
(a) (b) / m
Figure 4. (a) Fiber-optic aperture converter, and (b) spectral characteristics of an IR fluoride fiber (after Le Verre Fluore).
as shown in Figure 4(a). The fiber bundle is arranged to and preferential wet etching to pattern the blazed grating
provide a circular aperture at the interface to the input opti- elements. After grating fabrication, the patterned Si wafers
cal signal. The fibers are arranged in a linear array at the are coated with Au by vacuum evaporation to provide
input slit to the spectrometer. Since the IOSPEC fabrica- a highly reflective surface in the IR. Subsequently, each
tion technology does not have any intrinsic limitations on processed wafer is diced into rectangular strips, each about
the waveguide core size, hw , fairly large fiber-optic bundles 2 by 20 mm in size, providing a large batch of gratings
can be accommodated to provide a relatively large input per wafer. Efficient gratings with periods below 3 m are
aperture. The main current limitation is the low packing achievable using this technique. Each resulting strip is
density of the IR fibers. The typical attenuation charac- aligned with the convex end of a polished slab waveguide,
teristics of IR fluoride fiber are shown in Figure 4(b).15 and affixed using an IR-transmissive epoxy. This provides
This fluoride fiber exhibits an attenuation that is below each spectrometer effectively with a master grating.
0.1 dB m1 to about 4 m, increasing to about 4 dB m1 The gratings are blazed for diffraction in first order
at 5 m. (m D 1). The blaze wavelength is determined by the crystal-
lographic orientation of the Si wafer, the grating period, and
the refractive index of the waveguide. Figure 5(a) shows an
3.4 Spectral dispersion and resolution SEM micrograph of a second generation grating, blazed for
3.6 m. The fabrication technique can yield blazed grating
In the hybrid IOSPEC design, the wavelength separating elements with optically smooth walls to minimize optical
elements are fabricated in thin Si wafers of the appropriate scattering. Figure 5(b) shows the corresponding measured
crystallographic orientation using a sequence of mask diffraction efficiency of a sample grating. The Si-based grat-
layer deposition, photolithography, reactive-ion etching, ings offer fairly broad spectral operation with an efficiency
Experimental
100
Theoretical
Diffraction efficiency (%)
80
60
40
20
0
2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
(a) (b) Wavelength / nm
Figure 5. (a) SEM micrograph of an echellette grating fabricated in 50 m thick Si, and (b) diffraction efficiency of a thin Si diffraction
grating blazed at 3.6 m.
exceeding 40% over most of the operating spectral range. peak at x D 4.46 mm, and the m D 3 peak at x D 6.42 mm
The main limitation to the grating efficiency is the linewidth are resolvable. The spectrometer exhibits relatively linear
of the photolithographic process employed to define the dispersion across the output plane with a value of about
grating elements. A new generation of gratings has been (l/x D 0.79 nm m1 that is in close agreement with
designed employing a 0.5 m linewidth. This has yielded theoretical calculations. The actual horizontal output spot
m D 1 peak diffraction efficiencies of about 85 to 90%. width, as measured using a 2-D IR camera, was about 50 m
The output dispersion angle ao is determined by the near 3.1 m (m D 2 peak).
grating equation10 [equation (6)]:
ml
D sinao sinai 6 3.5 Output focal plane and imaging
nTg
where m is the diffraction order, n is the refractive index The design of the output plane is critical to successful
of the waveguide core, Tg is the grating period and ai is integration with a linear detector array. In the Rowland
the angle of incidence. The channel bandwidth, lCH , is geometry, the output focal plane lies on a circle of radius
determined by the output spectral dispersion, (l/x), of Rg /2. Using optical design software, a model of the
the spectrometer, optical aberrations and the width of the grating and optical layout of IOSPEC has been developed.
detector pixel, wdet , such that [equation (7)]: This is employed to design the shape of the input and

l output planes of the IOSPEC spectrometer in order to
lCH D wdet 7 optimize the linearity of the output plane within the required
x
spectral range and to control the position of the focal
For a given grating pitch, the spectral resolution can plane. As indicated in Kruzelecky et al.,6 it is possible
be increased by increasing the grating radius of curva- to obtain linear focal planes over relatively wide spectral
ture (Rg ) and/or reducing the refractive index of the slab bandwidths using high-index slab waveguide structures.
waveguide. Figure 6 shows the output characteristics of For example, an output root mean square spot radius of
a moderate resolution spectrometer (model IOS4) with between 24 m and 30 m is feasible over the range of 1
Rg D 30 mm, as designed for the 1.5 to 5 m spectral to 5 m for a spectrometer design in ZnSe employing a
range. IOS4 was manufactured using a 7 m pitch grat- 50-m input slit and 7-m grating pitch. Finer focal spots
ing blazed at 3.6 m, with a 0.5 mm waveguide core size. are possible for Si-based slab waveguide spectrometers.
The waveguide was anti-reflection (AR) coated and pig- Depending on Rg , the usable output focal plane width can
tailed to a 320-m IR fluoride fiber (NA D 0.25) with exceed 20 mm.
an intermediate 50-m wide slit. Testing was conducted At the output of the spectrometer waveguide, a cylindri-
using a 1-mW laser diode operating at 1537 nm. The out- cal lens can be polished onto the output face to condense
put was scanned by linearly translating a 250-m core the output signal in the vertical plane. A factor of about 4
chalcogenide fiber (NA D 0.6) across the output plane. Its to 5 is possible. Therefore, for a waveguide of thickness
position was not optimized for focal spot size. In addi-
hw , the detector pixel height can be hw /4 or smaller and
tion to the main peak at x D 2.5 mm (m D 1), the m D 2
still collect most of the dispersed signal. This minimizes
the detector noise equivalent power (NEP) and maximizes
0.10 its responsivity.
Figure 7(a) shows the m D 1 output response of IOS35 to
a laser diode at 1537 nm. IOS35 was fabricated with a 1-mm
Detector signal (mW)
0.08
thick Si-based waveguide, a 4-m pitch grating and an
0.06 output dispersion of 0.15 nm m1 . The laser illumination
was expanded and focused using a f D 40 mm ZnSe lens
0.04 into the spectrometer through a 50-m wide slit. The output
signal was measured with a 256-channel PbSe array with
0.02 42-m wide pixels, spaced 12 m apart. The output imaging
is very good, with the peak full width at half maximum
0.00
(FWHM) less than two pixels wide. With the proper slit
2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 width, the spectral resolution can be equal to the pixel
Position / mm bandwidth. As shown in Figure 7(b), the residual signal
Figure 6. Output dispersion characteristics of IOS4 with Rg D beyond the main peak is less than 0.05% of the peak value,
30 mm, for a 7 m grating pitch. indicating very low stray light and internal scattering.
0.10
8
0.08
Detector output (V)
Detector output (V)

6
0.06
4
0.04
2 0.02
0 0.00
50 52 54 56 58 60 62 64 66 100 110 120 130 140 150 160 170 180
(a) Pixel number (b) Pixel number
Figure 7. (a) Output spectral scan of the m D 1 diffraction peak of a laser diode operating near 1537 nm, as measured by IOS35 with
a 50 m input slit using a 256-channel PbSe detector array, and (b) detector signal outside main peak.
3.6 Linear detector arrays the incident optical power, the detector optical absorption
coefficient [al], and the electrical characteristics of the
IOSPEC employs an electronically scanned detector array detector semiconductor material [carrier lifetime (t), intrin-
that can provide relatively high spectral scan rates, exceed- sic dark carrier concentration [ndark T], carrier mobility
ing several hundred complete scans per second, to facilitate (m), and the photocarrier generation quantum efficiency
real-time on-line analysis of a process. The elimination of (h)]. This determines the photo-induced change in the con-
moving components and integration of the optical system ductivity of the detector. Therefore, for photoconductive
provides more reliable long-term performance in non-ideal detectors of different areas, Adet , the generic output voltage
environments. The detector dark signal and its stability are responsivity (in units of volts per watt) can be repre-
important concerns for IR spectroscopy. IOSPEC employs sented by equation (9):
a proprietary detector configuration in which the detector Vdet Vbias Co l, T
dark signal is actively measured and dynamically corrected Vr l, T D D 9
Pin l Adet
during each scan.
In the current detector mode of operation, the detector where Vbias is the bias voltage applied to the detector, and
photocurrent is converted to a charge using a dedicated Co l, T is given by equation (10):
integration capacitor for each pixel. The photocurrent
charge is integrated in parallel for each pixel, and is m(ill) l a l
Co l, T D hlt 10
transferred sequentially to the high-speed preamplifier using mdark hc ndark T
a low-noise multiplexer. A multiplex advantage is gained Here, h is Plancks constant and c is the speed of light. This
by having the pixel integration time, tint , almost equal to equation indicates that as the area of the detector increases,
the complete scan time, tscan [equation (8)]: more total optical power is required to obtain the same
Np 1 detector output voltage.
tint D tscan 8 In order to determine the S/N, the various noise sources
Np
have to be considered. One source is the statistical or
A time period of tscan /Np is available to sequentially random variations in the source optical signal at a given
transfer the charge from a given pixel capacitor to the l, Pis l. A second noise source is the background optical
preamplifier. signal within the spectrometer itself due to scattered light
In the broad-band operation of a dispersive spectrometer, and thermal self-emission by the spectrometer optics,
it is possible for several diffraction orders to overlap. This PiB l. In the case of a very compact slab waveguide
can be eliminated through the use of a sequential, multi- spectrometer, the thermal self-emission can be reduced
element long-pass filter at the output plane, integrated with substantially by cooling the waveguide. One additional
the linear detector array, to provide first-order operation advantage of a guided-wave spectrometer is that stray or
over the entire spectral range. scattered light tends to be rapidly absorbed in the cladding.
The intrinsic detector responsivity is a function of the Therefore, it is much less of a factor than for bulk-optic
input optical power density, Pin l/Adet , where Pin l is spectrometers.
The detector itself contributes noise due to the statistical Vnet T, T D VTH T, T Vp T, T
nature of the generation and recombination of carriers
0.1VTH T, T 15
within semiconductors, as specified by its specific detecti-
vity D l. The corresponding NEP is given by equa- where VTH and Vp are of the form given by equa-
tion (11): tion (14). The ability to track the detector thermal fluctua-
Adet BWd 0.5 tions and to compensate for them is limited by the attainable
NEP D 11
D pixel uniformity, which is typically about 90 to 95%. The
where Adet is the area of the detector pixel and BWd is the active dark-signal correction circuit can also reduce the dark
bandwidth of the detector signal. In IOSPEC , the dwell background detector signal to below 5 mV at the preamp-
time per pixel, or pixel integration time, during a scan lifier input. This allows a relatively high, direct current
is approximately equal to the scan period, tscan D 1/fscan . coupled preamplifier gain (22 to 100) to be employed to
This parallel signal integration can offer a significant reduce noise coupling and digitization errors with weaker
S/N advantage over techniques that employ sequential signals.
measurements. Hence [equation (12)]: The various random noise sources are additive, resulting
in a net equivalent noise at the system input as given by
[Adet IOS fscan ]0.5 equation (16):
NEP(IOS) D 12
D
Vnp D
The NEP can be related to an equivalent voltage using 0.5
equation (13): V2DAQ
V2is C V2iB C V2TH C V2NEP C V2e C
Vbias Co l, T A2v
VNEP IOS D NEP(IOS) 13 16
Adet IOS
Here, VDAQ is the additional noise and digitization
A second major source of the detector noise is associated introduced by the data acquisition system, Ve is the
with the thermal stability of the detector itself. If the equivalent input noise of the electronics, and Av is the
detector temperature changes by T due to changes in system gain. Averaging the measured data over several
the ambient temperature or the temperature controller scans can reduce the contributions of random noise sources
regulation, there will be a corresponding variation in the to Vnp .
detector dark signal. This is exponentially dependent on the
energy bandgap, Eg , of the detector semiconductor material.
For a detector operating in the charge integration mode 4 REVIEW OF THE PERFORMANCE OF
with a pixel integration capacitor Cp and a detector pixel MINIATURE SPECTROMETERS
dark resistance Rd biased at Vbias , a change in the detector
temperature of T will result in a corresponding change in Miniaturization of the optical spectrometer system need
the detector dark output voltage [equation (14)]: not entail a reduction of its performance. The technologies
described in this article facilitate relatively large slit heights,
Vdet T, T D VTH T, T hw , typically 1 mm or greater, and a large NAi > 0.3

tint Eg T to provide relatively efficient input optical coupling. The
D Vbias 14
Rd Cp 2kB T T root mean square output focal spot width at the output
plane can be less than 60 m for a 50 m input slit. If
where kB is the Boltzmann constant and VTH T, T we assume a typical pixel width of wdet D 60 m and
is the effective detector noise due to thermal fluctuations. hw D 3 mm to capture the most intense part of the focal
For PbSe at 250 K, a detector temperature jitter of T D spot, then Li (IOS at hw D 3 mm D 0.00056 sr cm2 . The
0.001 K can result in a corresponding detector signal fluctu- internal transmittance, Tl , corresponding to the signal
ation of about 45 V, corresponding to an equivalent noise throughput from the input face of the waveguide to the
power of 2.3 1010 W. In contrast, the detector NEP (for detector, is typically about 0.2 to 0.4 for m D 1 diffraction
BWd D 1 Hz) is only about 1.5 1012 W. In the case of between 1.0 and 5 m (IOSPEC ) and between 8 and
the detectors employed with IOSPEC , we have designed 12 m.7 If we assume an incident power density Pi in
a proprietary pixel configuration that provides active tem- W nm1 sr1 cm2 , then the resulting power density at the
perature tracking to reduce the effects of VTH T, T by detector of the diffractive microspectrometer (mD) will be
providing a compensating voltage variation, Vp T, T, given by equation (17):
such that the net output voltage variation of a detector pixel,
Vnet T, T, is given by equation (15): Pd mD D Tl Li Pi lCH 17
with a corresponding detector voltage given by equa- must be inverse transformed. For 2N points, the inverse
tion (18): transformation requires at least 2N mathematical operations
[equation (22)]:
Co
Vdet mD D Vbias Li Tl Pi lCH 18

Adet 2p dM
Il D Vdet dM cos dM 22
l
Since the spectral data are measured directly, the corres-
ponding total noise per spectral point is simply the noise Each measured sample point, Vdet dM has a noise,
associated with one detector pixel, Vnp mD, as given by Vnp FT, associated with it. In the inversion, the noise is
equation (16). The resulting S/N is given by equation (19): additive as the square root of the sum of the squares of
the noise of a single data point, Vnp FT, such that the net
Vdet mD
S/NmD D 19 noise is given by equation (23):
Vnp mD
Vn (FT) D 2N0.5 Vnp (FT) 23
For diffractive spectrometers, the S/N can be improved by
increasing the input slit width to increase lCH . where 2N is the number of sample points in Fourier space.
FT-IR spectrometers offer the advantage of a much larger The corresponding S/N is given by equation (24):
input aperture than miniature diffractive spectrometers.4
Vdet FT
However, all optical systems have some form of limiting S/N(FT-IR) D 24
aperture. In the case of FT-IR, it is in the form of the limited Vn FT
input NA and the size of the detector required to accept the Since the detector in FT-IR sees the full interference signal,
output optical signal (typically about 1.5 mm square). In a the optical noise affecting the determination of Il is
FT-IR spectrometer,4 the input signal is split into two com- determined not only by the noise at l, Pin l, but also by
ponents by an IR beamsplitter. Assuming an ideal beam- the overall background noise in the input optical signal at
splitter, the internal transmittance of each beam is about the Fourier frequencies.
Tl D 0.5 0.955 D 0.37, relative to the input beam intensity. In the case of a typical sequential FT-IR measurement,
The two-beam interference pattern [equation (20)]: the dwell time per sample point is given by equation (25):

LFT tscan
Vdet dM D Vbias Co lTl (FT)lPi l tdwell FT D 25
Adet (FT) 2N

2pdM where tscan is the total period for one complete spectral
1 C cos dl 20
l scan from dM D 0 to dM D dM max and 2N is the total
number of sample points. At a given scan rate, the NEP of
is measured sequentially by a detector as a function the detector for a typical FT-IR system is larger than that
of the mirror spacing, dM , representing the optical path for a miniaturized diffractive spectrometer, as indicated by
difference in the interferometer. The important signal is the equation (26):
interferogram4 [equation (21)]:

0.5 Adet FT 0.5
LFT NEP(FT) D 2N NEPmD 26
VFT dM D Vbias Co lTl (FT)lPi l Adet mD
Adet (FT)
due to the larger detector area and the shorter dwell time
2pdM
cos dl 21 per sample.
l
Table 1 summarizes the relevant characteristics of mini-
as obtained by subtracting the constant term in equa- ature slab-waveguide spectrometers and compares them
tion (20) from the measured signal, Vdet dM . with those of a typical FT-IR instrument. While FT-IR
There are significant differences in the effective noise instruments have largely replaced bulk-optic dispersive
contributions for FT-IR and dispersive spectrometers. The spectrometers for most IR spectral measurements, the
FT-IR does not measure the spectral composition of the advent of high-speed linear detector arrays has created new
input optical signal directly, but rather a transform of opportunities for the miniature dispersive spectrometers.
that signal, as represented by VFT dM . By the Nyquist As shown in Table 1, the miniature spectrometers can
criterion, to determine N points in wavelength requires offer IR spectral performance approaching that of a typical
the sampling of 2N points in the frequency domain. FT-IR instrument. The calculation of the ideal S/N assumes
The FT-IR noise per measurement point corresponds to a that the system noise is determined by the detector NEP.
single point in the Fourier frequency space. In order to The total input optical power was taken as N Pi lCH
obtain the actual spectral information, the measured data to enable an approximate comparison of the S/N of a
Table 1. Summary of the characteristics of IOSPEC and typical FT-IR spectrometers.
Parameter IOSPEC FT-IR
Input aperture (Ai ) w hw D 0.006 hw cm2 5.07 cm2

Input numerical aperture (NAi ) 0.3 to 0.5 0.03
Input luminosity (sr cm2 ) 0.00056 (for h D 0.3 cm) 0.015
Internal transmittance 0.2 to 0.4 about 0.4
Intrinsic resolution (l/l) >1/5000 1/15 000
Adet 0.0375 mm2 1.75 mm2
Pixel bandwidth lCH (about 2 nm limit for m D 1 About 0.5 cm1 in wavenumber,
with 60 m slits) variable in wavelength
Detector responsivity (PbSe, 250 K) 2 105 V W1 4.3 103 V W1
NEP 0.0194 f0.5
s /D

0.132Nfs 0.5 /D
Optical noise Pis C PiB lCH Pis C PiB NlCH
Noise per sample point Vnp (IOS) Vnp FT-IR
Total noise (Vn ) Vnp (IOS) 2N0.5 Vnp (FT-IR)
Ideal S/N D Vdet /Vn 0.006D Pi lCH /f0.5
s 0.023D Pi lCH /f0.5
s
Scan rate >500 scans s1 <60 scans s1
FT-IR spectrometer with that of a miniature guided-wave 1.0

spectrometer. The calculation of the S/N for the miniature
dispersive spectrometer assumes typical values of NAi D 0.8
0.3 and hw D 0.3 cm.
Transmittance
0.6
5 APPLICATIONS OF MINIATURE
0.4
SPECTROMETERS
0.2
IR spectroscopy has a broad range of applications out-
side the laboratory that spans most human activities and
0.0
concerns; from agriculture, transportation, petrochemicals, 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200
plastics and pharmaceuticals to health and the environment. (a) Wavelength / nm
The potential applications are too numerous to list compre-
hensively in a short article, but can be seen in Volumes 4 0.95
and 5 of this Handbook. One example is the analysis of IR
transmissive materials such as plastics and polymers.
0.90
Figure 8 shows the NIR transmission spectra of poly-
Transmittance
carbonate and a plastic sheet employed to produce overhead

transparencies. The measurements were performed by pass- 0.85
ing collimated light from a 50 W quartz lamp through
the samples. The light was subsequently collected by a
f D 40 mm ZnSe lens and focused into IOS35 through a 0.80
50 m wide laser-machined slit. The resulting output signal
was detected using a 256-element PbSe array cooled to
253 K. The measurements were performed at a scan rate of 0.75
1200 1400 1600 1800 2000 2200 2400 2600
125 scans s1 , averaging over three successive scans.
Even with the PbSe detector array, uncoated ZnSe optics (b) Wavelength / nm
and quartz lamp, the S/N was typically greater than 2000 in
Figure 8. Transmittance spectra of two plastics as measured by
the 1.1 to 2.8 m range, allowing the resolution of relatively IOS35 with a 50-m input slit using a 256-channel PbSe detector
small changes in sample transmittance. With a PbS detector array at 253 K: (a) polycarbonate; and (b) overhead plastic-film
array, which is an order of magnitude more sensitive than transparency.
PbSe in the 1 to 3 m range, and a Xe arc source, which as PbS/PbSe or InAs/InSb can be employed to optimize the
facilitates a much smaller spot size than the quartz lamp, responsivity in the 1 to 5 m range. For detection of very
the S/N can be further improved by a factor of about 50. low thermal signals, optical chopping of the input signal can
There is sufficient spectral resolution to resolve the fine be employed to improve the attainable S/N.17 The compact
structure in the absorption peaks of the plastics. Since the size of the waveguide spectral radiometer is ideal for space
transmittance is exponentially dependent on the thickness applications such as atmospheric monitoring of greenhouse
of the sample, the thickness variations across a sheet of gases from a microsatellite, or planetary geological surveys
the plastic may be readily monitored. The detailed peak on a robotic rover.
structure is indicative of the chemical composition of the
plastic. This can be employed to monitor and adjust the
fabrication of the plastic to meet the required specifications. 6 CONCLUSIONS
In the recycling of plastics, sorting is required since certain
plastics, such as poly(vinyl chloride)s must be handled Technologies for slab waveguide spectrometers can yield
differently. This can be accomplished by differentiating the compact, yet relatively high-resolution IR spectrometers
spectral signatures of different plastics. that are ruggedized in a monolithic structure. Despite the
Spectral reflectance measurements can be used to probe scaling in the size of the input aperture for miniature spec-
the composition of materials that are not IR transmissive. trometers, there is a compensating increase in the detector
For diffuse surfaces, the typical reflectance can be only 5 responsivity and decrease in the detector noise to facilitate a
to 10%. This requires a high S/N (10 000) to resolve the net high S/N. Fiber-optic pigtailing of the input optics pro-
peak structure in the reflectance data. This is achievable vides long-term optical alignment under various operating
with IOSPEC using a xenon arc source. These sources conditions. The use of a solid-state optical waveguide elimi-
can be obtained with sapphire windows that facilitate nates the requirements for vacuum sealing or gas purging
strong illumination beyond 5 m. An example of field and provides new design opportunities. Integration with a
applications is the analysis of soil and rock samples. linear detector array facilitates high-speed spectral acquisi-
In oil sand specimens, the water bands near 1420 and tion, facilitating time-resolved measurements of processes.
1940 nm, hydrocarbon bands near 1730 nm and 2310 nm, The performance of the slab waveguide spectrometers is
and clay mineral absorption peaks near 2160 and 2205 nm becoming comparable to that of FT-IR spectrometry. More-
can be employed to grade the oil sand sites. In wood over, since the spectral information is measured directly,
processing, diffuse reflection can be employed to monitor no data transforms are required, providing more CPU time
the moisture content for optimal drying of the wood for data identification and correlation. The slab waveguide
using the characteristic water band near 1420 nm. The spectrometer modules, weighing less than 2 kg, offer signi-
cellulose band near 2000 nm can be employed for species ficant advantages in terms of size, weight and operational
differentiation in wood chip sorting. Using an optical switch simplicity.
and fiber-optic reflection probes, several points at a site can
be probed using a single spectrometer.
In agriculture, spectroscopy can be employed to mon- ACKNOWLEDGMENTS
itor the health of foliage and the ripeness of fruit16 in
order to optimize plant spraying and minimize the usage The authors would like to acknowledge the technical
of chemicals. Most of this work is currently in the vis- contributions of INO, including C. Tremblay, J. Lauzon
ible and short-wavelength NIR due to the availability of and C. Paquet, as well as the support of MPB Technologies
low-cost spectrometers. In transportation applications, IR Inc. and the assistance of Suzanne Goulbourne in the
spectroscopy can be employed to monitor the combustion preparation of the figures.
of an engine to optimize its performance and minimize fuel
consumption, or to check the quality of a lubricant. The
high scan rates, up to 1000 scans s1 , afforded by linear ABBREVIATIONS AND ACRONYMS
detector arrays facilitate time-resolved spectral measure-
ments of transient phenomena such as pyrotechnics and AR Anti-reflection
various chemical processes. ECR Electron Cyclotron Resonance
Waveguide spectrometers are also capable of providing MEMS Micro Electromechanical Systems
useful IR spectral radiometric measurements. Their com- NA Numerical Aperture
pact size allows the optics to be cooled to reduce internal NEP Noise Equivalent Power
thermal signals. Special detector array configurations such OSU Optical Spectrometer Unit
QWIP Quantum-well Intrinsic Photoconductor 6. R.V. Kruzelecky, A.K. Ghosh, C. Tremblay and C. Paquet,
rf Radio Frequency Proc. SPIE, 3416, 13 (1998).
RIE Reactive Ion Etching 7. S. Chadha, C. Stevenson and L.E. Curtiss, Proc. SPIE, 3533,
SEM Scanning Electron Microscopy 103 (1998).
S/N Signal-to-noise Ratio 8. P. Krippner, J. Mohr, C. Muller and C. van der Sel, Proc.
UVvis UltravioletVisible SPIE, 2783, 277 (1996).
WDM Wavelength-division Multiplexing 9. Ocean Optics, Data Sheets, Ocean Optics, Inc., Dunedin, FL.
10. M. Young, Optics and Lasers, Springer Series in Optical
Science, ed. T. Tamir, Springer-Verlag, New York (1984).
REFERENCES 11. M. Tabib-Azar, Microactuators: Electrical, Magnetic, Ther-
mal, Optical, Mechanical, Chemical & Smart Structures,
Kluwer Academic, Boston (1998).
1. G. Herzberg, Infrared and Raman Spectra of Polyatomic
Molecules, Van Nostrand, New Jersey (1964). 12. J.L. Cecchi, J.E. Stevens, R.L. Jarecki, Jr and Y.C. Huang, J.
Vac. Sci., B9, 318 (1991).
2. M.H. Brodsky, M. Cardona and J.J. Cuomo, Phys. Rev. B:
Solid State, 16, 3556 (1977). 13. E.S. Koteles, Crit. Rev., CR71, 3 (1999).
3. W.T. Welford, Aberration Theory of Gratings and Grating 14. H.W. Yen, H.R. Friedrich, R.J. Morrison and G.L. Tangonan,
Mountings, in Progress in Optics, ed. E. Wolf, North Opt. Lett., 6, 639 (1981).
Holland Publishing, Amsterdam, 243280, Vol. IV (1965). 15. Le Verre Fluore, Data Sheets, Le Verre Fluore, Campus Ker
4. G.A. Vanasse and H. Sakai, Fourier Spectroscopy, in Lann, Brittany.
Progress in Optics, ed. E. Wolf, North Holland Publishing, 16. C.H. Blazquez, H.N. Nigg, L.E. Hedley, L.E. Ramos, R.W.
Amsterdam, 261327, Vol. VI (1967). Sorrell and S.E. Simpson, Proc. SPIE, 2744, 673 (1996).
5. R.V. Kruzelecky, S. Paquet, A.K. Ghosh, C. Tremblay, 17. R.V. Kruzelecky and A.K. Ghosh, Proc. SPIE, 4205, 25
J. Lauzon and N. Landry, Proc. SPIE, 2744, 684 (1996). (2000).
Tunable Filter and Discrete Filter Near-infrared
Spectrometers
David L. Wetzel, Arnold J. Eilert and Joseph A. Sweat
Kansas State University, Manhattan, KS, USA
1 NEAR-INFRARED FILTER acousto-optic TFS finally has become established as a com-

SPECTROMETERS petitive means of spectroscopy, at least in the NIR region
of the spectrum. Although crystals for performing acousto-
Filter spectrometers have traditionally served well for rou- optic TFS have been produced for the ultraviolet (UV),
visible, and mid-infrared regions of the spectrum, the utility
tine analysis for everyday use in industry. Even now prob-
of this technique is perhaps best realized in the NIR region.
ably 85% of near-infrared (NIR) instruments in use are
In the UV and visible regions, heat dissipation is required.
equipped with discrete multilayer interference filters to per-
Time delays required for cooling are inconvenient, and
form wavelength selection. For NIR, multiple filters are
competitive means such as silicon photodiode arrays were
required. Additionally, circular or linear variable interfer-
detrimental to early development of acousto-optic tunable
ence filters have been available for some time. Tunable
filter (AOTF) for this spectra region. It is reported that at
filters perform the same spectroscopic function as the inter-
the time when the first UV diode array detectors were being
ference filters. Tunable filters using acousto-optic or liquid
produced by Hewlett Packard for high-performance liquid
crystal technology require no moving parts, but allow pro-
chromatography (HPLC), a technique internally referred to
grammed random electronic wavelength access switching
as ATOF was considered and rejected in favor of the diode
that is useful for either spectroscopy or imaging. In this arti-
array for that purpose. In the mid-infrared region of the
cle we describe tunable filter spectrometers first, followed
spectrum, a ternary (Tl3 AsSe3 ) crystal transmitted radiation
by the more traditional discrete filter systems that require
and exhibited the acousto-optic effect. The disadvantage of
sequential physical interpositioning of the filter between the
this particular material is that it has a very low thermal
source and the sample (or alternately prior to the detector).
conductivity, and the energy that was input to tune the fil-
ter took time to dissipate, and, therefore, could be operated
only in a pulsed mode. The pulsed operation resulted in
2 ACOUSTO-OPTIC TUNABLE FILTERS
a very low-duty cycle, which was a disadvantage of that
solid-state device. Tellurium dioxide (TeO2 ) transmits out
2.1 Introduction
to the 3.5 m range and is quite useful in the NIR region
at shorter wavelengths.
Tunable filters provide a way of changing wavelengths for
either spectroscopy or spectroscopic imaging. Electronic
wavelength switching is an important feature of tunable 2.2 Background and theory
filters. Tunable filter spectroscopy (TFS) can be performed
with switching caused by the acousto-optic effect or the Interactions of electromagnetic radiation with acoustic
use of liquid crystals. After a slow and faltering start, waves in an optical medium (photonphonon interaction)
commonly referred to as acousto-optic interactions were
John Wiley & Sons Ltd, 2002. demonstrated by Debye and Sears.1 These interactions were
limited to low-frequency acoustic waves in liquids and band of optical frequencies (or wavelengths) that maintain
gases with incoherent light sources. More recently, acousto- proper phase-matching conditions. The majority of incident
optic interactions with crystalline materials as the optical broadband radiation passes through the crystal unaffected.
medium were explored.2 At the present time, devices in This selective diffraction occurs cumulatively as the radi-
everyday use such as laser printers depend upon the inter- ation traverses the acoustically perturbed optical medium.
action of radio frequency (RF) with light waves to produce The radiation affected by the diffraction process thus is
acousto-optic modulation of the light beam being used.3 7 separated from all nonaffected radiation. This separation is
In contrast to an acousto-optic modulator, the changing of possible by utilizing changes in the polarization orientation
wavelength transmitted with an AOTF is more complicated. as well as the deflection angle that takes place as a result of
Harris et al. used LiNbO3 and CaMoO4 crystals with visi- selective diffraction. The acousto-optic interaction is wave-
ble light.8 11 Investigations were extended with appropriate length restrictive in combination with electronic control of
crystals from the UV through the mid-infrared.10 21 which wavelengths are affected. By choosing the acoustic
Modern usage of AOTFs, unlike other acousto-optic frequency input, tunable acousto-optic radiation filtering is
devices, is dependent on the special case of acoustic diffrac- accomplished. Such an acousto-optic device can be used as
tion in an optically anisotropic medium. First described by the wavelength-selective component (monochromator) that
Dixon in 1967,22 an AOTF is produced by starting with a will behave essentially as an optical bandpass filter with
proper crystal of birefringent material. After the axes of the a spectral bandpass that can be electronically tuned very
crystal (using X-ray techniques) are accurately determined, rapidly over relatively large spectral regions.
it is properly cleaved or cut, and a piezoelectric transducer The acoustic wave consists of a sinusoidal perturbation
is bonded to a predetermined face. Opposite the transducer of the density of the material that travels at the velocity
is an RF absorber. The tunable filter in operation accepts of sound. Changes in the density of the medium result
RF input of appropriate frequency, in the megahertz range, in approximately proportional changes in the index of
with regard to the desired tuned output optical wavelength. refraction. The moving phase grating that results diffracts
The incident RF is transmitted by way of the transducer, portions of an incident radiation beam. In a particular
which acts as a piston source driver generating an acous- crystalline device, for a given acoustic frequency, only a
tic wave in the crystal at the transducercrystal interface limited band of optical frequencies can satisfy the phase-
that propagates through the bulk of the crystal. An acous- matching condition or momentum conservation relation for
tic absorber opposite at the termination face of the crystal the photonphonon particle theory and be cumulatively
minimizes unwanted reflection of the RF wave. diffracted (Bragg diffraction). As the frequency of the RF
Compression and rarefaction that occur as the longitu- acoustic wave in the crystal is changed, the band of optical
dinal sound wave propagates causes periodic differences frequencies that satisfy the phase-matching condition is
in the refractive index within the crystalline material. shifted. Consider the sound phonons with a momentum
Figure 1 schematically shows the uniform refractive indices equal to the product of a propagation wave vector k and
across the crystal without RF and those with RF hav- Plancks constant h. Diffracted wavelengths are selected by
ing designated alternate rarefaction and condensations. The acousto-optic interaction within an AOTF in order to satisfy
efficiency of such a process is dependent on the magni- momentum conservation requirements between incident ki
tude of the n (refractive index difference) produced in and diffracted kd photon wave vectors and the phonon (or
a particular device. The periodic differences in the refrac- acoustic) wave vector ka . This requires that the momentum
tive index within the crystalline material act as a moving of the colliding particles hki C ka equals the momentum
phase grating that will selectively diffract only a narrow of the diffracted photon hkd .23,24 This momentum-matching
condition can be expressed as:
n
n n n kd D ki C k a 1
n 2pnd
n n + n where kd D 2
n l
n n n
n 2pni
n n + n and ki D 3
n l
(a) (b) where nd and ni are the indices of refraction for the incident
and diffracted radiation, and l is the wavelength. Because
Figure 1. Schematic diagram showing (a) uniform index of
refraction with an unperturbed crystalline lattice and (b) periodic the energy of the acoustic waves is much smaller than
differences in the index of refraction that occur when an acoustic the energy of the photons involved in the acousto-optic
wave is propagating through the lattice. interactions, kd and ki are essentially equal in optically
Tunable Filter and Discrete Filter Near-infrared Spectrometers 3
isotropic media (ni D nd ). However, this is not the case The advantages of collinear devices is that they will include
with acousto-optic interactions occurring in an optically relatively large acceptance angles and produce high spectral
anisotropic medium. resolution.27 The disadvantage is that only uniaxial crystals
When electromagnetic radiation is diffracted by acoustic with both incident and diffracted radiation and a plane
shear waves, the displacement of matter is perpendicular to normal to the optic axis of the crystal can be used.28,29
the direction of propagation. In most cases, the interaction Noncollinear configurations are used most often for prac-
will result in a 90 change in the polarization of the tical AOTF devices. In these, the incident and diffracted
diffracted wave. The shear wave disturbance causes this wave vectors are noncollinearly aligned, and the acoustic
polarization rotation by inducing a birefringence that acts and optical waves are propagated in quite different direc-
on the incident radiation as a birefringent plate, thus rotating tions. If n represents birefringence, then the relationship
the plane of polarization. What occurs is that radiation between the acoustic frequency and optical angle of inci-
propagating as an extraordinary ray (e-ray) is converted into dence qi is given by the following equation:
an ordinary ray (o-ray) by diffraction, upon interaction with

the acoustic wave. In an optically anisotropic medium, the vs n 1 4 1
fa D sin qi C sin2 2qi 2 9
indices of refraction for the incident ni and the diffracted lo lo
nd waves, therefore, are unequal, resulting in a significant
change in momentum and wavelength.23,24 The associated With the noncollinear interaction, the diffraction results
wave vectors that occur when the incident radiation is an in an angular deflection in addition to polarization rota-
e-ray and the diffracted radiation is an o-ray are listed tion.30,31 This is an added advantage of noncollinear inter-
below. action, because the angular separation that occurs between
2pni the diffracted and nondiffracted radiation upon exiting
ki D 4 the crystal provides a means for selective radiation fil-
lo
tering. Figure 2 shows the separation of the ordinary ray,
2pnd the extraordinary ray, and the zero ray in a noncollinear
kd D 5
lo anisotropic AOTF device.
The propagation angles of incident qi and diffracted qd The conditions that must be met by the solid-state mate-
radiation are unequal and are related to the acoustic fre- rial to be used in an acousto-optic device are that it must
quency fa , the indices of refraction for the incident and be birefringent and transparent optically in the region for
diffracted waves, the diffracted radiation wavelength, and which it is to be used and have a good acousto-optic figure
the acoustic velocity vs by the following relationship: of merit. This latter item refers to the ability to produce a
2 maximum localized change in the index of refraction (n)
1 l o fa vs 2 2 from a minimum input of RF energy. The acousto-optical
sinqi D ni 1C ni nd 6
2 vs l o fa figure of merit, M, is related to the index of refraction, the
photoelastic constant (p), the density (r), and the sound
and velocity (vs ) by the following expression:
2
1 l o fa vs
sinqd D nd 1C n2i n2d 7 n 6 p2 r
2 vs l o fa MD 10
v3s
Alteration of the momentum-matching condition is achieved
in the tuning process. The acoustic frequency is changed by Tuned ordinary
varying the frequency at which the piezoelectric transducer Untuned
Unpolarized
Tuned extraordinary
is driven. Only radiation at a specific wavelength will be input
diffracted from the input of a particular acoustic frequency
to a particular crystal. The acousto-optic diffraction process Untuned
and the AOTF devices can involve collinear or noncollinear Horizontally Tuned extraordinary
polarized input
interaction of incident radiation, diffracted radiation, and
acoustic waves.24 26 In the case where all three wave vec- Tuned ordinary
tors ki , kd , and ka involve collinear interactions (when Untuned
Vertically
qi D 90 and qd D 90 ), the acoustic frequency expression polarized input
for collinear diffraction is as follows:
[vs ni nd ] Figure 2. Separation of tuned versus untuned rays as a function
fa D 8 of polarization upon exiting an AOTF.
lo
Another very important characteristic for the solid-state The relationship between the optical and acoustic frequen-
material is that because energy is being pumped into the cies is dependent on the architecture and composition of
crystal, it must have a good thermal conductivity so that the particular solid-state device. This means that in addition
the heat that is generated can be dissipated. One of the first to sweeping through a range of wavelengths, AOTF-based
materials tried commercially for the mid-infrared region instrumentation can be used to access only specific wave-
was thallium arsenic selenide (Tl3 AsSe3 ). This material lengths of interest quickly and in any desired sequence.
functions throughout the mid-infrared spectrum from 2.5 to For instance, during development of a calibration, informa-
17 m. Poor thermal characteristics prevented continuous tion over the entire spectral range of the instrument can
acoustic wave operation, and, thus, pulsed operation was be collected. Once the calibration is developed, acquisition
required to accommodate high power input.16 The Westing- parameters can be optimized for that specific calibration.
house Company put a lot of effort into such a device to be With acousto-optic TFS the instrument can be tailored to
used as an industrial stack monitor, which appears to have suit a particular analytical problem without making any
very high performance. Regrettably, regulatory agencies in hardware changes to the system.
the USA required periodic calibration of this single-beam Extremely high sample throughput rates are achievable
device in the field, and this prohibitively costly process pre- based on this design, and the potential of this technology
cluded its practical usage, even though the system was a is rivaled only by multisource or multidetector (photodiode
technical success. Throughout the entire visible and NIR array) techniques. The design provides another factor that
and into part of the mid-infrared, TeO2 is an excellent may be used to further increase the duty cycle of the spec-
medium for noncollinear acousto-optic interactions. It has trometer. This factor is optimization of the percentage of
an exceptionally high acousto-optical figure of merit and time that the instrument is engaged in obtaining the most
good thermal characteristics. It is most often regarded as spectroscopically useful information per sampling cycle.
the material of choice for these spectral regions.32 34 It is Photometric data processing from monochromatic radiation
not usable at wavelengths shorter than 350 nm or longer detection is relatively simple and low in time consumption.
than 4.5 m. Crystalline quartz works for UV spectroscopy Because no Fourier or Hadamard mathematical transforma-
and is expected to be utilized for fluorescence and in the tions need to be performed, the delay time required between
entertainment industry. acquiring data for a given sampling cycle and subsequent
reporting of analytical results based on that data is minimal.
2.3 Features of an acousto-optic tunable filter Unlike most conventional spectrometers, acousto-optic
spectrometer TFS instrumentation can access wavelengths throughout its
entire spectral range without requiring mechanical motion.
Spectrometry with an AOTF monochromator is achieved When necessary, intensity modulation can be achieved elec-
with electronic wavelength switching. Switching speed, tronically, either by amplitude or frequency modulation of
wavelength precision, accuracy, and resolution are excel- the RF signal, instead of through the use of a mechanical
lent. The dynamic range of operation is good, and it optical chopper. This makes it possible for the system to
offers stability, efficiency, durability, and a reasonable cost. be very rugged, durable, and reliable. It is also exceedingly
Unlike a grating monochromator, no slit is required for this compact and has been referred to as the monochromator
filtering process; thus, the high throughput of an efficiently in a matchbox, because the AOTF device is housed in a
operated system has the ability to provide a favorable metal case that is the size of a typical matchbox. The wave-
signal-to-noise ratio (S/N). Because it is not necessary to length access is unaffected by vibration or physical shock,
sweep through a series of wavelengths to get from one and physical wear is avoided completely. The only other
wavelength to another, the random wavelength access and types of spectroscopic instrumentation in use that can truly
the instantaneous switching between wavelengths provide claim an advantage of no moving parts involve either mul-
a high-duty cycle operation in the continuous wave mode. tiple electronically switched sources (light-emitting diodes,
Until recent fast photovoltaic detectors became available, LEDs), selective masking of polychromatic radiation with
the electronic wavelength switching of the monochromator multiple solid state (liquid crystals), beamshutters (with
greatly exceeded the detection speed. Hadamard transformation), or a post-sample grating poly-
The wavelength access is flexible, and programmable chromator diode array with parallel multichannel detection.
random access allows high sampling rates. Tuning of the Acoutsto-optic TFS has an advantage over multiple
optical frequency of the filter is dependent upon the ultra- source techniques in that the number of sources that can
sonic RF input into the device. The tuning RF is from either be used typically is limited, giving only discrete or low-
an oscillator (voltage controlled by a computer via digital- resolution wavelength coverage, whereas with this tech-
to-analog (D/A) converter) or from a digital RF synthesizer. nique, continuous coverage is furnished over its entire
spectral range. For the grating polychromator diode array, rejection characteristics make it possible to start with an
the intensity of radiation at each specific wavelength is slit- extremely intense source of radiation in order to have a
restricted, whereas the slit is not used in this technique. high signal without degradation in filtering characteristics
With each of the other techniques mentioned, the max- over time.
imum number of distinct wavelength windows available Acousto-optic TFS provides a combination of flexible
ultimately is limited by the total number of source, shutter, wavelength access and intense illumination that can be used
or detection elements that can be accessed independently, not only for clear transmitting samples, but for rapid analy-
whereas acousto-optic TFS is dependent upon the frequency sis of samples of all types, including scattering samples. A
switching resolution of the RF signal generator over the variety of radiation collecting devices can be incorporated
operational frequency range of the device. after the sample to collect transmitted, scattered, or reflected
The optical wavelength precision of acousto-optic TFS radiation. Further signal enhancement could be realized
exceeds wavelength reproducibility of a typical mechani- through the use of multiple bandpass filtering. Multiplexing
cally positioned diffraction grating monochromator and can capability could be achieved by construction of an AOTF
rival or surpass any other existing technologies (including device with multiple piezoelectric transducers, each with its
Fourier transform interferometry with laser fringe track- own electronic driver. Independent control of these drivers
ing of mirror position). Electronic frequency synthesizers would permit simultaneous selection of multiple wave-
with high (24-bit) digital resolution allow extremely minute lengths. In the NIR region of the spectrum, this approach
adjustments of the passband peak wavelength exits. Thus, has had very little exploration. It is reasonable to assume
TFS instruments can achieve spectroscopic analysis based that such multiplexing may provide an advantage in the case
on subtle wavelength shifts that instruments with impre- of fluorescence for use of multiple excitation wavelengths.
cise mechanically operated wavelength selection cannot An important characteristic of the tuned radiation
perform. The precision achievable allows valid spectral throughput of AOTF devices is the image integrity that is
subtraction procedures to be performed. Long-term wave- preserved even throughout the filtering process. The optical
length precision can be achieved with the acousto-optic TFS characteristics with the tunable filter are similar to those that
with confidence, avoiding the necessity of continual wave- would be observed with a fixed-wavelength interference
length verification over time as is commonly done with filter. Two-dimensional wavelength-specific imaging of a
monochromators. broadband source illuminated object can then be recorded
Throughput characteristics of AOTF devices are remark- readily on image detectors such as a charge-injection or a
able, based on their optical efficiency, the apertures achiev- charge-coupled device (CCD). This capability is simply not
able, and the avoidance of the restriction of a slit. The achievable with a grating monochromator system.
architecture of the solid-state device used determines the
aperture. A 1 cm2 aperture is readily feasible. Not only
2.4 A typical acousto-optic tunable filter
can large aperture crystals be used, but with certain appro-
spectroscopy scheme
priate designs, large acceptance angles (as high as 20
half-angle for a noncollinear device) can be achieved. Thus,
Figure 3 shows the optical diagram of the monochromator
throughput from the combination of a large solid angle of
portion of the spectrometer. In this particular case, polariz-
acceptance and a high input aperture typically can be quite
ers are included. (It is acknowledged that in most commer-
high. The efficiency is wavelength dependent, and, thus,
cial acousto-optic TFS systems, crossed polarizers are elim-
the device must be designed for optimization in the region
inated as a cost-cutting feature.) Rotation of the polarized
desired. Peak efficiencies exceeding 90% are achievable,
plane of the refracted wave in an anisotropic acousto-optic
which means that over 90% of the incident radiation at the
diffraction results, and in theory, polarization orientation
tuned wavelength can be utilized by the spectrometer, and
alone could be used for selection of acousto-optic radiation
diffraction efficiencies can be kept well above 50% over a
large spectral range. Thus, even with the use of a polarizer RF
reducing the incident radiation by 50%, the AOTF efficien- Polarizer 1 input Polarizer 2
cies are comparable to those typically achieved with simple Tuned
interference filters, i.e. 2050%. output
Rejection of unwanted radiation is not a problem. It Untuned AOTF
can be removed from a broadband source, either by spa- unpolarized
tial separation of the tuned and untuned components or by input Blocker
using polarizers based upon polarization-dependent trans- Figure 3. Diagram of an optical design for the monochromator
mission/reflection of the tuned/untuned radiation. These portion of an AOTF spectrometer.
Micro-
Digital RF computer, A/D
RF signal
frequency timer, and converter
amplifier control
synthesizer
hardware
Regulated
power
supply
AOTF
Amplifier
monochromator
Source Collimating Sample Focusing Detector

optics optics
Figure 4. Block diagram of the overall layout of an AOTF spectrometer.
filtering. When two polarizers that are orthogonal to each Figure 4 shows a block diagram of the electronic com-
other are placed before and after the AOTF crystal, and they ponents, in addition to the optical components. In Figure 3
are in correct orientation with respect to the geometry of the AOTF monochromator block in the figure symbolizes
the AOTF device, then only a select band of diffracted (or the polarizers as well as the AOTF device itself. Figure 4
tuned) radiation that corresponds to the acoustic frequency is a composite block diagram that could stand for either
will remain. The success of this means of selection depends one involving modulation or one involving direct current
on conformity to ideality of operation of the different com- (DC) operation. It also could be described as one in which
ponents. Ideality requires that the polarizers have high a voltage-ramp-driven oscillator would provide the RF or
transmission characteristics, and that the crossed polarizers a digital frequency synthesizer would supply the RF. A
exhibit excellent extinction efficiency throughout the spec- microcomputer controls the RF source. Alternatively, this
tral range of concern with the particular AOTF device. may be a digital frequency synthesizer accepting a digital
Ideality also requires high efficiency in terms of the redirec- signal directly from the microcomputer (central processing
tion of unfiltered radiation by the AOTF device. If sufficient unit, CPU), or an oscillator driven by voltage input coming
deflection is caused by noncollinear acousto-optic diffrac- from a D/A converter and driven by the CPU. In either case,
tion of radiation, spatial separation may be an adequate RF is input through the piezoelectric transducer into the
method of rejecting untuned radiation. AOTF crystal to provide the tuning. AOTF devices obtained
For commercial instruments in which polarizers are elim- independent of the driver must have their own response
inated for cost reduction, separation in space will allow curve provided to determine what input RF is required to
both the ordinary ray and extraordinary ray to be used in produce a particular optical filtering transmission.
a double-beam mode. The configuration shown in Figure 3 The source is a quartz tungsten halogen lamp of approx-
illustrates a case where both the spatial and polarization imately 20 W. In the authors laboratory (Kansas State
means of optimal rejection of untuned radiation are used. University, USA), an enhanced source was utilized wherein
It should be pointed out that spatial separation is not a nearly 4p spherical reflector returned radiation going in
an option for collinear acousto-optic diffraction; therefore, other directions back to the filament to improve the ratio
cross-polarization techniques must be relied on for elimi- of photons delivered per unit heat generated from the resis-
nation of untuned radiation. In Figure 3, note the radiation tance source. Radiation from the source was collected and
source, collimation lens system, the polarizer, the AOTF, collimated to direct it toward the polarizer and throughout
the second polarizer (with the deflected ray coming out of the rest of the optical system. The entrance to the AOTF
the polarizer at an angle different from the incident ray), a device was masked to avoid overfilling the crystal.
sample device, and then a detector. Note the RF input to The detector and amplifier used were dependent on
the AOTF device from an oscillator or frequency synthe- whether modulation was required for a lead sulfide detector
sizer by way of a piezoelectric transducer. Polarizer 1 and or DC as used with a photodiode such as InGaAs. Using
polarizer 2 are shown in orthogonal orientations. the lead sulfide required a lock-in amplifier, which then
was utilized between the lead sulfide detector circuit and

the A/D (analog-to-digital) converter that was the input to
2.0
the CPU. When a detector that requires modulation was
used a modulated source was needed. The 1986 experi-
mental instrument from which quantitative data were first 1.6
Absorbance
reported35 used a mechanical chopper. Subsequently in
order to avoid a moving mechanical component modula- 1.2
tion was accomplished by alternately detuning to the point
at which no diffraction occurred and tuning it to a particular 0.8
desirable wavelength to provide alternating current (AC)
operation.36,37 Alternatively, for use with a photovoltaic
detector in a DC mode, a simplified operational amplifier 0.4
was all that was required between the circuitry of the detec-
tor and the A/D converter input to the CPU. 0.0
Quantitative analysis in the NIR region of the spectrum is 800 1200 1600 2000 2400
based on small differences of a few milliabsorbance units; Wavelength / nm
therefore, limiting the noise level to a few microabsorbance Figure 5. Series of spectra obtained from monitoring a linear
units is necessary. To make analytical decisions based binary gradient going from 100% cyclohexane to 100% benzene.
on these small photometric differences requires not only
stability, but also referencing to provide a background for industrial online monitoring. Figure 5 shows transitions
spectrum for ratioing. With the exception of instruments occurring in the chemical monitoring of one chemical sub-
of a double-beam design, single-beam operation usually stance being replaced by a second chemical substance as the
entails collecting the reference spectrum just prior to the flow of two different pumps is controlled to make the tran-
actual analysis. For continuously operating flow detectors sition from 100% fluid A (cyclohexane) to 100% fluid B
or process monitoring devices, the dual-wavelength spectro- (benzene). This figure shows that the absorbing species
meter concept generally is used (in which a measurement at (shown in the absorbance spectrum) characteristic of the
a wavelength where no change occurs serves as an internal initial fluid diminishes throughout the process, whereas the
standard) to observe changes in intensity at the wavelengths bands characteristic of the substituted fluid increase. The
that are being monitored as a function of time or as function spectra taken on this figure required 200 ms each, with the
of flow in a processing stream. This also may be used to entire transition process of 100% A to 100% B taking place
observe differences in individual specimen units, such as over a period of 2 min.3639 As an online process monitor,
pharmaceutical tablets, polymer resin granules, or single not only is this used in large-scale operations found in a
seeds of plant material being analyzed for inspection or chemical plant, but it serves also for small-scale flow as
sorting purposes. found in an extraction process, in chromatography, or in
monitoring a chemical reaction occurring even under high
pressure.
2.5 Operation of an acousto-optic TFS
This discussion has stressed that in addition to the versatility 2.6 Performance of an acousto-optic TFS
of having a whole field of wavelengths that can be selected, designed for quantitative analysis
many scans of a selected region of the spectrum that is rich
in information can be taken, while avoiding parts of the From the preceding section it is evident that the acousto-
spectrum that provide only noise. Such operation allows optic TFS has great potential for online monitoring. Fig-
many opportunities and advantages. To take advantage of ure 5 shows the changes as a function of time in a binary
the speed of electronic wavelength switching that is possible system. Before discussing performance in an online func-
with the acousto-optic TFS, an equally high-speed detector tion, let us consider the aspect of wavelength reproduci-
such as the InGaAs was utilized. Limitations for speed then bility. Because the solid-state filter is digitally controlled,
were simply A/D conversion (with a sufficient number of and because for a given filter the optical wavelengths
bits to do justice to the analytical purposes), the computer are reproducibly controlled from the input of RF oscil-
inputoutput (I/O) capability, and the processor speed. Fast lation, wavelength reproducibility is not a problem. A
computation makes real-time analysis possible with the simple test used to demonstrate wavelength reproducibil-
output of real-time results and automated decision-making ity is whether or not the spectra obtained were worthy of
capability included. The ruggedness of the system is ideal spectral subtraction. Figure 6 shows two spectra. Spectrum
0.7 quantitative analysis. Because the operation of the AOTF

A instrument is software controlled, the measurement param-
0.6
A
eters can be readily changed interactively depending on the
0.5 resulting optical effect of either the sample or the absorption
characteristics of the optics of the instrument itself. On any
Absorbance
0.4
NIR instrument, the region with the weakest S/N is between
0.3 2400 and 2500 nm, because the black body radiation inten-
sity of the tungsten filament declines somewhat in that
0.2
region. In nearly all grating monochromator instruments,
B
0.1 B diode array grating polychromator instruments, and dis-
crete filter instruments, the amount of time spent taking
0.0 data at any one wavelength is constant. When the S/N is
1200 1400 1600 1800 2000 2200 2400
insufficient for quantitation from absorption bands being
Tuned wavelength / m
observed, extending the time of measurement at all wave-
Figure 6. Demonstration of spectral subtraction capability of lengths to allow coaddition of a large number of successive
AOTF spectrometer showing (A) a benzene spectrum and (B)
a 50 : 50 mixture of benzene and toluene with the toluene contri-
scans is often impractical. Custom or interactive data acqui-
bution removed by spectral subtraction. sition is easily programmed in the software-controlled TFS
instrument, and random wavelength access operation allows
A is that of 100% benzene obtained with the acousto-optic the instrument to take data only at wavelengths of inter-
TFS. Spectrum B was obtained by recording the spectrum est. Where energy (observed in a single-beam spectrum) is
of a 50 : 50 mixture of benzene and toluene and from that adequate at the reference wavelength and at the indicator
spectrum, subtracting the spectrum of pure toluene until (analytical) wavelength, there is no problem with getting
only the spectrum of benzene remained. The alignment of good quantitative data. If in fact the signal at a particular
peaks from the 100% benzene and from the subtracted spec- region of the spectrum of interest is very low, either because
trum is excellent. It is evident also from the figure that to of absorption of some optical component in the system, a
a first approximation Beers law holds and that the absorp- low intensity of the source, or low sensitivity of the detec-
tion bands of the 50% benzene are approximately half those tor, there may be a problem. Isophotonic data acquisiton
of the 100% benzene.37,39 The wavelength reproducibility provides an advantage. Isophotonic operation of the random
of this type of instrument is superior to that of nearly any wavelength access instrument means that when the filter is
grating monochromator system, most of which require mul- tuned to a particular wavelength, data will be taken at that
tiple scans and averaging or insertion of a standard between wavelength until a certain number of data registers have
scans to make correction possible. Wavelength reproducibil- been filled. The instrument then steps to the next wave-
ity of the AOTF-based spectrometer rivals that of Fourier length, which may have a higher or lower intensity, and
transform near-infrared (FT-NIR), which is assured by trac- accumulates data for a sufficient time to fill an equal num-
ing the mirror movement with a heliumneon laser to allow ber of data registers. As a result, averaging reduces the noise
correction or provide a tilt message. as the square root of the measurement time while accumu-
Electronic wavelength switching occurs very rapidly. lating the signal in a region of the spectrum where the S/N
When coupled with a rapid and highly sensitive detector, can use enhancement.40 The same system will avoid wast-
the speed of electronic wavelength switching is used to ing unnecessary time by accumulating excessive data where
advantage to perform high-speed analysis. The response to noise reduction is not necessary. A specific time could be
transients was tested by injecting different materials into a programmed for a particular analytical problem to take full
flowing stream or by digitally driven pumping of different advantage of the square root function of the noise in com-
chemicals simultaneously, as was done for Figure 5. To parison to the linear function for the signal. Experience has
further challenge the capability of the AOTF instrument to shown that a significant benefit occurs when the analyti-
observe transients, a wheel of acetate with single, double, cal wavelength and the reference wavelength are both in
or triple thicknesses was rotated in the beam with a motor. a region of the spectrum where there is considerable noise
It was possible to obtain 480 two-wavelength analyses in and a limited signal.34,37,40 A lesser benefit in precision
3 s, and the average analysis time was 8.3 ms.37 39 This from isophotonic operation occurs when only one of these
illustrates the potential for online analysis in a stream whose wavelengths has a shortage of signal.
composition is changing rapidly. A high-duty cycle is another feature that is achieved
Another feature that allows maximizing the S/N enhances readily with acousto-optic TFS. Unlike a grating monochro-
the ability of an instrument of this type to perform mator instrument that takes data only in one direction of the
grating rotation and must rotate through angles that produce

wavelengths that will never be used, the random wavelength 1.0
access of the acousto-optic TFS allows a higher-duty cycle.
A grating monochromator instrument that has a lead sulfide
0.8 D
detector loses 50% by chopping and 50% of that 50% by A
Absorbance
the return of the grating. If only 7 of 700 wavelengths that B
A
are obtained are used, the duty cycle then is 1/100 of 1/2 of 0.6
1/2, or 0.25%. It has been demonstrated with the acousto- D
optic TFS that a large increase in the precision in the ratio
0.4
two absorption bands results from increasing duty cycles
from 0.25 to 10, 20, and 50% due to improved S/N.41 A
From quantitative analysis of binary mixtures used to 0.2 C
test an instrument, the error in reference data is removed C E
by using programmed, digitally controlled HPLC pumps to
pump fluids through a flow cell starting with 100% A with 0.0
1300 1500 1700 1900 2100 2300
a linear gradient to 100% B. At any step, it is readily pos-
Wavelength / nm
sible to calculate the percentage of fluid B and perform a
regression in this progression, e.g. on all-odd spectra to pro- Figure 7. Superimposed spectra of the pure organic chemicals
used to evaluate AOTF spectrometer performance with multicom-
duce an analytical equation. Subsequently, the even spectra ponent mixtures: (A) benzene; (B) cyclohexane; (C) 1-octanol;
can be used for validation. When this is done, a multi- (D) butylamine; (E) paraffin oil.
ple correlation coefficient of 0.9999 results accompanied
with a very low standard error of calibration.36 39 Such a Preceding the entrance to this cuvette was a sample vial
procedure is useful to test an instrument, because the labo- with the inlet from a pressurized solvent reservoir and an
ratory variables and the reference data variables have been outlet from the bottom of the sample vial into the flow-
removed or at least minimized. Using binary mixtures is through cuvette. This miniature extraction device coupled
unrealistic, because obviously a complete intercorrelation with NIR spectroscopy of the extract constituted a cold
between the two components of the mixture occurs in the batch extraction lipid analysis system.42 Per cent oil by
closed system. Even with a ternary system, avoiding inter- weight was determined from the peak at 2302 nm minus
correlation between the different components is difficult. the baseline absorbance at 2222 nm. From the calibration
In order to break the intercorrelation, five-component mix- curve, it was possible to quantitate the oil extracted from a
tures of organic liquids were prepared gravimetrically. The sample of snack food placed in the sample vial by simply
liquids were chosen to be miscible and to represent differ- filling the sample vial with solvent, and after a reasonable
ent functional groups that would provide distinct spectral residence time, displacing the solvent in the sample vial
differences for each of the five components of the mixture and forcing it into the flow-through cuvette.
without serious interference by one or more of the other Online industrial monitoring is one of the greatest poten-
mixture components. To a paraffin oil base, benzene, cyclo- tial uses for this instrument with no moving parts. This
hexane, 1-hexanol, and butylamine were added in various includes high-pressure fluid systems. High-pressure NIR
amounts. Figure 7 shows superimposed spectra of organic spectra had been reported previously in a review by
chemicals used for multicomponent assessment of a spec- Whetsel43 for materials such as liquid CO2 . Supercritical
trometer performance. From spectra taken from each of fluid extraction procedures seldom employ monitoring. The
the members of the training set, wavelength searches were commonly used supercritical CO2 does not interfere with
performed that produced a three-wavelength equation from monitoring the pressure of organic materials in the extract.
multiple linear regression. Suitable analysis results were By use of a high-pressure cell with a quartz or sapphire win-
produced for benzene, cyclohexane, 1-hexanol, and buty- dow in the cell compartment of the acousto-optic TFS the
lamine with standard errors of calibration of 0.173, 0.254, extract flowing from a supercritical fluid extraction proce-
0.292, and 0.269, respectively.37,39 These values were com- dure can be monitored. Proof of principle was established,
parable to those obtained by analysis of the same mixtures as well as the ability to monitor transients of relatively
with a commercial grating monochromator instrument and short duration by injection of hydrocarbons into a flowing
an FT-NIR instrument, but were achieved in a significantly stream and monitoring at intervals of a few milliseconds.
shorter measurement time. In this case, it was readily apparent that the time of the
Performance of the acousto-optic TFS was tested with transient material was 3 s or less.37,38 Subsequent exper-
a flow-through cuvette having a pathlength of 2 mm. imentation with the acousto-optic TFS in series with the
extractor during actual extractions of lipids from food mate- the point where a steady state was apparent. In actuality,
rials showed the utility of such a device in determining the limiting factor was immobilization of the product once a
completeness of extraction for a given fluid. In a stepwise certain level of hydrogenation produced a solid. The system
extraction, where two different components were extracted performed well under pressure.
at different CO2 densities, peaks from cholesterol and lipids
extracted and sequentially measured in real time, further
demonstrating the utility of this as an online monitor at 2.7 Temporal resolution with an acousto-optic
high-pressure conditions.37,44,45 NIR acousto-optic TFS has TFS
also been used as a detector for flow injection analysis.46
The system also was tested with industrial waste mixtures The study of polymers and the conditions under which
containing both polar and nonpolar organic compounds and they are treated in the drawing process during production
aqueous solutions.47,48 With this material, a considerable lends itself to the use of a TFS with its speed and the
amount of scattering occurred, and although the analysis versatility of its software control. The use of oscillatory
techniques whereby a perturbation of low amplitude was
was successful with most specimens, the scattering reduced
imposed on a polymer film under tension was introduced
the signal so much in certain cases that it was difficult
and popularized by Noda et al.52 Their initial work with a
to get good data. Approaches to handling scattering sam-
dispersive system required a whole series of lock-in ampli-
ples while using a single small phosphorus-doped InGaAs
fiers, relatively complicated electronics, an optical system,
detector were developed. Attempts were made to use dif-
and mechanical motion to produce oscillatory data. From
ferent optical configurations to maximize the amount of
these data, relaxation times were accessible as well as other
energy diffusely transmitted by the sample that would hit
features associated with the perturbation and recovery from
the detector, even though much of it was scattered. Use
the imposed orientation. More recently, in the mid-infrared,
of a large-area InGaAs detector and placement of the sam-
step-scan Fourier transform infrared (FT-IR) spectrometers
ple slurry very close to the detector resulted in a distinct
were synchronized to a function generator, mechanical per-
improvement over use of a lens after the sample and before
turbation, and digital signal processing (DSP) was used to
the small detector.37,39,49 In order to retain the use of a small
avoid a series of lock-in amplifiers, shortening the exper-
area (1 mm2 ) phosphorus-doped InGaAs detector a lens was
imental time and allowing more widespread use of the
required to focus the transmitted rays onto the detector.
technique.
Online monitoring of reaction mixture during an organic
A dedicated NIR oscillatory system allows the use of
reaction in the presence of a catalyst50,51 also has been
thicker specimens, totally avoids the use of a step-scan
demonstrated. Specifically the disappearance of double-
FT-IR instrument, and makes it possible to obtain poly-
bonded materials during the hydrogenation of an oil was
mer oscillatory data in the NIR that correspond to much
monitored under pressure and in the presence of the cat-
of the information formerly obtained in the fundamental
alyst. Figure 8 shows the response of the system up until
region. The heart of this temporally resolved technique is
the TFS.51,53 55 This instrument, because of its speed and
0.16
the ability to control the mechanical oscillation, the data
acquisition timing, and the oscillatory waveform all with the
same computer, provides advantages and convenience. In
Single beam intensity
0.12
this instrument, data at one wavelength are collected at each
point along the waveform of each oscillation and the wave-
0.08 length is switched until a full spectrum is obtained.51,53
Data acquired at the same point of successive waveforms
are coadded. The static spectrum is collected to show the
0.04 wavelengths of interest. In-phase spectra and quadrature
spectra are compared with each other for the same speci-
men. The temperature of the specimen is controlled and the
0.00 frequency of the perturbing oscillation also is controlled.
0 40 80 120
Time / min
By comparing the pair of in-phase and quadrature spec-
tra obtained at different oscillatory frequencies, information
Figure 8. Real-time compositional monitoring of an organic reac-
pertinent to the relaxation characteristics of the polymer are
tion using an AOTF spectrometer. The disappearance of double-
bonded materials during hydrogenation of an oil was measured readily obtained. Other information includes the direction
under pressure and in the presence of a catalyst until a steady of the functional group orientation and the resolution of
state was reached. the overlapping bands if the direction or rate of orientation
is different for the same functional group in different sub- From the above cases that documented where the
molecular environments. Software control of synchronized performance, speed, versatility, high duty cycle, random
perturbation, wavelength switching, and data acquisition wavelength access, and programed intermittent events the
with the same computer provides temporal resolution. With- acousto-optic TFS appears uniquely suited to carry out the
out the necessity of an external function generator to serve tasks described.
as a clock, the imposed mechanical perturbation (stretching)
is initiated by way of a piezoelectric actuator. Thus, due to
the convenience of TFS, an industrial friendly, simplified, 3 LIQUID CRYSTAL TUNABLE FILTERS
and relatively low-cost oscillatory instrument is available
to the polymer science industry.54
3.1 Introduction
Software-based random-access electronic wavelength
The AOTF previously described retains the image intact
switching lends itself to data acquisition that is inter-
after it passes through the filter. This function has been used
rupted by intermittent mechanical or optical events. Such
for NIR microspectroscopic imaging,63 65 Raman micro-
is the case with a dedicated polymer stretching device that
scopic imaging,66 fluorescence imaging,67 macroanalysis,68
measures dichroic ratios (DRs).51,55 A series of DRs is
and also remote sensing.69 The transmission characteris-
measured after a series of programmed elongations. The tics of the crystal and the speed of electronic wavelength
software-controlled electronic wavelength switching allows switching are excellent. The ideal filter characteristics for
the instrument to obtain scanning data or discrete wave- imaging have been described as a broad spectral range,
length data via random wavelength access, while a polymer high acceptance angle, large physical aperture, and high
film is present between two jaws of the stretching device transmittance in the region of interest. A high rejection
in the beam. The acousto-optic TFS used has polarizers in of out-of-band radiation, electronic wavelength switching,
the system, and a polarized and tuned beam is utilized to mechanical ruggedness and physical compactness, and rel-
obtain polarized spectra or random wavelength access data ative independence of polarization are other features of
points. Once the program has performed its optical func- concern. The AOTF functions extremely well for perform-
tion, an automated rotation of the sample is performed by ing spectroscopy, whether for quantitative or qualitative
way of a pneumatically actuated rack and pinion system. purposes. For high magnification imaging work such as
The software initiates the optical exercise at the other polar- used in fluorescence microscopy, the use of the AOTF
ization orientations, and for the bands designated, the DR anisotropic crystal causes slight image blurring. This is one
is calculated and stored in the computer. After return of the reason for turning to liquid crystal tunable filters (LCTFs)
sample holder to the home position, a measured amount of for Raman imaging and various other imaging duties.
stretching of the sample is performed by way of a digi- Typically an LCTF would be comprised of several cas-
tally driven stepping motor also programmed in the overall caded stages, in accordance with the Lyot birefringent filter
automated instrument design. Under the stretched condi- concept.70 Each stage consists of a birefringent element
tion, data are obtained at both polarization orientations and and a pair of nematic liquid crystal waveplates sandwiched
the DR is calculated. This process is continued until elon- between a pair of aligned linear polarizers as illustrated in
gation has been produced, e.g. 8 in the case of a wheat Figure 9. In any multiple cascade stage configuration, the
gluten. The difference between the DR after elongation and exit polarizer from one stage serves as the entrance polar-
before elongation is referred to as DR. This optical mea- izer for the next stage. These calcite or quartz birefringent
surement that is based on molecular orientation allows the elements are oriented so that incident light is normal to
user to judge the ease with which the polymers involved the optic axis and rotated 45 relative to their linear polar-
become oriented. Those specimens that resist orientation ization direction. Incident linearly polarized light is divided
have a low DR, and those that are readily oriented have into two equal amplitude paths (o-ray and e-ray). These two
a higher DR. In the case of gluten, DR relates to the rays travel at different phase velocities through the birefrin-
functional characteristics of the polymers for end use. gent material. In the multiple element cascaded LCTF, each
Various acousto-optic TFS research activities have been successive thickness (of the birefringent element) increases
reported including low loss,56 high speed,57 narrow band,58 by a power of two. Each (successive) stage transmits with
micro Raman,59 fiber optic notch filter,60 fiber optic polymer half the bandpass of that coming from the previous stage
monitoring,61 and miniature LED source low power require- and half the free spectral range. The architecture of the
ment for NASA Mars rover and other space exploration.62 first stage determines the overall free spectral range. The
Commercial spectrometers for NIR acousto-optic TFS are first stage is the one with the thinnest (birefringent) ele-
presently available not only as OEM monochromators or ment. The last stage with the thickest birefringent element
custom devices but as off-the-shelf instruments. determines the bandpass. Later results are reported using a
Nematic relationship is described by:

liquid crystal
Birefringent
waveplate nd
V
l D 2p 11
element l
45
where n is the birefringence of the material, and d is
Linear 45 Linear the thickness of the material. For each Lyot filter stage, the
polarizer polarizer
polarization analyzer is placed parallel to the input linear
polarizer, allowing only in-phase wavelenths (of light) to
be transmitted (by the linear polarizer) and directed to the
next stage. The transmittance of stage n is give by:

n l
Tn l D cos2 12
2
The nematic liquid crystal waveplates act as electron-
ically controlled phase retarders. The waveplates are ori-
ented with a crystal axis rotated 45 relative to the direction
defined by the input linear polarizer. A large applied elec-
trical potential results in relatively little retardance, whereas
zero applied electric potential produces maximum retar-
dance. Thus the wavelength is continuously tunable. There
is one liquid crystal waveplate pair per filter stage. The total
retardation is expressed as follows:
n l D Fn l C cn l 13
when Fn l is retardation of a fixed birefringent element,

Figure 9. Diagram of a multistage LCTF with expanded view and cn l is additional retardation contributed from the
of an individual stage. [Reproduced from H.R. Morris, C.C. Hoyt liquid crystal waveplate.
and P.J. Treado, Appl. Spectrosc., 48(7), 857 (1994) by permission Note that the n appearing in the first of these equa-
of the Society of Applied Spectroscopy.]
tions is varied by changing the applied electrical potential to
10-stage Lyot LCTF. The transmission after any given stage produce the liquid crystal retardation. The bandpass wave-
is a product of the transmission of the preceding individual length lB is controlled by varying the electrical potential
stages that follows a replicated sine function. Thus, there applied to the liquid crystal retarders. Indiumtin oxide
are some trade-offs in the design of the LCTF just as there (ITO) transparent electrodes coated on the inner faces of
were in the design of the AOTF. The narrower bandpass optically polished parallel glass windows are in contact
is accompanied with a narrower free spectral range and a with nematic liquid crystals. The electrode face is treated to
significantly reduced transmittance characteristic. produce preferential alignment in a selected orientation of
LCTFs are tunable replacements from the visible through the rod-like molecules. The molecules exhibit electrical and
the NIR region for discrete multilayer dielectric interfer- optical anisotropic characteristics, and the molecular polar-
ence filters. The discrete filters are large and have a high izability is also anisotropic. Polarizability from retardation
fidelity optical quality and a wide field of view, and their along the crystal axis is larger than that from retardation
transmittance is typically from 40 to 80%. When specifi- across the crystal axis (o-ray). As a result, the e-rays have
cations are written for filters of this type to be used in an a larger index of refraction than the o-rays.
instrument, the maximum band width is specified, as is the
desired transmittance. This is then fixed in the manufactur-
ing process and thus is rigid. Such rigidity, if necessary, 3.3 Imaging with liquid crystal tunable filters
can be overcome with LCTF.
When an electric field is applied to the ITO electrode,
an electric field is produced that is parallel to the direc-
3.2 Nematic liquid crystal tunable filter theory tion of propagation of incident light. The dipole that is
induced causes the liquid crystal molecules to become
The wavelength retardance, l, is the result of the optical aligned with the field. The intermolecular spring constant
path difference between the o-ray and the e-ray. This (restoring force) of the liquid crystal opposes the alignment.
Spectral tuning (dependent on the applied field strength) is are produced by Chemicon (Pittsburgh, PA, USA) and Spec-
optically equivalent to rotating the crystal axis of a uniaxial tral Dimensions (Olney, MD, USA).
crystal. The result is a change in birefringence that pro- In the comparison work done by Maris et al.70 described
duces retardance. The nematic LCTFs have the following in the previous paragraph, a dispersive spectrometer was
advantages for spectral imaging: they have a high accep- placed after the tunable filter and before the CCD detec-
tance angle, acceptable transmittance, moderately narrow tor. More recently, Raman imaging has been done by
bandpass, and moderately rapid switching speed, although Christensen et al.71 using a dual-stage FabryPerot inter-
nowhere near as fast as the corresponding AOTF. Most ferometer filtering system. Nematic liquid crystals were
importantly, the image quality is excellent. It is possible to installed in the interferometer cavities for tuning via optical
take advantage of the diffraction limited performance of the path alteration caused by selective refractive index varia-
microscope. In addition, the LCTF is physically compact tion. A stage having low spectral resolution but a broad
and mechanically rugged. spectral range is paired with a second stage having high
Some proponents of using an LCTF instead of an AOTF spectral resolution but narrow free spectral range, hence
for imaging did a thorough comparison of one particular a dual-stage FabryPerot filter (DFPF). When used as a
10-stage LCTF and one particular AOTF positioned in the system, the resulting filter provides both narrow band-
same fluorescence microspectrometer and, subsequently, a pass (11 cm1 full width at half maximum) and large free
Raman microspectrometer. The useful part of this com- spectral range (6070% transmission over a 150200 nm
parison is that both of these devices were commercially range). The Raman images produced from this procedure
available at the time.70 In comparison to the AOTF, the were improved considerably from background subtraction
LCTF had a somewhat lower peak transmittance, compa- to remove the effect of fluorescence. The out-of-band rejec-
rable amplitude stability and tuning reproducibility, and an tion of the DFPF was approximately 103 . Nematic liquid
acceptable spectral range, although it was limited to 741 nm crystals used in an LCTF or DFPF provide an alternative
vs 1120 nm of the AOTF. The tunability of the 10-stage for NIR or Raman imaging just as they do for fluorescence
LCTF had an increment half the size of the particular AOTF imaging.
with which it was compared. However, the bandpass of
the LCTF was five times greater than the bandpass of the
AOTF at the visible wavelength at which the comparison 4 DISCRETE MULTILAYER
was made. The acceptance angle for the LCTF was slightly INTERFERENCE FILTER
greater, and the response time was slower by a factor of PHOTOMETERS
two. The sensitivity was lower by an order of magnitude.
In terms of image quality, the LCTF was unquestionably Interference filters have become well established in the
clearer based on a US Air Force test image. The blur is last four or five decades as very useful optical devices
from the spectral dispersion and the TeO2 crystal. With the particularly for filter photometry or discrete filter spec-
AOTF diffraction, there is a narrow but finite distribution troscopy. The filters are mechanically relatively rugged.
of wavelengths around the center band. As a result, the No slit is required, and, therefore, the throughput is reason-
angle at which the image is diffracted also exhibits a dis- ably high. The art of manufacturing these multilayer devices
tribution. This contributes to blurring of the image. The has been refined. As mentioned previously, transmittances
blurring is compounded from tuning. The image shift from range from 40 to 80% and bandwidths at half-height are
this test is 0.01 mrad nm1 resulting in a spatial displace- typically 10 nm in the NIR range. Broader bandpass fil-
ment of 0.03 pixels per nanometer. With an LCTF, spatial ters have a greater throughput and narrower bandpasses are
resolution is diffraction limited. With an AOTF, a minimum available for making measurements of an absorption that
image degradation of 15% occurs, compounded by repro- has a rather sharp peak. There is a sacrifice of throughput
ducible tuning-related image shifting. However, the image for the narrowed bandpass. NIR spectroscopy experienced a
shifting is reproducible and computationally correctable. In renaissance in the early 1970s because commercial instru-
this comparison of the two systems, the wavelength pro- ments that incorporated interference filters were released
file was also compared. The AOTF produced a tall spike that were useful for direct analysis of commodities by
with small side bands on either side, whereas the LCTF diffuse reflectance with only minimal sample preparation
produced a much flatter and broader transmission profile. by obtaining data at multiple wavelengths and utilizing
For spectroscopy the sharp band of the AOTF is obviously mathematical expressions derived from statistical means
superior, but for imaging the area under the broad band is that include multiple linear regression. Unlike most opti-
larger, and this measurement provides a greater overall optical instruments produced by a classical instrument com-
cal change in signal. Commercial NIR imaging instruments pany that are introduced at the Pittsburgh Conference on
Analytical Chemistry and Applied Spectroscopy or meeting of a turret device. Unlike the original carousel, which
of the Federation of Analytical Chemistry and Spectroscopy was in constant motion, the turret or filter wheel indexes
Societies, this instrument was considered to be a grain test- to the position of one filter and remains at that position
ing device so it was shown instead at the Illinois State long enough to accumulate data at that wavelength. Data
Fair in Springfield, IL, USA and demonstrated for people at that wavelength may include the ratio of intensities
involved directly in the agricultural industry. The primary reflected from the sample and from a standard reflector to
concern was for determining the protein or oil in soybeans generate a reflectance quotient, or as the log (1/reflectance).
and the protein or moisture in wheat. The stepping motor then advances the turret to the next
Although discrete interference filter photometers may be filter position at the wavelength that requires measurement.
used for clear liquids or slurries, much effort was expended Programming allows stepping to anywhere from 219
initially to allow measurement of diffuse reflectance and filter positions. Specialized instruments may have a greater
make this testing device user friendly to nonspectroscopists. number of filters, but these are not in common usage. To
The main interest in NIR reflectance analysis was that it avoid the motion of a turret or filter wheel, it is possible
could be done in a matter of 1520 s after grinding a to split radiation from a common source at different angles
sample instead of being done after time-consuming sample into multiple pathways, each through its own filter, and
preparation. Historically these first interference filter NIR each through its own path of the sample. The disadvantage
instruments all had essentially the same task to perform. of such a parallel configuration is that multiple detectors
All filter instruments had lead sulfide detectors and similar are required, and multiple phase sensitive amplifiers are
filters, and they all used a quartz tungsten halogen source. also required. Thus, in commercial instruments a different
For the most part this is still true, but two of the instrument filter is often interposed into the common path at different
companies have chosen to have a thermoelectrically cooled points of time. In contrast to the indexing turret design, the
lead sulfide detector. continuously moving discrete filter system reminiscent of
There are three considerations for comparing NIR inter- the original Dickey-John carousel mount has been brought
ference filter instruments. These include first the method of back into limited usage.
wavelengths selection, i.e. how to interpose the right fil- Diffuse reflectance NIR filter instruments at the present
ter in the beam of radiation that is going to the sample. time include those marketed by Dickey-John, Bran C
Second, the mechanism for referencing is very important Luebbe (Buffalo Grove, IL, USA), Perten Instruments
in reflectance filter instruments, because the response of a (Springfield, IL, USA), LECO Inc. (St Joseph, MI, USA),
lead sulfide detector will change with time. The third fea- and Kett (Villa Park, CA, USA). Of these instruments, the
ture that varies among the instruments is the mechanism Bran C Luebbe and the Perten Instruments models use an
for collecting the radiation that is diffusely reflected off integrating sphere, referencing at each wavelength, and a
the sample. Historically the first commercial filter instru- turret indexing wavelengths one at a time. The Dickey-John
ment patented was the Grain Analysis Computer (GAC instrument has one thermoelectrically cooled lead sulfide
II) produced and marketed by Dickey-John of Auburn, IL, detector mounted at 45 , the sample cup rotates, and wave-
USA that had the filters inserted in a rotating disk described lengths are changed with a turret mechanism.72 Referencing
as a carousel. The carousel was in continuous motion and at all wavelengths proceeds in sequence prior to measur-
each time a given filter came around, data from that filter ing the sample at all wavelengths in sequence. The LECO
was coadded with data obtained at the same filter in the instrument manufactured by Oxford Instruments (Witney,
previous rotations. This instrument was marketed subse- Oxfordshire, UK) employs a turret that indexes for each
quently by Technicon Industrial Systems (Tarrytown, NY, filter and has a pair of oscillating detectors. The Textron
USA) as the InfraAnalyzer model 2.5. Another early entry Systems (Wilmington, MA, USA) instrument uses a linear
into the commercial market was the grain quality analyzer variable filter coupled with a linear silicon photodiode array
by Neotec Instruments (Silver Spring, MD, USA). Neither (CCD) detection.
of these models are available at the present time. They were In addition to the general use reflectance filter NIR instru-
superseded by filter instruments initially manufactured by ments several have been produced for specialized use.
Dickey-John and Technicon Industrial Systems, and others Because overtones and combinations of OH fundamen-
that followed. Probably 85% of all NIR analyses carried out tal stretching vibrations are quite prominent in the NIR,
are performed with filter instruments. Most of these analy- one of the constituents easiest to determine is moisture.
ses are performed by technicians who have had very little Independent of the grain industry, Moisture Systems Inc.
spectroscopic education or training. (Chelmsford, MA, USA) has produced a noncontact mois-
Common mechanisms for wavelength selection include ture meter for suspension above conveyer belts. As many
the mounting of interference filters at different positions as four filters are arranged to be mechanically placed into
the beam. With this instrument, incident radiation is pro- REFERENCES

jected toward the sample and reflected radiation is collected
at a typical distance of a few inches. Typically with such a 1. P. Debye and F.W. Sears, Proc. Natl. Acad. Sci. USA, 18, 409
system, an average of a number of readings taken while the (1932).
conveyer belt is moving is used before making any process 2. E.I. Gordon, Appl. Opt., 5(10), 1629 (1966).
decisions. Such is the case with many online processes, 3. M.G. Cohen and E.I. Gordon, Appl. Phys. Lett., 5(11), 181
but it is particularly important in a noncontact situation, (1964).
because the solid angle of collection may change with the 4. R.W. Dixon, J. Appl. Phys., 38(13), 5149 (1967).
height of the material being conveyed. 5. D.A. Pinnow, IEEE Trans. Sonics Ultrason., SU-18(4), 209
Infrared Engineering Ltd (Maldon, Essex, UK) produces (1971).
a family of filter-based monitors for a variety of industrial 6. A.W. Warner, D.L. White and W.A. Bonner, J. Appl. Phys.,
applications from brewing to plastics to tobacco inspection. 43(11), 4491 (1972).
For the brewing industry, inline alcohol and original gravity 7. A. Goutzoulis, D. Pape and S. Kulakov (eds), Design and
are measured by transmission across a 1 mm gap in an opti- Fabrication of Acousto-optic Devices, Marcel Dekker, New
cal fiber equipped stainless steel probe. Solid samples such York (1994).
as tobacco leaves or shredded or powdered materials are 8. S.E. Harris and R.W. Wallace, J. Opt. Soc. Am., 56(6), 744
measured using a noncontact filter-based instrument. The (1969).
Kett Company makes such an instrument that projects an 9. S.E. Harris, S.T.K. Nieh and D.K. Winslow, Appl. Phys.
incident beam of NIR radiation from the instrument onto Lett., 15(10), 325 (1969).
a surface. An optical targeting and range finder simplify 10. S.E. Harris, S.T.K. Neih and R.S. Feigelson, Appl. Phys.
adjustment of the position of the testing device to provide a Lett., 17, 223 (1970).
constant solid angle of collection and the scattered radiation 11. S.E. Harris, US Patent 3 679 288, July (1972).
then is collected with this noncontact device. For most ana- 12. J. Jacob and I.C. Chang, SPIE Int. Soc. Opt. Eng., 202, 40
lytical tasks, a customized discrete wavelength instrument (1979).
can be configured once the important wavelengths have 13. M. Gottlieb, J.D. Feichtner and J. Conroy, SPIE Int. Soc. Opt.
been identified and filters are purchased for the application. Eng., 232, 33 (1980).
At the present time most NIR analyses in fact are performed 14. K.B. Steinbruegge, M. Gottlieb and J.D. Feichtner, SPIE Int.
using filter instruments. Even in the present era of sophis- Soc. Opt. Eng., 268, 160 (1981).
ticated scanning instruments an analytical monitoring task 15. J.B. Houston, M. Gottlieb, S. Yao, I.C. Chang, J. Tracy, L.M.
that had been designed for a FT-IR or grating monochroma- Smithline and G.J. Wolga, Opt. Eng., 20(5), 712 (1981).
tor scanning instrument was reportedly adapted73 to discrete 16. R.L. Nelson, ISA Trans., 25(4), 31 (1986).
filter instruments to save on the cost and provide mechan- 17. I. Kurtz, R. Dwelle and P. Katzka, Rev. Sci. Instrum., 58(11),
ical stability.74,75 The authors concluded that even in the 1996 (1987).
twenty-first century, discrete filter photometry is still a com- 18. B. Bates, D.R. Halliwell, S. McNoble, Y. Li and M. Catney,
petitive means of routine industrial monitoring. Appl. Opt., 26(22), 4783 (1987).
19. J.D. Fiechtner, M. Gottlieb and J.J. Conroy, Appl. Phys. Lett.,
34, 1 (1979).
20. M. Gottlieb, N.B. Singh, R.H. Hopkins and R. Mazelsky,
ABBREVIATIONS AND ACRONYMS Opt. Eng., 33, 2503 (1994).
21. M. Gottlieb and A. Wachtel, US Patent 4 602 342, July
AC Alternating Current (1986).
A/D Analog-to-digital 22. R.W. Dixon, IEEE J. Quant. Elect., QE-3(2), 85 (1967).
CPU Central Processing Unit 23. A. Yariv, Optical Electronics, 3rd edition, Holt, Rinehart
D/A Digital-to-analog and Winston, New York, 385400 (1985).
DC Direct Current 24. C.D. Tran, Anal. Chem., 64, 971A (1992).
DFPF Dual-stage FabryPerot Filter 25. J.A. Kusters, D.A. Wilson and D.L. Hammond, J. Opt. Soc.
DSP Digital Signal Processing Am., 64(4), 434 (1974).
I/O InputOutput 26. I.C. Chang, Appl. Phys. Lett., 25(6), 323 (1974).
ITO IndiumTin Oxide 27. S.T.K. Nieh and S.E. Harris, J. Opt. Soc. Am., 62(5), 672
LED Light-emitting Diode (1972).
RF Radio Frequency 28. M.S. Kharusi and G.W. Farnell, Can. J. Phys., 47, 2719
TFS Tunable Filter Spectroscopy (1969).
29. Y. Ohmachi, N. Uchida and N. Niizeki, J. Acoust. Soc. Am., 48. D.L. Wetzel and J.A. Sweat, Mikrochim. Acta., 14, 325
51(1), 164 (1972). (1997).
30. I.C. Chang, Appl. Phys. Lett., 25, 370 (1974). 49. A.J. Eilert and D.L. Wetzel, Solutions to Difficult Sampling
31. I.C. Chang, US Patent 4 052 121, October (1977). Problems with Near-IR Acousto-optic TFS, presented at the
Pittsburgh Conference on Analytical Chemistry and Applied
32. N. Uchida and Y. Ohmachi, J. Appl. Phys., 40(12), 4692 Spectroscopy, Chicago, IL, paper 101 (1994).
(1969).
50. D.L. Wetzel and J.A. Sweat, Interactive Software Controlled
33. T. Yano and A. Watanabe, Appl. Phys. Lett., 24(6), 256 Electronic Wavelength Switching of NIR Acousto-optic TFS
(1974). Enhances Quantitative Capability, presented at the 24th
34. G.J. Kemeny and D.L. Wetzel, US Patent 4 883 963, April Annual Conference of the Federation of Analytical Chemistry
1985/November (1989). and Spectroscopy Societies, Providence, RI (1997).
35. D.L. Wetzel, G.J. Kemeny and A.J. Eilert, Using and 51. J. Sweat, Instrumentation for Time Resolved Dynamic and
Acousto-optic Tunable Filter in Near-infrared Spectroscopy, Static Dichroic Measurements of Polymers with a Near-IR
presented at American Chemical Society Meeting, Denver, Acousto-optic Tunable Filter, PhD Dissertation, Kansas State
CO (1987). University, Manhattan, KS (1999).
36. A.J. Eilert and D.L. Wetzel, Acousto-optic Tunable Filter 52. I. Noda, A.E. Dowrey and C. Marcott, Appl. Spectrosc., 42,
Spectroscopy in the Near-infrared, presented at the Pitts- 203 (1988).
burgh Conference on Analytical Chemistry and Applied Spec- 53. J.A. Sweat and D.L. Wetzel, Dynamic Polymer Examination
troscopy, Chicago, IL, paper 217 (1991). with an AOTF Spectrometer, presented at the 50th Annual
37. A.J. Eilert, Acousto-optic Tunable Filter Spectroscopic Ins- Pittsburgh Conference on Analytical Chemistry and Applied
trumentation for Quantitative Near-IR Analysis of Organic Spectroscopy, Orlando, FL, paper 1195 (1999).
Materials, PhD dissertation, Kansas State University, Man- 54. J.A. Sweat and D.L. Wetzel, Rev. Sci. Instr., 72, 2153 (2001).
hattan, KS (1995).
55. D.L. Wetzel and J.A. Sweat, Automated Differential Di-
38. D.L. Wetzel, Fastest Gun in the West!, presented at the chroic Ratio Polymer Orientation Stretching Response Gauge
Pittsburgh Conference on Analytical Chemistry and Applied via Near-IR, presented at the 51st Annual Pittsburgh Con-
Spectroscopy, New Orleans, LA, paper 1135 (1993). ference on Analytical Chemistry and Applied Spectroscopy,
39. A.J. Eilert, J.A. Sweat and D.L. Wetzel, J. Near-Infrared New Orleans, LA, paper 1163 (2000).
Spectrosc., 8, 239 (2000). 56. D.O. Culverhouse, S.H. Yun, D.J. Richardon, T.A. Birks,
40. D.L. Wetzel and A.J. Eilert, Isophotonic Data Acquisition S.G. Farwell and P.St.J. Russel, Opt. Lett., 22, 96 (1997).
for Quantitative Multiple Discrete Wavelength Near-infrared 57. A. Barocsi, L. Jakab, P. Richter, C. Schrodter and C. Hage-
Spectroscopy, presented at the 16th Annual Meeting of dorn, Opt. Eng., 32, 2569 (1993).
the Federation of Analytical Chemistry and Spectroscopy 58. E.G. Paek, J.Y. Choe and T.K. Oh, Opt. Lett., 23, 1322
Societies, Chicago, IL (1989). (1998).
41. D.L. Wetzel and A.J. Eilert, Near-infrared Analysis with 59. E.N. Lewis, P.J. Treado and I.W. Levin, Appl. Spectrosc., 47,
a High Duty Cycle Acousto-optic Tunable Filter Random 539 (1993).
Wavelength Access System, presented at the 14th Annual
Meeting of the Federation of Analytical Chemistry and Spec- 60. H.S. Kim, S.H. Yun, I.K. Kwang and B.Y. Kim, Opt. Lett.,
troscopy Societies, Detroit, MI, paper 284 (1987). 22, 1476 (1997).
42. D.L. Wetzel and A.J. Eilert, Cereal Foods World, 37(7), 552 61. M. Huhne, U. Eschenauer and H.W. Sielser, Appl Spectrosc.,
(1992). 49, 177 (1995).
43. K.B. Whetsel, Appl. Spectrosc. Rev., 2(1), 1 (1968). 62. X. Wang, D.E. Vaughan, V. Pelekhaty and J. Crisp, Rev. Sci.
Instrum., 65, 3653 (1994).
44. D.L. Wetzel and J.A. Sweat, Near Infrared Acousto-
optic Tunable Filter Spectrometer Extractogram of Carbon 63. P.J. Treado, I.W. Levin and E.N. Lewis, Appl. Spectrosc., 46,
Dioxide Supercritical Fluid Extraction, in Near Infrared 553 (1992).
Spectroscopy: The Future Waves, eds A.M.C. Davies and 64. J. Romier, J. Selves and J. Gastellu-Etchegorry, Rev. Sci.
P. Williams, NIR Publications, Chichester, 8691 (1996). Instrum., 69, 2859 (1998).
45. D.L. Wetzel, J.A. Sweat and M. Zhao, Combined Extraction, 65. E.S. Wachman, W. Niu and D.L. Farkas, Appl. Opt., 35, 5220
Separation and Determination by Near-IR (AOTF) Detection (1996).
for Selective SFE of Cholesterol and Lipids, presented at the 66. E.N. Lewis, P.J. Treado and I.W. Levin, Appl. Spectrosc., 47,
Pittsburgh Conference, Atlanta, GA (1997). 539 (1993).
46. M.S. Baptista, C.D. Tran and G.H. Gao, Anal. Chem., 68, 971 67. C.D. Tran and R.J. Furlan, Anal. Chem., 64, 2775 (1992).
(1996).
68. Y. Cui, D. Cui and J. Tang, Opt. Eng., 32, 2899 (1993).
47. D.L. Wetzel, Fourier Transform Near Infrared Spectroscopy:
Its Place Among Other Quantitative Near Infrared Instruments 69. K.M. Nield, A. Bittar and J.D. Hamlin, Appl. Opt., 36, 7939
and Its Utility for Qualitative Analysis, in Near Infrared (1997).
Spectroscopy: The Future Waves, eds A.M.C. Davies and 70. H.R. Morris, C.C. Hoyt and P.J. Treado, Appl. Spectrosc.,
P. Williams, NIR Publications, Chichester, 5765 (1996). 48(7), 857 (1994).
71. K.A. Christensen, N.L. Bradley, M.D. Morris and R.V. Mor- 74. D.L. Wetzel, Anal. Chem., 55(12), 1165A (1986).
rision, Appl. Spec., 49(8), 1120 (1995). 75. D.L. Wetzel, Contemporary Near-infrared Instrumentation,
72. J.T. Knepler, US Patent 3 828 173 (1974). in Near Infrared Technology in the Agricultural and Food
73. D.S. Goldman and N. Lytle, Spectroscopy, 15(2), 40 Industries, eds P.C. Williams and K. Norris, Am. Assoc.
(2000). Cereal. Chem., St Paul, MN, in press.
Hadamard Transform Near-infrared Spectrometers
Robert M. Hammaker1, Richard A. DeVerse2, Daniel J. Asunskis1 and

William G. Fateley3
1
Kansas State University, Manhattan, KS, USA
2
Three LC, Inc.-Kona, Kailua-Kona, HI, USA
3
Three LC, Inc., Manhattan, KS, USA
1 INTRODUCTION TO HADAMARD is passed through the exit slit to the single-element detector.
TRANSFORM SPECTROMETRY Consider the simple example of three spectral resolution
elements (N D 3) represented by Ia with Ia D 3.0 0.1,
Ib with Ib D 4.0 0.1 and Ic with Ic D 5.0 0.1 where li
Hadamard transform spectrometry (HTS) is a unique com-
and Ii represent the wavelength and irradiance, respectively,
bination of dispersive and multiplexing spectrometries
of the three spectral resolution elements. We assume noise
where the choice of transparent or opaque for the ele-
to be independent of irradiance so the errors in measuring
ments of a multi-slit array (i.e. a mask), located on a focal
the irradiance are taken to be equal. Figure 1 presents a
plane, provides an encoding of spectral information that
schematic representation of the operation of a scanning
may be decoded into a conventional spectrum by Hadamard
dispersive spectrometer using a weight matrix for N D 3.
mathematics.1 25 As in a dispersive spectrometer, the radi-
A dispersive spectrometer can operate as a HT spectrom-
ation from a source is collected and separated into its
eter if the exit slit is replaced by a spatial light modulator
individual spectral resolution elements by a spectral separa-
(SLM) encoded by Hadamard encoding logic. Whichever
tor and then is collected and focused for spatial presentation
spectral resolution elements are selected by the Hadamard
on a focal plane. Unlike a scanning dispersive spectrometer,
encoding mask are combined and presented to a single-
which possesses a single exit slit, the Hadamard trans-
element detector. For our case of three spectral resolution
form (HT) spectrometer employs the multi-slit array as a
elements, each of three Hadamard encodements would con-
Hadamard encoding mask. This arrangement allows for the
sist of a different combination of two of the three spectral
simultaneous measurement of a multitude of spectral reso-
resolution elements. Figure 2 presents a schematic repre-
lution elements at a single-element detector and produces
sentation of the operation of a HT spectrometer using a
a multiplexing spectrometer. To recover the spectrum of N
weight matrix that is a Hadamard encoding matrix. The sub-
resolution elements requires measurement of the detector
ject of Hadamard encoding matrices has been extensively
response for N different encodements of (N C 1)/2 open investigated and three methods are known for generating
mask elements and (N 1)/2 closed mask elements. The the square cyclic Hadamard encoding matrices normally
primary data recorded as a plot of detector response versus represented by SN for the case of N spectral resolution
encodement number is called an encodegram. Hadamard elements.2 A table of results is available for many values
transformation of the encodegram yields the spectrum.1,22 of N.3,6 (The S3 matrix used here has its third row not its
A dispersive spectrometer presents individual spectral first row identical to the entry on page 202 of Harwit and
resolution elements on a focal plane and, by scanning the Sloane3 and page 435 of Hammaker et al.,6 but since the
spectral separator, one spectral resolution element at a time SN matrices are cyclic the S3 chosen here and the S3 gener-
ated using the entry on page 202 of Harwit and Sloane3 and
John Wiley & Sons Ltd, 2002. page 435 of Hammaker et al.,6 must give identical results.)
Weight matrix 100 => Weight matrix 011 =>
b + c Ib + Ic = irradiance
a Ia = irradiance =(4.0 + 5.0) 0.1
= 3.0 0.1
Source Mask detector
Source Sample
Mask detector
Sample
Weight matrix 110 =>
a + b Ia + Ib = irradiance
=(3.0 + 4.0) 0.1
Ib = irradiance
= 4.0 0.1 Source Mask detector
b
Sample
Source Mask detector
Sample Weight matrix 101 =>
a + Ia + Ic = irradiance
=(3.0 + 5.0) 0.1
Ic = irradiance Source Mask detector

c = 5.0 0.1 Sample
Figure 2. A schematic illustration of the measurement of
Source Mask detector (N C 1)/2 spectral elements at a time using a HT spectrometer
Sample for three spectral resolution elements (N D 3). [Reproduced by
Figure 1. A schematic illustration of the measurement of one permission of International Scientific Communications, Inc. from
spectral resolution element at a time using a scanning desper- R.A. DeVerse, R.M. Hammaker, W.G. Fateley, J.A. Graham and
sive spectrometer for three spectral resolution elements (N D 3). J.D. Tate, Am. Lab., 30(21), 112S (1998).]
[Reproduced by permission of International Scientific Communi-
cations, Inc. from R.A. DeVerse, R.M. Hammaker, W.G. Fateley, D1 Ib + Ic 0 1 1 Ia
J.A. Graham and J.D. Tate, Am. Lab., 30(21), 112S (1998).] [D] = D2 = Ia + Ib S3 = 1 1 0 [I] = Ib
D3 Ia + Ic 1 0 1 Ic
The operation of the HT spectrometer and the reduction
0 1 1 1 1 1 1/2 1/2 1/2
in error in Ii that leads to an improved signal-to-noise ratio 1
ST
3 = 1 1 0 T3 = 1 1 1 S3 = 1/2 1/2 1/2
(S/N) compared to a scanning dispersive spectrometer and 1 0 1 1 1 1 1/2 1/2 1/2
constitutes the multiplex advantage can be described by
matrix algebra as follows. The Hadamard encodement for Ia D1 1/2 (D1 + D2 + D3)
N spectral resolution elements is described by [D] = SN [I], [I] = Ib = S31 D2 = 1/2 (D1 + D2 D3)
Ic D3 1/2 (D1 D2 + D3)
where [D] is a column vector of N detector responses, SN
is the square cyclic Hadamard encoding matrix, and [I] Figure 3. Matrices and matrix equations for N D 3. [Repro-
is a column vector of the irradiances of the N spectral duced by permission of International Scientific Communica-
tions, Inc. from R.A. DeVerse, R.M. Hammaker, W.G. Fateley,
resolution elements. The encodegram from [D] = SN [I] is J.A. Graham and J.D. Tate, Am. Lab., 30(21), 112S (1998).]
Hadamard transformed to the spectrum by [I] = S 1
N [D],
1
where SN is the inverse of the Hadamard encoding matrix. elements are all C1, detailed treatments give the error in
p
S 1
N D [2/N C 1]TN , where TN is a matrix of C1 and 1 Ii from a HT spectrometer as [2 N/N C 1]s, where
elements. STN is the transpose of SN (interchange columns s is the standard deviation when Ii is measured using
and rows) and TN has C1 and 1 where STN has 1 and ap scanning dispersive 8
p spectrometer. For N D 3 we have
0, respectively.4,7 For our example of N D 3 the various [2 N/N C 1] D 3/2 D 0.866 as the factor reducing the
matrices and matrix equations are as shown in Figure 3. error in Ii , i D a, b, c from 0.1 to 0.0866 when using a
We define ei as the error in Ii using a scanning dispersive HT spectrometer. For large N the factor becomes significant
spectrometer and make the following three assumptions and, for example, N D 255 gives a factor of 0.125. This
using the notation EfAg for the expected value of A: (1) the is an improvement in S/N of approximately eight times
average error is zero so Efei g D 0; (2) the errors in different over conventional raster scan (CRS) methods. In practice
measurements are independent so Efei ek g D 0, (i 6D k); and it is not necessary to invert SN to obtain the spectrum
(3) ei has a variance s2 (and standard deviation s) so since a fast Hadamard transform (FHT) algorithm has been
Efe2i g D s2 . As a consequence of assumption (2) and the available for some time.24,25 It is important to note that both
fact that all TN elements are 1 so then squares of TN spectrometry and imaging realize the same theoretical S/N
Hadamard Transform Near-infrared Spectrometers 3
improvement based on the number of spectral or spatial crystal (LC) technology. The first generation stationary
resolution elements (N), respectively. Hadamard encoding mask was a cholesteric LC array with
If a conventional one-dimensional (1D) Hadamard encod- N D 127 mask elements and used polarization as its oper-
ing mask is folded in some manner, the result is the ating phenomenon.12,22 Two parallel polarizers and rotation
generation of a two-dimensional (2D) Hadamard encoding or lack of rotation of the polarized radiation generated
mask. Similar to the 1D masks used for spectral multiplex- the opaque and transparent states, respectively. The sec-
ing, 2D masks can be used for the spatial multiplexing done ond generation stationary Hadamard encoding mask was
in Hadamard transform imaging (HTI), an area of current fabricated using a polymer dispersed liquid crystal (PDLC)
interest22 24,26 36 but outside the subject of this article. material with N D 255 mask elements and used light scat-
In summary, some important features of HTS are: tering as its operating phenomenon.22 The PDLC contained
LC droplets dispersed in a polymer matrix whose index of
it is a dispersive technique using a spectral separator; refraction matched the index of refraction in one direction
it is a multiplexing technique using a single-element in the birefringent LC droplet. Alignment of the LC droplet
detector; optical axis under an applied voltage removed discontinu-
it uses a Hadamard encoding mask (multi-slit array) in ities in index of refraction at the polymer matrix/LC droplet
the focal plane; interface to generate a good transparent state while random
it sends one more than half the spectral resolution ele- orientation of LC droplets in the polymer matrix generated
ments to the single-element detector in an encodement; the opaque state from light scattering by the discontinu-
it uses a number of encodements equal to the number ities in index of refraction at the polymer matrix/LC droplet
of spectral resolution elements desired and the number interface.
of mask elements (pixels) in each encodement; Although we have been successful in doing HTS with
it has each encodement contain a different combination the first and second generation stationary Hadamard encod-
of one more than half the spectral resolution elements; ing masks,9,11 17,19,22 there were problems.34 The transi-
it has as its primary data the encodegram, a record of tion time between on and off states was slow. The
detector responses versus encodement number; absorbance spectrum of the LC masks limited their use
it uses a FHT of the encodegram to generate the to the visible and short wavelength near-infrared spectral
spectrum.25 regions. The transmission properties were not ideal. When
the spatial resolution element was selected to be closed or
off some radiation still reached the detector (transmit-
2 THE HADAMARD ENCODING MASK tance of To 6D 0) and when the spatial resolution element
was selected to be open or on some radiation was not
Two choices for the Hadamard encoding mask are the transmitted (transmittance of Tt 6D 1).
moving (mechanical) mask and the stationary (electro- We have not used moving masks in HTS but have done
optic) mask. The moving mask for N resolution elements so successfully in HTI30 33 with some limitations such
requires 2N 1 mask elements that are either completely as slow translation times for moving the mask from one
open (transmittance of 1) or completely closed (transmit- encodement to the next, the structural necessity of spacers
tance of 0). The stationary mask for N resolution elements between the mask elements, fixed mask dimensions, fixed
requires only N mask elements that are selected to be either mask element size, and general wear-positioning error in
transparent (transmittance of Tt
1) or opaque (transmit- moving mechanical systems.34
tance of To 0). The resulting S/N depends on (T D Our experience with both stationary and moving Hada-
Tt To ).10 The advantage and disadvantage of the moving mard encoding masks generated the following mask devel-
mask are no optical transmission problems and potential opment wish list: (1) maximize transmission in on
moving parts problems, respectively.5,22 The advantage state; (2) minimize transmission in off state; (3) maxi-
and disadvantage of the stationary mask are no moving mize number of resolution elements; (4) minimize size of
parts problems and potential optical transmission problems, resolution elements; (5) minimize transition time between
respectively.9,10,13,18,22 two states; (6) reliable operation.
The HT spectrometer developed by Decker in the early
1970s employed a moving mask.37 Significant problems
such as misalignment and jamming associated with a mov- 3 THE DIGITAL MICRO-MIRROR
ing mask led to a reputation of poor reliability and con- ARRAY
tributed to a dormant period in the development of HTS.
Our initial contribution to HTS was the introduction of The development of new technology is necessary to attain
the stationary Hadamard encoding mask based on liquid or approach more closely all six items on the mask
development wish list. Micro-optoelectromechanical sys- reflecting light away from the sample plane to an optical
tems (MOEMS) is a new technology that shows promise for stop. The DMA can be used as a SLM to develop a num-
advances in several areas. A commercial MOEMS device ber of unique adaptive spectrometers (and hyperspectral
that is a natural choice for a Hadamard encoding mask is imaging systems). Some features of the DMA that make it
a digital micro-mirror array (DMA) developed by Texas attractive as a SLM for spectrometry (and imaging) includ-
Instruments for projector display applications. ing but not limited to HTS (and HTI) are as follows:34,35,38
Figure 4 provides the details of a typical DMA.
This DMA incorporates 508 800 micro-mirrors in a Reflectivity. The mirror surface is aluminum, which is
848 column 600 row array that is 14.4 mm wide by highly reflective over broad spectral regions.
10.2 mm high. An individual micro-mirror is 16 m square Speed. The mirrors can move from the on (C10 ) to
and adjacent micro-mirrors are separated by a 1 m gap. off (10 ) position and come to rest in less than 20 s
The micro-mirrors are built using complementary metal providing the possibility for very rapid data collection.
oxide semiconductor (CMOS) construction techniques on Reliability in position. The mirrors are permanently
top of the integrated circuits used to control their move- fixed to the surface of the CMOS substrate and are
ments. Figure 4 also illustrates the 10 , bistable micro- rotated to exact locations determined by mechanical
mirror operation by rotation about its diagonal axis. The stops.
micro-mirrors in the DMA are individually addressable by Variable mask dimensions. Only the number of micro-
writing digital logic into the memory cells located directly mirrors in the array limits the number of useful
below them. They can rapidly switch and settle from one mask elements or pixels. The number of micro-mirrors
of the bistable conditions in less than 20 s. selected to form an individual optical mask element or
For the mask to provide full transmission, on or rejec- pixel dictates the spectral (or spatial) resolution. The
tion off of the signal from a given micro-mirror would ability to select any number of available micro-mirrors
be ideal. The on state reflects light onto the sample plane for each optical mask element or pixel provides vari-
with Tt approaching 1. Achieving Tt D 1 would require able spectral resolution in spectrometry and a digital
100% reflection over the entire spectral range considered. zoom feature in imaging. Changes in multiplexing and
The off state, however, achieves the ideal of To D 0 by data collection algorithms do not require physical alter-
ations of the optics.
on Virtual solid state device. The DMA is manufactured
+10
using CMOS construction techniques and is subject to
off the same physical advantages and limitations of solid
10 state devices.
Mirror drive options. The DMA was designed to accept
both digital data and video input for mirror control.
New mirror-driver technology may allow the micro-
CMOS base mirrors to be driven to the limits of their physical
performance.
17 m
High radiation tolerance. Built for high-luminosity
projector systems, the DMA is constructed to tolerate
the intense irradiance from the arc lamps associated
with these applications.
1 m Although the DMA is very useful in HTS, there are some

difficulties using the DMA in spectrometry (and imaging).38
Glass cover plate. The glass cover plate is less than

848 600 80% transmissive for wavelengths longer than 700 nm
DMA Aluminum and shorter than 2600 nm. The optical transmission
reflecting drops rapidly off to 0% for wavelengths longer than
surface
2750 nm.
Figure 4. DMA architecture showing detail for two micro- Bulky optical mounting. The large electronics board
mirrors in the DMA. Micro-mirrors rotate 10 from the normal
on a diagonal axis. [Reproduced by permission of Applied Spec- mounting limits optical flexibility in some applications.
troscopy from R.A. DeVerse, R.M. Hammaker and W.G. Fateley, Mirror dimensions. Although adequate for visible and
Appl. Spectrosc., 54, 1751 (2000).38 ] NIR spectral regions the small micro-mirror dimensions
may generate some optical problems at wavelengths Off

axis
longer than 16 m or wavenumbers less than 625 cm1 . Sample
paraboloid
Overhead video electronics. The video processing
requirements hinder some direct access to the DMAs Entrance
micro-mirrors. slit
Most of these problems for spectrometry (and imaging) Source

purposes have been addressed.38 The glass plate may pos-
sibly be replaced with a suitable optical material for the
longer wavelength regions of the spectrum. An extension Holographic
that removes the DMA from the bulky circuit board has concave
been developed to provide for flexibility in optical posi- grating
tioning. The potential diffraction problems at longer wave-
lengths may possibly be avoided by appropriate pre-DMA
optical systems. Overhead electronics have been eliminated
by the availability of new generation DMA drivers. PC
The fact that each individual micro-mirror in the DMA
tor
is under computer control provides a multitude of ways DMA interface
tec
to present the radiation distributed over the focal plane
De
to the single-element detector of any DMA dispersive
spectrometer. Once the instrument has been designed and
Figure 5. A general schematic for a flat-field NIRDMAS with
constructed any combination of any number of whatever the DMA providing a multitude of ways to present the radiation
spectral (or spatial) resolution elements are present on distributed over the focal plane to the single-element detector.
the focal plane can be sent to the single-element detector [Reproduced by permission of International Scientific Communi-
without physical alteration of the optics. The CRS scan cations, Inc. from R.A. DeVerse, R.M. Hammaker, W.G. Fateley,
and the HTS scan described later are only two of many J.A. Graham and J.D. Tate, Am. Lab., 30(21), 112S (1998).]
possibilities.
at the focal plane of a holographically ruled concave
A single DMA dispersive spectrometer, acting as an
grating and a general schematic for a NIRDMAS appears in
adaptive programmable filter spectrometer, can replace
Figure 5. This simple dispersive spectrometer has no mov-
many different optical filter spectrometers since the DMA
ing parts other than the micro-mirrors in the DMA. The
can select or reject any set of desired frequencies from
DMA allows the measurement of all available spectral res-
the range of frequencies provided by the source. Rejection
olution elements individually and consecutively as in a CRS
of unwanted spectral information is easily accomplished
scan or individually in any conceivable sequence or in sets
as done in eliminating the Rayleigh line from a Raman
of two or more in any conceivable sequence (including a
spectrum23,24 or selecting only the atomic emission lines
for a single element in a multi-element atomic emission HTS scan and a multitude of other possibilities) without
spectrum23,24 or eliminating any spectral region that con- physical alteration of the other optics in the instrument.
tains more unwanted noise than desired signal. The 2D The 2D nature of the DMA provides a choice of any
nature of the DMA may permit an echelle emission spec- section of the available entrance aperture in any type of
trometer with dispersion in 2D to be operated with a single- scan as illustrated in Figure 6. The columns and rows of
element detector rather than a multi-channel detection sys- the individual micro-mirrors forming the DMA correspond
tem. The future applications of the DMA in spectrometry to the dispersion dimension of the focal plane and the spa-
(and imaging) might be limited only by the imagination, tial dimension of the entrance aperture, respectively. The
interests, and enthusiasm of individual investigators. rectangular slit image of width x and height y in Figure 6
represents a spectral resolution element and illustrates the
ability to select the desired number of the available columns
4 THE NEAR-INFRARED DIGITAL and rows of individual micro-mirrors for choosing the spec-
tral resolution and slit height, respectively, for a particular
MICRO-MIRROR ARRAY experiment.
SPECTROMETER Figure 7 illustrates how the ability to select the number
of individual micro-mirrors in the 2D array corresponding
A flat-field near-infrared digital micro-mirror array spec- to a spatial resolution element in an imaging application
trometer (NIRDMAS) has been constructed with a DMA leads to a digital zoom feature. In the z spatial resolution
1 800 the feature of interest is imaged with a 10 magnification

1 in the higher spatial resolution 10 scan. The procedures
x illustrated in Figures 6 and 7 are implemented entirely by
software and do not require physical alterations of the
1 2 n optics.
Rows
y Multiplexing techniques provide the maximum multi-

plex advantage only if the noise is independent of signal,
i.e. noise D (signal)m with m D 0 (detector noise limit).
600 Values of m other than m D 0 (e.g. m D 1/2 (photon or
Columns shot noise limit) or m D 1 (fluctuation noise limit)) can
negate the multiplex advantage or produce a multiplex dis-
Figure 6. The DMA as a SLM at the focal plane for the NIRD-
MAS with the columns and rows corresponding to the dispersion advantage. In the m D 0 case the Fourier transform (FT)
dimension and the spatial dimension of the entrance aperture, spectrometer and the p HT spectrometer
p have limiting mul-
respectively. [Reproduced by permission of Applied Spectroscopy tiplex advantages of N and N/2, respectively, for N
from R.A. DeVerse, R.M. Hammaker and W.G. Fateley, Appl. spectral resolution elements compared to a CRS on a scan-
Spectrosc., 54, 1751 (2000).38 ]
ning dispersive spectrometer for equal measurement time.
The FT spectrometer hasp a twofold
p S/N advantage over the
17 columns
HT spectrometer (i.e. N/ N/2 D 2) for equal through-
put of incident radiation. However, in practice the HT
spectrometer may require fewer optical surfaces than the
SLM
at z spatial FT spectrometer leading to a larger throughput of incident
15 rows
resolution radiation and making any multiplex advantage for the two
spectrometer types more nearly equal.
The NIRDMAS is capable of performing a rigorous
comparison of a CRS and a HTS scan to determine the
Hadamard multiplex advantage with a non-photon noise
limited single-element detector. The fact that no physical
alterations should be required between the CRS data col-
lection and the HTS scan data collection should allow a
direct evaluation of the Hadamard multiplex advantage to
be made. The NIRDMAS schematic in Figure 5 was mod-
ified by replacing all components prior to the entrance slit
by a mercuryargon calibration lamp optically coupled by
17 columns
3 feet of 200 m diameter fiber optic to a 1 cm cuvette sam-
ple cell holder that was optically coupled to the entrance
15 rows
slit of the NIRDMAS by 3 feet of 800 m fiber optic.

SLM
at z /10
The CRS and HTS scans used 800 columns of the DMA
spatial resolution for spectral resolution and 30 of the 600 rows of the DMA
Figure 7. Illustration of the convenience of variable mask dimen- for slit height to allow both the CRS scan measurement
sions in providing a digital zoom feature that uses software only and the HTS scan measurement to fit within the dynamic
and requires no physical alterations in the optics. range of the single-element InGaAs detector operating at
ambient temperature. The ability to select the necessary slit
arrangement larger numbers of the available individual height by computer control of the DMA without physical
micro-mirrors are grouped into larger 2D arrays to form alterations of the optics is critical to the success of this
a smaller number of spatial resolution elements used in a comparison of the CRS and HTS scans. The total data
lower spatial resolution z scan to quickly locate features collection time was 26.2 min. The two types of scans
of interest. In the z/10 spatial resolution arrangement an were done sequentially under identical conditions so that
array forming a spectral resolution element in the z spa- the only difference between the CRS scan and the HTS
tial resolution arrangement is divided into arrays forming scan other than their inherent difference was temporal.
spatial resolution elements of 1/10 the size of those in the The results appear in Figure 8 where the two scans are
lower spatial resolution z scan. Then the spectral resolution compared to show the dramatic improvement in S/N for
element in the lower spatial resolution z scan containing the HTS scan approximating the theoretical improvement
3000 ABBREVIATIONS AND ACRONYMS
2000
CMOS Complementary Metal Oxide
Semiconductor
CRS Conventional Raster Scan
1000 DMA Digital Micro-mirror Array
FHT Fast Hadamard Transform
HT Hadamard Transform
0
Arbitrary
HTI Hadamard Transform Imaging

(a) 1000 1200 1400 1600
HTS Hadamard Transform Spectrometry
LC Liquid Crystal
MOEMS Micro-optoelectromechanical Systems
2000 NIRDMAS Near-infrared Digital Micro-mirror Array
Spectrometer
PDLC Polymer Dispersed Liquid Crystal
1000 SLM Spatial Light Modulator
0 REFERENCES
1000 1200 1400 1600
(b) Wavelength (nm) 1. M. Harwit and N.J.A. Sloane, Hadamard Transform Optics,
Figure 8. A comparison of (a) the CRS and (b) the HTS scan Academic Press, New York, 119 (1979).
illustrating the dramatic improvement in S/N for the HTS scan. 2. M. Harwit and N.J.A. Sloane, Hadamard Transform Optics,
p Academic Press, New York, 200209 (1979).
in S/N D 800/2 D 14.1. The HTS scan S/N, the CRS 3. M. Harwit and N.J.A. Sloane, Hadamard Transform Optics,
scan S/N, and the ratio of the HTS scan S/N to the CRS Academic Press, New York, 202203 (1979).
scan S/N, respectively, for the four bands are: 85.7, 7.4, 4. M. Harwit and N.J.A. Sloane, Hadamard Transform Optics,
11.6 at 1530.5 nm, 152.2, 10.4, 14.6 at 1368.2 nm, 77.0, Academic Press, New York, 215 (1979).
6.4, 12.0 at 1129.4 nm, and 132.4, 10.1, 13.1 at 1014.7 nm. 5. R.M. Hammaker, J.A. Graham, D.C. Tilotta and W.G. Fate-
ley, What is Hadamard Transform Spectroscopy?, in Vibra-
tional Spectra and Structure, ed. J.R. Durig, Elsevier, Ams-
terdam, 401485, Vol. 15 (1986).
5 SUMMARY
6. R.M. Hammaker, J.A. Graham, D.C. Tilotta and W.G. Fate-
ley, What is Hadamard Transform Spectroscopy?, in Vibra-
The DMA promises to be the best Hadamard encoding tional Spectra and Structure, ed. J.R. Durig, Elsevier, Ams-
mask yet developed. However, its potential applications in terdam, 435436, Vol. 15 (1986).
spectrometry and imaging are by no means limited to HT 7. R.M. Hammaker, J.A. Graham, D.C. Tilotta and W.G. Fate-
techniques. The NIRDMAS has no moving parts other than ley, What is Hadamard Transform Spectroscopy?, in Vibra-
the micro-mirrors in the DMA and promises to provide tional Spectra and Structure, ed. J.R. Durig, Elsevier, Ams-
a compact and robust instrument for operation in poten- terdam, 439, Vol. 15 (1986).
tially hostile environments ranging from process control to 8. R.M. Hammaker, J.A. Graham, D.C. Tilotta and W.G. Fate-
outer space. Some of the options for DMA-based instru- ley, What is Hadamard Transform Spectroscopy?, in Vibra-
tional Spectra and Structure, ed. J.R. Durig, Elsevier, Ams-
mentation are a HT spectrometer, a HT imager, a HTI (one terdam, 434444, Vol. 15 (1986).
spatial dimension) spectrograph, a variable filter spectro-
9. D.C. Tilotta, R.M. Hammaker and W.G. Fateley, Appl. Spec-
meter, and a HT imager with spectral resolution (from a trosc., 41, 727 (1987).
second DMA in a HT spectrometer or a FT spectrometer
10. D.C. Tilotta, R.M. Hammaker and W.G. Fateley, Appl. Opt.,
or some other spectral separator) to generate a hyperspectral 26, 4285 (1987).
data cube using a single-element detector. We believe that
11. D.C. Tilotta, R.D. Freeman and W.G. Fateley, Appl. Spec-
the combination of a DMA with a single-element detector trosc., 41, 1280 (1987).
will provide an important advance in spectroscopic instru-
12. A.P. Bohlke, D. Lin-Vien, R.M. Hammaker and W.G. Fate-
mentation and that instrumentation based on the DMA will ley, Hadamard Transform Spectrometry: Application to Bio-
lead to a host of environmental, industrial, medical, and logical Systems, A Review, in Spectroscopy of Inor-
military applications. ganic Bioactivators, Theory and Applications Chemistry,
Physics, Biology, and Medicine, NATO ASI Series C, ed. 25. E.D. Nelson and M .L. Fredman, J. Opt. Soc. Am., 60, 1664
T. Theophanides, Kluwer, Boston, 159189, Vol. 280 (1989). (1970).
13. A.P. Bohlke, J.D. Tate, J.S. White, J.V. Paukstelis, R.M. 26. S.L. Wright, K.J. Latas, A.N. Mortensen, E.A. Orr, J.V.
Hammaker and W.G. Fateley, J. Mol. Struct. (Theochem.), Paukstelis, R.M. Hammaker and W.G. Fateley, SPIE Proc.,
200, 471 (1989). 1857, 135 (1993).
14. W.G. Fateley, R.M. Hammaker, J.V. Paukstelis, A.P. Bohlke, 27. W.G. Fateley, R.M. Hammaker, J.V. Paukstelis, S.L. Wright,
J.D. Tate, J.S. White and J.M. Jarvis, SPIE Proc., 1320, 39 E.A. Orr, A.N. Mortensen and K.J. Latas, Appl. Spectrosc.,
(1990). 47, 1464 (1993).
15. R.M. Hammaker, A.P. Bohlke, J.M. Jarvis, J.D. Tate, J.S. 28. W.G. Fateley, R. Sobczysnki, J.V. Paukstelis, A.N. Mor-
White, J.V. Paukstelis and W.G. Fateley, SPIE Proc., 1336, tensen, E.A. Orr and R.M. Hammaker, The Application of
124 (1990). Hadamard Masking for Multi-dimensional Studies in the
16. A.P. Bohlke, J.M. Jarvis, J.S. White, J.D. Tate, J.V. Pauk- Visible and Near-infrared Spectral Regions, in Spectropho-
stelis, R.M. Hammaker and W.G. Fateley, J. Mol. Struct., tometry, Luminescence and Colour; Science and Compli-
247, 293 (1991). ance, Analytical Spectroscopy Library, eds C. Burgess and
D.G. Jones, Elsevier, Amsterdam, 315335, Vol. 6 (1995).
17. R.M. Hammaker, A.P. Bohlke, J.M. Jarvis, J.D. Tate, J.S.
White, J.V. Paukstelis and W.G. Fateley, Whats n in Near- 29. R.M. Hammaker, A.N. Mortensen, E.A. Orr, M.K. Bellamy,
infrared Hadamard Transform Spectroscopy, in Analyti- J.V. Paukstelis and W.G. Fateley, J. Mol. Struct., 348, 135
cal Applications of Spectroscopy II, eds A.M.C. Davies (1995).
and C.S. Creaser, The Royal Society of Chemistry, London, 30. M.K. Bellamy, A.N. Mortensen, R.M. Hammaker and W.G.
260268 (1991). Fateley, NIR News, 6(6), 10 (1995).
18. S.A. Dyer, R.M. Hammaker and W.G. Fateley, Hadamard 31. M.K. Bellamy, A.N. Mortensen, E.A. Orr, T.L. Marshall,
Methods in Signal Recovery, in Computer-enhanced Ana- J.V. Paukstelis, R.M. Hammaker and W.G. Fateley, Mikro-
lytical Spectroscopy, ed. P.C. Jurs, Plenum, New York, chim. Acta, Suppl. 14, 759 (1997).
3167, Vol. 3 (1992).
32. M.K. Bellamy, A.N. Mortensen, R.M. Hammaker and W.G.
19. S.L. Wright, R.M. Hammaker and W.G. Fateley, Appl. Spec- Fateley, Appl. Spectrosc., 51, 477 (1997).
trosc., 47, 272 (1993).
33. R.A. DeVerse, T.A. Mangold, R.M. Hammaker and W.G.
20. J.D. Tate, J.V. Paukstelis, R.M. Hammaker and W.G. Fateley, Fateley, Chemical Mapping Using Two-dimensional Hada-
J. Mol. Struct., 300, 385 (1993). mard Transform Raman Spectrometry, in Proceedings of
21. R. Sobczynski, J.L. Porter, R.M. Hammaker and W.G. Fate- the 11th International Conference on Fourier Transform Spec-
ley, Laser Focus World, 31(3), 75 (1995). troscopy, ed. J.A. de Haseth, American Institute of Physics,
22. R.M. Hammaker, A.P. Bohlke, R.D. Freeman, J.M. Jarvis, Woodbury, NY, 443446 (1998).
K.J. Latas, D. Lin-Vien, A.N. Mortensen, E.A. Orr, J.D. Tate, 34. R.A. DeVerse, R.M. Hammaker and W.G. Fateley, Vib. Spec-
D.C. Tilotta, J.S. White, S.L. Wright, J.V. Paukstelis and trosc., 19, 177 (1999).
W.G. Fateley, Hadamard Transform Raman Spectrometry,
35. R.A. DeVerse, R.M. Hammaker and W.G. Fateley, J. Mol.
in Modern Techniques in Raman Spectroscopy, ed. J.J.
Struct., 521, 77 (2000).
Laserna, John Wiley & Sons, Chichester, 143226 (1996).
36. W.G. Fateley, R.M. Hammaker and R.A. DeVerse, J. Mol.
23. R.A. DeVerse, R.M. Hammaker, W.G. Fateley, J.A. Graham
Struct., 550551, 117 (2000).
and J.D. Tate, Am. Lab., 30(21), 112S (1998).
24. R.A. DeVerse, Multiplexed Hyperspectral Imaging and Spec- 37. J.A. Decker, Appl. Opt., 10, 510 (1971).
trometry Using Spatial Light Modulators, PhD Dissertation, 38. R.A. DeVerse, R.M. Hammaker and W.G. Fateley, Appl.
Kansas State University, Manhattan, KS (1999). Spectrosc., 54, 1751 (2000).
Infrared-emitting Diodes for Near-infrared
Spectrometry
Todd C. Rosenthal
Zeltex, Inc., Hagerstown, MD, USA
1 THE EARLY DAYS OF the detector. All heat generated by the rest of the source
NEAR-INFRARED SPECTROMETRY energy must be dissipated by the spectrometer. A further
disadvantage to the use of QTH lamps is the fact that their
lifetime is limited to between about 1000 and 10 000 hours.
The application of near-infrared (NIR) spectroscopy for
All the original instruments made in the 1970s worked
the rapid quantitative analysis of commodities started in
in the longer wavelength NIR region, between 1100 and
the 1970s when Karl Norris and his colleagues at the US
2500 nm, and the grain analyzers based on these spectrom-
Department of Agriculture demonstrated that this technique
eters operated in the diffuse reflection mode. This required
could be used for determining protein and moisture in wheat
the grain sample to be first ground to a coarse powder,
and barley. The first commercial NIR analyzers were used
with the typical particle size being between 50 and 200 m.
at grain elevators and terminals where conventional wet
In these spectrometers, light from the source was typically
chemistry laboratories could not be used.
passed through the filter or reflected off a grating to pro-
The first generation of NIR analyzers were based
duce the desired analytical wavelength. The radiation of
on the use of a broadband NIR light source, usually
these wavelengths was then reflected from the powdered
a quartztungstenhalogen (QTH) lamp. The analytical
sample and focused on a lead sulfide detector.
wavelengths were selected by spinning narrow-band-pass
filters located in wheels (see Figure 1) or tilting filter wheels
(see Figure 2). These instruments were quite large and 2 NIR SPECTROMETERS BASED ON
cumbersome in design and relatively slow, since a motor
INFRARED-EMITTING DIODES
was needed to turn the filter wheel to the exact location
before the energy at that wavelength could be measured.
A major breakthrough in the development of small portable
The QTH bulb was continuously illuminated while the
NIR spectrometers was made in 1979 when it was shown
instrument was on. This caused a considerable amount of that the QTH lamp and the spinning filter wheel, tilting fil-
heat that had to be dissipated by the instrument. Although ter wheel and grating monochromator could be replaced by
the introduction of grating monochromators (see Figure 3) low-power infrared emitting diodes (IREDs). IREDs oper-
improved the resolution of filter-based instruments, they did ated at a much shorter wavelength (8001100 nm) than
not eliminate the need for moving optics. had been used in previous grain analyzers. This allowed
Whether or not the instruments incorporate spinning very sensitive and inexpensive silicon detectors to be used
filters, tilting filters or grating monochromators, they are instead of the relatively expensive and insensitive PbS
all based on the QTH lamp, which radiates in the infrared, detectors.
visible and ultraviolet regions of the spectrum, with only Originally developed because of a need by the telecom-
a very small fraction of desired NIR radiation reaching munications industry, IREDs emit light over a much smaller
band-pass than the QTH lamp. The typical full width at
John Wiley & Sons Ltd, 2002. half-height (FWHH) of the emission profile of an IRED
Light is 50 nm. The wavelength resolution emitted by a typical

source
IRED is too low to obtain good analytical results. For
ground grains, for example, the resolution of an NIR spec-
trometer should be less than 15 nm if accurate quanti-
tative results of ground grains are to be found by NIR
spectroscopy. Thus the output of each IRED is passed
Disk with
optical filters through an appropriate band-pass optical filter to limit the
spectral range to 12 nm. The amount of heat needed to
be dissipated by an IRED-based NIR spectrometer is far
less than that of a conventional NIR spectrometer equipped
Optical detectors
with a QTH lamp. The emission profile of several IREDs
is shown in Figure 4.
The smaller the FWHH of the instrument line-shape func-
Figure 1. Schematic of an NIR analyzer incorporating a number
of rotating narrow-band fixed filters.
tion, the greater is the accuracy at which the analytical
wavelength must be set and the less energy that must be dis-
sipated by the spectrometer. Most IRED-based instruments
Light
are currently made by mounting a very small interference
source
filter with a band-pass of about 12 nm directly in front of
Lens the IRED (Figure 5). The optimum FWHH of the filter
Wideband or monochromator band-pass involves a trade-off between
light beam tight wavelength control, the effect of resolution errors
and the need to pass the maximum optical power to the
Tilting filter
wavelength varies
as filter turns 1.0
Narrowband
light beam 0.8
Relative output
Optical detector 0.6
0.4
Figure 2. Schematic of an NIR analyzer incorporating a number
of narrow-band fixed filters. 0.2
0.0
800 900 1000 1100
Exit slit
Wavelength /nm
Figure 4. Normalized emission profiles of typical IREDs.
Collimating Narrow band-pass Sample holder

mirror filters
Detector
Folding x Zel
Grating Tex
mirrors Inc x xxx xxx
xxxxx x xxxx
xxxx x
x Thermistor
IRED
Diffuser
Entrance slit
Micro computer 93.2
Light source
Figure 5. Schematic of an NIR analyzer based on a number of
Figure 3. Schematic of a monochromator-based NIR analyzer. combinations of IREDs and narrow band-pass filters.
Infrared-emitting Diodes for Near-infrared Spectrometry 3
IREDs illuminate a limited region of the spectrum

Control
at only the desired wavelengths, reducing heat and
Diffraction electronics deterioration of filters or gratings.
LED array
grating The power consumption of IREDs is very low, enabling
their use in battery-powered, portable instruments.
Exit slit 3 APPLICATIONS OF IRED

Sample SPECTROMETERS
Spectrometer module Detector
Figure 6. Schematic of an NIR analyzer based on a number of Particle size has an important role in diffuse reflection or
combinations of IREDs and a grating monochromator. (Repro-
duced by permission of Jaiko Malinen.) transmission spectroscopy. The weaker the absorptivity of
a sample, the greater the pathlength that is needed. In the
short-wavelength region of the NIR, where most IREDs
detector in order to obtain the highest possible signal-to-
operate, absorption of radiation by organic samples is due
noise ratio. The instrument is made by having a number
to the third and fourth overtones of CH, OH and NH
of the IRED/filter combinations with each IRED being
stretching bands and their associated combination bands.
switched on and off sequentially. The typical time in which
Although spectra in this region are not as easily interpreted
each IRED is illuminated is 18 ms, so even when 32
IRED/filter units are being used, the total measurement time as mid-infrared or long-wavelength NIR spectra, they are
is about 0.6 s. Admittedly the limited wavelengths of IREDs just as unique and can be used for quantitative analysis
do not allow a complete NIR spectrum to be measured, but in the same way. Moreover very long pathlengths can be
allow the instrument to concentrate on the wavelengths that used for the measurements of the weak bands in the short-
are known to be analytically important. wavelength NIR.
Small diffraction gratings are also being used for this pur- The limited number of wavelengths makes the knowledge
pose on an experimental basis (Figure 6). In this approach, of the specific constituents to be analyzed by the instrument
light from an array of IREDs is reflected off a grating onto a critical in determining which wavelengths to install in an
silicon detector to allow the analytical signal to be generated instrument. Instruments utilizing IRED narrow-band-pass
by turning on different IREDs at specific intervals. filter instruments are best when used for specific analysis,
Diodes are available for many different wavelengths rang- not research. They are not designed to analyze constituents
ing from the visible (where they are called light emitting in trace amounts and all possible constituents in products.
diodes) to over 2900 nm for laser diodes. IREDs are available A limited number of filters means a limited number of
relatively inexpensively for many different wavelengths. wavelengths which in turn means limited analysis capa-
In summary, the unique characteristics of IREDs include bility. Nonetheless, all NIR emitting diode spectroscopy
the following: instruments can use multiple linear regression, partial least
squares or any other types of mathematics.
IREDs can be rapidly switched on and off allowing The limitation at the present time is the number of
different wavelengths to be examined without any samples needed for calibration. For example, needing a
moving filters or gratings. minimum of six samples per wavelength per calibration
IREDs permit spectroscopic measurements to be made means that, if you are using a 14-filter instrument, you
in the 800 to 1100 nm region using a single silicon must have a minimum of approximately 85 samples; 10
detector, which is both sensitive and inexpensive. samples per wavelength, 140 in all, are recommended.
The combination of high optical power in the spectral The requirement of large sample sets for calibration has
region of interest combined with the use of a silicon slowed the progress of these instruments towards becoming
detectors extreme sensitivity (up to 7.0 absorbance a standard product with consumers.
units), allows IRED-based spectrometers to analyze As of this writing, all IRED-based commercial NIR ana-
samples that are normally considered opaque. IRED- lyzers are individually calibrated. Universal transfer of cal-
based instruments provide as much as three orders of ibration constants is needed to be achieved for these instru-
magnitude more sensitivity than a conventional NIR ments to reach their full potential. This calibration transfer
spectrometer with PbS detectors. research is on-going and hopefully will be solved shortly.
The lifetime of IREDs is nearly unlimited (100 000 to The original IRED-based spectrometer was introduced
1 000 000 h), in contrast to the normal lifetime of 1000 in 1979 by Trebor Industries (Gaithersburg, MD) for use
to 10 000 h for a QTH bulb. as a whole grain analyzer, which eliminated the need to
capability to run either on batteries or on alternating current

it allowed the analyzer to be brought to the sample rather
than the sample having to be brought to the analyzer.
The FUTREX-1100 was the first NIR analyzer specifi-
cally designed to be used by the public instead of by trained
professionals. The FUTREX-1100 had a sale price of less
than $100. It was designed to allow an individual, in the pri-
vacy of his home, to measure and analyze his percent body
fat. There are approximately 100 000 of these NIR personal
percent body fat analyzers in use in homes worldwide.
The development of these IRED instruments led to porta-
bility for the application where this technology was first
applied, namely the grain industry. Whereas only a few
wavelengths are needed for measuring body fat, the mea-
surement of multiple components in grain and food products
requires 12 to 14 wavelengths. In the 1990s, the telecom-
munications industrys need for IREDs for fiber-optic com-
Figure 7. The TREBOR 90 whole grain analyzer.
munication continued to increase, and smaller diodes with
higher power than before became available. For IRED-
grind the wheat prior to the measurement. The TREBOR 90
based NIR analyzers, this allowed more IRED/filter combi-
allowed the NIR radiation to be transmitted through a sam-
nations (as many as 39) to be packaged than was previously
ple of unground wheat (2 cm wide) instead of diffusely
possible. This new capability led to the development of a
reflecting the radiation from the ground sample (Figure 7).
portable device for the rapid measurement of the octane
From the late 1980s, the potential of IRED-based NIR
number of gasoline, the Zeltex ZX101C Portable Octane
analyzers started to be realized. The first portable NIR Analyzer (Figure 9). This device has 14 narrow band-pass
analyzer designed for consumer use was developed by filter/IRED combinations. It is currently being used by a
Futrex, Inc. This device was designed to measure percent majority of the states in the United States to screen gaso-
body fat in the human body. Using a simple design with line for octane levels for compliance with state and federal
only two wavelengths, the water to fat ratio of 940 nm regulations. It has also achieved international approval from
and 950 nm was used to predict the percent body fat the Russian State Committee for Standards, Center of Stan-
(Figure 8). The instrument was sold commercially as the dardization, Metrology and Certification of N. Novgorod.
FUTREX-5000 Body Fat Tester for a cost of under $3000. With portability came a new possibility for the quality
This instrument was the first truly portable NIR analyzer. control of a variety of products using this technology.
Currently there are over 17 000 of these instruments in
use in the medical and fitness industries worldwide. The
0.4
Low fat
High fat person
person
0.3
Log (1/I )
0.2 Pure Pure

fat water
0.1
0.0
750 800 850 900 950 1000 1050 1100
Wavelength /nm
Figure 8. Comparison of body-fat measurements. (Reproduced
by permission of Robert Rosenthal.) Figure 9. Zeltex ZX-101C portable octane analyzer.
Infrared-emitting Diodes for Near-infrared Spectrometry 5
While other portable NIR analyzers have been reported, identification of solvents and other raw materials and prod-
it is probably true to say that the only ones that can operate ucts on a loading dock. Such an instrument has never been
from low-power batteries (AA in the USA) are based sold commercially but illustrates one possible future direc-
on IREDs. Most other portable instruments still require tion for this technology.
either alternating power or a large gel-cell direct current
power supply in order to supply sufficient power for a
QTH bulb. ABBREVIATIONS AND ACRONYMS
IRED Infrared Emitting Diode

4 SUMMARY QTH QuartzTungstenHalogen
The use of IREDs as light sources for NIR analyzers is

still in its infancy. The number of possible applications for REFERENCES
the future for a low-cost, low-energy illumination source is
enormous. For example, Bonanno and Griffiths1,2 demon- 1. A.S. Bonanno and P.R. Griffiths, J. Near Infrared Spectrosc.,
strated the feasibility of identifying 30 different solvents 1, 13 (1993).
rapidly and unequivocally using this technology. An instru- 2. A.S. Bonanno and P.R. Griffiths, Appl. Spectrosc., 49, 1590,
ment based on this concept could be used for rapid, safe 1595 (1995).
Raman Monochromators and Polychromators
Michael J. Pelletier
Kaiser Optical Systems, Inc., Ann Arbor, MI, USA
1 INTRODUCTION The focal ratio, or f -number, is the distance from a focal

point to a collimating (or focusing) optic divided by the
Light collected from a sample needs to be processed in utilized diameter of that optic. It is written as f /#, where #
some way so that wavelength and intensity information can is the value of the focal ratio. For example, a lens having
be extracted from the detected signal. The three types of a focal length of 100 mm and a diameter of 25 mm has a
instrument commonly used to do this are monochromators, focal ratio of f /4. If only a diameter of 10 mm of that f /4
polychromators, and interferometers. The use of interferom- lens is being used, the f /4 lens is functioning at f /10.
eters is discussed in detail in a separate article (see Fourier Numerical aperture (NA) and solid angle () are also
Transform Near-infrared Raman Spectroscopy). This used to describe the optic diameter and its focal length.
article will describe monochromators and polychromators The relationship between f -number, numerical aperture,
that have been used for Raman spectroscopy. and solid angle are given by equations (13):
D 2p1 cos q 1

L 1
1.1 Terminology f# D D tan q 2
2R 2
Monochromators block all but a narrow spectral region NA D n0 sin q D [4f#2 C 1]0.5 3
from reaching the detector. Spectra can be built up by
where is the solid angle between focal point and optic in
measuring the transmitted intensity as the spectral region
steradians, q D half-angle between focal point and the optic
transmitted by the monochromator is changed with time.
in degrees, f# D f-number of the optic, L is the distance
Polychromators separate different wavelengths of light
between focal point and the optic, R is the aperture radius
and deliver them to different detectors for simultaneous
of the optic, NA is the numerical aperture of the optic, and
measurement. Spectra can be built up by making a histo-
n0 is the refractive index of the medium between the focal
gram of the signals from the different detectors. The
point and the optic.
term spectrometer is used in this article to include both
monochromators and polychromators. Spectrometers can
be classified as dispersive and nondispersive. Dispersive
1.2 Article layout
spectrometers deliver light to a position that varies con-
tinuously with wavelength. A dispersive polychromator is The rest of this article is organized as follows. Section 2
called a spectrograph. Nondispersive spectrometers include discusses the critical performance requirements that guide
everything else, such as interferometers, dichroic beam the design of Raman spectrometers. Section 3 describes the
splitters, optical filters, and energy-selective detectors. The optical components that largely determine the performance
only nondispersive spectrometers considered in this article of Raman spectrometers. Special emphasis is placed on
are those based on dichroic beam splitters and optical filters. diffraction gratings because they are the key component
of most modern Raman monochromators and polychro-
John Wiley & Sons Ltd, 2002. mators. Section 4 discusses some design concepts common
2 Instrumentation for Raman Spectroscopy
to dispersive monochromators and polychromators. Finally, detector element leads to an additional loss in transmission.
Sections 5 and 6 discuss specific monochromator and poly- Spectrometer imaging quality is best described by the
chromator designs. point spread function. The point spread function is the
spatial energy distribution at the spectrograph exit plane
that is created by an ideal point source at the entrance
2 CRITICAL PERFORMANCE NEEDS aperture. A convolution of the entrance aperture with the
point spread function gives the image at the exit aperture.
The most important properties of spectrometers used for Complete point spread function information for Raman
Raman spectroscopy can be summarized in a small number spectrometers is rarely available. However, the limited
of critical performance specifications. Ideally, all of these qualitative information about it, which usually is available,
properties would be simultaneously optimized to the funda- is sufficient for most needs.
mental limits imposed by physics. Unfortunately, this is not Other spectrometer transmission loss mechanisms include
yet possible. Furthermore, added capability and operational vignetting and polarization dependent loss. Vignetting is
flexibility usually lead to added cost and maintenance. the loss of light delivered to an optical element that occurs
The art in building or selecting Raman instrumentation because some of the light is outside the acceptance aper-
is in finding the right balance between capability, perfor- ture of the optical element. In other words, part of the light
mance, and cost, as defined by the needs and value of the directed to an optical element misses the target, and there-
intended use. Selecting research instruments is especially fore never reaches the detector. Spectrometer vignetting
difficult because they are often purchased to do things often causes transmission at the ends of a spectrum to be
that cannot be defined in advance. This section reviews less than that in the middle of the spectrum. Transmission
the properties that are generally most important for Raman loss that depends on the polarization of the light is primarily
spectrometers. due to the polarization dependence of grating diffraction
efficiency.
Etendue, the product of image area and the solid angle
2.1 Efficient use of photons of light at a focal plane, is another factor that determines
how much light from the sample can be imaged onto the
The signal-to-noise ratio of Raman spectra is often limited detector. Magnification or demagnification of an image
by the number of Raman photons that can be detected. changes the image area and the solid angle of light leaving
Raman spectrometers should, therefore, deliver a large frac- the image, but not the etendue. It can be proved that the
tion of the Raman photons collected from the sample to etendue of the entire optical system, from the sample to the
the detector. Two specifications, transmission and etendue, detector, cannot exceed the smallest etendue at any point in
indicate how efficiently Raman spectrometers do this. the optical system. This limiting etendue plays a major role
Transmission is the fraction of the light intensity prop- in Raman spectrometer design. It is worth noting that the
erly coupled into the input of the spectrometer that reaches concept of etendue also holds in non-focal plane locations,
the spectrometer output. Etendue is a combined mea- but is harder to calculate.
sure of the sample image size and the solid angle of Consider the etendue of a Raman microscope using a
light from the sample that can be transmitted through the 0.85 NA objective lens. The illuminated spot on the sample
spectrometer. may be 1 m or less in diameter. If the sample image at
Ideally, the transmission is the product of the optical the entrance slit of a spectrometer is magnified 25 times
efficiencies of the optical components in the optical path. to fill a 25-m slit, light exits the slit at a focal ratio of
For example, a spectrograph consisting of a diffraction greater than f /12. A lower focal ratio spectrometer, with
grating with 50% diffraction efficiency, two mirrors with greater etendue, would provide no more light to the detector
90% reflectivity each, and a lens having 95% transmission because the sample has the limiting etendue. The situation
would ideally have a transmission of 0.50 0.90 0.90 is different when analyzing light exiting an optical fiber.
0.95 D 38.5%. A typical optical fiber may have a diameter of 100 m and
Imperfect imaging inside a spectrometer may cause the emit light into an f /2 cone. If the image of the fiber were
transmission to be less than that predicted by the efficiencies focused with unity magnification onto a 100-m entrance
of the optical components. For example, a blurred image slit of an f /4 spectrometer, much of the light from the
of the entrance aperture at the exit plane could be too optical fiber would be lost. Here, reducing the focal ratio
large to be fully transmitted by a monochromator exit slit of the spectrometer would increase the amount of light
or to be fully captured by a small spectrograph detector reaching the detector because the spectrometer has the
element. The light blocked by the exit slit or missed by the limiting etendue.
Raman Monochromators and Polychromators 3
2.2 Stray light type and accuracy of the assumptions, the signal-to-noise
ratio of the measured spectrum, and the calibration stabi-
Light that does not go where it is supposed to go not only lity of the spectrometer. Raman band positions and widths
reduces spectrometer transmission, but may also appear in can also be measured to much greater precision than the
the spectrometer where a different color of light should be. Nyquist-limited resolution.4,5 Measurements at or beyond
This misdirected light, called stray light, distorts Raman the Nyquist resolution limit are susceptible to artifacts, so
spectra. Stray light may diffusely illuminate the spectro- it is generally desirable to have five or more sample points
meter exit plane, causing the addition of a broad back- to define each Raman band.
ground to Raman spectra. Stray light from specular reflec-
tions can add structured artifacts to Raman spectra. Even
diffuse stray light can add structured artifacts to Raman 2.4 Calibration stability
spectra owing to shadow formation on the spectrometer exit
plane. Laser light from the sample is the most important An important characteristic of Raman spectrometers is how
source of stray light in Raman spectrometers because it is well they remain calibrated. Unfortunately, quantitative
usually 104 1012 times more intense than the Raman light. specifications for comparing the calibration stability of dif-
Elimination of stray light artifacts is crucial for Raman ferent Raman spectrometers are rarely available. Raman
spectrometers. Unfortunately, stray light specifications for spectrometer calibration can be degraded over time by
Raman spectrometers can be confusing because of the dif- several factors including temperature change, vibration,
ferent ways stray light can be described. contamination of optical surfaces, and even the normal
operation of the spectrometer itself. Some spectrometer
optical design aspects are inherently more sensitive to
2.3 Spectral resolution these factors than others. For example, reflective optics
are generally much more sensitive to angular misalign-
The spectral resolution of a spectrometer is the difference ment than refractive optics. Movable optics tend to be less
in wavelength needed by two equally intense spectral lines positionally stable than permanently mounted optics. Focal
to be recognized as two separate lines, rather than just plane positional stability tends to decrease with increas-
one. Rayleighs criterion for this was a local minimum ing distance between the entrance aperture and the focal
between the two spectral lines of 8/p2 D 0.811 times the plane. Following good design practices minimizes cali-
maximum intensity. This corresponds to the maximum of bration drift and failure over time for any type of spec-
one sinc2 function overlapping the first minimum of another trometer. Good design practices include strict adherence
equally intense sinc2 function. This criterion was physically to kinematic design principles, minimization of moving
meaningful for Rayleighs work because he used prism parts, attention to material thermal properties, attention to
or grating spectrometers. The point spread function for an component natural vibrational frequencies, and isolation of
ideal prism or grating spectrometer is a sinc2 function. The critically positioned parts from external forces and sources
point spread function of modern Raman spectrometers is of contamination.
rarely a sinc2 function, however, so the choice of Rayleighs
criterion today is somewhat arbitrary. Precise definitions
of spectral resolution are elusive,1 and often misused, so 2.5 Data acquisition speed
quantitative published resolution specifications must be
used with caution. Raman measurements used to be very time consuming.
The best possible spectral resolution of a spectrometer is Each spectral point had to be measured sequentially. The
often determined by the diffraction limit of light. Section 3 only array detector was photographic film, but even there
gives an equation for this optically limiting spectral re- the positions of bands on the developed film had to be
solution when diffraction gratings are used. The Nyquist measured sequentially. Modern parallel electronic detection
sampling theorem places an additional constraint on the has changed all that. Now 2000 or more spectral points
maximum possible resolution for a Raman spectrometer. can be measured simultaneously. If all other things were
The resolution cannot be better than the wavelength inter- equal, such a parallel measurement would be 2000 or more
val covered by twice the wavelength sampling interval. times faster than the corresponding set of sequential mea-
Interpolation combined with mathematical super-resolution surements. For example, a 1-s measurement today using
techniques,2,3 along with certain assumptions about the a 2000-channel parallel detector would take 2000 s (more
spectrum, can improve the spectral resolution beyond the than 33 min) using a single detector and sequential measure-
Nyquist limit. The spectral resolution then depends on the ments. In fact, the benefit of parallel detection should be
even greater because of the unavoidable dead time between 2.6 Operational flexibility
sequential measurements. All other things are not necessar-
ily equal, however. Increasing the number of different things a Raman spectro-
The benefit of the faster data acquisition speed of a meter can do will usually compromise its performance on
multichannel spectrometer, compared with a single-channel any one of those things. For example, the use of protected
spectrometer, can be reduced if the etendue of the mul- aluminum for reflective surfaces in a Raman spectrom-
tichannel system is less than that of the single-channel eter will allow the spectrometer to be used in spectral
spectrometer. Reduced etendue can lead to a reduced num- regions ranging from the ultraviolet (UV) through to the
ber of detected photons per second per channel, which in near-infrared (NIR). Gold reflective surfaces do not per-
turn results in a longer acquisition time to collect the same form well at wavelengths shorter than about 550 nm, but
number of photons per channel. The reason for reduced they significantly outperform protected aluminum surfaces
etendue in multichannel spectrometers is that parallel detec- for wavelengths longer than about 700 nm. Industrial ana-
tors have smaller detector elements than single detectors. lyzer applications tend to be very specific and repetitive,
For example, the effective size of a detector element so the design of their Raman spectrometers can be opti-
of a charge-coupled device (CCD) detector, the standard mized for a very limited range of operating parameters (e.g.
parallel detector for Raman spectroscopy, is often about spectral coverage, spectral and spatial resolution, etendue).
0.025 mm 0.1 mm. The photocathode of an R928 pho- Research instruments need to be more flexible so they can
tomultiplier, a commonly used single channel detector for adapt to evolving needs. This flexibility usually degrades
Raman spectroscopy, is about 8 mm 24 mm (a factor of optical throughput, though.
76 800 larger in area!). Even though Raman spectrometers
have never fully utilized the possible etendue of an R928
photomultiplier, Raman spectrographs using parallel detec- 2.7 Cost
tion have to be very well designed to match the etendue
typical of single-channel systems. The reduced etendue of The current cost of a spectrometer suitable for making a
the spectrometer may not have any effect on data acquisi- Raman measurement ranges from a few dollars in the case
tion speed, however, if the other parts of the Raman system of a glass filter to over $200 000 for some types of state-of-
have less etendue than the spectrometer. This is usually the the-art triple spectrographs. Market forces tend to eliminate
case for Raman microscopes. spectrographs that violate the principle of you get what you
Etendue loss caused by small detector size can be recov- pay for. The challenge, then, is to select the spectrometer
ered by using a larger solid angle of light. This is the that provides the needed capabilities and support, but does
motivation for building low f /# spectrographs. Optical aber- not have added features that are not likely to be used.
rations increase rapidly, however, with decreasing f /#. If
optical aberration causes the point spread function to exceed
the size of the detector elements, adjacent detector elements 3 KEY COMPONENTS
are no longer independent. The effective number of detec-
tor elements is reduced, leading both to a reduction of the 3.1 Diffraction gratings
possible multichannel data collection speed benefit and to
a loss of spectral resolution. A spectrograph point spread Before high-quality diffraction gratings were available,
function tends to be worse near the ends of a multichannel Raman spectrometers used prisms to separate different
detector than at its center, so a calculation of the effec- wavelengths of light. The prisms have been replaced
tive number of independent detection channels based on by diffraction gratings in modern Raman spectrometers.
the point spread function at the center of the detector may Diffraction gratings6 provide a much better combination
significantly overestimate the true multichannel benefit of of optical throughput and spectral resolution than prisms
the multichannel Raman system. do. For example, a 50 mm 50 mm diffraction grating with
Finally, the multichannel benefits only apply to those 1800 grooves per millimeter provides 16 times the max-
channels that provide useful information. The advantage imum spectral resolution of a typical 50-mm dispersing
of multichannel detection over single-channel detection is prism at 550 nm.7 More important, the grating also provides
greatly reduced when only a small part of the Raman much greater angular dispersion (discussed below).
spectrum is needed. This is why single-channel detection There are many different kinds of diffraction grating.
is often used for dedicated gas-phase applications. There is They can be classified as reflection or transmission, sur-
little benefit to using multichannel detection when only a face or volume, ruled or holographic, and flat or concave.
few channels need to be measured. Reflection gratings have the incident and diffracted light
on the same side of the grating surface, while transmission (electric vector of light parallel to grating grooves) because
gratings have the incident and diffracted light on opposite diffraction efficiency is often very sensitive to the optical
sides of the grating surface. Surface gratings have grooves polarization.
on the surface of a reflecting or transmitting material, while Imperfections and contamination on the grating surface
volume gratings use variations in refractive index and/or cause light at any wavelength to scatter over a wide range
absorbance through the depth of a material in place of of angles. This scattered light adds a small background to
grooves on the surface to produce diffraction. Ruled grat- the spectrum measured by a Raman spectrometer. Even a
ings are manufactured by mechanically cutting or emboss- small background from scattered laser light can overwhelm
ing grooves into the substrate surface, while holographic the Raman signal, though, since the laser intensity may be
gratings (surface or volume) are manufactured by crossing more than 1010 times greater than the Raman intensity. A
two laser beams to produce an interference pattern on some chief concern in Raman spectrometer design is minimizing
kind of photosensitive material to create the grooves. Holo- the impact of light scattered from the diffraction grating.
graphic gratings tend to have more precise groove spacing Detailed analyses of diffraction gratings are presented
and groove shape than ruled gratings. Surface holographic elsewhere6,9,10 and are beyond the scope of this article.
gratings also tend to produce less scattered light and can Many of the results of such analyses that are important
have more complex groove patterns than ruled gratings. Flat for Raman spectrometer performance are summarized by
gratings have no optical power. Collimated light incident on the seven equations in Table 1. Equation (4) is the basic
a flat grating will produce collimated diffracted light whose equation describing diffraction from a grating when the
direction depends on wavelength. Concave gratings do have incident light is in the plane normal to the grating grooves.
optical power. Light diverging from a point source that is When the incident and diffracted light are on the same
incident on a concave grating can produce diffracted light side of the grating surface normal, the plus sign is used;
that converges to a point image whose location depends on otherwise, the minus sign is used.
the optical wavelength. Diffraction from a grating is described by equation (5)
Diffraction grating performance can be described by how when the incident light makes a nonzero angle, g, with
much of the input light goes into the desired diffraction the plane normal to the grating grooves.11 Only one point
order, and where the rest goes. The distribution of incident along the entrance slit height of a spectrometer can have a
energy between the different diffraction orders is largely g value of zero, where equation (4) may correctly replace
determined by the shape of the grating groove. The groove equation (5). The change in g with position in slit height
shape used by modern diffraction gratings directs most of causes a corresponding change in diffraction angle for a
the diffracted light into a single order. Gratings with these given wavelength of light. The image of the spectrometer
shaped grooves are called blazed gratings for historical rea- entrance slit aperture at the spectrometer exit plane is
sons. In the early 1900s, R.W. Wood made gratings with therefore curved. The curvature increases with increasing
shaped grooves that would light up or blaze when viewed slit height and decreasing spectrometer f /#. This effect can
at the correct angle.8 A particular blazed grating will work be significant for high-throughput spectrometers.
better at some wavelengths than others, so a wide range Equation (6) gives the upper limit for spectral resolution
of diffraction gratings are commercially available. Grat- from a diffraction grating. This resolution limit increases
ing manufacturers provide plots of diffraction efficiency with increasing grating ruled width, angle of incidence,
versus wavelength for their diffraction gratings. Separate and angle of diffraction. Raman spectrometer resolution is
curves are usually given for s-polarization (electric vector usually much less than that possible according to equa-
of light normal to grating grooves) and for p-polarization tion (6). Entrance and exit apertures are almost always
Table 1. Equations describing diffraction grating performance.
ml D asin a sin b Diffraction, traditional mount Equation (4)

ml D asin a sin b cos g Diffraction, conical mount Equation (5)
R D l/l D Wsin a sin b/l Maximum resolution Equation (6)
db/dl D sin a sin b/l cos b Angular dispersion, general case Equation (7)
db/dl D 2 tan b/l Angular dispersion, Littrow mount Equation (8)
Ma D cos a/ cos b Anamorphic magnification Equation (9)
F D s/m Free spectral range Equation (10)
Definitions: m D grating diffraction order, l D wavelength of light, a D distance between grating grooves,
a D angle of incidence from grating normal in plane perpendicular to grooves, b D angle of diffraction
from grating normal, g D angle of incidence from grating normal in plane parallel to grooves, W D ruled
width of grating, s D 1/l, optical frequency in wavenumbers, cm1 .
larger than optimal for maximum resolution in order to are actually sinusoidal variations in refractive index that
improve spectrometer throughput. The degraded spectral are produced holographically. These grooves are formed
resolution, however, is still proportional to the angular dis- throughout the thickness of the grating. Diffraction from
persion of the diffraction grating. The angular dispersion such a grating is actually Bragg scattering from the
of the diffraction grating, given as db/dl in equations (7) three-dimensional refractive index structure. This scattering
and (8), therefore has greater practical importance than the nevertheless follows the grating equations given in Table 1.
maximum possible spectral resolution. The intensity of the diffracted light depends on the depth of
When diffracted light nearly retraces the path of the inci- the refractive index modulation and on the phase matching
dent light, a ' b, and the grating is said to be in a Littrow condition of Bragg scattering. Just as X-ray diffraction
mount. Equation (7) then simplifies into equation (8). Most lines from crystals (produced by Bragg scattering) become
grating spectrometers operate in a Littrow mount, or nearly sharper as the crystal size increases, the bandwidth of light
so. Deviations from the Littrow condition of a D b cause efficiently diffracted by a volume holographic transmission
the image of the slit in the exit plane to be magnified in grating decreases with increasing grating thickness. The
width, but not in length. This anamorphic magnification is grating manufacturer can therefore adjust the spectral width
described by equation (9). Notice that anamorphic magnifi- that is diffracted by the grating. Optical wavelengths
cation effects are much greater for large angles of incidence outside this width are transmitted, rather than diffracted.
and diffraction, the conditions that maximize angular dis- Other grating properties, such as diffraction efficiency and
persion and spectral resolution. polarization dependence, can be controlled as well using
Equation (4) indicates that for a given grating and angle the remarkably flexible holographic processing techniques.
of incidence, different wavelengths can have the same angle Phase gratings, whether they are volume holographic
of diffraction. For example, 1000-nm light diffracted in gratings or surface transmission gratings, transmit light that
the first order (ml D 1 1000) will have the same angle is not diffracted. This transmitted light can be diffracted
of diffraction as 500-nm light diffracted in the second by a different phase grating having a different groove fre-
order (ml D 2 500). A single-grating Raman spectro- quency. If the grooves in one phase grating are not parallel
meter measuring light intensity in the 7851100-nm region to the grooves in the other phase grating, the diffrac-
would be unable to distinguish between 500-nm light and tion from the two gratings, when focused, can produce
one spectrum stacked on top of the other spectrum. A
1000-nm light. Some sort of order sorting filter would
single-volume holographic optical element, called a holo-
be needed to prevent wavelengths shorter than 1100/2 D
plex grating, composed of two thin transmission gratings is
550 nm from reaching the detector. A simple glass filter is
now commercially available.13,14
often used for this purpose. A more powerful order sorting
Diffraction gratings are usually made on optically flat
approach is described in Section 6. The continuous range
substrates. They can also be made on concave substrates.
of wavelengths free from overlap of adjacent diffraction
Such gratings are called concave gratings. These grat-
orders is called the free spectral range, F, and is described
ings combine the spectrometer functions of collimation,
by equation (10).
diffraction, and focusing into a single optical element. Holo-
Diffraction gratings are usually operated in their first or
graphic concave surface diffraction gratings can be made
second order. This maximizes their free spectral range.
that also correct optical aberrations such as spherical aber-
Echelle gratings are an important exception. They are
ration and coma. The correction can be perfect only at a
blazed to operate at high angles of incidence and diffraction,
single wavelength, but is reasonably good over an extended
yet have a relatively low groove density. As a result, they spectral region. Concave surface holographic gratings nor-
have high spectral resolution and angular dispersion, but mally have a sinusoidal groove shape. These gratings can
a very small free spectral range. For example, a typical in some cases have diffraction efficiencies as high as 85%.
echelle grating may be 128 mm wide, have 79 grooves per Less efficient gratings can be blazed using ion-etching tech-
mm, and be blazed for operation at 63.5 . This grating niques. Concave gratings are especially useful in the deep
would have a maximum resolution at 550 nm of 0.0013 nm UV spectral region because the efficiency of collimating
or 0.04 cm1 in the 41st diffraction order and a free spectral and focusing optics is poor in this spectral region.
range of only 440 cm1 . Unlike gratings that operate in their
first or second diffraction order, echelle gratings are blazed
for efficient operation at all optical wavelengths. 3.2 Mirrors
Volume holographic transmission gratings are relatively
new optical components for the design of Raman spectro- Mirrors in Raman spectrometers are made of thin films of
meters.12 The diffraction efficiency of these gratings can aluminum, silver, or gold on a suitable substrate. Bare alu-
in some cases be very close to 100%. Their grooves minum is the best general-purpose reflecting surface with
good reflectivity from the deep UV through to the NIR. their cost is extremely low given their sophistication and
Freshly deposited aluminum has good reflectivity down to performance. Many camera lenses have been designed for
90 nm, but the formation of an oxide layer causes the reflec- a 35-mm-wide focal plane, which is significantly better
tivity to fall below 220 nm. Aluminum mirrors also have a than required by most Raman spectrometers. Higher-quality
weak absorption band near 800 nm. Bare aluminum mir- camera lenses have focal plane resolutions better than
rors are soft and must be handled carefully. An aluminum 80 lines per millimeter, which is a good match for typical
mirror is usually coated with a 0.2-m-thick film of sili- CCD detector pixel sizes and monochromator exit slit sizes.
con oxide to retard aluminum oxide formation and to make The optical transmission of camera lenses tends to be
the mirror more durable. These protected aluminum mirrors in the 9096% range. The limitations of camera lenses
are the most commonly used mirrors for Raman spectrom- compared with concave mirrors are much higher optical
eters. They have good reflectivity from about 300 nm to the scattering, smaller aperture, and a more limited wavelength
infrared. range over which good performance can be obtained. The
Silver mirrors have a higher reflectivity than aluminum useful aperture of a multielement camera lens can be less
mirrors in the visible and NIR spectral regions, but tarnish- obvious than that of a simple concave mirror. Spectrometer
ing can quickly degrade that reflectivity. A thin dielectric vignetting caused by camera lenses can therefore easily be
film over the silver mirror (along with a good edge seal) pre- overlooked. A Raman spectrometer using low f /# camera
vents tarnishing and makes the mirror more durable. Gold lenses rather than concave mirrors was first reported in
mirrors have excellent reflectivity in the NIR, but are poor 1989.15 Most dispersive Raman vendors now offer lens-
in the blue and green spectral regions. Gold mirrors do not based Raman spectrographs.
tarnish, but are very soft and easily scratched. A thin dielec-
tric film is often deposited on top of gold mirrors to make
them more durable. Gold does not adhere well to glass, 3.4 Focal reducers
so gold mirrors are made by first depositing a chromium
layer on the glass and then depositing the gold onto the A focal reducer is a system of lenses and/or mirrors that
chromium. There may be a step change from chromium collect the large, high f /# output of a high focal ratio spec-
to gold or a graded increase in gold concentration from 0 trograph and produce a smaller, lower f /# image. Focal
to 100%. reducers are commonly used in astronomy, but are rarely
Concave mirrors are often used in Raman spectrometers found on Raman instruments. They may become more pop-
to collimate light for the diffraction grating and to focus ular as Raman instruments become more sophisticated. One
extremely simple example of a focal reducer on a Raman
light for the detector. Concave mirrors have no chromatic
spectrograph is the use of a cylindrical lens to reduce the
aberrations. They have very low optical scattering and are
height of the slit image in the spectrograph exit plane in
available with large apertures. Optical aberrations increase
order to match better the detector elements of an array
rapidly with decreasing f /#, though, so Raman spectro-
detector.
meters relying only on concave mirrors for collimation and
focusing rarely have focal ratios less than f /4. Toroidal
mirrors are sometimes used in Raman spectrometers in
3.5 Laser blocking filters
place of concave mirrors. They correct the astigmatism
that would be present when spherical concave mirrors are Laser blocking filters reduce the intensity of laser light
operated off-axis. Off-axis parabolic mirrors are rarely used entering a Raman spectrometer. This relaxes the stray light
in dispersive Raman spectrometers. They focus collimated requirements placed on the spectrometer. Two important
light to a point on axis with no aberration. Their aberrations types of laser blocking filters are notch filters and edge
increase more rapidly than those of spherical mirrors as the filters. Notch filters strongly attenuate a narrow spectral
focus moves off-axis. They are also more expensive than region (ideally by more than a factor of 1010 ) and transmit
spherical mirrors. the rest. Edge filters transmit a spectral region on one side
of a given wavelength and block on the other side. Most of
the different notch filters and edge filters luminesce to some
3.3 Multielement lenses degree at some wavelengths. Improper use of notch or edge
filters can therefore lead to excessive background light in
Unlike concave mirrors, multielement lenses have good Raman spectra. The many types of optical filter that have
aberration correction even for off-axis and at low f /#. been used as laser blocking filters for Raman spectrometers
Camera lenses are especially attractive because they have were reviewed recently.16 This section will focus briefly on
been highly engineered and are mass produced. As a result, the most important of these filters.
Atomic vapor filters17 20 have rejection linewidths of a of these edge filters tends to be rather poor, but they
few wavenumbers or less and can easily exceed laser line are simple, inexpensive, and sometimes work in difficult
attenuation of 1010 . They consist simply of a container with spectral regions, such as the deep UV, where better alter-
metal vapor inside. The laser is attenuated by absorption natives are not known.
caused by atomic transitions that originate in the ground
state. Atomic vapor filters are readily available at only a
few specific wavelengths such as 852.1 nm, 766.5 nm, and 3.6 Bandpass filters
253.7 nm. They are also somewhat difficult to work with, Bandpass filters pass a narrow wavelength region and block
so their use in Raman spectrometers is not common outside the rest of the spectrum. Bandpass filters can be used
specialized laboratories. as simple monochromators. The most important bandpass
Multilayer dielectric filters are much easier to use and are filters used in Raman spectroscopy are thin film dielectric
available at any wavelength from the near-UV through the bandpass filters, zero dispersion double monochromators,
NIR. Their typical rejection linewidths of several hundred birefringent filters, and acousto-optic tunable filters.
wavenumbers and typical laser wavelength attenuation fac- Thin film dielectric bandpass filters are commonly used
tors of 103 105 are not ideal, but are useful in many cases. on Raman light from point objects, planar surfaces, and
The main problem with these filters is their oscillating trans- three-dimensional volumes. They can be used as simple,
mission with changing wavelength in the Raman spectral low-cost, high-throughput Raman monochromators. A typ-
region. This spectral signature from the dielectric filter dis- ical filter that has a clear aperture of 21 mm, a 50%
torts the measured Raman spectrum. In principle the spec- transmission, a 2030 cm1 transmission bandwidth full-
tral distortion can be corrected, but the temperature depen- width-at-half-maximum, and an out-of-band transmission
dence of the signature makes robust correction difficult. of less than 0.01%, costs about $300. The transmission
Volume holographic notch filters are the most popular wavelength of these filters shifts rapidly to shorter wave-
type of notch filter for Raman spectrometers. Their typical lengths as the incident angle of the light increases. Tilting
rejection linewidths of a few hundred wavenumbers and of the filter can therefore be used to adjust the transmission
typical laser wavelength attenuation factors of 105 108 are wavelength. Tilt angles greater than about 10 cause the
well suited to most Raman applications. Like dielectric transmission wavelengths for the s- and p-polarizations to
filters they are easy to use, but unlike dielectric filters they separate significantly and cause overall filter performance
have almost no spectral signature in the Raman spectral to degrade. As a result, dielectric bandpass filters retain
region. Volume holographic notch filters are available at their maximum performance only in collimated light. There
any wavelength from the near-UV through to the NIR. are commercially available dielectric bandpass filters made
Edge filters serve the same function as notch filters using high-refractive-index films that are less sensitive to
when anti-Stokes information is not required. The most the angle of incident light.
popular edge filter is the subtractive double spectrograph. The double subtractive spectrograph described above can
A first spectrograph disperses input light and a second also serve as a bandpass filter by using an appropriate inter-
spectrograph reverses the process. An aperture (usually a mediate slit. Maximum throughput and spectral resolution
wide slit) in the focal plane between the two spectrographs for point or line sources is achieved by imaging the point or
blocks unwanted wavelengths of light. Light exits the line source onto the entrance slit. Two-dimensional images
second spectrograph through an exit slit, which serves as can be filtered by focusing the aperture of the collection lens
the entrance slit to a spectrograph. A subtractive double on the spectrograph entrance slit, and focusing the image
spectrograph can attenuate the laser frequency by a factor on the first diffraction grating.26
of 1010 1012 when the transmitted pass-band is several Birefringent bandpass filters27 use an optical retarder
hundred wavenumbers away from the laser frequency. It between two polarizers. The retarder rotates the polar-
also allows Raman spectra a few wavenumbers from the ization of light by a wavelength-dependent amount. The
exciting line to be measured, but with much less attenuation transmission spectrum of this simple filter is sinusoidal,
at the laser wavelength. Subtractive double spectrographs with transmission minima occurring at odd multiples of 90
are extremely versatile, but tend to have relatively poor polarization rotation and transmission maxima occurring at
transmission. Their transmission can exceed 50%, however, even multiples of 90 polarization rotation. Larger amounts
when they are optimized for a limited wavelength region.21 of polarization rotation, produced by using thicker retarders,
The optical absorption from carefully chosen condensed cause the transmission spectrum to oscillate more rapidly
phase materials can provide a Raman edge filter function. with changing wavelength. Bandpass filters that have a nar-
Certain semiconductor crystals,22 colored glasses,23 and row transmission band and an acceptably large free spectral
liquids24,25 have been used in this way. The performance range can be made by stacking simple birefringent filters
that have different retarder thicknesses. Electrically tunable section in order to avoid repetition in the separate sections
birefringent bandpass filters can be made by using liquid on monochromators and polychromators.
crystal films as the retarder material. For a given transmis-
sion bandwidth, birefringent filters are less sensitive to the
incident angle of light than are dielectric filters. This prop- 4.1 Optimal slit width
erty makes birefringent filters attractive for Raman imaging
applications. The optimal entrance slit width for a dispersive spectro-
Acousto-optic tunable bandpass filters (AOTFs) use an meter needed to maximize spectral resolution is the width
acoustic grating in a birefringent crystal to create a trans- where the divergence of the zero order diffraction produced
mission pass-band.28,29 AOTFs can be tuned much more by the slit matches the f /# of the spectrograph. Narrower
rapidly than birefringent bandpass filters. AOTFs can also slits have little effect on spectral resolution but reduce the
provide more complex transmission curves, such as mul- spectrometer optical throughput. Wider slits degrade the
tiple pass-bands, because the acoustic grating structure is possible spectral resolution of the spectrometer. Disper-
determined by the radiofrequency spectrum used to drive sive Raman spectrometers are usually in the f /4 to f /8
the transducers on the AOTF crystal. Like birefringent range, and so have optimal spectral resolution when the
bandpass filters, AOTFs are less sensitive to the incident slits are a few micrometers wide. In nearly all cases Raman
angle of light than are dielectric filters. AOTFs have also spectrometer slits are much wider than this. Widths of
been used for Raman imaging applications. 50500 m are common, depending on the specific spec-
trometer. The reason for using wider slits, despite the loss
in spectral resolution, is to increase the throughput of the
3.7 Slits spectrometer.
Wider slits not only decrease spectral resolution, but
Slits, or more generally, entrance and exit apertures, are also introduce a potential wavelength calibration error. The
extremely important optical components in Raman spec- entrance slit width is no longer a limiting aperture when the
trometers. A good slit can be thought of as a spatial image focused onto the slit is narrower than the slit itself.
bandpass filter that has 100% in-band transmission and In this situation a monochromatic image could move within
0% out-of-band transmission. Slits can have a fixed or the dimension of the slit width, causing a corresponding
adjustable width. Fixed-width slits are often laser-cut in movement of the image at the spectrometer exit plane.
metal sheet or scribed through a vacuum-deposited metal Such a movement in the exit plane would appear to the
on a dielectric surface. The dimensions of the most com- detector to be a change in wavelength. The optical energy
monly used fixed-width slits in Raman spectrometers range distribution across the width of the entrance slit needs to
from about 20 m to 500 m in width and from 3 mm to be the same during wavelength calibration as it is during
25 mm in height. Adjustable slits consist of two smooth a sample measurement in order to prevent this type of
metal edges whose separation is continuously adjustable. calibration error.
Slit edges are always very thin in order to minimize reflec-
tion of light from the inside edges of the slit thickness. The
slit edges need to be parallel and free of edge distortion 4.2 Dispersion
and dust so that the slit transmission does not vary along
the height of the slit. The linear dispersion of a spectrometer measures how far
Spectrometer slits greatly reduce detected stray light. apart different wavelengths are in the exit focal plane. It is
This is why the conversion of a dispersive monochromator the product of the grating (or prism) angular dispersion and
to a polychromator by removing the exit slit degrades the the focal length of the spectrometer focusing optic. Linear
stray light performance. Slits also improve intermediate dispersion is usually expressed in units of millimeters in the
image quality by blocking light that extends beyond the focal plane per nanometer (or cm1 ) change in the optical
desired point spread function. wavelength. The reciprocal linear dispersion of spectro-
meters, expressed in reciprocal centimeters per nanometer,
is more commonly reported than linear dispersion, however.
4 PROPERTIES COMMON TO BOTH
DISPERSIVE MONOCHROMATORS
AND SPECTROGRAPHS 4.3 Focusing tolerance
Dispersive monochromators and spectrographs share seve- The focusing tolerance of a spectrometer is the maximum
ral important properties. Some of these are discussed in this distance between the exit aperture and focal plane that does
not significantly degrade the spectrometer performance.

Ideally the exit aperture is exactly in the focal plane. A P
L3
large focusing tolerance, however, makes a spectrometer
easier to align and more robust to environmental factors
P
such as vibration and temperature changes. The focusing
F
tolerance for a spectrometer, Tf , is given by L4
P
f L1 L2
Tf D s D sf/# 11
DCs (a) (b)
where Tf is the distance between focal plane and exit
aperture that causes the width of the entrance slit image
G2
to double at the exit aperture, s is the width of the in- M1
focus entrance slit image near spectrometer exit plane, f is G1
the focal length of spectrometer focusing optic, D is the M2
diameter of spectrometer focusing optic, and f /# is the
spectrometer focal ratio or f -number. (c) (d)
A typical f /8 double monochromator with unity magni- Figure 1. Monochromators. The filter monochromator (a) and the
fication between the entrance slit and the exit slit might CzernyTurner monochromator (c) are the most commonly used
use an entrance slit width of 500 m, giving a focal tol- monochromators for Raman spectroscopy. L1, L2, L3, and L4
are lenses; F is a filter; P are dispersive prisms; M1 and M2 are
erance of 4 mm. An f /1.8 spectrograph, again with unity
concave mirrors; G1 is a planar diffraction grating; and G2 is
magnification, using a 25-m slit width would have a focal a concave diffraction grating. (b) shows a prism monochromator
tolerance of 0.045 mm. and (d) shows a concave grating monochromator.
analysis, where Raman spectra are simple and analytical

4.4 Polarization effects bands are well separated in wavelength.
Images can be transmitted through a bandpass filter. A
The transmission of modern dispersive Raman spectro- two-dimensional detector can then be used to map Raman
meters is polarization dependent. One way to eliminate intensity as a function of horizontal and vertical position
this dependence is to randomize the polarization before in the sample. This approach to Raman image collection
the light enters the Raman instrument. Birefringent wedges is called global imaging.30 Images at several different
called polarization scramblers carry out this function. The wavelengths are usually collected by tuning or exchanging
polarization state of light transmitted by a location of the the bandpass filter. Thin-film dielectric filters have been
polarization scrambler is determined by the polarization used for Raman global imaging, but birefringent filters and
scrambler thickness at that location. The overall polari- acousto-optic tunable filters are more often used because of
zation of light transmitted by the polarization scrambler their superior imaging and tuning capabilities.
is essentially random if the polarization state of the trans-
mitted light goes through 16 or more cycles across the
changing thickness of the polarization scrambler. 5.2 Dispersive single-stage monochromators
Early Raman spectrometers used prisms rather than diffrac-

5 MONOCHROMATORS tion gratings to provide wavelength dispersion. This ap-
proach is now obsolete. As technology evolves, however,
5.1 Nondispersive monochromators spectrometer design constraints change and old design con-
cepts find new life. Figure 1(b) shows a schematic diagram
A single-element bandpass filter is a simple, inexpen- of a prism-based Raman monochromator made in the mid-
sive monochromator. One configuration is illustrated in 1950s.31 Three 60 prisms made of highly dispersive, extra
Figure 1(a). Thin-film dielectric filters are most commonly dense, flint glass provided the wavelength dispersion. These
used, but certain colored glass filters are sometimes suffi- prisms were 73 mm tall and had base widths of 134 mm,
cient. These filters often have very high throughput. This is 141 mm, and 153 mm. The prisms needed to be very large
especially true for colored glass filters because their perfor- so that the monochromator could have both a long focal
mance is only weakly affected by the incidence angle of the length for adequate linear dispersion, and a low f /# for high
light. Thin-film dielectric filters are most often used for gas throughput. The focusing optic in the monochromator was
a modified Schmidt camera (concave mirror with spherical modification of the Rowland monochromator, where the
correction plate). It had a focal ratio of f /3, a focal length entrance and exit slits form a 70 angle from the grating
of 345 mm, and a 7% loss caused by the central obscu- surface. A drawing of this monochromator is shown in
ration. Its useful focal plane was 8 mm 100 mm. The Figure 1(d). The transmission wavelength can be changed
monochromator collimating lens was an achromatic doub- simply by rotating the grating.
let with a focal length of 920 mm. Optical surfaces were
antireflection coated with MgF2 . The adjustable entrance
slit ranged from 4 to 16 mm high and from 3 to 1500 m 5.3 Dispersive multiple-stage monochromators
wide. The entrance slit image in the focal plane was demag-
nified by a factor of 920/345 D 2.7. Wavelength tuning Stray light levels can be greatly reduced by sending the
was accomplished by replacing the photographic film with output of one monochromator into the input of a second
a scanning slit. monochromator. The stray light level of the resulting
The most popular dispersive monochromator for Raman double monochromator is equal to the product of the
spectroscopy is the CzernyTurner monochromator. Its stray light levels of the two individual monochromators.
optical diagram is shown in Figure 1(c). Light from the Laser light intensity transmitted by a reasonably good sin-
entrance slit is collimated by a concave mirror, diffracted gle monochromator set at a wavelength a few hundred
by the plane grating, and focused through an exit slit by wavenumbers away from the laser wavelength will often
a second concave mirror. The transmitted wavelength is be about 106 times as strong as the laser light intensity
changed by rotating the diffraction grating. The degree to sent into the monochromator. Weak Raman bands would
which each mirror is operated off-axis can be indepen- be overwhelmed by this level of stray light. A double
monochromator has a corresponding stray light level of
dently varied in order to reduce optical aberration (field
about 1012 . The stray light level for a triple monochro-
curvature and coma). Spherical mirrors can be replaced by
mator would be about 1018 . Stray laser light levels in
toroidal mirrors to reduce astigmatism. Even with these cor-
monochromators increase rapidly as the wavelength trans-
rections, optical aberration generally limits CzernyTurner
mitted by the monochromator approaches the laser wave-
monochromator focal ratios to values of f /4 or greater. In
length. A good double monochromator provides adequate
principle, the use of aspheric corrections to toroidal mirrors
stray light rejection to within 5 or 10 cm1 of the laser
could substantially reduce optical aberrations at low aper-
line for most Raman samples. Multiple-stage monochro-
ture ratios, but such mirrors would be expensive.
mators are usually made with CzernyTurner monochro-
A monochromator can be made from an entrance slit,
mator stages, but double SeyaNamioka monochromators
a concave grating, and an exit slit. Concave gratings can
and prism/grating hybrid multiple-stage monochromators
sometimes increase monochromator transmission by replac-
are also commercially available (McPherson, Acton, Mas-
ing optical losses from three optical elements with the
sachusetts, USA).32
optical loss from just one. This transmission benefit is The linear dispersions of the individual monochroma-
very significant in the deep UV spectral region, because tors used to make a multiple-stage monochromator add
good reflective coatings are not available. The transmission together. The resulting spectral resolution is improved
benefit is not very important in the visible and NIR spec- without reducing the slit width. The transmission of the
tral regions, where most Raman measurements are made, multiple-stage monochromator is equal to the product of
because very good reflective coatings allow the use of mul- the individual monochromator transmissions and is there-
tiple optical components with acceptably low losses. More fore relatively low. Multistage monochromators also tend
important, the aberration correction that can be accom- to be large, mechanically complex, and expensive. Stray
plished with one optical surface (the concave grating) is light performance, however, often outweighs all of these
more limited than that which can be accomplished with the limitations.
three optical surfaces of a CzernyTurner monochromator.
Consequently, monochromators using concave gratings are
rarely used for Raman spectroscopy. This could change as 6 POLYCHROMATORS
interest in deep UV Raman spectroscopy grows.
The first concave grating monochromator developed was Monochromators transmit a small part of the Raman spec-
the Rowland monochromator, named after the inventor of trum to a detector and reject all the rest of the Raman
the concave grating. The entrance slit, grating surface, and light. Instead of wasting this light, polychromators direct it
exit slit were all placed on a circle, called Rowlands circle, to different detectors for simultaneous measurement. This
whose diameter equaled the radius of curvature for the gives polychromators several important advantages over
grating. The SeyaNamioka monochromator is a modern monochromators:
1. Better use of collected Raman light leads to faster data

acquisition and a better signal-to-noise ratio. D
F1
2. Variations in collected Raman intensity affect the entire
L1 L2
spectrum in the same way. Common causes for such
variations include bubbles, laser power changes, and F2
L3 L4
movements of the sample in and out of focus.
3. Spectra can be collected without moving any of the
optical components. This leads to improved precision
and reduced maintenance. D D
Polychromators do have some disadvantages, however, (a)
when compared with monochromators:
1. Detector elements are never identical, so polychroma- D M1

tor systems have some level of fixed pattern noise.
G2
2. Elimination of the monochromator exit slit increases
stray light. M2
G1 M3
3. Polychromator imaging requirements are more severe
than those of monochromators. L5 M4 M5
L6 G3
6.1 Multiple-filter polychromators M6
Just as a simple bandpass filter can perform as a monochro- (b) D (c)

mator, so several different filters together can perform
Figure 2. Polychromators. Additional types of polychromator can
as a nondispersive polychromator. The polychromator can be made from the monochromators shown in Figures 1(b), 1(c),
be made by pointing several independent nondispersive and 1(d) by replacing the exit slit by an array detector. L1
monochromators, along with their collection optics, at the through L6 are lenses; F1 and F2 are reflective filters; G1 is a
same sample from different angles. A better way to make transmission grating; D is an exit focal plane; M1, M2, M5, and
a nondispersive polychromator is to use bandpass filters M6 are concave mirrors; M3 and M4 are flat mirrors; and G2
and G3 are planar diffraction gratings. (a) Filter polychromator,
as dichroic beam splitters. This is shown in Figure 2(a). (b) axial transmissive spectrograph, (c) zero-dispersion double
Light from the sample can then be reflected from one spectrograph.
filter to the next, with a different spectral band being
transmitted to a detector by each filter. Nondispersive with a curved photographic film holder. The focal plane
polychromators are rugged, inexpensive, and have very was 8 100 mm. The spectral coverage was from 400 nm
high throughput. They are well suited to gas-phase anal- to 650 nm with a spectral resolution ranging from 2.5 cm1
ysis, where Raman bands are often well isolated from at 400 nm to 6.6 cm1 at 550 nm. Spectra from different
each other. excitation wavelengths were recorded simultaneously.
The CzernyTurner monochromator, described in Sec-
tion 5, is also commonly used as a spectrograph. Its exit
6.2 One-dimensional dispersive spectrographs focal plane can be made reasonably flat by using an asym-
metric geometry where the off-axis angles of the two
A one-dimensional dispersive spectrograph uses only one mirrors are not equal. A modified CzernyTurner spectro-
direction in its exit plane for the separation of different graph using an aberration-optimized design with a refractive
wavelengths. The other direction may be used to maximize focal reducer can provide line image widths of 1015 m
spectrograph etendue or for spatial resolution. Spatial reso- at f /2.32 Another modified CzernyTurner instrument using
lution allows one-dimensional imaging and/or simultaneous three off-axis aspheric imaging mirrors achieved pixel lim-
measurement of multiple samples. A dispersive monochro- ited imaging over the entire 1037 1340 pixel field of a
mator can be converted into a one-dimensional dispersive CCD detector.33
spectrograph by replacing the exit slit with an array detector A spectrograph using a concave grating can be made by
if the exit focal plane is flat enough. replacing the exit slit of a Rowland monochromator with
The prism-based monochromator described in Section 5 a focal plane detector. This type of spectrograph is called
was operated as a spectrograph by replacing the exit slit a PaschenRunge spectrograph. The field curvature owing
to Rowlands circle can be corrected by using an array then combines these segments. Variations between seg-
detector with a curved surface (a bent photographic plate). ments tend to be averaged out and fixed pattern noise
In principle a focal reducer could correct the field curvature is reduced.
and thereby make the PaschenRunge spectrograph better A similar technique for extending the spectral cover-
suited to modern CCD detectors. A better approach would age of a spectrograph is time-delayed integration (TDI)
be to use an aberration-corrected concave grating. scanning.37 TDI scanning moves photogenerated charge in
The axial transmissive spectrograph34 is a relatively new CCD detector columns synchronously with a given wave-
design. Its schematic diagram is shown in Figure 2(b). length of light as the spectrograph grating is rotated. When
Light from the entrance slit is collimated with a camera the charge reaches the end of the detector, it is digitized and
lens, diffracted at about 90 by a volume transmission stored. Like SMT, TDI scanning averages out discontinu-
grating, and focused onto a CCD detector with another ities between segments and fixed pattern noise. At least one
camera lens. Both camera lenses have short focal lengths, Raman spectrograph using TDI scanning is commercially
typically 7585 mm, to keep aperture sizes reasonable and available.
focal ratios below f /2. High-quality camera lenses provide a
point spread function that is smaller than 20 m even with
focal ratios of f /1.8 and f /1.4. The focusing lens needs 6.3 Two-dimensional spectrographs
to collect diffracted light at 8.4 to have a 25-mm exit
plane, so it must be located close to the grating in order to Two-dimensional spectrographs use both dimensions of a
minimize vignetting. The large angle between incident and two-dimensional detector to extend spectral coverage. Dif-
diffracted light from the grating makes this possible. The ferent spectral segments are dispersed horizontally and
main limitation of the axial transmissive spectrograph is its stacked one above the other in the vertical direction at the
limited range of operating wavelengths caused by its use of detector. These spectral segments can then be combined
multielement camera lenses. to make a single spectrum with greater spectral coverage
One of the most popular spectrographs for Raman at a given spectral resolution. The combination of seg-
spectroscopy is the triple spectrograph. It is simply a ments is not required for analytical method development
CzernyTurner spectrograph that disperses the output of and deployment, but is more visually appealing and is a
a zero dispersion double spectrograph. The zero dispersion good diagnostic for correctness of wavelength and intensity
double spectrograph was described in Section 3; key com- calibration. The etendue of the spectrograph is decreased by
ponents and its schematic diagram are shown in Figure 2(c). this approach, however, because the usable slit height must
It is used as a laser wavelength rejection filter. Triple be reduced in order to avoid overlap of the different spectral
spectrographs have very low stray light and are capable segments.
of measuring Raman bands located as close as a 2 cm1 The first two-dimensional Raman spectrograph used an
from the laser wavelength. They are also very versatile. echelle grating.15,38,39 Unlike echelle spectrographs com-
Gratings are easily changed to set spectral coverage and monly used in atomic spectroscopy, this spectrograph oper-
spectral resolution. A single instrument can be operated ated only in part of the visible spectrum. As a result, highly
with laser wavelengths ranging from the UV to the NIR. aberration corrected, low f /#, multielement lenses could be
The transmission of a triple spectrograph is low, however, used instead of high f /# concave mirrors. The relatively
because of the large number of optical elements in the poor stray light performance of the echelle grating and
optical path. This problem is often made worse by the multielement lenses was not a limiting factor because the
use of broadband optics that work at many wavelengths, laser wavelength was filtered out before light from the sam-
but do not have the high transmission of more special- ple entered the spectrograph. A schematic diagram of this
ized optics. Triple spectrographs are also large, complex, spectrograph is shown in Figure 3. Light transmitted by the
and expensive. entrance slit was collimated by camera lens L1. The colli-
Spectrograph gratings can be moved in order to increase mated light was reflected by mirror M1 to cross-dispersing
spectral coverage without decreasing spectral resolution. grating G1. This grating dispersed light horizontally onto
Normally spectral segments are collected sequentially and the echelle grating G2. The echelle grating dispersed the
mathematically combined to create the larger spectrum. light vertically. Lens L2 focused the twice-dispersed light
There are usually discontinuities at the joining points caused onto a two-dimensional CCD detector. Eight diffraction
by wavelength calibration error, spectrograph throughput orders from the echelle grating were imaged onto the CCD
variations, or time-varying background such as fluores- detector. The cross-dispersing grating caused each order
cence. The scanning multichannel technique (SMT)35,36 to be separated from the others in the vertical direction.
collects spectral segments that have large overlap and An echelle spectrograph for Raman spectroscopy has been
spectrograph design works best when beam splitting filters

L2 G2
with at least one spectrally sharp transition from highly
FP reflecting to highly transmitting are used.
The axial transmissive one-dimensional spectrograph
shown in Figure 2(b) can be made into a two-dimensional
S
spectrograph14,42 by replacing the volume transmission
grating with the holoplex transmission grating described in
Section 3. The f /# of the resulting spectrograph remains the
G1
L1 M1 same, but the usable slit height is reduced by a factor of 2.
A holoplex grating can be built from transmission grat-
ings because a transmission grating disperses light in one
spectral segment, and leaves the rest available for a dif-
Figure 3. Echelle spectrograph designed for Raman spectro- ferent optical element. In the case of a holoplex grating,
scopy. S is the entrance slit; L1 and L2 are camera lenses; G1 is
the cross-dispersing diffraction grating; G2 is the echelle diffrac-
the other optical element is another transmission grating.
tion grating; and FP is the exit focal plane. Reflection gratings can also be configured to stack differ-
ent spectral segments on a two-dimensional detector, just
as a holoplex grating does. The simplest way to do this is
commercialized.40 The fully integrated Raman instrument is
to arrange two separate reflection gratings to each receive
portable and has been used in nonlaboratory environments.
half of the light from the entrance slit. When the relative
Another way to implement a two-dimensional spectro-
angles between the two gratings are properly adjusted this
graph has been patented.41 Beam splitting filters were used
composite grating assembly can replace the standard plane
to divide the input light into spectral segments. Each spec-
grating in a one-dimensional spectrograph to create a two-
tral segment was delivered to a diffraction grating at a
dimensional spectrograph. This aperture-sharing approach
unique angle so that the diffracted light, when focused,
effectively degrades the spectrograph f /# and reduces the
formed spectra stacked vertically on top of each other. usable slit height by a factor of 2.
Figure 4 shows a schematic diagram of the optical config- There is a second approach to making a two-dimensional
uration. The angles of incidence to the grating for the three spectrograph using reflection gratings that does not require
spectral segments are all different such that the range of the the spectrograph f /# to be degraded. Reflection gratings
angles of diffraction for each spectral segment is approxi- diffract light until the wavelength becomes too long. Then,
mately the same. The angle between the incident spectral the grating simply acts like a nondispersive mirror. A
segment beams and the plane normal to the grating grooves two-dimensional spectrograph has been constructed that
(g in equation 5) is different for each spectral segment beam uses a concave grating to disperse short-wavelength light
causing the spectra to be displaced from each in the direc- and reflect long-wavelength light to a second grating that
tion normal to the dispersion axis. This two-dimensional disperses the long-wavelength light.43,44 This configuration
is to reflection gratings what the holoplex grating is to
L1 transmission gratings.
G1
NA Numerical Aperture
FP SMT Scanning Multichannel Technique
TDI Time-delayed Integration
F3
F2
REFERENCES
F1
1. A.J. den Dekker and A. van den Bos, J. Opt. Soc. Am. A, 14,
547557 (1997).
Figure 4. Stacking of spectral segments using dichroic beam
splitters and a diffraction grating. F1, F2, and F3 are dichroic 2. P.A. Jansson, Deconvolution of Images and Spectra, Aca-
beam splitters; G1 is a diffraction grating; L1 is a camera lens; demic Press, New York (1996).
and FP is the exit focal plane. 3. W.F. Maddams, Appl. Spectrosc., 34, 245267 (1980).
4. A. Mestari, R. Gaufres and P. Huguet, J. Raman Spectrosc., 27. W.I. Kaye, Liquid Crystal Tuned Birefringent Filter, US
29, 353358 (1998). Patent 4 394 069, assigned to Beckman Instruments, Inc.,
5. M.J. Pelletier, Appl. Spectrosc., 53, 10871096 (1999). Fullerton, CA (1983).
6. S. Singh, Opt. Laser Tech., 31, 195218 (1999). 28. C.D. Tran, Anal. Chem., 64, 971A981A (1992).
7. M. Born and E. Wolf, Principles of Optics, 6th edition, 29. I.C. Chang, SPIE (International Society for Optical Engineer-
Pergamon Press, Oxford, 406407 (1980). ing) Acousto-Optics, 90, 1222 (1976).
8. R.W. Wood, Phil. Mag., 20, 770771 (1910). 30. L. Markwort, B. Kip, E. Da Silva and B. Roussel, Appl. Spec-
trosc., 49, 14111430 (1995).
9. E. Hecht and A. Zajac, Optics, Addison-Wesley, Menlo
Park, chapter 10 (1974). 31. Advertising literature for the High Speed Raman Spectro-
graph Assembly 9100, manufactured by Applied Research
10. M. Born and E. Wolf, Principles of Optics, 6th edition, Laboratories, Glendale, California, approximate date (1953).
Pergamon Press, Oxford (1980).
32. J.F. James, J. Mod. Opt., 41, 20332042 (1994).
11. P. Lemaire, Appl. Opt., 30, 12941302 (1991).
33. C. Feng and A. Ahmad, Opt. Eng., 34, 32173221 (1995).
12. J.M. Tedesco, H. Owen, D.M. Pallister and M.D. Morris, Anal.
Chem., 65, 441A449A (1993). 34. D.E. Battey, J.B. Slater, R. Wludyka, H. Owen, D.M. Pallis-
ter and M.D. Morris, Appl. Spectrosc., 47, 19131919 (1993).
13. Holoplex transmission grating, Kaiser Optical Systems, Inc.,
Ann Arbor, MI. 35. P. Knoll, R. Singer and W. Kiefer, Appl. Spectrosc., 44,
776782 (1990).
14. D.E. Battey, H. Owen and J.M. Tedesco, Spectrograph with
Multiplexing of Different Wavelength Regions onto a Single 36. V. Deckert and W. Kiefer, Appl. Spectrosc., 46, 322328
Opto-Electric Detector Array, US Patent 5 442 439, assigned (1992).
to Kaiser Optical systems, Inc., Ann Arbor, MI (1995). 37. R.B. Bilhorn, Scanning Multichannel Spectrometry Using a
15. Chem. Eng. News, March 20, 41 (1989). Charge-Coupled Device (CCD) in Time-Delay Integration
(TDI) Mode, US Patent 5 173 748, assigned to Eastman
16. M.J. Pelletier, Analytical Applications of Raman Spectro- Kodak Company (1992).
scopy, Blackwell Science, Oxford, 8692 (1999).
38. M.J. Pelletier, Appl. Spectrosc., 44, 16991705 (1990).
17. M.F. Fink, J.C. Robinson and W.F. Buell, Raman Spectro-
meter, US Patent 5 786 893, assigned to the University of 39. M.J. Pelletier, SPIE Raman and Luminescence Spectroscopies
Texas System Austin, TX, July 28 (1998). in Technology II, 1336, 152158 (1990).
18. M.J. Pelletier, Appl. Spectrosc., 46, 395400 (1992). 40. M.M. Carrabba, J.M. Bello, K.M. Spencer and J.W. Haas, III,
Echelle Spectroscopy and CCDS, An Ideal Union for Fiber
19. M.J. Pelletier, Appl. Spectrosc., 47, 6974 (1993). Optic Raman Systems, in Recent Developments in Sci-
20. R. Indralingam, J.B. Simeonsson, G.A. Petrucci, B.W. Smith entific Optical Imaging, eds M.B. Denton, R.E. Fields and
and J.D. Winefordner, Anal. Chem., 64, 964967 (1992). Q.S. Hanley, Royal Society of Chemistry Information Ser-
vices, London (1996).
21. T.D. Harris, M.L. Schnoes and L. Seibles, Anal. Chem., 61,
994998 (1989). 41. B.J.E. Smith, D.N. Batchelder and K.J. Baldwin, Spectrum
Splitting Spectroscopic Apparatus, US Patent 5 638 173,
22. A. Schulte, Appl. Spectrosc., 46, 891893 (1992). assigned to Renishaw Transducer, Gloucestershire, GB (1997).
23. M.P.F. Bristow, Appl. Opt., 18, 952955 (1979). 42. H. Owen, D.E. Battey, M.J. Pelletier and J.B. Slater, SPIE
24. J.A. Kleimeyer, J.C. Fister, J. Zimmerman and J.M. Harris, Practical Holography IX, 2406, 260267 (1995).
Appl. Spectrosc., 50, 15971602. 43. J.P. Laude, SPIE (International Society for Optical Engineer-
25. R.L. McKenzie, Opt. Lett., 18, 995997 (1993). ing), 701, 9397 (1986).
26. P. Dhamelincourt, F. Wallart, M. Leclercq, A.T. NGuyen and 44. J.P. Laude, SPIE (International Society for Optical Engineer-
D.O. Dandon, Anal. Chem., 51, 414A421A (1979). ing), 673, 455458 (1986).
Volume Phase Holographic Optical Elements
H. Owen
Kaiser Optical Systems, Ann Arbor, MI, USA
1 INTRODUCTION resolution of electron microscopy. Gabor recognized that

recording the interference pattern of two wavefronts would
Holograms can be thought of as a kind of three-dimensional produce a complex recording which, when illuminated by
laser photograph which when viewed using a suitable point one of the original wavefronts, would reconstruct the other
light source produced remarkable images with outstanding wavefront by diffraction. Low-coherence light sources lim-
depth and resolution. Volume phase holographic optical ele- ited the optical set-ups to in-line geometry which, in con-
ments are functional rather than visual but, when used as junction with the silver halide recording materials at that
components within a Raman spectrometer, they provide sig- time, could only produce noisy low-efficiency holograms.
nificant performance advantages when used as alternatives The introduction of the laser provided a light source with
for traditional optical components. Bragg or volume phase the spatial and temporal coherence necessary to create inter-
diffraction operates on different principles when compared ference patterns capable of producing modern holograms.
against traditional surface relief gratings or thin film dielec- Using the laser, Leith and Upatnieks3 in 1962 demonstrated
tric interference filters. the off-axis reference beam technique that created the first
Included in this article is a look at the historical back- practical transmission holograms. During the same year
ground, materials and fabrication techniques, physical and Denisyuk4 introduced the reference beam from the oppo-
optical properties, and a detailed description of the three site side of the photographic plate and produced the first
major holographic components used for Raman spec- modern reflection holograms.
troscopy, notch filters, laser bandpass filters, and volume As the field of holography progressed, limited diffrac-
phase holographic gratings. tion efficiency and problems with scattering and absorption
associated with silver halide-based photosensitive emul-
sions led to the need for improved chemical processing
2 BACKGROUND of silver halide emulsions or new better materials. In 1968
Shankoff5 proposed that dichromated gelatin was a photo-
Lippmann1 in 1891 using a high-resolution photographic sensitive material that offered significantly better resolution
emulsion, a mercury-vapor line as his light source, and and diffraction efficiency, but required significant further
a mercury mirror as the reflective background, produced research and development before it could be widely used.
color photographs that could be viewed in reflected white Dichromated gelatin has been used since the beginning of
light. The limited coherence of the mercury vapor line was the nineteenth century as a photosensitive material to gen-
still sufficient to set up an interference pattern within the erate relief images for use in the printing industry.
emulsion; this was recorded to produce the first reflection Improving the performance of dichromated gelatin
hologram. attracted a lot of attention and significant developments
The principle of wavefront reconstruction was proposed continued throughout the 1970s on material and process
by Gabor2 in 1948; his original aim was to improve the improvements.6 11 In the early to mid-1980s researchers
within this field focused more on using dichromated
John Wiley & Sons Ltd, 2002. gelatin to develop new applications and products.12 15 The
significant advantages of using holographic combiners in 3.2 Laser exposure

the avionics heads-up display of fighter aircraft provided
the incentive for several major companies to make the large The laser wavelength, angle of incidence and source
investments required to fully develop this technology and position with respect to the sensitized film generally deter-
make it available for the development of new spectroscopic mine the spectral and angular replay characteristics of the
products in the 1990s. final element. For notch filters, changing the exposure angle
results in changes in the fringe frequency, which in turn
determines the replay wavelength. As both filters and grat-
3 HOLOGRAPHIC MATERIALS AND ings generally require uniform replay characteristics, the
FABRICATION PROCESS laser beams used to generate the interference pattern need
to be collimated. Other key parameters that need to be
Dichromated gelatin is now well established as the material carefully selected and controlled during the laser exposure
of choice for producing high-quality holographic compo- process include the laser intensity and uniformity across
nents. Its ability to provide high diffraction efficiency, the sensitized film, the exposure level, and the interference
high clarity, low scatter, low absorption, and long lifetime pattern stability.
when properly sealed and protected against moisture has Figure 1 outlines the typical optical geometries used to
secured its dominant position as the material of choice. construct both notch filters and volume phase gratings. The
Being difficult to use due to its extremely complicated notch filter requires interference fringes to be parallel with
and demanding process requirements has limited its use respect to the sensitized film so the laser beams need to
in the past to mainly military applications. The fabrica- enter the sensitized film from opposite directions. By index
tion process can be divided into four parts: film prepara- matching a mirror onto the sensitized film the reflection
tion, laser exposure, chemical processing, and drying and from the incident beam by the mirror generates the second
sealing. beam and sets up the required interference pattern. Volume
phase gratings require interference fringes to be perpendic-
ular with respect to the sensitized film so the laser beams
3.1 Film preparation need to enter the sensitized film from the same direction.
During the laser exposure a latent image is formed within
Gelatin is derived from the collagen found in animal tissues. the sensitized film by recording the interference pattern set
It is produced by breaking down the fibrous collagen into a up by the two mutually coherent laser beams. The latent
colloid of various amino acids linked together by polypep- image results from cross-linking between the molecular
tide links. Gelatin quality and variability in properties need chains within the gelatin emulsion and the photosensitive
to be accommodated if product quality and commercially dichromate ions. The photosensitive emulsion records the
viable process yields are to be achieved. variation in energy associated with the laser interference
The emulsion used to prepare the photosensitive film con- pattern and sets up a small sinusoidal variation in refrac-
sists of a solution of photographic grade gelatin sensitized tive index throughout the film thickness. At this stage of the
with ammonium dichromate. This emulsion is coated onto
a suitable substrate using established coating techniques to Filter exposure Grating exposure
produce uniform films of the required thickness. The sub-
strate is typically made from a high-quality optical glass Laser 1 Laser 1
such as BK7, the requirements and specifications of which,
such as size, thickness, flatness, and surface finish, relate
to the needs of the final element.
Gratings typically require films of a few micrometers in
thickness, whereas high-performance notch filters require
films of well over 100 m in thickness. The parameters Laser 2 Laser 2
that need to be controlled accurately, in addition to film
thickness and uniformity, include film hardness and refrac-
tive index, and emulsion sensitivity to the laser wavelength
used during the exposure. As dichromated gelatin is nor- Substrate Mirror Substrate Cover plate
mally a blue-sensitive material, the laser lines from an argon Emulsion Emulsion
ion laser at 457 nm, 488 nm, and 514 nm are the preferred Figure 1. Diagram of laser exposure configuration for filter and
wavelengths for the laser exposure. grating-type volume phase holographic optical elements.
Volume Phase Holographic Optical Elements 3
process the diffraction efficiency associated with the latent diffraction , the wavelength of the light l, and the semi-
image is of the order of 12%. angle q between the incident and diffracted ray. The law
is expressed analytically by 2 sin q D ml where m is an
integer.
3.3 Chemical processing Kogelnik16 was the first to apply the coupled wave theory
to volume phase holographic elements to provide analytical
The purpose of the chemical process is to develop the latent solutions to diffraction efficiency, spectral characteristics
image to generate the almost 100% diffraction efficiency and polarization effects; his two-wave first-order method
required by notch filters. neglects higher order waves and the second derivatives of
The first step of the chemical process is to soak the the field amplitude.
exposed film in water, which causes the film to swell sig- Dielectric filters typically consist of a stack of discrete
nificantly and allows residual sensitizer and other unwanted layers with alternating higher then lower refractive index.
chemical byproducts to be removed. The average refractive index is generally greater than 1.5
The second step is the controlled removal of the excess due to the materials used to create the filter. The square-
water from within the film by using a series of alcohol wave variation in refractive index in a relatively thin film
baths with increasing proportions of alcohol to water. This leads to a filter with a wide spectral bandwidth and strong
controlled chemical dehydration process reduces the film ringing effects.
thickness rapidly and causes the significant increase in The differences in refractive index profiles between a
refractive index modulation within the film. holographic notch filter and a conventional dielectric filter
This simplified description of the process fails to describe are compared schematically in Figure 2.
the degree of process knowledge and control required to Holographic notch filters are constructed from signifi-
reduce the swollen film thickness by a factor of 510 while cantly thicker films (typically 50150 m) and have a lower
maintaining fringe spacing to fractions of a wavelength index modulation when compared to dielectric filters. The
throughout the film thickness and over large film areas. average refractive index is close to 1.5, corresponding to
the gelatin used to create the filter. The sinusoidal vari-
ation in refractive index in a thick film produces a filter
3.4 Drying and sealing with high optical density (OD), narrow spectral bandwidth,
sharp spectral edges and very little spectral ringing. A spec-
Gelatin is very sensitive to moisture and readily absorbs tral plot of transmission versus wavelength comparing these
around 10% by weight in a typical room environment. differences between holographic notch filters and dielectric
Therefore to protect the final holographic element a suitable filters is shown in Figure 3.
coverplate is laminated onto the substrate and a protective What was initially a sinusoidal variation in refractive
optical cement barrier added to the perimeter to act as an index within the latent image becomes an almost sinusoidal
edge seal. Once the element has been dried and sealed it variation in refractive index within the final holographic
can be handled and cleaned as you would a conventional element. The spectral characteristics close to the laser
optical element. line show how the sidebands are smoothed due to this
difference. The spectral sidebands can also be skewed on
4 PHYSICAL AND OPTICAL Dielectric interference filter
PROPERTIES
>1.5
Refractive index
The underlying principle of holography is that both the

amplitude and phase associated with an object wave-
front are recorded together. As only the amplitude can be
Holographic notch filter
recorded directly a second reference wavefront is intro-
duced to set up an interference pattern that, when recorded, 1.5
contains both amplitude and phase information. We can
then use the reference wavefront to reconstruct the original
object wavefront by diffraction. 0 30 >60
For best reconstruction the wavelength and incident angle Film thickness / m
must satisfy the Bragg condition. Braggs law of diffraction Figure 2. Diagram of refractive index changes with film thick-
is a relationship between the spacing of the planes of ness for dielectric and holographic notch filters.
100 band positioned at approximately half the wavelength of

the primary band.
90
80
70 5 HOLOGRAPHIC NOTCH FILTERS

Transmission (%)
60
5.1 Background
50
Holographic filters are ideally suited as a prefilter when
40
used with a single-stage grating spectrograph for Raman
30 spectroscopic applications. Single-stage instruments have
20 Holographic better throughputs, are more compact and can be less
notch filter expensive than multigrating-based systems. The first pub-
10 Dielectric filter
lication outlining the benefits of using holographic filters
0 for Raman spectroscopy was by Carrabba et al.17 when the
450 470 490 510 530 550 570 590 holographic edge filter was shown to have useful trans-
Wavelength / nm mission characteristics down to 200 cm1 . A more detailed
Figure 3. Comparison of holographic notch and dielectric analysis of the holographic edge filter was presented by Pel-
filters. letier and Reeder.18 Second-generation holographic notch
filters where described by Yang et al.19 Filters with sig-
either the blue or red side of the main spectral band. The nificantly narrower spectral bandwidth and sharper spec-
skewing is a measure of how well you have maintained the tral edges allowed both Stokes- and anti-Stokes-shifted
final l/2 spacing between the hundreds of fringes that form data to be collected below 100 cm1 by angle tuning the
the holographic notch filter. The wavelength range available filter.
for the use of holographic optical elements is determined A holographic notch filter used as the beamsplitter in a
by the transmission characteristics of the gelatin used to Raman microprobe was described by Pallister et al.20 The
construct the element. The spectral plot from 250 nm to combination of high laser reflectivity and high transmission
2500 nm for a 785 nm notch filter is shown in Figure 4. of Stokes-shifted data make it very attractive for use with
The thinner films used to produce volume phase gratings low-power laser sources.
have lower absorption characteristics and can be used from The performance of an improved version of the holo-
below 300 nm to over 2500 nm. Also present is a harmonic graphic notch filter was described by Schoen et al.21 With
100
90
80
70
Transmission (%)
60
50
40
30
20
10
0
250 450 650 850 1050 1250 1450 1650 1850 2050 2250 2450
Wavelength / nm
Figure 4. Plot of transmission versus wavelength for a 785 nm notch filter.

a film thickness of twice that of the original holographic therefore the filter needs to be located between collimating
notch filter, Stokes shifted data were collected below optics.
50 cm1 .
5.3 Optical characteristics

5.2 Operating characteristics
Holographic notch filters are readily available for laser
The ability to rotate the filter from 0 to 10 allows the user wavelengths within the range 3501064 nm. The gelatin
to select a position for the laser line within the reflection films used to produce notch filters have strong absorption
band to optimize the filter performance for the required below 350 nm, which excludes their use for UV applica-
experiment. Figure 5 identifies three possible locations for tions, but at longer wavelengths custom notch filters have
the laser line within the laser line rejection band that provide been successfully produced and used at 1300 nm for FT
attractive tradeoffs in performance: (Fourier transform) Raman experiments using diode lasers.
Laser attenuation is measured as optical density (absorb-
1. la Locating the laser line close to the blue edge ance), spectral bandwidth is measured in wavenumbers
allows for the collection of anti-Stokes data close to between the 0.3 optical density points, and spectral edge-
the laser line. width is measured in wavenumbers between the 0.3 and
2. lb For highly Rayleigh-scattering samples, angle 4.0 optical density points (Table 1). The filter transmission
tuning the filter to position the laser line at the center of away from the notch is typically greater then 80%.
the filters rejection band will maximize the rejection Laser attenuation at an optical density value of 4 is
of unwanted Rayleigh line scattering. (Optical density equivalent to 99.99% of the incident laser light being
typically greater than 6.) reflected. An edgewidth of 150 cm1 between the 0.3 and
3. lc This is the most widely used location. By position- 4.0 optical density points describes the distance between the
ing the laser line close to the red edge, Stokes-shifted laser being reflected by 99.99% and the Raman data being
Raman data can be collected closer to the laser line transmitted at 50%. The rate of change of optical density
while a still-acceptable level of laser line rejection with filter angle is shown in Figure 6.
is provided. (Optical density typically greater than 4,
Raman spectra down to below 50 cm1 .)
Table 1. Comparison of some commercially available holo-
graphic notch filter.
The best filter performance is achieved when the light
passing through the filter is collimated. However, because Specification Laser attenuation Spectral Spectral
excellent laser line rejection over 6 is provided, posi- (optical density) bandwidth edgewidth
tioning the laser line close to the blue edge of the fil- (cm1 ) (cm1 )
ter allows it to be used at the entrance of an f/4 or
slower spectrographs, without the need for collimating Notch >4 <700 <300
optics. Because light is converging through the filter over Notch-Plus >6 <700 <300
SuperNotch >4 <350 <150
a range of different angles the performance is reduced by
SuperNotch-Plus >6 <350 <150
the averaging effect of the filter. For spectrographs faster
than f/4, insufficient angular coverage is provided and
7
100
90 6
80
Optical density
5
Transmission (%)
70
4
60
c
50 3
40
b 2
30 a
20 1
10
0
0
760 770 780 790 800 810 0 5 10 15 20 25
Wavelength / nm Filter angle
Figure 5. Holographic notch filter. Figure 6. Plot of optical density versus filter angle.
4.0
I
3.0
Optical density
Laser
Laser
input
output
2.0
Volume phase
holographic I
grating Spatial filter
1.0
Figure 8. Diagram showing the operation of a laser bandpass
40 20 10 0
filter.
0.0
675 685 695 705 715 725 735 745 755 765 775 785 795
grating at 45 ; the grating diffracts the laser line through
Wavelength / nm 90 so that the output beam is nominally perpendicular to
Figure 7. Plot of optical density versus wavelength for different the exit face of the cube. The grating is spectrally tuned so
filter angles. only the laser line is diffracted at 90 . Undiffracted light
passes through the grating and exits out of the back face
Good laser rejection performance is shown to exist out
of the cube. Unwanted light with wavelengths close to the
to 6 off-normal.
laser line are diffracted by the transmission grating but at
slightly different angles. Unwanted diffracted light can then
5.4 Polarization properties be removed by adding a spatial filter (slit or aperture) and
suitable optics located beyond the exit face of the cube. As
The transmission of a holographic filter is independent of only the laser line of interest is transmitted, the out-of-band
polarization at normal incidence. As the filter is turned attenuation of all other wavelengths is excellent. Also, as
off-normal it becomes polarization dependent; up to 10 off- the holographic grating only passes or diffracts the incident
normal the effects are quite small. Plots of optical density light, unwanted temperature effects due to light absorption
versus wavelength for a range of filter angles between 0 are minimized.
and 40 are shown in Figure 7. Dielectric bandpass filters can absorb out-of-band wave-
For angles between 10 and 40 the optical density lengths and the resulting increase in filter temperature can
drops from 4 down to 1, corresponding to the less efficient cause the main transmission band to shift from its specified
reflection of p-polarized light. wavelength position. This spectral shift reduces the filter
transmission at the laser wavelength and therefore reduces
the laser power through the filter. This effect makes it very
6 LASER BANDPASS FILTERS difficult to use dielectric bandpass filters when variable laser
power is required by the experiment. The values for com-
6.1 Description
mercially available holographic laser bandpass filters are
The function of the holographic laser bandpass filter is to given below.
remove unwanted wavelengths from the plasma discharge
in laser tubes, super radiance generated in laser diodes,
Laser throughput >90% of s-polarized laser line
and silica Raman generated in long runs of optical fiber.
Spectral <2 nm at OD 0.3 or 50% points
Using a volume phase holographic grating a filter with
bandwidth with user supplied spatial filter
narrow bandpass, high transmission and excellent out-of-
(aperture)
band attenuation can be produced. For maximum dispersion
Laser damage >5 W CW unexpanded at 514 nm
the volume transmission grating can be located between two
threshold and 1064 nm
prisms and for ease of use the diffraction angle is designed
Wavelength range Standard wavelengths (nm): 442,
to be 90 . For custom applications the angular dispersion
457, 476, 488, 514, 532, 568, 633,
and diffraction angle can be designed to meet the needs of
647, 752, 785, 1064
the optical design. A diagram outlining the operation of a
Custom wavelengths: between
holographic laser bandpass filter is shown in Figure 8.
400 nm and 1064 nm
The incoming, collimated, s-polarized laser beam is inci-
Size 0.95 in 0.95 in 0.8 in
dent normal to one of the prism faces. The laser beam
enters the cube and intersects the volume phase holographic CW, continuous wave.
7 VOLUME PHASE HOLOGRAPHIC CCD array detector
GRATINGS
Volume phase holographic gratings are now being used as Spatially separated
spectral channels
the dispersive element in high-performance imaging spec-
trographs for Raman spectroscopy. Also, they are now
being evaluated for astronomical spectroscopy and wave-
length division multiplexing applications. Higher diffrac- Raman signal
tion efficiencies, lower optical scattering, easy cleaning and
long-term durability are several of the benefits provided by
Collimated incident light
volume phase holographic gratings over traditional surface
relief gratings. As the diffraction is a volume phase and not
Volume phase holographic grating
a surface relief phenomenon it does not suffer from Woods
Figure 9. Diagram showing the operation of a dual volume phase
anomaly normally associated with surface relief gratings.
holographic transmission grating.
Due to the grainless nature of the dichromated gelatin
emulsion, grating frequencies of 6000 lines per millime-
modulation. Evaluating the grating equation in conjunction
ter can be recorded, gratings with angular dispersion rates
with the Bragg conditions determines the performance and
of 0.45 per nanometer are being used in optical spec-
identifies the Bragg wavelength and Bragg angle, terms that
trum analyzers operating at 1.5 m for wavelength division
exactly satisfy the Bragg condition.
multiplexing applications. The first spectrograph utilizing
More complex volume phase holographic gratings can
a volume phase holographic grating was initially described
be designed to provide unique performance advantages for
and discussed in a review article on the spectroscopic appli-
Raman spectroscopy.
cations of volume holographic optics by Tedesco et al.22
A schematic representation of a dual volume phase holo-
The first commercial product was introduced in 1993 when
graphic grating is outlined in Figure 9. By tuning each
the axial transmissive f/1.8 imaging Raman spectrograph
was described in detail by Battey et al.23 grating to diffract a different spectral slice at a different
The replacement of single surface reflective elements diffraction angle, two separate spectra can be imaged onto
by complex refractive lenses allows for more compact the same CCD array detector without sharing the aperture.
and faster optical designs with significantly better imag- Combining the spectra effectively extends the spectral cov-
ing performance across large image fields. The image of erage, without any reduction in spectral resolution. This
a 10 m slit illuminated by an atomic line onto a charge property is particularly important for Raman spectroscopy
coupled device (CCD) array detector has been measured because the full spectral range can be collected, simultane-
for an f/1.8 axial transmissive spectrograph and an f/4 ously averaging out effects such as laser power fluctuations
CzernyTurner spectrograph. The f/1.8 axial transmissive and changes in the strength of the Raman scattered light
spectrograph broadened the image of the slit by only 4 m, due to changes in sampling conditions that might occur
whereas the f/4 CzernyTurner broadened the image by during the data acquisition period. Internally stable spec-
30 m. The high-quality imaging capability of the axial tra are essential if a quantitative analysis is required on the
transmissive design makes it ideal for multichannel spec- collected data set. A commercial Raman spectrometer using
troscopy when large numbers of closely packed optical this type of HoloPlex grating was introduced in 1995.26
fibers are located at the entrance of the spectrograph and The combined advantages of using a complex volume
imaged onto a large-format CCD array detector located at phase holographic grating within the axial transmissive
the image plane. spectrograph can be seen in Figure 10.
Throughput is significantly better because the f/1.8 axial The eight line images correspond to four input optical
transmissive has a large acceptance angle, increasing its fibers split by the HoloPlex grating into low-frequency
light collection efficiency by almost five times more than a and high-frequency regions. One grating diffracts the low-
CzernyTurner spectrograph operating at f/4. wavenumber region to the top half of the CCD array
Design considerations and a first-order analysis of a detector; the second grating is designed to diffract the
simple volume phase holographic grating were presented high wavenumber region to the bottom half of the CCD
by Arns and colleagues.24,25 To design a volume phase array detector. The dark regions on the top four images
holographic grating to meet a particular specification the correspond to the positions of the Raman wavelengths
wavelength, grating frequency, and construction angles are rejected by the holographic notch filter located in the
used to calculate the required film thickness and index prefilter part of the spectrograph. The images located to the
1 2. D. Gabor, Nature, 161, 777 (1948).

2
3. E.N. Leith and J. Upatnieks, J. Opt. Soc. Am., 52, 1123
3
(1962).
4
1 4. Y.N. Denisyuk, Soviet Phys. Doklady, 7, 543 (1962).
2 5. T.A. Shankoff, Appl. Opt., 7, 2101 (1968).
3
4
6. L.H. Lin, Appl. Opt., 8, 963 (1969).
7. M. Chang, Appl. Opt., 10, 2250 (1971).
Figure 10. Image from a CCD array detector illuminated by an 8. D. Meyerhofer, RCA Rev., 33 (1972).
input array of four optical fibers using a dual volume phase
holographic transmission grating. 9. A. Graube, Opt. Commun., 8, 251 (1973).
10. B.J. Chang, Opt. Commun., 17, 270 (1976).
left of the dark regions on the top four images corresponds 11. T. Kubota, T. Ose, M. Sasaki and K. Honda, Appl. Opt., 15,
to the position where anti-Stokes data would be collected. 556 (1976).
12. B.J. Chang, Opt. Eng., 19, 642 (1980).
8 SUMMARY 13. H. Owen, SPIE, 2406, 260 (1985).
14. R.D. Rallinson, SPIE, 2406, 292 (1985).
Volume phase holographic optical elements are making 15. J.R. Margarinos and D.J. Coleman, SPIE, 2406, 203 (1985).
a major contribution to Raman spectroscopy by provid-
16. H. Kogelnik, Bell Syst. Tech. J., 48, 2909 (1969).
ing notch filters, laser bandpass filters, and volume phase
holographic gratings that provide significant performance 17. M.M. Carrabba, K.M. Spencer, C. Rich and D. Rauh, Appl.
Spectrosc., 44, 1558 (1990).
advantages over previous technologies. New applications
of volume phase holographic optical elements in astronom- 18. M.J. Pelletier and R.C. Reeder, Appl. Spectrosc., 45, 765
(1991).
ical spectroscopy and optical spectrum analyzers are also
providing significant advantages over earlier technologies. 19. B. Yang, M.D. Morris and H. Owen, Appl. Spectrosc., 45,
New, more complex volume phase holographic optical 1533 (1991).
elements are likely to continue to be developed as the needs 20. D.M. Pallister, K.L. Liu, A. Goval, M.D. Morris, H. Owen
and benefits of new applications are better understood. and T.R.Harrison, Appl. Spectrosc., 46, 1469 (1992).
21. C.L. Schoen, S.K. Sharma, C.E. Helsley and H. Owen, Appl.
Spectrosc., 47, 1913 (1993).
ABBREVIATIONS AND ACRONYMS 22. J.M. Tedesco, H. Owen, D.M. Pallister and M.D. Morris,
Anal. Chem., 65, 441 (1993).
CW Continuous Wave 23. D.E. Battey, J.B. Slater, R.S. Wludyka, H. Owen, D.M.
OD Optical Density Pallister and M.D. Morris, Appl. Spectrosc., 47, 1913 (1993).
24. J.A. Arns, SPIE, 2404, 174 (1995).
REFERENCES 25. J.A. Arns, W.S. Colburn and S.C. Barden, SPIE, 3779, 313
(1999).
1. G. Lippmann, Comtes Rendures Acad. Sci. Paris, 112, 274 26. H. Owen, D.E. Battey, M.J. Pelletier and J.B. Slater, SPIE,
(1891). 2406, 260 (1995).
Continuous Lasers for Raman Spectrometry
Ming-Wei Pan, Robert E. Benner and Lee M. Smith
Process Instruments, Inc., Salt Lake City, UT, USA
1 INTRODUCTION 2 IMPORTANT LASER

CHARACTERISTICS FOR RAMAN
Although the Raman effect was first discovered in 1928, it
SPECTROMETRY
remained little more than a laboratory curiosity until the
advent of the laser in 1960. After relatively high-power
The laser characteristics of interest for designing a Raman
monochromatic laser sources became available, interest in
system fall into two categories. The first category includes
Raman spectroscopy grew rapidly. For many years ineffi-
parameters associated with the laser output, such as wave-
cient water-cooled ion lasers were used in conjunction with
length, linewidth, output power, polarization, and beam
large double spectrometers and photomultiplier detectors for
quality. The second category includes physical characteris-
the collection of most Raman spectra. Since the mid-1980s,
tics of the laser system, such as size, reliability, and power
a variety of instrumentation advances have greatly reduced
consumption. Laser output parameters are discussed in this
the size and complexity of the equipment required for the
section. Physical characteristics of specific laser systems
collection of Raman data. These developments include the
are discussed in Section 3.
use of cooled charge coupled device (CCD) array detectors
for parallel detection of Raman spectra with high quan-
tum efficiency and low noise;1 3 the availability of vol-
2.1 Laser wavelength
ume holographic filters and gratings to provide improved
overall throughput and elastic light rejection for Raman
The Raman effect relies on the interaction of monochro-
spectrographs;4,5 the use of excitation sources in the near- matic light with the vibrational and rotational modes of
infrared (NIR) to reduce fluorescence emission compared to molecules to produce scattered light, which is shifted in
Raman scattering;6 10 the availability of economical, long frequency from that of the incident light by an amount cor-
life, laser diodes of sufficient power for Raman applications; responding to the energies of the scattering molecules. Con-
and optical fiber probes for remote chemical analysis by sequently, the measured quantity in a Raman experiment
Raman scattering.11 14 As a result of these advancements, is a frequency shift from the laser excitation wavelength
Raman scattering is now becoming a general-purpose ana- rather than an absolute wavelength. Thus, in principle, any
lytical technique suitable for on-line chemical monitoring. excitation wavelength can be used to obtain a Raman spec-
This article considers the issues associated with the choice trum. However, the Raman scattering efficiency varies with
of excitation source for continuous wave (CW) Raman the excitation frequency to the fourth power (!4 ) which
experiments, recognizing the advances which have occurred favors the use of ultraviolet (UV) and visible lasers oper-
in Raman instrumentation since the late 1980s. The dis- ating at short wavelengths. The problem with using short-
cussion is limited to CW lasers and spontaneous Raman wavelength lasers for excitation of Raman spectra is that
scattering under nonresonant conditions, which is the most the high-energy photons often induce fluorescence which is
commonly applied form of Raman spectroscopy today. much more intense than the Raman emission from the sam-
ple under study. For some fluorescing samples, a deep blue
John Wiley & Sons Ltd, 2002. excitation wavelength can be chosen that shifts the portion
of the Raman spectrum of interest to a spectral region with Useful laser excitation wavelengths for Raman spec-
no fluorescence emission, but these samples are rare. If the troscopy range from the UV (200 nm) to the NIR
fluorescence lifetime is relatively long, the amount of fluo- (1100 nm).17 In the UV region, available CW lasers include
rescence that appears in a collected Raman spectrum can be argon ion and krypton ion, which both produce fun-
reduced by using a pulsed laser for excitation in conjunc- damental emission lines in the UV spectral range with
tion with time-gated detection. This is possible because the relatively high powers available from the large-frame
spontaneous Raman process is quite fast. However, the most models. In addition, CW frequency-doubling techniques
general approach for reducing fluorescence is to use NIR can be used to produce numerous additional UV wave-
wavelengths for excitation of Raman spectra. The NIR pho- lengths from argon ion, krypton ion, and CW frequency-
tons do not have sufficient energy to populate the excited doubled Nd : YAG lasers. It is also possible to use CW
electronic states within the sample which give rise to fluo- frequency-doubling methods to achieve wavelength-tunable
rescence, and the Raman spectrum can be detected without outputs in the UV from tunable dye lasers which have
a strong fluorescence background. Using NIR excitation it fundamental outputs in the visible region. In the visi-
becomes possible to record Raman spectra from laser dyes ble region, the primary CW laser candidates for Raman
such as rhodamine with a high signal-to-noise ratio (S/N).15 excitation include frequency-doubled Nd : YAG (532 nm),
The primary penalties for using NIR excitation are the !4 argon ion (458, 488, and 514 nm), krypton ion (476,
losses as well as a loss in detection efficiency if the Raman 520, 531, 568, 647, and 676 nm), HeNe (heliumneon;
emission is shifted too far into the infrared (IR) range. 633 nm), semiconductor (400410 and 630700 nm), and
Several approaches are available for collecting Raman dye lasers (400700 nm). In the NIR region krypton (752
spectra with NIR excitation. Dispersive Raman systems and 799 nm), Ti : sapphire (6701100 nm), semiconduc-
normally employ a diffraction grating spectrograph in con- tor (7001100 nm), Nd : YAG (1064 nm), and Nd : YLF
junction with a silicon CCD detector with a typical response (yttrium lithium fluoride; 1047 and 1053 nm) lasers are
curve, as shown in Figure 1. As these CCD-based systems available. The wavelengths in parenthesis represent major
are limited to detecting wavelengths shorter than about lines, but are not all inclusive.
1050 nm, they are often used with excitation at 785 nm
which gives a Raman shift of about 3200 cm1 at 1050 nm.
Some back-thinned and deep-depletion CCD designs are 2.2 Output power
becoming available to improve the NIR response, but these
are still fundamentally limited by the bandgap of silicon. Firm rules for the amount of power required to produce
Fourier transform (FT) Raman systems typically utilize a Raman spectrum are difficult to formulate because the
excitation at 1.06 m from a neodymium-doped yttrium amount of laser power is dependent on the type of sample
aluminum garnet (Nd : YAG) laser to avoid fluorescence,16 being studied, the laser excitation wavelength, the nature of
but normally require high laser power and high-throughput the focusing and collection optics used, the overall through-
FT interferometers to achieve good S/Ns because of their put and detection efficiency of the Raman instrument, the
lower-sensitivity IR detectors. available data acquisition time, and the desired S/N. Nev-
ertheless, some general guidelines can be provided. First,
100
as the Raman signal is proportional to excitation power,
high laser powers will decrease the required data acqui-
80 sition time for a particular S/N as long as the sample
Quantum efficiency (%)
is not damaged. Depending upon whether the sample is

absorbing or transparent at the excitation wavelength, the
60
damage threshold might be as low as a few milliwatts if
the excitation light is tightly focused onto the sample using
40 microscope optics. In contrast, for Raman studies of trans-
parent gases the laser power used is typically limited by
20
that available from the laser source and can easily exceed
tens of watts without sample damage. Most CW Raman
applications utilize laser powers between these extremes
0 and probably average in the range from 100 mW to 1 W.
200 300 400 500 600 700 800 900 1000 1100
With these typical power levels, Raman spectra can often
Wavelength / nm
be obtained from pure bulk samples in integration times
Figure 1. Typical response curve of a silicon CCD detector. of 110 s with good S/N. Obtaining recognizable spectra
Continuous Lasers for Raman Spectrometry 3
from minor components in mixtures, thin films, or materials 2.5 Polarization

with weak Raman cross-sections can require much longer
integration times. Although not essential for collecting Raman spectra, utiliz-
ing a laser with a linearly polarized output beam is often
convenient for a variety of reasons. When collecting spectra
2.3 Linewidth from crystalline materials, the laser polarization can be used
in combination with the polarization of the scattered light
Three factors affect the overall resolution of the Raman to determine crystal orientation based on selection rules for
spectrum: laser linewidth, spectrograph resolution, and the crystal symmetry. Even when recording Raman spec-
the inherent linewidth of the vibrational transition being tra from noncrystalline solids and thin films, a polarized
probed. Ideally, the laser linewidth should be narrower than laser output can be used to maximize the coupling of laser
the other two factors. Fortunately, most commercial CW light to the sample by taking advantage of Brewsters angle
lasers have narrower linewidths than usually required. For effects. However, to achieve maximized Raman signals, the
example, an argon-ion laser operating at 488 nm in a single laser polarization direction must be considered in conjunc-
line, but with multiple longitudinal modes, has a linewidth tion with the scattering geometry to avoid minima in the
of 4.5 GHz or 0.15 cm1 . Most modern Raman spectro- dipole radiation pattern.
graphs employing CCD detectors have spectral resolutions If a polarized laser beam is used with an optical fiber
of 26 cm1 . One commonly used CW laser which does not probe to collect Raman spectra, the laser polarization will
inherently have a sufficiently narrow linewidth for Raman be scrambled by a multimode excitation fiber. Similarly,
studies is the multimode semiconductor laser, which typi- any polarization of the scattered light will tend to be scram-
cally has a linewidth of 1.5 nm (25 cm1 ). As discussed bled by the collection fiber(s). As most of the commercially
later in this chapter, an external-cavity design can be used available probes operate in a backscattering geometry, the
with semiconductor lasers to reduce their linewidths and loss of polarization does not degrade the signal, but it does
stabilize their output wavelengths.18,19 A laser linewidth of prevent using the polarization characteristics to determine
1 cm1 is suitable for most applications. sample orientation.
2.4 Beam quality 3 COMMONLY USED CW LASERS FOR

RAMAN SPECTROMETRY
For free-space and microscopy Raman applications, a
nearly diffraction-limited transverse electromagnetic Laser technology has developed to the extent that essen-
(TEM)00 laser beam is preferred (the subscript denotes tially any laser wavelength can be obtained in the range
the transverse mode of the lowest order in the cavity). of interest for Raman applications, from 200 to 1100 nm,
A typical focused spot diameter for a Raman microprobe using a combination of discrete laser lines and tunable lasers
system is 12 m, which is difficult to achieve without a with and without frequency doubling. Furthermore, essen-
TEM00 laser beam. However, many Raman spectra are now tially any laser with suitable linewidth and power can be
being collected using commercially available optical fiber used for Raman spectroscopy because the Raman spectral
probes which utilize a multimode fiber with a diameter of bands are always shifted by a constant frequency (energy)
50300 m for the excitation fiber. As long as the out- relative to the excitation frequency; i.e. under nonresonant
put can be coupled efficiently to these multimode fibers, conditions Raman spectra are independent of the excita-
an excitation laser does not need to have especially good tion wavelength. For these reasons, only those CW lasers
beam quality for Raman applications that use optical fiber mostly commonly used for Raman scattering are surveyed
probes. here. More complete surveys are readily available in other
Most of the commercially available CW gas, dye, solid- texts.20 26
state, and low-power (1 3 m emitting area) semicon-
ductor lasers can operate with nearly diffraction-limited
beam quality. However, high-power (broad stripe >10 m) 3.1 Gas lasers
semiconductor lasers have poor beam quality with strong
astigmatism. These high-power semiconductor lasers can Gas lasers have become very common since their first
be coupled efficiently into multimode optical fibers, but appearance in 1961. The gaseous media are relatively
they are difficult to use with Raman microprobe inexpensive and have a much higher damage threshold
systems. than crystals and semiconductors. Depending upon the gain
medium, good beam quality and high output power are design a HeNe laser for operation at an output power
often easy to achieve. In addition, the commonly used gas greater than 50 mW because of a need for wall collisions
lasers have multiple lasing transitions which can be used to depopulate the lower level of the lasing transition. Most
to obtain Raman excitation wavelengths from the UV to HeNe lasers are designed to operate in the fundamental
the NIR. Although diode-pumped solid-state (DPSS) lasers TEM00 mode, but the ratio of the length of the gas tube
have displaced a portion of the gas laser market recently, to its diameter is critical to obtain TEM00 operation. The
gas lasers are still widely used, especially to obtain visible emission linewidth of a HeNe laser operating at 633 nm is
and shorter wavelengths. about 1.5 GHz full-width at half-maximum (FWHM), which
allows several longitudinal modes to lase simultaneously
3.1.1 Heliumneon laser for a cavity length of 0.150.5 m. HeNe lasers stabilized
to operate in a single longitudinal mode can be purchased,
One of the most widely used lasers today, the HeNe laser but this is not necessary for most Raman applications.
is based on a gain medium of helium and neon mixed at a An inconvenience of using HeNe lasers for excitation
ratio of about 10 : 1. This laser can provide a maximum CW of Raman spectra is that the neon has a large number of
power of about 50 mW at the most utilized wavelength in spontaneous emission lines which can often be observed
the red part of the spectrum at 633 nm, although other lasing in collected Raman spectra. These lines can usually be
output wavelengths (including 543, 1152, and 3391 nm) removed by passing the laser beam through a laser-line-
can be obtained.27 In general, HeNe lasers are compact, transmitting interference filter before it enters the sample,
lightweight, and reliable. The power supply for a HeNe or by using some dispersive element such as a prism or
laser typically must provide a high voltage of up to 1 kV grating to remove the spontaneous emission lines from the
in order to supply the necessary current density (about excitation beam path. Either of these approaches result in
100 mA cm2 ) for excitation. some loss in excitation power.
The excitation mechanism for a HeNe laser begins Overall, todays HeNe lasers represent well developed
with excitation of the helium atoms to upper energy states and mature technology with ease of use and good reliability.
through collisions with electrons in the electrical dis- For Raman applications, the available output power is low,
charge process. Population inversion in the neon atoms is but the linewidth and beam quality are ideal. Collecting
achieved by resonant energy transfer through atomic colli- Raman spectra with excitation at 633 nm provides a good
sions to upper energy levels of the neon atoms. The helium match to the quantum efficiency curve of a typical CCD
metastable energy levels accumulate population because (Figure 1), but this wavelength is often not long enough
they do not have radiative decay to the ground state. The to suppress sample fluorescence. A summary of the basic
simplified energy level diagram of Figure 2 illustrates the characteristics of HeNe lasers is given in Table 1.
excitation and emission processes.
The output power of HeNe lasers ranges from about
0.5 mW to about 50 mW. Unfortunately, it is difficult to 3.1.2 Argon-ion lasers
The argon-ion laser is a major workhorse for conventional
Atomic
He 2 s1
collision Ne 3s dispersive Raman spectroscopy. Although they are less effi-
3391.2 nm cient than many other lasers (about 0.1% efficiency) and
3p they require water cooling at higher output powers, argon
Atomic lasers are probably the most commonly used lasers in lab-
3
2 s collision 2s 632.8 nm
oratories for Raman spectroscopy because they emit in the
1152.3 nm 2p blue-green spectral region and have relatively high power
Table 1. Typical HeNe laser characteristics.

1s Fast
decay
Parameter Value
Electron
excitation Collision Major laser line 632.8 nm
with walls
Spontaneous emission linewidth 1.5 GHz
Output power 0.550 mW
Ground states Transverse mode Single mode (TEM00 )
Laser cavity length 0.150.5 m
Helium Neon
Power stability 5% per hour
Figure 2. Energy level diagram for a HeNe laser showing laser Lifetime 20 000 h
transitions.
Table 2. Major emission lines of a the argon-ion ground state. The laser transition linewidth of
5 W argon-ion laser. a typical CW argon laser is in the order of 4.0 GHz. This
allows about 27 longitudinal modes to lase simultaneously
Wavelength (nm) Power (W)
for a laser cavity length of 1.0 m.
The argon laser has higher gain than the HeNe laser
333.6363.8 1.0
454.5 0.15 and can produce much higher output powers. The output
457.9 0.35 power is proportional to the square of the current density,
465.8 0.15 and current densities as high as 100 A cm2 can be used
472.7 0.25 in argon lasers.27 This high current density also produces
476.5 0.6
erosion and heating of the gas tube. The high-temperature
488.0 1.6
496.5 0.6 plasma tube usually has a bore diameter of 12 mm and
501.7 0.35 lengths that range from 0.1 m to approximately 1.8 m. High
514.7 2.1 power (above 3 W) argon-ion lasers also utilize a magnetic
528.7 0.35 field in the direction of the discharge current to prevent
electrons from hitting the tube walls. The argon-ion laser
(>2 W) in multiple lines. The argon-ion laser can provide is rather inefficient. For example, a medium-power laser
more than 35 wavelengths, ranging from 275 to 686 nm. with an output power of 5 W all-lines and 2 W single-
In the blue-green spectral region, CW powers of up to line needs a 240 V three-phase power supply and at least
25 W are available with two major strong lines at 488.0 and 35 A for pumping. To remove the heat generated in the
514.5 nm. Also, an intracavity frequency-doubled argon-ion gas tube, water cooling is necessary with a cooling rate of
laser can emit wavelengths as short as from 229 to 257 nm. at least 2 gal min1 and at a pressure of 25 lb in2 . High
Table 2 shows the major lasing lines of a 5 W (CW) argon electrical power consumption, the need for water cooling,
laser in the visible region and their output powers. and a typical lifetime that ranges from 2000 to 5000 h are
For a CW argon laser, the excitation mechanism involves characteristics that detract from the popularity of ion lasers.
a two-step process (Figure 3). Initially, the electrons collide The basic characteristics of the argon-ion laser are given in
with ground-state neutral argon atoms, and this produces Table 3.
ground-state argon ions. These argon ions then collide with
the electrons again to raise themselves to the upper laser 3.1.3 Krypton-ion lasers
states. The laser transitions thus occur between the argon-
ion ground state and the argon-ion upper laser states. The Krypton-ion lasers are very similar to argon-ion lasers in
population inversion is also achieved by the rapid decay of excitation mechanism and construction. The krypton-ion
laser provides many additional laser wavelengths that range
4p2 from 406.7 to 799.0 nm. The wavelengths toward the red
part of spectrum are often used for Raman excitation to
4p4 454.5, 465.8, 457.9, 476.5, reduce fluorescence. This is a major advantage of using a
472.7, 488.0 nm 496.5 nm
krypton-ion laser over an argon-ion laser. Table 4 shows
the major lasing lines in the visible region of a 14 W
514.5, 528.7 nm (CW) krypton ion lasers in the visible region and their
4s2 output powers. As argon and krypton lasers were discovered
Electron shortly after the first laser invention in 1960, they have been
collisions very good research tools in Raman laboratories because
Table 3. Typical argon-ion laser characteristics.

3p5 Ar+ ground state
Parameter Value
Pump Electron collisions

Major laser lines 488.0, 514.5 nm
Output power 0.150 W
Transverse mode TEM00 or multimode
Laser cavity length 0.12 m
3p6 Ar ground state
Power stability 0.2% per hour
Figure 3. Energy level diagram for an argon-ion laser showing Lifetime 20005000 h
the two-step excitation process and laser transitions.
Table 4. Major emission lines of a already reached a mature stage and continues to grow
14 W krypton-ion laser. rapidly. New materials are being found and novel designs
are invented so frequently that it is impossible to include
Wavelength (nm) Power (W)
them all here only the solid-state lasers most commonly
used for Raman spectrometry are considered here.
337.5356.4 2.0
406.7 0.9
413.1 1.8 3.2.1 Nd : YAG lasers
415.4 0.3
468.0 0.5 The Nd : YAG laser has been by far the most commonly
476.2 0.4 used solid-state laser. The fundamental laser line at 1064 nm
482.5 0.4 is currently used as the major excitation wavelength for FT
520.8 0.7
530.9 1.5 Raman spectroscopy. Its frequency-doubled version emit-
568.2 1.1 ting at 532 nm is also an important excitation line for
674.1 3.5 dispersive Raman. The Nd : YAG lasers provide several
676.4 0.9 advantages over other solid-state lasers: the Nd : YAG mate-
rial is stable, hard, and easy to polish to high quality optical
they have numerous lasing lines in the visible region and finishes;28 for some compact designs, high-reflectivity coat-
provide high output power. In fact, lasers with a mixture ings can even be employed on the end surfaces of the
of argon and krypton gases are commercially available. laser rod to act as cavity mirrors; Nd : YAG has good ther-
When they are used as sources for Raman experiments, mal conductivity that allows high CW power operation; as
both lasers typically require some type of plasma line filter Nd : YAG lasers do not require the constant refill of con-
to remove nonlasing spontaneous emission lines. The basic sumable gases, maintenance costs are reduced compared to
characteristics of the krypton-ion laser are summarized in gas lasers.
Table 5. The Nd : YAG laser involves the four-level system shown
in Figure 4. Its upper energy level lifetime is relatively
long (230 s) which makes population inversion easy to
achieve. Also, it has a much lower pumping threshold than
3.2 Solid-state lasers
that of the ruby laser (a three-level system). Absorption
occurs from the ground state 4 I9/2 to the upper states and
The active media of solid-state lasers consist of a small
then the energy relaxes rapidly to the metastable state
percentage of metal ions in an insulating host material 4
F3/2 . A large amount of heat is generated from the laser
such as a glass which can be readily fabricated into rods
rod because of nonradiative decay to this state. Thus, for
with two polished ends. Two optical excitation methods
are typically used, flash-lamp pumping and laser pumping. Pumping bands
Flash lamps provide an inexpensive source for pumping,
but their operating lifetime is short, and they are less
efficient in power conversion. Laser pumping provides Nonradiative
better optical-to-optical conversion efficiency, but has a decay
higher cost. The recent availability of diode laser pumping
makes solid-state lasers more attractive than ion lasers 4
F3/2
for some applications because they are compact, efficient,
reliable, and air cooled. Solid-state laser technology has
Laser transition
1064 nm
Table 5. Typical krypton-ion laser characteristics. Pump
4
I11/2
Parameter Value
Major laser lines 530.9, 568.2, 674.1 nm Nonradiative

decay
Output power 0.120 W 4
I9/2
Laser cavity length 0.12 m Nd : YAG ground state
Power stability 0.2% per hour
Lifetime 20005000 h Figure 4. Energy level diagram of the Nd ion and its excitation
process.
flashlamp-pumped Nd : YAG lasers, the total efficiency is laser rod act like a thick positive lens.30 The Nd : YLF laser
low, only about 0.11.0%. The laser transition occurs from is much less susceptible to the thermal lensing problem.
the metastable state 4 F3/2 to the 4 I11/2 state, producing Depending on the polarization, two different wavelengths,
an emission band which peaks at 1064 nm. This emission 1047 and 1053 nm, are possible. Another advantage of
band has a FWHM linewidth of 0.45 nm. The laser Nd : YLF is that it can store energy about twice as long
rod of Nd : YAG typically has a length of 10 cm and a as Nd : YAG.
diameter of 1.2 cm. The host material, yttrium aluminum A relatively new solid-state laser material based on
garnet (YAG) (Y3 Al5 O12 ), is usually doped with 0.725% of neodymium doping is Nd : YVO4 .31 33 It has the same emit-
neodymium by weight. ting wavelength as Nd : YAG at 1064 nm, but the Nd : YVO4
The Nd : YAG lasers can be pumped either by broad material has broader and stronger absorption bands than
band flashlamps or by other lasers. Diode pumping,29 a Nd : YAG. For the intracavity frequency-doubled laser,
recently developed technology, provides the best over all single-mode operation can be easily achieved by using the
efficiency of about 58%. A high power multimode diode anisotropic emission cross-section of the Nd : YVO4 and
laser emitting at 810 nm matches the absorption peak of a nonlinear crystal such as potassium titanyl phosphate
the Nd : YAG very well and provides very good optical-to- (KTP; KTiOPO4 ) acting as a birefringent filter. An intra-
optical efficiency. This conversion efficiency of the pump cavity DPSS laser usually produces chaotic output when
light to the fundamental (TEM00 ) laser light can be as multiaxial modes co-exist in the laser cavity, but single
high as 53%. Although diode-pumped Nd : YAG lasers have longitudinal mode operation provides stable output power.
not yet reached the maximum power (over 100 W) of the This instability is believed to arise from nonlinear cou-
flashlamp-pumped lasers, output powers greater than 30 W pling of axial modes via sum-frequency generation. Output
at 1064 nm and 5 W at 532 nm are commercially available. power greater than 3.2 W for a single-mode diode-pumped
Diode-pumped solid state lasers also have several advan- Nd : YVO4 /KTP green laser has been demonstrated.
tages over the lamp-pumped lasers. They offer 110 V oper-
ation, high efficiency, air cooling, small footprints, better
lifetimes, and portability. Some commercial DPSS devices
3.3 Tunable lasers
are now equipped with field-replaceable diode pumping
units that are fiber coupled to the laser head, thereby making
Laser tunability in CW nonresonant Raman spectroscopy
maintenance simple. The basic characteristics of diode-
is usually not necessary as there are many discrete laser
pumped Nd : YAG lasers are summarized in Table 6.
lines available with high power, from the UV to the NIR.
However, tunability is essential in resonance Raman spec-
3.2.2 Nd : YLF and Nd : YVO4 lasers troscopy and can be helpful for finding the best excitation
Although YAG has been one of the most important laser wavelength for an unknown sample or for adjusting the
host materials, the neodymium ion can also be doped into a laser wavelength to match bandpass and rejection filters.
variety of other host materials. Two host materials, yttrium Only the two most commonly used tunable lasers are dis-
lithium fluoride (YLF) and yttrium vanadate (YVO4 ), are cussed here: titanium sapphire and dye lasers.
becoming increasingly popular.
The Nd : YLF material is preferred when high beam qual- 3.3.1 Titanium sapphire lasers
ity and high power are both necessary. When pumping a
laser medium with high power, a radial temperature gradi- The titanium sapphire (Ti : Al2 O3 ) laser is a widely used
ent is formed that causes thermal lensing which makes the tunable laser because of its broad-gain bandwidth. The
tuning bandwidth typically ranges from 670 to 1100 nm in
Table 6. Typical diode-pumped Nd : YAG laser characteristics. the NIR region and frequency-doubling extends the tuning
range from 335 to 525 nm. The Ti : Al2 O3 lasers have a large
Parameter Value
stimulated emission cross-section, but a rather short lifetime
(3.2 s) of the upper laser level. Because of this unusually
Major laser lines 1064 nm, 532 nm (frequency doubled)
Spontaneous emission 120 GHz (1064 nm) short lifetime, Ti : Al2 O3 lasers are usually pumped by other
linewidth lasers. Commercial Ti : Al2 O3 lasers are typically pumped
Output power 2.520 W (1064 nm), by an argon ion laser (488 nm) or a frequency-doubled
0.210 W (532 nm) Nd : YAG laser (532 nm) because its absorption band ranges
from about 400 nm to 630 nm and peaks at around 490 nm.
Power stability <2% per hour
Lifetime 5000 h (limited by laser diodes) As this absorption band is currently not in the range
where semiconductor lasers operate, it is not suitable for
1.0 tuning range of other tunable lasers, does not occur. Both
Ti : sapphire the ground state, 2 T2 state, and the excited state, 2 E state,
Absorption/emission (arb. units)
0.8
have a broad range of vibrational levels extending from
the bottom to the top. Laser pumping first excites the Ti3C
ion from lower vibronic levels of the 2 T2 state to the broad
0.6 range of vibronic levels of the 2 E state. The population then
relaxes rapidly to the lower vibronic levels of the 2 E state.
0.4
Laser emission occurs when the population decays (radi-
ates) back to any one of the vibronic levels of the 2 T2 state.
After the laser emission, the population remaining in the
0.2
Absorption Emission upper vibronic levels of the 2 T2 state will then again relax
spectrum spectrum to its lower vibronic levels with a distribution based on a
0.0
Boltzmann relationship. Thus, this is a four-level laser sys-
300 400 500 600 700 800 900 1000 1100 tem involving phonon-assisted processes. The basic char-
Wavelength / nm acteristics of Ti : Al2 O3 lasers are summarized in Table 7.
Figure 5. The emission and absorption spectra of Ti : sapphire.
3.3.2 Dye lasers
semiconductor laser pumping. The emission and absorption
spectra of Ti : Al2 O3 are depicted in Figure 5. The laser Dye lasers are among the few liquid lasers that are com-
can produce a maximum CW output power of 50 W and a mercially available. The major advantage of the dye lasers
diffraction-limited beam quality. Although Ti : Al2 O3 lasers is their broad tunability ranging, from 320 to 1200 nm.34
are robust and tunable, they are expensive and not as However, to achieve this tuning range requires using many
compact as DPSS lasers. different dyes. For some applications that need both narrow
The Ti : Al2 O3 laser is a vibronic laser in which both linewidth and fine tuning of the wavelengths, dye lasers can
the electronic and vibrational states participate in laser be a good choice. Although gas and solid-state lasers cover
transitions.30 The energy level structure and the optical a large number of discrete wavelengths ranging from the
transitions are illustrated in Figure 6. For the 3d ! 3d elec- UV to the NIR region, dye lasers provide continuous tun-
tronic transitions, the Ti3C ion interacts very strongly with ing in the same region and fill substantial portions of the
the host, Al2 O3 . Also, as the Ti3C ion only has a single d optical bands that are not covered by those lasers. A list of
electron (3d1 ) in the valence shell, it splits into two states, commonly used dyes and their tuning ranges is shown in
2
E and 2 T2 . Because of the simplicity of the energy-level Table 8. Dye lasers are, however, not as compact as solid
structure, excited-state absorption, an effect that reduces the state lasers, and require regular dye changes. In addition,
some dyes are health hazards.
The absorption and emission processes of a dye laser
can be explained by a simple harmonic oscillator model.
2 As dye molecules are typically dissolved in liquid solvents,
E
their electronic states involve both vibrational and rotational
Electronic potential energy
states that broaden because of collisions with solvents. A

Vibronic
states typical energy-level diagram for a dye laser, a four-level
laser system, is shown in Figure 7. Each of those energy
Absorption
states contains many vibrational and rotational sublevels
Emission
2
Table 7. Typical Ti : Al2 O3 laser characteristics.
T2
Parameter Value
Vibronic
states Major laser lines 6701100 nm
Spontaneous emission linewidth 100 THz
Output power Up to 50 W
Lattice coordinate Power stabilitya 0.22% per hour
Lifetimea 30005000 h
Figure 6. Energy level structure and optical transitions of
Ti : sapphire lasers. a Limited by pumping sources.
Table 8. Commonly used dyes and their tuning ranges. Table 9. Typical dye laser characteristics.
Dye Peak wavelength (nm) Tuning range (nm) Parameter Value
Polyphenyl 2 384 370406 Major laser lines 3201200 nm

Stilbene 1 415 403428 Spontaneous emission linewidth 3040 nm
Stilbene 3 435 410485 Output power up to 2 W
Coumarin 102 482 463515 Transverse mode TEM00 or multimode
Coumarin 6 535 510550 Power stabilitya 0.22% per hour
Rhodamine 110 550 530580 Lifetimea 30005000 h
Rhodamine 6G 575 560625
a Limited by pumping sources.
Sulforhodamine B 625 598650
DCM special 645 610695
Pyridine 2 720 675783 large amount of the population, and this reduces the useful
Rhodamine 700 740 690785 population for laser operation. Another triplet state (T2 ) also
Styryl 9 830 785900 can interfere with the laser transitions via the absorption of
the emitting laser energy which takes place between the T1
and T2 states. To avoid these competing processes in a CW
T2 dye laser, the dye is often required to be circulated with a
Fast circulation time long enough for the molecules in the T1
S1
decay state to decay back to the S0 state. The typical concen-
tration for a dye in solvents is about 104 to 103 molar.
The tuning range depending on dyes is typically 3040 nm.
Absorption
Collisions The basic characteristics of dye lasers are summarized in
Table 9.
Pump Emissions T1
3.4 Semiconductor lasers
Semiconductor lasers have several advantages over other

types of lasers because they are compact, robust, air-cooled,
Slow decay efficient, and easy to use. Near-IR semiconductor lasers are
also excellent candidates as excitation sources for Raman
applications because of the fluorescence reduction with
S0
NIR excitation.35 With excitation from 785 to 850 nm,
Raman shifts in the fingerprint region (502000 cm1 )
Figure 7. Typical energy levels and optical transitions of dye can be detected with high quantum efficiency using sen-
lasers.
sitive silicon CCD arrays. However, semiconductor lasers
also exhibit some significant drawbacks. Single-mode diode
and forms a continuous band. Thus, the absorption and lasers have a narrow linewidth which is more than adequate
emission spectra are also broad. When the dye is pumped, for Raman measurements, but they are limited in output
some of the ground-state (S0 ) molecules are excited to state power, have a strong tendency to mode hop causing changes
S1 and then relax quickly to the lowest sublevel of S1 . Heat in Raman peak positions, and are prone to damage from
is generated during these nonradiative relaxation processes. feedback when used in conjunction with optical fibers with-
Because the upper sublevels of the S0 are initially empty out optical isolation. Conversely, broad-stripe (high power)
according to Boltzmann statistics, population inversion is diode lasers have good output powers (greater than 1 W),
achieved between the lower sublevels of S1 and the upper but their emission linewidths (2 nm) are too broad for
levels of S0 . Raman applications and their beam quality is relatively poor
There are, however, competing processes that can reduce because of the large source size (typically 1 100 m).
the stimulated emission probability, and thus the output Finally, the amplified spontaneous emission (ASE) from a
power. A triplet energy state T1 will act as a trap for those semiconductor laser can be significant (0.11% of the laser
excited molecules in the sublevels of S1 because some of line) and very broad (2030 nm). Thus, a good bandpass
them will make transitions to the T1 state through collisions. filter with sufficient rejection of the ASE is necessary.
The T1 state has a very long lifetime and it has nonradia- Although diode lasers with output powers less than
tive decay to the S0 state. As a result, the T1 state obtains a 200 mW are usually single transverse mode and generally
operate in a single longitudinal mode, modest environmen- increases the spectral and temperature stability. The laser
tal fluctuations or the presence of even a small amount linewidth is typically less than 4 MHz, which is more than
of optical feedback can induce longitudinal mode hopping, adequate for most Raman applications.
causing the effective laser spectral width to increase by The DBR lasers also have some drawbacks. First, they
orders of magnitude. Such optical feedback can be avoided only have a maximum power of 150 mW. Second, as
through the use of optical isolators, but these isolators are they provide single transverse-mode output, they are very
still expensive. However, by stabilizing the laser in an sensitive to external optical feedback which will often
external or internal cavity containing a diffraction grating, cause facet damage on the laser. An optical isolator is
the sensitivity to optical feedback can be greatly reduced, usually needed which adds cost to the system. Third, only
and the laser can be forced to operate with a greatly nar- one wavelength at 852 nm is available commercially. At
rowed and much more stable optical spectrum.18 Such laser this wavelength, the useful detectable range for Raman
devices emitting at NIR wavelengths (780850 nm) are spectroscopy is limited because of the wavelength response
becoming popular for use in turnkey Raman analyzers. Two of the silicon CCDs. Also, the cross-section for Raman
leading types of semiconductor lasers useful for Raman scattering is about 28% less than with 785 nm excitation
spectroscopy are discussed here: distributed Bragg reflec- based on the !4 scattering law. In the future, DBR lasers
tor (DBR) lasers and external-cavity semiconductor lasers are likely to become available at shorter wavelengths and
(ECSLs). higher power. The basic characteristics of DBR lasers are
summarized in Table 10.
3.4.1 Distributed Bragg reflector lasers
3.4.2 External-cavity semiconductor lasers
Compared to an index-guided (FabryPerot) semiconductor
laser, a DBR laser consists of two additional sections The ECSL devices are more expensive than DBR lasers,
on each side of the active region. These built-in sections but they provide good frequency stability and higher out-
not only provide a linewidth narrowing function but also put power (greater than 500 mW).37 41 They also have a
act as mirrors whose reflectivity is maximized at one much broader wavelength range available, from 630 nm
particular wavelength. Figure 8 shows a typical DBR laser to about 850 nm. An ECSL uses the semiconductor chip
structure. Fabrication of DBR semiconductor lasers requires only as a gain medium and employs an external grating
advanced and sophisticated technology. Holographic or as the reflective mirror and also the frequency selector.
electron-beam lithographic techniques are usually used to High-power ECSLs (around 300 mW) have been built and
form a grating with submicrometer periodicity. Once the used successfully for Raman monitoring of industrial pro-
grating pattern has been etched onto the substrate, multiple cess control lines. These lasers are chosen for industrial
epitaxial layers are then grown by using standard epitaxy uses because they are compact, efficient (low electrical-
methods. Although DBR lasers are difficult to make, they power consumption), reliable (lifetime above 10 000 h), air-
have been commercialized for several years. However, they cooled, and good for fluorescence reduction (excitation at
are currently commercially available only in one fixed about 800 nm). In todays Raman industry, a few spe-
wavelength at 852 nm and with a maximum output power cific wavelengths, such as 670, 785, 830, and 852 nm,
of 150 mW. are gaining popularity as standard excitation wavelengths,
The DBR lasers are superior to other free-running semi- and some commercial Raman fiber probes have been built
conductor laser devices in elimination of mode hopping, and designed for use at these wavelengths. Commercial
frequency hysteresis, and broad-band ASE.36 The built-in ECSLs have linewidths ranging from a few megahertz
grating not only provides optical feedback to the laser- to about 30 GHz which is again more than adequate for
active region in a very narrow spectral range but also most Raman applications. Recent data also show that the
intensity stability of an ECSL can be better than 0.2% and
Table 10. Typical DBR diode laser characteristics.

DBR p-type DBR
Parameter Value
n-type Major laser line 852 nm

Laser linewidth 4 MHz
Output power 150 mW
Active region Transverse mode nearly diffraction limited
Lifetime 10 000 h
Figure 8. Typical DBR laser structure.
the frequency stability (jitter) can be better than 0.2 cm1 sent directly back to the laser to provide optical feedback
over a 7-day period.42 only over a narrow wavelength range. However, the output
There are three basic cavity configurations used to con- beam is the light that is emitted from the front facet. This
struct ECSLs with a diffraction grating: Littrow type-I, type of design is not as convenient as the type-I design as it
Littrow type-II, and Littman configurations. For the Lit- requires access to and antireflection coatings on both facets.
trow type-I design (Figure 9a), light from the laser diode is Cost is another drawback of this design as the antireflection
collimated and sent to the grating at an angle of incidence coatings require custom fabrication. In addition, this design
qLittrow . The first-order diffracted light is sent directly back is more sensitive to external optical feedback.
to the laser to provide optical feedback only over a narrow In the Littman design (Figure 9c), the collimated laser
wavelength range. The output beam is the light which is light strikes the grating at near grazing incidence, so that the
specularly reflected from the grating (the zeroth order), so diffracted order does not return to the laser directly. Instead,
that the output is at an angle which is approximately twice this diffracted light is reflected by a mirror, diffracted by the
qLittrow from the optical axis. Because of the simple design, grating a second time, and then returned as optical feedback
the external-cavity length (i.e. the distance from laser to to the laser diode. The advantage of the double-pass geom-
grating) can be made quite short, which in turn causes the etry is that the grating bandwidth is less than half of what it
external-cavity modes to be widely spaced. A wide mode is in the Littrow case. However, the external-cavity length
spacing facilitates obtaining single-frequency operation, tends to be somewhat longer. Therefore, the two configura-
where the goal is to have only one external-cavity mode fall tions are trading off cavity length (and hence mode spacing)
under the bandwidth of the diffraction grating. However, with grating bandwidth. In practice, the Littman configura-
the bandwidth of the grating is relatively large because the tion is usually chosen when narrower frequency operation
Littrow configuration only employs a single-pass geometry, is desired. The grazing incidence angle also enhances the
i.e. the light only interacts with the grating once. dispersion and further narrows the bandwidth. The other
For the Littrow type-II design (Figure 9b), light from advantage of the Littman configuration is that the tuning
the back facet of the laser diode is collimated and sent is achieved by moving the mirror rather than the grating,
so that the output beam, which is again the zeroth-order
to the grating at an angle of incidence qLittrow . Similar to
reflection, does not rotate when the wavelength is tuned. A
the Littrow type-I, the first-order diffracted light is again
disadvantage of the Littman geometry, other than its slightly
more complicated design and longer length, is that the out-
Laser
diode Littrow put power tends to be lower compared to Littrow geometry.
This power reduction arises because of the double-pass
geometry in order to obtain a given amount of optical
Lens feedback, the Littman grating efficiency must be the square
Diffraction
grating root of that required for the Littrow configuration, leaving
significantly lower power available as output. Moreover, in
(a) Output beam
order to obtain this high efficiency at near grazing incident
Diffraction
angle, the incident light polarization often must be made
grating Laser perpendicular to the grating grooves which necessitates a
diode
polarization rotator inside the cavity.
Output The spatial and spectral modes of semiconductor lasers
beam
are coupled together and this coupling can be quite compli-
Littrow Lens Lens
cated. For instance, if a gain-guided narrow-stripe laser is
(b) designed to be multitransverse mode, then it will definitely
not lase in a single longitudinal mode. However, index-
Tuning mirror guided lasers are easily designed for single transverse mode
Output beam
operation, but may or may not lase in a single longitudi-
Laser nal mode, or they will only do so in a very narrow range
diode
of operating parameters. A semiconductor laser is normally
considered to be predominantly homogeneously broadened
Lens as the output spectrum is considerably narrowed when the
(c) Diffraction grating transition from nonlasing to lasing is made. However, the
Figure 9. Three basic configurations of external-cavity diode situation is more complex for broad-stripe lasers because of
lasers: (a) Littrow type-I , (b) Littrow type-II, and (c) Littman. the coupling between spatial and spectral modes. The fact
that it is possible to reduce the bandwidth of a broad-stripe with an amplitude about 0.11% of the main laser line.
laser by a factor of more than 50 when operating it in an Consequently, additional filtering to remove the ASE is nor-
external cavity, while still retaining more than 50% of the mally required. Thus, a good bandpass filter with at least
initial power, indicates that homogeneous broadening is still 104 rejection at ASE will be useful. Finally, the basic
dominant. However, if an etalon is inserted into the HPECL characteristics of ECSLs are summarized in Table 11.
(high-power external cavity laser) in order to select a sin-
gle external cavity mode, there is a very large power loss.
Although one mode can be selected, the gain which was 4 COMPARISON OF CW LASERS FOR
being used by the other modes becomes wasted and is not RAMAN APPLICATIONS
used by the selected mode. This unused gain is character-
istic of inhomogeneous broadening. This section presents some selected Raman data obtained
Another parameter of interest for ECSL designs is the with excitation from a variety of different laser sources. The
amount of ASE. Figure 10 plots the frequency profile of goal is to illustrate some of the tradeoffs associated with the
a high-power ECSL operating at 785 nm on a logarith- various laser parameters, as well as to provide an indication
mic scale at three power levels ranging from just above of the signal levels which can be expected in CW Raman
threshold to full power. The laser exhibits an ASE pedestal measurements from different types of materials. All of the
data were recorded using a single grating spectrograph with
Amplified spontaneous emission f/1.4 camera lenses for collimation and focusing of the
105
dispersed Raman signal onto a CCD array detector. Most of
the data were collected using a Santa Barbara Instruments
Group (SBIG) 7i CCD camera which utilizes a Kodak KF-
400 CCD, but a few spectra were recorded with an Andor
Camera that included an EEV 30-11 CCD. Spectra from
104
Output intensity (arbitrary units)
High power
liquid samples were collected using a Visionex optical fiber
Medium power immersion probe which has seven 300 m collection fibers
Low power around a single 400 m excitation fiber. The collection
fibers are terminated in a slit shape at the entrance of
1000
the spectrograph. When necessary either a dielectric or
volume holographic filter was used within the spectrograph
to provide increased elastic light rejection.
100 4.1 Gas-phase Raman data
Because of their high power at visible wavelengths and

good beam quality, argon-ion lasers are good candidates for
excitation of gas-phase Raman spectra. Figure 11 displays
10 a Raman spectrum of ambient air recorded with a single
770 775 780 785 790 795 800 805 810
pass of laser light from an argon-ion laser operating at
Wavelength / nm 488 nm with a power level of 2 W. With an integration
Figure 10. Typical ASE spectrum of a high-power ECSL oper- time of 120 s, nearly 1.6 million counts of peak signal are
ating at 785 nm. obtained from the nitrogen (about 80% of mixture) at a
frequency shift of 2331 cm1 , and more than 0.5 million
Table 11. Typical ECSL characteristics.
counts of signal are obtained from the oxygen (20% of
Parameter Value mixture) at a frequency shift of 1555 cm1 . Under these
conditions, gas concentrations at the 0.1% level and below
Major laser lines 630860 nm can be accurately determined. Most gases are transparent at
Laser linewidth <30 GHz 488 nm and can have their Raman spectra recorded without
Output power up to 1 W significant interference from fluorescence, but there are
Transverse mode nearly diffraction-limited, or multimode
some important exceptions such as NO2 . In addition, as
Power stability <0.1% per hour
Lifetime 10 000 h the transparent gases have relative Raman scattering cross-
sections that vary by only about an order of magnitude, the
2 106 costs. Nevertheless, they represent one of the few lasers

488 nm, 2 W Nitrogen capable of providing high CW output powers in the blue-
SBIG 7i camera at 0 C (1 597 700
120 s integration counts)
green spectral range.
2 17 binning
1.5 106
4.2 Attributes of NIR excitation
Raman intensity
For many years NIR excitation of Raman spectra was not

1 106
considered a viable option because it resulted in large sig-
nal losses based on the !4 scattering law and because
Oxygen
(564 040 counts)
the available photomultiplier detectors had limited red
response. However, the availability of silicon CCDs with
5 105
high quantum efficiency beyond 900 nm and high-power
diode lasers operating in the NIR has greatly expanded
the types of materials from which good quality Raman
spectra can be obtained. Figure 12 shows Raman spec-
0
1500 2000 2500 tra obtained from reagent-grade toluene using an argon-
Frequency shift / cm1 ion laser (Coherent I-90) operating at 514 nm and three
Figure 11. Raman spectrum of ambient air using an argon-ion external-cavity-stabilized diode lasers (Process Instruments
laser with a power level of 2 W. ECSL series) operating at 670 nm, 785 nm, and 852 nm. All
minimum detection levels for all of the gases are similar.

Table 12 summarizes the relative Raman cross-section and 400
frequency shifts for some common atmospheric gases. 784 cm1 1000 cm1
To obtain gas detection limits in the ppm range, the good
beam quality of ion lasers can be exploited in a variety 350
of multipass cell configurations to enhance the effective 852 nm
excitation power. Signal enhancements by several orders of
magnitude are achievable by using multiple passes of light
300
Raman intensity (arbitrary units)
to excite the sample and by optimizing the collection optics 784 cm1
to collect Raman signals over as large a solid angle as pos- 1000 cm1
sible. Even larger enhancements can be achieved by using 785 nm
250
an argon laser in an intracavity configuration for excitation.
The primary drawbacks of using argon-ion lasers for gas-
monitoring applications include their high cost, large size,
200
need for water cooling, and relatively high maintenance
1000 cm1
Table 12. Relative Raman cross-section and frequency shifts for 784 cm1
environmental gases. 150 670 nm
Compound Raman shift (cm1 ) Raman cross-section

(relative to N2 )
100
514 nm
SO2 1151 5
NO2 1320 1 1000 cm1
SO3 1330 5 50 784
CO2 1388 1.4 cm1
CH4 1535 6
O2 1555 1.3
NO 1877 0.4 0 100 200 300 400 500 600 700
CO 2143 1
Pixel number
N2 2331 1
NH3 3334 5 Figure 12. Raman spectra of toluene using four different excita-
H2 4156 2.4 tion wavelengths of 514, 670, 785, and 852 nm (1200 gratings per
millimeter).
of the spectra were recorded using a Visionex optical fiber Raman features can be observed above the strong fluores-
immersion probe with no filters, a single-grating spectro- cence. With HeNe excitation at 633 nm, a few Raman
graph with 1200 gratings per millimeter, and an SBIG 7i features begin to appear above the background, but only
CCD detector. by exciting at 785 nm or 852 nm is it possible to obtain a
The spectra were normalized by setting the maximum good quality Raman spectrum from the gasoline. This gaso-
value of each spectrum to a value of 100, and the spectra line example is typical of how NIR excitation can extend
were offset from each other by 100 units. Two features of the capabilities of Raman spectroscopy from pure spectral-
the data are noteworthy. First, the spectral region imaged grade materials to real-world industrial applications.
onto the CCD decreases significantly with increasing excita-
tion wavelength. This decrease in spectral coverage results
in better resolution of the spectral features. In addition, all 4.3 Identifying the components of organic liquid
of the spectra are obtained with good S/N, but there is mixtures
increased fluorescence background with decreasing wave-
length of excitation. The capability of NIR excitation to Many potential applications of Raman spectroscopy for
suppress fluorescence is shown much more dramatically in monitoring an industrial process stream require determining
Figure 13, which shows Raman spectra from an unleaded the concentrations of components in a mixture. One such
premium gasoline obtained with excitation wavelengths application is measuring the meta, ortho, and para isomers
between 514 nm and 852 nm. The spectra were again nor- of xylene in a process stream. To provide an example of the
malized by setting the maximum value of each spectrum accuracy which might be expected using 785 nm excitation
to 100, and the spectra were offset from each other by 100. from a stabilized diode laser, an optical fiber immersion
With excitation at 514 nm from an argon-ion laser, no probe, and a single-grating spectrograph in conjunction with
CCD detection, we recorded Raman spectra from a series
of 20 samples containing varying amounts of the xylene
1000 cm1 isomers. The spectra of the pure xylene isomers are shown
500
in Figure 14. Although there is some overlap of the spectral
features, there is at least one distinct Raman peak for each
852 nm isomer. The mixture data were analyzed using a partial
least squares (PLS) chemometric model.43 Fifteen spectra
1000 cm1 were used to make the model, and then the model was
400
used to predict concentration values for the five spectra
Raman intensity (arbitrary units)
785 nm
1.4 105
300 1000 cm1 1.2 105

m-Xylene
p-Xylene
670 nm
1 105
Raman intensity
200
o-Xylene
1000 cm1 8 104
633 nm
6 104
100
4 104
514 nm
2 104
0 100 200 300 400 500 600 700

0 100 200 300 400 500
Pixel number
Pixel number
Figure 13. Raman spectra of unleaded premium gasoline using
excitation wavelengths from 514 nm to 852 nm. Figure 14. Raman spectra of pure m-, o-, and p-xylene.
70 3500
PI 100785 nm excitation Andor CCD (EEV 30-11)
PLS modelfour factors 120 s integration
60 15 sample training set 3000 785 nm excitation
Mean-centered data
5 sample validation set
Predicted p-xylene concentrations
2500 900 ppm

50
800 ppm
700 ppm
Raman intensity
2000 600 ppm
40 500 ppm
400 ppm
300 ppm
1500 200 ppm
30 100 ppm
SD = 0.055 % 0 ppm
1000
20
500
10
y = 0.055953 + 1.0018x ; R = 1 0
1000 1020 1040 1060 1080 1100
0 10 20 30 40 50 60 70 Frequency shift / cm1
Mixed p-xylene concentrations Figure 16. Raman data of nitrate in aqueous solutions at the
Figure 15. Predictions of the p-xylene concentrations using a 100 ppm levels using a high-sensitivity and low-noise CCD
PLS chemometric model. detector.
which were not included in the model. Figure 15 compares

Andor CCD (EEV 30-11)
the p-xylene concentrations predicted from the model for
2500 120 s integration
the five samples not used in forming the model with the 785 nm excitation
nominal concentration placed in the mixtures. A standard Visionex probe
deviation (SD) of 0.055% between predicted and nominal
values was achieved. The SD values for m- and o-xylene 2000
were 0.13% and 0.10%, respectively. These SD values are
Raman peak intensity
of the order of 0.1%, and are representative of the accuracy

which can be expected using Raman techniques for mixture
1500
analysis.
4.4 Applications which require high sensitivity 1000
The SBIG CCD detector used for collecting most of the data
presented as examples can only be cooled to approximately 500
10 C. Although this CCD is adequate for many Raman
y = 314.14 + 2.0923x ; R = 0.99404
applications, it does not have the low-noise performance of
more expensive CCDs cooled to much lower temperatures.
To illustrate the detection levels that can be achieved with 0 200 400 600 800 1000
Raman methods using 785 nm excitation and a better CCD, Nitrate concentration (ppm)
we replaced the SBIG camera on our single spectrograph Figure 17. Raman peak intensities versus nitrate concentrations
with an Andor camera with an EEV 30-11 CCD cooled from 0 to 1000 ppm showing good linearity.
to 65 C. Figure 16 shows Raman data from nitrate in
aqueous solutions at the 100 ppm level, and Figure 17 also completed a similar study to determine how small an
illustrates how the Raman peak intensity increases linearly amount of toluene could be detected in nearly pure benzene.
with nitrate concentration from 0 to 1000 ppm of nitrate. We Again, we were able to see differences at the 100 ppm level.
Thus, 100 ppm probably represents a reasonable estimate ABBREVIATIONS AND ACRONYMS
for the detection limit that can be expected using the
Raman techniques described. There will, of course, be many ASE Amplified Spontaneous Emission
examples where minor component concentrations will not CCD Charge Coupled Device
be detectable at this 100 ppm level because of spectral CW Continuous Wave
interferences with the major component or simply weak DBR Distributed Bragg Reflector
Raman cross-sections. DPSS Diode-Pumped Solid-state
ECSL External-cavity Semiconductor Laser
FT Fourier Transform
5 CONCLUSIONS FWHM Full-width at Half-maximum
HeNe HeliumNeon
Applications of Raman spectroscopy have increased signif- HPECL High-power External Cavity Laser
icantly because of new advances in laser technology, high- IR Infrared
sensitivity CCDs, holographic gratings, and optical fibers.
KTP Potassium Titanyl Phosphate
With these advances, Raman spectroscopy has moved out
Nd : YAG Neodymium-doped Yttrium Aluminum
of the laboratory into the real-world of chemical pro-
Garnet
cess analysis. There are now many CW laser lines ranging
NIR Near-infrared
from the UV to the NIR available commercially for use in
PLS Partial Least Squares
Raman experiments. In the UV region, frequency-doubled
SD Standard Deviation
versions of argon-ion, krypton-ion, Ti : sapphire, and dye
TEM Transverse Electromagnetic
lasers are most useful. These CW UV lines are especially
YAG Yttrium Aluminum Garnet
important for resonance Raman spectroscopy. In the visible
YLF Yttrium Lithium Fluoride
region, argon-ion, krypton-ion, dye, and frequency-doubled
YVO4 Yttrium Vanadate
Nd : YAG lasers are commonly used. Although fluorescence
is a major concern for visible excitation, these lasers typ-
ically have good beam quality and high power which are REFERENCES
very beneficial for gas-phase Raman spectroscopy. Near-
IR excitation for Raman spectroscopy has gained much
1. C.A. Murray and S.B. Dierker, J. Opt. Soc. Am. A, 3, 2151
attention since the late 1980s because of its success in flu- (1986).
orescence reduction. Using FT Raman spectroscopy with a
2. Y. Wang and R.L. McCreery, Anal. Chem., 61, 2647 (1989).
1.06 m Nd : YAG laser as the excitation source, researchers
have obtained good quality Raman spectra from a wide vari- 3. D.A. Gilmore and M.B. Denton, Spec. Publ. R. Chem. Soc.,
194, 145 (1996).
ety of materials without fluorescence interference. Newly
improved semiconductor laser technology in the NIR region 4. D.E. Battey, J.B. Slater, R.W. Wludyka, H. Owen, D.M. Pal-
lister and M.D. Morris, Appl. Spectrosc., 47, 1913 (1993).
makes it possible to reduce the size and cost of Raman
instrumentation and enables the manufacture of turnkey 5. J.M. Tedesco, H. Owen, D.M. Pallister and M.D. Morris,
Anal. Chem., 65, 441A449A (1993).
and compact Raman analyzers. These NIR semiconductor
lasers typically operate near 785 nm and allow Raman spec- 6. D.A. Gilmore, D. Gurka and M.B. Denton, Appl. Spectrosc.,
49, 508 (1995).
tra to be detected using sensitive silicon CCD arrays. Two
leading types of NIR semiconductor lasers are currently 7. J.M. Williamson, R.J. Bowling and R.M. McCreery, Appl.
Spectrosc., 43, 372 (1989).
employed in Raman instrumentation: DBR and external-
cavity lasers. Frequency-stabilized ECSL have excellent 8. Y. Wang and R.L. McCreery, Anal. Chem., 61, 2647 (1989).
frequency stability, intensity stability, and long lifetimes. 9. C.D. Allred and R.L. McCreery, Appl. Spectrosc., 44, 1229
These recent improvements in laser technology have made (1990).
Raman spectroscopy not only a qualitative tool but also a 10. R.L. McCreery, Proc. SPIE, 1637, 208 (1992).
quantitative tool. Although modern Raman technology has 11. I.R. Lewis and P.R. Griffiths, Appl. Spectrosc., 50, 12A
already attracted much interest in the process-monitoring (1996).
community, the acceptance of Raman process analyzers 12. M.L. Myrick, S.M. Angel and R. Desiderio, Appl. Opt., 29,
depends strongly on their cost. With lower cost instrumenta- 1333 (1990).
tion, more and more applications will be found. Thus, better 13. M.L. Myrick and S.M. Angel, Appl. Spectrosc., 44, 565
and lower-cost laser sources are still needed for Raman (1990).
spectroscopy to become a widely accepted analytical tech- 14. S.M. Angel, M.L. Myrick and T.M. Vess, Proc. SPIE, 1435,
nique in the chemical analysis instrumentation market. 72 (1991).
15. M. Majoube and M. Henry, Spectrochim. Acta, Part A, 47A, 29. R. Burnham and A.D. Hays, Opt. Lett., 14, 27 (1989).
1459 (1991). 30. J.F. Ready, Industrial Applications of Lasers, Academic
16. P.J. Hendra, Fourier Transform Raman Spectroscopy in Press, Chestnut, MA (1997).
Modern Technology in Raman Spectroscpy, ed. J.J. Laserna, 31. Y.F. Chen, T.M. Huang, C.L. Wang and L.J. Lee, Appl. Opt.,
John Wiley & Sons, Chichester, 73108 (1996). 37, 5727 (1998).
17. M.J. Weber, Handbook of Laser Wavelengths, CRC Press, 32. T. Sasaki, T. Kojima, A. Yokotani, O. Oguri and S. Nakai,
New York (1999). Opt. Lett., 16, 1665 (1991).
18. M.W. Pan, D.J. Evans, G.R. Gray, L.M. Smith, R.E. Benner, 33. S. Helmfrid and K. Tatsuno, J. Opt. Soc. Am. B, 11, 436
C.W. Johnson and D.D. Knowlton, J. Opt. Soc. Am. B, 15, (1994).
2531 (1998).
34. T.F. Johnston, Tunable Dye Lasers, in Encyclopedia of
19. R. Claps, J. Sabbaghzadeh and M. Fink, Appl. Spectrosc., 53, Lasers and Optical Technology, ed. R.A. Meyers, Academic
491 (1999). Press, San Diego, 663710 (1991).
20. A.E. Siegman, Lasers, University Science Books, Mill Val- 35. S.M. Angel, M. Carrabba and T.F. Cooney, Spectrochim.
ley, CA (1986). Acta, Part A, 51, 1779 (1995).
21. O. Svelto, Principles of Lasers, Plenum Press, New York 36. J.B. Cooper, P.E. Flecher, S. Albin, T.M. Vess and W.T.
(1989). Welch, Appl. Spectrosc., 49, 1692 (1995).
22. P.W. Milonni and J.H. Eberly, Lasers, John Wiley & Sons, 37. T.F. Cooney, H.T. Skinner and S.M. Angel, Appl. Spetcrosc.,
New York (1988). 49, 1846 (1995).
23. J. Hecht, The Laser Guidebook, McGraw-Hill, New York 38. Y. Li, C.K. Wu, M.B. Snipes and J.G. McInerney, Proc.
(1986). SPIE, 1634, 532 (1992).
24. J. Verdeyen, Laser Electronics, Prentice Hall, Englewood 39. P. Gavrilovic, A.V. Chelnokov, M.S. ONeill and D.M.
Cliffs, NJ (1995). Beyea, Appl. Phys. Lett., 60, 2977 (1992).
25. A. Katzir, Lasers and Optical Fibers in Medicine, Academic 40. T. Day, M. Brownell and I.F. Wu, Proc. SPIE, 2378, 35
Press, San Diego, CA (1993). (1995).
26. B.E.A. Saleh and M.C. Teich, Fundamentals of Photonics, 41. P. Zorabedian, Tunable Lasers Handbook, Academic Press,
John Wiley & Sons, New York (1991). New York (1995).
27. W.T. Silfvast, Laser Fundamentals, Cambridge University 42. M.W. Pan, G.R. Gray, L.M. Smith, R.E. Benner, C.W. John-
Press, New York (1996). son and D.D. Knowlton, Appl. Opt., 37, 5755 (1998).
28. W. Koechner, Solid-State Laser Engineering, Springer- 43. D.M. Haaland and E.V. Thomas, Anal. Chem., 60, 1193
Verlag, Heidelberg, Germany (1996). (1988).
Array Detectors for Raman Spectroscopy
Carolyn J.S. Pommier, Linda K. Walton, Trent D. Ridder and
M. Bonner Denton
University of Arizona, Tucson, AZ, USA
1 INTRODUCTION laser and background flicker noise, sample and back-

ground shot noise, detector dark current, and readout noise.
Although Raman spectroscopy has been used for many Because the excitation sources used for modern instruments
years, only recently has the technique become a relatively are more stable than hitherto, laser and background flicker
common tool for chemical analysis. Despite its early discov- noise are insignificant and are ignored. Sample shot noise
ery, limitations in practical sources, high throughput optics, and background shot noise, although not directly related
and sensitive multichannel (MC) detectors have restricted to the detector, are affected by the multiplex advantage of
the practical application of Raman spectroscopy. Several array detectors. Consequently, a brief description of those
recent advances now allow fast, routine analysis of Raman noise sources is presented; the remainder of this section
scattering to be performed. These include stable, relatively deals with detector dark current and read noise.
inexpensive lasers, new optical components such as volume-
phase holographics, and a variety of new high-performance
array detectors. This article discusses the role of focal plane 2.1 Sample shot noise
detectors and their impact on Raman spectroscopy.
High noise and poor sensitivity limited early detectors Sample shot noise is composed of fluctuating analytical sig-
used for Raman spectroscopy. In low-light-level applica- nals caused by photons from the sample arriving randomly
tions, every source of noise is critical and must be con- at the detector. Sample shot noise is limiting when the
sidered and minimized. The recent development of very detector itself contributes little significant noise and obtains
high performance array detectors with extremely low noise its strongest light source from analyte light scattering. The
characteristics and high quantum efficiency (QE) repre- number of photons (S ) reaching the detector from the
sents a major advance in Raman spectroscopy. In addition, analyte is given by equation (1),
the inherent multiplex advantage of array detectors allows
every wavelength to be sampled simultaneously, resulting S D LAT 1
in short integration times and high sample throughput. To
begin the article, an overview is given of noise characteris- where L is the scattered light intensity per unit of scattering
tics of past and present detectors for Raman spectroscopy. area, A is the area of scattered light collected, is the solid
angle collected, and T is the transmission of the optical path
of the spectrometer. The electronic output signal S from the
2 NOISE detector is given by equation (2),
S D S Qt 2
Detectors suitable for Raman spectroscopy must have low
noise characteristics. There are six main sources of noise: where Q is the QE of the detector and t is the exposure
time of the detector to the photon flux. QE is a measure
John Wiley & Sons Ltd, 2002. of device sensitivity, characterized by the percentage of
incident photons on the device that result in the creation of shot noise limited. Increasing the integration time or the QE
measurable electrons. of the detector increases the S/N by t0.5 or Q0.5 , respectively.
In a single-channel (SC) detector, sample shot noise is In MC detectors, t is replaced with T from equation (4) and
proportional to the square root of the signal, S0.5 , equa- the S/N increases by n0.5 for the same sampling time.
tion (3): In Raman spectroscopy, a fluorescent sample or a flu-
shot noiseSC / S Qt 3 orescent impurity in the sample may emit such a large
background signal that background shot noise limits the
Under sample shot-noise-limiting conditions, the signal
S/N. The S/N decreases as the background signal increases,
to noise ratio (S/N) increases with the square root of time.
making it advantageous to minimize fluorescence. In the
For a SC detector, t is the time spent integrating each
background shot-noise-limited case, a small Raman signal
wavelength. To collect an entire spectrum of n resolution
will be superimposed on a large fluorescence signal. Here,
elements, the entire time T required to collect the spectrum
B is large and the noise contribution from the sample is
is given by equation (4),
negligible. In some cases, the fluorescence shot noise is so
T D nt 4 large that the Raman spectrum is completely obscured by
fluorescent shot noise.
An array detector enables detection of all resolution
elements at once. Thus, t in equation (3) can be substituted
for T in MC detectors. The S/N then becomes equation (5),
2.3 Detector noise
S
(S/N)MC /p 5
Qnt In many applications, shot noise from fluorescence is the
The ratio of equation (5) to equation (3) shows that the limiting factor in the S/N. In certain Raman applications,
S/N of a MC detector is higher than that of a single channel such as Raman spectroscopy on monolayers2 and UV
detector by a factor of n0.5 if time, QE, and photon flux (ultraviolet) Raman, however, detector dark current and
are equivalent. Thus, using the total sampling time, a MC read noise inhibit the efficiency of the instrument.
device will yield a greater S/N than a SC device. This
increase in S/N is referred to as the MC advantage. This 2.3.1 Dark current
advantage has been demonstrated.1
Dark current, defined as any signal generated by a detec-
tor not exposed to photons, is constant over time for a
2.2 Background shot noise given temperature. Because its major source is the ther-
mal production of electrons, many detectors are cooled to
In Raman spectroscopy, any light not caused by inelastic decrease the amount of dark current. Electrons from dark
scattering that reaches the detector is considered back- current can fill an integrating detector to capacity before a
ground noise. Background shot noise has many different significant amount of analytical electrons are collected, lim-
sources, with fluorescence, solvent scattering, and stray iting the length of time available to collect analytical signal.
light being the most common in Raman spectroscopy. Some Raman experiments deal with such low light levels
Because it is another kind of shot noise, background shot that integration times of an hour or more can be necessary.
noise can be treated in the same manner as sample shot Consequently, one of the most important parameters of an
noise. Total background signal B can be considered pro- array detector is its dark current.
portional to the product of the amount of photons from
the background reaching the detector B , the QE of the 2.3.2 Read noise
detector, and the integration time t. Background shot noise
and sample shot noise are additive. System noise, including Read noise is inherent in the electronics that carry the sig-
both background and sample shot noise, can be expressed nal from the detector to the data processor. Some sources
mathematically, as follows: of read noise include signal amplification, analog-to-digital
conversion, signal processing, and detector readout cir-
N / B C S 6 cuitry. Read noise is not important when photomultiplier
Dividing signal by noise to obtain the S/N yields tubes (PMTs) are used but is significant in charge-transfer
devices and photodiode arrays (PDAs). Read noise is inde-
S S S Qt pendent of the signal amplitude and integration time, such
/p Dp 7
N B C S S Qt C B Qt that it is not a problem when high signal levels are present
If the signal is large compared to the background, equa- but may limit the S/N when signal intensities are low. Aver-
tion (7) reduces to equation (3), and the Raman signal is aging multiple reads can minimize the contribution of read
Array Detectors for Raman Spectroscopy 3
noise to the total detector noise. The magnitude of the composition of the sample. The primary disadvantage of
decrease in read noise is equal to the square root of the PMTs is that to achieve the same S/N as MC detectors, an
number of reads. increase in sampling time is necessary.
Another disadvantage of using SC detectors is that
because each wavelength is collected sequentially rather
2.3.3 Pixel-to-pixel response
than simultaneously, it is impossible to obtain a spectrum
Array detectors contain noise not present in other MC that represents a single moment in time for a dynamic
devices. Deviations in pixel response, varying as much system. An array of one-dimensional detectors can be con-
as 10% from pixel to pixel, cause artificially high or low structed by aligning PMTs to coincide with the spatial
measurements. This type of noise is difficult to eliminate distribution of wavelengths. The resolution of these detec-
but can be compensated for by flat field correction, which tion systems is poor, however, and in practice alignment
requires the collection of a dark spectrum and a spectrum and calibration are difficult. Array detectors do not suffer
of a uniform white light source. A normalized response from these disadvantages.
algorithm is created and used for the correction of all
spectra thereafter. The algorithm A is created as shown in
equation (8): 4 MULTICHANNEL DETECTORS
W n Dn
AD 8
W Davg MC detectors are transducers with multiple resolution
where Wn is the white light response at the nth pixel and Dn elements, allowing spatial resolution in one or two dimen-
is the dark response at the same pixel. The denominator is sions. The following section discusses common MC detec-
the average intensity of the dark-subtracted signal. Pixel-to- tors, including photographic emulsions, PDAs, and charge-
pixel variations can be corrected by dividing the spectra by transfer devices.
the normalized response algorithm. This correction yields
a difficulty because there is no truly flat white-light source,
so that the light source introduces a response curve into the 4.1 Photographic emulsions
algorithm. In Raman spectroscopy, white light generated by
a blackbody has a smooth response curve over the relatively The photographic emulsion is worthy of a brief mention for
narrow spectral range of the Raman scattering, so this error historical reasons. A photographic plate is used to image an
is usually ignored. entire spectrum at once by placing undeveloped film at the
In addition to the traditional noise sources mentioned focal plane of a spectrograph while the exposure time is
above, array detectors suffer from a noise source referred controlled with a shutter. This technique is simple, inex-
to as cosmic ray events (CREs). These events are discussed pensive, and sensitive to light levels as low as 10100
in Section 4.3. photons.3,4 The photographic emulsion was replaced by
Noise sources critical in Raman spectroscopy can be electronic transducers because of the time required to
dark current and read noise. The remainder of this article develop film, the difficulty of processing for quantitative
considers each type of detector and the parameters affecting applications, and the relatively narrow dynamic range.
the overall S/N. First, a brief discussion of SC detectors is
necessary in order to demonstrate the advantages of array
detectors. 4.2 Resistive anode photomultiplier tubes
Resistive anode (RA) devices are the equivalent of MC

3 SINGLE-CHANNEL DETECTORS PMTs. Although not an array detector, the RA is a position-
sensitive detector. An RA is made of a single sheet of
Before the invention of array detectors, the PMT was the a resistive anodic material with electrodes attached to the
most common detector for Raman. However, SC detectors sides (to collect a spectrum) or to the corners to (acquire an
have several disadvantages, and their use is being phased image). The centroid position of a photon signal is deter-
out in favor of MC detectors. A SC detector only detects mined by taking the ratio of the current at the electrodes.
a single resolution element at a time, so a monochromator The precision with which the position of the electron event
must be scanned through the desired wavelength region to can be measured and the number of wavelengths that fall
obtain a spectrum. Spectra collected by scanning monochro- across the detector determine the resolution of this device.
mators are susceptible to changes in signal intensity, caused RAs must be coupled to a microchannel plate (MCP) in
by source or sample fluctuations unrelated to chemical order to function as low-light-level detectors. Please refer to
Section 4.4 for a discussion of MCPs. An extra electrode is per second.7 High levels of dark current in PDAs limit the
often included in RA devices so that noise from thermionic maximum possible integration time because the dark current
emission from the MCP can be neglected.5 The electrode rapidly causes complete pixel saturation. In one experiment,
only records signals above a threshold value. The minimal the pixels of a PDA were saturated after 1 h of integration,
threshold value corresponds to a single photon event at the whereas a charge coupled device (CCD) and an RA PMT
photocathode and its signal amplification in the microchan- had little problem from dark current after 3 h.5 PDAs have
nels. Thermionic emission originating in the microchannels significant read noise, typically from 1700 to over 3000
can therefore be eliminated, although thermionic emission electrons.6
from the photocathode remains a cause of dark current.
Although no read noise is associated with RA detectors, Intensified photodiode arrays. In Raman spectroscopy,
their use for Raman spectroscopy has been limited because the signal is often too low to be easily detected with a
other MC detectors have a higher QE, better resolution, and PDA because of high dark current and read noise. To
a greater linear dynamic range. compensate, a MCP is coupled to a PDA, in the same
manner as described for RAs, to increase the number of
photons reaching the detector. To maximize performance,
4.3 Array detectors the MCP is chosen such that the photons emitted from
the phosphor are near the wavelength of highest silicon
To permit thorough discussion, it is necessary to define response. MCPs may also be employed with PDAs for gated
terms common to all array detectors. An array detector is detection.
composed of an array of pixels where each pixel is one
resolution element of the array. The maximum amount of Summary. PDAs have high read noise and high dark cur-
charge a pixel can hold is defined as the full well capacity rent compared to that of other array detectors for Raman
(FWC). Array detectors are integrative devices, collecting spectroscopy. Although cooling the detector can minimize
electrons for as long as the detector is exposed to light. dark current, detectors such as CCDs have better per-
After exposure, the device is sampled to read the signal. formance characteristics. Because of these disadvantages,
PDA detectors are becoming less common for Raman spec-
troscopy.
4.3.1 Photodiode arrays
A linear PDA consists of a single row of photodiodes on
4.3.2 Charge coupled devices
a silicon support. Modern spectroscopic arrays typically
have 1024 diodes spaced at 25 m intervals; the photoactive CCDs became commercially available in 1973 and now
area of each diode is 13 m wide by 2.5 mm high. The largely dominate the detector field in Raman spectroscopy.
PDA is positioned so the spectrum falls across it, causing A CCD consists of a two-dimensional array of pixels on
each diode to collect a different wavelength region. Photons a silicon chip.8 Each pixel is composed of metal-oxide
incident upon a photodiode are converted to electronhole semiconductor (MOS) electrodes, called gates, on top of
pairs. These electrons and holes discharge a finite amount the chip (Figure 1). A potential well is produced under
of charge from a dedicated capacitor. The corresponding these gates and stores electrons created by incident photons.
signal is determined by measuring the amount of charge The array is read by shifting the electrons from pixel
necessary to recharge the capacitor. to pixel across the chip into a register consisting of an
integrating capacitor and amplifier (Figure 2). A series of
Quantum efficiency. The QE of PDAs is high compared analog amplifiers and an analog-to-digital converter (ADC)
to that of RAs and PMTs. The most critical factor affecting convert the analog signal from the chip into a computer-
the QE of a PDA is the photoresponse of silicon. The QE readable format.
of silicon peaks near 70% around 550 nm, and is above
20% from 200 nm to 900 nm.6 From 900 to 1100 nm, the Array formats. The CCD is available in a variety of for-
photoresponse drops from 18% to about 1%. This response mats, allowing selection of the most appropriate array shape
makes PDAs better suited as detectors in the visible red and size for a particular Raman spectroscopic or imaging
and near-infrared (NIR) regions than RA PMTs. application. The number of necessary pixels is determined
by the spatial resolution required by the application. The
Dark current and read noise. Dark current in PDAs is largest available array may not be the best choice, because
primarily due to pixel capacitor leakage. Typical dark optical components in a system can limit resolution. Conse-
current for a PDA at room temperature is 40 electrons quently, increasing the number of pixels may not increase
1
2
3
Gate electrodes
SiN2
Insulators 20 nm
SiO2
20 m epitaxy
Potential
wells +
+ + + +
+ + 500 m substrate
Figure 1. Cross-section of a CCD, demonstrating the location of electrodes and the formation of potential wells. The locations of the
electrons and holes in the diagram show where charge is collected.
Serial register Raman spectroscopy is extremely important, because highly

Serial sensitive detectors are necessary to maximize the S/N. As
Parallel shift for seen with other detectors such as the PDA, the QE of CCDs
shift readout varies with wavelength. The reflectivity of silicon, ranging
4 3 2 1 4 3 2 1 from 30% at 1090 nm to 50% at 400 nm, can limit the QE of
R R To amplifier CCDs.9 However, the QE can be improved by application
(a) (b) (c) of an antireflective (AR) coating.
Figure 2. Demonstration of charge readout on a CCD. (a) Charge Applying an AR coating to frontside-illuminated CCDs
as collected on the CCD immediately following integration. is not easily done because of the varying nature of the CCD
(b) Readout begins charge from columns in (a) are shifted one gate structure. The photosensitive Si comprising the epitaxy
column towards the readout such that the charge from column 1
in (a) is moved to the readout in (b). (c) Charge in the readout is partially covered by a silicon dioxidesilicon nitride
is measured and the process is repeated until all columns have insulating layer and gate electrodes (see Figure 1). The
been read. varying, overlapping polycrystalline silicon gate electrodes
have a different refractive index from that of the silicon
the system resolution. In such cases, a CCD with fewer epitaxy, making conventional AR coatings on the frontside
pixels is desirable, because a smaller array with the same of the device ineffective.
pixel read rate and noise characteristics will produce faster The backside of a CCD is uniform and not covered
image reads and will be substantially less expensive than a by gate structure, allowing backside AR coating and
larger array. illumination to be highly efficient. However, the thick
Rectangular arrays are frequently used in Raman spec- substrate below the epitaxy must be etched away
troscopy, with formats matched to the usable focal plane through mechanical and chemical processes prior to
of the spectrograph. Array sizes can range from 512 64 AR coating. Backside-thinned CCDs have much higher
pixels, where each pixel is 24 m square, to 2000 800 QE than frontside-illuminated devices (Figure 3), but the
pixels, where each pixel is 15 m square. When a rect- etching process renders them significantly more costly
angular CCD is placed at the flat focal plane of the and fragile. This expense is often justified for low-
spectrometer, wavelengths are dispersed horizontally along light-level experiments requiring high sensitivity, such
the long axis of the CCD to obtain the widest spec- as Raman spectroscopy. Optimized backside-illuminated,
tral coverage with the best spatial resolution. All pixels antireflection-coated CCDs can yield QEs approaching 98%
in a particular vertical column record the same wave- (see Figure 3).
length. In Raman spectroscopy, the second dimension of
spatial resolution is usually not required, so the vertical Open electrode structure Technology employing an
columns are summed (see Binning) to increase the S/N. If open electrode structure can increase the QE for frontside-
a square device is used for spectroscopic purposes, often illuminated devices. In this format, the electrode on top
only data from a rectangular section of the device will be of the device is open, leaving a large amount of the
employed. frontside of the pixel uncovered and suitable for AR
coating (Figure 4), so more photons are incident on the
Quantum efficiency. CCDs are currently widely used in epitaxy. Open-electrode devices are easier to make and
Raman spectroscopy because they can offer very high less fragile than backside-thinned devices. The shifting
QE from the UV to the NIR. The QE of CCDs used in electrodes reduced size, however, reduces the imposed
100 200
90 180
90% absorption depth (m)

80 160
70 140
60 120
QE (%)
50 100
40 80
30 60
20 40
10 20
0 0
200 300 400 500 600 700 800 900 1000 1100 350 450 550 650 750 850 950
Wavelength (nm) Wavelength (nm)
Figure 3. The QE varies widely for different CCD structures and Figure 5. Graph of photon penetration depth versus wavelength
devices. Graph of QE versus wavelength for some types of CCDs: in silicon, demonstrating that UV photons are rapidly absorbed,
() normal frontside illuminated; () frontside illuminated with whereas NIR photons have deep penetration.
organic phosphor to convert UV photons into the visible region;
) frontside illuminated in deep depletion mode; () backside
(C The temperature at which the CCD is operated has
illuminated with antireflection coating optimized in the visible; an effect on QE in the NIR region. Scientific devices
() backside illuminated with antireflection coating optimized for
are operated at temperatures where approximately 1.14 eV
the UV (special thanks to Mike Lesser); () backside illuminated
and UV optimized with AR coating. is required to generate an electronhole pair. Therefore,
only photons with wavelengths shorter than 1090 nm have
electric field, lowering the full-well capacity by a factor enough energy to be detected. The energy required to create
of two. electronhole pairs is affected by the vibrational energy
of the silicon lattice. At lower temperatures, this lattice
Photon penetration depth QE is also limited by pho- has less vibrational energy, and therefore slightly higher-
ton penetration depth, defined as the depth at which 90% energy photons are required to generate electronhole pairs.
of incident photons create electronhole pairs. Wave- Consequently, the QE in the NIR is higher when the CCD
length dependence exists for photon penetration depth. is warmer; however, there is a corresponding increase in
Shallow penetration is observed in the UV region, and dark current.
deeper penetration occurs as wavelength increases. Pen- Photon penetration depth is also a problem in the UV
etration depths range from 2 nm at 275 nm9 to 30 m region. On front-illuminated CCDs, the incident photon is
at 800 nm and up to several hundred micrometers at often absorbed in the gate structure before it can reach
1000 nm (Figure 5).8 The QE in the NIR region is reduced the epitaxy to create an electronhole pair. On backside-
because the deep penetration of NIR photons causes most illuminated devices, the electron is created at such a shallow
to pass through the typically 1020 m epitaxy with- depth that it cannot be trapped by a well unless the epitaxy
out creating an electronhole pair (Figure 6).8 Sensitivity is extremely thin. Therefore, QE is a few percent at wave-
can be increased in the NIR region by using a modi- lengths less than 300 nm for frontside-illuminated devices.
fied CCD with a thicker epitaxy (Figure 7) and optimized Backside-illuminated devices can achieve a QE in excess
doping. of 40% in the same wavelength region.
Phase 1 Phase 2 Phase 3 Phase 1
Oxide Oxide
Nitride Open phase Nitride
8 m 10 m
Figure 4. By eliminating the electrodes over phase 3 and increasing its size, it is possible to realize part of the QE gains of a backside
device without the added expense of thinning. Unfortunately, the smaller phase 1 and phase 2 structures reduce FWC.
Frontside Backside
900 nm photon
500 nm photon +
Substrate
e
Trapping
200 nm photon
Well
Gate structure 1020 m epitaxy
(a)
Frontside Backside
Antireflection
900 nm photon coating 12 nm
+ 500 nm photon
e 200 nm photon
Trapping
Well
1020 m epitaxy
(b)
Figure 6. In typical 1020-m epitaxy devices, NIR photons often pass through the active epitaxy region. (a) In frontside devices, UV
photons are absorbed by the electrodes. The complex nature of the gates prevents effective AR coating of the frontside. (b) Backside
illumination allows effective AR coating and effectively traps charge from UV photons.
Backside,
Frontside antireflection
coated
900 nm photon + 900 nm photon
500 nm photon + e
e
Trapping
Well
50 m epitaxy
Gate structure
Figure 7. Deep depletion thick epitaxy can be configured in either frontside or backside illumination for improved NIR performance.
Backside configuration allows for effective AR coating, further enhancing NIR performance.
The most common means of improving the QE in the in loss of half the emitted light, frontside CCDs with
UV for frontside-illuminated devices is to thinly coat the phosphors are versatile for wavelengths from the vacuum
CCD with an organic phosphor (Figure 3). The phosphor is UV to the NIR regions.
selected so that UV photons are absorbed and fluorescence In order to be detected, an electron must be trapped
emission occurs at approximately 500 nm, where the QE for in a well created by gate electrodes. In many CCDs the
CCDs is high. Although fluorescence is isotropic, resulting depletion depth, i.e. the depth to which the well can trap
electrons in the epitaxy, is often only 25 m from the will alleviate the problem of flat peak tops at the cost of
gate electrode.9 Photons absorbed in the epitaxy below the resolution for low-intensity signals.
depletion depth create electronhole pairs that recombine Blooming is a phenomenon that occurs when a pixel is
before the electron can be trapped and detected. Further- exposed to too many photons, resulting in the creation of
more, if an electron is not quickly trapped in a well it may more electrons than the pixel can contain. When FWC
diffuse into an adjacent pixels well, causing decreased resis exceeded, excess electrons in the pixel spill over, or
olution. To minimize this problem, the depletion depth may bloom, into neighboring pixels along the transfer register.
be increased, as seen in deep depletion CCDs. Blooming is inherent in CCDs, and the extent to which it
occurs is contingent on the CCD design. CCDs with larger
Window material Window material indirectly affects FWCs are less likely to bloom, as their FWC is not as
the QE of a CCD. In almost every instrument, the CCD easily exceeded. Blooming causes serious problems when
chip is located behind a window through which photons high-intensity features occur next to low-intensity ones. The
must pass before reaching the CCD. The transmission and exposure time required to detect the low-intensity feature
reflection properties of the window material vary with may be sufficient to cause the pixels corresponding to the
wavelength. AR coatings are often applied to the window high-intensity feature to bloom, completely obscuring the
to increase throughput. The wavelength of the excitation low-intensity feature. To prevent blooming, some CCDs
source, and therefore the wavelength range of the Raman have antiblooming drains that bleed off excess electrons.
spectrum, should dictate the choice of window material. However, such a drain reduces the amount of photosensitive
area in a pixel, which reduces the effective QE of the
Dynamic range and resolution. Once a CCD has been
device.
read, the analog signal from the CCD chip must be con-
In Raman spectroscopy, high- and low-intensity signals
verted to a digital signal by an ADC. The conversion of
are common in the same spectrum, so using an ADC with
electrons within a CCD to a digital output signal is an
the highest available grayscale resolution will allow res-
important parameter in determining CCD dynamic range
olution of low intensities with sufficient range at high
and grayscale resolution. Grayscale resolution, in this sense,
intensities to eliminate flat-topped peaks. For Raman spec-
refers to the ability to accurately measure the number of
troscopy, therefore, a 16-bit ADC performs better than a
electrons within a pixel. The number of bits available in
12-bit ADC. Having more ADC resolution than necessary
an ADC determines its resolution. For scientific applica-
is not desirable, however, because higher resolution ADCs
tions, ADCs typically range from 12 to 18 bits.9 A 12-bit
are more expensive and require more time to convert the
ADC can resolve 212 (4096) individual electron packets,
signal from analog to digital.
where each packet is referred to as an analog-to-digital
unit (ADU). Similarly, an 18-bit ADC can resolve 218 , or
262 144 ADU. Noise considerations. The CCD QE is one determining
The FWC of a pixel, determined by the electric field factor for the magnitude of signal observed from incident
imposed on the gate structure of the pixel, typically ranges light of a given intensity. However, true measurement qual-
from 10 000 to 500 000 electrons. When the FWC is greater ity is determined by the S/N. In many areas of spectroscopy,
than the number of ADUs in the electronics package, each CCD noise is so low that other instrumental or experimental
ADU must represent more than one electron. The gain parameters limit the S/N. However, in Raman spectroscopy,
of the system is defined as the number of electrons per dark current and read noise can be of critical importance.
ADU. A device set at a higher gain has fewer electrons per
ADU, providing better resolution for low-intensity signals. Dark current In a CCD, the dark current is charge
However, a gain equal to fewer electrons than the read noise derived from sources other than photons that collects in
does not enhance resolution at low light levels, but serves potential wells. The primary source of dark current is
to reduce the dynamic range. Consequently, the readout temperature-dependent thermionic emission occurring at
gain is often set equal to the number of electrons of read surface defect sites. The amount of dark current is depen-
noise per pixel. For example, if the CCD had five electrons dent on the temperature of the CCD chip. Dark current
of read noise, each ADU would represent five electrons. decreases by approximately 50% for every 6 C decrease in
High gain limits the dynamic range because the maximum temperature. Scientific CCDs are operated well below room
number of electrons represented by the ADC may be less temperature in order to minimize dark current.
than the number of electrons in the well of a pixel. Thus, all CCDs are cooled mechanically, using a thermoelectric
pixels containing electrons in excess of the ADC dynamic or refrigerant cooler, or cryogenically, using liquid nitro-
range will have the same digital value, resulting in high- gen. Thermoelectric coolers lower the temperature of the
intensity signals exhibiting flat tops. Decreasing the gain chip to 30 C to 60 C, where the dark current is a few
electrons per pixel per second. In this temperature range, in an abnormally high photon count for a single pixel
however, dark current may still limit spectroscopic applica- within an image. Because long exposure times are com-
tions requiring long integration times, so further reduction monly employed in Raman spectroscopy, many CREs may
is necessary. In these cases, the CCD is cooled to 100 C be detected in an acquisition. However, these events can
or below, where dark current is often less than 1 electron generally be removed by software correction.
per pixel per second. The material used to construct a CCD-based detector
Charge-transfer efficiency, a measure of the percentage of system may contribute to the number of detectable CREs.
charge transferred from one pixel to another during readout, Stainless steel is often used in CCD camera construction
is degraded if the device is cooled too much. Thermal in both dewars and cooling devices. A thoriated tungsten
energy is needed to move the charge from one potential electrode is commonly used to weld stainless steel, resulting
well into a neighboring, deeper potential well. Without in a weld that incorporates a small amount of thorium. If
sufficient energy, charge may be left behind or trapped the weld is in the line of sight of the CCD, a-particles
in matrix imperfections. Low charge-transfer efficiencies emitted by the thorium will cause the CCD to detect many
render a device useless, limiting the temperature to which large-magnitude CREs. Another possible source of CREs
a device may be cooled to decrease noise. in CCD-based detectors is the window located in front
To further reduce dark current, CCDs operating in a of the CCD. Thorium oxide is often incorporated into
unique mode called multipinned phase (MPP) may be optical windows to provide a high refractive index and
employed. These devices contain an n-type buried channel low dispersion.11 In order to minimize CREs in silicon
within the p-type epitaxy. Slight p doping of the buried detectors, care must be taken to ensure that window and
channel beneath one gate of a pixel raises the potential dewar materials are not radioactive.
energy under the gate. Raising the potential energy under
one gate of a pixel creates a barrier that allows all gates to Etalon interference patterns A final type of noise,
be negatively biased during integration while maintaining observed with backside-illuminated CCDs, is an etalon
separate potential energy wells for each pixel. The negative interference pattern. This phenomenon occurs when inci-
bias creates a charge inversion, trapping holes underneath dent light is reflected between the surfaces of the epitaxy,
the gates, which nullifies dark current electrons created at creating interference patterns that appear as waves in the
surface defect sites. Dark current is reduced by a factor of image. The spacing between the waves depends on epi-
1030,10 but the FWC can be reduced by as much as half. taxy thickness. A white-light source can be used to map
the etalon interference pattern, and the map can be used
Read noise As with other integrative detectors, read
to implement a software correction. AR coatings optimized
noise in CCDs is produced by readout electronics. At 220
for the wavelength range of interest can also be used to
electrons per pixel, the read noise of CCDs is much lower
minimize or eliminate this effect.
than that of PDAs. One way that scientific devices reduce
read noise is by correlated double sampling, accomplished
by measuring the potential of the output both before and Specialized readout modes. Specialized readout modes
after a charge packet has been shifted onto the control node are employed to increase readout speed and gather spe-
of the output preamplifier. Taking the difference of the cific information. Subarray readout, charge binning, time-
two measurements provides a value due only to the signal. delayed integration (TDI), and spectral framing are exam-
However, correlated double sampling increases readout ples of specialized readout modes.
time. Many CCD systems have several available readout
rates, allowing the user to choose the best compromise Charge binning The most common specialized read-
between read noise levels and readout speed. In practice, out mode is charge binning. Binning refers to the process
the magnitude of CCD read noise is often proportional to of combining charge from adjacent pixels, which can be
the readout rate: faster readout rates produce more read accomplished in either a hardware or a software bin. To
noise. bin a column in the hardware, charge is transferred from
every pixel in the column to the pixel in the serial regis-
Cosmic ray events (CREs) These events, named by ter (Figure 8), effectively transforming the column into one
astronomers using CCDs, give rise to a type of noise com- long pixel. This type of charge binning increases read speed
mon to all array detectors. A CRE results from random and the S/N. Read noise is reduced because fewer reads are
background radiation caused by radioactive decay of ele- required to record the total charge on the CCD, and read
ments and random cosmic events. High-energy background noise is directly proportional to reduction in the number
radiation incident on the photosensitive area of the detec- of reads. For example, if each read resulted in five elec-
tor creates a large number of electronhole pairs, resulting trons of noise and 250 pixels were binned, only five total
read by advancing charge in the serial and parallel registers,

sampling and digitizing only pixels desired and discarding
Parallel register
information from regions not of interest.14 Following read-

out, the entire array is clear and ready for the next exposure.
The primary advantage of this technique is speed; however,
information is lost in regions where the charge is not read.
Subarray readout is frequently employed when a rectangu-
lar portion of a square device is used for a spectroscopic
application.
Time-delayed integration TDI is used to improve sen-

Parallel shift sitivity of an image of a moving sample. In TDI operation,
the CCD columns shift at the same velocity as the sample
in order to track the sample as it moves. TDI mimics the
collection of photons from one position. As a result, the
integration time for pixels tracking the sample motion can
be extended to provide increased sensitivity. Ultimately,
TDI has the effect of making a moving sample appear sta-
tionary. The increase in sensitivity is proportional to the
increase in integration time per pixel, or the number of
pixels along the axis of the samples movement. The dif-
ficulty with TDI read mode is in synchronizing the shift
of CCD registers with the motion of the sample. Sample
movement must be known precisely in order to program
Serial register appropriate CCD shifts for the experiment.
Figure 8. Demonstration of the hardware binning process in TDI has been applied to Raman spectroscopy in a method
spectroscopic CCDs. known as continuous extended scanning.12 In this tech-
nique, the grating in the spectrometer is scanned such that
electrons of noise would be associated with the read, rather the wavelength range covered is continuously changing.
than 5 electrons 250 separate reads. Charge can also be The CCD clocking is matched to the scan rate of the grat-
binned in the computer software. However, noise reduction ing so that charge for a specific wavelength is collected
is proportional to the square root of the number of pixels in one packet and not throughout multiple pixels across
binned, rather than to the number of pixels binned. the CCD. Scanning the grating and moving the wavelength
CCD technology has improved, and read noise levels across the chip effectively increases the wavelength range
have become negligible; the primary reason for binning, collected with a single exposure.
therefore, is to decrease readout time. Unfortunately, spatial
resolution is lost when binning is employed. In the above Spectral framing Spectral framing is used to increase
example, where the column is binned, spatial resolution in the number of exposures that may be taken per second.
the vertical dimension is lost. However, two-dimensional When in spectral framing mode, part of the CCD is exposed
spatial resolution is not important in most spectroscopic to light while the other part is covered, or masked off, to
applications. prevent integration. Information gathered during an expo-
Some potential problems with binning include the detec- sure is shifted to the masked region as another exposure
tion and removal of CREs and the limited FWC of the is taken. Once information from multiple exposures has
serial register. It is difficult to remove a CRE after bin- filled the chip, the entire chip is read. In this manner, many
ning because CREs, easily detected when reading individual exposures can be collected but the chip need only be read
pixels, can be difficult to detect after binning. Another dis- once. Software is used to extract information derived from
advantage to binning is that the FWC of the serial register different exposures.
limits the dynamic range and effective FWC per pixel.
Therefore, integration times must be limited so that the Intensified CCDs for time-resolved Raman spectroscopy. In
FWC of the serial register is not exceeded. time-resolved Raman spectroscopy, only a few nanosec-
onds are collected per exposure. The short integration time
Subarray readout Subarray readout is used to collect makes time-resolved Raman spectroscopy a very low-light-
information from specific parts of the array. A subarray is level technique. As a result, very low levels of noise are
necessary to obtain the best S/N. Although CCDs are usu- The UV QE of CIDs is higher than the QE of frontside-
ally sufficiently sensitive to detect low light levels, the illuminated CCDs, because CIDs have more epitaxy not
primary reason for coupling an MCP intensifier to a CCD covered by electrodes. The photoresponse of the epi-
is to provide time resolution not attainable with gated CCD taxy still follows the wavelength response of silicon, as
electronics or fast shutter speeds. However, the dark current with CCDs. Unfortunately, only one manufacturer produces
of an MCP is much greater than that of a CCD, and cou- CIDs, so to date there is a limited number of design choices.
pling the two devices effectively negates the dark current The CID FWC is approximately three times that of a
reduction obtained by cooling the CCD. Another disadvan- CCD, resulting in a larger linear dynamic range.8 In addi-
tage of coupling MCPs to CCDs is the loss of resolution tion, CIDs are inherently antiblooming, because a thick
caused by the isotropic emission of photons from the MCP channel stop surrounds each pixel. The channel stop pre-
phosphor. Therefore, CCDs are usually only coupled to vents charge from leaking into adjacent pixels if FWC is
MCPs when high-speed gating is required for time-resolved exceeded.
spectroscopy. Charge-transfer efficiency is not an important param-
eter in CID operation, where charge is transferred over
Summary. CCDs are versatile array detectors with many extremely short distances within a single pixel. Therefore,
applications in Raman spectroscopy. The combination of CIDs can be operated at liquid nitrogen temperatures to
low read noise, low dark current, high QE, and readout reduce dark current to negligible levels.
flexibility yield a detector well suited to low-light-level The CID output signal is approximately two orders of
experiments. The literature shows that Raman spectroscopy magnitude lower (proportional to the charge per pixel
has greatly benefited from the use of CCD detectors. divided by the capacitance of the electronics) than that of
CCDs.8 Consequently, the relative readout noise for CIDs
4.3.3 Charge-injection devices is higher. Because the CID can take advantage of multiple,
A charge-injection device (CID) is a silicon-based charge- nondestructive reads, noise is reduced by the square root of
transfer device composed of pixels similar to those of a the number of reads. Read noise of the CMOS CID38, a
CCD. However, CIDs operate in a very different man- complementary metal oxide semiconductor CID produced
ner. Each pixel has two gate electrodes, a collection gate by CID Technologies, Inc. (Liverpool, NY), can be reduced
and a sense gate. Photons incident on the device create from about 150 electrons to less than five electrons by
electronhole pairs, and the holes accumulate under the performing numerous reads. In practice, the read noise level
collection gate, which is biased to a potential approximately is much lower than that of PDAs but is generally higher
twice as negative as that of the sense gate.8 A blank mea- than that of high-quality scientific CCDs.
surement is made of charge under the sense gate, and the A new method collective read mode has been devel-
potential on the collection gate is raised to move collected oped to minimize the read noise limitation of CIDs. The
charge under the sense gate. The charge under the sense CID electronics are configured such that the charge in all
gate is measured again, and the blank measurement is sub- pixels selected by the controller can be read simultaneously.
tracted from the measured charge (i.e. a double correlated Collective read mode results in the same readout signal
readout). as that obtained by reading each pixel independently and
CIDs have an advantage over CCDs in that readout does adding the charge, but the read noise for collective read
not destroy or change the amount of charge in the pixel, mode is decreased by the number of pixels read. This tech-
allowing one pixel to be read while others continue their nique is analogous to binning in a CCD, where all pixels in
integration. Because the readout mode is nondestructive, the one row of the detector can be read as one, resulting in a
pixel may also be read hundreds or thousands of times to decrease in read noise by the number of pixels binned. As
increase precision and reduce read noise. Following read- with binning, collective read mode results in loss of spatial
out, the CID is prepared for another exposure by biasing information. Operating in collective read mode is nonde-
both gates positive, to inject the charge into the substrate. structive, so averaging multiple reads can further reduce
CIDs are also capable of random access, in which any read noise. Unfortunately, this read mode is not yet com-
one pixel can be rapidly selected. In Raman, this random mercially available.
access may be useful because it enables varying integration Although CIDs are not currently used for detection in
times, using several short integrations for strong lines and Raman spectroscopy, new technology should enable them
one long integration for weak lines, dramatically increasing to become an effective tool. They combine the advan-
the usable dynamic range. This technique has been widely tages of high QE, large FWC, and low dark current with
used in atomic spectroscopy and is referred to as random an ability to read pixels randomly and nondestructively.
access integration. The only major problem high read noise could be
overcome by using collective read mode. Therefore, CIDs fabricated with up to 512 pixels. These detectors have a
may soon become an attractive detector choice for Raman wider spectral range than Ge detectors, commonly ranging
spectroscopy. from approximately 0.8 m to 1.7 m or to 2.2 m, depend-
ing on detector composition. The QE is high, above 80% for
4.3.4 Nonsilicon devices much of its spectral response. Dark current is less than one
electron per pixel per second at cryogenic temperatures.17
Raman spectroscopy is often plagued by fluorescence The spectral range changes as the detector cools, however,
caused by the high energy of visible radiation. Excita- shifting the response region to shorter wavelengths. As with
tion at longer wavelengths can minimize fluorescence. For Ge arrays, each pixel is read individually, so read noise is
example, the Nd : YAG (neodymium:yttrium aluminum gar- substantial: 650 electrons at high gain.17
net) laser, operating at 1.064 m, is commonly used as
an excitation source; however, it yields Raman signal at Platinum silicide detectors. Other nonsilicon array de-
wavelengths longer than 1.064 m. As previously noted, tectors are available that have not yet been used to detect
silicon detectors have low QE at wavelengths longer than Raman scattering in the NIR region. One such detector is
1.09 m, because photons do not have enough energy to the platinum silicide (PtSi) array, which covers a spectral
create an electronhole pair. In order to detect photons at region of 15 m. As with Ge or InGaAs detectors, pix-
wavelengths longer than 1.09 m, array detectors have been els in PtSi arrays are individually read. Two-dimensional
fabricated incorporating nonsilicon material sensitive in the arrays are available that allow for resolution in the second
NIR region. Two nonsilicon materials commonly used in dimension. Unfortunately, todays PtSi detectors have poor
array detectors for NIR Raman spectroscopy are germa- QE, below 6% for most of their spectral range.13,18
nium and indiumgallium arsenide. In addition, platinum
silicide and HgCdTe arrays appear to be feasible detectors
Other nonsilicon array detectors. Array detectors using
for Raman spectroscopy; however, the relatively low QE
other detector materials are emerging for use in the IR
of todays PtSi devices will limit their sensitivity.
(infrared). These include lead sulfide (PbS), lead selenide
Germanium detectors. Germanium (Ge) detectors, cover- (PbSe), indium antimonide (InSb), and mercury cadmium
ing the spectral range 0.8 m to 1.6 m with an average telluride (MCT). Although not yet widely applied to Raman
QE of 60%,13 are available in formats similar to those spectroscopy, MCT arrays show promise as detectors for
of PDAs. These detectors operate in capacitive discharge Raman excited in the IR. New MCT arrays have a spectral
mode, where each 25 m by 2.5 mm element in the array response to 2550 nm, high QE (greater than 50% over most
is connected to its own readout electronics. Common array of the spectral range), very low read noise (less than 10
sizes are 128 1 or 256 1 elements.13 These small array electrons), and low dark current (less than 0.1 electron per
sizes restrict wavelength coverage and spatial resolution. second).19 Their ability to detect photons in the IR may
In order to counteract this limitation, the polychromator lead to the wide application of these detectors to IR-excited
grating is often rotated to expose different regions of the Raman spectroscopy.
spectrum to the detector, and the spectra of the different Much Raman spectroscopy currently excited at 1.064 m
regions are then spliced together to make one complete is performed using a Fourier transform spectrometer. When
spectrum. the data is transformed, this noise is averaged over the entire
Dark current for Ge array devices is approximately 24 spectrum, resulting in degradation of the S/N of weaker
electrons per pixel per second when cooled to liquid nitro- bands. The incorporation of noise over an entire spectrum is
gen temperatures.14 A consequence of cooling is that the known as the multiplex disadvantage. No such disadvantage
longest wavelength detected by the array is reduced from is observed in MC, dispersive instruments, because noise
1.6 to 1.45 m. in the large bands is detected only in pixels monitoring that
To compensate for the high read noise, a number of band and, therefore, is not averaged in with the weaker
scans can be acquired and averaged, improving the S/N signals.
but increasing experiment time.15,16 To best exploit the low A disadvantage to working deeper into the NIR is that
dark current and high read noise of cooled Ge detectors, high noise from background thermal radiation is caused
it is preferable to maximize integration time to reduce the by the dewar materials, the window, and the general envi-
number of total scans needed to achieve the desired S/N. ronment. The detector must be cold shielded to avoid the
incidence of thermally generated photons on the chip. Opti-
Indium gallium arsenide detectors. Indium gallium ar- cal filtering is used to reduce background noise from the
senide (InGaAs) detectors have been developed as an alter- optical path, but the filter must be cold shielded to prevent
native to Ge detectors. Linear array devices have been reemission of the photons absorbed.13 A final disadvantage
to working deeper into the NIR is that there are several the detection system. Some loss of spatial resolution often
absorption bands for water that may affect the sensitivity results from slight mismatches between the various stages.
of Raman experiments and the relative intensities of Raman To gate off the device, the voltage drop across the channels
bands. can be set to zero or inverted so that no acceleration takes
The primary reason for investigating the use of nonsilicon place. Depending on configuration, MCPs provide photon
detectors is to decrease background noise from fluores- gains of 10 000 or more.20
cence by exciting Raman scattering in the NIR and IR Although the QE of a detector coupled to an MCP is
regions. Furthermore, array detectors do not suffer from the limited by the response of the photocathode on the MCP,
multiplex disadvantage of the FT devices commonly used a low QE can be offset by signal amplification. Peak QE
with Nd : YAG lasers. However, detectors sensitive in the generally occurs between 400 and 500 nm and is reported
NIR have drawbacks: current array sizes limit their spec- to be between 10% and 15%.5,21 The QE rapidly drops to
troscopic spatial coverage, and read noise and background less than 1% in the red and NIR regions. Because of their
from environmental IR emissions can substantially impair low QE in the red and NIR regions, MCPs are less suited
measurements. for detection in Raman systems, because many lasers for
Raman spectroscopy excite in the red or NIR regions to
decrease sample fluorescence.
4.4 Microchannel plates (MCPs)
MC detectors can be coupled to MCPs for signal inten- 4.5 Applications

sification and gating to provide time resolution for the
study of reaction dynamics. It is not possible to gate at 4.5.1 Chemical imaging
the nanosecond timescale using conventional shutters and
detector electronics; however, modern MCPs can often Chemical imaging by Raman spectroscopy provides spec-
obtain better than 1.5 ns resolution. To collect photons dur- troscopic information over a large sample area.22,23 Raman
ing a certain time frame, voltage can be applied across the imaging is employed to differentiate among chemical
microchannels for the specified time. regions of a heterogeneous sample. A tunable filter deter-
An MCP intensifier is a photocathode coupled to an array mines the wavelength of the image collected by the CCD.
of closely packed hollow glass fibers that act as channel By changing the wavelength of light collected and imaged
multipliers. Incident photons cause the photocathode to emit to a wavelength corresponding to a Raman emission band,
electrons that are accelerated into the hollow glass fibers. sample regions containing different chemical information
These channels are internally coated with a conductive, sec- can be distinguished.
ondary electron emissive material, and the electrode on the
output of the channel is positively biased with respect to the
4.5.2 Simultaneous spectra collection
electrode on the input of the channel. The resulting acceler-
ation and multiplication of electrons through the channel is The CCDs unique array format allows simultaneous col-
analogous to the operation of a PMT. Electrons emerging lection of spectra from several different samples under
from the channel impinge upon a phosphor screen, causing different experimental conditions. This technique can be
the emission of photons. These photons are measured by useful for monitoring drug quality at several points along
Sample spatial resolution
Sample 1
Sample 2
Intensity
Sample 3
Sample 4
Wavelength resolution Pixel

Figure 9. A representation of the appearance of spectra from several resolved samples as an image on a CCD. Larger spots indicate
more collected electrons.
a production line. In this application, the output of a high- PDA Photodiode Array
powered laser is split and directed into several single fiber PMT Photomultiplier Tube
optic probes for excitation and collection of Raman scat- QE Quantum Efficiency
tering. Fibers from the collection bundle are arranged such RA Resistive Anode
that a stack of fibers is aligned vertically along the entrance SC Single-channel
slit of the spectrometer. The CCD records a Raman spec- TDI Time-delayed Integration
trum of each bundle in the horizontal direction, with each
sample separated in the vertical dimension. In Figure 9, the
CCD records the spectra of four different samples resolved REFERENCES
spatially in the vertical direction.
This configuration is complicated by the fact that many 1. R. Savoi, M. Pezolet, S. Dallaire and C. Simard, Can. J.
Appl. Spectrosc., 39, 164 (1994).
polychromators have optical aberrations. Spectra from each
sample must be spatially separated in the focal plane 2. N.E. Schlotter, S.A. Schaertel, S.P. Kelty and R. Howard,
Appl. Spectrosc., 42, 746 (1998).
of the spectrometer. Unfortunately, most commonly used
3. J.D. Ingle and S.R. Crouch Spectrochemical Analysis, Pren-
CzernyTurner polychromators suffer from an astigma-
tice Hall, NJ (1998).
tism that transforms a point at the slit into a line at the
4. R.K. Chang and M.B. Long, Optical Multichannel Detec-
focal plane, thereby degrading the spatial separation of
tion, in Light Scattering in Solids II, eds M. Cardona and
spectra.24 In this instance, information resolved in the verti- G. Guntherodt, Springer-Verlag, Berlin, 179205 (1982).
cal dimension at the entrance slit is overlapped at the detec- 5. W.P. Acker, B. Yip, D.H. Leach and R.K. Chang, J. Appl.
tor. The astigmatism can be minimized in CzernyTurner Phys., 64, 2263 (1988).
monochromators by specially designed optics. 6. Technical Data Sheet, PDA 1024 linear array, Princeton
Instruments.
7. R.L. McCreery, CCD Array Detectors for Multichannel
5 CONCLUSION Raman spectroscopy, in Charge-transfer Devices in Spec-
troscopy, eds J.V. Sweedler, K.L. Ratzlaff and M.B. Den-
ton, VCH Publishers, New York, 227279 (1994).
Interest in Raman spectroscopy has recently been height-
ened because of improved laser and array detector tech- 8. G.H. Sims, Principles of Charge-transfer Devices, in
Charge-transfer Devices in Spectroscopy, eds J.V. Sweed-
nology. In addition to the MC advantage, the low-noise
ler, K.L. Ratzlaff and M.B. Denton, VCH Publishers, New
characteristics of array detectors have yielded increased S/N York, 958 (1994).
in low-light-level experiments. CCD array detectors are par- 9. J.H. Giles, T.D. Ridder, R.H. Williams, D.A. Jones and M.B.
ticularly sensitive to low levels of light and have extremely Denton, Anal. Chem., 70, 663A (1998).
low read noise and dark current. The two-dimensional res- 10. G.H. Sims, Brief History of Charge-transfer Devices, in
olution available on some array detectors is an additional Charge-transfer Devices in Spectroscopy, eds J.V. Sweed-
advantage. All of these characteristics have led to the ler, K.L. Ratzlaff and M.B. Denton, VCH Publishers, New
development of exciting new applications, such as Raman York, 17 (1994).
imaging and multisample spectra collection. 11. D.R. Lide (ed.), CRC Handbook of Chemistry and Physics,
74th edition, CRC Press, Boca Raton, FL (1993).
12. C. Dyer and B.J.E. Smith, J. Raman Spectrosc., 26, 777
ABBREVIATIONS AND ACRONYMS (1995).
13. B. Chase, Arrays for Detection Beyond One Micron, in
Charge-transfer Devices in Spectroscopy, eds J.V. Sweed-
ADC Analog-to-digital Converter ler, K.L. Ratzlaff and M.B. Denton, VCH Publishers, New
ADU Analog-to-digital Unit York, 109131 (1994).
AR Antireflective 14. J. Barbillat and E. da Silva, Spectrochim. Acta A, 53, 2411
CID Charge-injection Device (1997).
CRE Cosmic Ray Event 15. B. Chase and Y.Talmi, Appl. Spectrosc., 45, 929 (1991).
FWC Full Well Capacity
16. C. Engert, V. Deckert, W. Kiefer, S. Umpathy and H. Hama-
MC Multichannel guchi, Appl. Spectrosc., 48, 933 (1994).
MCP Microchannel Plate 17. Technical Data Sheet, OMAV, Princeton Instruments.
MOS Metal-oxide Semiconductor
18. J.W. MacKenty, MACSQIID: A package for the reduction
MPP Multipinned Phase of data from the SQIID Infrared Camera, in Proceedings of
Nd : YAG Neodymium:Yttrium Aluminum the Astronomical Society of the Pacific Conference Series,
Garnet eds R.A. Shaw, H.E. Payne and J.J.E. Hayes, Vol. 77 (1995).
19. Technical Data Sheet, Hawaii Focal Plane Array, Rockwell 22. J.J. Andrew, M.A. Browne, I.E. Clark, T.M. Hancewicz and
Scientific. A.J. Millechope, Appl. Spectrosc., 52, 790 (1988).
20. M. Lampton and F. Paresce, Rev. Sci. Instrum., 45, 1098 23. M.D. Scaeberle, H.R. Morris, J.F. Turner, II and P.J. Treado,
(1974). Anal. Chem., 71, 175A (1999).
21. H. Tanino and H. Okushi, Jpn. J. Appl. Opt., 29, L2133 24. R.D. Jalkian, R.S. Pomeroy, J.D. Kolczynski, M.B. Denton,
(1990). J.M. Lerner and R. Grayzel, Am. Lab., 21, 80 (1989).
Fourier Transform Near-infrared Raman
Spectroscopy
Bruce Chase
DuPont Experimental Station, Wilmington, DE, USA
1 INTRODUCTION photomultipliers) and the loss in scattering cross-section

dictated by the n4 law that governs scattering intensities.
The decade of the 1990s has been witness to a veritable Very soon after the advent of FT-Raman spectrometry,
explosion in the popularity and utility of Raman scattering developments in array-based detectors occurred and Raman
spectrometry. Biomedical studies,1 as well as applications spectrometers based on focal plane array (FPA) detec-
in chemistry2 and materials science,3 are all readily appar- tors were investigated.7,8 These instruments, now based
ent, and the technique has even made its way into the on single monochromators, holographic notch filters, and
process monitoring environment.4 Low-cost instruments charge-coupled device (CCD) detectors, combine with FT-
are available that have allowed undergraduate programs to Raman spectrometers to account for most of the Raman
introduce students to the instrumentation and concepts of instrumentation being used today. However, it was the
Raman scattering. While infrared (IR) instruments are still initial observation of fluorescence-free Raman scattering
dominant in numbers, it is quite clear that Raman scat- made with FT-Raman instrumentation that catalyzed this
tering has become a significant component of analytical tremendous growth in Raman spectroscopy resulting in the
spectroscopic measurements. The driving force behind this high-quality instrumentation available today.
phenomenon has been the development of Raman instru- The basic FT-Raman instrumentation currently in use
mentation, which is easy to use, has high sensitivity, and today is really quite similar to the instruments discussed
is able to overcome the problems associated with fluores- in the first paper. We can conveniently break down the
cent interference that have plagued Raman measurements instrumentation in to four parts: lasers, sampling and associ-
for the last 50 years. ated filters, interferometers, and detectors. There have been
The most crucial aspect of these developments has been significant advances in the performance for each of these
the use of near-infrared (NIR) lasers to avoid the excitation components. However, before discussing these instrumen-
of fluorescence. In 1986, Hirschfeld and Chase5 published tal components, it is important to digress to a discussion on
the nature of the noise in a interferometric measurement.
their first paper on Fourier transform (FT)-Raman spec-
This will have impact on the choice of laser and detector
trometry, where excitation at 1.064 m was used. Though
for the FT-Raman experiment.
this paper was entitled FT-Raman Spectrometry, the most
important aspect was the recognition that by going to suit-
ably long wavelength excitation, most fluorescence could be
2 NATURE OF NOISE
avoided a concept that was discussed more than 25 years
earlier in a paper by Gebbie and co-workers,6 then forgot- The performance of any interferometric spectrometer is crit-
ten. The use of a Michelson interferometer was required ically dependent on the nature of the dominant noise source.
to compensate for the poor-quality detectors (relative to Noise can be characterized as one of three different types.
Detector noise is independent of signal level. This is the
John Wiley & Sons Ltd, 2002. noise figure characterized by the noise equivalent power
(NEP), D , or dark count of the detector. Often this detector is shot noise limited. For a dispersive experiment, under
noise can include contributions from the preamplifier and shot noise limited conditions, both the noise and the signal
filtering electronics when one is using low noise detec- will vary as a function of the intensity of the band.
tors, but it is still independent of signal level. The next
type of noise would be shot or statistical noise, which is
due to the uncertainty in counting statistics. This noise is 3 LASERS
clearly dependent on the signal level. As a higher photon
flux occurs, the absolute noise will increase as the square The choice of laser to use in FT-Raman measurements is by
root of the flux. Since the signal increases linearly with flux definition a compromise. Long wavelength is necessary to
(in the absence of nonlinearities), the signal-to-noise ratio avoid exciting fluorescence, but long wavelength results in
(S/N) will increase by the square root of the flux. If either a loss in scattering cross-section since the Raman scattering
the detector noise is reduced or the flux increased, even- cross-section is proportional to n4 . In addition, as the laser
tually one will transition from detector noise dominance wavelength is increased, it forces the use of longer wave-
to shot noise dominance. For detectors such as photomul- length sensitive detectors. For a laser operating at 1.064 m,
tipliers or CCDs, the detector noise is so low that given the detector must be sensitive out to 1.67 m (6000 cm1 )
any photon flux at all on the detector, one is immediately in order to get a full spectral range Raman spectrum.
in the shot noise limit. It is obvious that if the experiment These detectors will carry with them increased noise since
results in a photon flux where shot noise dominates detector they are sensitive to the longer wavelengths where thermal
noise, further improvements in detector noise performance background fluctuations can become the limiting noise fac-
are irrelevant. The final possible noise type is source fluc- tor. The best compromise to date has been the first laser
tuation or flicker noise. In this case, the noise is linearly chosen for FT-Raman. A neodymium : yttrium aluminum
proportional to the flux. Source fluctuation noise can arise garnet (Nd : YAG) laser operating at 1.064 m provides
from many sources. The most obvious is a varying flux from the optimum fluorescence minimization for most chemi-
the radiation source in the experiment. For Raman scatter- cal systems, while maintaining the ability to operate with
ing this would be a noisy laser. It could also arise from the most sensitive NIR detectors, InGaAs and germanium.
sample fluctuations. For example, a liquid sample with par- A very similar laser based on neodymium : yttrium lithium
ticulates that cause random changes in scattering, or a liquid fluoride (Nd : YLF) is also employed in some commercial
sample that undergoes heating and fluctuations in refractive FT-Raman systems. This unit lases at 1.047 m. The subtle
index can result in source fluctuation noise. Any factor that shift does affect the FT-Raman spectrometry, and care must
causes the detected signal to vary randomly in time can be taken to specify the exact wavelength when replacing the
cause source fluctuation noise to become dominant. If the Rayleigh line filters (see next section). Figure 1 shows the
multiplex advantage is to be significant, then the domi- spectral ranges relevant to a discussion around FT-Raman
nant noise must be detector noise. For the interferometric spectrometry. The respective wavelength regions are shown
measurement, we want to start in a detector noise limited for excitation at 532 nm, 785 nm and 1064 nm, along with
situation, where the multiplex advantage is N1/2 , where N the regions where various detectors have reasonable sen-
is the number of spectral elements. If the detector is really sitivity. In terms of fluorescence excitation, even the most
noisy, we may never transition into a shot noise limited highly conjugated aromatic systems, such as coronene or
experiment. However, as detectors improve, and high flux substituted anthraquinones, have absorption maxima near
experiments are considered, one must always be aware of 600 nm. It is quite clear that excitation at 1.064 m should
the possibility of being shot noise limited. In this case, the avoid the fluorescence from any purely organic material.
multiplex advantage reduces to one. Finally, we never want However, inorganics, especially transition metal ion and
to be in a source noise limited condition. Under these con- lanthanide ion systems, which have excited state energies
ditions, the multiplex advantage turns into a disadvantage in the NIR, can still cause problems.
(N1/2 ), and we would be better off doing the experiment Alternative lasers for FT-Raman have been explored,
with a dispersive system and single-channel detection. both to shorter and longer wavelength. It is quite feasible
A final point about noise concerns the distribution of to carry out FT-Raman at shorter wavelengths using diode
noise throughout the spectrum. For an interferometric lasers or tunable Ti/sapphire lasers. However, given the
experiment, to first order, the noise in the interferogram relative performance of CCD detectors in a multichannel
is distributed equally throughout the spectrum by the FT. spectrometer, with respect to InGaAs or Ge, it makes no
This gives rise to the interesting observation that the rela- sense. If one can work with a shorter-wavelength laser
tive S/N for a given spectral feature is a function only of the with no sample fluorescence, the best approach is to use a
intensity of that feature. This is true even if the experiment dispersive single monochromator with CCD detection. The
Fourier Transform Near-infrared Raman Spectroscopy 3
InSb
Ge or IGA
CCD/Si
5
3
7
2 1
8
0
5 500 1000 1500 2000 2500 3000
6
4 Raman shift (cm1)
Figure 2. FT-Raman spectrum of copper phthalocyanine excited
at 1.064 m. (Reproduced by permission from Asselin and
600 800 1000 1200 1400 1600 1800 2000 Chase.9 )
Nanometers
Figure 1. Spectral range for Raman spectra excited at 532 nm,
785 nm, and 1064 nm. Optimum sensitivity spectral range for
CCD, indium gallium arsenide (IGA), Ge, and InSb detectors.
superior performance of these instruments when there is no

background present is due to the larger scattering cross-
section and the much lower noise level obtained with CCD
detectors. The use of FPA detectors offers a multichannel
advantage equivalent to the multiplex advantage of FT-
Raman in the detector noise limit. This holds even when the
experiment is shot noise limited. Since the Raman scattering
intensity is now dispersed over many pixels, any individual 200 400 600 800 1000 1200 1400 1600
pixel sees less intensity, and thus the shot noise, which Raman shift (cm1)
varies as the square root of the intensity, is less. Figure 3. FT-Raman spectrum of copper phthalocyanine excited
Operation with a longer-wavelength laser has also been at 1.339 m. (Reproduced by permission from Asselin and
demonstrated. Raman scattering at 1.3 m is even weaker Chase.9 )
than at 1.064 m, and the response of a germanium detec-
tor limits the spectral bandwidth in the Raman spectrum superior results using a single-channel detector and a scan-
to 1800 cm1 Stokes shift. However, for specific cases ning dispersive instrument (assuming all else is equivalent
where fluorescence is still a problem at 1.064 m, this is an such as throughput and efficiency). In the earlier versions
effective approach. Figure 2 shows the full Raman spec- of Nd : YAG lasers used in FT-Raman, this was a seri-
trum of copper phthalocyanine9 obtained with excitation ous consideration. These mainframe, flash-lamp pumped
at 1.064 m. Figure 3 shows the Raman spectrum of the systems often had an output fluctuation of 12% within
same material obtained with excitation at 1.339 m. The the frequency range of the FT-Raman system. Though this
reduced Raman spectral range is due to the detector cut- was not the limiting noise, it did contribute to the overall
off limiting the range. The fluorescence under excitation noise level. Power stabilization is required for such lasers,
at 1.064 m is clearly seen in Figure 2. Moving to longer and very effective results were obtained using a Cam-
wavelengths (1.3 m) results in the dramatically improved bridge Instruments Model LPC controller. The introduction
spectral quality shown in Figure 3. of diode-pumped Nd : YAG and Nd : YLF lasers overcame
After the laser wavelength has been chosen, it is impor- the need for the stabilization unit. These lasers have power
tant to consider the noise characteristics of the laser. For a output fluctuation of less than 0.2% in the appropriate fre-
multiplexing instrument, the worst type of noise is source quency range, which brings the laser noise well below the
fluctuation. If source fluctuation, or flicker, noise is the other noise sources in the system.
limiting noise, then the spectrum actually suffers from a One of the advantages of a Nd : YAG laser is the available
multiplex disadvantage. In other words, one could achieve power. Commercially available lasers can easily deliver
150 mW
150 mW
40 mW
40 mW
20 mW
20 mW
500 1000 1500 2000 2500 3000 200 400 600 800 1000 1200
Raman shift (cm1) Raman shift (cm1)
Figure 4. FT-Raman spectra of MoO3 excited at 1.064 m and Figure 5. Expanded region of Figure 4 showing the effects of
varying power levels. power levels on the spectral features.
several watts to the sample. It was realized early in FT- system with the laser pulsing synchronized to the data
Raman spectrometry that many samples could withstand acquisition.
such high power levels since there is usually no optical An alternate approach for using pulsed excitation in-
absorption in this spectral region. In early publications and volves asynchronous sampling developed by Sakamoto
instrument company application notes it was not unusual et al.11 Sampling theory allows for unsynchronized oper-
to see spectra taken at 12 W. The user should exercise ation if certain frequency/bandwidth constraints are met.
extreme caution in applying these powers routinely to sam- Asselin and Chase12 have demonstrated equivalent sup-
ples. In most laboratories now, the routine power levels pression of thermal background using this asynchronous
should rarely exceed several hundred milliwatts. Higher approach. The use of pulsed excitation in FT-Raman also
powers often produce thermal effects that distort the spec- allows for time-resolved measurements. Sakamoto et al.13
trum and can induce subtle and not so subtle chemical have demonstrated the ability to record excited-state Raman
changes in samples that are thermally sensitive. Figure 4 spectra using asynchronous sampling.
shows the effect of laser power and subsequent heating
on the Raman spectrum of MoO3 . Higher power results in
increased thermal background as seen by increased signal 3.1 Collection and filtering
at large Stokes shifts. Figure 5 shows the spectral changes
due to thermally induced chemistry. Since the FT-Raman In any Raman experiment, efficient collection of the scat-
spectrum is itself sensitive to elevated temperatures as the tered photons is critical, since the scattering efficiency is
black-body emission intrudes into the Raman spectrum, it so low to start with. FT-Raman is no different from con-
is clear that FT-Raman studies of samples at elevated tem- ventional Raman scattering experiments in this area, and
peratures will be difficult. both reflective and refractive optic collection systems can
The use of pulsed lasers in FT-Raman spectrometry be employed. However, one must exercise caution in the
has for the most part been ignored, for the simple rea- choice of lens, since some multielement lenses utilize opti-
son that interferometric measurements are extremely sen- cal cements that can absorb in the NIR. The safest choice
sitive to source fluctuations. There is no worse fluctua- is to use air gap lenses.
tion than a source turning on and off. However, if the A major difference between conventional, dispersive
pulsing of the laser can be synchronized with the data Raman and FT-Raman lies in the focusing requirements
acquisition it is possible to use pulsed sources. Such an for the laser. When dispersive spectrometers are utilized,
approach has been explored by Cutler and Petty.10 They the f number matching required to couple f/1 collection
demonstrated the utility of pulsed excitation for minimiza- optics with f/4 (or maybe f/2) spectrometers dictates a
tion of constant thermal background using an FT-Raman magnification factor of 24. Since the entrance slit (or fiber
optic) is usually no greater than 100 m (dictated by res- absorbance at the laser wavelength must be used. However,
olution requirements), this means that the laser must be the transmittance of the filter over the Stokes shift range
focused to less than 2550 m at the sample. If the spot must be high, in order to prevent the loss of the scattered
size at the sample is any larger than this, collected light photons. As with the laser filters, these elements can be
will be wasted by overfilling the spectrometer or fiber optic either dielectric filters or holographic elements. Since the
entrance aperture. Such tight focusing invariably produces actual power collected in the scattered channel is relatively
high power densities with the concomitant possibility of low, damage thresholds are less relevant here. However,
thermal degradation. For a typical Michelson interferome- because of the distributive nature of a multiplexed measure-
ter, the equivalent entrance aperture is closer to 4 mm. Thus ment, it is critically important to filter the Rayleigh line to
a focused spot size of 1 mm is sufficient to allow for full an intensity less than the strongest line in the Raman spec-
utilization of the collected scattered photons. FT-Raman trum. If this is not accomplished, the noise associated with
measurements can be made at much lower power densities the Rayleigh line, whether shot or flicker, will be redis-
for the same total photon flux, resulting in less chance for tributed over the entire spectrum, reducing the S/N for the
thermal degradation. It should be noted that photochemical Raman peaks.14 The effects of insufficient filtering can be
degradation does not benefit from the increased spot size seen easily in the spectra shown in Figures 6 and 7. These
since it is a function of photon flux and not power density. are FT-Raman spectra of bis methyl styryl benzene obtained
Of course, this argument cannot be applied if the collec- with three different levels of optical filtering corresponding
tion optics in the FT-Raman experiment are fiber optically to an optical density at the laser line of 10, 9, and 8, respec-
coupled to the spectrometer. Then the fiber optic will likely tively. As the filtering is reduced the intensity of the laser
become the limiting aperture. line observed in the spectrum increases, but the noise level
There are significant requirements around the optical fil-
tering necessary for both the incident laser and scattered
light. As with any laser used for Raman scattering, the
nonlasing lines must be optically filtered. For an Nd : YAG
laser the most severe nonlasing lines occur near 2000 cm1
Stokes shift. These are easily filtered using either dielectric
or holographic filters. Since higher powers may occasion-
ally be employed, the holographic filters are preferable
due to higher damage threshold. One particular aspect to
watch for during operation is leakage of these nonlasing
lines into the spectrum as the filter degrades. This can be
most conveniently detected by looking at the scattering of
KBr powder. The appearance of sharp line features near (c)
2000 cm1 is indicative of aging of the laser line filter.
The second set of filters relevant to FT-Raman are the
laser line filters on the HeNe laser, which is used in
the drive systems of most Michelson interferometers for
retardation referencing and data acquisition triggering. This
laser has numerous neon emission lines in the NIR. In (b)
order to avoid the presence of these lines in the FT-
Raman experiment, the output of the HeNe laser must
be filtered using a notch transmission element with a center
wavelength of 632.8 nm. In addition there must be a long-
pass filter with a cut-on wavelength near 800 nm in front of
the detector to avoid saturation effects from the scattered (a)
632.8-nm light from the reference laser.
Finally, and most importantly, the Rayleigh scattered
light from the Nd : YAG laser must be filtered before the 500 1000 1500 2000 2500 3000
detector. This was the most challenging task in the ini- Raman shift (cm1)
tial FT-Raman experiments. Since the Rayleigh scattered Figure 6. FT-Raman spectra of bis methyl styryl benzene with
intensity can be 810 orders of magnitude more intense varying degrees of laser line filtering. Optical density: (a) 10;
than the Raman scattered intensity, a filter with very high (b) 9; (c) 8. (Reproduced by permission from Chase.14 )
optimal operation in the 11.6 m region. Most of the com-

mercially available instruments today operate with quartz
or calcium fluoride beamsplitters. Either one can achieve
modulation efficiencies of greater than 70%. Beamsplitter
alignment is more critical for FT-Raman in the NIR than in
the mid-infrared (MIR) since the wavelengths are shorter.
The dedicated instruments have fixed beamsplitters that
minimize realignment problems. However, most systems
are reasonably stable in terms of long-term fluctuations.
(a)
The other important variable is the overall transmittance
of the interferometer. A dedicated instrument can minimize
the number of reflecting surfaces. In the optimum dedi-
cated FT-Raman instrument, there are only two mirrors
(one collection and one detector focusing) in addition to
(b)
the Rayleigh line filters and beamsplitter. The weakness of
this particular design in the lack of a true J-stop system.
The optical aperture is set by the smaller of the detec-
500 1000 1500 2000 2500 3000
tor area or the illuminated sample area. This can lead to
Raman shift (cm1)
significant sensitivity to sample position, as the half-angle
Figure 7. FT-Raman spectra of Figure 6(a) and (b) with the scale divergence of the optical system may be determined at the
expanded. (Reproduced by permission from Chase.14 )
sample and is subject to variability as a function of focus-
ing. As discussed in the article on artifacts (see Anomalies
far away from the laser line also increases. This can be
and Artifacts in Raman Spectroscopy), this can cause
seen most clearly in Figure 7 where the two lower curves
frequency shifts that are a function of sample position. The
are shown after scale expansion.
strength of the design is high transmittance due to the low
The proper choice of holographic filter can also provide
number of reflecting surfaces.
for acquisition of both Stokes and anti-Stokes spectra. Since
The alternative approach is to use an accessory that
the anti-Stokes spectrum is a mirror image of the Stokes
serves as the illumination and collection unit for the Raman
spectrum, with changes in band intensity due to Boltzman
scattering. The collected beam is then introduced into
populations, there is no obvious need to acquire these data.
the interferometer by coupling mirrors. An appropriate
However, it can be very useful for fine tuning the frequency
scale. When there is enough anti-Stokes intensity to observe beamsplitter is inserted in place of the conventional MIR
several bands in both Stokes and anti-Stokes scattering, it is beamsplitter, and an alternative detector is chosen. Early
a simple matter to adjust the frequency base to give identical versions of FT-Raman accessories had up to eight mirrors
frequencies for both Stokes and anti-Stokes bands. This in addition to the collection and detector focusing elements.
correction removes linear frequency shifts and leaves only The real strength of this approach was that it allows
higher-order errors, which for a Michelson interferometer the same instrument to serve multiple needs for both IR
are negligible. In addition, acquisition of both Stokes and and Raman measurements. Most manufacturers have now
anti-Stokes spectra can be used to calculate an approximate recognized the need to reduce the number of reflecting
sample temperature, using the Boltzman relationship. surfaces in the FT-Raman accessory. A comparison of these
two approaches is illustrated in Figures 8 and 9.
With auto-alignment features, component interchange is
3.2 Interferometers not difficult. Most commercial systems now offer gold-
coated mirrors within the interferometer to help minimize
This is the one area where the user has very little flex- the transmission losses from the additional reflecting sur-
ibility. One is basically stuck with the interferometer as faces. These are usually fabricated using hard-coat gold and
delivered. So the only real decision to make is whether to seldom need cleaning. In fact, there is really only one mirror
use a dedicated FT-Raman instrument or an accessory to that may need occasional cleaning. The detector focusing
a research grade instrument. Clearly, when an instrument mirror in instruments that have a cryogenic, down-looking
is asked to perform a variety of measurements over differ- detector can occasionally develop deposits from conden-
ing spectral ranges, there is almost always a compromise sation dripping off the detector housing. This can reduce
of some sort. The dedicated instruments have the advan- the transmittance of this mirror. Gentle cleaning with an
tage of form following function. They are designed for ethanol stream can usually remove these deposits.
OF
EQUINOX
APT
Laser
Aperture
Control electronics
Detector
Sample compartment IR FRA

106/S
Rayleigh filter
Objective
XYZ
table
Sample compartment Raman
Figure 8. Optical diagram of an FT-Raman accessory with base interferometer. (Courtesy of Bruker Instruments.)
The choice between dedicated instruments and acces- When choosing a detector for FT-Raman, there are sev-
sories is often dictated simply by cost. The accessories eral considerations. First, we assume that an Nd : YAG
are cheaper than the full instrument assuming one already laser, operating at 1.064 m is being used. Therefore, to
has a research-grade interferometer capable of operating in get the maximum coverage in the Stokes region, the detec-
the NIR. If there are no budgetary constraints, a dedicated tor needs to be sensitive out to at least 1.6 m. However,
instrument is preferable due to less component interchange it should be kept in mind that the longer-wavelength oper-
and no conflict between IR and Raman measurements. ation brings with it a higher noise floor. For detectors that
are background limited (BLIP condition), the noise is due to
statistical fluctuations of the thermal background in the field
3.3 Detectors of view of the detector.15 As the detector responds to longer
wavelengths, the detected blackbody emission from the
As with the other components of the FT-Raman spectrome- background will necessarily increase. Once the operating
ter, the detector situation has also stabilized over time with range has been set and appropriate detectors identified, then
a current choice of two elements, InGaAs or germanium. the criteria become NEP, linearity and ease of use.
Laser Clearly, if long-wavelength response is needed the element

must be no colder than dry ice temperatures. However, the
optimum sensitivity is achieved at 77 K. Comparisons of
various InGaAs detectors have established that the effective
Control electronics
NEP is really dominated by preamplifier performance. It
has been possible to get an NEP D 1015 W Hz1/2 from
an InGaAs detector, but significant work in preamplifier
design and cooling is involved. Since these detectors are
operated with very high feedback resistance, the frequency
J-stop
Filter module I response can be limited. A very small amount of stray
Laser out capacitance, such as one might encounter in the leads from
External the detector element to the preamplifier, can couple with the
Filter module II input feedback resistor to produce significant low-pass filtering.
These detectors are usually run with modulation frequencies
of less than 5 kHz. This in turn limits the optical veloc-
ity of the interferometer to less than 0.25 cm s1 . Most
commercially available InGaAs detectors designed to run
at 77 K have dewar hold times of greater than 8 h, so day-
Sample compartment long operation is feasible. Overnight runs are difficult due
Figure 9. Optical diagram of a dedicated FT-Raman system.
to loss of cooling fluid. InGaAs detectors can be obtained
(Courtesy of Bruker Instruments.) in both down-looking and side-looking configurations for
easy use in any detector module. The usual InGaAs ele-
ment is a 1-mm2 element. In several commercial FT-Raman
spectrometers, this becomes the limiting aperture for the
optical system.
The current version of germanium detectors must be
cooled for proper noise performance and are usually oper-
ated at 77 K. Earlier versions of germanium detectors were
operated as simple low-bias diodes. The noise performance
was significantly inferior to InGaAs. The newer version
of germanium detector operates with high-purity germa-
nium elements biased near 100 V. The spectral coverage
(a) (b) (c)
goes out to 3400 cm1 Stokes shift, and the sensitivity is
excellent. The NEP can be less than 1015 W Hz1/2 . As
with the InGaAs detector the frequency response is limited.
Most systems operate at less than 5 kHz for a sampling
6000 6800 7600 8400 9200
frequency. There are high-gain versions of the germanium
Wavenumber (cm1)
detector that have an NEP near 1016 W Hz1/2 but even
lower frequency response. These detectors are well suited
Figure 10. Spectra of a white-light source showing the tempera-
ture dependent spectral response of an InGaAs detector. (a) Room
to step-scan operation with a phase modulation frequency
temperature (NEP 1.3 1013 W Hz1/2 ; (b) 198 K (NEP 4.4 below 1 kHz. The hold times for the dewars are in excess
1014 W Hz1/2 ); (c) 77 K (NEP 2.0 1014 W Hz1/2 ). (Repro- of 8 h, allowing for full-day operation. One particular char-
duced by permission from Chase.16 ) acteristic of this detector is the sensitivity of the noise
performance to the quality of the vacuum in the dewar.
InGaAs was one of the first detectors to be used. When We have found that as the dewar vacuum degrades, one
cooled to 77 K the spectral range becomes somewhat lim- observes an increase in background noise. Repumping the
ited, with a long wavelength cut-off at 3000 cm1 Stokes dewar to 107 torr will alleviate this problem. It is believed
shift. This eliminates observation of aromatic CH stretch- that the degradation in noise performance is due to sur-
ing modes as well as OH and NH stretching modes. face adsorption of gas molecules on the germanium element
However, cooling does produce a better NEP for the detec- changing the surface state energy of the germanium. The
tor. Figure 10 shows the effect of element temperature pump-out procedure can be enhanced by heating the detec-
on the long-wavelength cut-off of an InGaAs detector.16 tor during pumping. This is most easily done by filling the
dewar with hot water. One should always confirm with the
detector supplier that this treatment is acceptable.
A comparison of these two detectors really requires
that the detectors are on the same interferometer to avoid
measuring differences in instrument performance. In such
a case, one must always accept the fact that one or the
other detector may not exhibit the ultimate performance that
that element is capable of. We have made a comparison of
a germanium detector versus a room temperature InGaAs (a)
detector on a Nicolet 910 interferometer. Using both cyclo-
hexane and anthracene as samples, 100 scans acquisition,
and 300 mW power, the S/N was measured for the strongest
band in the spectrum. The noise was measured by acquiring
a spectrum with the laser off. This neglects the contribu-
tion from shot noise and flicker noise. Figure 11 shows the
two spectra of cyclohexane scaled with respect to the noise
floor for each element, i.e. the Raman spectra were scaled
by the ratio of the noise spectra. The germanium detector
had a S/N approximately five times greater than the InGaAs (b)
detector. This comparison is roughly in agreement with the
prediction based on manufacturers stated NEP. Clearly,
improved performance with the InGaAs detector could be
achieved by cooling. Similar results are shown in Figure 12
for anthracene, a much more strongly scattering sample. 0 500 1000 1500 2000 2500 3000 3500
Raman shift (cm1)
Figure 12. FT-Raman spectra of anthracene obtained with (a) an
InGaAs detector and (b) a germanium detector.
Figure 13 shows the effect of signal level on noise. For the

anthracene spectra obtained with either detector, duplicate
runs were made. A one-to-one difference between dupli-
cate runs should produce a noise spectrum with all sample
features eliminated. Figure 13(a) shows the noise level for
the anthracene spectrum taken with the InGaAs detector.
(a) Figure 13(b) shows the noise level under the same con-
ditions but with the laser turned off, i.e. the dark noise
spectrum. There is a slight increase in noise level as a
function of illumination, indicating that we are not fully
in the detector noise limit for a strongly scattering sample,
even with the InGaAs detector. Figure 13(c) and (d) show
similar results for the germanium detector. Clearly, there
is a component of shot noise observed in this experiment.
These results confirm the conclusion stated earlier, that for
strongly scattering samples, further improvements in detec-
tor performance will not produce any improvement in the
spectral S/N.
(b)
0 500 1000 1500 2000 2500 3000 3500 3.4 Standard operations
Raman shift (cm1)
Figure 11. FT-Raman spectra of cyclohexane obtained with The objective in any spectroscopic measurement should be
(a) an InGaAs detector and (b) a germanium detector. the measurement of a spectrum that is independent of the
for sample alignment. Optimization of position is accom-

(a)
plished by monitoring the center-burst of the interferogram
and adjusting the sample position for a maximum signal.
When switching between liquid and solid samples, one will
observe significant differences in optimal sample position.
This is due to the relatively large depth of focus and field.
Liquid samples benefit from a focus inside the volume.
(b)
Solid samples that are highly scattering will give the max-
imum signal when the focus is held near the surface of
the sample. For weak Raman scatterers contained in glass
(c) containers, it is sometimes possible to end up optimizing
the focus on the signal from the glass container rather than
from the sample inside. In these cases one should also moni-
tor the spectrum after Fourier transformation to ensure that
the proper signal is being maximized. Glass containers will
contribute a spectral signature for silica at 250 cm1 . As the
focus is moved into the sample and away from the glass
(d) container wall, this peak will disappear.
As mentioned earlier in the discussion of lasers, there
is significant power available with Nd : YAG systems. The
temptation to use all that power is always there, since
the Raman signal improves linearly with power. However,
1600 1800 2000 2200 2400 2600 2800 3000
since the possibility of thermal effects is always present,
Raman shift (cm1) one should start at low powers: 50 mW is a reasonable
Figure 13. Relative noise spectra for anthracene Raman spec- starting point, and then the power is increased slowly if
tra. (a) InGaAs detector laser on; (b) InGaAs detector laser the S/N requirements are not met. As pointed out earlier,
off; (c) germanium detector laser on; (d) germanium detector
laser off. the FT-Raman experiment does not require tight focus of
the laser on the sample. If it is possible to vary the focus
instrument and conditions used for acquisition. This concept of the laser on the sample, it should be adjusted to give
was originally put forth as a Coblentz Class I spectrum.17,18 as broad a spot as possible without losing the signal due
In practice, this is still not achievable, but our process for to the Raman scattering. In the area of thermal effects, it
acquiring a spectrum should strive to eliminate instrumental is sometimes possible to trigger thermal backgrounds by
effects. This process will include reproducible sampling, focusing, since thermal effects are proportional to power
optimization of instrument acquisition, and elimination of density at the sample. As one goes through the focusing
the instrument response function. For both liquids and solid step described in the previous paragraph, the interferogram
powders, convenient sample holders include NMR tubes intensity may start to increase dramatically. It will also have
and melting point capillaries. Before using them, however, the appearance associated with a continuum source. If this
it is a good precaution to prescan them for fluorescence occurs the power should be reduced and a quick spectrum
background signals. Some glass samples contain transition obtained to look for evidence of thermal decomposition.
metal ions as impurities and these can be the source of The acquisition time will be determined by the S/N
significant interference even when exciting at 1.064 m. requirements. The linearity of S/N with the square root
Most commercial FT-Raman spectrometers have sample of the measurement time has been clearly established by
holders that allow for reproducible sample alignment. This several groups, so one can effectively estimate the required
becomes critical if spectral subtractions are to be carried out measurement time. Occasionally the Raman interferogram
since subtle changes in sample position can induce small can be corrupted by a spike in the data. This can occur
frequency shifts causing the subtraction process to fail. For when germanium detectors are used, since they have a large
this reason, sample focusing should not be done between area and are sensitive to cosmic rays. If the acquisition
measurements that are going to be subtracted. software cannot detect these events and reject the offending
Sample focusing itself is quite easy, assuming that suffi- scan, then the sinusoidal fringes will be seen in the
cient signal is present to see an interferogram. Weak Raman transformed spectrum. If co-addition has occurred, thereby
scatterers can be a problem, so in these cases a strongly incorporating the spike into the averaged interferogram,
scattering sample in an identical holder should be used then the spectrum must be discarded.
After acquisition of the Raman spectrum, it is necessary differences arise from the intensity units that are measured
to acquire a reference spectrum that can be used to cor- in each experiment. The CCD detector is a photon detector
rect for instrumental response effects that include detector with intensity units of photons s1 . The FT-Raman detec-
response, beamsplitter efficiency, and overall transmittance. tor is a power detector with intensity units of watts. Since
The total instrument response function can vary signifi- photons per watt is a function of wavelength, the relative
cantly in the spectral region from 1.0 to 1.6 m. The usual intensity ratios between the FT-Raman measurement and
reference source is a simple tungsten bulb or some other the CCD Raman measurement have a wavelength depen-
white-light source with a known spectral distribution in the dence. In order to compare the observed intensities across
NIR. The output is focused either on to the same sample, the two measurements, one must divide the FT-Raman
or a sample of a highly scattering, nonabsorbing powder intensities by the absolute wavelength. This produces rea-
such as KBr. Using a nonabsorbing powder, one generates sonable agreement between the two measurements.19
a reference spectrum that is a direct measure of the instru- The final step involves the detection of spectral artifacts.
ment response function. Using the sample itself will result An accompanying article (see Anomalies and Artifacts
in a reference spectrum that is the product of the instru- in Raman Spectroscopy) has discussed several potential
ment response function and any sample absorption bands spectral artifacts encountered in FT-Raman spectrometry.
that occur in this spectral range. This effect is most obvious The FT-Raman spectrum should be examined for the pos-
with aqueous solutions. Water has a strong absorption band sible presence of such artifacts.
at 1.3 m arising from the first overtone of the OH stretch. After this entire operation, one should have a spec-
This absorption will attenuate any Raman scattered intensity trum that approaches a Coblentz Class I spectrum. The
in this region, reducing the observed intensity. Unfortu- relative intensities should be transferable from instrument
nately, the correction for sample absorption obtained by to instrument. FT-Raman really does provide the capabil-
using the sample as a reference is not exact since the scatter- ity for routine acquisition of Raman spectra with minimal
ing volume and the absorption depth are not necessarily the interference from fluorescent impurities. There is certainly
same. For this reason, it is preferable to use a nonabsorbing sufficient sensitivity to acquire Raman spectra on almost
powder for the reference spectrum, and accept any relative all pure samples in very reasonable measurement times. In
intensity effects due to sample absorption. Fortunately, this short, we have a spectroscopic tool for routine sample char-
effect is minimal, except for aqueous solutions. There is a acterization that approaches and in some cases exceeds the
good example of this effect in the article on artifacts (see capability of routine IR spectroscopy.
Anomalies and Artifacts in Raman Spectroscopy).
Once the reference spectrum is acquired, it must be
converted to an instrument response function, Irf , using the ABBREVIATIONS AND ACRONYMS
following relationships. The observed white-light spectrum,
Wobs , is the product of the instrument response function and IGA Indium Gallium Arsenide
the true white-light spectrum, Wtrue . Nd : YAG Neodymium : Yttrium Aluminum Garnet
Nd : YLF Neodymium : Yttrium Lithium Fluoride
Wobs D Irf Wtrue
1 NIST National Institute for Standards
and Technology
Knowing the true white-light spectrum [a National Institute
for Standards and Technology (NIST) traceable source],
we can calculate Irf . Using these definitions it is sim-
REFERENCES
ple to correct the observed Raman spectrum to the true
Raman spectrum by simply dividing by Irf . Since vari-
1. C.G. Kontoyannis, N. Bouropoulos, H.H. Dauher, C. Bouro-
ous instruments will have different response functions due poulos and N.V. Vagenas, Appl. Spectrosc., 54(2), 225 (2000).
to differences in beamsplitters, detectors, and filters, it is
2. A. Gbureck, U. Gbureck, W. Keifer, U. Posset and R. Thull,
imperative that intensities be corrected. Appl. Spectrosc., 54(3), 390 (2000).
There is one aspect of the measured Raman intensities
3. K. Sano, M. Shinoyama, M. Ohgane, H. Higashiyama,
that must be recognized. When Raman spectra are measured M. Watari, M. Tomo, T. Ninomiya and Y. Ozaki, Appl.
with a CCD-based dispersive spectrometer, a similar instru- Spectrosc., 53(5), 551 (1999).
ment response correction is applied. Presumably, after both 4. C. Bauer, B. Amram, M. Agnely, D. Charmot, J. Sawatzki,
the FT-Raman spectrum and the CCD-Raman spectrum of N. Dupuy and J. Huvenne, Appl. Spectrosc., 54(4), 528
the same material have been corrected, the relative inten- (2000).
sities should be in close agreement. However, frequently 5. T. Hirschfeld and D.B. Chase, Appl. Spectrosc., 40(2), 133
one finds significant differences across the spectrum. These (1986).
6. G.W. Chantry, H.A. Gebbie and C. Helsum, Nature, 203, 13. A. Sakamoto, H. Okamoto and M. Tasumi, Appl. Spectrosc.,
1052 (1964). 52(1), 76 (1998).
7. R. McCreery and Y. Wang, Anal. Chem., 62, 2647 (1989). 14. B. Chase, J. Am. Chem. Soc., 108, 7485 (1986).
8. J. Pemberton and R. Sobocinski, J. Am. Chem. Soc., 111, 432 15. E.R. Huppi, Cryogenic Instrumentation, in Spectrometric
(1989). Techniques, ed. G.A. Vanasse, Academic Press, New York,
9. K. Asselin and B. Chase, Appl. Spectrosc., 48(6), 699 (1994). 155158 (1977).
10. D.J. Cutler and C.J. Petty, Spectrochim. Acta, 47A, 1159 16. B. Chase, Mikrochim. Acta, 3, 81 (1987).
(1991). 17. Coblentz Society Task Group, Anal. Chem., 38(9), 27A
11. A. Sakamoto, Y. Furakawa, M. Tasumi and K. Masutani, (1966).
Appl. Spectrosc., 47, 1457 (1993). 18. Coblentz Society Task Group, Anal. Chem., 47, 945a (1975).
12. K. Asselin and B. Chase, J. Mol. Struct., 347, 207 (1995). 19. R. McCreery, personal communication (2000).
Resonance Raman Spectroscopy
Jeanne L. McHale
1 INTRODUCTION The RR spectrum, like an ordinary Raman spectrum, is

recorded as a function of the frequency shift n D n0 ns
The resonance Raman (RR) effect comes into play when the between the incident (0) and scattered (s) radiation, where
frequency of the incident radiation falls within the electronic peaks are observed at the vibrational frequencies of the
absorption band.1 4 The resonance condition introduces molecule. Although the selection rules for resonance and
additional experimental considerations as well as changes off-resonance Raman differ, the peak frequencies in either
in the selection rules and depolarization ratios. Theoreti- case are those for the molecule in its ground electronic state.
cal approaches to the analysis of Raman intensities differ However, the intensities of the RR-active vibrational modes
from those which apply in nonresonance Raman. An impor- depend on the nature of the resonant excited electronic
tant aspect of resonance Raman spectroscopy (RRS) is the state. The Raman excitation profile (REP) is the variation
intensity enhancement of some normal modes, by factors in Raman intensity of one of these modes as a function
as large as 106 , compared to Raman spectra excited off- of excitation frequency. This profile can be analyzed to
resonance. When the exciting radiation is resonant with an determine the change in geometry of the molecule along
electronic transition, Raman spectra are dominated by totally the normal coordinate on going from the ground to the
symmetric vibrations which mimic the change in molecular excited electronic state. As illustrated in Figure 1, RRS
geometry that accompanies electronic excitation. In some is essentially a two-dimensional experiment, where the
cases, nontotally symmetric vibrations are also enhanced intensity is recorded as a function of frequency shift n (for
as a result of symmetry changes which accompany the fixed excitation frequency) to obtain the RR spectrum, and
electronic transition, or by means of vibronic coupling of as a function of excitation frequency n0 to obtain the REPs
excited electronic states. Resonance enhancement presents for the various modes (at fixed frequency shifts). REPs
the opportunity for increased sensitivity and selectivity in are closely connected to electronic absorption spectra and
the detection of dilute components of mixtures. Conven- provide additional mode-specific information which cannot
tional Raman instrumentation supplemented by a tunable be obtained from the absorption spectrum alone.
source of excitation can be employed in analytical and bio- Theoretical approaches to the interpretation of Raman
chemical applications in which the vibrational spectrum of intensities take advantage of the common underpinnings
a chromophore is enhanced relative to its nonabsorbing of Raman and absorption profiles, discussed in Section 3.
surroundings. Intensities in RRS reveal structural informa- The full power of these approaches is only realized when
tion about the excited electronic state. As will be shown, absolute Raman cross-sections, not just relative intensi-
RR enhancement also depends on the dynamics in the ties, can be measured. Compared to conventional Raman
excited electronic state. As a result RR spectra excited with measurements, this introduces additional experimental con-
continuous-wave sources can provide subpicosecond time- siderations to be discussed below. Depolarization ratios
domain information about the excited chromophore and the in RRS depend on the symmetry of the normal mode as
solvent motion which couples to the electronic transition. well as the degeneracy and number of resonant electronic
states. The range of values for depolarization ratios dif-
John Wiley & Sons Ltd, 2002. fers from that observed in off-resonance Raman and proper
Intensity
23 766
23 139
1
22 849 0 (cm )
22 284
700 900 1100 1300 1500 1700
(cm 1)

Figure 1. Raman spectra of 4-[2-(4-dimethylaminophenyl)ethenyl]-1-methylpyridinium iodide in water, at several excitation
frequencies.
conclusions about molecular symmetry require consider- that in favorable cases the excitation frequency can be
ation of the theoretical basis for polarization in Raman chosen to minimize interference from fluorescence emis-
scattering, examined in Section 3.2. sion. Electronic transitions typically span the ultraviolet
(UV), visible, and near-infrared (NIR) regions of the spec-
trum. The samples of interest will determine the nature
2 EXPERIMENTAL CONSIDERATIONS of the source required. A large number of applications
of RRS exploit the visible electronic transitions of con-
The instrumentation for RRS is not fundamentally differ-
jugated organic molecules and transition metal complexes,
ent to that for off-resonance Raman spectroscopy, but the
thus dye lasers and Tisapphire lasers are widely used.
frequency of the source is required to fall within the elec-
Colorless samples such as small aromatics and molecules
tronic absorption band. Additional sample-handling con-
lacking p p transitions require UV sources, whereas the
straints apply in order to minimize the deleterious effects
intervalence transitions of binuclear metalligand com-
of absorption, such as sample heating and photochemical
plexes may fall in the NIR. Tunable UV sources have been
decomposition. In addition, the intrinsic fluorescence of the
employed based on Raman shifting of the output of dye
sample may be unavoidable and require instrumental or
lasers pumped by doubled or tripled output of a Ndyttrium
chemical approaches to fluorescence rejection. Intensity cal-
aluminum garnet (YAG) laser.5 7 Excimer laser sources are
ibration is of special interest in the determination of REPs.
also employed.8 Laser sources in RR may be either con-
In order to determine the total Raman cross-section sR by
tinuous wave or pulsed, but use of the latter requires care
referencing to a standard, it is necessary to measure the
to avoid high peak powers which may lead to saturation,
depolarization ratios. Depolarization ratios are also mea-
photodecomposition, and other undesired nonlinear effects.9
sured in order to make conclusions about the symmetry of
Tunability imposes additional considerations when emp-
the RR-active mode and the nature of the resonant state.
loying single monochromators with array detectors.10 Such
an experimental arrangement requires notch filters to reject
2.1 Instrumentation the strong Rayleigh scattering at the laser frequency. If
unrejected, high levels of stray light obscure the Raman
The determination of REPs requires a source which can signal from lower frequency Raman lines. Each excitation
be tuned over the range of the absorption spectrum. Even frequency requires its own filter. This consideration is
the determination of a single RR spectrum requires tun- obviated when using polychromators with higher stray light
ability in order to achieve resonance with the absorption rejection. Choices for laser line rejection filters have been
band. Another advantage of having a tunable source is described.10
Resonance Raman Spectroscopy 3
2.2 Self-absorption the molar absorptivity at the scattered frequencies of the

two bands, the effective path distance is determined from
The appropriate laser power at the sample is an impor- a graph of (I1 /I2 ) versus concentration. Any Raman band
tant parameter in RRS because the incident radiation is can then be corrected for self-absorption using equation (1).
absorbed, leading to sample heating. Liquid solutions are Alternatively, the relative intensities of a number of sol-
typically circulated using flow cells, jets, or rotating sample vent lines can be compared to that observed in the neat
holders. The optimum flow rate depends on the laser power, solvent, and a frequency-dependent correction curve can
the sample concentration and the molar absorptivity. It is be determined. This empirical approach depends on having
a good idea to monitor the concentration of the sample a sufficient number of solvent Raman bands that span the
(usually by measuring the absorption spectrum) before and frequency range of interest without overlapping with the
after the Raman spectrum is determined, to ensure that no Raman bands of the chromophore.
sample degradation takes place. Even without decomposi-
tion, sample heating may perturb the measured spectrum by
decreasing the local concentration of the analyte in the irra- 2.3 Fluorescence
diated volume. The optimum concentration of the sample is
a compromise there must be enough of the chromophore Fluorescence emission is a challenging problem in many
present to observe its spectrum but, if the sample is too applications of Raman spectroscopy. The problem is more
dark, the incident and scattered radiation will be attenuated prevalent under resonance conditions, of course, as the radi-
too strongly. Back-scattering geometries permit darker sam- ation used to excite the Raman spectrum may also excite
ples to be employed, compared to 90 scattering, because the intrinsic fluorescence of the chromophore. The less seri-
the path distance through the sample is minimized. ous problem of fluorescent impurities is best handled by
Differential self-absorption of the scattered radiation sample purification, though the approaches discussed below
alters the relative intensities of the bands in a RR spectrum. may be applied whether or not the fluorescence is intrin-
Correction factors for this effect have been considered by a sic. The typical quantum yield for fluorescence emission is
number of authors.11 13 According to Beers law, the inten- several orders of magnitude larger than that for RR scat-
sity of the incident radiation I0 is attenuated exponentially tering so, depending on the frequency range of the emitted
as it traverses the sample: and scattered radiation, the fluorescence signal may com-
I D I0 exp[2.303enCx] 1 pletely swamp the Raman signal. As the source frequency
is varied, the frequency shifts n D n0 ns of the Raman
In equation (1), I is the intensity of the light after traveling bands remain constant, whereas the relaxed fluorescence
the distance x, en is the Beer law molar absorptivity at is observed at fixed absolute frequency ns . Relaxed flu-
the frequency n, and C is the concentration of the absorb- orescence is commonly observed for large molecules in
ing species. Equation (1) applies to the attenuation of the condensed phases, and leads to a diffuse emission spec-
laser radiation on its way to the sampled volume, as well as trum which is readily distinguished from the sharper Raman
to the scattered radiation as it emerges. Attenuation of the emission. The distinction between fluorescence and Raman
incident radiation affects all Raman lines the same way, but emission is less clear cut in the case of resonance fluores-
self-absorption of the scattered radiation alters the relative cence, which may be encountered in small molecules in gas
intensities in the Raman spectrum due to the dependence phase experiments.14 If the Stokes shift between the absorp-
of the molar absorptivity on the frequency of scattered tion and fluorescence maxima is sufficiently large, the RR
radiation (differential self-absorption). The effect of self- spectrum can be excited at wavelengths outside the range
absorption can be minimized by keeping the path lengths of the fluorescence spectrum. In favorable cases, the RR
small, but quantitative work requires that the effects of spectrum may be excited at wavelengths on the blue side
differential self-absorption be considered. Exact treatment of the absorption band where there is little overlap of the
of the self-absorption correction is quite complicated and absorption and fluorescence profiles.15 An example of this
requires consideration of the scattering geometry, intensity approach is shown in Figure 2, where the Raman spectrum
profile of the source, and the depth of focus of the collec- of the laser dye LDS 750 is observed along with the flu-
tion optics.11 A simple approach to the problem is based orescence. For very low fluorescence intensity, the Raman
on equation (1) and the approximation that there is a single bands appear on top of the broad fluorescence, which may
effective path length traversed by the scattered radiation. be subtracted by treating it as a background.
This effective path length can be determined by measuring Smaller fluorescence Stokes shifts and higher fluores-
the relative intensity of two Raman lines, I1 /I2 , as a func- cence quantum yields make measuring the RR spectrum
tion of the concentration of the absorbing species. Knowing very difficult. Although quenching agents have been used
Emission intensity
Absorption and emission intensity
3000 2500 2000 1500 1000 500 0

(cm 1)
max, fluor = 694 nm max, abs = 568 nm
488 nm
Excitation
10 000 12 000 14 000 16 000 18 000 20 000 22 000 24 000 26 000

Frequency / cm1
Figure 2. Absorption and fluorescence plus RR emission spectra of LDS 750. The expanded view in the inset shows the Raman peaks
on top of the fluorescence.
to suppress the fluorescence signal,16,17 this approach has instrumentation requirements incorporate signal processing
the potential to perturb the Raman spectrum of interest. A tricks, such as Fourier transformation of the spectrum and
number of fluorescence-rejection techniques take advantage subtraction of the low-frequency features.24,25
of the longer timescale for fluorescence, which for typi-
cal strong emitters is in the nanosecond regime, compared
to the virtually instantaneous Raman scattering. Pulsed 2.4 Absolute and relative intensities
sources and gated detection enable the bulk of the fluo-
rescence spectrum to be detected separately and then sub- The full power of RRS is only realized when intensities
tracted from the total spectrum.18 21 A related method is the are correctly determined. For quantitative work at a sin-
use of intensity modulation of the laser source and phase- gle excitation frequency, it is often necessary to determine
sensitive detection of the Raman scattering with a lock-in intensities relative to internal or external standards, or per-
amplifier.22 Techniques such as these become more chal- haps the relative intensities of the various Raman bands
lenging when the fluorescence lifetime is short and there of one component of the sample. This requires that the
is more temporal overlap of the fluorescence and Raman measured spectrum be corrected for the variation in the
emission. There also exist approaches which are based on efficiency of the instrument (collection optics, monochro-
the much weaker wavelength dependence of the fluores- mator, and detector) as a function of the frequency ns of the
cence intensity, compared to the Raman signal. Shreve scattered light. This instrument response can be measured
et al.23 have developed the method of shifted excitation using a standard lamp,26,27 or by measuring the emission or
Raman difference spectroscopy (SERDS), in which two Raman spectrum of a material for which the true relative
spectra are recorded at slightly different excitation frequen- intensities are known. The rotational Raman lines of D2 and
cies and then subtracted. The diffuse fluorescence spectrum the emission spectra of various fluorescent dyes have been
is effectively subtracted and the Raman signal shows up used for this purpose.28 30 The problem of determining the
as a difference spectrum. The SERDS technique requires a instrument response is a little more involved when using a
continuously tunable laser and the high signal-to-noise ratio single monochromator system, typically with a CCD detec-
and fast acquisition time afforded by charge coupled device tor, as laser line rejection filters are required to minimize
(CCD) detection. Related approaches having less stringent stray light. If data are to be obtained at various excitation
wavelengths, each filter will have its own response function direction, and polarized in either the Z direction (polarized
and thus the instrument response must be determined for spectrum) or X direction (depolarized spectrum). The depo-
each excitation frequency n0 . McCreery and colleagues larization ratio is defined as the ratio of the intensities for
have made available data for emission standards conve- the depolarized and polarized spectra:
nient to excitation wavelengths in the red (Kopp glass) and
Idep ds/d
?
the blue (coumarin dye) regions of the spectrum.28,31 rD D 3
Analysis of excitation profiles requires not just relative Ipol ds/d
jj
intensities, but absolute cross-sections for Raman scattering For other scattering geometries, the measured value of r
as a function of excitation frequency. The cross-section depends on the angle q between the direction in which the
sR determines the total intensity of light scattered into scattered light propagates and the direction in which the
a sphere of 4p steradians (sr) surrounding the sample. incident radiation is polarized:
The measured intensities depend on the solid angle
r
subtended by the detector, and are therefore proportional rq D 4
to the differential cross-section ds/d
, which has units of 1 1 r cos2 q
cm2 sr1 . The measured intensity Is , in photons per second, where r is the depolarization ratio for 90 scattering. From
is given by equation (4), it is apparent that 90 scattering and back-
ds scattering (180 ) geometries yield the same depolarization
Is D I0
N 2
d
ratio, as both have q equal to 90 . Accurate determination
where I0 is the incident intensity of the source in photons of r requires some means to randomize (or scramble) the
per second per square centimeter,
is the solid angle of polarization of the scattered light after it passes through the
scattered light seen by the detector, and N is the number of analyzer. Sources of experimental error in the determination
scatterers in the sampled volume, which is the intersection of r have been described.32,33
of the irradiated volume with that seen by the detector. The cross-section for total scattering is related to the
The conversion between sR and ds/d
requires that the differential cross-section as follows:
depolarization ratio r be known. The depolarization ratio is 8p1 C 2r ds
sR D 5
measured by exciting the Raman spectrum with polarized 31 C r d
incident radiation, and separately detecting the scattered

The differential cross-section ds/d
in equation (5) is the
radiation having polarization vector parallel and perpen-
sum of the contributions for parallel and perpendicular
dicular to the incident polarization. As discussed in the
scattering:
next section, depolarization ratios in RRS provide useful

information about the enhancement mechanism. The depo- ds ds ds ds
D C D 1 C r 6
larization ratio is frequently measured using the conven- d
d
? d
jj d
jj
tional 90 scattering arrangement, depicted schematically in
Figure 3. In this figure, the incident radiation is considered If an internal standard of known differential cross-section
to propagate along the X direction with electric field vec- is employed, the differential cross-section of the analyte can
tor oscillating in the Z direction. A polarization analyzer is be determined from the measured relative intensities:
used to detect scattered radiation, propagating along the Y I1 C1 ds1 /d
D 7
I2 C2 ds2 /d
Z
where Ci is the molar concentration of species i and dsi /d
its differential cross-section for a particular Raman band.

The intensities in the above equation should be determined
from the area under the band rather than the peak intensity.
If an external standard is used, it should be taken into
Source account that the collection angle
and thus the measured
e^ i radiation
e^s intensity is inversely proportional to the square of the
X
refractive index n:34 37
I1 n2 C1 ds1 /d
D 22 8
Polarization I2 n1 C2 ds2 /d
Y
analyser
Raman cross-sections have been reported for a num-
Figure 3. Ninety degree scattering geometry. ber of convenient intensity standards.38 42 Trulson and
Mathies38 used an integrating cavity to measure sR for is some experimental evidence that the local field factor
Raman modes of benzene, cyclohexane, and cacodylate of equation (11) does an adequate job of accounting for
ion. Asher and colleagues39,40 have reported Raman cross- variation in Raman intensities with solvent.47,48 However,
sections for selected modes of acetonitrile, benzene, and the molecular details of solvent dynamics, not accounted for
aqueous standards sulfate, nitrate, and perchlorate ions. Li by continuum theory, have been shown to have a strong
and Myers41 have measured absolute intensities of acetoni- effect on the Raman cross-section in some systems.50 53
trile, cyclohexane, and water at UV wavelengths. Measure- Mode-dependent local field corrections may also influence
ments using UV/RR of gases often employ methane as an Raman intensities.54 These correction factors depend on the
internal standard.7,43 normal-coordinate dependence of the dipole moment dif-
Raman cross-sections are proportional to the square of ference between the ground and excited states, a quantity
the transition polarizability, discussed in Section 3, and to not generally available from experiment. The nature of the
the fourth power of the incident frequency. Early reports of resonant transition may also vary with solvent, causing a
absolute Raman intensities of transparent standards consid- perturbation to the RR intensity50,51,55 that is not captured
ered the transition polarizability (defined below in equa- by equation (11). Unless there is a need to compare gas
tion 13) to be independent of frequency.44,45 It is now phase and solution phase Raman intensities, or the disper-
recognized that even for colorless materials, preresonance sion in the refractive index is significant, it is preferable to
enhancement via electronic states well into the UV causes consider the uncertain local field factor to be an integral
the frequency dependence of the cross-section to deviate part of the measured cross-section.
from the trivial n4 factor. When employing photon counting
detection, this n4 factor should be replaced by n0 n3s . The fre- 3 THEORY OF RESONANCE RAMAN
quency dependence of the Raman cross-section for intensity
standards is often well characterized by the equation46 INTENSITY
sR n0 n2e C n20 2 The Raman cross-section for the transition i ! f depends
D K 9 on the quantum mechanical transition polarizability aif
n0 n3s n2e n20 4
connecting the initial i and final f states:
Frequency-dependent data for total cross-sections have been
fit to equation (9) using the scaling factor K and the origin dsif 16p4 n3s n0
D jOes aif eO 0 j2 12
of the resonant transition ne as adjustable parameters.38,41 d
s,0 c4
This enables published cross-section data to be applied at In the above equation, cgsesu units are employed, and eO 0
any excitation wavelength. and eO s are unit vectors in the direction of the polarization
Comparison of Raman intensities in the liquid phase to of the incident and scattered radiation, respectively. These
those in the gas requires the local field factor to be taken unit vectors project out the relevant components of the tran-
into account.35,37,47,48 There is no rigorous way to make sition polarizability aif . This transition polarizability is a
this correction, because solution-phase Raman intensities second-rank tensor which is defined in the laboratory frame
may depend on specific intermolecular interactions in addi- in equation (12). The usual starting point for discussing
tion to bulk dielectric properties. Treating the solution as the polarizability tensor, however, is the molecule frame,
a dielectric continuum49 leads to the prediction that mea- as this frame permits the molecular symmetry to be taken
sured Raman intensities are proportional to the local field into account. In Section 3.2, the projection of the molecule
factor L, frame polarizability onto the laboratory coordinate system
2 4
n C2 will be considered, in order to predict the depolarization
LD 10
3 ratio. In the present section, theories for the transition polar-
izability are given, and it will be convenient to consider this
In strongly absorbing solutions, there is dispersion in the
quantity in the molecule frame of reference. Upper case
refractive index, and a more accurate local field factor is
(XYZ ) and lower case (xyz ) letters denote directions in the
given by 3 laboratory and molecule frame, respectively.
n20 C 2 n2s C 2
LD 11
3 3
3.1 Sum-over-states expression for the
where n0 is the refractive index at the incident frequency polarizability
and ns that at the scattered frequency. When Raman data are
collected over a wide range of excitation frequencies, dis- The transition polarizability is expressed by the Kramers
persion in the local field factor can be significant.35 There HeisenbergDirac (KHD) equation as follows:56

1 hijmr jnihnjms jfi hijms jnihnjmr jfi state e:
ars if D C
h n n0 nni C in n0 nnf C in 1 hgv0 jmr jevihevjms jgv00 i
13 ars if D 14
h v n0 nev,gv0 C ie
The subscripts i, n, and f represent initial, intermedi-
ate, and final vibronic states, respectively, and the sum is Next the BornOppenheimer approximation is made, per-
over all intermediate states. Note that the states jii and mitting the wavefunction for a vibronic state such as jevi
jfi are typically vibrational levels within the ground elec- to be factored into vibrational and electronic parts: jevi D
tronic state, and the sum over n, in principle, runs over jeijvi. The electronic transition moment mge D hgjmjei,
all excited electronic states and all the vibrational sublevels which is a function of the normal coordinates Qi , is
within those states. The terms mr and ms are components of also expanded in a Taylor series about the equilibrium
the dipole moment operator, where the subscripts r and s geometry:

denote Cartesian directions (x, y, z) in the molecule frame. 3N6
The numerator of each term in the KHD equation is related @mge
mge Q D m0ge C Qi C 15
to the intensity of the transitions i ! n and n ! f, through iD1
@Qi 0
the transition dipole moments, where, for example, hijmr jni
is the component of the transition dipole moment along The first term, m0ge , is the transition moment evaluated at the
equilibrium geometry of the molecule, and the second term
the r direction, connecting states i and n. The intensity in
takes into account the dependence of the transition moment
absorption or emission is proportional to the square of this
on the normal coordinate Qi . As will now be shown, if
transition moment, and the RR intensity is proportional to
equation (15) is truncated after the first term (the Condon
the fourth power of this quantity. Thus strong enhancement
approximation), only totally symmetric vibrations are per-
is expected for resonance with strong electronic transitions,
mitted. The part of the transition moment which depends
but this is not the only factor which determines the RR
on the normal coordinate, however, permits nontotally sym-
cross-section, as will be shown. The term n0 is the inci-
metric vibrations to be active in RR. These two terms are
dent frequency, and nni D En Ei /h is the frequency of referred to as the Albrecht A and B terms, respectively, and
the transition from state i to n. The quantity n is the the transition polarizability is expressed as their sum:
homogeneous linewidth of the transition to state n. Equa-
tion (13) applies strictly only to isolated molecules, where ars v0 v00 D Av0 v00 C Bv0 v00 16
the linewidth is limited by the rate of population relax-
ation. In solution, the linewidth comprises that due to pure
3.1.1 Totally symmetric modes
dephasing (solvent-induced fluctuations in the transition fre-
quency) as well as population relaxation. In principle, The Condon approximation leads to the A term
depends on the vibronic sublevels attached to states i and n,
1 0 hv0 jvihvjv00 i
in the first term, and to states f and n, in the second term, of Av0 v00 D mge r m0ge s 17
equation (13). In practice, is a phenomenological damp- h v nev,gv0 n0 ie
ing constant considered independent of vibrational sublevel. In equation (17), hnev,gv0 is the energy of the transition
Increasing damping, whether via dephasing or population between the intermediate jevi and initial jgv0 i vibronic
relaxation, serves to decrease the Raman intensity. states. Thus resonance is achieved by tuning the incident
When the excitation frequency is far from resonance with frequency n0 to approach the frequency of any vibronic
an allowed transition, an indefinite number of intermediate transition within the electronic absorption spectrum. The
states contribute to the sum, and we resort to referring to numerator of equation (17) contains products of vibrational
an intermediate state as a superposition of stationary states, overlaps, such as hvjv00 i, which is the overlap of the vibra-
that is, a virtual level. However, in RRS one chooses n0 to tional state v within the excited electronic state manifold,
be close to a transition frequency nni . The first term in the and the vibrational state v00 belonging to the ground elec-
KHD equation then becomes large, and the smaller second tronic state. The squares of these same vibrational overlaps,
term is dropped. The sum-over-states in equation (13) then also known as FranckCondon (FC) factors, determine the
runs over the allowed vibrational sublevels of a given intensities of vibrational sub-bands contributing to vibra-
electronic transition. Taking the initial state to be the tional progressions in electronic absorption and emission
vibrational level v0 within the ground electronic state, jgv0 i, spectra. These are also referred to as FC progressions. The
and the final state to be some higher vibrational level overlap integral is a very strong function of the change
jgv00 i, also within the ground electronic state, the sum now in the equilibrium geometry on going from the ground to
runs over vibrational levels v within the excited electronic the excited electronic state. To illustrate this, consider a
ground vibrational state. It turns out that the geometry

change is much more effective than a frequency change
in contributing to an FC progression, thus the occurrence
Absorption energy
e of an FC progression is generally associated with a change
07 in equilibrium bond distance.
06
05 =5 To consider polyatomic molecules, the preceding dis-
04 =4
cussion can be cast in terms of normal coordinates Qi
03 =3
rather than the bond distance R. The vibrational modes
02 =2
01 =1 that mimic the geometry change of the molecule con-
00 =0 tribute to the FC progression. Put another way, a vibrational
g progression is expected for a mode if the excited state
Absorption
intensity potential energy is displaced relative to the ground state
along that normal coordinate. If this is not the case, then
= 4 Av0 v00 vanishes unless v0 D v00 . We reach the conclusion
= 3
that the A-term part of the Raman intensity is nonzero
= 2
V (R )
only for normal modes that are displaced in the excited

= 1
= 0 electronic state. Unless the molecule undergoes a change
in symmetry in the excited electronic state, the A-term
enhancement mechanism permits only totally symmetric
Re Re + R modes because nontotally symmetric modes cannot be dis-
Figure 4. Displaced potential energy curves leading to an FC placed in the electronic state. Figure 5 illustrates this by
progression. comparing the ground and excited potential energy surfaces
for a totally symmetric mode to those for a nontotally sym-
diatomic molecule for which the equilibrium bond distance metric mode.
goes from the value Re in the ground state to Re C R in the Figure 5(a) depicts the surfaces for a symmetric stretch,
excited state, as shown in Figure 4. The most intense band for example the ring-breathing vibration of benzene. In the
in the vibrational progression is that for which the overlap p p excited state of benzene the CC bond distances
integral is largest. This is the so-called vertical transition increase by a few hundredths of an Angstrom, preserving
corresponding to a fixed internuclear distance, as shown by the sixfold symmetry of the ground electronic state. The
the vertical arrow in Figure 4. The minimum energy or 00 excited state potential is thus displaced along the normal
transition, at E00 , is the one connecting the ground vibra- coordinate for the ring-breathing mode, and the vibration
tional levels of the ground and excited electronic states. is FC active. However, a nontotally symmetric vibration
If the displacement is large there is little overlap of the (Figure 5b), such as an out-of-plane vibration or asym-
ground vibrational wavefunctions and thus little intensity metric CC stretch, cannot be displaced in the excited
at this energy. The internuclear distance in the initial state state because the equilibrium geometry is at the symmet-
is distributed over a range of values according to the wave- ric configuration in both the ground and excited electronic
function for the ground vibrational levels. The breadth of states. Not all totally symmetric vibrations are displaced.
the absorption spectrum reflects the slope of the upper For example, CH bond distances in aromatic hydrocarbons
potential energy surface in the vicinity of R D Re , known are not changed much in p p excited electronic states, so
as the FC geometry.
Figure 4 illustrates the reflection principle, in which the
absorption profile is sketched over a range of energies
spanned by the slope of the upper potential surface across e e
the zero-point motion of the initial (ground) vibrational
V (Qa)
V (Q s)
state. The greater the value of R, the larger the range of g

transition energies between E00 and the vertical transition. g
If the ground and excited state vibrational frequencies are
the same, and there is no change in bond distance, then
all the overlap integrals are zero except those connecting
states with the same vibrational quantum number. All the (a) Q0 Q0 + Q (b) Q0
absorption intensity is then concentrated in the v0 D 0 ! Figure 5. Ground and excited state potentials along (a) a totally
v D 0 transition, assuming the absorption originates in the symmetric and (b) a nontotally symmetric normal coordinate.
vi
little resonance enhancement of symmetric CH stretches 1 i 2
is observed when the excitation frequency is resonant with h0i jvi i D p exp i 20
vi ! 2 4
the p p transition.
The vibrational overlaps in equation (17) are multimodal where i is the displacement of normal mode i in dimen-
quantities which depend on all the displaced vibrational sionless normal coordinates:
modes. In the harmonic approximation, the 3N 6 dimen- 2 1/2
4p ni
sional vibrational wavefunction is factored into a product of i D Qi 21
h
one-dimensional harmonic oscillator wavefunctions. Con-
sider the case of a fundamental transition involving normal where Qi is the shift in the equilibrium mass-weighted
mode a. The initial state jv0 i is a product of these one- normal coordinate between the ground and excited elec-
dimensional wavefunctions, all in the ground vibrational tronic states, and ni is the frequency of the mode. When
state. The final state jv00 i is a product in which the wave- Qi is zero, equation (20) gives zero values for all over-
function for the a mode has one quantum of excitation and laps except h0j0i, as discussed previously. Note that the
all other modes remain in the ground vibrational state. The Raman intensity of any particular mode also depends on
sum over intermediate states is then a sum over all possi- the displacements of all other Raman active modes.
ble combinations of vibrational quantum numbers, v D v1 , Resonance excitation can lead to the appearance of over-
v2 , v3 , . . . v3N6 in the resonant electronic state. The FC tones which are not observed off-resonance. The basis for
overlaps are then factored as follows: this is revealed by equation (17). Provided that the FC fac-
tors are significant, there is no selection rule requiring that
h0jvi D h01 jv1 ih02 jv2 i . . . h0a jva i . . . h03N6 jv3N6 i 18 v0 D v00 1. However, high-order overtones are generally
only observed in the spectra of small molecules. See for
and example the spectrum of CH3 I reproduced in Figure 6.57
hvj1i D hv1 j01 ihv2 j02 i . . . hva j1a i . . . hv3N6 j03N6 i 19 In larger molecules with many displaced modes, overtones
are less likely to be observed. Although this statement is
For harmonic modes having no frequency change in the perhaps more readily appreciated within the formalism to
excited electronic state, the one-dimensional overlap is be discussed in Section 4.2, equations (18)(20) hint at
36
31
32
37
33 1t
I 38
34 35
310
2
39
21
270 280 nm 290 300
310
39 311 312 314
313 315 316 317
318
2137 2138
310 320 nm 330 340
319
320
321 322 324
323 325
4
326
2.5
327 328 329
350 360 370 nm 380 390 400
Figure 6. RR spectrum of CH3 I excited at 266 nm, showing the fundamental and 28 overtones of the CI stretch, nQ 3 500 cm1 .
[Reprinted with permission from Imre et al., Journal of Physical Chemistry, 88, 3956 (1984), copyright 1984, American Chemical
Society.]
why this is so. Vibrational progressions merely redistribute tensor are nonzero. Equation (22) then permits activity of
the total absorption intensity, which is constant. When a nontotally symmetric modes which would not be allowed
large number of normal modes are displaced, the intensity by the A term because they are undisplaced in the excited
invested in any one mode is less, and overtone progressions electronic state. Consider such a transition starting from
are less readily observed. the v0 D 0 level of the ground electronic state. The overlap
hv0 jvi D h0jvi in the first term of the numerator of equa-
tion (22) is nonzero for the v D 0 intermediate state. The
3.1.2 Nontotally symmetric modes
quantity hvjQi jv00 i D h0jQi jv00 i vanishes unless v00 D 1. The
Nontotally symmetric vibrations can be active in RR by result is enhancement of the fundamental transition 0 ! 1
virtue of the second term in equation (15), from which the when the radiation is resonant with the g0 ! e0 transition.
Albrecht B term is derived: Similarly, the second term in equation (22) results in activ-
ity of the fundamental when the radiation is resonant with
m0ge r m0gr s hv0 jvihvjQi jv00 i
the g0 ! e1 transition.
1 Hier 0 0 0
Cmge s mgr r hv jQi jvihvjv i
00
Bv0 v00 D
h v,r Er Ee !ev,gv0 !0 ie
3.2 Polarization in resonance Raman spectra
22
Equation (22) applies to the v0 ! v00 transition of mode In off-resonance Raman spectroscopy, the form of the
i. B-term intensity is a result of vibronic coupling between polarizability tensor reflects the symmetry of the molecule
two allowed excited electronic states e and r. This coupling as well as the symmetry of the normal mode. In RR
is a consequence of a breakdown in the BornOppenheimer spectroscopy, however, the symmetry properties of the
approximation, which occurs when the matrix element Hier resonant states determine which elements of this tensor are
is nonzero: nonzero. We begin by considering the case of off-resonance

@H O Raman and then specialize to the case of RR.
i
Her e 23 The polarizability tensor for any totally symmetric vibra-
@Qi r
0 tion is diagonal:
This matrix element is also responsible for intensity borrow-
a 0 0
ing in electronic spectra resulting from vibronic coupling a D 0 b 0 totally symmetric vibration 24
of excited electronic states.58 Such intensity borrowing, and 0 0 c
also the magnitude of the B term, increases when the cou-
pled states are close in energy, as indicated by the energy In a spherically symmetric molecule, such as CCl4 or SF6 ,
denominator in equation (22). the three diagonal components of a are equal. Cylindri-
The B-term part of the transition polarizability accounts cally symmetric molecules, such as CH3 CN, C6 H6 , and all
for RR activity of vibrational modes which couple excited linear molecules, have two equal diagonal components in
electronic states, and symmetry arguments can be employed equation (24). Less symmetric molecules have a 6D b 6D c.
to predict which vibrations are capable of connecting a The polarizability for a nontotally symmetric mode may
given pair of BornOppenheimer electronic states. If e , have off-diagonal elements, and the sum of the diagonal
i , and r denote the irreducible representations to which elements (the trace) is zero. Group theoretical arguments
electronic state e, mode i, and state r correspond, respec- can be employed to deduce the form of the Raman tensor
tively, then in order for the matrix element Hier to be for nontotally symmetric modes.59
nonzero, the triple direct product e i r must equal The Raman intensity is observed in a laboratory coor-
or contain the totally symmetric representation. In the com- dinate system. In the typical 90 scattering geometry
mon case where the ground state is totally symmetric, an depicted in Figure 3, the polarized and depolarized inten-
allowed electronic state must have the same symmetry as sities depend on the relevant laboratory-frame components
one or more of the Cartesian coordinates. For example, if of the polarizability as follows:
g ! e is polarized in the x direction and g ! r is polar-
ized in the y direction, then normal modes which transform Ipol / jaZZ j2 25
according to the same representation as xy are capable of Idep / jaZX j2 26
vibronically coupling the two electronic states e and r. Thus
vibronic coupling of two electronic states of different sym- The problem is to express the laboratory-frame compo-
metry results in RR enhancement of nontotally symmetric nents of the polarizability, aZZ and aZX , in terms of the
modes, for which off-diagonal elements of the polarizability molecular-frame components axx , axy , etc.60,61 This is done
with the help of three rotational invariants of the polariz- be one-third for a totally symmetric mode enhanced via a
ability J , for J D 0, 1, 2. (Note that the polarizability is nondegenerate electronic transition. If the resonant excited
a second-rank tensor having three rotational invariants. An electronic state is doubly degenerate, then two equal diago-
example of a first-rank tensor having a single rotational nal components, for example axx D ayy , are nonvanishing.
invariant is the dipole moment , for which the combina- This results in the prediction that r equals one-eighth. When
tion m2x C m2y C m2z D m2 is invariant, i.e. it is independent two differently polarized (but nondegenerate) electronic
of reference frame.) These rotational invariants are com- transitions contribute to resonant enhancement, then two
binations of tensor elements which are independent of the diagonal elements of the polarizability tensor are nonzero
reference frame. They are: and have different frequency dependence. The depolariza-
tion ratio is then a function of excitation frequency and
0 D 13 jaxx C ayy C azz j2 deviates from the value of one-third. Frequency-dependent
1 D 12 [jaxy ayx j2 Cjaxz azx j2 Cjayz azy j2 ] depolarization ratios have been used to deduce solvent-
induced symmetry breaking62 and overlapping electronic
2 D 12 [jaxy C ayx j2 Cjaxz C azx j2 Cjayz C azy j2 ] transitions.55,63
C 13 [jaxx ayy j2 Cjaxx azz j2 Cjayy azz j2 ] 27 B-term enhancement of nontotally symmetric modes
results from vibronic coupling of excited electronic states
0 is called the isotropic part of the polarizability, pro- belonging to different symmetry species of the molecular
portional to the square of the trace of the tensor. It van- point group. Far from resonance, equation (22) predicts that
ishes for nontotally symmetric vibrations. 1 , the antisym- axy and ayx are equal, but close to resonance they may dif-
metric anisotropy, is zero in most cases, as the Raman fer. When axy 6D ayx , 1 is nonzero and consequently, in
tensor is generally symmetric. Exceptions arise in the contrast to off-resonance Raman, the depolarization ratio
case of B-term enhancement, which can lead to anoma- can exceed 0.75. When the excitation is midway between
lously polarized Raman lines as discussed below. The the 00 and 01 resonances, then axy D ayx . The result is
2 term is called the symmetric anisotropy. In general anomalous polarization 2 vanishes and the depolariza-
it can be nonzero for both totally and nontotally sym- tion ratio becomes infinite. When this happens, the Raman
metric vibrations. In the case of totally symmetric vibra- band is observed in the depolarized but not the polarized
tions of spherically symmetric molecules, however, 2 is spectrum.64,65
zero.
The laboratory-frame components required in equa-
tions (25) and (26) are found as follows:60 4 ANALYSIS OF RAMAN EXCITATION
1 0 2 2 PROFILES
jaZZ j2 D C
3 15
1 1 2 The analysis of REPs provides information about the geom-
jaZX j2 D 1 C 28 etry of a molecule in the excited electronic state. In
6 10
principle, the vibrational structure of the electronic absorp-
The depolarization ratio is thus tion band can also reveal this information, but there are
Idep 51 C 32 complications which make such analysis impractical. The
rD D 29 vibrational structure is not always resolved in absorption
Ipol 100 C 42
spectra of large molecules, due to the large number of FC-
In off-resonance Raman, 1 is zero, and equation (29) gives active transitions and to solvent-induced broadening. The
r 0.75. The value of 0.75 is observed for nontotally individual vibrations coupled to the electronic transition
symmetric vibrations and for totally symmetric vibrations are readily observed in the RR spectrum, however, and by
r is predicted to be less than 0.75. employing different excitation frequencies the variation in
Next consider the RR spectrum, and totally symmetric intensity for each mode can be analyzed to obtain the mag-
vibrations allowed by the A-term enhancement mechanism. nitude (but not the sign) of the normal mode displacements.
The form of the polarizability tensor depends on the direc- RRS enables the influence of solvent spectral broaden-
tion of the transition moment and the degeneracy of the ing to be distinguished from broadening due to unresolved
resonant state. If the excited state is nondegenerate, then a vibrational progressions. Analysis of RR intensities permits
single diagonal element of a is nonzero. For example, if solvent reorganization energies, the energy lost by the sol-
the transition is polarized in the x direction in the molecule vent molecules on adapting to the charge distribution of the
frame, then all elements except axx are zero. Employing chromophore in its excited electronic state, to be separated
equations (27) and (29), the depolarization ratio is found to from internal reorganization energy, which is the energy
lost by the chromophore on relaxing from the FC geom- where NA is Avogadros number, e is in liters per mole per
etry to the equilibrium geometry of the excited electronic centimeter (L mol1 cm1 ), and sA has units of centimeter
state. squared per molecule. A!0 is found from an integral over
Two approaches have been widely used to analyze RR the absorption band:
intensity. The transform theory approach66 73 takes advan- 1
tage of the information common to both the absorption s A ! sA !0
A!0 D P d! C ip 31
and the RR profile. The experimental absorption spec- 0 ! ! !0 !0
trum is used to predict the shape of the REP, which is
then scaled to match experiment by adjusting the value Here, !0 D 2pn0 is the excitation frequency, and P indi-
of the squared displacement 2a . In the time-dependent cates the principal part of the integral. The above transform
approach,74 78 however, the absorption and Raman pro- is also calculated at a frequency !0 !a , where !a is the
files are simultaneously modeled using a common set of vibrational frequency in angular units. The Raman cross-
parameters, including the mode displacements. Both meth- section for the fundamental transition of mode a is predicted
ods are based on a set of standard assumptions which are from the difference in these two transforms:
not always realized in practice, as will be shown. The work-
!0 !3s n2 2
ing equations presented here are applicable only to totally s R !0 D jA!0 A!0 !a j2 32
symmetric vibrations. 8c2 p3 a
Both transform theory and the time-dependent theory Here !s D !0 !a is the angular frequency of the scat-
are equivalent to the sum-over-states calculation of the A- tered radiation, and a is the dimensionless displacement,
term contribution to the Raman intensity. However, these defined previously in equation (21).
approaches bypass the need to sum over a large number
Equation (32) suggests that resonance enhancement is
of intermediate vibrational states on the upper potential
stronger for molecules having structured, rather than dif-
surface. In the transform theory, advantage is taken of the
fuse, absorption profiles, because the Raman cross-section
fact that the absorption spectrum inherently contains the
depends on the difference in the absorption spectrum for
relevant information concerning these states. In the time-
frequency shifts equal to the vibrational frequency. To
dependent theory, details of the excited state potential are
some extent, this is a disappointing conclusion, because
taken as parameters and adjusted to obtain agreement with
molecules with resolved vibrational structure are likely to
the experimental absorption profile and the REPs for all
have smaller Stokes shifts between the absorption and flu-
RR-active modes. In practice, the two approaches do not
orescence. Although the resonance enhancement may be
always yield the same results, perhaps due to a failure of
strong, the fluorescence interference problem may also be
one or more of the standard assumptions on which they are
based. greater. Molecules having diffuse absorption spectra have
a larger amount of internal and solvent reorganization, and
thus larger Stokes shifts, but less resonance enhancement.
4.1 Transform theory The transform theory also leads to the prediction that,
for similar displacements, lower-frequency vibrations have
Transform theory is based on the relationship between smaller Raman cross-sections than high-frequency vibra-
the real and imaginary parts of the polarizability, which tions. Thus there may be low-frequency vibrations such as
relate to scattering and absorption, respectively. For the torsions which contribute to the excited state reorganization
working equations presented here to be valid, it is required but are difficult to observe in RR.
that the excitation frequency be resonant with a single Equation (32) is based on the Condon approximation.
electronic state, and that the vibrational modes be harmonic When it is valid, the calculated and experimental REPs
and have the same frequency in the ground and excited have the same shape and maximize at the same excita-
electronic states. Inhomogeneous broadening is considered tion frequency. The dimensionless displacement can then
absent, although it can be taken into account as described be determined by scaling each calculated profile to match
elsewhere.67 the experiment. When the Condon approximation fails,
The absorption cross-section, sA , is taken from experi- equation (32) can be modified to account for the linear
ment and the appropriate transform A! is calculated. The dependence of the transition moment on the normal coor-
absorption cross-section is related to the Beer law molar dinate:
absorptivity e as follows:
[A!0 A!0 !a ] DDD) [1 C Ca A!0
1000 ln 10e
sA D 30 1 Ca A!0 !a ] 33
NA
The non-Condon coefficient Ca is defined by Wave packet evolves with time

e
p
2@mge /@Qa 0 Qa
Ca D 34
m0ge a
where Qa is the root-mean-square displacement of the

normal mode a.
The above equations are applicable to fundamentals.
Transform theory expressions for overtones and combina-
tions are discussed elsewhere.69,71 Extensions of the theory
to consider anharmonicity,73 thermal effects, and inhomo-
g
geneous broadening67 have also been presented.
4.2 Time-dependent theory
Time-dependent theory, developed by Heller and collea- |f> v =1

gues,74 78 relates the absorption and REPs to dynamics
taking place on the excited state potential surface. In addi- |i> v =0
tion to providing a practical means to calculate spectra, the
approach yields valuable physical insight into the molec-
Figure 7. Wavepacket dynamics on a displaced excited state
ular basis for observed intensities and lineshapes, and potential surface.
takes advantage of classical concepts to describe the time-
evolution of wavepackets, or wavefunctions, on excited t1
state potential surfaces. Figures 7 and 8 illustrate the basic 1
principles. We imagine a system which starts out in ini- t2
tial state jii, shown in Figure 7 as the ground vibrational t3
state within the ground electronic state. Absorption of the
|<i |i (t )>|
incident photon promotes the system to the upper potential

surface. At time t D 0 the initial wavefunction looks just
like the ground state wavefunction jii, but since the upper
state potential surface is displaced, jii is not an eigenfunc-
tion of this potential and it evolves in time. The overlap
of this time-dependent wavefunction jiti with the ini- (a) 0 100 200 300
tial wavefunction, hijiti, evolves on a timescale dictated
by the slope of the upper potential and the excited state 1
vibrational frequency, as shown in Figure 8. The absorp-

tion cross-section sA is found from the Fourier transform
|<f |i (t )>|
(FT) of the damped time-dependent overlap:

2pm0ge 2 !0 1
s A !0 D hijiti exp[i!i
3hcn 1
C !0 t] expjtj dt 35
where the damping factor, expjtj, contains the same 0 100 200 300
homogeneous linewidth found in the sum-over-states (b) t (fs)
expression, equation (13). Equation (35) reveals the recip- Figure 8. (a) Damped time-dependent overlaps relevant to
rocal relationship between dynamics on a particular time- absorption and (b) REP for a mode at about 1000 cm1 .
scale and frequency features in the absorption profile.
The earliest dynamic feature is the initial motion of the increases with the slope of the upper potential surface. In
wavepacket out of the FC region, taking place on the accord with the reflection principle mentioned previously,
timescale t1 in Figure 8. The width of the overall spectrum the overall width of the FC progression is a function of
is inversely related to this timescale. The rate of this motion the slope of the upper state potential in the FC region.
Each time the wavepacket returns to the FC region, the If there is no vibrational frequency shift on excitation, the
overlap maximizes. These recurrences, spaced by the vibra- one-mode overlaps in the above equations are obtained as
tional period (t2 in Figure 8) in the time domain, result in follows:
vibrational sub-bands spaced by the excited state vibrational
2k i !k t
frequency in the absorption spectrum. The timescale t3 for h0k j0k ti D exp [1 expi!k t] i!eg t
damping is inversely related to the broadening of these indi- 2 2
vidual vibrational peaks. Faster damping rates lead to more k
diffuse absorption spectra. h1k j0k ti D p [expi!k t 1]h0k j0k ti 39
2
The overlaps in Figure 8 were calculated for a vibrational
mode with frequency nQ of about 1000 cm1 . The time- where k is the displacement and !k the frequency of mode
dependent overlap in equation (35) is actually a multimodal k. (Analogous expressions which apply to modes having a
quantity which depends on all the displaced (FC-active) frequency change in the excited electronic state are given
vibrational modes. Note that h!i is the energy of the initial elsewhere.4 ) The 00 frequency !eg appears only once in
state, frequently taken to be the zero-point vibrational level. the product of overlaps. Variation of this parameter shifts
The Raman cross-section for the i ! f transition is the calculated profiles along the frequency axis. Note that
calculated as a function of excitation frequency in a similar the collection of terms expi!k t/2 cancels the expi!i t
fashion: term in equations (35) and (36). Undisplaced modes, for
which k D 0, make no contribution to the time-dependent
8p!3s !0 m0ge 4 1 overlaps. Implementation of equations (35) and (36) con-
sR,i!f !0 D hfjiti exp[i!i
9h2 c4 sists of finding a common set of parameters, including
0
2 the displacements for the RR-active modes, the transition

C !0 t] expt dt 36 moment, the 00 energy, and damping factor, which can
simultaneously account for the absorption profiles and the
The overlap integral in equation (36) is that of the final REPs of all modes. The peak intensity of the REP for a
state and the time-evolving initial state. This overlap starts given mode is proportional to the square of the displacement
out at zero at time t D 0, as shown in Figure 8. Thus the for that mode, whereas increasing values of the displace-
Raman intensity depends on the displacement for this over- ments tend to broaden the calculated absorption profile.
lap to build up as the wavepacket moves out of the FC In agreement with the predictions of transform theory,
region. Unlike the absorption cross-section, which can in the time-dependent approach reveals that larger damping
principle be back-Fourier transformed to obtain the damped rates, associated with more diffuse absorption spectra, result
overlap, the Raman cross-section is obtained by first cal- in weaker Raman scattering. It is also apparent that low-
culating a half-FT (to get the transition polarizability, or frequency modes are less readily observed, because longer
amplitude) and then squaring the amplitude. The damping vibrational periods result in more sensitivity to the overall
factor has a very different effect on the Raman intensity damping. Any dynamics that move the wavepacket out of
compared to absorption. Increasing the value of serves the FC region contribute to damping. The presence of more
to broaden the absorption profile while keeping the inte- than one displaced mode causes the overlap relevant to
grated absorbance constant. However, the calculated Raman the intensity of any one mode to decay due to motion
cross-section decreases when is increased. This provides in the direction of the other displaced coordinates. The
a way to disentangle the damping dynamics from the time time-dependent equations can be applied to overtones and
dependence of the overlap integrals. combinations as well as fundamentals. For higher-order
These multimodal overlap integrals may be calculated overtones, compared to the fundamental, the amplitude
from the frequencies and displacements of the FC-active of the final state wavefunction moves toward the turning
modes. Assuming the absorption originates in the ground points of the ground state potential surface, and increasingly
vibrational state of all modes, the overlap needed to calcu- more time is required for the jiti to move to a region
late sA is where the overlap hfjiti is more favorable. Thus higher
order overtones are more susceptible to damping, whether
hijiti D h01 j01 tih02 j02 ti . . . h03N6 j03N6 ti 37 due to dephasing, homogeneous broadening, or multimodal
dynamics. This is the basis for less prevalent overtone
The overlap needed to calculate the fundamental transition activity in larger molecules.
of mode a is A severe limitation of the working equations presented
above is the assumption of exponential damping. This
hfjiti D h01 j01 tih02 j02 ti . . .
functional form may be appropriate in the case where
h1a j0a ti . . . h03N6 j03N6 ti 38 the homogeneous linewidth is limited by the lifetime of
the excited state. In the condensed phase, however, pure where kB is Boltzmanns constant, and T is the absolute
dephasing also influences the linewidth, and this contribu- temperature. Application of the Brownian oscillator model
tion to the damping need not be exponential. This pure consists of adjusting the parameters s and to obtain
dephasing is the result of solvent-induced fluctuations of agreement of the calculated absorption and Raman profiles
the transition frequency !eg , and it plays a critical role with experiment. Increasing either the amplitude or rate of
in the partitioning of the total emitted radiation into RR solvent-induced frequency fluctuations leads to broadening
and fluorescence.14,41 Due to random perturbations from of the absorption profile and diminished Raman cross-
the solvent, the transition frequency is considered to be a sections.
stochastic function: The time-dependent theory can also accommodate true
inhomogeneous broadening, viewed as a Gaussian dis-
!eg t D !0eg C !eg t 40 tribution of 00 transition frequencies which is static
on the Raman timescale. (This timescale is determined
The integrand in equations (35) and (36) should incorporate
by the pure dephasing time for the Raman frequencies,
the function exp[gt], where
which in turn is reciprocally related to the linewidth of
t
0 0 0
a Raman band.) The Raman and absorption cross-sections
exp i !eg t t dt D exp[gt] 41 are then convoluted with a Gaussian distribution of tran-
0
sition frequencies. The calculated cross-sections can also
The angle brackets in equation (41) represent an equilib- include thermal effects by summing over of a Boltzmann-
rium average over the solution degrees of freedom, assumed weighted distribution of initial states i.4 Making these
to be independent of the electronic state of the solute. modifications to equations (35) and (36), the working equa-
One model for gt is the Brownian oscillator model pre- tions are
sented by Mukamel and colleagues.79,80 This model is 1
predicated on the assumption that the correlation function
sR,i!f !0 / Pi d!G! hfjiti
for the solvent-induced frequency fluctuations relaxes expo- i 0
nentially: 2

exp[i!i C !0 C !t] exp[gt] exp[t] dt 46
h!eg 0!eg ti D 2s expt 42
The parameter s represents the amplitude and the rate and

of solvent-induced frequency fluctuations. Experimental 1
evidence81 83 argues that, on the subpicosecond timescale s A !0 / Pi d!G! hijiti exp[i!i
i 1
relevant to the RR experiment, the correlation function
h!eg 0!eg ti is a Gaussian, rather than exponential, C !0 C !t] exp[gjtj] exp[jtj] dt 47
function of time. However, the Brownian oscillator model
has the advantage that it leads to a simple closed-form where G! is a normalized Gaussian distribution func-
expression for gt. The line-broadening function, which tion whose width can be taken as an adjustable param-
is the FT of exp[gt], takes the form of a Lorentzian eter, and Pi is the fraction of molecules in the ini-
when s , and a Gaussian when s . The Brow- tial state i. In equations (46) and (47), is the inverse
nian oscillator model has been extended to permit gt to lifetime and gt accounts for solvent-induced dephas-
have both real and imaginary parts, gt D gR t C igI t, ing. The scaling factors for the absorption and Raman
where the latter depends on the solvent-induced frequency cross-sections are the same as those in equations (35)
shift of the absorption spectrum: and (36).
2
s
gR t D [t 1 C expt] 43
5 APPLICATIONS OF RESONANCE
Esoly RAMAN SPECTROSCOPY
gI t D [1 expt] 44

Intensity analysis of RR spectra has been applied to the
The quantity Esolv is the solvent reorganization energy,
determination of excited state structure and dynamics in a
related to the amplitude s of the solvent-induced fre-
wide variety of organic, inorganic and biochemical systems.
quency fluctuations:
The following survey is admittedly incomplete, but will
2s hopefully provide an overview of the kinds of problems
Esolv D 45 amenable to solution by RRS.
2kB T
5.1 Excited state dynamics and structure RR intensity analysis has been applied to the study of
proton transfer.109 Recent studies have applied RRS to
It has long been recognized that RRS provides the opportu- the analysis of photodissociation of I 3 in solution.
110,111
112,113
nity for the separation of homogeneous and inhomogeneous Reid and colleagues have employed RRS to study
contributions to the spectral linewidth. A study by Strauss solvent effects on the photochemistry of ClO2 . Phillips
and colleagues accomplished this separation for I2 in sev- and colleagues have applied RRS to photodissociation of
eral solvents, using sum-over-states expressions for both the haloethanes,114,115 and Spiro and colleagues have looked at
absorption and Raman cross-sections.84 86 The data were photoisomerization of N-methylacetamide.116
consistent with a dephasing time T2 of about 0.3 ps. Math-
ies and colleagues have employed time-dependent theory to
determine broadening mechanisms and dephasing times in 5.3 Biological applications
a number of systems.87 89 RR spectra excited in the vac-
uum UV have been applied to studies of small molecules Although many biologically active molecules have visible
such as benzene90,91 and ethylene.92 Excited state torsional chromophores which are RR active, the development of
dynamics and photoisomerization have been emphasized in UV/RR techniques has considerably broadened the scope
many RR studies.52,87,93,94 of RRS for biological applications.5,6 The primary pro-
Applications of RRS to the determination of excited state cess of visual transduction, photosynthesis, and structure
geometry are numerous, but one limitation of the approach and function of various heme proteins are examples of
is that only the magnitudes, and not the signs, of the problems which have been addressed. The photoisomer-
normal mode displacements are obtained. Furthermore, a ization of visual pigments has been widely studied by
normal coordinate calculation must be performed in order Mathies and colleagues.117 120 Loppnow and colleagues
to determine the magnitudes of the displacements along the have examined excited state dynamics in blue copper
internal coordinates (bond lengths and angles). Ab initio proteins.121,122 Photosynthetic reaction centers have been
calculations have been employed to supplement the infor- the subject of many RR studies.123 130 Champion and col-
mation from RRS.95,96 In addition, chemical intuition is leagues have focused on myoglobin,131 134 and other heme
often helpful in determining the sign of a normal mode proteins have been investigated with RRS.64,135 138 Also,
displacement. For example, we employed transform theory RRS has been applied to studies of hemoglobin,139 142 pep-
to deduce the normal mode displacements in the electron tide conformation and dynamics,143 145 protein secondary
donoracceptor (EDA) complex between hexamethylben- structure,146 and structure and interactions of DNA.147,148
zene (HMB) and tetracyanoethylene (TCNE).97 Based on Kincaid and colleagues have used RRS to investigate enzy-
quantum calculations, the nodal structure of the TCNE matic mechanisms.149 152 Porphyrins have been the subject
acceptor orbital is consistent with an increase in the CDC of many RRS investigations.153 158 Finally, UV/RRS has
and CN bond lengths in the excited state of the complex. even been applied to studies of bacteria.159
With this assumption, the calculated excited state geome-
try of TCNE, based on RR intensity analysis, was found
to be in good agreement with the structure of the anion 5.4 Electron transfer
TCNE determined from X-ray crystallography, in accord
with the charge-transfer nature of the excited electronic Photoinduced electron transfer has been the subject of
state: HMBC TCNE . many applications of RRS. In inorganic metal complexes,
RRS is advantageous for probing metal-to-ligand, ligand-
to-metal, and intervalence electron transfer. Hupp and
5.2 Photochemistry colleagues160 163 have used RR intensities to derive reorga-
nization energies for intervalence charge-transfer in mixed-
Many RRS studies have focused on photochemical reac- metal and mixed-valence inorganic complexes, and in inter-
tions such as dissociation, ring opening, and proton transfer. facial electron transfer and metalligand charge transfer
Imre and colleagues57,98 applied UV resonance Raman and (MLCT) in ruthenium complexes. The Hupp group was
time-dependent theory to study the photodissociation of perhaps the first to appreciate the connection between spec-
CH3 I and O3 . Excited state ring-opening dynamics99 105 troscopically derived reorganization energies and the rate
and hydrogen migration106 have been explored by Math- of return electron transfer, a subject on which Myers and
ies and colleagues. Kinsey and colleagues have applied colleagues have written widely.164,165 Zink and colleagues
time-dependent theory to the analysis of RR intensities have used the time-dependent theory of RRS to explore
in a number of photodissociating systems.107,108 Also, excited state properties of transition metal complexes166 169
and to interpret interference effects arising from coupling donorsolvent and acceptorsolvent interactions are com-
of excited electronic states.170 172 The MLCT transitions parable in strength to the DA interaction. The equilibrium
of platinum complexes have been studied using RRS.173 geometry of the complex may vary with solvent.184 The
We recently interpreted REPs in a solvatochromic ruthe- excited state complex may undergo a change in symmetry,
nium(II)tetramminebypridine complex with the help of with the result that the normal modes and displacements
quantum chemical calculations.55,174 which contribute to the fluorescence spectrum, and to return
Organic intermolecular EDA complexes have been con- electron transfer, are different from those relevant to the
sidered good candidates for RRS, because they are often absorption and REP analysis. As a result, there remain
weakly fluorescent and may be employed as model sys- numerous questions concerning the solution phase structure
tems for photoinduced electron transfer. Some EDA com- and dynamics in these systems.
plexes between TCNE as acceptor and various substi-
tuted benzenes as donors were the subject of early RR
experiments.175,176 We reported resonance-enhanced low- 5.5 Solvent dynamics and resonance Raman
frequency vibrations of both 1 : 1 donoracceptor (DA) and intensities
2 : 1 donoracceptordonor (DAD) complexes of HMB
and TCNE.177,178 These low-frequency features were We have employed time-dependent theory to investigate
assigned to intermolecular DA vibrations deriving enhance- the solvent relaxation coupled to the electronic transi-
ment from the change in DA distance on going from tions of solvatochromic molecules. The dye betaine-30 has
the weakly bound DA ground state to the ionic DC A been widely used as a probe of solvent polarity,185 and in
excited state. Later, we used both transform theory54 and time-resolved spectroscopic studies of the rate of back elec-
time-dependent theory97 to deduce normal mode displace- tron transfer.186,187 The large difference in dipole moment
ments in the charge-transfer excited state of HMB and between the ground and excited state permits the visi-
TCNE in methylene chloride. We have also observed inter- ble S0 ! S1 transition to be viewed as an intramolecular
ference effects in the REPs of a 2 : 1 DAD complex that charge-transfer excitation. We measured REPs of betaine-30
result from the coupling of two charge-transfer excited in the solvents CH3 CN, CD3 CN, CH3 OH, and CD3 OD and
states.179 Markel et al. also employed time-dependent the- employed time-dependent theory in the analysis.50,51 The
ory in the analysis of REPs for HMBTCNE, but in car- visible absorption spectrum of betaine-30 is quite different
bon tetrachloride solvent.165 They later reported the very in methanol from that in acetonitrile, but there is no depen-
weak charge-transfer emission,180 which was also analyzed dence on solvent isotopomer. Raman intensities, however,
using time-dependent theory. Recently, Phillips et al. have are dramatically different in methanol compared to acetoni-
applied time-dependent theory to analysis of RR intensities trile, and in addition vary with solvent isotopomer. In ace-
in covalently bonded EDA systems.181 tonitrile, the Raman intensities of betaine-30 normal modes
It is now clear that weakly bound intermolecular EDA are greater in CH3 CN than in CD3 CN, whereas in methanol
complexes are fairly challenging systems for the analysis the Raman intensities are greater in CD3 OD than in
of Raman intensity. Quantum chemical calculations sug- CH3 OH. This is illustrated by the REPs for the 1622 cm1
gest that complexes of TCNE and aromatic donors have mode of betaine-30 shown in Figure 9. The results of this
poorly defined ground state geometries, strong dependence study can be interpreted in terms of the influence of solvent-
of the transition moment on the intermolecular geome- induced dephasing. The much greater amplitude of sol-
try, and overlapping excited states.182,183 Solvent effects vent relaxation in acetonitrile than in methanol leads to
in these kinds of systems are especially important because depressed Raman intensities in the former. On the basis of
2.90 0.24
Methanol Acetonitrile
per molecule)
per molecule)
2.32 0.19
2
2
R(107
R(107
1.74 0.14
1.16 0.10
0.58 0.05
0.00 0.00
16 000 18 000 20 000 22 000 16 000 17 000 18 000 19 000 20 000
(a) (cm1)
(b) (cm1)

0 0
Figure 9. REPs for the 1622 cm1 mode of betaine-30 in (a) methanol and (b) acetonitrile. The triangles are for the undeuterated
solvent and the squares are for the deuterated solvent. The symbols are experimental data and the curves are calculated.
the absorption spectra alone, the larger transition moment in is lost, because simultaneous modeling of the absorption
acetonitrile compared to methanol would have led to larger and Raman profiles using a common gt is not valid.
Raman intensity in acetonitrile. The solvent isotope effects Another startling result of this work is the observation
reflect the very sensitive dependence of the Raman intensity that different solute Raman modes are perturbed differ-
on the amplitude of the solvent relaxation that takes place ently by isotopic substitution of the solvent, which leads
after the transition to the excited electronic state. In acetoni- us to conclude that solvent-induced dephasing is mode
trile, the slightly higher collision rate in the undeuterated dependent. This is in contrast to the basic assumption
solvent, compared to deuterated, results in smaller ampli- that the dephasing rate (i.e. in equation 13) is indepen-
tude of inertial motion and larger Raman cross-sections. In dent of the vibrational state. Further evidence for mode-
methanol, however, the amplitude of solvent motion is lim- dependent solvent effects was found in our study of the
ited by hydrogen-bonding dynamics. The smaller zero-point hemicyanine dye 4-[2-(4-dimethylamino-phenyl)ethenyl]-
energy and amplitude of these low-frequency vibrations 1-methylpyridinium iodide (HR) in H2 O and D2 O.52 In this
in deuterated methanol, compared to CH3 OH, results in study, the magnitude and direction of the solvent isotope
less damping and larger Raman cross-sections. Particularly effect were observed to vary with the solute normal mode,
strong solventisotope effects were observed for some of as shown in the example profiles of Figure 10. It is not
the lower-frequency vibrations of betaine-30, which may possible, in this system, to account for the solvent isotope
be assignable to torsional modes coupled to the electronic effect merely by using a different gt in water and deuter-
transition. ated water. Application of the standard time-dependent
Although the physical conclusions from this work seem equations requires different mode displacements i to be
clear, the implications for the use of standard theoretical employed for HR in H2 O and D2 O. This variation of the
approaches to RR intensity are troubling. If the solvent i values with solvent isotope is probably artefactual and
lineshape function defined in equation (41) is independent reflects failure of the standard assumptions of the theory.
of the electronic state of the solute, then the absence of Unfortunately, both transform theory and time-dependent
the solvent isotope effect on the absorption spectrum is theory are based on the assumption that the homogeneous
inconsistent with the observation of such an effect in all linewidth or dephasing rate is independent of vibrational
the Raman profiles. The solvent relaxation coupled to the state.
ground electronic state, relevant to the calculation of the The homogeneous linewidth may also depend on solvent
absorption profile and apparently independent of solvent isotope due to variation in the lifetime of the excited elec-
isotope, reflects solvent dynamics already equilibrated to tronic state, if the lifetime is sufficiently short to matter
the ground state charge distribution. Immediately after the on the Raman timescale. Nonradiative relaxation rates can
electronic transition, however, the solvent molecules find depend on solvent isotope. We have recently observed the
themselves in a highly nonequilibrium configuration, due fluorescence yields of the dye LDS 750 in CH3 OH and
to the large change in solute dipole moment. The sol- CD3 OD, and find them to be about 30% higher in deuter-
vent relaxation on which the Raman cross-section depends ated methanol.188 However, the Raman intensity of LDS
proceeds from this nonequilibrium state, and the larger 750 is the same in CH3 OH as in CD3 OD. These observa-
amplitude motion may be more sensitive to solvent iso- tions suggest that the pure dephasing rate is dominated by
tope effects than the equilibrium fluctuations. Thus one an intramolecular process and is thus independent of the
of the major advantages of the time-dependent approach, solvent isotope. Further analysis of the Raman, absorption,
separation of internal and solvent reorganization energies, and emission profiles in this system are in progress.
R(107 per molecule)
R(107 per molecule)
0.20
0.020
0.15
0.10
2
0.010
0.05
0.00 0.000
15 000 20 000 25 000 30 000 15 000 20 000 25 000 30 000
(a)
0 (cm1) (b)
0 (cm1)
Figure 10. REPs for (a) the 1588 cm1 mode and (b) the 1642 cm1 mode of HR in H2 O (empty symbols) and D2 O (filled symbols).
The symbols are experimental data and the curves are calculated.
5.6 Future directions FC FranckCondon

FT Fourier Transform
Predicting future directions of research is even more haz- HMB Hexamethylbenzene
ardous than attempting to review the work that has already HR 4-[2-(4-Dimethylamino-phenyl)ethenyl]-
been done. Clearly the ideas presented here reflect the 1-methylpyridinium Iodide
authors preferences and prejudices. It is my opinion that KHD KramersHeisenbergDirac
both instrumental and theoretical advances in the field of MLCT MetalLigand Charge Transfer
RRS are needed. A major instrumental advantage will be NIR Near-infrared
provided when REPs can be measured routinely and directly REP Raman Excitation Profile
by simultaneously scanning the excitation and emission RR Resonance Raman
frequencies. Kinsey and colleagues have demonstrated the RRS Resonance Raman Spectroscopy
viability of this approach by measuring continuous REPs for SERDS Shifted Excitation Raman Difference
iodobenzene excited from 216 to 233 nm.189 The approach Spectroscopy
is technologically challenging, but affords the opportu- TCNE Tetracyanoethylene
nity to obtain more detailed information in the frequency UV Ultraviolet
domain. The advantages would be especially appreciated YAG Yttrium Aluminum Garnet
in studies where excited state couplings lead to interfer-
ences in the Raman profiles, because RR data gathered
using coarsely spaced excitation wavelengths could easily REFERENCES
miss de-enhancement effects or attribute real structure in
the REP to mere noise. A major theoretical obstacle is the 1. P.M. Champion and A.C. Albrecht, Ann. Rev. Phys. Chem.,
33, 353 (1982).
loss of phase information in the time-dependent expression
2. B.B. Johnson and W.L. Peticolas, Ann. Rev. Phys. Chem.,
for the Raman cross-section, which is found from the abso-
27, 465 (1976).
lute square of the transition polarizability, which in turn is
3. W. Siebrand and M. Zgierski, Resonance Raman Spec-
calculated from a half-FT of the time correlation function.
troscopy A Key to Vibronic Coupling, in Excited
Unlike the absorption spectrum (equation 35), which can States, ed. E.C. Lim, Academic Press, New York, Vol. 1,
be inverse Fourier-transformed to calculate the dynamical 136 (1979).
information directly, the Raman profile cannot be inverted 4. A.B. Myers and R.A. Mathies, Resonance Raman Intensi-
to obtain dynamical information. Some progress has been ties: A Probe of Excited State Structure and Dynamics,
made in developing a maximum entropy approach to inver- in Biological Applications of Raman Spectroscopy, ed.
sion of REPs,190 192 but the method has yet to be widely T.G. Spiro, Wiley, New York, Vol. 1, 158 (1988).
applied to experimental Raman profiles. Future applications 5. S.A. Asher, Ann. Rev. Phys. Chem., 39, 537 (1988).
of RRS to systems with strong solvent interactions will 6. S.A. Asher, Anal. Chem., 65, 201A (1993).
depend on completely new theoretical approaches to treat 7. L.D. Ziegler, J. Chem. Phys., 86, 1703 (1987).
nonlinear (solute-dependent) solvent relaxation and mode- 8. C.M. Jones, V.L. Devito, P.A. Harmon and S.A. Asher,
dependent dephasing and relaxation dynamics. Appl. Spectrosc., 41, 1268 (1987).
9. C.R. Johnson, M. Ludwig and S.A. Asher, J. Am. Chem.
Soc., 108, 905 (1986).
ACKNOWLEDGMENTS 10. M.J. Pelletier, Raman Instrumentation, in Analytical
Applications of Raman Spectroscopy, ed. M.J. Pelletier,
Support from the National Science Foundation is gratefully Blackwell Science, Oxford, 5397 (1999).
acknowledged. I also owe thanks to my colleagues Fritz 11. M. Ludwig and S.A. Asher, Appl. Spec., 42, 1458 (1988).
Knorr, Mark Wall, and Xihua Zhao for their support and 12. T.C. Strekas, D.H. Adams, A. Packer and T.G. Spiro, Appl.
assistance during the preparation of this manuscript. Spectrosc., 28, 324 (1974).
13. D.F. Shriver and J.B.R. Dunn, Appl. Spectrosc., 42, 520
(1988).
ABBREVIATIONS AND ACRONYMS 14. L.D. Ziegler, Acc. Chem. Res., 27, 1 (1994).
15. M.K. Lawless and R.A. Mathies, J. Chem. Phys., 96, 8037
CCD Charge Coupled Device (1992).
DA DonorAcceptor 16. M. Kasha, J. Chem. Phys., 20, 71 (1962).
DAD DonorAcceptorDonor 17. J.M. Friedman and R.M. Hochstrasser, Chem. Phys. Lett.,
EDA Electron DonorAcceptor 33, 225 (1975).
18. R.P. Van Duyne, D.L. Jeanmaire and D.F. Shriver, Anal. 47. N. Abe, M. Wakayama and M. Ito, J. Raman Spectrosc., 6,
Chem., 464, 213 (1974). 38 (1977).
19. J.M. Harris, R.W. Chrisman, F.E. Lytle and R.S. Tobias, 48. G. Eckhardt and W. Wagner, J. Mol. Spectrosc., 19, 407
Anal. Chem., 48, 1937 (1976). (1966).
20. T.L. Gustafson and F.E. Lytle, Anal. Chem., 54, 634 (1982). 49. L. Onsager, J. Am. Chem. Soc., 58, 1486 (1936).
21. K. Kamogawa, T. Fujii and T. Kitagawa, Appl. Spectrosc., 50. Y. Zong and J.L. McHale, J. Chem. Phys., 106, 4963 (1997).
42, 248 (1988). 51. Y. Zong and J.L. McHale, J. Chem. Phys., 107, 2920 (1997).
22. J.N. Demas and R.A. Keller, Anal. Chem., 57, 538 (1985). 52. X. Cao and J.L. McHale, J. Chem. Phys., 109, 1901 (1998).
23. A.P. Shreve, N.J. Cherepy and R.A. Mathies, Appl. Spec-
53. G.D. Scholes, T. Fournier, A.W. Parker and D. Phillips, J.
trosc., 46, 707 (1992).
Chem. Phys., 111, 5999 (1999).
24. C.K. Mann and T.J. Vickers, Appl. Spectrosc., 41, 427
54. B.M. Britt, J.L. McHale and D.M. Friedrich, J. Phys. Chem.,
(1987).
99, 6347 (1995).
25. P.A. Mosier-Boss, S.H. Lieberman and R. Newberry, Appl.
55. J.H. Streiff and J.L. McHale, J. Chem. Phys., 112, 841
Spectrosc., 49, 630 (1995).
(2000).
26. F.J. Purcell, R. Kaminski and E. Russavage, Appl. Spec-
56. A.C. Albrecht, J. Chem. Phys., 34, 1476 (1961).
trosc., 34, 323 (1980).
57. D. Imre, J.L. Kinsey, A. Sinha and J. Krenos, J. Phys.
27. M. Fryling, C.J. Frank and R.L. McCreery, Appl. Spectrosc.,
Chem., 88, 3956 (1984).
47, 1965 (1993).
28. K.G. Ray and R.L. McCreery, Appl. Spectrosc., 51, 108 58. J.L. McHale, Molecular Spectroscopy, Prentice-Hall,
(1997). Upper Saddle River, NJ (1999).
29. H. Hamaguchi, Appl. Spectrosc. Rev., 24, 137 (1988). 59. L.A. Nafie and W. Peticolas, J. Chem. Phys., 57, 3145
(1972).
30. K. Iwata, H. Hamaguchi and M. Tasumi, Appl. Spectrosc.,
42, 12 (1988). 60. O. Sonnich-Mortensen and S. Hassing, Adv. Infrared Raman
Spectrosc., 6, 1 (1980).
31. http://www.chemistry.ohio-state.edu
61. H. Hamaguchi, Adv. Infrared Raman Spectrosc., 12, 273
32. J.R. Scherer Experimental Considerations for Accurate (1985).
Polarization Measurements in Analytical Raman Spec-
troscopy, eds J.G. Grasselli and J.B. Bulkin, John Wiley 62. H.B. Lueck, J.L. McHale and W.D. Edwards, J. Am. Chem.
& Sons, New York, 4557. Soc., 114, 2342 (1992).
33. V. Teboul, J.L. Godet and Y. LeDuff, Appl. Spectrosc., 46, 63. Q.-y. Shang and B.S. Hudson, Chem. Phys. Lett., 183, 63
467 (1992). (1991).
34. M.D. Ediger, R.S. Moog, S.G. Boxer and M.D. Fayer, 64. T.G. Spiro and T.C. Strekas, Proc. Natl. Acad. Sci. USA, 69,
Chem. Phys. Lett., 88, 123 (1982). 2622 (1972).
35. K.T. Schomacker, J.K. Delaney and P.M. Champion, J. 65. H. Hamaguchi, I. Harada and T. Shimanouchi, Chem. Phys.
Chem. Phys., 85, 4240 (1986). Lett., 32, 103 (1975).
36. J.N. Demas and G.A. Crosby, J. Phys. Chem., 75, 991 66. P.M. Champion and A.C. Albrecht, Chem. Phys. Lett., 82,
(1971). 410 (1981).
37. J.R. Nestor and E.R. Lippincott, J. Raman Spectrosc., 1, 305 67. B.R. Stallard, P.M. Champion, P.R. Callis and A.C. Al-
(1973). brecht, J. Chem. Phys., 78, 712 (1983).
38. M.O. Trulson and R.A. Mathies, J. Chem. Phys., 84, 2068 68. V.V. Hizhnyakov and I.J. Tehver, Opt. Commun., 32, 419
(1986). (1980).
39. S.A. Asher and C.R. Johnson, J. Phys. Chem., 89, 1375 69. D.L. Tonks and J.B. Page, Chem. Phys. Lett., 66, 449
(1985). (1979).
40. J. Dudik, C.R. Johnson and S.A. Asher, J. Chem. Phys., 82, 70. D.L. Tonks and J.B. Page, J. Chem. Phys., 76, 5820 (1982).
1732 (1985). 71. L. Chinsky, A. Laigle, W.L. Peticolas and P.-Y. Turpin, J.
41. B. Li and A.B. Myers, J. Phys. Chem., 94, 4051 (1990). Chem. Phys., 76, 1 (1982).
42. H.H. Eysel, J. Raman Spectrosc., 19, 223 (1988). 72. D.C. Blazej and W. Peticolas, J. Chem. Phys., 72, 3134
43. R. Fan, T. Kalbfleisch and L.D. Ziegler, J. Chem. Phys., (1980).
104, 3886 (1996). 73. T.W. Patapoff, P.-Y. Turpin and W.L. Peticolas, J. Phys.
44. M.J. Colles and J.E. Griffiths, J. Chem. Phys., 56, 3384 Chem., 90, 2347 (1986).
(1971). 74. S.-Y. Lee and E.J. Heller, J. Chem. Phys., 71, 4777 (1979).
45. J.E. Griffiths, J. Chem. Phys., 60, 2556 (1974). 75. E.J. Heller, R.L. Sundberg and D. Tannor, J. Phys. Chem.,
46. A.C. Albrecht and M.C. Hutley, J. Chem. Phys., 55, 4438 86, 1822 (1982).
(1971). 76. E.J. Heller, Acc. Chem. Res., 14, 368 (1981).
77. D.J. Tannor and E.J. Heller, J. Chem. Phys., 77, 202 (1982). 104. P.J. Reid, M.K. Lawless, S.D. Wickham and R.A. Mathies,
78. A.B. Myers, R.A. Mathies, D.J. Tannor and E.J. Heller, J. J. Phys. Chem., 98, 5597 (1994).
Chem. Phys., 77, 3857 (1982). 105. M.K. Lawless, S.D. Wickham and R.A. Mathies, Acc. Chem.
79. S. Mukamel, Principles of Nonlinear Optical Spectroscopy, Res., 28, 493 (1995).
Oxford University Press, New York (1995). 106. P.J. Reid, A.P. Shreve and R.A. Mathies, J. Phys. Chem.,
80. B. Li, A.E. Johnson, S. Mukamel and A.B. Myers, J. Am. 97, 12 691 (1993).
Chem. Soc., 116, 11 039 (1994). 107. B.R. Johnson, C. Kittrell, P.B. Kelly and J.C. Kinsey, J.
81. M. Cho and G.R. Fleming, Adv. Chem. Phys., 107, 311 Phys. Chem., 100, 7743 (1996).
(1999). 108. R.E. Stevens, C. Kittrell and J.L. Kinsey, J. Phys. Chem.,
82. S.J. Rosenthal, X. Xie, M. Du and G.R. Fleming, J. Chem. 99, 11 067 (1995).
Phys., 95, 4715 (1987). 109. L.A. Peteanu and R.A. Mathies, J. Phys. Chem., 96, 6910
83. R.M. Stratt and M. Maroncelli, J. Phys. Chem., 100, 12 981 (1992).
(1996). 110. A.E. Johnson and A.B. Myers, J. Phys. Chem., 100, 7778
84. R.J. Sension and H.L. Strauss, J. Chem. Phys., 85, 3791 (1996).
(1986). 111. A.E. Johnson and A.B. Myers, J. Chem. Phys., 104, 2497
85. R.J. Sension, T. Kobayashi and H.L. Strauss, J. Chem. Phys., (1996).
87, 6221 (1987). 112. P.J. Reid, A.P. Esposito, C.E. Foster and R.A. Beckman, J.
86. R.J. Sension, T. Kobayashi and H.L. Strauss, J. Chem. Phys., Chem. Phys., 107, 8262 (1997).
87, 6233 (1987).
113. A.P. Esposito, C.E. Foster, R.A. Beckman and P.J. Reid, J.
87. A.B. Myers, M.O. Trulson and R.A. Mathies, J. Chem. Phys. Chem., 101, 5309 (1997).
Phys., 83, 5000 (1985).
114. X. Zheng, Y.F. Cheng and D.L. Phillips, Chem. Phys. Lett.,
88. A.B. Myers, M.O. Trulson, J.A. Pardoen, C. Heeremans, 292, 295 (1998).
J. Lugtenburg and R.A. Mathies, J. Chem. Phys., 84, 633
(1986). 115. X. Zheng and D.L. Phillips, Chem. Phys. Lett., 268, 79
(1998).
89. M.K. Lawless and R.A. Mathies, J. Chem. Phys., 96, 8037
(1992). 116. Y. Wang, R. Purrelo and T.G. Spiro, J. Am. Chem. Soc., 111,
8274 (1989).
90. R.J. Sension, R.J. Brudszynski and S. Li, J. Chem. Phys.,
96, 2617 (1992). 117. S.P.A. Fodor, R.A. Bogomolni and R.A. Mathies, Biochem-
istry, 26, 6775 (1987).
91. R.J. Sension, R.J. Burdzynski and B. Hudson, J. Chem.
Phys., 94, 873 (1989). 118. G.R. Loppnow and R.A. Mathies, Biophys. J., 54, 35 (1988).
92. R.J. Sension and B. Hudson, J. Chem. Phys., 90, 1377 119. G.G. Kochendoerfer, Z. Wang, D.D. Oprian and R.A.
(1989). Mathies, Biochemistry, 36, 6577 (1997).
93. A.B. Myers and R.A. Mathies, J. Chem. Phys., 81, 1552 120. S.W. Lin, M. Groesbeek, I. van der Hoef, P. Verdegem,
(1984). J. Lugtenburg and R.A. Mathies, J. Phys. Chem., 102, 2878
(1998).
94. M.O. Trulson and R.A. Mathies, J. Chem. Phys., 94, 5741
(1990). 121. E. Fraga, M.A. Webb and G.R. Loppnow, J. Phys. Chem.,
95. B.S. Hudson and L.M. Markhan, J. Raman Spectrosc., 29, 100, 3278 (1996).
480 (1998). 122. M.A. Webb, C.M. Kwong and G.R. Loppnow, J. Phys.
96. W.L. Peticolas and D.C. Blazej, Chem. Phys. Lett., 63, 604 Chem., 101, 5062 (1997).
(1974). 123. J.B. Ames, S.R. Bolton, M.M. Netto and R.A. Mathies, J.
97. B.M. Britt, H.B. Lueck and J.L. McHale, Chem. Phys. Lett., Am. Chem. Soc., 112, 9007 (1990).
190, 528 (1992). 124. S.W. Lin, S.P.A. Fodor, L.J.W. Miercke, R.F. Shand, M.C.
98. D. Imre, J.L. Kinsey, R. Field and D. Katayama, J. Phys. Betlach, R.M. Stroud and R.A. Mathies, Photochem. Photo-
Chem., 86, 2564 (1982). biol., 53, 341 (1991).
99. M.O. Trulson, G.D. Dollinger and R.A. Mathies, J. Am. 125. A.P. Shreve, N. Cherepy, S. Franzen, S.G. Boxer and R.A.
Chem. Soc., 109, 586 (1987). Mathies, Proc. Natl. Acad. Sci. USA, 88, 11 207 (1991).
100. M.A. Trulson, G.D. Dollinger and R.A. Mathies, J. Chem. 126. J.B. Ames, J. Raap, J. Lugtenburg and R.A. Mathies, Bio-
Phys., 90, 4274 (1989). chemistry, 31, 12 546 (1992).
101. P.J. Reid, S.J. Doig and R.A. Mathies, Chem. Phys. Lett., 127. N.J. Cherepy, A.P. Shreve, L.J. Moore, S. Franzen, S.G.
156, 163 (1989). Boxer and R.A. Mathies, J. Phys. Chem., 98, 6023 (1994).
102. M.K. Lawless and R.A. Mathies, J. Chem. Phys., 100, 2482 128. N.J. Cherepy, A.R. Holzwarth and R.A. Mathies, Biochem-
(1994). istry, 34, 5288 (1995).
103. M.K. Lawless, S.D. Wickham and R.A. Mathies, J. Am. 129. N.J. Cherepy, A.P. Shreve, L.J. Moore, S.G. Boxer and R.A.
Chem. Soc., 116, 1593 (1994). Mathies, J. Phys. Chem., 101, 3250 (1997).
130. N.J. Cherepy, A.P. Shreve, L.J. Moore, S.G. Boxer and R.A. 157. A. Salehi, W.A. Oertling, H.N. Fonda, G.T. Babcock
Mathies, Biochemistry, 36, 8559 (1997). and C.K. Chang, Photochem. Photobio., 48, 525
131. D. Morikis, P. Li, O. Bangcharoenpauron, J.T. Sage and (1988).
P.M. Champion, J. Phys. Chem., 95, 3391 (1991). 158. V.A. Walters, J.C. DePaula, G.T. Babcock and G.E. Leroi,
132. A.V. Wel, J.T. Sage, D. Morikis, P.M. Champion, M. Chiu J. Am. Chem. Soc., 111, 8300 (1989).
and S.G. Sligar, J. Am. Chem. Soc., 113, 9655 (1991). 159. W.H. Nelson, R. Manoharan and J.F. Sperry, Appl. Spec-
133. P.M. Champion, J. Raman Spectrosc., 23, 557 (1992). trosc. Rev., 27, 67 (1992).
160. H. Lu, V. Petrov and J.T. Hupp, Chem. Phys. Lett., 235, 521
134. L. Zhu, J.T. Sage and P.M. Champion, Biochemistry, 32,
(1995).
11 181 (1993).
161. J.T. Hupp, C. Wang, B.K. Mohney, R.D. Williams and G.C.
135. P.G. Wright, P. Stein, J.M. Burke and T.G. Spiro, J. Am.
Walker, J. Am. Chem. Soc., 120, 5848 (1998).
Chem. Soc., 101, 3531 (1979).
162. S.K. Doorn and J.T. Hupp. J. Am. Chem. Soc., 111, 1142
136. G.A. Schick and D.F. Bocian, J. Am. Chem. Soc., 106, 1682
(1989).
(1984).
163. R.D. Williams, V.I. Petrov, H.P. Lu and J.T. Hupp, J. Phys.
137. C. Varotsis, G.T. Babcock, J.A. Garcia-Horsman and R.B.
Chem. A, 101, 8070 (1997).
Gennis, J. Phys. Chem., 99, 16 817 (1995).
164. A. Myers Kelley, J. Phys. Chem. A, 103, 6891 (1999).
138. J.P.M. Schelvis, Y. Zhao, M.A. Marletta and G.T. Babcock,
Biochemistry, 37, 16 289 (1998). 165. F. Markel, N.S. Ferris, I.R. Gould and A.B. Myers, J. Am.
Chem. Soc., 114, 6208 (1992).
139. M. Nagai, H. Wajcman and T. Kitagawa, Biochemistry, 38,
1243 (1999). 166. J.I. Zink and J.L. Wooton, J. Am. Chem. Soc., 119, 1895
(1997).
140. L.A. Dick, G. Heibel and T.G. Spiro, Biochemistry, 38,
6406 (1999). 167. K.S. Shin, R.J.H. Clark and J.I. Zink, J. Am. Chem. Soc.,
111, 4244 (1989).
141. S. Nagatome, N. Nagai and T. Kitagawa, Biochemistry, 38,
9659 (1999). 168. J.L. Wooten and J.I. Zink, J. Phys. Chem., 99, 7251
(1995).
142. J.M. Friedman, S. Huan and E.S. Peterson, Biochemistry,
36, 6197 (1997). 169. D. Wexler and J.I. Zink, Inorg. Chem., 35, 4064 (1996).
143. S.A. Asher, P. Li and X.G. Cheng, J. Am. Chem. Soc., 119, 170. C. Reber and J.I. Zink, J. Chem. Phys., 96, 2681 (1992).
1116 (1997). 171. C. Reber and J.I. Zink, J. Phys. Chem., 96, 571 (1992).
144. S. Song and S.A. Asher, J. Am. Chem. Soc., 111, 4295 172. K.-S.K. Shin, J.I. Zink, J. Am. Chem. Soc., 112, 7148
(1989). (1990).
145. G. Sieler, R. Schweitzer-Stenner and S.A. Asher, J. Phys. 173. W.M. Kwok, D.L. Phillips and V.W.-W. Yam, Chem. Phys.
Chem. B, 103, 372 (1999). Lett., 262, 699 (1996).
146. Z. Chi, X.G. Chen and S.A. Asher, Biochemistry, 37, 2854 174. J.H. Streiff, W.D. Edwards and J.L. McHale, Chem. Phys.
(1998). Lett., 312, 369 (1999).
147. H. Takehuchi and J. Sasamori, Biopolymers, 35, 359 (1995). 175. K.H. Michaelian, K.E. Rieckhoff and E.-M. Voigt, Proc.
148. N. Cho and S.A. Asher, J. Am. Chem. Soc., 115, 6349 Natl. Acad. Science USA, 72, 4196 (1975).
(1993). 176. P.W. Jensen, Chem. Phys. Lett., 39, 138 (1976).
149. J.R. Kincaid, Y. Zheng, J. Al-Mustufa and K. Czarnecki, J. 177. M.L. Smith and J.L. McHale, J. Phys. Chem., 89, 4002
Biol. Chem., 271, 28 805 (1996). (1985).
150. K. Czarnecki, S. Nimri, Z. Gross, L.M. Proniewicz and J.R. 178. J.L. McHale and M.J. Merriam, J. Phys. Chem., 93, 526
Kincaid, J. Am. Chem. Soc., 118, 2929 (1996). (1989).
151. M. Simianu and J.R. Kincaid, J. Am. Chem. Soc., 117, 4628 179. B.M. Britt and J.L. McHale, Chem. Phys. Lett., 270, 551
(1995). (1997).
152. C. Rajani and J.R. Kincaid, J. Raman Spectrosc., 26, 969 180. K. Kulinowski, I.R. Gould, N.S. Ferris and A.B. Myers,
(1995). J. Phys. Chem., 99, 17 715 (1995).
153. K. Czarnecki, L.M. Proniewicz, H. Fujii and J.R. Kincaid, 181. D.L. Phillips, I.R. Gould and A.B. Myers, Chem. Phys.
J. Am. Chem. Soc., 118, 4680 (1996). Lett., 258, 87 (1996).
154. L. Proniewicz, J. Golus, K. Nakamoto and J.R. Kincaid, J. 182. W.D. Edwards, M. Du, J.S. Royal and J.L. McHale, J. Phys.
Raman. Spectrosc., 26, 27 (1996). Chem., 94, 5748 (1990).
155. W.A. Oertling, A. Salehi, C.K. Chang and G.T. Babcock, J. 183. L.C. Emery, J.M. Sheldon, W.D. Edwards and J.L. McHale,
Phys. Chem., 91, 3114 (1987). Spectrochim. Acta A, 48, 715 (1992).
156. W.A. Oertling, G.T. Babcock and C.K. Chang, Inorg. Chem., 184. L.A. Walker II, S. Pullen, B. Donovan and R.J. Sension,
26, 4296 (1987). Chem. Phys. Lett., 242 (1995).
185. C. Reichardt, Chem. Rev., 92, 2319 (1994). 189. C.-Y. Kung, B.-Y. Chang, C. Kittrell, B.R. Johnson and
186. A.E. Johnson, N.E. Levinger, W. Jarzeba, R.E. Schlief, J.L. Kinsey, J. Phys. Chem., 97, 2228 (1993).
D.A.V. Kliner and P.F. Barbara, Chem. Phys., 176, 555 190. F. Remacle and R.D. Levine, J. Chem. Phys., 99, 4908
(1993). (1993).
187. P.F. Barbara, G.C. Walker and T.P. Smith, Science, 256, 191. F. Remacle and R.D. Levine, J. Phys. Chem., 97, 12 553
975 (1992). (1993).
188. M.H. Wall, F.J. Knorr and J.L. McHale, J. Phys. Chem. A, 192. F. Remacle, R.D. Levine and J.L. Kinsey, Chem. Phys.
104, 9494 (2000). Lett., 205, 267 (1993).
Ultraviolet Raman Spectrometry
Sanford A. Asher
University of Pittsburgh, Pittsburgh, PA, USA
1 INTRODUCTION mid-infrared absorption measurements. Raman excitation

used Hg arc emission lines and photographic detection. The
Raman spectroscopy measures the magnitudes and intensi- use of the Hg arc allowed the first ultraviolet (UV) Raman
ties of frequency shifts that occur because of the inelastic measurements, some of which were actually resonance
scattering of light from matter.1 4 The observed shifts Raman measurements within the absorption bands of
can be used to extract information on molecular structure analyte molecules.
and dynamics. In addition, important information can be After 1945 the advent of relatively convenient IR instru-
obtained by measuring the change in the electric field ori- mentation decreased enthusiasm for Raman spectroscopy,
entation of the scattered light relative to that of the incident and it became a specialized field. Much of the work before
exciting light. 1960 centered on examining molecular structure and molec-
The experiment is usually performed by illuminating a ular vibration dynamics. Major advances occurred during
sample with a high-intensity light beam with a well-defined this period in the theoretical understanding of Raman spec-
frequency and a single linear polarization (Figure 1); the troscopy and in the resonance Raman phenomenon, where
scattered light is collected over some solid angle and mea- excitation occurs within an electronic absorption band.
sured to determine frequency, intensity, and polarization. The first commercial Raman instrument became available
The inelastic scattering Raman phenomenon, which leads in 1953.
to frequency shifts, is phenomenologically distinct from The advent of the photomultiplier tube and the HeNe
the relaxed emission denoted as fluorescence (Figure 2) laser caused a resurgence of interest. In 1928, 58 papers
because the inelastic scattering is a single event, and a real were published on Raman spectroscopy; by 1937, the total
emitting excited state is never created. was 1785. Between 1987 and 1989, for example, more than
In contrast, relaxed fluorescence emission (Figure 2) 6000 papers were published with the word Raman in their
occurs through the population of an excited state, which title.6 Raman spectroscopy has become a major tool for
can subsequently undergo vibrational relaxation prior to flu- fundamental studies in physics and physical chemistry as
orescence back to the ground state. The distinction between well as for applied studies in the biological sciences and
fluorescence and Raman scattering is more subtle for exci- analytical chemistry.
tation within the absorption bands of small molecules in the During the 1970s most Raman instrumentation used ArC
gas phase, where single vibrational-level fluorescence can lasers with excitation between 450 and 520 nm a region
occur from well-defined vibronic levels.5 that, unfortunately, is especially prone to interference from
Much of the early work centered on vibrational Raman fluorescent impurities. The interferences degrade signal-
scattering; these studies were central to developing a to-noise ratios (S/Ns), resulting in poor spectra for many
fundamental understanding of molecular structure and complex samples. Many Raman studies involved pure com-
vibrational motion. Before 1945 it was much easier to pounds or those samples that proved amenable to visible
take Raman spectroscopic measurements than to take excitation.
The usefulness of Raman spectroscopy for studying
John Wiley & Sons Ltd, 2002. aqueous samples, especially biological samples, was
of the resonance Raman effect in selective studies of

es
the chromophoric hemes of hemoglobin and myoglobin.
Spectra of very high S/N were obtained from heme protein
I(0 v)
samples at relatively low concentrations. This technique
was quickly adopted by those researchers interested in
e0 biological problems. Amazing progress has been made over
the past 30 years using this technique to, for example, study
heme protein structure and function8,9 and to elucidate
the photochemical mechanisms of visual pigments such
I(0 )
as rhodopsin.10 Until recently, however, resonance Raman
Figure 1. Raman inelastic light-scattering process showing exci- spectroscopy was limited to the small subset of compounds
tation of a sample with light of an intensity In0 and frequency
absorbing in the visible or near-UV spectral regions where
n0 . The incident light has an electric field of magnitude and direc-
tion shown by eo . Raman scattered light at a frequency n0 nv (nv laser sources typically have been available.
is the Raman active molecular vibrational frequency) and inten- The 1980s saw even more rapid advancements, such as
sity In0 nv ) is collected over the solid angle . A ray of this the development of UV-Raman spectroscopy and of Fourier
light is shown with its electric field polarized along es parallel to transform (FT)-Raman spectroscopy,11 which uses near-
the incident field. [Reprinted with permission from S.A. Asher,
Anal. Chem., 65(2), 60A (1993). Copyright (1993) American
infrared excitation. These extensions into spectral regions
Chemical Society.] away from the visible have dramatically increased the utility
and applicability of Raman spectroscopy.
appreciated early on. Because the Raman scattering of In FT-Raman spectroscopy, the multiplex advantage
water is very weak, it is easy to study species in aqueous of the Michelson interferometer is used to increase the
environments. IR measurements, in contrast, often suffer spectral S/N that is limited by IR detector background
from water interferences that cannot be overcome. In 1972 noise. FT-Raman spectroscopy has the advantage of exci-
a paper by Strekas and Spiro7 demonstrated the usefulness tation in a spectral region that is much less susceptible
S2
0 0a
Excited
state
S1
0
0v
0 0v
0 0v
v
Ground IR Non-Resonance Resonance Absorption
state Raman Raman fluorescence
(a) (b) (c) (d)
Figure 2. Infrared (IR) absorption and Raman scattering. (a) Vibrational IR absorption. (b) Inelastic scattering of light of frequency
n0 to a frequency of n0 nv . The shift in frequency corresponds to a vibrational frequency in this case, and the scattering is known
as vibrational Raman scattering. Because the excitation occurs at a frequency far removed from any absorption bands associated
with excitations to electronic excited states, the phenomenon is known as normal, or non-resonance, Raman scattering. (c) Excitation
occurring within an electronic transition produces resonance Raman scattering. (d) Absorption and fluorescence emission in which an
excited state is created. Typical relaxed fluorescence occurs subsequent to the population of an excited state. Prior to emission, this
excited state quickly relaxes to the lowest vibrational level of the lowest singlet excited electronic state. Emission is broad and the
molecule undergoes a transition to different vibronic levels of the ground state. [Reprinted with permission from S.A. Asher, Anal.
Chem., 65(2), 60A (1993). Copyright (1993) American Chemical Society.]
Ultraviolet Raman Spectrometry 3
than visible excitation to fluorescent interferences. Non- vibrational motion causes an oscillation of the electron
resonance Raman measurements of samples result in Raman cloud at frequency nv . The electron cloud oscillation at n0
vibrational intensities that are roughly proportional to con- couples to the vibrationally induced electron cloud oscilla-
centrations of those species in the sample; the Raman-active tion at nv to give rise to a beat oscillation at n0 nv , which
vibrations have roughly similar intensities. For example, radiates Raman scattered light at this frequency. Those
aromatic ring vibrations have intensities similar to those of vibrations that most efficiently Raman scatter are those that
aliphatic CC stretching vibrations, which have intensities couple most effectively to the oscillating electric dipole
similar to those of CH stretching vibrations. moment induced by the exciting electromagnetic field.
In contrast, UV-Raman spectroscopy uses selective exci- There are particular excitation frequencies that are natural
tation in the UV absorption bands of molecules to produce frequencies of oscillation of specific electron oscillators of
spectra of particular analytes and chromophoric segments the molecular electron cloud. These natural frequencies are
of macromolecules. Resonance excitation has the advan- the molecular electronic absorption band frequencies. Exci-
tage of selectivity. In addition, UV-Raman measurements tation at these frequencies is said to be in resonance with
of condensed-phase samples excited below 260 nm are not the electronic transition (Figure 2); therefore, the Raman
plagued by fluorescent interferences.12 scattering is said to be resonance Raman scattering. This
resonance excitation at the natural frequency of electron
2 PHENOMENOLOGY cloud oscillation results in an increased oscillating charge
displacement and a corresponding increase in the induced
Figure 3 illustrates the Raman scattering phenomenon. An dipole moment; this, in turn, results directly in an increased
incident electromagnetic field drives the electron cloud of scattering or reradiation efficiency for Raman scattering.
the molecule at the incident frequency n0 . An oscillating The enhancement factor of resonance Raman scattering
dipole moment is created by the resulting displacement of compared with that of normal Raman scattering can be as
electronic charge. Because the charge is accelerating, it high as 108 .
radiates energy in the form of electromagnetic radiation. The observed Raman band intensities at a particular
The radiated frequency n0 is identical to the excitation excitation frequency depend on the degree to which a vibra-
frequency; light is scattered elastically. If the scattering tion modulates the molecular polarizability. Classically, the
species is small compared with the wavelength of light, the magnitude of the induced dipole moment m depends linearly
scattering is known as Rayleigh scattering. Because charge on the molecular polarizability a and the incident electro-
acceleration induces the radiation, the Rayleigh scattering magnetic field E, giving m D aE. In the case of Raman
efficiency increases with the fourth power of the excitation scattering, the effective polarizability is not the static polar-
frequency. This phenomenon explains why the sky is blue izability a; it is the Raman polarizability aR associated with
and the setting (or rising) sun is red. perturbation of the static polarizability by the Raman active
Other dynamic processes modulate the induced oscil- vibration aR D @a/@Q0 Q, where Q is the normal vibra-
lating dipole moment. As shown in Figure 3, a nuclear tional mode displacement. For excitation at frequency n0 ,
the Raman intensity Imn observed over a unit solid angle
for a transition between vibronic levels m and n in the
electronic ground-state manifold is13 Imn D sR NI0 W,
an
where sR (cm2 per molecule sr1 ) is the total differential
m
Molecule Ra Raman cross-section for a single gas-phase molecule inte-
)
grated over the Raman peak bandwidth, I0 is the incident

Incident v
photon 0
excitation intensity (photons cm2 s1 ) into a particular
h( sample volume element of area dA and length dl, N is
v
h0 the number of molecules within that volume element, and
Ra
yle W is a parameter that details the optical geometry and
igh
Vibrational
mode h includes factors such as the collected solid angle and the
0 spectrometer efficiency.
The resonance Raman effect can lead to important selec-
Figure 3. Raman scattering phenomenon. The presence of a tivity in Raman spectral measurements. Figure 4 shows
vibration at frequency nv is shown by the spring. Light scattered the absorption spectrum of a sample with two absorption
at n0 nv is known as Stokes Raman scattering; light scattered at
n0 C nv is known as anti-Stokes Raman scattering. [Reprinted with bands. Raman spectra are shown that are excited at a wave-
permission from S.A. Asher, Anal. Chem., 65(2), 61A (1993). length longer than any absorption band (normal Raman)
Copyright (1993) American Chemical Society.] or within the two absorption bands (resonance Raman).
chromophoric segment of a single molecule, the spectra will

(a)
differ because of differences in the structure of the excited
states associated with the two electronic transitions; the dif-
I ferent excited states couple differently with the ground-state
(b) vibrational motion.
(c)
We can easily estimate the number of non-resonance
Raman photons that can be detected from a 0.1 M solution

Absorbance
I I of an analyte with a Raman cross-section of 1 b per

molecule sr1 , illuminated by an incident laser beam of 1 W
of 514.5 nm light (2.6 1018 photons s1 ). For a typical
case, the limiting aperture is the height of the intensified

Reticon detector (2.5 mm), and the spectrometer f/#. Let
us assume the spectrometer is f/7.0, with a slit width of
200 m, and the collection optic has a magnification of 6.
For this case, the laser beam can be efficiently imaged
into a column with a 33-m (maximum) diameter and a
417-m height that contains 2.2 1013 molecules. Ideally,
we can collect 0.86 sr of the scattered light. According to
Wavelength the equation, if the spectrometer were 100% efficient, a
Figure 4. Selectivity available from resonance Raman spec-
total of 5.6 106 Raman photons s1 would be detected
troscopy. The absorption spectrum is shown for a sample with for this vibration.
two absorption bands. These bands could derive from two differ- In contrast, the fluorescence cross-section of an analyte
ent analytes, two different chromophoric segments of a macro- with a molar absorptivity of 104 and a fluorescence quantum
molecule, or two different electronic transitions of a single yield of 0.1 is equal to 3 1019 cm2 per molecule sr1 ;
molecule. Three different Raman spectra are shown that derive
from resonance Raman excitation within either of the two absorp- 2 1018 fluorescence photons would be observed using
tion bands (a and b) or from excitation away from the absorption the same instrument (assuming no self-absorption of the
(c). Excitation within resonance leads to much more intense emitted light occurs).
Raman spectra. The most intense feature is the Rayleigh scat- Obviously, the fluorescence experiment is much more
tering band. [Reprinted with permission from S.A. Asher, Anal. sensitive than the Raman experiment and we can observe
Chem., 65(2), 62A (1993). Copyright (1993) American Chem-
ical Society.] the same amount of fluorescence signal as the Raman
signal using a concentration 3 1011 -fold lower. How-
Excitation away from resonance results in comparable ever, in the case of resonance Raman scattering, Raman
Raman intensities from all analytes. The typical scatter- cross-sections as high as 1022 cm2 per molecule sr1
ing cross-section is sR D 1030 cm2 per molecule sr1 (or can occur. The resonance Raman sensitivity, while always
1 b per molecule sr1 , where 1 b equals 1024 cm2 ). Res- smaller, can approach within a 1000-fold that of fluo-
onance excitation within an absorption band can lead to an rescence. Furthermore, the combination of selectivity and
immense Raman intensity increase. For example, we have sensitivity can be greater because the Raman spectra have
measured a Raman cross-section of 50 b per molecule higher resolution.
sr1 for pyrene.14 Generally, numerous Raman bands exist, and all can be
Figure 4 shows an increased Raman intensity occurs for used to determine concentrations. The narrow Raman spec-
resonance excitation. Different Raman spectra are observed tral bands carry a great deal of information on molecular
with excitation in resonance versus not in resonance. In structure, in contrast to the broad fluorescence emission.
addition, different spectra are observed with resonance exci- Also, the Raman cross-sections are much less dependent
tation within different sample absorption bands. If these on environment than are the fluorescence quantum yields.
absorption bands derive from different analytes in the sam- Modest changes in solvent composition can cause dramatic
ple, we selectively enhance the vibrational Raman spectrum changes in fluorescence quantum yields, because the quan-
of different analytes as we change the excitation wave- tum yields are very sensitive to relatively small changes
length. The selectivity factor can be as large as 108 . in fluorescence lifetimes. Thus, resonance Raman measure-
If the different absorption bands derive from different ments can be very sensitive, the spectra can provide a great
segments of a molecule, by changing excitation wave- deal of molecular structural and environmental information,
lengths we change the segment of the molecule studied. In and an increased selectivity results from the ability to excite
contrast, if the two absorption bands derive from the same particular species.
3 INSTRUMENTATION The major difficulty with these laser sources is that they
are large and expensive. A new hollow cathode UV laser
UV-Raman instrumentation has evolved rapidly. The orig- has recently been demonstrated that utilizes sputtered silver
inal laser source used a low repetition rate (20 Hz) neo- or copper in a noble gas to give rise to 224-nm and 248-nm
dymium doped yttrium aluminum garnet (Nd : YAG) laser excitation.24 The output is quasi continuous at 1 kHz with
with 5-ns frequency pulses that were doubled or tripled to a duty cycle of 1%. The advantage of this laser is that it
pump a dye laser whose frequency was doubled or mixed is small and inexpensive (< $10 K), it is air-cooled and it
with the Nd : YAG 1.06-m fundamental wavelength to gen- has a low power consumption.
erate light between 200 and 400 nm.15,16 The minimum UV-Raman sample handling requires flowing the sample
average output power available in this region is 20 mW; in through the laser beam as a liquid jet, through a capillary, or
some spectral regions, hundreds of milliwatts are available. along a waveguided stream. With pulsed lasers, it is essen-
Raman shifting of the Nd : YAG harmonics in H2 gas can tial to exchange each illuminated sample volume before the
also generate numerous wavelengths in the UV, and this arrival of the next laser pulse to avoid sample heating; also,
approach can be used to construct a simpler, inexpensive the high-energy UV photons might cause photochemical
UV laser source. degradation in the samples. Most samples studied to date
The low-duty cycle (107 ) of the Nd : YAG laser repre- have not shown significant monophotonic photodegradation
sents a major limitation because the laser pulses cannot be during the typical 10-min Raman measurement time.
focused into a sample volume that can be efficiently imaged The collection optics use either an ellipsoidal collection
into the Raman spectrometer. The Nd : YAG output occurs mirror15,16 for achromatic imaging or quartz lenses (when
as 5 ns pulses, at a repetition rate of 20 Hz. An aver- accurate relative intensity measurements are not essential).
age power of 20 mW corresponds to a peak power level The typical spectrometer is a triple grating instrument in
of 200 kW, which can induce nonlinear optical phenom- which the initial double monochromator operates in a sub-
ena in matter. A power flux value of 2.5 GW cm2 would tractive configuration to filter out the Rayleigh scattered
occur if this beam were focused into a 100-m-diameter light. The third-stage spectrograph disperses and images
spot size. the light onto a multichannel detector such as an intensi-
A partial solution to this low-duty cycle is to use a higher fied Reticon array. This multichannel detector is essential
repetition rate YAG laser and to Raman shift it into the for simultaneous collection of the entire spectrum. A scan-
UV. The high temporal and special mode structure of the ning system would have a far inferior S/N because most of
Coherent Infinity 100 Hz YAG lasers results in extremely the noise comes from the low-stability UV laser source.
efficient harmonic generation and H2 Raman shifting. Intensified charge-coupled device (CCD) detectors have
Another approach is to use high repetition rate excimer proven to be superior to the intensified Reticon detec-
lasers to directly pump a dye laser whose output can be fre- tors.
quency doubled into the UV region.17 High-power excimer The triple spectrometer is inefficient; at best, only 7%
lasers can have repetition rates between 200 and 500 Hz; of the Raman scattered light entering the entrance slit is
at the highest repetition rates the 16-ns laser pulses result transferred to the detector. We recently developed a highly
in a duty cycle of 8 106 . This partially alleviates the efficient new single grating UV-Raman spectrometer, which
nonlinear optical phenomena and the sample ground-state utilizes dielectrically coated mirrors and a dielectrically
depletion18 that both complicates the Raman spectral mea- coated long pass filter to remove the Rayleigh scattered
surements and necessitates defocusing the laser excitation light.24 The spectrometer utilizes a solar blind intensified
beam at the sample. An alternative approach uses a mode- CCD detector to optimize signal to noise. We further
locked Nd : YAG laser to create a quasi-continuous wave developed an efficient UV-Raman microscope spectrometer
(CW) UV source with a duty cycle of 0.01.19,20 A new tun- for studying small sample areas.25
able quasi-CW UV laser source has been demonstrated that
utilizes a titanium sapphire laser that is frequency quadru-
pled to generate light tunable from 205 to 230 nm.21 4 FUNDAMENTAL APPLICATIONS
True CW UV laser excitation can now be obtained by
intracavity doubling of Ar and Kr lasers. The intracavity Ar Investigations of the structure and dynamics of excited
laser gives five excitation wavelengths in the UV region: states of small molecules are an especially fertile area
257, 248, 244, 238, and 228.9 nm,22 while an intracavity for UV-Raman investigations. Examination of the vibra-
doubled Kr laser gives 206-nm light.23 These lasers give tions most enhanced by excitation in resonance with an
rise to a major improvement in spectral S/N and, for the first electronic transition gives information on the excited-
time, allow UV-Raman studies of solid absorbing samples. state structure and dynamics. The simplest rule of thumb
is that resonance Raman-enhanced vibrations are those Figure 5 compares the theoretically calculated and the
that distort the ground-state geometry toward the excited- measured UV-Raman spectrum for the 13 Ca isotope of
state geometry. If the atomic motions involved in the NMA. The major features include a weak 1642 cm1 shoul-
enhanced vibrations are known, theoretical models can be der for the amide I vibration, strong 1582 and 1311 cm1
applied to the measured Raman intensities to calculate the bands for the amide II and III vibrations, and a moderate-
excited-state distortion. Numerous workers are using this intensity 1371 cm1 CH3 unbrella mode band. The other
approach to examine the excited states of small molecules differences observed derive from the lack of an internal
such as oxygen,26 ammonia,27 methyl iodide,28,29 carbon standard band at 932 cm1 and the lack of a cis NMA
disulfide,30 ethylene,31 33 and other species. component at 1495 cm1 .
For methyl iodide,28,29 the excited state is dissociative This theoretical study found that the amide excited state
and the Raman data give information on the short time was mainly expanded along the C(O)N bond, with a
dynamics of the photodissociation. For ethylene, excita- smaller expansion along the CDO bond and with smaller
tion in the lowest allowed p ! p transition results in the contractions along the CC(O) and NC bonds. In the case
unique enhancement of the overtone of a torsional vibration, of diglycine derivatives they discovered that the lowest-
in addition to the CDC double-bond stretch and the CH2 energy electronic transition was a charge transfer transi-
bending vibration; the enhancement of this torsional vibration from the carboxylate to the amide bond.37 39 This
tion confirms that the ethylene excited state is twisted.32,33 accounted for the anomalous enhancement of carboxylate
Similar experiments examine the vibrational enhancement stretching vibrations for the penultimate carboxylates of
of stilbenes,32,33 for example, to determine the photochem- proteins and peptides.
ical mechanism of photoisomerization. Interested readers
are urged to examine the rapidly expanding literature on
these insightful studies. 5 ANALYTICAL APPLICATIONS
UV-Raman spectroscopy has been applied extensively
to biological molecules, especially to probe the struc- The ability of UV resonance Raman spectroscopy to selec-
ture and function of proteins. Early applications explored tively examine the vibrational spectra of particular species
the excited states of aromatic amino acids, nucleic acids in complex mixtures makes the technique uniquely impor-
and amides.34 For example, recent studies examined the tant for analytical applications. Selectivity is determined
excited states of amides in order to determine the ori- by the relative Raman cross-sections of the analyte com-
gin of the resonance Raman enhancement mechanism.34 36 pared with other species present in the sample. Sensitivity
To investigate the prototypical amide bond, Asher et al. depends on the magnitude of the analyte Raman cross-
examined the resonance Raman enhancement mechanism sections compared with those of overlapping, interfering
in N-Methylacetamide (NMA). Raman bands and emission from the sample. Fluorescence
interference does not normally occur in condensed phases
with UV excitation below 260 nm; those species with their
lowest singlet states below 260 nm have vanishingly low
1582
932
fluorescence quantum yields because they are flexible and

1311
return to the ground state through nonradiative processes.12

1496 cis Am II
Raman intensity
Polycyclic aromatic hydrocarbons (PAHs) are the most

intense Raman scatterers.40 43 The UV-Raman spectra of
1371
1642
a series of PAHs in Figure 6 show numerous bands that

are indicative of fused-ring systems. The vibrations that
1421
621
879
are enhanced are mostly symmetric in-plane ring-breathing

vibrations. The spectra, though similar, differ enough to
identify the ring system. Although aliphatic ring sub-
stituents perturb the spectra, the enhanced vibrations remain
500 1000 1500 2000 predominantly ring-breathing modes; peripheral substituent
vibrations are not generally observed unless they are conju-
Wavenumber /cm1
gated into the aromatic p network.40 The major changes that
Figure 5. Comparison of the measured and calculated 218 nm occur upon ring substitution involve alterations in relative
excited spectrum of trans 13 CH3 CONHCH3 in water (42 mM).
See text for details. [Reprinted with permission from S.A. Asher, intensities and frequency shifts, which depend on the sub-
J. Am. Chem. Soc., 117(10), 2893 (1995). Copyright (1995) stitution patterns in the different compounds. The Raman
American Chemical Society.] cross-sections are large enough that it is easy to study these
analytes at parts per million (ppm) and parts per billion
1400
(ppb) and even parts per trillion (ppt) concentrations.40,43
The large cross-sections of these compounds has allowed
the Stair group to use UV-Raman to study coke formation
753
in catalysts such as ZSM-5 and USY zeolites.44,45 They
1562
1010
1257
were able to selectively examine the formation of PAHs
1155
(a) and demonstrated the selective formation of PAHs on USY
1563
1398
1354
755
zeolites.
1012
1258
1599
Similarly the Asher group demonstrated the ability to
1151
545
(b) selectively study non-diamond carbon in chemical vapor

1337
deposition (CVD) diamond-grown films.23,46 48 They con-

structed a UV-Raman microscope to examine the spacial
1457
distribution of non-diamond carbon in these CVD films.

696
1059
They found that the non-diamond carbon localized at

1392
1226
1607
1752
(c) the interstices between diamond crystallites.25 The Raman

spectra of diamond show high S/Ns with UV excitation
1624
below 229 nm. Excitation at 206 nm showed strong first-

1562 1594
order and higher-order Raman bands. In addition, strong

1405
narrow luminescent bands from defects were also observed.

1367
1237
1028
1066
1495
571
589
The intensities were sufficiently high that monolayer cov-

802
(d)
erage of diamond could be observed on a silicon substrate.
1611
Intensity
On the basis of the high signal-to-noise spectra the Asher

group constructed a UV-Raman spectrometer to examine in
1480
1579
situ the growth of diamond in a microwave plasma CVD

1294
1155
1343
1232
1015
1094
843
740
diamond reactor (Figure 7). A 1.5-kW ASTeX microwave

633
(e)
1629
plasma reactor was used for CVD diamond growth. The in

situ UV-Raman spectra were excited with a 244-nm light
1577
beam, which was expanded and focused onto the growth

763
1371
1036
substrate/sample inside the reactor.

1145
1241
(f) 255
Gas flow controller
(g)
235
CW UV laser
(h)
Sample
Stage Array
220 detector Spectrograph
400 800 1200 1600 2000

Wavenumber /cm1
Figure 6. Resonance Raman spectra of 5 mM solutions of PAHs
dissolved in acetonitrile and excited at 255 nm. (a) Anthracene.
(b) Phenanthrene. (c) Triphenylene. (d) Pyrene. (e) Fluorene. Stack
Napthalene excited at (f) 255 nm, (g) 235 nm, and (h) 220 nm. filters
[Reprinted with permission from S.A. Asher, Anal. Chem., 65(4) Figure 7. Schematic diagram showing the UV-Raman spectrom-
(1993). Copyright (1993) American Chemical Society.] eter coupled to the ASTex plasma CVD diamond growth chamber.
The scattered light was collimated by using a parabolic spectra for relatively short 10 min spectral accumulation
mirror. Two 244-nm dielectric stack filters were used to times. We calculate that the detection limit for the growing
reject Rayleigh scattering and the Raman scattered light CVD diamond films is a film of 0.8 nm thickness.
was reimaged onto the slit of a modified Spex 1701, single The spectra display the well-known frequency decrease
monochromator and detected by an intensified photodiode of the 1332 cm1 first-order phonon band with temper-
array. Approximate confocal imaging was used to help min- ature. We were surprised to see no non-diamond carbon
imize interference from the plasma emission in the reactor. impurity bands in these spectra since they were clearly
Although the plasma emission intensity in the visible is suf- observed in spectra of these same CVD films when they
ficiently high to prevent Raman measurements, it decreases were cooled down to room temperature. We discovered that
dramatically in the UV spectral region at 260 nm. these non-diamond carbon bands are enhanced by a nar-
Figure 8 shows the temperature dependence of the row electronic resonance whose frequency is temperature
first-order Raman band of the growing diamond films dependent. This electronic resonance shifts away from this
(1332 cm1 at room temperature). We see very high S/N 244-nm excitation wavelength at elevated temperatures.
The use of UV-Raman for remote detection of chemical
1321.4
species has recently been demonstrated by the construction
of a UV-Raman Lidar system for studying organic species
400 000
remotely in the atmosphere and on surfaces.49,50 This Lidar
system used a telescope and 266-nm excitation from a
200 000 quadrupled YAG laser. They demonstrated detection limits
(a) of 500 g m2 at distances of 0.5 km.
0
As evident from the work above, UV-Raman spec-
1315.1 troscopy is a highly sensitive and selective spectral probe of
200 000 dilute species. The power of this approach derives from the
fact that the majority of species have strong UV absorption
100 000
bands and show strong resonance enhancement. Further,
the UV-Raman spectra show high S/Ns due to the lack of
(b)
0 luminescence interference for UV excitation below 250 nm.
1308.9
200 000
6 BIOLOGICAL ULTRAVIOLET-RAMAN
100 000
STUDIES
(c) Major advances have occurred in UV-Raman measurements
0
of biologically important molecules and for biological
800 000 1307.9 assemblies such as viruses, bacteria and algae. Over the
600 000 last decade the resonance Raman enhancement profiles have
been almost completely characterized for aromatic amino
400 000
acids, peptide and protein amide vibrations, nucleic acids,
200 000 DNA and visual pigments.
(d)
0 For example, excitation of tyrosine, tryptophan and
phenylalanine roughly follows their UV absorption bands
600 000 1305.7 (Figures 9 and 10). Figure 9 shows the UV-Raman bands
400 000 of tryptophan and tyrosine. All of the bands observed are
symmetric ring in-plane breathing modes.
200 000 Most UV-Raman aromatic amino acid studies occur with
(e) excitation within the 220-nm S2 absorption band of tyrosine
0
and the S3 absorption band of tryptophan, since excitation in
1000 1500 2000 2500 their longer wavelength bands at 270 nm results in weaker
Raman shift /cm1 enhancement and interference from fluorescence.
Figure 8. Temperature dependence of the in situ 244-nm UV- The lifetimes of the S1 excited state of tyrosine and the
Raman spectra of CVD diamond films grown within the plasma S1 /S2 excited states of tryptophan are in the 10 ns time
reactor: (a) 578 C; (b) 766 C; (c) 930 C; (d) 955 C; (e) 1008 C. regime. The Asher group used the fact that nanosecond
3.0
760 cm1
0.0
1550
222 nm
2.0
1009 cm1
1009
760
1345
932 ClO4
0.0
Raman cross-section
(b per molecule sr 1)
2.0
877
1345 cm1
Raman Intensity
0.0
248 nm
4.0
1550 cm1
0.0
1614
1.1
Absorbance
1550
0.0
600 1000 1400 1800 190 210 230 250
(a) Wavenumber /cm1 (b) Wavelength /nm
Figure 9. (a) Resonance Raman spectra of tryptophan excited at 222 and 248 nm. The 932 cm1 band derives from the internal standard.
Obviously, the relative intensity compared with the internal standard band increases with 222 nm excitation. This is more evident from
(b) the resonance Raman excitation profiles of the vibrational modes of tryptophan. The experimental data (diamonds) are red-shifted
4 nm from the absorption maximum of the Bb transition. The broken curve represents the basic Condon transform calculation of the
RREP. The solid curve is the result of introducing non-Condon coupling into the transform calculation. [Reprinted with permission
from S.A. Asher, J. Phys. Chem., 94(12), 47856 (1990). Copyright (1990) American Chemical Society.]
resonance laser pulse excitation occurs in a timescale pulse energy monitors depletion of the ground state and
similar to that required for the excited state to return can determine the tyrosine and tryptophan T1 ground state
back to the ground state to develop a technique called recovery rates. These Raman saturation measurements were
Raman saturation spectroscopy to study the photophysics able to measure picosecond and nanosecond Forster trans-
of tyrosine and tryptophan residues.59 61 fer rates from tyrosine to tryptophan in peptides, and to the
Nanosecond UV pulsed laser excitation simultaneously heme group in heme proteins. These measurements comple-
excites the Raman spectra and transfers ground-state ment fluorescence lifetime measurements, and are unique in
molecules into their excited states. As a result, the Raman that they can measure tyrosine excited-state relaxation, in
intensities increase sublinearly with the excitation energy. the common cases where tyrosine does not fluoresce, but
Thus, the dependence of the Raman intensity on the incident transfers its energy to tryptophan.
1006 (W16)
1611 (Y8a)
1553 (W3)
753 (W18)
1172 (Y9a)
1353 (W7)
.0
874 (W17)
250
.0 pH
245
.0 10
24 0 1.50
0
235. 0
gth
10
Raman intensity
.0 3.11
230
elen
0.0 0
Wav
.0 270
225 .0
300
2
0.0 1 4.05
.0 190 10
220 cm
150
0.0 b er / 0
0 ve num
00.
11 Wa
.0 4.80
(a) 215 10
0
1.0 1618 cm1
and
1601 cm1 7.12
Absorbance Raman cross-section (1024/cm2 per molecule sr1)
10
0.5 0
600 800 1000 1200 1400 1600 1800
(a) Raman shift /cm1
0.0
1183 cm1
(b per molecule sr1)
Raman cross-section
0.5 1.5
1.0
0.0
850 cm1
0.5
0.5
0.0
1.00 3.00 5.00 7.00
0.0 (b) pH
Figure 11. (a) 229-nm excited Raman spectra of horse aquomet-
myoglobin at various pH values.65,66 (b) pH dependence of
229-nm excited myoglobin tryptophan band intensities. The
500 cm1 water band was used as an internal standard to deter-
mine the Raman cross-sections. , W18 759 cm1 , 0.574; ,
W17 879 cm1 , 0.130; , W16 1012 cm1 , 0.509; X, W31
210 230 250 551 cm1 , 0.383 b per molecule sr1 . [Reprinted with permission
from S.A. Asher, Biochemistry, 37(9), 2868 (1998). Copyright
(b) Wavelength / nm (1998) American Chemical Society.]
Figure 10. (a) Zwitterionic tyrosine (pH 6.0) Raman spectra
excited between 217 and 250 nm. The strongest band is the studies of tyrosine and tryptophan Raman cross-sections
1620 cm1 ring breathing mode. (b) Zwitterionic tyrosine (pH 6.0)
and their vibrational frequencies can monitor exposure of
absorption spectrum and the total differential Raman cross-section
excitation profiles. [Reprinted with permission from S.A. Asher, the aromatic ring side chain to the aqueous environment62
J. Am. Chem. Soc., 110(4), 10078 (1998). Copyright (1998) and the hydrogen bonding of these residues.63
American Chemical Society.] For example, Figure 11(a) shows the dependence of the
229-nm excited Raman spectra of horse aquomyoglobin as
Numerous UV-Raman studies have taken advantage of a function of pH.64 66 As the pH decreases, the intensity
the strong, selective enhancement of the tyrosine and tryp- of the tryptophan bands at 1611, 1353, 1006, 879 and
tophan in proteins with 230-nm excitation. UV-Raman 753 cm1 decreases (Figure 11b). This signals unfolding of
the A helix, which contains the two myoglobin tryptophan Linear fitting of these basis spectra to the observed
residues. The cross-section changes are consistent with protein and peptide Raman spectra can be used to determine
the exposure to the tryptophan residues to the aqueous the protein and peptide solution secondary structure. UV-
environment. The tyrosine Raman cross-sections do not Raman is now arguably the most sensitive method to
change, indicating a lack of conformational changes of the determine dilute solution secondary structure.
helices that contain the tyrosine residues. Combining the information obtained from 206-nm exci-
Excitation of proteins and peptides with 205-nm excitation of the amide bands with information obtained from
tation selectively excites the peptide backbone amide vibra- 229-nm excitation of the aromatic amino acid bands
tions (Figure 12). With great similarity to the enhance- makes it possible to spatially resolve unfolding of par-
ment in NMA (Figure 5), the protein spectra show large ticular segments of proteins and peptides. For example,
enhancements for the amide II and III bands and weaker a 206-nm study of the acid denaturation of myoglobin65
enhancement of the amide I band. By examining a library and apomyoglobin66 determined the pH dependence of
of proteins we calculated64 the Raman spectra contributed the protein a-helical content, while a study at 229 nm
by the pure a-helix, b-sheet and random coil conformations found that the tyrosine environment remained intact as the
(Figure 13). protein unfolded. In contrast, in myoglobin both tryptophan
These basis spectra demonstrate strong enhancement of residues (on the A helix) became fully solvent exposed with
the amide III (1267 cm1 ), amide II (1560 cm1 ), Ca -H a pKa of 3.5 (Figure 11). These data demonstrated that only
bending (1386 cm1 ) and amide I (1665 cm1 ) in the the G and H helix remained intact by pH 2. In contrast, for
unordered or random coil form. The amide III band is apomyoglobin this study demonstrated a three-stage acid
downshifted in the b-sheet form, while the Ca -H bending unfolding where the A helix underwent two-stage melting.
band is broadened. The amide III band is upshifted and the These results were used to create a model for myoglobin
Ca -H bending mode disappears in the a-helix conformation, denaturation where helices in the heme pocket melted upon
which shows a strong amide I band at lower frequency. removal of the heme in apomyoglobin. As the pH decreased
Am II
Am II
Am III
Am III
Am I
Am I
C-H
C-H
h. Chy
1 a. Con
1
0 0
b. Trypsin i. Elas
1
1
0
c. RNase
0
j. Lyso
0.4
Raman intensity
0 1
d. Cyt c
0
0.4 k. Carb
0 1
e. BSA
0
0.4
l. Trio
0
f. Hb 1
0.4 0
m. Ins
0
g. Mb 0.5
0.4
0
0
1200 1400 1600 1800 1200 1400 1600 1800
Raman shift /cm1
Figure 12. 206.5-nm excited measured protein Raman spectra at pH 7.0 water used for modeling the spectra and their residuals
calculated as the difference between the measured and calculated spectra.
R1 Pump beam, 1.9 m

Unordered
P1
1560
1267
1386
1665
2.0 Y
A P2
G R2
1.0
FC
Probe
0.0 beam,
1647
-Helix SP 204 nm
Raman intensity
1545
1299
0.5 DL
0.0
Figure 14. The T-jump spectrometer consists of a Nd : YAG laser
1551
-Sheet
(YAG), two H2 Raman shifters (R1 and R2), a thermostated flow
1654
1235
cell sample circulator (FC), and a subtractive double monochro-

mator with a solarblind ICCD detector (SP). The heating pump
1.0
1386
beam is obtained by Raman shifting the YAG fundamental in H2

0.5 to 1.9 m (first Stokes). The probe beam is obtained by Raman
shifting the YAG third harmonic to 204 nm (fifth anti-Stokes). The
0.0 time delays between the pump and probe pulses are provided by
the variable delay line (DL).
1200 1400 1600 1800
Raman shift /cm1 1.06 m fundamental of the YAG was Stokes shifted one
Figure 13. Calculated pure secondary structure Raman spectra harmonic in hydrogen to generate 1.9 m IR light, which
(PSSRS) obtained from aqueous solution 206.5-nm protein Raman was absorbed by water to generate a T-jump. This T-jump
spectra. The regions around 1610 and 1450 cm1 are excluded was followed by a delayed 204-nm pulse to excite the
due to interference from aromatic amino acid ring vibrations and
amide vibrations. The 204-nm pulse was generated by shift-
side-chain vibrations.
ing the 355-nm third harmonic of the YAG 5 anti-Stokes
the A helix melted in two stages and by pH 2 the structure harmonics in hydrogen (Figure 14). The first applications
of myoglobin and apomyoglobin became identical. of this transient spectrometer was to examine the dynam-
Transient UV-Raman measurements are a powerful tool ics of peptide unfolding of the 21 amino acid residue
for studying protein dynamics. An important biological tar- mainly polyalanine peptide AP. This peptide is mainly
get for these studies is the development of a molecular level a-helical at room temperature and unfolds completely by
understanding of the mechanism of hemoglobin oxygen 50 C to the random coil form. Figure 14 shows the tem-
binding. Hemoglobin enhances oxygen transport by altering perature dependence of the 204-nm UV-Raman spectra.
the heme ligand affinity of one oxygen binding subunit in The spectra evolve with temperature, showing a downshift
response to oxygen binding by another subunit. The Spiro67 in the amide III frequency and an increased 1400 cm1
and Kitagawa groups34 used nanosecond visible excita- Ca H bending band intensity. These changes could have
tion to partially photolyze the carbon monoxy ligand off been analyzed with the basis spectra of Figure 13 to cal-
the heme in cabonmonoxy hemoglobin. They then exam- culate the peptide secondary structure at any delay time.
ined changes in the aromatic amino acid sidechain spectra However, we utilized a-helix and random coil basis spec-
with delayed nanosecond 230-nm excitation pulses. These tra we independently calculated for AP (Figure 15). These
studies were able to resolve hydrogen bonding changes of AP a-helix and random coil basis spectra are much nar-
a functionally important tyrosine residue at the hemoglobin rower than those of the proteins since AP is almost an
subunit interface that appears to serve as the conduit for oligopeptide. Steady-state UV-Raman measured melting
actuating the protein cooperative structural changes that curves accurately determined the temperature dependence
alter the hemoglobin subunit ligand affinities. of the a-helical composition; UV-Raman measurements
The Asher group has been utilizing UV-Raman to probe appear to be more linear in a-helix composition than is
the early stages of protein and peptide unfolding. They CD spectroscopy.
developed a T-jump transient UV-Raman technique for The transient UV-Raman spectra (Figure 15) indicate
studying the first steps in protein folding.17,68 Two YAG that negligible conformational changes occur at short delay
lasers were used to excite peptide and protein solutions. The (<20 ns). The short time spectral changes were shown to
The peptide has been isotopically labeled at each amino

Am II
Am III acid along the chain by deuterating the Ca position. The
amide III band is significantly blue-shifted for the deuter-
C-H ated derivative, while the Ca bending vibration is absent.
Am I
Thus, the difference spectra between the selectively labeled
AP peptide and the fully hydrogenated peptide selectively mon-
itors the secondary structure and dynamics of the labeled
70 C
1 position. Future studies will examine the dynamics of these
2 30 C
particular sites due to unfolding induced by a T-jump. The
3
results of these studies will test the existing theoretical
5.5 C
Raman intensity
models, which predict the dynamics and mechanisms of

A5-A3
a-helix melting.
0 C
Numerous other proteins have recently been subjects
4
of UV-Raman investigations. An important group is the
visual pigments. These species, such as rhodopsin and
Random coil bacteriorhodopsin, were extensively investigated by visual
(calc.) Raman measurements with the objective of determining
5 5.5 C
the photochemistry of the retinal pigment. Further inves-
6 -Helix tigations using UV-Raman are examining the coupling
(calc.) of retinal photochemistry with the aromatic amino acid
T-jump environments.69,70
4 C 64 C While this review has emphasized structural and dynami-
95 ns
7 cal studies of proteins, numerous other studies have focused
8 4 C on nucleic acids and DNA.71 The field is rapidly expanding
with numerous reports appearing monthly.
1000 1200 1400 1600 1800
A number of groups are attempting to use UV-Raman
Wavenumber /cm1
to examine more complex biological systems such as
Figure 15. Comparison between steady-state and time-resolved cells, viruses and tissues. For example, the Nelson group
204-nm UVRS of AP (15 mg mL1 ). AP UVRS measured
at (1) C70 C, (2) C30 C, and (3) C5.5 C. (4) The 0 C UV- has examined the utility of UV-Raman to study micro-
Raman difference spectrum between penta-alanine and tri-alanine; organisms,71,72 while the Thomas group has examined DNA
(5) pure random coil (T D 5.5 C); and (6) a-helix (T D and proteins in viruses.73 Sureau et al.,74 Richards-Kortum
5.5 C) basis Raman spectra of AP (see text). Transient UVRS et al.75 and Feld et al.71,76 have attempted to differentiate
of AP measured at (7) 95 ns after a T-jump of 60 C and (8) a cancerous from noncancerous cells.
static UVRS of AP at 4 C in the absence of a T-jump.
result only from a temperature jump of the static AP initial

conformations. The AP peptide unfolds at room temperature 7 CONCLUSIONS
with a 200-ns relaxation time.
The measured temperature dependences of the relaxation Obviously, the field of UV-Raman is flourishing. The major
rates were combined with the measured temperature depen- limitation to the use of UV-Raman is the high cost of
dences of the equilibrium constants for unfolding in order instrumentation, especially the laser sources. Further, until
to calculate the Arrhenius activation barriers. The unfold- most recently, UV-Raman instruments were commercially
ing activation barrier energy was 8 kcal mol1 , a value unavailable.
that is similar to that determined by other methods. In However, continuing advances in laser technology are
contrast, the folding barrier showed a negative activation rapidly decreasing the cost of UV excitation sources and
barrier, which clearly indicates that even this simple peptide commercial instruments are emerging. The field of UV-
system (where amide conformations change only between Raman is expected to dramatically expand its scope as
a-helix and random coil states) cannot be considered as a more fundamental and applied studies demonstrate its high
simple two-state equilibrium. In fact, the energy landscape information content. Resonance Raman spectroscopy is
is highly complex and a more sophisticated view of the extraordinarily sensitive to molecular structural details since
unfolding is required. it directly probes the coupling of molecular vibrations to
In fact, UV-Raman measurements are providing a much the frontier electrons involved in the UV electronic transi-
more detailed examination of the unfolding mechanism. tions of a molecule. The selectivity of resonance excitation
allows the spectroscopist to focus on the dynamics of 11. B. Chase, Anal. Chem., 59, 881A (1987).
particular frontier electrons of the analyte species of inter- 12. S.A. Asher and C.R. Johnson, Science, 225, 311 (1984).
est. Resonance enhancement allows studies of low concen- 13. J.M. Dudik, C.R. Johnson and S.A. Asher, J. Chem. Phys.,
trations of analyte. The ultimate limitation in UV-Raman 82, 1732 (1985).
measurements is the resistance of the sample to photochem- 14. C.M. Jones and S.A. Asher, J. Chem. Phys., 89, 2649 (1988).
istry induced by the UV excitation. To date this has not 15. S.A. Asher, C.R. Johnson and J. Murtaugh, J. Rev. Sci.
posed an insurmountable problem to any UV-Raman inves- Instrum., 54, 1657 (1983).
tigation to my knowledge. 16. L.D. Ziegler and B.S. Hudson, J. Chem. Phys., 74, 982
(1981).
17. I.K. Lednev, A.S. Karnoup, M.C. Sparrow and S.A. Asher,
ACKNOWLEDGMENTS J. Am. Chem. Soc., 121, 8074 (1999).
18. C.M. Jones, V.L. DeVito, P.A. Harmon and S.A. Asher,
I am pleased to acknowledge the excellent work of the post- Appl. Spectrosc., 41, 1268 (1987).
doctorate and graduate students who have developed UV- 19. K.P.J. Williams and D.J. Klenerman, Raman Spectrosc., 23,
Raman in my laboratories. I would also like to acknowledge 191 (1992).
the long-term support of NIH in the development of UV- 20. T.L. Gustafson, Opt. Commun., 67, 53 (1988).
Raman through grant GM30741. 21. X. Zhao, R. Chen, C. Tengroth and T.G. Spiro, Appl. Spec-
trosc., 53, 1200 (1999).
22. S.A. Asher, R.W. Bormett, X.G. Chen, D.H. Lemmon,
ABBREVIATIONS AND ACRONYMS N. Cho, P. Peterson, M. Arrigoni, L. Spinelli and J. Cannon,
Appl. Spectrosc., 47, 628 (1993).
CVD Chemical Vapor Deposition 23. J.S.W. Holtz, R.W. Bormett, Z. Chi, N. Cho, X.G. Chen,
CW Continuous Wave V. Pajcini, S.A. Asher, L. Spinelli, P. Owen and M. Arri-
Nd : YAG Neodymium Doped Yttrium goni, Appl. Spectrosc., 50, 1459 (1996).
Aluminum Garnet 24. M.C. Sparrow, J. Jackovitz, C.H. Munro, W. Hug and S.A.
NMA N-Methylacetamide Asher, Appl. Spectrosc., 55, 66 (2001).
PAH Polycyclic Aromatic Hydrocarbon 25. V. Pajcini, C.H. Munro, R.W. Bormett and S.A. Asher, Appl.
PSSRS Pure Secondary Structure Raman Spectra Spectrosc., 51, 81 (1997).
26. Y.P. Zhang and L.D. Ziegler, J. Phys. Chem., 93, 6665
(1989).
REFERENCES 27. L.D. Ziegler, J. Chem. Phys., 86, 1703 (1987).
28. D.L. Phillips and A.B. Myers, J. Chem. Phys., 95, 226 (1991).
1. D.A. Long, Raman Spectroscopy, McGraw-Hill, New York
29. D.L. Phillips, A.B. Myers and J.J. Valentini, J. Phys. Chem.,
(1977).
96, 2039 (1992).
2. S.A. Asher, Ann. Rev. Phys. Chem., 39, 537 (1988). 30. B. Li and A.B. Myers, J. Chem. Phys., 94, 2458 (1991).
3. I. Harada and H. Takeuchi, Raman and Ultraviolet Reso- 31. X. Ci and A.B. Myers, Chem. Phys. Lett., 158, 263 (1989).
nance Raman Spectra of Proteins and Related Compounds,
in Spectroscopy of Biological Systems, eds R.T. Clark and 32. R.J. Sension and B.S. Hudson, J. Chem. Phys., 90, 1377
R.E. Hester, J. Wiley & Sons, New York (1986). (1989).
4. B. Hudson and L. Mayne, Methods Enzymol., 130, 331 33. L.D. Ziegler and B.S. Hudson, J. Chem. Phys., 79, 1134
(1986). (1983).
5. P.A. Harmon and S.A. Asher, J. Chem. Phys., 88, 2925 34. S.A. Asher, Anal. Chem., 65, 59A, 201A (1993).
(1988). 35. X.G. Chen, R. Schweitzer-Stenner, S. Krimm, N. Mirkin and
6. F.A. Miller and G.B. Kauffman, J. Chem. Educ., 66, 795 S.A. Asher, J. Am. Chem. Soc., 117, 2884 (1995).
(1989). 36. S.A. Asher, Z. Chi and P. Li, J. Raman Spectrosc., 29, 927
7. T.C. Strekas and T.G. Spiro, Biochim. Biophys. Acta, 263, (1998).
830 (1972). 37. X.G. Chen, P. Li, J.S.W. Holtz, Z. Chi, V. Pajcini, S.A. Asher
8. T.G. Spiro, G. Smulevich and C. Su, Biochemistry, 29, 4497 and L.A. Kelly, J. Am. Chem. Soc., 118, 9705 (1996).
(1990). 38. V. Pajcini and S.A. Asher, J. Am. Chem. Soc., 121, 10 942
9. S.A. Asher, Methods Enzymol., 76, 372 (1981). (1999).
10. R. Mathies, S.O. Smith and I. Palings, in Biological Appli- 39. V. Pajcini, X.G. Chen, R.W. Bormett, S.J. Geib, P. Li, S.A.
cations of RS, ed. T.G. Spiro, J. Wiley & Sons, New York, Asher and E.G. Lidiak, J. Am. Chem. Soc., 118, 9716 (1996).
Vol. II (1987). 40. S.A. Asher, Anal. Chem., 56, 720 (1984).
41. C.R. Johnson and S.A. Asher, Anal. Chem., 56, 2258 (1984). 59. J. Teraoka, P.A. Harmon and S.A. Asher, J. Am. Chem. Soc.,
42. S.A. Asher and C.M. Jones, UV Resonance Spectroscopy: 112, 2892 (1990).
A New Technique for Speciation of Aromactics in Complex 60. P.A. Harmon, J. Teraoka and S.A. Asher, J. Am. Chem. Soc.,
Matrices, in New Applications of Analytical Techniques to 112, 8789 (1990).
Fossil Fuels, eds M. Perry and H. Retcofsky, ACS Sym-
61. J. Sweeney, P.A. Harmon, S.A. Asher, C.M. Hutnik and
posium Series 31, American Chemical Society, Washington,
A.G. Szabo, J. Am. Chem. Soc., 113, 7531 (1991).
DC, 170180 (1986).
62. Z. Chi and S.A. Asher, J. Phys. Chem. B, 102, 9595 (1998).
43. R. Rummelfanger, S.A. Asher and M.B. Perry, Appl. Spec-
trosc., 42, 267 (1988). 63. M. Matsuno and G. Takeuchi, Bull. Chem. Soc. Jpn., 71(4),
44. C. Li and P.C. Stair, Catalysis Today, 33, 353 (1997). 851 (1998).
45. Y.T. Chua and P.C. Stair, J. Catal., 196, 66 (2000). 64. Z. Chi, X.G. Chen, J.S.W. Holtz and S.A. Asher, Biochem-
istry, 37, 2854 (1998).
46. R.W. Bormett, S.A. Asher, R.E. Witkowski, W.D. Partlow,
R. Lizewski and F. Pettit, J. Appl. Phys., 77, 5916 (1995). 65. Z. Chi and S.A. Asher, Biochemistry, 37, 2865 (1998).
47. C.D. Zuiker, A.R. Krauss, D.M. Gruen, J.A. Carlisle, L.J. 66. Z. Chi and S.A. Asher, Biochemistry, 38, 8196 (1999).
Terminello, S.A. Asher and R.W. Bormett, Mater. Res. Soc. 67. D. Wang, X. Zhao and T.G. Spiro, J. Phys. Chem. A, 104,
Symp. Proc., 437, 211 (1996). 4149 (2000).
48. V.I. Merkulov, J.S. Lannin, C.H. Munro, S.A. Asher, V.S. 68. I. Lednev, A. Karnoup, M. Sparrow and S.A. Asher, J. Am.
Veerasamy and W.I. Milne, Phys. Rev. Lett., 78, 4869 (1997). Chem. Soc., 123, 2388 (2001).
49. M.D. Ray, A.J. Sedlacek and M. Wu, Rev. Sci. Instrum., 71, 69. G.G. Kochendoerfer, S. Kaminaka and R.A. Mathies, Bio-
3485 (2000). chemistry, 36, 13 153 (1997).
50. M. Wu, M. Ray, K.H. Fung, M.W. Ruckman, D. Harder and 70. S. Hashimoto, K. Obata, H. Takeuchi, R. Needleman and
A.J. Sedlacek, Appl. Spectrosc., 54, 800 (2000). J.K. Lanyi, Biochemistry, 36, 11 583 (1997).
51. S.A. Asher, M. Ludwig and C.R. Johnson, J. Am. Chem. Soc., 71. E.B. Hanlon, R. Manoharan, T.-W. Koo, K.E. Shafer, J.T.
108, 3186 (1986). Motz, M. Fitzmaurice, J.R. Kramer, I. Itzkan, R.R. Dasari
52. C.R. Johnson, M. Ludwig and S.A. Asher, J. Am. Chem. Soc., and M.S. Feld, Phys. Med. Biol., 45, 1 (2000).
108, 905 (1986).
72. E. Ghiamati, R. Manoharan, W.H. Nelson and J.F. Sperry,
53. M. Ludwig and S.A. Asher, J. Am. Chem. Soc., 110, 1005 Appl. Spectrosc., 46, 357 (1992).
(1988).
73. Z.Q. Wen and G.J. Thomas Jr, Biochemistry, 39, 146 (2000).
54. J. Sweeney and S.A. Asher, J. Phys. Chem., 94, 4784 (1990).
74. F. Sureau, L. Chinsky, C. Amirand, J.P. Ballini, M. Du-
55. C.R. Johnson, M. Ludwig, S. ODonnel and S.A. Asher, J. quesne, A. Laigle, P.Y. Turpin and P. Vigny, Appl. Spec-
Am. Chem. Soc., 106, 5008 (1984). trosc., 44, 1047 (1990).
56. C. Su, Y. Wang and T.G. Spiro, J. Raman Spectrosc., 21, 75. Y. Yazdi, N. Ramanujam, R. Lotan, M.F. Mitchell, W. Hit-
8179 (1990). telman and R. Richards-Kortum, Appl. Spectrosc., 53, 82
57. S. Fodor, R. Copeland, C. Grygon and T.G. Spiro, J. Am. (1999).
Chem. Soc., 111, 5509 (1989). 76. N.N. Boustany, J.M. Crawford, R. Manoharan, R.R. Desara
58. R. Rava and T.G. Spiro, J. Phys. Chem., 89, 1856 (1985). and M.S. Feld, Lab. Invest., 79(10), 1201 (1999).
Introduction to Nonlinear Raman Spectrometry
A.M. Zheltikov
M.V. Lomonosov Moscow State University, Moscow, Russia
1 INTRODUCTION approach and first experiments to a widely accepted tool

for numerous engineering applications and routine measure-
Nonlinear Raman spectroscopy brings us to the domain of ments in many areas of modern physics, chemistry, and
nonlinear optics, where the properties of a medium are no biology. The main ideas and approaches in nonlinear Raman
longer independent of the intensity of incident light, and spectroscopy are now described in comprehensive and illu-
the light, in its turn, is modulated by vibrations coher- minating books and reviews.1 7
ently excited in the medium. Historically, it was a nonlinear This article provides a brief introduction to the main prin-
Raman process, namely stimulated Raman scattering (SRS), ciples of nonlinear Raman spectroscopy, giving a general
that was among the first nonlinear optical effects discov- idea of how the measurements are performed and how the
ered at the dawn of the laser era. (Another nonlinear optical spectroscopic data are extracted from the results of these
effect of the same order of nonlinearity resulting in the measurements. Following these, a very brief introduction
appearance of an electric-field-induced birefringence in an will be given to the basic concepts of nonlinear Raman spec-
initially isotropic medium quadratic in the applied field troscopy. Then, various modifications of coherent Raman
was discovered by J. Kerr in the 19th century, i.e. long FWM spectroscopy will be considered, including the stan-
before the advent of lasers. The Raman-induced Kerr effect dard CARS scheme, SRS, RIKE, degenerate four-wave mix-
is one of the efficient methods of nonlinear spectroscopy, ing (DFWM), and coherent hyper-Raman scattering. We will
which is now widely employed for the spectroscopic inves- also briefly describe polarization techniques of nonlinear
tigation of Raman-active media.) Many other methods of Raman spectrometry and coherent ellipsometry, allowing
nonlinear Raman spectroscopy making use of four-wave selective investigation of multicomponent molecular and
mixing (FWM) processes involving coherent excitation of atomic systems and permitting the sensitivity of nonlinear
Raman-active vibrations in nonlinear media were soon Raman spectrometry to be radically improved. Finally, in
developed. Some of these techniques, such as coherent the context of the growing interest in the applications of
anti-Stokes Raman scattering (CARS) and coherent Stokes short-pulse spectroscopy for the investigation of ultrafast
Raman scattering (CSRS), were based on measuring the processes, we will provide an introduction to time-resolved
amplitude of Raman-shifted signals, while other approaches, nonlinear Raman spectroscopy. Some aspects of the histori-
cal perspective of nonlinear Raman spectroscopy and the
such as coherent polarization ellipsometry and spectroscopy
outlook for the future will also be discussed.
based on the Raman-induced Kerr effect (RIKE), implied
the measurements of polarization characteristics of Raman-
shifted signals (see Appendix 1 for the glossary of acronyms
2 THE BASIC PRINCIPLES
and abbreviations that have grown out of the nonlinear
Raman field). The years that followed witnessed an impres- Nonlinear Raman spectroscopy is based on the processes
sive progress of nonlinear Raman spectrometry, which has of nonlinear optical interactions in Raman-active media.
traveled a long way from a conceptually new physical The nonlinear character of light interaction with a medium
implies that molecular, atomic, or ionic vibrations in a
John Wiley & Sons Ltd, 2002. medium are no longer independent of the light field. Instead,
1 solution of Maxwell equations for the field of the signal.

Below, we restrict our brief introduction to the theory of
3
1 2
a I (a),,, nonlinear Raman processes to the description of the basic
2 notions and terminology, which will be employed later
to explain the main concepts of various nonlinear Raman
methods, including frequency- and time-domain CARS,
3
coherent ellipsometry, and RIKE.
Figure 1. Coherent excitation and probing of Raman-active
modes by the method of nonlinear Raman spectrometry: pump 2.1 Excitation of Raman vibrations
light waves with frequencies !1 and !2 modulate Raman-active
vibrations at frequency in the medium (!1 !2 ), which
Within the framework of a simple, but physically instruc-
are then probed by the third laser beam (generally with a
frequency !3 ). tive, semiclassical model of a nonlinear medium consisting
of noninteracting molecules with a Raman-active vibration
pump light waves with frequencies !1 and !2 modulate with frequency , the interaction of light with molecules
Raman-active vibrations in a medium at the frequency (atoms or ions) can be described in terms of the electronic
!1 !2 , which can then be probed with another light polarizability of molecules a depending on the generalized
beam, generally having a frequency !3 (Figure 1). This normal coordinate Q (e.g. defined as the distance between
wave-mixing process involving the inelastic scattering of the nuclei in a molecule):1
the probe wave by molecular vibrations gives rise to coher-
@a
ent Stokes- and anti-Stokes-frequency-shifted signals (the aQ D a0 C Q C 1
@Q 0
process of CARS is shown in Figure 1), whose ampli-
tude I, polarization (the ellipticity c and the tilt angle y of where a0 is the equilibrium polarizability of a molecule,
the principal axis of the polarization ellipse), and phase j @a/@Q0 is the derivative of the electronic polarizability
carry the spectroscopic information concerning the medium in the normal coordinate taken for the equilibrium posi-
under study. This is the general idea of nonlinear Raman tion of nuclei, and we restrict our consideration to terms
spectrometry illustrated by Figure 1. in this expansion linear in Q. The term @a/@Q0 Q in
In practical terms, to undertake a simple three-color CARS equation (1) is responsible for the modulation of light by
experiment, one generally needs three laser sources generat- molecular vibrations, as it gives rise to new frequency com-
ing radiation with the frequencies meeting the requirements ponents in the induced polarization of the system, whose
specified above. In the most popular scheme of two-color frequency shift is determined by the frequency of molecu-
CARS, where the anti-Stokes signal is generated through lar vibrations. This can be seen from the expression for the
the frequency-mixing scheme !a D 2!1 !2 , the number polarization of a medium:
of lasers required is reduced to two. The light beams have to P D Np, 2
be brought into coincidence in space to excite Raman-active
transitions in a medium and to generate the anti-Stokes sig- where N is the number density of Raman-active molecules
nal. Introducing some delay time between pumping and
@a
probing pulses, one can also perform time-resolved CARS p D aQE D a0 E C QE C 3
@Q 0
measurements to keep track of the temporal dynamics of
excitation in the system under investigation. is the dipole moment of a molecule, and E is the electric
While spontaneous Raman scattering often suffers from field strength.
the low quantum yield, which eventually results in the loss The energy of a molecule in a light field is written as
of sensitivity, coherent Raman scattering allows much more H D pE D aQE2 4
intense signals to be generated, thus allowing very high
sensitivities to be achieved. Coherent Raman scattering also Thus, the light-induced force driving molecular vibrations
offers several other very important advantages that stem is given by
@H @a 2
from the coherent character of the signal, which is thus well FD D E 5
collimated and generated in a precisely known direction. @Q @Q
Generally, the analysis of the amplitude, phase, and As can be seen from equation (5) the force acting on
polarization of the signal resulting from a nonlinear wave- a molecule in a light field may result in a resonant exci-
mixing process in a Raman-resonant medium involves the tation of Raman-active vibrations with a frequency if
calculation of the nonlinear response of the medium in the field involves frequency components !1 and !2 with
terms of the relevant nonlinear susceptibilities and the !1 !2 .
Introduction to Nonlinear Raman Spectrometry 3
Obviously, equations (15) do not provide a detailed anti-Stokes frequency !a is given by1
understanding of all the properties of nonlinear Raman 2
processes, giving only a general idea of the physical N @a
P3 !a D EE1 E2 9
mechanism of such processes. A more detailed, quantum- 4MD, !1 !2 @Q 0
mechanical description of nonlinear Raman processes where
can be found in the classical textbooks on nonlinear
optics.1,2,8 10 D, !1 !2 D 2 !1 !2 2 2i!1 !2
derivatives are taken in the equilibrium position, is the
phenomenologically introduced damping constant, M is the
2.2 Wave mixing in Raman-active media
reduced mass of a molecule, N is the number density of
molecules, an asterisk indicates a complex conjugate, Q is
The propagation of light waves in a nonlinear medium is
the amplitude of the Raman-active molecular vibration,
governed by the wave equation
which can be expressed in terms of the density matrix r
E
n2 @2 E E nl
4p @2 P of an ensemble of molecules,1,9
E
E D 6
2
c @t 2 c @t2
2
Q D Sp rq D rab qba 10
where EE is the field in the light wave, n is the refractive
E nl is the nonlinear where Sp is the trace operator and q is the operator of the
index, c is the speed of light, and P
vibrational coordinate.
polarization of the medium.
Introducing the third-order nonlinear optical susceptibi-
Consider the CARS process, as an example of a non-
lity of the medium,
linear Raman process, assuming that it involves plane and
2
monochromatic waves, N @a
c3R D 11
E 1 expi!1 t C ikE1 Er C E
E r , t D E E 2 expi!2 t C ikE2 Er 24MD, !1 !2 @Q 0
EE
Er C E and solving equation (8), we arrive at the following expres-
E expi!t C ikE
CE E a expi!a t C ikEa Er
sion for the intensity of the CARS signal Ia :
C c.c. 7 2
3 2 2 sinkl/2
Ia / jc !a ; !, !1 , !2 j II1 I2 l 12
where !1 and !2 are the frequencies of pump waves, kl/2
! is the frequency of the probe wave, !a D ! C !1 !2 , where I, I1 , and I2 are the intensities of the pump and probe
the field envelopes E, E E E 1, E
E 2 , and E
E a are slowly varying
beams, l is the length of the nonlinear medium, and k D
functions of coordinates Er and time t, k, E kE1 , kE2 , and kEa
jkj D jka k k1 C k2 j is the wave-vector mismatch.
are the wave vectors of light waves with frequencies !, As can be seen from equation (12), the anti-Stokes wave
!1 , !2 , and !a , respectively, and c.c. stands for a complex is generated especially efficiently in the direction of phase
conjugate. Then, representing the nonlinear polarization in matching, where k D 0.
the medium as a superposition of plane waves, we can write Generally, the cubic nonlinear optical susceptibility of
the equation for the amplitude of the anti-Stokes wave as a medium is represented as a fourth-rank tensor c3 ijkl .
Ea
na @E @EEa 2p!a The knowledge of the form of this tensor is important, in
C Di E nl !a expika z
P 8 particular, for understanding the polarization properties of
c @t @z cna
the signal of nonlinear Raman scattering. The form of the
3
where PE nl !a , na , and ka are the amplitude of the nonlinear cijkl tensor is determined by the symmetry properties of a
polarization, the refractive index, and the z-component of medium. In particular, in an isotropic medium, only 21 of
the wave vector at the frequency !a . The pump wave 81 tensor components of c3 ijkl are nonvanishing. Only three
amplitudes will be assumed to be constant. Now, the of these components are independent of each other, as the
nonlinear polarization of a medium has to be found with relations1,9
the use of a model of the nonlinear medium. 3 3 3
c1111 D c2222 D c3333
3 3 3 3 3 3
c1122 D c1133 D c2211 D c1122 D c3311 D c3322
2.3 Nonlinear polarization and nonlinear
3 3 3 3 3 3
susceptibilities c1212 D c2121 D c1313 D c3131 D c2323 D c3232 13
3 3 3 3 3 3
c1221 D c2112 D c1331 D c3113 D c2332 D c3223
Within the framework of our model of a nonlinear medium,
3 3 3 3
the third-order nonlinear polarization of the medium at the c1111 D c1122 D c1212 D c1221
hold true for an isotropic medium. The form of the c3 Equations (14) and (15) show that the information on
ijkl
tensor (i, j, k, l D x, y, z D 1, 2, 3) for all the crystallo- the nonlinear cubic susceptibility in the CARS signal may
graphic classes can be found in many books on nonlinear be distorted by phase-matching effects. This problem was
optics.1,10 12 analyzed both theoretically and experimentally for different
modifications of nonlinear Raman spectrometry.14 17 The
influence of absorption at the wavelength of pump and
2.4 Phase matching probe waves and at the frequency of the FWM signal can be
taken into account by including the imaginary parts of the
Generally, in the case of focused beams, phase-matching relevant wave vectors. The integral in equation (15) can
effects in nonlinear Raman scattering are taken into account be calculated in an analytical form for several particular
through the phase-matching integral. In particular, for Gaus- cases, giving a clear physical understanding of the role of
sian beams, the expression for the overall power of the phase-matching and absorption effects in coherent FWM
two-color CARS signal occurring in accordance with the spectroscopy and imaging.16,17
scheme !a D 2!1 !2 is written as1,13 15
2 2 1
3p !a 3 2 2 32 b 2 3 THE MAIN SCHEMES OF NONLINEAR
Pa D jc j P1 P2 3 2prjJj2 dr
c 2 na p w03 RAMAN SPECTROSCOPY
0
14
Here, P1 and P2 are the powers of pump beams with In this section, standard and widely-used schemes of
frequencies !1 and !2 , respectively, w0 is the waist size nonlinear Raman spectroscopy will be briefly considered
of the focused pump beams, b is the confocal parameter, (Figure 2), including coherent Raman and hyper-Raman
C2 scattering, SRS, and RIKE, and a brief introduction into
expr 2 /bH
JD dx0 15 a vast area of DFWM will be given.
1 C ix0 k 00 ik 0 x0 H
C1
where C1 and C2 are the coordinates of the boundaries

of the nonlinear medium,
3.1 Stimulated Raman scattering
1 C x02 x0 z 00 SRS, i.e. a nonlinear Raman process where a high-power
HD i , k D 2k1 C k2 , k 0 D 2k1 k2
k 00 ik 0 x0 k0 laser beam generates new Raman-shifted spectral compo-
z D 2z f/b is the normalized coordinate along the nents in a Raman-active medium due to the scattering of
z-axis (f is the coordinate of the beam waist along the pump radiation by coherently excited Raman-active vibra-
z-axis). tions in the medium, was discovered in 1962 by Woodbury
2
1 2 3 CARS
1 2 3 FWM
2 DFWM
(a) 1 (b) 1 (c) 1

pump2()
CARS(+)
pump2()
DFWM()
pump1(+)
pump1()
CSRS()
pump3()
pump1()
pump2()
pump3()
pump3()
(d)
Figure 2. The main schemes of nonlinear Raman spectroscopy: (a) CARS; (b) FWM with a hyper-Raman resonance; (c) DFWM; and
(d) DFWM/CARS/CSRS with femtosecond pulses.
and Ng,18 who observed the generation of a new Raman- high-resolution molecular spectroscopy,7,30 39 and other
shifted frequency component in a Raman-active liquid irra- applications.
diated by ruby-laser light. The widespread use of frequency-domain CARS as a
The idea of using SRS as a spectroscopic technique is method of spectroscopy of Raman-active media has been
based on the measurement of the frequency dependence vigorously promoted by the demonstration of the capa-
of the SRS small-signal gain, which is proportional to the bilities of CARS with frequency-tunable lasers. The first
imaginary part of the nonlinear cubic susceptibility of a experiments carried out in this direction by Akhmanov
Raman-active medium.1,9 The power of the pump wave et al.40 have employed an optical parametric oscillator
in such measurements has to be chosen in such a way as (OPO) as a frequency-tunable source. Radiation of an
to avoid uncontrollable instabilities and to obtain notice- Nd : YAG laser with a wavelength of 1.06 m and radiation
able SRS gain. The SRS-based approach was successfully produced by a lithium niobate-based OPO, whose wave-
employed, in particular, for high-resolution spectroscopy of length was tuned around 0.95 m, were used as pump waves
Raman transitions.19,20 in the CARS scheme, while the second harmonic of the
Limitations of SRS as a spectroscopic technique are due Nd : YAG laser was used as a probe. With this approach,
to instabilities arising for light intensities exceeding the CARS spectra of a CaCO3 crystal were detected and dis-
threshold SRS intensity. These instabilities build up under persion properties for different components of the cubic
conditions when several nonlinear processes, including self- susceptibility tensor of calcite were measured (Figure 3).
focusing and self-phase modulation, compete with each To emphasize the common nature of various modifications
other, often rendering the SRS method impractical for of stimulated-scattering spectroscopy, Akhmanov et al.40
spectroscopic applications. called the spectroscopic approach they implemented active
Raman scattering spectroscopy (ARSS).
The widely employed geometry of nonlinear Raman
3.2 Coherent anti-Stokes Raman scattering measurements implies the use of collinear focused laser
beams (Figure 4a). While focusing allows the intensity
Instead of measuring the gain of one of two waves, as it sufficient to ensure a reliable detection of the nonlinear
is done in SRS, Maker and Terhune21 proposed to extract Raman signal to be achieved, the collinear geometry of
spectroscopic information on Raman-active transitions by wave mixing increases the length of nonlinear interaction.
measuring the frequency dependence of the intensity of However, such an approach is reasonable as long as the
a new wave generated at the anti-Stokes frequency !a D spatial resolution along the propagation coordinate is not
2!1 !2 in the presence of two light waves with frequen- important. The scheme of nonlinear-Raman spectroscopy
becomes resolvable along the propagation coordinate as
cies !1 and !2 chosen in such a way as to meet the condi-
tion of a Raman resonance with a Raman-active transition in
a medium: !1 !2 (Figure 2a). This approach, called 106
coherent anti-Stokes Raman scattering, became one of the a
most widespread nonlinear Raman methods, allowing many
urgent spectroscopic problems to be solved successfully
IARSS (arbitrary units)
and stimulating numerous engineering applications of non- 104

linear laser spectroscopy (the so-called three-color CARS
with !a D !1 !2 C !3 is shown in Figure 2a). Similar to
the SRS process described in the previous section, CARS b
involves the stimulated scattering of light in a Raman-active
medium. However, in contrast to the standard SRS scheme, 102
where a Stokes wave is generated or amplified, the CARS
process gives rise to the appearance of a new frequency
component, suggesting a spectroscopic approach that is free
of instabilities typically arising in SRS due to the competi-
100
tion of different nonlinear processes. 700 1100 1450
Due to its high spatial, temporal, and spectral resolution, [(1 2)/2c ] (cm1)
the possibilities of studying highly luminous objects, and a
Figure 3. Spectra of different components of the nonlinear opti-
rich variety of polarization methods, the CARS technique cal susceptibility tensor c3 of a CaCO3 crystal measured by
has gained a wide acceptance for temperature and con- Akhmanov et al.40 with the use of polarization-sensitive ARSS:
centration measurements in gases,3 6,22 29 gas analysis,1 (a) jc3 2 3 2
1122 j ; and (b) jc1123 j .
k1 k2 (Figure 2b). Four-photon processes with Raman-type reso-

nances are not always convenient for diagnostics of early
stages of laser-plasma expansion because of the consid-
(a) k3 kFWM erable lowering of the FWM signal intensity due to the
decrease in the difference of populations in the levels
kFWM involved in a Raman resonance, which usually have close
k1 energies. FWM with hyper-Raman resonances seems to be
advantageous in such a situation. Amplitude and polariza-
tion properties of the nonlinear Raman signal are governed
k2
by invariants of the tensor of hyper-Raman scattering in this
(b) k3 case. Hyper-Raman-resonant FWM can be also efficiently
employed for two-dimensional mapping of relative popu-
kFWM lations of excited-state atoms and ions in low-temperature
k1
plasmas.47,48
k2
3.4 Raman-induced Kerr effect

k3
RIKE9 is understood as an optical birefringence induced
(c) in an initially isotropic medium due to an anisotropic
Figure 4. Diagrams of light beams in (a) collinear, (b) noncol- Raman-resonant third-order polarization of a medium. In
linear, and (c) folded broad-beam coherent FWM of light beams this scheme, a nonlinear medium is irradiated with a pair of
with wave vectors k1 , k2 , and k3 . light beams with frequencies !1 and !2 whose difference
is tuned, in accordance with the general idea of probing
soon as collinear beams are replaced by a noncollinear Raman-active vibrations, to a resonance with Raman-active
one (Figure 4b). In CARS, this technique is called the transitions in a medium. Then, the polarization of a probe
BOXCARS geometry.41 The interaction area is confined wave at the frequency !1 becomes perturbed due to the
in this case to the region where the beams intersect each anisotropic nonlinear polarization induced in the Raman-
other, allowing a high spatial resolution to be achieved. active medium, which can be detected with the use of a
The next step is to abandon using focused beams by polarization analyzer and a detector. The RIKE technique
replacing them with broad or sheetlike beams (Figure 4c). provides us with a convenient method for measuring the fre-
This approach allows the nonlinear Raman signal to image quency dependence of the cubic susceptibility c3
ijkl around
whole areas of a nonlinear medium on a charge-coupled a Raman resonance. The transmission coefficient, T, of a
device (CCD) camera. The idea of broad-beam CARS polarization analyzer aligned in such a way as to block the
imaging, which was discussed by Regnier and Taran22 in probe beam in the absence of the pump beam is given by1
1973, has proved to be efficient for the solution of many
problems of gas-phase and plasma diagnostics. Signifi- Tl !1 !2 / sin2 gjc3
1122 !1 ; !2 , !1 , !2
cant progress in extracting the data concerning parameters
C c3 2 2
1221 !1 ; !2 , !1 , !2 j I2 16
of a gas medium was achieved, allowing CARS signals
from molecules of different types to be simultaneously in the case of a linearly polarized pump (I2 is the intensity
detected, with the development of a dual broadband CARS of the pump beam and g is the angle between the polariza-
scheme42,43 and angularly resolved CARS.44,45 tion vectors of the pump and probe waves) and
Tc !1 !2 / jc3R
1122 !1 ; !2 , !1 , !2
3.3 Four-wave mixing with a hyper-Raman c3R 2 2
1221 !1 ; !2 , !1 , !2 j I2 17
resonance
in the case of a circularly polarized pump. As can be seen
In an FWM scheme with a hyper-Raman resonance,46 pump from equation (17), the coherent background is completely
fields with frequencies !1 , !2 , and !3 , chosen in such a suppressed in the case of a circularly polarized pump.
way that their combination !1 !2 C !3 is close to a one- Importantly, the RIKE is one of the FWM processes
photon-allowed transition in a nonlinear medium, give rise where the phase-matching condition is satisfied automa-
to an FWM signal at the frequency !FWM D !1 !2 C !3 tically regardless of the arrangement of the wave vectors of
pump and probe waves, since the phase-matching condition permits one to separately measure the real and imagi-
k1 D k2 C k1 k2 becomes an identity in this case. nary parts of the relevant third-order nonlinear optical
susceptibility,56 58 to resolve closely spaced lines in FWM
spectra of molecules1,59 and atoms,46 to increase the con-
3.5 Degenerate four-wave mixing trast of the dispersion curve corresponding to the cubic
nonlinear optical susceptibility near Raman resonances,60,61
Although, rigorously speaking, DFWM does not employ to analyze the interference of vibrational Raman resonances
Raman transitions and the models used to describe DFWM with one- and two-photon electronic resonances in CARS
may sometimes differ from the standard ways of CARS spectra,62 to determine invariants of atomic and molecular
description,9,10 it seems reasonable to take at least a very Raman and hyper-Raman scattering tensors,1,63 to measure
quick look at this rapidly growing area of modern FWM the characteristics of Raman resonances and to perform con-
spectroscopy, as the DFWM technique very frequently formational analysis for complex organic molecules.64 The
offers a useful alternative to CARS, allowing valuable methods of polarization-sensitive four-photon spectroscopy
data on a medium to be obtained in a convenient and have been actively developed for more than a quarter of a
physically clear way. A close relation between DFWM century.65 70 A comprehensive review of polarization tech-
and CARS is due to the fact that both processes are niques employed for molecular spectroscopy was provided
associated with the third-order nonlinearity of a medium, by Akhmanov and Koroteev.1
while the main difference between these techniques is that
CARS implies the use of a two-photon Raman-type reso-
nance (Figure 2a), while DFWM is a frequency-degenerate 4.1 Polarization properties of the coherent
process (Figure 2c), involving either four one-photon reso- four-wave mixing signal
nances or a pair of two-photon resonances. However, with
modern lasers capable of generating very short pulses, hav- When analyzing polarization properties of the FWM signal,
ing large spectral widths, the DFWM signal can be detected one has to take into account the interference of reso-
simultaneously with CSRS and CARS in the same experi- nant components of FWM related to various (molecular
mental geometry with the same molecular system by simply or atomic) transitions in the medium and the nonresonant
tuning the detection wavelength49 (Figure 2d). The combi- coherent background. In particular, it is the interference
nation of these nonlinear optical approaches allows a more of the resonant FWM component with the nonresonant
elaborate study of molecular relaxation and photochemistry coherent background that ensures the possibility to record
processes, providing a much deeper insight into the ultrafast complete spectral information concerning the resonance
molecular and wave-packet dynamics.49,50 under study, including the data on the phase of resonant
The main advantages of DFWM as a spectroscopic tech- FWM.
nique are associated with the fact that this approach requires The polarization ellipse of a Raman-resonant FWM sig-
only one laser source and allows phase-matching conditions nal is characterized by its ellipticity c (which is defined
to be automatically satisfied regardless of the dispersion as c D arctanb/a, where b and a are the small and
of the medium under study. Broadband DFWM51 makes principal semiaxes of the polarization ellipse, respectively)
it possible to measure the temperature of excited gases, and the tilt angle y of its principal axis (Figure 5a). These
including atomic gases,52 with a single laser pulse. Folded parameters are related to the Cartesian components of the
broad-beam DFWM schemes are employed in several con- third-order polarization of a medium Px and Py by the fol-
venient and elegant methods for two-dimensional imaging lowing expressions:1
of spatial distributions of gas parameters.53 55
tan2y D tan2b cos
sin2c D sin2b sin 18
4 POLARIZATION TECHNIQUE
where b and are defined as
Methods of polarization-sensitive four-photon spectroscopy
provide an efficient tool for the solution of many prob- Py
tanb D
lems arising in the investigation of Raman resonances. In Px
particular, the polarization technique is a standard method D argPy argPx 19
to suppress the coherent background in CARS measure-
ments, which makes it possible to considerably improve the In contrast to the spectrum of the FWM intensity, the
sensitivity of four-photon spectroscopy.1 Under certain con- frequency dependences of the parameters of the FWM
ditions, the polarization technique in FWM spectroscopy polarization ellipse, as can be seen from equations (18)
y In the case of an isotropic medium, the relations shown

Pr
in equation (13) are satisfied for both resonant and nonres-
onant components of the nonlinear optical susceptibility.
However, only the nonresonant part of the cubic suscepti-
bility satisfies the Kleinman relations,1,9,10
x
Pnr 3NR 3NR 3NR 3NR
a b c1111 D 3c1122 D 3c1221 D 3c1212 21
(a) while the resonant part of the cubic susceptibility is usually

characterized by a considerable dispersion near a Raman
resonance, which implies that the resonant cubic suscep-
Pr tibility tensor components are not invariant with respect
to the permutation of their frequency arguments. Taking

e2 into account the relations in equation (21) for the non-
Pnr
resonant part of the cubic susceptibility, we arrive at the
P following expression for the nonresonant and resonant com-
ponents of the third-order polarization of a Raman-active
medium:
e1
E nr D c3NR
P e1 Ee1 Ee2 C Ee2 Ee1 Ee1 ]
1111 [2E 22
(b)
Figure 5. Polarization technique of nonlinear Raman spectrome-
Er D
P 3c3R
1111 [1 rN Ee1 Ee1 Ee2 C rN Ee2 Ee1 Ee1 ] 23
try: (a) coherent ellipsometry; and (b) polarization suppression of
the nonresonant background in CARS spectroscopy. where rN D c3R 3R
1221 /c1111 , and E
e1 and Ee2 are the unit polariza-
tion vectors of the light fields with frequencies !1 and !2 ,
and (19), provide information concerning the phase of respectively (Figure 5b).
the resonant FWM component, allowing a broad class of Suppose that the pump fields with frequencies !1 and
phase measurements to be performed by means of nonlinear !2 are linearly polarized and their polarization vectors Ee1
Raman spectroscopy. and Ee2 are oriented at an angle m with respect to each
other, as shown in Figure 5(b). The vectors P E nr and PE r,
in accordance with equations (22) and (23), generally have
4.2 Suppressing the nonresonant background different orientations in space making a nonzero angle q
with each other. Therefore, by orienting the polarization
Polarization suppression of the nonresonant background in analyzer in such a way as to suppress the nonresonant
CARS is one of the most useful, practical, and widely component of the CARS signal (by setting the angle e
measured from the direction perpendicular to the vector P E nr
employed polarization techniques in nonlinear Raman spec-
troscopy. Physically, the possibility of suppressing the non- equal to zero, see Figure 5b), one can analyze background-
resonant background in coherent Raman spectroscopy is free CARS spectra.
due to the fact that the resonant and nonresonant compo- In many situations, nonlinear Raman study of Raman-
nents of the nonlinear polarization induced in a Raman- active media would be simply impossible without this tech-
nique. This is the case, for example, when the CARS signal
active medium are generally polarized in different ways.
from a resonant gas under investigation is too weak to be
Let us illustrate this technique for the CARS process !a D
reliably detected against the nonresonant CARS signal from
2!1 !2 , where !a is the frequency of the anti-Stokes sig-
the windows of a gas cell. Another example is the CARS
nal and !1 and !2 are the frequencies of pump waves, in
spectroscopy on low concentrations of complex biological
the case of an isotropic medium. The third-order polariza-
molecules, when the coherent background due to solvent
tion responsible for the generation of the signal with the
molecules may be so strong that it leaves no way to detect
frequency !a is then written as
the CARS signal from the molecules being studied with-
E 3 D [P
P E nr ]E21 E2
Er C P 20 out polarization suppression of the nonresonant nonlinear
Raman signal. Even small deviations of the orientation of
where E1 and E2 are the amplitudes of the light fields, and the polarization analyzer from the background-suppression
E r and P
P E nr are the resonant and nonresonant components, position may be crucial, leading to dramatic changes in the
respectively of the third-order polarization induced in the signal-to-noise ratio.71 Another important conclusion that
Raman-active medium. can be made from equations (22) and (23) is that measuring
the ratio of the CARS signals for different polarization polarization ellipse respectively reproduce the spectral or
arrangements is a convenient way of determining the prop- temporal dependences of the real and imaginary parts of
erties of the nonlinear susceptibility tensor (Figure 3) and, the resonant component of the nonlinear polarization of a
thus, characterizing the symmetry properties of molecular medium.
transitions under investigation. Thus, the data obtained by means of coherent ellipso-
metry enable one to extract complete information concern-
ing the resonant component of the nonlinear polarization of
4.3 Coherent ellipsometry a medium, including the information on its phase.72,73 Note
that no assumptions regarding the shape of the line observed
Coherent ellipsometry, i.e. the measurement of the para- in a nonlinear Raman spectrum were made in our consid-
meters of the polarization ellipse corresponding to the eration, which means that this approach can be applied to
FWM signal, is one of the widely used modifications of a broad class of spectral lines. The procedure described
polarization-sensitive four-photon spectroscopy. Below, we above can also be extended to the time domain, allow-
consider the main physical principles and discuss the main ing not only the spectra but also time dependences of the
ideas of coherent ellipsometry. real and imaginary parts of the nonlinear polarization of a
The possibility of reconstructing the real and imaginary medium to be reconstructed. Finally, close molecular and
parts of the nonlinear-optical susceptibility of a medium atomic lines unresolvable in amplitude nonlinear Raman
as functions of frequency and time is due to the interfer- spectra can also be resolved in certain cases with the use of
ence of the resonant FWM component with the nonresonant the phase information stored in coherent nonlinear spectra.
coherent background, which ensures the recording of the
phase information for the resonance being studied. For a
broad class of problems, the nonlinear polarization of a 5 TIME-RESOLVED COHERENT
medium responsible for coherent FWM processes can be ANTI-STOKES RAMAN SCATTERING
represented as a sum of the nonresonant and resonant com-
ponents described by a real vector PE nr and a complex vector The method of time-resolved FWM spectroscopy implies
E r , respectively. Choosing the x-axis along the vector P
P E nr that the information on the parameters of atomic or mole-
(Figure 5b), we can write the Cartesian components of the cular systems are extracted from an impulse response of
total polarization of a medium cubic in the external field as a coherently excited system rather than from the fre-
quency dispersion of nonlinear susceptibilities, as done in
Px D Pnr C Pr eij cosq 24
frequency-domain FWM spectroscopy. The original idea of
ij
Py D Pr e sinq 25 time-resolved CARS is that a light pulse with a duration
shorter than the characteristic transverse relaxation time
where j is the phase of the resonant component of the non- T2 induces coherent molecular vibrations with amplitude
linear polarization and q is the angle between the resonant Qt, and the decay kinetics of these vibrations are ana-
and nonresonant components. lyzed with the use of another, probe light pulse, which is
Knowing the parameters of the polarization ellipse from delayed in time with respect to the pump pulses (Figure 6).
ellipsometric measurements, we can reconstruct, after some The complete set of equations governing the processes
not very complicated algebra with the use of equations (18), related to time-domain CARS includes the nonlinear wave
(19), (24), and (25), the real and imaginary parts of the res- equation (6) and the equations for the amplitude of coher-
onant nonlinear polarization as functions of frequency and ent molecular vibrations Qt, defined in accordance with
time from the experimental data of coherent ellipsometry. equation (10), and the normalized population difference
In the important particular case when the resonant com- between the levels involved in the Raman resonance, n D
ponent of the FWM signal can be considered as a small raa raa :
correction to the nonresonant component, the general pro-
cedure of separating the real and imaginary parts of the @2 Q 2 @Q 1 @a 2
C C 2 Q D nE 27
nonlinear polarization of a medium becomes especially sim- @t2 T2 @t 2M @Q
ple. One can easily verify that the relations1 @n n 1 1 @a 2 @Q
C D E 28
y D b cosj / ReP @t T1 2h @Q @t
c D b sinj / ImP 26 where T1 is the population relaxation time.

In many cases, the set of equations (6), (27), and (28) can
are satisfied in this case, showing that the spectral or tem- be simplified with the use of the slowly varying envelope
poral dependences of the parameters y and c of the FWM approximation. In this approximation, the energy of the
can be extracted through polarization measurements and

coherent ellipsometry (see the discussion above), and the
Q 2,I level of coherent nonresonant background can be sup-
pressed by means of the relevant polarization technique (see
1 2 3
equations (22) and (23) and Figure 5b). In time-domain
CARS, on the other hand, the nonresonant background
appears only at zero delay time between the pump and probe
pulses (Figure 6), having no influence on the transient sig-
nal, while the impulse-response measurements may directly
provide the information not only on the amplitude, but also
on the phase of a molecular or atomic resonance.
The development of femtosecond laser systems resulted
t
in the impressive technical and conceptual progress of
Figure 6. Diagram of pumping and probing of a medium in time-
time-domain FWM spectroscopy, allowing photochemistry
resolved CARS: (1) pump pulses, (2) decay of coherent vibrations
Qt, and (3) probe pulse; t is the delay time between pump and processes and molecular dynamics to be monitored in real
probe pulses. time.49,50,79 87 While picosecond pulses allow molecular
and atomic excitations to be studied with a high selectivity,
CARS signal as a function of the delay time t of the probe femtosecond pulses, characterized by large spectral widths,
pulse in the scheme of time-resolved CARS with short light lead to the excitation of molecular vibrations within a broad
pulses and k D 0 is given by1 spectral range, permitting the dynamics of various molecu-
1 lar vibrational subsystems to be simultaneously probed by
Wa t / jQtAt tj2 dt 29 detecting the beats in the impulse response79 or by using
1
several FWM processes simultaneously and distinguish-
ing between them by changing the detection wavelength49
Since the intensity of the CARS signal in such a scheme (Figure 2e).
is determined by equation (29), the use of sufficiently short
probe pulses makes it possible to measure the kinetics of
Qt. The possibility of such measurements was discussed 6 CONCLUSIONS AND OUTLOOK
by Akhmanov et al. in 1970.74 Experiments by Alfano and
Shapiro75 and von der Linde et al.76,77 have demonstrated Thus, nonlinear Raman spectroscopy is one of the most
the possibility of using the time-domain CARS technique powerful techniques of nonlinear spectroscopy, which has
to directly measure the T2 time for Raman-active modes in found numerous applications in condensed- and gas-phase
crystals and organic liquids. analysis, plasma diagnostics, investigation of molecular
The development of time-resolved approaches in CARS relaxation processes, temperature and concentration mea-
spectroscopy was stimulated by the progress of high-power surements, condensed-phase studies, and, very recently, in
picosecond laser systems, which made it possible to vir- femtochemistry. While many modifications of nonlinear
tually instantaneously (on the timescale of characteristic Raman spectroscopy have become a routine tool of modern
relaxation times) excite Raman-active transitions and then optical experiments, giving rise to many successful engi-
probe the decay kinetics of such excitations (Figure 6), neering applications, some nonlinear Raman experiments
thus obtaining the data on the molecular or atomic system carried out in the last decade have shown that the potential
under study. This approach proved to be efficient for study- of this technique for many topical and sometimes inter-
ing various nonstationary processes in molecular systems, disciplinary problems of modern physics, chemistry, and
permitting characteristic relaxation times and spectroscopic biology is far from being completely realized. Similar to
constants to be determined from the impulse response and frequency-tunable sources of coherent radiation, which rev-
the shape of the quantum beats.78 olutionized nonlinear optics in its early days, allowing
Equations (1012) and (29) show the relation between many delicate spectroscopic experiments, including nonlin-
the information that can be obtained by frequency- and ear spectroscopic studies, to be performed, the impressive
time-domain CARS spectroscopy. In fact, this information progress of femtosecond lasers in the 1990s has resulted in
is essentially the same, and frequency- and time-domain the breakthrough of nonlinear Raman spectroscopy to new
CARS methods can successfully complement each other unexplored areas, giving rise to several elegant new ideas
in studies of complex inhomogeneously broadened spectral and approaches, permitting more complicated systems and
bands. For example, in frequency-domain CARS spec- problems to be attacked, and leading to the measurements
troscopy, the phase information on molecular resonances of fundamental importance. This new phase of nonlinear
Raman spectroscopy also promoted the development of new 16. A.M. Zheltikov, N.I. Koroteev and A.N. Naumov, Quantum
spectroscopic concepts, including time-resolved schemes, Electron., 24, 1102 (1994).
broadband spectroscopy, polarization measurements, and 17. A.N. Naumov, N.I. Koroteev and A.M. Zheltikov, Laser
CARS generalizations based on higher-order nonlinear pro- Phys., 5, 965 (1995).
cesses. This conceptual and technical progress achieved in 18. E.J. Woodbury and W.K. Ng, Proc. IRE, 50, 2347 (1962).
the last decade shows us some very important features of 19. A. Owyoung and E.D. Jones, Opt. Lett., 1, 152 (1977).
what nonlinear Raman spectroscopy is going to be in the 20. A. Owyoung, Opt. Commun., 22, 323 (1977).
near future, encouraging the application of new ideas, tech-
21. P.D. Maker and R.W. Terhune, Phys. Rev. A, 137, 801 (1965).
niques, and methods in this area of spectroscopy.
22. P.R. Regnier and J.P.-E. Taran, Appl. Phys. Lett., 23, 240
(1973).
ABBREVIATIONS AND ACRONYMS 23. L.A. Rahn, L.J. Zych and P.L. Mattern, Opt. Commun., 30,
249 (1979).
ARSS Active Raman Scattering Spectroscopy 24. A.F. Bunkin and N.I. Koroteev, Sov. Phys. Uspekhi, 24, 394
CARS Coherent Anti-Stokes Raman Scattering (1981).
CSRS Coherent Stokes Raman Scattering 25. J.H. Bechtel and A.R. Chraplyvy, Proc. IEEE, 70, 657 (1982).
DFWM Degenerate Four-wave Mixing 26. R.J. Hall, Combust. Flame, 35, 47 (1979).
OPO Optical Parametric Oscillator 27. V.V. Smirnov and V.I. Fabelinskii, Pisma Zh. Eksp. Teor.
RIKE Raman-induced Kerr Effect Fiz., 28, 461 (1978).
SRS Stimulated Raman Scattering
28. J.W. Nibler, J.R. McDonald and A.B. Harvey, Opt. Com-
mun., 18, 371 (1976).
REFERENCES 29. D.V. Murphy and R.K. Chang, Opt. Lett., 6, 233 (1981).
30. D.N. Kozlov, A.M. Prokhorov and V.V. Smirnov, J. Mol.
1. S.A. Akhmanov and N.I. Koroteev, Methods of Nonlinear Spectrosc., 77, 21 (1979).
Optics in Light Scattering Spectroscopy, Nauka, Moscow 31. A. Owyoung, C.W. Patterson and R.S. McDowell, Chem.
(1981) (in Russian). Phys. Lett., 59, 156 (1978).
2. G.L. Eesley, Coherent Raman Spectroscopy, Pergamon, 32. V.I. Fabelinsky, B.B. Krynetsky, L.A. Kulevsky, V.A. Mis-
Oxford (1981). hin, A.M. Prokhorov, A.D. Savelev and V.V. Smirnov, Opt.
3. A.C. Eckbreth, Laser Diagnostics for Combustion Tempera- Commun., 20, 389 (1977).
ture and Species, Abacus, Cambridge, MA (1988). 33. M.R. Aliev, D.N. Kozlov and V.V. Smirnov, JETP Lett., 26,
4. A.B. Harvey and J.W. Nibler, Appl. Spectrosc. Rev., 14, 101 27 (1977).
(1978). 34. V.I. Fabelinsky and V.V. Smirnov, Pisma Zh. Eksp. Teor.
5. A.C. Eckbreth, Combust. Flame, 39, 133 (1980). Fiz., 27, 131 (1978).
6. S.A.J. Druet and J.-P.E. Taran, Progress Quantum Electron., 35. W. Nitsch and W. Kiefer, Opt. Commun., 23, 240 (1977).
7, 1 (1981). 36. W. Nitsch and W. Kiefer, J. Mol. Struct., 45, 343 (1978).
7. W. Kiefer, J. Mol. Struct., 59, 305 (1980).
37. M. Duncan and R.L. Byer, IEEE J. Quantum Electron.,
8. N. Bloembergen, Nonlinear Optics, Benjamin, New York QE-15, 63 (1979).
(1965).
38. S.A. Akhmanov, N.I. Koroteev, R. Yu. Orlov and I.L. Shu-
9. Y.R. Shen, The Principles of Nonlinear Optics, J. Wiley & may, Pisma Zh. Eksp. Teor. Fiz., 23, 276 (1976).
Sons, New York (1984).
39. S.A. Akhmanov, F.N. Gadjiev, N.I. Koroteev, R. Yu. Orlov
10. P.N. Butcher and D. Cotter, The Elements of Nonlinear and I.L. Shumay, Appl. Opt., 19, 859 (1980).
Optics, Cambridge University Press, Cambridge (1990).
40. S.A. Akhmanov, V.G. Dmitriev, A.I. Kovrigin, N.I. Koro-
11. F. Zernike and J.E. Midwinter, Applied Nonlinear Optics, teev, V.G. Tunkin and A.I. Kholodnykh, JETP Lett., 15, 425
J. Wiley & Sons, New York (1973). (1972).
12. G.G. Gurzadyan, V.G. Dmitriev and D.N. Nikogosyan, Non- 41. A.C. Eckbreth, Appl. Phys. Lett., 32, 421 (1978).
linear-optical Crystals: Their Properties and Applications in
Quantum Electronics, Radio i Svyaz, Moscow (1991) (in 42. A.C. Eckbreth and T.J. Anderson, Appl. Opt., 24, 2731
Russian). (1985).
13. J.F. Reintjes, Nonlinear Optical Parametric Processes in Liq- 43. M. Alden, P.-E. Bengtsson and H. Edner, Appl. Opt., 25,
uids and Gases, Academic Press, New York (1984). 4493 (1986).
14. J.F. Ward and G.H.C. New, Phys. Rev., 185, 57 (1969). 44. G. Laufer and R.B. Miles, Opt. Commun., 28, 250 (1979).
15. G.C. Bjorklund, IEEE J. Quantum Electron., QE-11, 287 45. G. Laufer, R.B. Miles and D. Santavicca, Opt. Commun., 31,
(1975). 242 (1979).
46. A.M. Zheltikov and N.I. Koroteev, Phys. Uspekhi, 42, 321 67. A.F. Bunkin, S.G. Ivanov and N.I. Koroteev, Sov. Phys. Dok-
(1999). lady, 22, 146 (1977).
47. D.A. Akimov, A.B. Fedotov, N.I. Koroteev, A.N. Naumov, 68. J.R. Nestor, J. Raman Spectrosc., 7, 90 (1978).
D.A. Sidorov-Biryukov and A.M. Zheltikov, Opt. Commun., 69. M.D. Levenson, J. Raman Spectrosc., 10, 9 (1981).
140, 259 (1997).
70. N.G. Khadzhiiski and N.I. Koroteev, Opt. Commun., 42, 423
48. D.A. Akimov, A.B. Fedotov, N.I. Koroteev, R.B. Miles, A.N. (1982).
Naumov, D.A. Sidorov-Biryukov and A.B. Fedotov, Opt.
Lett., 24, 478 (1999). 71. A.Y. Chikishev, G.W. Lucassen, N.I. Koroteev, C. Otto and
J. Greve, Biophys. J., 63, 976 (1992).
49. M. Schmitt, G. Knopp, A. Materny and W. Kiefer, Chem
Phys. Lett., 280, 339 (1997). 72. A.B. Fedotov, N.I. Koroteev, A.N. Naumov, D.A. Sidorov-
Biryukov and A.M. Zheltikov, J Nonlinear Opt. Phys. Mater.,
50. S. Meyer, M. Schmitt, A. Materny, W. Kiefer and V. Engel, 6, 387 (1997).
Chem. Phys. Lett., 301, 248 (1999).
73. N.I. Koroteev, Sov. Phys. Uspekhi, 30, 628 (1987).
51. D.R. Meacher, A. Charlton, P. Ewart, J. Cooper and G. Alber,
Phys. Rev. A, 42, 3018 (1990). 74. S.A. Akhmanov, K.N. Drabovich and I.I. Nurminskii, Digest
of the International Quantum Electronics Conference, Tokyo,
52. P. Ewart and P. Snowdon, Opt. Lett., 15, 1403 (1990). Japan, 43 (1970).
53. P. Ewart, P. Snowdon and I. Magnusson, Opt. Lett., 14, 563 75. R.R. Alfano and S.L. Shapiro, Phys. Rev. Lett., 26, 1247
(1989). (1971).
54. P. Ewart, P.G.R. Smith and R.B. Williams, Appl. Opt., 36, 76. D. von der Linde, A. Laubereau and W. Kaiser, Phys. Rev.
5959 (1997). Lett., 26, 854 (1971).
55. D.J. Rakestraw, R.L. Farrow and T. Dreier, Opt. Lett., 15, 77. D. von der Linde, A. Laubereau and W. Kaiser, Phys. Rev.
709 (1990). Lett., 27, 802 (1971).
56. J.-L. Oudar, R.W. Smith and Y.R. Shen, Appl. Phys. Lett., 78. S.A. Akhmanov, N.I. Koroteev, S.A. Magnitskii, V.B. Moro-
34, 758 (1979). zov, A.P. Tarasevitch and V.G. Tunkin, J. Opt. Soc. Am. B,
57. L.S. Aslanyan, A.F. Bunkin and N.I. Koroteev, Sov. Tech. 2, 640 (1985).
Phys. Lett., 4, 473 (1978). 79. R. Leonhardt, W. Holzapfel, W. Zinth and W. Kaiser, Chem.
58. L.S. Aslanyan, A.F. Bunkin and N.I. Koroteev, Opt. Spec- Phys. Lett., 133, 373 (1987).
trosc., 46, 89 (1979). 80. M. Fickenscher and A. Laubereau, J. Raman Spectrosc., 21,
59. N.I. Koroteev, M. Endemann and R.L. Byer, Phys. Rev. Lett., 857 (1990).
43, 398 (1979). 81. T. Joo, M.A. Dugan and A.C. Albrecht, Chem. Phys. Lett.,
60. H. Lotem, R.T. Lynch and N. Bloembergen, Phys. Rev. B, 14, 177, 4 (1991).
1748 (1976). 82. M. Fickenscher, H.-G. Purucker and A. Laubereau, Chem
61. L.S. Aslanyan, A.F. Bunkin and N.I. Koroteev, Sov. Tech. Phys. Lett., 191, 182 (1992).
Phys. Lett., 4, 473 (1978). 83. C.C. Hayden and D.W. Chandler, J. Chem. Phys., 103, 10465
62. A.F. Bunkin, S.G. Ivanov and N.I. Koroteev, Pisma Zh. (1995).
Eksp. Teor. Fiz., 24, 468 (1976). 84. M. Motzkus, S. Pedersen and A.H. Zewail, J. Phys. Chem.,
63. A.B. Fedotov, O.S. Ilyasov, N.I. Koroteev and A.M. Zhelti- 100, 5620 (1996).
kov, Nuovo Cimento D, 14, 1003 (1992). 85. M. Schmitt, G. Knopp, A. Materny and W. Kiefer, J. Phys.
64. S.A. Akhmanov, A.A. Ivanov, N.I. Koroteev, S.F. Mironov Chem. A, 102, 4059 (1998).
and A.I. Fishman, J. Mol. Liquids, 53, 111 (1992). 86. T. Chen, V. Engel, M. Heid, W. Kiefer, G. Knopp,
65. M.D. Levenson and N. Bloembergen, Phys. Rev. B, 10, 447 A. Materny, S. Meyer, R. Pausch, M. Schmitt, H. Schwoerer
(1974). and T. Siebert, Vib. Spectrosc., 19, 23 (1999).
66. M.D. Levenson and N. Bloembergen, J. Chem. Phys., 60, 87. T. Lang, K.-L. Kompa and M. Motzkus, Chem. Phys. Lett.,
1323 (1974). 310, 65 (1999).
APPENDIX 1
Glossary of widely used nonlinear Raman CARS Coherent anti-Stokes Raman scattering
acronyms and abbreviations CAHRS Coherent anti-Stokes hyper-Raman
scattering
BOXCARS Coherent anti-Stokes Raman scattering CREM Coherent Raman ellipsometry
where beam crossing is used to improve CSRS Coherent Stokes Raman scattering
spatial resolution DFG Difference-frequency generation
DFWM Degenerate four-wave mixing OHD-RIKE Optically heterodyne detected

FWM Four-wave mixing Raman-induced Kerr effect
Helicopter, Polarization configurations in four-wave PARS Photoacoustic Raman spectroscopy
submarine, mixing allowing coherent background PCARS Partially coherent anti-Stokes Raman
asterisk suppression scattering
HRS Hyper-Raman scattering RIKE Raman-induced Kerr effect
IDSRS Ionization-detected stimulated Raman SBS Stimulated Brillouin scattering
spectroscopy SFG Sum-frequency generation
IGSRS Ionization-gain stimulated Raman SHG Second-harmonic generation
spectroscopy SRGS Stimulated Raman gain spectroscopy
ILSRS Ionization-loss stimulated Raman SRS Stimulated Raman scattering
spectroscopy THG Third-harmonic generation
LIGS Laser-induced gratings spectroscopy TPA Two-photon absorption
Coherent Anti-Stokes Raman Spectroscopy
Laszlo Ujj and George H. Atkinson
University of Arizona, Tucson, AZ, USA
1 INTRODUCTION as a coherent-like beam which is easily separated spatially

from the pumping beams (the angles at which the pump-
Although the principles of Raman scattering have been ing beams enter the sample are well defined and specific
described in detail many times, a general understanding for a particular sample and set of wavelengths). The CARS
can be obtained from the schematic representations shown signals appear energetically at !a D !1 C !1 !s ) where
in Figure 1. Spontaneous Raman spectroscopy is derived !1 and !s are the h D 1 energies of the two pumping
from inelastic light scattering which results from the trans- beams (Figure 1b) and the conservation of energy defined
fer of radiative energy from an excitation beam at !1 to the by this relationship must be fulfilled if CARS signals are to
internal degrees of freedom (e.g. rotational and vibrational) be generated. In addition to the CARS signals assignable to
of a molecule (Figure 1a). When the excitation energy !1 internal molecular degrees of freedom, separate CARS sig-
is energetically near a transition (e.g. electronic) within nals can arise from the nonresonant background of the sam-
the molecule, the spontaneous Raman signal intensity is ple (e.g. solvent). In general, the CARS signals assignable
enhanced due to a coupling of the internal degrees of to specific molecular rotations and/or vibrations must be
freedom to the molecular transient moment. The resultant quantitatively separated from the nonresonant background
resonance Raman (RR) signals appear over a range of ener- signals. In analogy with RR, CARS signals are significantly
gies normally measured as displacements from !1 , !s D enhanced in intensity whenever !a , !1 or !s is energetically
!1 at lower energies (Stokes) and !as D !1 C at close to an electronic transition.
higher energies (anti-Stokes) where represents the inter- Applications of CARS encompass a variety of experi-
nal molecular degrees of freedom. The values and intensi- mental configurations and, thereby, a range of information
ties of can be utilized to characterize molecular structure can be obtained. Most simply, CARS signals can be gen-
in terms of the rotational and vibrational motions. Since erated from stable species (Figure 1b). In a time-resolved
RR signals normally appear as small, narrow-bandwidth configuration, the CARS signal can be generated after the
features on top of any sample fluorescence having a broad initiation of a chemical and/or physical event which results
bandwidth, the separation of RR signals from the back- in the formation of a transition species. Two types of tran-
ground fluorescence is a major experimental/analysis issue sient species are normally encountered: photophysical (e.g.
which often limits the resultant signal to noise. excited electronic states) and photochemical (e.g. dissocia-
Coherent anti-Stokes Raman spectroscopy (CARS) is tion products, isomers, etc.). Both types of transient species
derived from nonlinear optical phenomena in which coher- can be examined via time-resolved CARS measurements
ent resonances are generated within a sample by at least (Figure 1c and 1d).
two laser fields acting simultaneously. As in the case
of RR spectroscopy (see above), the resonances within
the sample are derived at least in part from the internal 2 NONLINEAR OPTICAL
molecular degrees of freedom (i.e. rotations and vibrations) SPECTROSCOPY
(Figure 1b). Importantly, the CARS signal exits the sample
The nonlinear electromagnetic properties of many mate-
John Wiley & Sons Ltd, 2002. rials are sufficiently large to generate a wide variety of
|0 |0 materials can be measured and utilized spectroscopically

|e0 |e0 by the high intensity radiation delivered by lasers operating
throughout the infrared, visible, and ultraviolet spectral
1 s 1 s 1 a regions.
Of particular interest has been the inelastic, nonlinear
|0
scattering in which the vibrational degrees of freedom in
|0
|g0 |g0
molecular systems are measured using Raman methodolo-
(a) Ground state (b) Ground state gies. Since the first observation of coherent Raman scat-
tering (CRS) via the mixing of four electromagnetic waves
|1
in 1965,1 several other nonlinear Raman phenomena have
|e1
|0
been observed and developed as spectroscopic methods.
|e0 These coherent Raman spectroscopies have been widely
1 s 1 a used for numerous applications focusing on analytical and
diagnostic results as well as determinations of molecular
exc
|1 structures and kinetic properties. This article focuses on
|g1 these broadly successful nonlinear Raman methodologies
|0 Photochemical and especially on those involving four-wave mixing phe-
|g0 product
(c) Ground state
nomena (interactions between four electromagnetic fields)
in a medium with nonlinear response.
|2 The mechanism underlying nonlinear, four-wave mix-
|e2 ing derives from the nonlinear polarization induced in
the medium by the three driving electromagnetic fields
which leads to the generation of a fourth electromagnetic
1 s 1 a
|0
wave at a combinatory frequency determined by the fre-
|e0 quencies of the three driving fields. CRS is a specific
|1
|e1
example of such a nonlinear interaction in which the com-
Excited bination of the frequencies of the driving and generated
exc electronic fields is selected to match at least one vibrational res-
state
onance of a molecular (nonlinear) medium. CARS is a
|1> special case of CRS in which the generated (signal) field
|g0> has the largest frequency among the four electromagnetic
(d) Ground state
waves involved. Term diagram representations of three
Figure 1. Term diagram representations of four different Raman types of CARS measurements from either stable or tran-
scattering processes for measuring molecular vibrations. (a) Spon-
taneous RR scattering from a ground-state species using excitation sient molecular species are presented in Figure 1 together
at !1 (in resonance with a ground-state absorption transition). with the corresponding representation for spontaneous RR
(b) Resonant CARS from a ground-state species using excitation scattering.
at !1 and !s (in resonance with an absorption transition in the Detailed theoretical descriptions and reviews of CRS and
ground-state species). (c) Time-resolved CARS of a photochem- CARS appear throughout the scientific literature.2 6 A con-
ical intermediate measured at a fixed time delay following the
optical excitation (!p ) of the ground state. The CARS signal is cise theoretical formalism of CARS, and its electronically
generated by !1 and !s (in resonance with an absorption transi- resonant variant, is presented here with an emphasis on
tion in the photochemical species). (d) Time-resolved CARS of points relevant to the interpretation of experimental spectra.
a photophysical intermediate (e.g. excited electronic state) mea- Particular attention is given to deriving vibrational signa-
sured at a fixed time delay following the optical excitation (!p ) tures from the Raman resonances observed in CARS data.
of the ground state. The CARS signal is generated by !1 and !s
(in resonance with an absorption transition in the photophysical The instrumental components required for obtaining typi-
intermediate). cal CARS data are illustrated in a brief description of an
experimental system used to record CARS spectra from bio-
nonlinear optical responses when the material is exposed physical molecules (photoactive protein reactions). Specific
to high intensity radiation such as commonly available examples of CARS data are taken from measurements of
from modern lasers. Such nonlinear optical responses pro- stable and transient intermediates comprising the photore-
vide the foundation for a diverse array of spectroscopic actions of the bacteriorhodopsin (BR) and rhodopsin (Rh)
phenomena and, thereby, an equally diverse group of spec- proteins. Finally, an overview of developments in this field
troscopic measurements. The properties of these nonlinear to be anticipated is presented.
Coherent Anti-Stokes Raman Spectroscopy 3
3 THEORETICAL specific vibrational normal mode. While both c3 1111 and
3
c1212 individually provide information about the vibra-
3.1 Principles of CRS tional degrees of freedom of the molecular medium, the
ratio of different c3 susceptibility components reflects the
Theoretically, phenomena derived from CRS, including differences between the molecular symmetry of selected
CARS, can be described in terms of the field-induced vibrational modes. Thus, PSCARS can be used to both
polarization of the nonlinear medium in which the driving selectively suppress a particular polarization component
radiation mixes. For CARS, induced polarization is created (i.e. vibrational symmetry) in favor of another, and measure
within the nonlinear medium by the coherent interaction of the differences in polarization properties of given vibra-
three radiation fields. In the general case, three different fre- tional modes. The first application increases the signal-to-
quencies are used (!0 , !1 (!0 ), and !s (<!1 )) which are noise ratio (S/N) of vibrational features from one molecular
commonly referenced to as the pumping (!0 and !1 ) and species in a sample relative to others by selectively con-
Stokes frequencies (!s ). When !0 D !1 degenerate scat- trolling the relative intensities of individual CARS features
tering occurs which is the case considered in detail here. (termed polarization suppression).7 9 Since the highest S/N
Formally, the nonlinear cubic polarization of the medium, can be experimentally obtained for c3 1111 (commonly termed
having a linear dependence on the amplitude of each of the the CARS signal), only this c3 component is consid-
external radiation fields, can be written as equation (1): ered here (subscript indices of c3 components are omitted
throughout the remainder of this paper). The second appli-
Pi !a D 2!1 !s cation measures the polarization properties of individual
3
D3 cijkl !a , !1 , !1 , !s E1j !1 E1k !1 Esl !s vibrational modes and thereby, characterizes the molecular
j,k,lDx,y,z structure.
1 In general, c3 is composed of three distinct parts:
where i, j, k, and l denote vector projections in Cartesian the nonresonant susceptibility, c3 NR , of the solvent, the
coordinates. The quantity c3 , a tensor of rank four, repre- electronically resonant susceptibility, c3 ER , of the solute,
sents the third-order, nonlinear susceptibility of the medium and a vibrational-resonant term, c3 VR , consisting of the sum
and describes its macroscopic polarization properties. The over individual vibrational resonances, c3 VRj , found in any
indexed number of each independent tensor component molecule in the solution (equation (3)):
reflects the macroscopic symmetry of the medium (i.e.
anisotropy). For a macroscopically isotropic medium (e.g. a 3 3

N
3
liquid or gas), c3 c3 D cNR C cER C cVRj 3
ijkl (!a , !1 , !1 , !s ) has only two indepen-
jD1
dent tensor components (c3
and
1111
3
c1212 ). All other tensor
3 3 Explicit expressions for the microscopic resonant Raman
components can be derived from c1111 and c1212 , equa-
tion (2), (assuming permutation invariance): 7 susceptibility have been derived from a third-order quantum
mechanical treatment (e.g. Feynman-like diagrams10 14 ).
3 3 3
cijkl !a , !1 , !1 , !s D c1122 ij kl C c1212 ik jl A microscopic expression for c3 VR , obtained from a
third-order perturbation theory is considered here. Each
C c3
1221 il jk 2 contribution from an individual Raman resonance can be
3 3
c1122 !a , !1 , !1 , !s D c1212 !a , !1 , !1 , !s represented as equation (4):
The susceptibility component measured experimentally is cVRj D aj fj 4

selected by choosing the appropriate polarizations for the
driving electric fields, e.g., c3 where aj is the strength of the resonance, f is its form
1111 is measured if the polar-
ization planes of the !1 and !s fields are parallel and c3 factor, and j describes the detuning. The typical reso-
1212
is measured if the polarization planes of the !1 and !s nance form factor is given by the complex, Lorentzian-type
fields are perpendicular. CARS measurements made as a function of equation (5):
systematic function of these polarization properties (e.g. 1 j !1 !s
ellipticity) can be utilized to determine the polarization fj D j D 5
j i j
properties assignable to specific vibrational modes within a
molecule. Such polarization sensitive coherent anti-Stokes where j j > 0 is the frequency of the jth vibrational
Raman spectroscopy (PSCARS) recognizes the distinct mode, j is a half-width (half width at half maximum,
polarization properties (e.g. dipole moment) reflected in HWHM), and i is the imaginary unit. The inhomogeneous
given vibrational degrees of freedom as represented by a broadening of the vibrational bands is neglected here and
therefore Lorentzian lineshapes are assumed to be accurate equation (7):8,17

for all bands. 2

For the case of resonance CARS (excitation occur- Ia
Sample

NA
A e ij
NB
A e ik
2 j k
ring at a frequency which generates resonances with an D m 1 C 1 h Ch
IReference
a jD1
j i kD1
k i
electronically excited state), the strength of the isolated
Raman resonance can be represented by equation (6): 7

N dgn dnj djm dmg where j and k are the differences between the vibra-
aj D 3 tional and background phases, m is a scaling factor, the
h n
!a !ng ing m !1 !mg img
summations run over the vibrational bands of species A
6
(NA ) and B (NB ), respectively. Aj Ak is the amplitude and
Equation (6) is essentially a product of one-photon resonant
j k is the detuning parameter for the jth (kth) vibra-
polarizabilities at !1 and !a where dng and dmg are molec-
tional mode having a frequency j k and a bandwidth
ular dipole moment matrix elements for the transition from
2j 2k where j D [j !1 !s ]/j . The relative
the ground state to the nth (mth) vibrational level of the
concentration of species A is given by h0 h 1, but
excited electronic state, !ng and !mg are the corresponding
h D 0 when only species A is present in the sample). A
frequencies of these transitions, ng and mg are the half-
more restricted function is found for retinal proteins where
widths of these transitions, N is the molecular density of
ground-state molecules, and hi denotes averaging over an the frequency dependence of the phase factors are explic-
ensemble of molecules and their respective dipole moment itly considered by introducing a second-order polynomial
orientations. frequency domain function into equation (7).17
The complex aj functions describing !a (equation (4)),
are different for each vibrational mode, j. Based on
3.2 Narrow and broadband (multiplex) CARS
the BornOppenheimer approximation, different aj val-
methodologies
ues appear in the overlap integrals between the nuclear
wave functions of the jth vibrational mode in the ground-
In general, two different experimental methods are used to
state and the vibrational modes of the first electronically
record the spectral distribution of CARS signals. Narrow-
excited state of the molecule (i.e. FranckCondon factors).
band CARS utilizes !1 and !s bandwidths !1 and !s
The dipole moment matrix element between two electronic
that are smaller than those of the vibrational transitions
states is the same for all vibrational modes.
monitored. Usually !s is tuned over the relevant spectral
It is convenient to represent aj as: Aj !a exp[ij !a ].
range and the CARS signal is recorded by wavelength
In the vicinity of an isolated vibrational resonance, the
scanning and a single channel detector (e.g. photomulti-
aj D Aj expij factor can be considered as frequency
plier tube). In broadband or multiplex CARS, !1 and !s
independent where Aj and j values are constant, real
bandwidths are respectively much smaller and much larger
numbers (0 j < 2p, and j D 0 is the nonresonant
than the vibrational bandwidths observed. The spectral dis-
case).
tribution of the !a over a broad (typically 300700 cm1 )
Variations of !1 can alter the CARS bandshapes through
spectral range is measured by multichannel detectors (with-
the frequency dependence of the electronically resonant
out wavelength scanning) and is determined by the !s
complex phase factor, j , which is a function of both
bandwidth. Attention is given here to multiplex CARS since
!1 and !s . When !1 is tuned relative to the absorp-
it can be optimized for rapidly recording larger parts of the
tion maximum, the relative contributions of dispersive line
vibrational spectra.
shapes increase, thereby altering the amplitudes of over-
The intensity of the observed multiplex CARS signal is
lapping features (i.e. changing the spectral shape). This
proportional to the square modulus of the induced polariza-
!1 dependence is commonly used, by the selection of
tion. If the spectral width of the narrowband laser, !1 , is
!1 and !s , to optimize the S/N of the CARS spectrum
small compared to the vibrational linewidths, 2, then the
for the molecule of interest. The dependence of CARS
steady-state approximation applies and the CARS spectral
bandshapes on !1 is an important issue when CARS spec-
intensity is given by equation (8):
tra are to be interpreted in terms of different molecular
species.15,16 Ia !a / jc3 !a , !1 , !1 , !s j2 I21 Is !s G 8
The successful modeling of the third-order suscepti-
bility (i.e. c3 or the symmetrical response function in where I1 and Is !s are respectively the intensity and
the frequency domain) can be derived from the the- spectral intensity of the driving radiation (laser) fields. The
oretical results used to interpret the vibrational spec- term G represents the phase-matching within the sample
tra from a reactive mixture of two species (A and B), (i.e. sample dispersion and absorption18,19 ). Since Ia !a is
proportional to Is !s , the spectral intensity of the observed where the function G describes the effect of both sample
CARS signal is proportional to the jc3 j2 of the sample. dispersion and absorption, D D L2a1 C as aa , K D
In the case of narrowband CARS, equation (8) has to be Lka j2k1 ks j, A D L aa , L is the sample thickness, and
integrated over !s . !1 , !s and !a D 2!1 !s are respectively, the frequencies
Since the molecular vibrations are excited coherently and of the narrowband laser, the broadband laser, and the
the scattering process is parametric, the resultant scatter- anti-Stokes Raman field. The absorption coefficients of the
ing from the sample is also coherent. As a consequence, sample at !1 , !s , and !a are given by a1 , as , and aa ,
coherent interferences occur not only between the vibra- respectively, while ka is the absolute value of the anti-
tionally resonant and nonresonant parts of the CARS signal, Stokes Raman wave vector, and k1 and ks are respectively,
but also between the coherent scattering itself as it is gen- the wave vectors of !1 and !s .
erated from individual vibrational modes.
In general, the relative amplitudes of the vibrational
bands for CARS and RR can be significantly different, but 4 CRS/CARS METHODOLOGIES
the experimental conditions described here ensure that the
relative amplitudes are similar. This result is evident since CRS/CARS methodologies can be applied to essentially
the cross-section for RR scattering is proportional to the any molecular system containing well-defined vibrational
imaginary part of c3
VR (equation (9)):
20
degrees of freedom. Samples in the gas, liquid, or solid
d2 s Im c3 phases can be examined and an extensive literature covering
VR
/ 9 a broad range of molecules in each of these phases exists.6
dod exph/kT 1
Experimentally, the primary criterion to be fulfilled
where d2 s/dod is the differential cross-section expressed involves the peak power of excitation needed to generate
in terms of the unit solid angle and unit frequency, h is CRS/CARS signals with high S/N. Typically, pulsewidths
the Planck constant, and T is the absolute temperature. By in the 10 ns21 to 1 ps22 range have been successfully uti-
contrast, the intensity of resonant CRS depends on both the lized since the peak powers required for CRS/CARS can be
real and imaginary parts of the medium. provided by commonly available laser systems. Except for
issues of photolability, the differences in pulsewidths are
not a primary consideration for studies of stable species.
3.3 Phase matching For time-resolved measurements from transient species,
however, these pulsewidths determine the maximum time
The spatial separation between the !1 and !s beams, and resolution achievable.
therefore the phase matching angle between the two focused The versatility with which CRS can be applied, as well as
beams in a sample, is controlled by the angle by which the type of vibrational spectra obtainable from its applica-
the !s beam enters the sample. The phase-matching angle tion, can be illustrated by CARS data recorded from several
expected for a sample is calculated for the solvent in related retinal proteins, specifically BR and Rh and their
dilute solutions. The effect of absorption on the CARS respective room-temperature photoreactions. These retinal
intensity must be considered, however, even though it is proteins store absorbed light energy with exceptionally
difficult to determine since it depends on the transverse high efficiency and support effective signal transduction
intensity distribution and the spatial overlap of the !1 and mechanisms which are utilized for a variety of biochem-
!s beams in the sample. Experimentally, these latter two ical processes including vision and transmembrane proton
effects can be normalized by dividing the CARS signal pumping.23,24 Containing respectively 248 and 348 amino
measured from the absorbing solution by the nonresonant acids arranged within seven a-helical structures spanning
signal measured from the solvent. This empirical method the protein membrane, both BR and Rh contain only a sin-
requires that the CARS signal from the solution contains gle chromophore which absorbs in the visible, namely reti-
no resonant vibrational features. nal. Since the retinal structure (especially its isomeric con-
The relationships between the picosecond resonance figuration) is thought to control both BR and Rh reactivity,
coherent anti-Stokes Raman scattering (PR/CARS) intensity picosecond time-resolved coherent anti-Stokes Raman scat-
and the dispersion and absorption properties of the sample tering (PTR/CARS) studies have focused on determining
are important to determine if the phase matching conditions the time-dependent structural changes in the retinal through-
are to be optimized. These relationships are well described out the picosecond/millisecond time intervals over which
theoretically by equation (10):18,19 both photoreactions function. The instrumentation and mea-
1 C eD 2eD/2 cosK A surement procedures used for CARS studies of BR and Rh
GD, K D e 10 exemplify those required for such measurements in general.
K2 C 0.25D2
4.1 Instrumentation and procedures
Monochromator
LN CCD
Lock-in amp
Nd : YLF
4.1.1 Laser system and optical configuration for
CW ML
Computer
laser
CARS
Although instrumentation using a variety of laser sys- PD
tems with different pulsewidths (typically nanosecond to F G
SHG
picoseconds) has been used successfully to obtain CARS M FM
data, this discussion focuses on picosecond instrumentation PTA M THG
signal CARS M BS
utilizing 28 ps pulsewidths. Such picosecond systems signals
L
permit the measurement of vibrational CARS spectra with A BS
bandwidths determined solely by the Fourier transform
SP dye laser P
Optical delay
properties of the driving electronic fields (for multiplex Sample Ref. lines
SP dye laser 1
CARS the !s bandwidth is large and does not signifi- L
SP dye laser S
cantly influence the vibrational bandwidths). When utilized
for time-resolved measurements from transient intermedi-
O
ates, such PTR/CARS experiments also provide picosecond
BE CD
time-resolution.
M CD
A typical, multiplex CARS system17,25 with 28 ps time
AC M CD
resolution uses the 1053-nm output (30-ps pulses (HWHM)
at 76-MHz repetition rate) of a continuous wave, mode- M BE
locked Nd : YLF laser to generate second- (527 nm) and M
BS PM
third- (351 nm) harmonic radiation from nonlinear optical BE
crystals [e.g. lithium triborate (LBO) and b barium borate M M BS M Chopper

(BBO)]. The 527-nm and 351-nm radiation is used to pump (a)
three, independently controlled dye lasers. A combination
Excitation beam translation
of polarizing beam splitters and half-lambda plates are used a signals
to adjust the output powers (0.61 W) for each dye laser.
Each dye laser is equipped with a cavity dumper, three of Reference
which are synchronized to the 76-MHz repetition rate of the
Nozzle Sample jet
Nd : YLF mode locker. The entire laser system is operated
at a 400-kHz repetition rate (Figure 2a) in order to match 1 s
the flow properties of the liquid sample jet (see below). 1 s
(b) p
For typical CARS experiments, one dye laser is operated
with a narrowband (<4 cm1 , full-width at half-maximum, Figure 2. (a) Instrumentation to measure PTR/CARS spectra
FWHM) using an intracavity, three-plate birefringent filter and picosecond transient absorption (PTA) data. Abbreviations:
SHG, second harmonic generator; THG, third harmonic generator;
and a variety of dyes (e.g. rhodamine 590 (560600 nm),
M, mirror; BS, beam splitter; CD, cavity dumper; BE, beam
rhodamine 610 (590625 nm), Kiton Red (625645 nm), expander; AC, auto-correlator; O, oscilloscope; PD, photo diode;
and 4-dicyanomethylene-2-methyl-6-p-diethylaminostyryl- A, aperture; PM, phase-matching adjustment; F, optical filter;
4-H-pyran (DCM) (640690 nm)) to produce 30 nJ per G, optical grating; and L, lens. (b) Schematic representation,
pulse at !1 . The temporal autocorrelation width (FWHM) based on a perspective view of the sample region (a), of the
laser beam arrangement used to generate PTR/CARS signals (!a ).
of the laser pulses, as measured by a rapidly scanned
The horizontal arrow indicates adjustable spatial positions of the
autocorrelator,17 is typically 7.5 ps. Assuming Gaussian pump laser beam. The spatial displacement is synchronized with
pulse shape, the corresponding pulse width is 5.5 ps. the timing sequence of the respective laser pulses. See text and
Broadband, picosecond pulse generation to obtain !s Figure 1 for definitions of !p , !1 , !s . The figure is adapted from
is achieved by synchronously pumping a second contin- Ujj et al.17
uous wave, mode-locked dye laser in which the intracavity
birefringent filter is removed. In some cases, the birefrin- pumped laser cavity. Such broadband outputs are needed to
gent filter is replaced by a 23-m thick, uncoated pellicle generate the CARS signals in the 7601750 cm1 region
etalon. The resultant broadband dye laser output is typi- of the vibrational spectrum typically examined. The exper-
cally 300700 cm1 (FWHM) depending on which dye imental conditions are selected to obtain the maximum
solution is used. The spectral width of the broadband laser output energy (typically 20 nJ per pulse) and the shortest
output also depends on the alignment of the synchronously pulse width (3.57.5 ps).
Beam expanding (5) telescopes are used to adjust beam same sample (using optical gratings as wavelength selectors
divergence and, therefore, to control the spatial overlap of and lock-in detection (Figure 2a)).
the !1 and !s beams in the sample. Expansion of the !1 and Procedurally, CARS spectra are recorded from stable
!s beams also minimizes the variation in beam diameter samples using only !1 and !s pulses (i.e. PR/CARS).
caused by the movement of optical elements composing PTR/CARS data are recorded from a reactive mixture
the delay line. The divergence of the CARS beam itself is of stable and transient species as a function of the time
typically 102 rad. delay (controlled by optical delay lines) between the pump
An aperture is used to spatially separate the CARS beam pulse (!p ) and the !1 and !s probe pulses. The spectral
(!a ) by blocking the !1 and !s beams emerging from the changes observed in these PTR/CARS data can be related to
sample. After collimation, the !a beam is focused onto the time-dependent structural changes in the transient species.
entrance slit of a filter stage (operating in subtractive dis- Nonresonant CARS signals from the solvent in the refer-
persion mode) of the triple monochromator. The adjustable ence compartment are obtained simultaneously for both the
bandpass (115 nm) of the monochromator is used to PR/CARS and PTR/CARS data.
reduce the residual Rayleigh scattering generated during
laser excitation. The radiation exiting the filter monochro-
mator enters the entrance slit of a 0.75-m single dispersion 4.2 Examples of CRS/CARS data
spectrometer (1800-l mm1 grating). The dispersed radia-
tion is detected by a multichannel charge coupled device CRS has been utilized in an extremely wide range of appli-
(CCD) (e.g. a 256 1024 CCD detector) attached to the cations, especially for gas and liquid samples. CRS spectra
exit slit of the spectrometer. The spectral resolution of the from gas samples have long been identified as important
entire spectrometer/detector system is 1.2 cm1 (HWHM). diagnostic methods in combustion and plasma chemistries26
The !1 and !s beams are linearly polarized with a depo- and such measurements continue to be an active area of
larization ratio of <103 . The optical elements throughout research. Gas samples offer a major experimental advantage
the system are selected to preserve this degree of polariza- for CRS measurements since the nonresonant background
tion. A GlanThomson polarizer is used to ensure that the signals are small. Unfortunately, this advantage is often
probe laser beams before emerging from the sample have offset by the small CRS signals associated with the low
the same polarization plane. number densities of scatters found in gas samples.
A different set of experimental parameters dominate the
case of CRS spectra measured in liquid samples. Although
4.1.2 Sample
the number density of scatters in liquids is significantly
The spatial configuration of the laser pulses within the larger, the analysis of the resultant CRS signal is often more
sample and reference regions, as well as their relation- complicated due to the correspondingly larger nonresonant
ships to the spectrometer/multichannel detection system, are background signal.
described here for PTR/CARS measurements. A schematic
representation of the five laser pulses used to generate
4.2.1 CRS from gas samples
respectively the PTR/CARS and nonresonant CARS sig-
nals from the sample and reference regions, is presented Applications of CRS methodologies in gas samples such as
in Figure 2(b). These CARS signals are generated simulta- those found in combustion mixtures26 29 and plasmas30,31
neously from separate sample and reference compartments. have focused on measurements of the steady-state concen-
The reference compartment is a capillary containing a static tration of a given gas species, primarily to determine the
amount of the solvent (e.g. water) placed next to the nozzle sample temperature. These techniques mostly utilize high
through which the sample solution flows. The simultaneous power (mJ per pulse), nanosecond lasers to measure CARS
measurement of the nonresonant CARS background signal data. Since CARS spectra can be analyzed in terms of vibra-
from solvent and the PTR/CARS signal from the sample tional and rotational band intensities, it is feasible to directly
itself minimizes any variability in the spectral and/or inten- obtain the associated vibrational and rotational populations
sity properties of the laser pulses or in the flow dynamics within the gas sample, and thereby the gas temperature.
of the sample. The signals from both compartments are CARS data, therefore, can be used as a real-time, diagnos-
collimated and focused onto the entrance slit of the triple tic tool for determining the in situ thermal conditions found
monochromator. The wavelength-dispersed CARS signals within a reactive chemical chamber. When CRS data with
are focused onto two separate, parallel stripes of a liquid- sufficiently high S/N can be recorded, spectral signatures
nitrogen cooled, CCD multichannel array. This instrumental also can be obtained which can be used to identify gaseous
configuration also permits PTA data to be recorded from the species within the chemical reaction.
The continuing advances in femtosecond laser tech- 1.5

nology has made it feasible to extend both types of
CRS measurements in gases to include the investiga-
tions of molecular properties in the sub-picosecond time
domain: observations of femtosecond dynamics associated (a)
with nuclear responses,32 34 measurements of vibrational 1.0
dephasing properties,35,36 and studies of intramolecular
vibrational energy redistributions.33,37
Rel. PR/CARS intensity

4.2.2 CRS from liquid samples
0.5
The major experimental challenges for CRS applications in
liquid samples remain those issues associated with the large
nonresonant background signal. Applications of CRS in liq-
uids continue to expand since this challenge has been suc-
cessfully addressed by experimental methodologies which
use electronic resonance enhancements and/or polarization 0.0
sensitive configurations to minimize the contributions of the
nonresonant signal. Both methods are particularly useful for
those samples in which large fluorescence background sig-
nals also exist (a major reason why it is difficult to derive (b)
vibrational information from RR data).38 41
Although applications of CRS to liquid samples in gen- 800 900 1000 1100 1200 1300
eral continue to grow, the examples selected here to illus- Raman shift /cm1
trate the type and quality of CARS currently obtainable Figure 3. (a) PR/CARS spectrum (8001350 cm1 ) of ground-
come from an especially challenging group of condensed state BR-570 (data: unfilled circles) together with a c3 fit
phase molecules, namely photoactive proteins. Photoactive (solid line). The non-resonant background signal from the water
proteins not only share all the normal properties of liquids solvent is shown as a horizontal dashed line. (b) Lorentzian
which make them challenging experimentally to examine by lineshape CARS spectrum of BR-570 derived from the c3 fit
to PR/CARS data. Vibrational bands in the 900950 cm1 region
CRS (e.g. large nonresonant signals), but in addition their are assigned to hydrogen out-of-plane modes while those in the
photo-induced reactivity results in mixtures of species hav- 11001300 cm1 region are assigned to CC stretching modes
ing distinct structural and conformational properties that can within the retinal chromophore.
be analyzed via their respective vibrational spectra. Inde-
pendently, of course, these same proteins are susceptible The c3 -fit to the PR/CARS signal (<2% error) provides
to denaturation when exposed to heat such as that derived a CARS spectrum of BR-570 exhibiting Lorentzian line-
from moderate to high power laser excitation. For these rea- shapes (Figure 3).
sons, measuring CRS data from photoactive proteins must Normal mode analyses of the resultant BR-570 spectrum
consider carefully the laser powers to which the protein is lead to the assignment of specific features to well-defined
exposed and the time period over which each measurement vibrational modes (e.g. involving CC or CDC stretching,
is made. hydrogen out-of-plane motions, or methyl group in-plane
Examples of CARS data from a stable photoactive pro- rocking, etc.). These mode assignments provide an oppor-
tein (BR-570 in bacteriorhodopsin) and from transient inter- tunity to associate vibrational band positions and intensities
mediates within the photoreactions of two proteins (K-590 with structural changes occurring in the transient interme-
from BR and batho and lumi from Rh) are used here to diates formed throughout the BR photocycle.
demonstrate the high quality (S/N) structural information
that can be obtained. BR photoreaction intermediate, K-590. The room temper-
ature BR photocycle is amenable to examination by time-
PR/CARS from ground-state BR-570. The CARS spec- resolved vibrational spectroscopy since it reforms its stable
trum (8001350 cm1 ) of ground-state species in the BR ground-state BR-570 within 5 ms of excitation23 at room
photocycle, BR-570 (Figure 3), can be obtained with pico- temperature. PTR/CARS data (8001350 cm1 ), recorded
second !1 /!s excitation tuned into resonance with the from the room temperature BR photocycle 50 ps after it
BR-570 absorption spectrum, i.e. PR/CARS (Figure 1c). is initiated, can be quantitatively (<2% residual error) fit
Probe only Rhodopsin

(a) (PR/CARS)
Batho-rhodopsin
Rel. PTR/CARS intensity

Rel. CARS intensity
0 20 ps delay
K-590 BR-570 (PTR/CARS)
Lumi-rhodopsin
2.5 m
(b) delay
(PTR/CARS)
(a) 800 1000 1200 (b) 800 1000 1200

1100 1150 1200 1250 1300 1350
Raman shift /cm1
Raman shift / cm1
Figure 5. (a, top to bottom) PR/CARS and PTR/CARS data
Figure 4. (a) PR/CARS spectrum (11001380 cm1 ) of ground- (7501280-cm1 ) from ground-state bovine Rh and two transient
state BR-570 (data: unfilled circles) and PTR/CARS spectrum (50- mixtures of species within the Rh photoreaction at time delays
ps delay time) of the 80/20 mixture of BR-570 and K-590 (data: of 20 ps and 2.5 s. The c3 fits (solid lines) for each set
filled circles) together with their respective c3 fits (solid lines). of data are also shown. (b, top to bottom) Lorentzian CARS
The nonresonant background signal from the water solvent is spectra (7501280-cm1 ) for ground-state Rh and the two Rh
shown as a horizontal dashed line. (b) Lorentzian lineshape CARS intermediates, batho and lumi. Vibrational bands in this region are
spectra of BR-570 and K-590 derived from the corresponding assigned primarily to hydrogen out-of-plane and CC stretching
c3 fits to PR/CARS and PTR/CARS data. Vibrational bands in modes. Comparisons of these CARS spectra reveal large changes
the 11001300-cm1 region, assigned to CC stretching modes in the retinal structure throughout this part of the irreversible,
within the retinal chromophore, are clearly different between BR- room temperature Rh photoreaction. Data are adapted from Ujj
570 and K-590 indicating significantly different retinal structures et al.17 and Jager et al.40
(all-trans versus 13-cis). The figure is adapted and modified from
Ujj et al.16
dissociation of the retinal chromophore from the opsin pro-
(Figure 4) using c3 relationships (see above) and used tein. As a consequence, vibrational spectra of high S/N
to obtain a CARS spectrum containing >20 bands with must be recorded from a sample which is continuously
Lorentzian lineshapes assignable to the BR intermediate dissociating.
K-590.16 This analysis must quantitatively account for PTR/CARS data (7501300 cm1 ) from the Rh pho-
the large (typically >80%) BR-570 population present toreaction recorded 20 ps and 2.5 s after excitation are
in the reactive mixture from which the PTR/CARS data presented in Figure 5 together with their respective c3
are generated. The distinct vibrational structure of K-590 fits. These time delays correspond respectively with the dis-
(13-cis retinal) is evident from comparisons (Figure 4) tinct Rh intermediates batho40 and lumi.17 The analogous
with the analogous CARS data from BR-570 (all-trans PR/CARS data from Rh are also presented. Comparisons
retinal). of the three respective Lorentzian CARS spectra (Figure 5)
clearly show that these three species have dramatically
Rh photo-reaction intermediates, batho and lumi. Mea- different vibrational degrees of freedom (i.e. retinal struc-
surements of vibrational spectra in Rh are much more tures). Since these vibrational spectra are recorded only as
challenging experimentally than those in BR since the a function of reaction time, PTR/CARS data provide a real-
room temperature Rh photoreaction does not cycle to time, in situ view of the room temperature Rh photoreaction
the starting ground-state species, but rather results in the through its respective retinal structures.
5 FUTURE DEVELOPMENTS PTA Picosecond Transient Absorption

PTR/CARS Picosecond Time-Resolved Coherent
The long historical connection between advances in CRS Anti-Stokes Raman Scattering
and the development of laser sources will likely continue Rh Rhodopsin
with respect to wavelength42 and polarization parameters41 RR Resonance Raman
used, but given the limitations imposed by the timeband-
width product (tn) of the excitation laser pulses, the
time resolution used to measure the frequency-resolved REFERENCES
CARS spectra designed to provide structural information is
less likely to increase. Since the bandwidths of the vibra- 1. P.D. Maker and R.W. Terhune, Phys. Rev. A, 67, 801 (1965).
tional features measured with laser pulsewidths <2 ps are 2. A.B. Harvey (ed.), Chemical Applications of Nonlinear
broad, much of the vibrational mode information required Raman Spectroscopy, Academic Press, New York (1981).
to identify and characterize structural changes is obscured. 3. S. Druet and J.P. Taran, Prog. Quant. Electr., 7, 1 (1982).
Of course, even broad vibrational features measured with 4. M.D. Levenson, Introduction to Nonlinear Laser Spec-
laser pulsewidths <2 ps still provide a view of how given troscopy, Academic Press, New York (1982).
vibrational degrees-of-freedom change during a reaction, 5. Y.R. Shen, The Principles of Nonlinear Optics, John Wiley
but their assignment(s) to a specific normal mode(s), and & Sons, New York (1984).
thereby to a particular structural change, becomes increas- 6. R.J.H. Clark and R.E. Hester (eds), Advances in Non-linear
ingly difficult. Spectroscopy, Advances in Spectroscopy, John Wiley &
Sons, New York, Vol. 15 (1988).
The expanding wavelength regions over which pulsed
lasers operate (with and without nonlinear optics) should 7. H.C. Andersen and B.S. Hudson, Coherent Anti-Stokes
Raman Scattering, in Molecular Spectroscopy, ed. D.A.
increasingly provide more opportunities to apply CRS/ Long, American Chemical Society, New York, 142, Vol. 5
CARS methods to a larger range of molecular species (1977).
(e.g. stable ground states, transient photochemical prod- 8. V.F. Kamalov, N.I. Koroteev and B.N. Toleutaev, Time
ucts, excited electronic states, radicals, ions, etc.). Of Resolved Spectroscopy, Advances in Spectroscopy, eds
special note are the applications to biochemically impor- R.J.H. Clark and R.E. Hester, John Wiley & Sons, New York,
tant, large molecular systems (e.g. proteins, polypeptide, Vol. 18 (1989).
DNA, etc.). 9. A. Voroshilov, G.W. Lucassen, C. Otto and J. Greve, J.
The expanded use of polarization control in CRS/CARS Raman Spectrosc., 26, 443 (1995).
also promises to provide new data sensitive to changes 10. N. Bloembergen, M. Lotem and R.T. Lynch, Jr, Indian J.
in the interactions present between the optically resonant Pure Appl. Phys., 16, 151 (1978).
chromophore (e.g. retinal) and its environment (e.g. the 11. Y. Prior, IEEE J. Quantum Electron, 20, 37 (1984).
amino acids composing the binding pockets in BR and Rh). 12. S.Y. Yee, T.K. Gustafson, S.A.J. Druet and J.-P.E. Taran,
The same type of information should be accessible from Opt. Commun., 23, 1 (1977).
chromophores in solutions and solids. 13. S.Y. Yee and T.K. Gustafson, Phys. Rev. A, 18, 1597 (1978).
In general, the results produced since 1980 have made 14. S. Mukamel, Principles of Nonlinear Optical Spectroscopy,
it evident that the examination of molecular systems by Oxford University Press, New York (1995).
nonlinear optical methodologies has only begun. There is 15. L. Ujj, A. Popp and G.H. Atkinson, Chem. Phys., 188, 221
much evidence to suggest that CRS/CARS applications (1994).
will continue to increase with respect to both the type 16. L. Ujj, F. Jager, A. Popp and G.H. Atkinson, Chem. Phys.,
and complexity of molecular systems examined and the 212, 421 (1996).
sophistication of the measurements themselves. 17. L. Ujj, F. Jager and G.H. Atkinson, Biophys. J., 74, 1492
(1998).
18. J.R. Andrews and R.M. Hochstrasser, Chem. Phys., 62, 87
19. L. Ujj, B. Volodin, A. Popp, J.K. Delaney and G.H. Atkinson,
BR Bacteriorhodopsin Chem. Phys., 182, 291 (1994).
CRS Coherent Raman Scattering 20. N.I. Koroteev, Usp. Fiz. Nauk., 152, 493 (1987).
PR/CARS Picosecond Resonance Coherent 21. M. Nissum, J.-M. Funk and W. Kiefer, J. Raman Spectrosc.,
Anti-Stokes Raman Scattering 30, 605 (1999).
PSCARS Polarization Sensitive Coherent Anti-Stokes 22. B.N. Toleutaev, T. Tahara and H. Hamaguchi, Appl. Phys. B,
Raman Spectroscopy 59, 369 (1994).
23. J.K. Lanyi, Int. Rev. Cytol., 187, 161 (1999). 34. T. Lang, K.L. Kompa and M. Motz, Chem. Phys. Lett., 310,
24. D.S. Kliger and J.W. Lewis, Israel J. Chem., 35, 289 (1995). 65 (2000).
35. F. Lindenberger, R. Stockl, B.P. Asthana and A. Laubereau,
25. G.H. Atkinson, F. Jager and L. Ujj, Laser Chem., 19, 1, 127
J. Phys. Chem. A, 103, 5655 (1999).
(1999).
36. Y. Zhou, S. Constantine, S. Harrel and L.D. Ziegler, J. Chem.
26. M. Alden, Combust. Sci. Technol., 149, 1 (1999).
Phys., 110, 5893 (1999).
27. P.H. Bouma and L.P.H. DeGoey, Combust. Flame, 119, 133
37. T. Lang and M. Motzkus, J. Raman Spectrosc., 31, 65
(1999).
(2000).
28. A.V. Mokhov and H.B. Levinsky, Combust. Flame, 118, 733
38. A. Voroshilov, C. Otto and J. Greve, Appl. Spectrosc., 50, 78
(1999).
(1996).
29. S.P. Kearney, R.P. Lucht and A.M. Jacobi, Exp. Therm. Fluid
39. A.Y. Chikishev, N.I. Koroteev, C. Otto and J. Greve, J.
Sci., 19, 13 (1999).
Raman Spectrosc., 27, 893 (1996).
30. V.N. Ochkin, S.Y. Savinov, S.N. Tskhai, U. Czarnetzki,
40. F. Jager, L. Ujj and G.H. Atkinson, J. Am. Chem. Soc., 119,
V. Schulz-von der Gathen and H.F. Dbele, IEEE T. Plasma
12 610 (1997).
Sci., 26, 1502 (1998).
41. A. Voroshilov, Polarised Coherent Raman and Saturation
31. L.X. Wang and W.E. Jones, Can. J. Phys., 76, 699 (1998). Micro-Raman Studies of Biological Macromolecules, PhD
32. S. Meyer and V. Engel, J. Raman Spectrosc., 31, 33 (2000). Dissertation, University of Twente, The Netherlands (1997).
33. G. Knopp, I. Pinkas and Y. Prior, J. Raman Spectrosc., 31, 42. P.C. Chen, C.C. Joyner and M. Burns-Kaurin, Appl. Opt., 38,
51 (2000). 5894 (1999).
Inverse Raman Spectrometry
Dionisio Bermejo
Instituto de Estructura de la Materia, Madrid, Spain
1 INTRODUCTION Contrarily to other coherent Raman spectroscopic tech-

niques such as coherent anti-Stokes Raman spectroscopy
The first experimental observation of an inverse Raman (CARS), SRS yields a spectrum that in line-shapes, intensi-
spectrum was reported in 1964 by Jones and Stoicheff.1 The ties and wavenumbers corresponds exactly to a spontaneous
emission of a ruby laser was transmitted through a cell con- Raman spectrum. The main advantage compared with the
taining toluene, placed near the focus, to produce spectrally spontaneous technique is realized when high resolution is
broad band anti-Stokes radiation. This radiation and the required as the instrumental resolution is only limited by
remaining original pulse at !0 were sent into a second cell the spectral bandwidth of the lasers used and no spectral
containing different liquids. Subsequent spectral analysis analysis device (refractive or interferometric) is required.
of the broad radiation showed absorptions at the positions Furthermore, the experimental setup can be easily arranged
!0 C !R , where !R corresponds to Raman-allowed transi- in order to observe the Doppler width, in spectral lines of
tions of the liquid contained in the second cell. In this way nearly isolated molecules, corresponding to the frequency
Raman spectra from several liquids (benzene, nitromethane, of the observed transition, that is, to infrared (IR) fre-
pyridine, etc.) were recorded. This experimental arrange- quencies for the case of vibrational transitions, even if the
ment was designed in order to show evidence for the lasers involved emit in the visible region. This is easily
absorption occurring during incoherent scattering at optical achieved in SRS by using a co-propagating configuration
frequencies. of the lasers, whereas in the spontaneous technique the
This pioneering work gave birth to a new spectroscopic much more problematic observation of the forward scat-
Raman technique based upon the previously discovered tering would be required. These facts explain why, unlike
stimulated Raman effect.2 Since the Raman spectrum was other Raman techniques, SRS finds its main application in
observed as an induced absorption, contrary to the usual the field of high-resolution spectroscopy of diluted gases.
observation in emission, the name Inverse Raman Scatter- No detailed description of the theory of SRS will be
ing was given to this technique. In subsequent experiments, given in this article. Those interested in the theory can refer
two coherent monochromatic beams of radiation, at least one to Section 1 in this volume of the Handbook, to any of
of them tunable, were used to record Raman spectra based the classical reference texts,3,4 or to some of the excellent
upon this principle. Since the observation of the spectrum reviews on coherent Raman spectroscopy.5 8
can be made as a gain on the lower frequency laser or as a
loss on the higher frequency one, depending on the experi-
2 EXPERIMENTAL
mental arrangement, the names of Raman gain spectroscopy
or Raman loss spectroscopy are currently used. In a more The first gas phase stimulated Raman spectrum was reported
general way this technique can be referred to as Stimulated in 1978.9 The technique used is a variation of SRS in which
Raman Spectroscopy (SRS) comprising spectroscopy based an analyzer is used to detect polarization changes in one
upon both, simultaneously occurring, effects. laser induced by a second when the frequency difference
matches a Raman-allowed transition of the medium. The
John Wiley & Sons Ltd, 2002. molecule under study was H2 , and this variation of SRS
overcame the main problem of the SRS technique, namely 10

0
the amplitude noise that masked the weak induced absorp- s
tions. Up to the present day, this technique called Raman 1
Induced Kerr Effect Spectroscopy (RIKES)10 is considered
the Raman technique with the highest potential sensitiv- (3)
ity. Nevertheless, the unavoidable birefringence of windows (R)
and other optical components and possible line-shape dis- ns
12
tortions, common to other polarization techniques, have
1 2 = R
limited the expansion of this spectroscopy. 2
Another early variant of SRS was developed by Rosasco
and Hurst11 in order to improve the ultimate sensitivity of Figure 1. Basic schematic of the quasi-CW stimulated Raman
this technique. The proposed approach was called phase- process.
modulated SRS and was based on the use of two pump
lasers, frequency-degenerated, whose relative phase was From equations (1) and (2), it is seen that sensitivity
modulated. Several spectra from 12 CH4 (n1 and n3 ) and scales as P [P!/F]1/2 and, therefore, the use of
D2 were reported. pulsed lasers can dramatically increase the sensitivity as
The first systematic attempts to develop a precise tool long as saturation of both the transition and the detector is
for high resolution spectroscopy based on the stimulated avoided.
Raman effect were reported by Owyoung et al. between Taking into account all these considerations, Owyoung
1977 and 1978.9,12 14 In all those early experimental setups proposed the experimental setup called quasi-CW SRS,16
two continuous wave (CW) lasers were used. Absorptions that, until today, is the most widely used scheme for SRS.
on a frequency-fixed laser were sought when a second, tun- A time-gated CW laser (usually an ArC or KrC laser) is
able, CW laser, was scanned. In order to improve the ulti- used as probe. The pulsed pump radiation is produced by
mate sensitivity attainable with this technique, a multipass amplifying the output of a single mode CW dye laser in
cell with a structure of a nearly confocal resonator was the way described below. An additional advantage of this
configuration is that, usually, the pulse duration of the pump
used. Although a gain enhancement up to a factor of 50
beam and therefore of the transient absorption is of the
was achieved, it soon became clear that SRS, in its CW
order of 10 ns. So, the region of 100 MHz in the noise
version, would always be a low sensitivity technique, even
spectrum becomes important, and the amplitude noise of the
taking into account foreseeable technical improvements of
probe laser, mainly dominated by the acoustic contribution,
the laser sources.
is nearly shot-limited.
A careful look at the expressions relating the induced
In a very schematic way this technique is illustrated in
Raman loss (gain) with the power of the lasers gives a hint
Figure 1. When the frequency difference between the pump
of the further improvement of the technique. For the usual
and probe lasers matches a Raman-allowed transition, an
case of two Gaussian collinear beams having the same ray
absorption is imprinted on top of the long duration pulse
parameters, the absolute power variation in the probe laser
of the probe laser. This long pulse is high-pass filtered
can be expressed as equation (1),15
in order to isolate the absorption that temporally follows

P! Nl2 p d2 s the pump laser profile. Then, it is integrated with a boxcar

P! hc2 d dl1 P 1 integrator and averaged over several pulses. The plot of this
averaged area versus the frequency difference between the
where N is the population difference between the lower pump and probe lasers when the former is scanned provides
and upper levels of the transition, P is the power of the Raman spectrum.
the pump laser and d2 s/d dl1 is the spectrally resolved A detailed schematic of the experimental setup for the
differential Raman cross-section. The signal-to-noise ratio quasi-CW stimulated Raman spectrometer at the Instituto
can be expressed as equation (2),15 de Estructura de la Materia in Madrid, Spain, is shown in
S G Figure 2.
D 2 2 2
N f G C ! C 2[h!F/QP!]g1/2
2
where G is the gain (or loss) expressed as power variation 2.1 Quasi-CW SRS spectrometer
in equation (1), and ! stand for root mean square ampli-
tude fluctuations of the pump and probe lasers, respectively, 2.1.1 Probe laser
within the detection bandwidth F and Q is the detector An ArC or KrC laser is used as probe laser, operating on the
quantum efficiency. emission lines 529 nm, 514 nm or 488 nm for the former and
Inverse Raman Spectrometry 3
Oscilloscope
Lock-in chanel A
0 9
HVA1
x
PID HVA2 9
v div
a b
amplifier
c d e f
chanel B
ajus
. 9 9 a b
e f c d
9
time
ajus
Etalon Ar + An.
I2 or Te2
Pol. /4
FPE AOM
FPE
Ar + SHG Chopper
Nd:YAG
Seeder Raman cell Prism
Unidirectional Dye amplifier

device
Ring dye laser
Dye stabilization Diffraction

station grating
I2 cell
FPE
Wavemeter
ADC Boxcar Preamplifier
PC
Figure 2. Schematic diagram of the quasi-CW stimulated Raman spectrometer in the Instituto de Estructura de la Materia, CSIC,
Madrid. FabryPerot etalon (FPE); Proportional integral differential (PID) amplifier; HVA, High voltage amplifier; AOM, Acousto-optic
modulator; SHG, Second harmonic generator; ADC, analog to digital converter; An., Analyzer, Pol., Polarizer.
647 nm for the latter. The choice of the laser and emission re-establish the original frequency. In this way, a first sta-
line depends mainly on the Raman shift of the region to be bilization loop is closed and the laser cavity is slaved to the
studied. An intracavity etalon ensures that the Transverse mechanically much more stable cavity of the FPE, resulting
Electric and Magnetic (TEM00 ) probe laser is operated in a in a jitter reduction down to a few hundreds of kilohertz.
single longitudinal mode. (TEM00 is the lowest order (0,0) In order to reduce the long term drift of the FPE and to
mode of a typical gas laser resonator. For this mode, the provide an absolute frequency reference, a second loop
field variation depends only on axial distance and distance is required. Saturated absorption18 or polarization19 sub-
from the axis of propagation.) To control the laser cav- Doppler signals of 127 I2 20 or 130 Te2 21 molecules are used
ity length, and therefore the frequency of the single mode, for this purpose. A general description of this type of fre-
one or both cavity mirrors are cemented to piezo-ceramic quency stabilization technique can be found elsewhere.22,23
elements. A fraction of the laser emission is directed into The hyperfine signal from these molecules is used to slave
a confocal FPE with a spectral range of 300 MHz and a the cavity of the FPE and therefore the frequency of the
finesse between 5 and 20. The emission of the laser is tuned laser, in a similar manner to that employed in the first loop
(by adjusting the high-voltage applied to the piezo-ceramic using a FPE transmission fringe. In this way the frequency
element) to the half height of a transmission fringe of the drift of the laser is eliminated, the absolute frequency of
FPE. Deviations in the frequency of the laser result in vari- the probe laser being precisely known from the frequency
ations in the FPE transmission and, consequently, in the of these spectral lines.20,21
signal output of a photodiode placed after it. This electri- The power of the probe laser has to be maintained
cal signal is conditioned by a PID amplifier.17 The output typically in the range between 200 and 500 mW to avoid
of the PID is high-voltage amplified and sent to the piezo- unwanted saturation effects on the detector, as explained
ceramic element controlling the cavity length in order to below.
2.1.2 Pump laser 2.1.3 Data acquisition and calibration

The spectral resolving power of the instrument depends Since the wavenumber of the probe laser is precisely known
directly on the spectral bandwidth of both pump and probe and its frequency drift is negligible, the final accuracy in
lasers, since the instrumental line shape (ILS) is the con- the determination of the spectral wavenumbers is mainly
volution of their spectral profiles. As the bandwidth of limited by the non-linearity of the pump laser scan and by
the probe laser is of only several hundreds of kilohertz, the precision in the wavenumber scale calibration.
it is the pulsed pump laser that limits the final resolv- Important deviations from linearity in the dye laser scan
ing power of the spectrometer. From equation (2) it is are usually present and can greatly degrade the accuracy
seen that pulses as short as possible are required in order in the determination of the line positions if a simple
to increase P , and therefore the signal-to-noise ratio. linear interpolation is performed. In order to cope with this
However, too short pulses are not compatible with the problem, a small fraction of the CW dye laser is taken and
requirement of a high resolving power since the spectral the transmission through a FPE, with similar characteristics
bandwidth of a pulsed laser is ultimately limited by the to the one mentioned above, is recorded simultaneously
Fourier transform of its temporal profile. A compromise is with the spectrum. The recorded fringes are used as event
required. The pump laser is produced by pulse amplifying markers. The spectral line positions are then interpolated
the output of a single mode, actively frequency stabilized between these marks, i.e. between 150 MHz intervals.25
CW dye laser (usually with a ring cavity), in a two- or For the wavenumber calibration two main different
three-stage dye amplifier. This amplifier usually follows approaches are possible. One possibility is to measure the
the scheme proposed by Drell and Chu,24 in which one wavenumber of the CW dye laser at the beginning and at
or two transverse amplification stages are followed by a the end of every scan with a high accuracy wavemeter.27
final longitudinal stage, in which the pumping laser co- or Another possibility is the simultaneous recording, using
counter-propagates with the amplified one. The radiation a fraction of the CW dye laser output, of an absorption
used for pumping the amplifier is the second harmonic of spectrum of the I2 molecule.28 In both cases the achieved
a Nd : YAG laser. The amplified dye follows the temporal accuracy is of the order of 0.001 cm1 .
and spatial shape of this laser, it being therefore impor- Once both lasers have interacted with the sample, the
tant that these characteristics are appropriate. Single mode loss (gain) imprinted on the intensity profile of the probe
operation of the Nd : YAG laser is required in order to laser is detected. First, the intense pump laser must be
avoid mode beating that would result in a complex spec- spectrally filtered out. As the weak Raman signal occurs
tral profile of the amplified dye, and therefore of the ILS in synchronism with the pump laser pulse, care must be
of the spectrometer. This is achieved by using an injec- taken to avoid photons from this laser, scattered from any
tion seeded laser. Commercial single mode Nd : YAG lasers laboratory surfaces, hitting the detector and masking the
usually produce temporally and spatially Gaussian pulses Raman signal. It must be emphasized that this spectral
of 10 ns (full width at half-maximum, FWHM) at a rep- filtering, performed by the usual means (prisms, gratings,
etition rate of 10 to 20 Hz and with a typical energy of spatial filters, etc.), has only the above mentioned purpose
about 300 mJ in the second harmonic. Sometimes the cav- and does not determine, to any degree, the resolving power
ity of the laser is extended in order to generate somewhat of the system.
longer pulses and therefore to increase the resolving power Once the pump laser has been removed, a silicon photo-
of the spectrometer. For a Gaussian pulse of 12 ns (FWHM), diode with a rise time of the order of 1 ns or less is used to
the spectral bandwidth of its Fourier transform corresponds detect the probe laser. Care must be taken in order not to
to 2.7 103 cm1 , compatible with the nearly Doppler- saturate the photodiode either in the low frequency range or
limited widths of the ro-vibrational spectral lines of most at the 100 MHz range. A too small illuminated area on the
light molecules. photodiode surface can result in a saturation at 100 MHz
In this way, tunable, Fourier-transform limited pulses of without any apparent effect at low frequency, resulting in a
a power of about 5 to 20 mJ are produced. Higher peak decrease or disappearance of the Raman signal, even if the
power may result in a degradation of the resolving power photodiode shows a normal behavior in detecting the long
because of the magnetic sublevel splitting due to the AC- probe laser pulse.
Stark effect.15,25,26 As the strength of the SRS signal does The signal from the photodiode, after passing the high-
not depend on the degree of focusing, long focal length pass filter is amplified with a gain between 10 and 100 and
lenses (500 mm) are used (unless a good spatial resolution boxcar averaged over about 10 laser shots.
is required) in order to decrease the power density in the A typical scan can take about 20 min and cover 1.2 cm1 .
focal region and, therefore, the above mentioned AC-Stark If the region to be investigated is wider, consecutive over-
broadening. lapping scans have to be added together.
3 APPLICATIONS than provide an extensive review, since this would not be

possible here. It is worth pointing out that CARS has been
For years, the classical application of high resolution mole- used in parallel with SRS for most of the studies refer-
cular spectroscopy has been the determination of molecular enced below. As this article deals only with SRS, most of
ro-vibronic constants to characterize the energy levels. the CARS work will not be mentioned, but the interested
The impressive development of IR absorption spectroscopy reader should take into account that a knowledge of the cor-
during the last years, due to the development of laser spec- responding CARS work is essential in order to complete the
troscopy techniques and, to a greater extent, of Fourier picture of the activity in this field. A common situation is
transform spectroscopy (FTS) has reduced the interest on that the advantage of clean line profiles provided by the
high resolution Raman spectroscopy for this kind of study. SRS technique is preferred in the laboratory in order to
FTS in its IR version provides the possibility of simul- derive the fundamental parameters needed for diagnostics
taneously recording a very wide spectral region with an in combustion media, while CARS is preferred for tests on
excellent internal coherence in the wavenumber scale, fea- actual engines as it is an easier to implement technique and,
tures not equaled by laser spectroscopy techniques. On the under some conditions, a more sensitive one.
negative side, the complex apparatus function has limited The rest of this section will be classified, in a somewhat
the applicability of this technique when a detailed know- arbitrary way, according to the symmetry of the molecules
ledge of the spectral profile is required. Unpleasant wings studied.
in the line profiles arising from distortions, imperfect inter-
ferometer alignment or from the unavoidable apodization,
make IR or Raman laser spectroscopy techniques more 3.1 Diatomic molecules
appropriate for this kind of study.
Nevertheless, even for classical studies aimed at the Diatomic homonuclear molecules do not show any IR
determination of molecular constants from the wavenum- electric dipole absorption but show Raman activity. Further-
bers of the spectral lines, a field of applicability for high more, for heteronuclear molecules, transitions correspond-
resolution Raman techniques, particularly SRS, has been ing to J D 0 are allowed in the Raman effect, giving rise
maintained. The reason for this stems from the different to a Q branch that is absent in the IR absorption spectrum.
selection rules of the Raman effect, as compared with the IR
electric dipole absorption. Although most vibrational levels
3.1.1 Nitrogen
with a few vibrational quanta are accessible from populated
levels with the appropriate symmetry by IR absorption spec- N2 has been widely investigated by SRS as it is a very good
troscopy, there are still some cases where the information monitoring molecule for temperature measurement in com-
provided by the Raman spectrum is irreplaceable. Inciden- bustion media. Already in 1981, using the less sensitive
tally, it is worth pointing out that, although at the time of CW technique, Owyoung reported an almost completely
writing this paper some promising projects are going on, resolved Q branch of 14 N2 .29 In 1987, Lavorel et al.30
Fourier transform Raman spectroscopy has not contributed reported a precise set of molecular constants by extrapo-
significantly, up to the present, to the field of high resolu- lating to zero density the spectral peak positions measured
tion, as has been the case for its IR counterpart. in the SRS spectrum at different densities up to about 8
Another interesting characteristic of Raman spectroscopy, amagat. (The amagat number is a dimensionless measure of
either in its spontaneous or in its nonlinear versions, includ- gas density expressed as the ratio of the measured density
ing SRS, is that the signal is produced in the focal region of divided by the gas density at standard temperature and pres-
the lasers, rather than along a more or less long path as is sure (i.e. 0 C, 1 atm). It is unique for each gas, but varies
the case in IR absorption. This characteristic makes Raman with temperature and pressure). In this work, pressure-
techniques and particularly coherent Raman techniques, induced shifts were also measured to be 5.5 103 cm1
very well adapted to measurements where spatial resolu- amagat1 . In 1990 a new, more accurate set of mole-
tion is required as in monitoring temperature or density in cular constants for this molecule was reported by Tabyaoui
hostile media like reactors, combustion engines and flames. et al.31 These data were provided also by the laboratory at
This characteristic feature also makes useful the combina- the Universite de Bourgogne, Dijon, France, using a quasi
tion of SRS and free expansion jet techniques in order to CW SRS spectrometer. The authors recorded the spectrum
obtain spectra of cooled or unstable molecular species. of this molecule at 1300 K and, in this way, they were able
In the rest of this paper some of the most interesting con- to observe the first hot band and propose a more accu-
tributions of SRS will be briefly mentioned. The selection rate set of constants for the band origins, B1 -B0 , B2 -B1
is intended to give a general overview of the field, rather and D1 -D0 and related Dunham constants. In 1987, Nibler
and Yang5 reported an SRS spectrum on the free expan- lasers was used. From these measurements in a range
sion jet of a mixture of He and N2 . A weak feature near between 25 and 190 kPa, collisional broadening coefficients
2326.4 cm1 , whose intensity and shape were dependent on were derived. One year later this study was extended to
the He proportion in the mixture, was attributed to small N2 higher temperatures, up to 1310 K.36
aggregates. This work was continued in a study of phase Line-mixing effects in the Q branch of this molecule
transitions of large N2 clusters in a free expansion jet.32 were also investigated in 1983 by Rosasco et al. in the
The high spectral resolution and point probing capabilities SRS spectrum.37 Pressure broadening and line-mixing para-
of the technique were combined in this study. Figure 3 meters were derived and the effect was described using
shows the evolution of the spectrum downstream in the a relaxation matrix formalism. In 1990, the shifting and
expansion. SRS spectra of molecular crystals of N2 and of broadening induced in the Q branch of N2 by collisions
matrix-isolated N2 and CO have also been recorded.33 with CO2 were reported.38 This study was performed in
Collisional effects on the lineshapes of this molecule the low density range (between 0.3 and 1.0 atm) and the
have also been investigated by SRS. In 1982 the effect derived parameters were extrapolated to the higher density
of Dicke narrowing34 on this relatively heavy molecule regime (up to 30 amagat). The validity of different models
was observed by Millot et al.35 In order to increase the to predict high density behavior from low density measure-
Doppler width and, therefore, make more evident the Dicke ments was discussed. In 1992 a study on pure N2 in the
narrowing effect, a counter-propagating, rather than the range from 150 to 6800 bar was reported.39 This paper also
included a similar study on the CO molecule. In 1994 a
usual co-propagating, configuration of the pump and probe
line coupling study on anisotropic Raman branches in the
N2 molecule was reported.40
Also the first observation of optical Stark splitting of
a molecular Raman transition was reported in the N2
molecule by Farrow and Rahn.26
X /D
1.2
3.1.2 Oxygen
2.0 Oxygen has also been widely investigated as a probe
4.7
molecule for temperature and density monitoring of com-
6.0
bustion media. It is a paramagnetic molecule having a
7.5 3
g triplet ground state and therefore a triplet structure
10
in the rotational lines of the transitions within the ground
13
electronic state. The corresponding satellite lines in the Q
21
27 branch, resulting from a change in the electron spin dur-
28 ing the transition, were observed for the first time by Hill
30 et al.41 These spectra were taken in an adiabatic free expan-
31 sion in order to cool the gas down to about 40 K. The
33 direct transition at about 2 cm1 between the sublevels of
35 the ground state was also observed in the SRS spectrum of
38 this molecule.42
44 In 1992, Rouille et al.43 derived a set of molecular con-
54 stants for the v D 1, 2 and 3 states combining microwave
73
data with the data obtained from their SRS spectra from
the S and O branches of the fundamental and from the
110
Q branch of the first and second hot bands. In order to
2326.0 2326.5 2327.0 observe these weak transitions with a good signal-to-noise
cm1 ratio, spectra were taken using a multipass cell and a sam-
Figure 3. SRS spectra at different points throughout the full ple pressure of 1 atm at 1350 K (for the hot bands). Effects
range of the neat N2 expansion with T0 D 160 K, P0 D 30 atm. of collisional shifting, broadening and line-mixing on the
X/D is the distance from the nozzle (measured in units of Q branch of this molecule have been widely investigated
nozzle diameter D) under identical expansion conditions. See
Beck et al.32 for full details. [Reproduced by permission of Ameri- by the Universite de Bourgogne, Dijon, and Universite
can Institute of Physics from D.R. Beck, M.F. Hineman and de Franch-Comte, Besancon, groups in France and several
J.W. Nibler, J. Chem. Phys., 92, 7068 (1990).] models have been used to fit the observed profiles and to
derive the shifting and broadening parameters. In particular, This work was aimed at the laboratory observation of this
self-broadening and broadening by N2 were investigated for rotationally excited line that had been observed in the
temperatures up to 1350 K and pressures up to 2 atm44 in Orion molecular cloud. The spectrum was recorded on an
1992. One year later, this work was extended to higher airmethane flame and the Q(27) line of the CO molecule
pressures, up to 1255 bar at 295 K and 176 bar at 446 K.45 that accidentally appears near the S(9) line of H2 was used
Since the ground state of this molecule has 3 g sym- as a reference for wavenumber calibration. A complete set
metry, additional broadening mechanisms resulting from of molecular constants for the ground state was derived
the electronic spin relaxation are possible. These effects from the laboratory and astrophysical data.
were investigated in the o O(J,5) triplet up to pressures of Since H2 and D2 are very light molecules, the rota-
4.4 atm.46 In this work, experimental profiles were fitted and tional structure of the corresponding Q branches can be
coupling parameters were derived by using a Rosenkranz easily resolved in the spontaneous Raman spectrum. There-
perturbation treatment.46,47 Figure 4 shows one of the spec- fore, the main interest of SRS in this molecule is not
tra published in this article showing the triplet structure of the determination of molecular constants but, rather, the
the above mentioned transition. A letter by the same group study of collisional effects where a very narrow apparatus
published later that very year summarized the fitting and function is called for. In 1989, Farrow et al.50 reported non-
scaling laws for high temperature Q branch collapse in the Lorentzian line profiles, due to inhomogeneous broadening
SRS spectrum of O2 and O2 H2 O mixtures.47 of the vibrational lines in the Q branch of H2 perturbed
by Ar, Kr and N2 at densities up to 27 amagat. They also
3.1.3 Hydrogen and deuterium reported the absence of such effects for light perturbers like
Hydrogen and deuterium have also been studied by SRS. He or H2 . In 1990 Rahn and Rosasco51 published the results
The former is a common optical test molecule for the of a study on the dependence on temperature (up to 1000 K)
combustion medium in rocket engines. As early as 1978 and rotational quantum number, J (up to J D 5), of the
the SRS spectrum of the Q01 (1) transition was reported Raman Q branch self-density-shift coefficients. Collisional
by Owyoung.9 Two years later the SRS spectrum of the broadening and shifting in the pure rotational spectrum in
Q01 (2) line of D2 was reported by the same author.29,48 the range between 1 and 35 amagat were reported in 1994
In this work a sub-Doppler signal was detected combining by Le Flohic et al.52 A general approach to these pheno-
SRS with saturation spectroscopy. An SRS spectrum of mena, including both speed- and phase-changing collisions
the D2 molecule has also been recorded by Rosasco and was published in the same year.53 Spectra of H2 perturbed
Hurst11 with the phase modulation technique mentioned by Ne, Ar, and Xe in the range between 395 and 800 K
above. In 1985 the SRS spectrum of the pure rotational were recorded and analyzed in this paper. This study was
S(9) line of H2 was reported by the Owyoung group.49 extended by the same group one year later to the case of
a lighter perturber, He, and to the range between 296 and
995 K and 10 to 20 amagat.54 One year later, this work
3 was further extended to the H2 N2 system,55 particularly
J 4 v = 1, N= 3
2 interesting for the study of combustion media. New, more
0
0
0 ++ precise peak positions for the Q(7) to Q(9) lines for pure H2
Signal (a.u.)
5 were also reported in this work. In 1998, collision effects

J 6 v = 0, N = 5
4 measurements on rotational lines of H2 , pure and perturbed
by He, were reported in a range between 300 and 1000 K at
a density of about 10 amagat.56 Corrections for frequency
shifts were included for the first time. Recently a compre-
0 hensive study of the line profiles of the Q branch of H2
3
perturbed by Ar in the range between 300 and 1000 K and
between 0.2 to 11 amagat has been published.57 A unique,
1530.10 1530.55 unified model, accounts for all the observed features in this
Raman shift /cm1 wide range of experimental conditions. A similar consis-
Figure 4. Schematic representation of the energy levels for the tency was reported when N2 was the perturber.58
o
O(J,5) transitions and spectrum of the 16 O2 molecule at 295 K
and 132.5 Torr. The lower trace shows the difference (multiplied
by 3) between the observed and the calculated spectrum, using a 3.1.4 Fluorine and chlorine
Galatry140 profile. [Reproduced by permission of Academic Press
from G. Millot, B. Lavorel and G. Fanjoux, J. Mol. Spectrosc., The fluorine spectrum has been investigated in SRS.
176, 211 (1996).] Martnez et al.59 recorded the spectrum of the Q branch and
several lines of the O and S branches of the fundamental levels by infrared spectroscopy have made this molecule
and the Q branch of the first hot band, deriving a set of a favorite for SRS. As stated before, an overview of
molecular constants for the v D 0, 1 and 2 vibrational states the work done on this molecule would not be com-
of this molecule. plete without the complementary work done using CARS.
Three different isotopomers of the chlorine molecule, Also some interesting work on this molecule has been
namely 35 Cl2 ,35 Cl37 Cl and 37 Cl2 , have been investigated done using a variation of SRS, Photoacoustic Stimu-
by the same group.60 Q branches of the fundamental, first lated Raman Spectroscopy (PARS), which is discussed in
hot bands and some O and S lines have been observed and Section 4.
a set of molecular constants derived for the v D 0, 1 and 2 In 1982, Baran et al.64 evidenced the motional narrowing
vibrational states. in the two components of the n1 /2n2 Fermi dyad by using
high resolution SRS. Almost simultaneously, a similar
study was done on the upper frequency component of
3.1.5 Carbon and nitrogen monoxides
the dyad.65 The lower component of this dyad was used
These heteronuclear molecules have IR absorptions and by the Universite de Bourgogne, Dijon, group to test
therefore have been less extensively investigated by the enhancement in sensitivity achieved by the use of
Raman spectroscopy. NO has been studied using Ionization a multipass cell.66 A very comprehensive study of the
Detected Stimulated Raman Spectroscopy (IDSRS). As a collisional effects on the shapes of the above mentioned
result of the interaction between light and matter in a Q branches can be found elsewhere.67,68 Spectra were
stimulated Raman process, molecules are pumped to the recorded in the range between 0.01 and 50 amagat at
upper state of the involved transition. This fact has been 295 K and 0.01 and 20 amagat at 500 K for the upper
advantageously used for increasing the sensitivity of the frequency component and between 0.2 and 50 amagat at
SRS technique by selective ionization of the molecules 295 K and 0.5 to 20 amagat at 500 K for the lower one.
in the excited state and ion detection. With this modified A complete modeling of the observed lineshapes was done
version of the SRS technique, the spectrum of the Q branch and vibrational relaxation coefficients for both bands were
of the fundamental of NO was recorded.61,62 Two series of derived. This work was extended to higher temperatures
transitions corresponding to the two sublevels of the ground (up to 900 K) and from these spectra together with the
state of this molecule, 2 1/2 , 2 3/2 were observed in the ones reported in the previous two references, values of
Q branch. the vibrational linewidth and shift of the lower component
Line mixing effects in the Q branch of CO perturbed by of the dyad were derived as a function of temperature.69
He have been investigated.63 Spectra have been recorded at This is valuable information for practical applications
77 and 195 K in a pressure range between 50 and 1300 mbar such as CARS thermometry. One year later the energy
with a CO concentration of only 1% in order to avoid self- corrected sudden (ECS) model was generalized to account
perturbation effects. Close coupling scattering calculations for the anisotropic SRS spectrum of both components
in a pure quantum mechanical frame show a very good of the Fermi dyad.70 These studies were pursued in the
agreement with the experiment without the use of any following years by the same group who developed a
adjustable parameters. double resonance RamanIR experiment.71 In this case
the pump process used to populate the v1 D 1 state was
a doubly pulsed CARS process followed by IR absorption
3.2 Linear molecules of a CO2 CW laser which probed the transition from this
state to the v3 D 1 state. A similar approach using time
Linear molecules with a center of symmetry have some resolved spectroscopy was followed 4 years later.72 This
transitions that show no electric dipole IR activity. There- work was generalized to study collisional relaxation in
fore, in some cases, high resolution Raman spectroscopy is CO2 Ar and CO2 He mixtures.73 The CARS technique
the only, or at least the more convenient, alternative for the was used to study the line-mixing effect in a type
study of these states. hot band, taking into account the vibrationrotation angular
momentum coupling.74
3.2.1 Carbon dioxide

3.2.2 Acetylene
This molecule has been widely investigated by SRS. The
presence of the very strong bands of the n1 /2n2 Fermi Acetylene is a centrosymmetric linear molecule and there-
dyad, its implication in atmospheric and combustion pro- fore some totally symmetric levels are not easily accessible
cesses and the difficulty of studying some vibrational by IR absorption spectroscopy. Vibrationally excited states
Virtual state
3

2
Virtual state
529 nm 591 nm 1
532 nm 594 nm
0
3
v2 = 2 2
J
1 3340 3341 3342
0
Wavenumber / cm1
3
Figure 6. Q branch of the n1 C n2 n2 band for 13 C2 H2 (2.6 kPa,
v2 = 1 2 J 195 K). The line marked with an asterisk corresponds to O(2) of
1
0 n1 . [Reproduced by permission of American Institute of Physics
from D. Bermejo, R.Z. Martnez, G. Di Lonardo and L. Fusina,
J. Chem. Phys., 111, 519 (1999).]
3
Ground 2
J from where the pumping process starts. The same technique
state 1
0
was used for recording the spectrum of the Q branch of
Pumping Spectroscopy
2n2 n2 of the 12 C2 D2 and 13 C2 D2 isotopomers.77 Spectra
Figure 5. Schematic of the pump and spectroscopy processes in of n2 , the associated hot bands (n2 C n4 n4 , n2 C 2n4
the high resolution double resonance RamanRaman technique. 2n4 , n2 C n5 n5 , n2 C 2n5 2n5 and n2 C n4 n5 n4
n5 ) for both isotopomers and n1 for 13 C2 D2 were also
of different isotopomers have been studied in our group in
recorded and analyzed and a complete set of ro-vibrational
Madrid by using a double resonance RamanRaman tech-
parameters of the involved levels was derived.
nique in which two consecutive SRS processes are chained
as schematically shown in Figure 5. The first one, dou-
bly pulsed, pumps a fraction of the population from the 3.2.3 Other linear molecules
ground to a vibrationally excited state. One of the lasers The Q branch spectra of n1 and n2 and associated hot
used for this process comes from a second doubling of bands of the cyanogen isotopomers 12 C2 N2 and 13 C2 N2
the Nd : YAG and has a bandwidth of about 3 103 cm1 have been recorded and analyzed yielding sets of molecular
while the other is a pulsed dye laser with a spectral width parameters of the involved ro-vibrational levels.78,79
of about 2 101 cm1 . The second one, therefore, deter- Molecular parameters of the dicyanoacetylene molecule
mines the fraction of the rotational levels of the ground (N
CC
CC
N) have been derived from electron diffrac-
state that are pumped to the vibrationally excited state. After tion data and the SRS spectrum.80
a delay of about 15 ns to allow for thermal redistribution The strongly perturbed region of the n2 Q branch of
among the rotational levels of the populated state and to diacetylene (H
CC
CC
H) has also been recorded and
avoid unwanted Stark broadening associated with the laser analyzed.81
fields, a second quasi-CW SRS process is used to record the
high resolution Raman spectrum from the populated level
to a higher excited vibrational state. With this technique the 3.3 Spherical top molecules
Q branches of the 2n2 n2 transitions of the 12 C2 H2 and
13
C2 H2 molecules have been recorded and sets of molecu- These molecules have been extensively studied with the
lar parameters for the v2 D 2 state of both molecules have SRS technique as their totally symmetric bands are IR-
been derived.75 In a similar experiment, the n1 C n2 n2 forbidden and, by contrast, exhibit a very high Raman
bands of 12 C2 H2 and 13 C2 H2 and the 2n2 C 2n5 n2 band of cross-section.
13
C2 H2 have been recorded.76 Figure 6 depicts the Q branch
of n1 C n2 n2 for 13 C2 H2 to show the good signal-to-noise
3.3.1 Methane
ratio achieved in this experiment that, paradoxically, corre-
sponds to a hot band and has been taken at dry ice temper- Methane was one of the first molecules studied by SRS.
ature to increase the population of the ro-vibrational levels As early as 1978 Owyoung published the first spectrum of
the n1 band of 12 CH4 recorded with the CW technique.14 3.3.2 Other spherical top molecules
It should also be mentioned that a CARS spectrum was
The SRS spectrum of CF4 was reported in 1981.96 The
published one year later by Kozlov et al.82 As it was taken
n1 and n1 C n2 n2 bands were recorded. The spectra
at 80 K with a CW system, it provided the best resolution were assigned and a set of molecular constants of the
attained for this spectrum and allowed a more precise involved levels was derived. This molecule was reinvesti-
assignment. One year later an SRS spectrum was recorded gated in 1994 combining CARS and SRS data to derive
in a free expansion jet of 12 CH4 achieving a temperature a set of molecular constants for the v1 D 1 and v2 D 2
of 13 K.83 Only three lines, corresponding to the three states.97
different nuclear spin isomers, remain in the spectrum at The SRS spectrum of the n1 band of 28 SiH4 was first
this temperature. In 1991, Millot et al.84 published an SRS investigated with this technique by Owyoung et al.98 as
spectrum of 2n2 reporting broadening coefficients for Ar, early as 1981. In 1984 the same band was studied again
He and self-perturbation. and some weak lines were identified as belonging to 29 SiH4
In 1992, Santos et al.85 reported spectra of this molecule present in natural abundance.99 IR and SRS data were
at 77 K and room temperature.85 In order to improve the combined100 to analyze the n1 /n3 system in a natural silane
accuracy of the wavenumber determination, the pump and sample. Collisional linewidths in the n1 band of 28 SiH4 self-
probe lasers were mixed in a LiNbO3 crystal to generate the perturbed at room temperature were measured in 1990.101
IR frequency difference radiation. This radiation was used Linemixing in the Q branch of the n1 band of this molecule
to record, simultaneously with the SRS spectrum, an IR in the 0.4 to 1 atm range was observed and interpreted using
absorption spectrum of some lines of n3 lying in the same a strong collision model.102
spectral region that were used as a standard for calibration. Spectroscopic constants for the v1 D 1 level for five
This calibration procedure and a careful evaluation of all isotopomers present in natural abundance in GeH4 have
factors affecting the wavenumber scale accuracy led to an been reported.103 IR absorption data were combined with
improved determination of the peak positions. The same SRS data in this work. More recently the SRS spectrum of
year, this work was extended to higher rotational lines, up the n1 band of 70 GeD4 has been recorded and analyzed.104
to J D 17, providing, in this way, the most extensive and An analysis including IR and SRS data was performed in
accurate list of wavenumbers for this band up to today.86 the 1900 cm1 region to derive spectroscopic constants for
In 1997, the previously mentioned RamanRaman dou- the (1,0,0,0) and (0,0,1,0) states of the 116 SnH4 molecule.105
ble resonance technique was used for recording the fully A first SRS spectrum of the n1 region of the SF6 molecule
resolved spectrum of 2n1 n1 of 12 CH4 .87 It is worth men- was published in 1982.106 More exhaustive work on this
tioning that the fraction of population in equilibrium at molecule was published by the same authors, including
room temperature in the v1 D 1 state is only 7 107 . The n1 spectra in a free expansion jet and the n1 C n3 n3
observed spectrum was analyzed by simultaneous fitting of hot band.107 The SRS technique was used to detect a
n1 and 2n1 n1 and, alternatively, with a model based on hole burned in the rotational distribution of the ground
the tetradecad-pentad system. state when a CW CO2 laser was tuned to the n3 band.
Other isotopomers of methane have also been studied by This experiment can be considered a precedent for other
SRS. The spectrum of n1 of 13 CH4 was reported in 1979 double resonance techniques such as RamanIR73 and
and 1980.16,88 In 1991 a more extensive study, including RamanRaman.87 Very recently, the SRS spectrum of
2n2 was reported.89 In this paper, IR and SRS data were this molecule has been recorded in the n2 region.108 A
analyzed simultaneously. simultaneous analysis of this spectrum together with the
The SRS spectrum of 12 CD4 was published for the first IR data for the n2 C n6 band allowed the determination of
time in 1983.90 In this paper, room temperature and free the molecular constants of the v6 D 1 state that is directly
expansion jet (at 45 K) spectra were reported. Two years accessible neither by Raman nor by IR spectroscopy.
later this work was extended to a larger spectral region and
69 lines were recorded and assigned to n1 and n2 C n4 .91 In
1992, self- and Ar-broadening coefficients were measured 3.4 Symmetric top molecules
for some of these lines.92 A first spectrum of the 13 CD4
molecule was published in 1988.93 A region of about Several ammonia isotopomers with 14 N, 15 N, H and D
50 cm1 was covered, including n1 , n3 , n2 C n4 , 2n2 , and have been extensively investigated using high resolution
2n4 . A complete study on the pentad region of this molecule nonlinear Raman techniques, but in all cases CARS, rather
was also published in 1988.94 than SRS was used.109 112
The CD stretching region of the 12 CH3 D molecule was Ethane has also been investigated by SRS. The rotational
also investigated by SRS, and the spectrum analyzed.95 structure in the vibrational spectrum of this simple and light
0.5
C6H6+ + e
74528
J = K =29
0.4
38086
J = K = 30
0.3
v = 0 1B2u
Ionization signal (pcoul / mJ3)

0.2
la 16
1600
0.1 lb 2 + 18
0.0 v = 0 1A1g
74528 0
J = K = 19
1.6 C6H6+ + e
J = K = 17
J = K = 18
1.2
Energy (cm1)
38086
v = 0 1B2u
0.8
0.4 la 16
1600
lb 2 + 18
0 v = 0 1A1g
0
1598.0 1598.2 1598.4 1598.6 1598.8 1599.0
Raman shift /cm1
Figure 7. IDSRS spectra of benzene in the region of overlap between the o O-branch transition of the upper (1609 cm1 ) vibrational
state and the s S-branch transition of the lower (1591 cm1 ) state. Upper field, UV wavelength tuned to 36 467 cm1 in the pP-branch
from the upper state and lower field UV wavelength tuned to 36 496 cm1 in the rR branch from the lower state. [Reproduced by
permission of American Institute of Physics from P. Esherick, A. Owyoung and J. Plva, J. Chem. Phys., 83, 3311 (1985).]
molecule is very complex due to the low torsional barrier of spectra were successfully analyzed. Figure 7 shows the two
the CH3 groups, that manifests itself as many low energy different spectra observed in the same spectral region with
states giving rise to many hot bands and splitting in the this technique. This work was extended by the same authors
rotational levels. The first report of the SRS spectrum of two years later.118
C2 H6 goes back to 1982.15 Esherick et al. recorded the very The IDSRS technique was improved in sensitivity by
complex spectrum of this molecule in the CH stretching using channel plate detectors and in resolution by using
region (n1 and n10 ) at a pressure of 3.5 Torr and in a longer pump laser pulses (45 ns).119,120 A further develop-
free expansion jet. In spite of the simplification in the jet ment of the IDSRS technique was proposed by implement-
spectrum the authors were unable at this time to analyze ing it by Fourier transform techniques.121 In this article, the
this complex structure. This experiment was redone under spectra of several molecular complexes between benzene
similar conditions in 1994.113 From the room temperature and different molecules were reported.
and 30 K spectra an analysis of this strongly perturbed The vibrational energy redistribution/predissociation in
region was successfully carried out this time. In 1992 a the benzene dimer was studied by using the IDSRS
similar study was reported on the CC stretching region of technique.122 Two years later this study was extended
the same molecule.114 to four deuterated isotopomers.123 Several benzeneAr
Spectroscopic constants for n1 , 2n2 , n9 C n7 C n14 and clusters were also detected by the same technique.124
n4 C n5 C n7 have been derived from the jet cooled SRS Mass-selective ionization detection was combined with SRS
spectrum of cyclopropane.115 to study numerous dimers of benzene isotopomers.125 The
Benzene has also been widely investigated by SRS. In same technique was used for detection of dimer, trimer and
1985 the already mentioned IDSRS technique was used to octamer126 and tetramer and pentamer.127 Benzenewater
study the n16 /n2 C n18 Fermi dyad in this molecule.116,117 clusters,128 and naphthalene clusters129 were also observed.
By tuning the UV, ionizing laser alternatively to the upper
and lower component of the dyad, the authors were able
to selectively record the spectrum of one or the other 3.5 Asymmetric top molecules
level that otherwise would appear mixed in the same,
very congested, spectral region. This amazingly elegant Water vapor has also been studied at high resolution by
experiment was conducted in a free expansion jet in order SRS. In a first work, spectra were observed at 7 Torr and
to achieve further rotational simplification and the observed 296 K and at 84 Torr and 404 K.130 Line positions, relative
intensities, frequency shifts and collisional broadening were

measured. The temperature and pressure dependence of the
water vapor spectrum has been studied by SRS up to 1200 K
in order to gain information on these aspects for practical
thermometry applications in combustion and atmospheric
studies.131
The ethylene SRS spectrum was recorded, in both a static
cell and a free expansion jet cooled to 25 K.132
530 531 532 533 534
Diborane has also been investigated by SRS.133,134 The
Wavelength / nm
Q branches of n4 in 10 B2 H6 and 11 B2 H6 were recorded and
analyzed. Figure 8. The pure rotational PARS spectrum of N2 O at a
pressure of 500 Torr. No signal is observed in the region near
zero Raman shift. [Reproduced by permission of Academic
Press from Photoacoustic Raman Spectroscopy of Gases, in
4 PHOTOACOUSTIC SRS Chemical Applications of Nonlinear Raman Spectroscopy, ed.
A.B. Harvey, Academic Press, New York, 89169, (1981).]
An interesting variation of SRS is its, already mentioned,
photoacoustic version (PARS). In this technique the energy to the Rayleigh scattering at 532 nm is apparent. In spite
transferred to the gas is detected, rather than the photon gain of the resolution/sensitivity limitations, this technique was
or loss. This detection possibility is a direct consequence successfully used in the study of the band profile of the
of the fact that molecules result in an excited state after the CO2 Fermi dyad at a pressure as low as 10 Torr.139
stimulated Raman process. Unlike the spontaneous Raman,
in the SRS process a selective excitation is produced in the
medium. In this technique, usually, both pump and probe ACKNOWLEDGEMENTS
lasers are pulsed and the collisional relaxation following
the excitation propagates as an acoustic wave in the gas. I am indebted to Raul Martnez and Jose Luis Domenech
(Also, CW PARS is possible by amplitude modulation for carefully reading the manuscript and for their interesting
of one of the lasers and phase sensitive detection.) A suggestions and to Miguel Angel Moreno for his valuable
simple microphone detects this acoustic wave, whose help in the preparation of the figures. I am also indebted
amplitude as a function of the frequency difference of to Professor J.W. Jones for kindly forwarding me a copy
both lasers provides the Raman spectrum. This variation of his recent review on non-linear Raman spectroscopy and
of the SRS technique was proposed by Barrett et al. in for reviewing this article.
1978.135,136 Barrett also demonstrated in an extensive way
this technique, including numerous examples some years
later.137,138 In Barrett,137 PARS spectra of 45 transitions of ABBREVIATIONS AND ACRONYMS
27 molecules were reported and many spectra reproduced.
A very interesting characteristic of this technique is that, CW Continuous Wave
as the elastic scattering does not leave any energy on the ECS Energy Corrected Sudden
sample, no signal is observed at zero Raman shift. By FPE FabryPerot Etalon
contrast, in spontaneous Raman scattering, the very strong FTS Fourier Transform Spectroscopy
wings of the Rayleigh line and stray light at the excitation IDSRS Ionization Detected Stimulated Raman
frequency, prevent the observation of low frequency lines. Spectroscopy
In SRS and other coherent techniques this drawback is PARS Photoacoustic Stimulated Raman Spectroscopy
not as severe as for spontaneous Raman, but nevertheless PID Proportional Integral Differential
separation of pump and probe lasers becomes difficult when RIKES Raman Induced Kerr Effect Spectroscopy
their frequency becomes very close. This fact makes PARS SRS Stimulated Raman Spectroscopy
an excellent technique for pure rotational spectroscopy. TEM00 Transverse Electric and Magnetic
On the negative side, as the signal scales linearly with
pressure,138 the sensitivity of this technique decreases when
low pressure has to be used in order to avoid pressure REFERENCES
broadening and subsequent degradation of the spectral
resolution. Figure 8 shows the pure rotational spectrum of 1. W.J. Jones and B.P. Stoicheff, Phys. Rev. Lett., 13, 657
N2 O. The absence of any signal at the position equivalent (1964).
2. G. Eckhardt, R.W. Hellwarth, F.J. McClung, S.E. Schwartz, 28. S. Gestenkorn and P. Luc, Atlas du Spectre dAbsorption
D. Weiner and E.J. Woodbury, Phys. Rev. Lett., 9, 455 de la Molecule dIode, CNRS, Paris (1978); Rev. Phys.
(1962). Appl., 14, 791 (1979).
3. G.L. Esley, Coherent Raman Spectroscopy, Pergamon, 29. A. Owyoung, CW Stimulated Raman Spectroscopy, in
Oxford (1981). Chemical Applications of Nonlinear Raman Spectroscopy,
4. M.D. Levenson, Introduction to Nonlinear Laser Spec- ed. A.B. Harvey, Academic Press, New York, 281320
troscopy, Academic Press, New York (1982). (1981).
5. J.W. Nibler and J.J. Yang, Ann. Rev. Phys. Chem., 38, 349 30. B. Lavorel, R. Chaux, R. Saint-Loup and H. Berger, Opt.
(1987). Commun., 62, 25 (1987).
6. J.W. Nibler and G.A. Pubanz, Coherent Raman Spectro- 31. A. Tabyaoui, B. Lavorel, G. Millot, R. Saint-Loup and
scopy of Gases, in Advances in Non-linear Spectroscopy, H. Berger, J. Raman Spectrosc., 21, 809 (1990).
eds R.J.H. Clark and R.E. Hester, John Wiley & Sons, 32. D.R. Beck, M.F. Hineman and J.W. Nibler, J. Chem. Phys.,
Chichester, 150, Vol. 15 (1988). 92, 7068 (1990).
7. J. Santos, Coherent Raman Spectroscopy, in Modern 33. R. Beck and J. Nibler, Chem. Phys. Lett., 159, 79 (1989).
Techniques in Raman Spectroscopy, ed. J.J. Serna, John 34. R.H. Dicke, Phys. Rev., 122, 1218 (1961).
Wiley & Sons, Chichester, 305342 (1996).
35. G. Millot, B. Lavorel, R. Saint-Loup and J. Berger, J.
8. S.A. Druet, J.-P. Taran and Ch.J. Borde, J. Physique, 40, Physique, 46, 1925 (1985).
819 (1979); 41, 183 (1980).
36. B. Lavorel, G. Millot, R. Saint-Loup, C. Wenger, H. Berger,
9. A. Owyoung, Opt. Lett., 2, 91 (1978). J.P. Sala, J. Bonamy and D. Robert, J. Physique, 47, 417
10. D. Heiman, R.W. Hellwarth, M.D. Levenson and G. Martin, (1986).
Phys. Rev. Lett., 36, 189 (1976). 37. G.J. Rosasco, W. Lempert, W.S. Hurst and A. Fein, Chem.
11. G.J. Rosasco and W.S. Hurst, J. Opt. Soc. Am. B, 2, 1485 Phys. Lett., 97, 435 (1983).
(1985). 38. M.L. Gonze, R. Saint-Loup, J. Santos, B. Lavorel, R. Chaux,
12. A. Owyoung and E.D. Jones, Opt. Lett., 1, 152 (1977). H. Berger, L. Bonamy, J. Bonamy and D. Robert, Chem.
13. A. Owyoung, Opt. Commun., 22, 323 (1977). Phys., 148, 417 (1990).
14. A. Owyoung, C.W. Paterson and R.S. McDowell, Chem. 39. B. Lavorel, B. Oksengorn, D. Fabre, R. Saint-Loup and
Phys. Lett., 59, 156 (1978). H. Berger, Mol. Phys., 75, 397 (1992).
15. P. Esherik and A. Owyoung, High Resolution Stimulated 40. L. Bonamy, J. Bonamy, D. Robert, G. Millot and B. Lavorel,
Raman Spectroscopy, in Advances in Infrared and Raman J. Chem. Phys., 101, 7350 (1994).
Spectroscopy, eds R.J.H. Clark and R.E. Hester, John 41. R.A. Hill, P. Esherick and A. Owyoung, J. Mol. Spectrosc.,
Wiley & Sons, Chichester, 130187, Vol. 9 (1982). 100, 119 (1983).
16. A. Owyoung, High Resolution Coherent Raman Spectro- 42. A. Owyoung, R.A. Hill and P. Esherick, Opt. Lett., 8, 425
scopy of Gases, in Laser Spectroscopy IV, eds H. Walther (1983).
and K.W. Rothe, Springer Verlag, Berlin, 175187 (1979). 43. G. Rouille, G. Millot, R. Saint-Loup and H. Berger, J. Mol.
17. K. Bystron, Technische Elektronik, Hanser Verlag, Mun- Spectrosc., 154, 372 (1992).
chen (1974). 44. G. Millot, R. Saint-Loup, J. Santos, R. Chaux and H. Berger,
18. T.W. Hansch, M.D. Levenson and A.L. Schawlow, Phys. J. Chem. Phys., 96, 961 (1992).
Rev. Lett., 26, 946 (1971). 45. G. Millot, C. Roche, R. Saint-Loup, R. Chaux, H. Berger
19. M. Raab and A. Weber, J. Opt. Soc. Am. B, 2, 1476 (1985). and J. Santos, Chem. Phys., 173, 505 (1993).
20. Metrologia, 19, 163 (1984). 46. G. Millot, B. Lavorel and G. Fanjoux, J. Mol. Spectrosc.,
21. P. Cancio and D. Bermejo, J. Opt. Soc. Am. B, 14, 1305 176, 211 (1996).
(1997). 47. G. Millot, G. Fanjoux and B. Lavorel, J. Chem Phys., 104,
22. W. Demtroder, Laser Spectroscopy. Basic Concepts and 5347 (1996).
Instrumentation, Springer-Verlag, Berlin (1981). 48. A. Owyoung and P. Esherick, Opt. Lett., 5, 421 (1980).
23. F. Spieweck, Appl. Phys. B, 29, 99 (1982). 49. D.E. Jennings, L.A. Rahn and A. Owyoung, Astrophys. J.,
24. P. Drell and S. Chu, Opt. Commun., 28, 343 (1979). 291, L15 (1985).
25. J. Santos, P. Cancio, J.L. Domenech, J. Rodrguez and 50. R.L. Farrow, L.A. Rahn and G.O. Sitz, Phys. Rev. Lett., 63,
D. Bermejo, Laser Chem., 12, 53 (1992). 746 (1989).
26. R.L. Farrow and L.A. Rahn, Phys. Rev. Lett., 48, 395 51. L.A. Rahn and G.J. Rosasco, Phys. Rev. A, 41, 3698 (1990).
(1982). 52. M.P. Le Flohic, P. Duggan, P.M. Sinclair, J.R. Drumond
27. G. Millot, B. Lavorel, R. Chaux, R. Saint-Loup, M. Terki- and A.D. May, Can. J. Phys., 72, 186 (1994).
Hassaine, J. Santos, G. Pierre and H. Berger, J. Physique, 53. J. Ph. Berger, R. Saint-Loup, H. Berger, J. Bonamy and
12, 763 (1987). D. Robert, Phys. Rev. A, 49, 3396 (1994).
54. J.W. Forsman, J. Bonamy, D. Robert, J.Ph. Berger, R. Saint- 77. D. Bermejo, J.L. Domenech, G. Di Lonardo and L. Fusina,
Loup and H. Berger, Phys. Rev. A, 52, 2652 (1995). High Resolution Raman Spectra of 12 C2 D2 and 13 C2 D2 :
55. P.M. Sinclair, J.Ph. Berger, X. Michaut, R. Saint-Loup, The n2 Fundamental and Associated Hot Bands, presented
R. Chaux, H. Berger, J. Bonamy and D. Robert, Phys. Rev. at The 16th International Conference on High Resolution
A, 54, 402 (1996). Molecular Spectroscopy, 217 (2000).
56. X. Michaut, R. Saint-Loup, H. Berger, M.L. Dubernet, 78. D. Bermejo, R. Escribano and J. Santos, J. Mol. Spectrosc.,
P. Joubert and J. Bonamy, J. Chem. Phys., 109, 951 186, 144 (1997).
(1998). 79. R. Escribano D. Bermejo, J.L. Calvente, P. Cancio and
57. F. Chaussard, X. Michaut, R. Saint-Loup, H. Berger, A. Maki, J. Mol. Spectrosc., 193, 174 (1999).
P. Joubert, B. Lance, J. Bonamy and D. Robert, J. Chem. 80. K.W. Brown, J.W. Nibler, K. Hedberg and L. Hedberg, J.
Phys., 112, 158 (2000). Phys. Chem., 93, 5679 (1989).
58. F. Chaussard, R. Saint-Loup, H. Berger, P. Joubert, 81. J. Santos and D. Bermejo, High Resolution Stimulated
X. Bruet, J. Bonamy and D. Robert, J. Chem. Phys., 113, Raman Spectrum of the n2 Q branch Region of Diacetylene,
4951 (2000). presented at Sixteenth Colloquium on High Resolution
59. R.Z. Martnez, D. Bermejo, J. Santos and P. Cancio, J. Mol. Molecular Spectroscopy, 393 (1999).
Spectrosc., 168, 343 (1994). 82. D.N. Kozlov, A.M. Prokhorov and V. Smirnov, J. Mol.
60. R.Z. Martnez and D. Bermejo, High-resolution Stimulated Spectrosc., 77, 21 (1979).
Raman Spectrum of the Q-branches of 35 Cl2 , 37 Cl2 and 83. J. Valentn, P. Esherick and A. Owyoung, Chem. Phys. Lett.,
35 37
Cl Cl, presented at 20th Europeam CARS Workshop, 75, 590 (1980).
30 (2001).
84. G. Millot, B. Lavorel and J.I. Steinfeld, J. Chem. Phys., 95,
61. P. Esherick and A. Owyoung, Chem. Phys. Lett., 103, 235
7938 (1991).
(1983).
85. J. Santos, P. Cancio, J.L. Domenech, J. Rodrguez and
62. P. Esherick and A. Owyoung, Ionization Detected Raman
D. Bermejo, Laser Chem., 12, 53 (1992).
Spectroscopy, in Laser Spectroscopy VI, eds H.P. Weber
and W. Luthy, Springer Verlag, Berlin, 66, 68 (1983). 86. D. Bermejo, J. Santos and P. Cancio, J. Mol. Spectrosc.,
63. F. Thibault, J. Boissoles, J.L. Domenech, D. Bermejo, 156, 15 (1992).
C. Boulet and J.M. Hartmann, J. Chem. Phys., to be 87. R.Z. Martnez, D. Bermejo, J. Santos, J.P. Champion and
published. J.C. Hilico, J. Chem. Phys., 107, 4864 (1997).
64. J. Baran, A. Grofcsik and W.J. Jones, Mol. Phys., 45, 1291 88. R.S. McDowell, C.W. Paterson and A. Owyoung, J. Chem.
(1982). Phys., 77, 1071 (1980).
65. R. Roy, D.S. Elliot, D. Meshede, F.M. Pipkin and S.J. 89. J.M. Jouvard, B. Lavorel, J.P. Champion and L. Brown, J.
Smith, Chem. Phys. Lett., 93, 603 (1982). Mol. Spectrosc., 150, 201 (1991).
66. R. Saint-Loup, B. Lavorel, G. Millot, C. Wenger and 90. P. Esherick, A. Owyoung and C.W. Paterson, J. Chem.
H. Berger, J. Raman Spectrosc., 21, 77 (1990). Phys., 87, 602 (1983).
67. B. Lavorel, G. Millot, R. Saint-Loup, H. Berger, L. Bonamy, 91. J.-E. Lolck, A. Owyoung and P. Esherick, J. Raman Spec-
J. Bonamy and D. Robert, J. Chem. Phys., 93, 2176 (1990). trosc., 16, 163 (1985).
68. B. Lavorel, G. Millot, R. Saint-Loup, H. Berger, L. Bonamy, 92. G. Millot, B. Lavorel and J.I. Steinfeld, J. Quant. Spectrosc.
J. Bonamy and D. Robert, J. Chem. Phys., 93, 2185 (1990). Radiat. Transfer, 47, 81 (1992).
69. B. Lavorel, G. Millot, G. Fanjoux and R. Saint-Loup, J. 93. H.W. Schrotter, H. Frunder, H. Berger, J.-P. Boquillon,
Chem. Phys., 101, 174 (1994). B. Lavorel and G. Millot, High Resolution CARS and
70. B. Lavorel, G. Fanjoux, G. Millot, L. Bonamy and F. Emon, Inverse Raman Spectroscopy, in Advances in Non-linear
J. Chem. Phys., 103, 9903 (1995). Spectroscopy, eds R.J.H. Clark and R.E. Hester, John
71. G. Roche, G. Millot, R. Chaux and R. Saint-Loup, J. Chem. Wiley & Sons, Chichester, 97147 (1988).
Phys., 101, 2863 (1994). 94. G. Millot, B. Lavorel, R. Chaux, R. Saint-Loup, G. Pierre,
72. G. Millot and G. Roche, J. Raman Spectrosc., 29, 313 H. Berger, J.I. Steinfeld and B. Foy, J. Mol. Spectrosc., 127,
(1998). 156 (1988).
73. A. Deroussiaux and B. Lavorel, J. Chem. Phys., 111, 1875 95. D. Bermejo, J. Santos, P. Cancio, J.L. Domenech,
(1999). C. Domingo, J.M. Orza, J. Ortigoso and R. Escribano, J.
Raman Spectrosc., 21, 197 (1990).
74. L. Bonamy, J. Bonamy, D. Robert, A. Deroussiaux and
B. Lavorel, J. Quant. Spectrosc. Radiat. Transfer., 57, 341 96. P. Esherik and A. Owyoung, J. Mol. Spectrosc., 86, 250
(1997). (1981).
75. D. Bermejo, P. Cancio, G. Di Lonardo and L. Fusina, J. 97. A. Tabyaoui, B. Lavorel, R. Saint-Loup and M. Rotger,
Chem. Phys., 108, 7224 (1998). J. Raman Spectrosc., 25, 255 (1994).
76. D. Bermejo, R.Z. Martnez, G. Di Lonardo and L. Fusina, 98. A. Owyoung, P. Esherick, A.G. Robiette and R.S. McDowel,
J. Chem. Phys., 111, 519 (1999). J. Mol. Spectrosc., 86, 209 (1981).
99. B. Lavorel, R. Saint-Loup, G. Pierre and H. Berger, J. 119. M. Crew, M. Leuchs and J.W. Nibler, High-resolution
Physique Lett., 45, L295 (1984). Ionization-detected Raman Spectroscopy, in Proceedings
100. B. Lavorel, G. Millot, Q.L. Kou, G. Guelachvili, K. Bou- of the Fifteenth International Conference on Raman Spec-
zoba, P. Lepage, Vl.G. Tyuterev and G. Pierre, J. Mol. troscopy, eds S.A. Asher and P. Stein, John Wiley & Sons,
Spectrosc., 143, 35 (1990). Chichester, 174175 (1996).
101. G. Millot, A. Bouthar, B. Lavorel, C. Wenger, R. Saint- 120. M. Leuchs, M. Crew, J. Harrison, M.F. Hineman and J.W.
Loup and H. Berger, J. Raman Spectrosc., 21, 803 (1990). Nibler, J. Chem. Phys., 105, 4885 (1996).
102. B. Lavorel, R. Pykhov and G. Millot, J. Quant. Spectrosc. 121. G.V. Hartland, B.F. Henson, V.A. Venturo, R.A. Hertz and
Radiat. Transfer., 49, 579 (1993). P.M. Felker, J. Opt. Soc. Am. B, 7, 1950 (1990).
122. B.F. Henson, G.V. Hartland, V.A. Venturo and P.M. Felker,
103. S.Q. Mao, R. Saint-Loup, A. Aboumajd, P. Lepage and
J. Chem. Phys., 91, 2751 (1989).
H. Berger, J. Raman Spectrosc., 13, 257 (1982).
123. B.F. Henson, G.V. Hartland, V.A. Venturo, R.A. Hertz and
104. G. Pierre, V. Boudon, E.B. Mkadmi, H. Burger, D. Bermejo
P.M. Felker, Chem. Phys. Lett., 176, 91 (1991).
and R. Martnez, Study of the Fundamental Bands of
70
GeD4 by Raman and Infrared Spectroscopy, Presented at 124. V.A. Venturo, P.M. Maxton and P.M. Felker, Chem. Phys.
Sixteenth Colloquium on High Resolution Molecular Spec- Lett., 198, 628 (1992).
troscopy, 428 (1999). 125. B.F. Henson, G.V. Hartland, V.A. Venturo and P.M. Felker,
105. A. Tabyaoui, B. Lavorel and G. Pierre, J. Mol. Spectrosc., J. Chem. Phys., 97, 2189 (1992).
148, 100 (1991). 126. B.F. Henson, V.A. Venturo, G.V. Hartland and P.M. Felker,
106. P. Esherick and A. Owyoung, J. Mol. Spectrosc., 92, 162 J. Chem. Phys., 98, 8361 (1993).
(1982). 127. M.W. Schaeffer, P.M. Maxton and P. Felker, Chem. Phys.
107. P. Esherick and A. Owyoung, High Resolution Inverse Lett., 224, 544 (1994).
Raman and Raman Gain Spectroscopy, in Non-linear 128. P.M. Maxton, M.W. Schaeffer and P. Felker, Chem. Phys.
Raman Spectroscopy and its Chemical Applications, Pro- Lett., 241, 603 (1995).
ceedings of the NATO Advanced Study Institute, eds 129. M.W. Schaeffer, W. Kim, P.M. Maxton, J. Romascan and
W. Kiefer and D.A. Long, D. Reidel, Dordrecht, 499517 P. Felker, Chem. Phys. Lett., 242, 603 (1995).
(1982).
130. L.A. Rahn and D.A. Greenhalgh, J. Mol. Spectrosc., 119, 11
108. D. Bermejo, R.Z. Martnez, E. Loubignac, V. Boudon and (1986).
G. Pierre, J. Mol. Spectrosc., 201, 164 (2000).
131. D.A. Greenhalgh and L.A. Rahn, J. Raman Spectrosc., 21,
109. R. Angst, H. Finsterholtz, D. Illig, D. Papousek, P. Pracna, 874 (1990).
K.N. Rao, H.W. Schrotter and S. Urban, J. Mol. Spectrosc.,
110, 454 (1985). 132. L. Abad, D. Bermejo, R. Escribano, V.J. Herrero, J. Santos,
I. Tanarro, G.D. Nivellini and L. Ramonat, Chem. Phys.
110. J. Jonuscheit, U. Lehner, D. Illig, S. Anders, K. Sarka and Lett., 227, 248 (1994).
H.W. Schrotter, J. Mol. Struct., 349, 389 (1995).
133. J.L. Domenech, D. Bermejo, J.-M. Flaud and W. Lafferty,
111. S. Anders, J. Jonuscheit, U. Lehrer, K. Sarka and H.W. J. Mol. Spectrosc., 194, 185 (1999).
Schrotter, J. Raman Spectrosc., 31, 711 (2000).
134. J.-M. Flaud, W.J. Lafferty, H. Burger, G. Pawelke, J.L.
112. M. Stamova, S. Anders, J. Jonuscheit, H.W. Schrotter, Domenech and D. Bermejo, J. Mol. Spectrosc., 203, 339
P. Pracna, S. Urban, S. Klee, M. Winnewieser and K. Sarka, (2000).
J. Mol. Struct., 482483, 481 (1999).
135. J.J. Barrett and M.J. Berry, Photoacoustic Raman Scatter-
113. L. Abad, D. Bermejo, P. Cancio, D. Domingo, V.J. Herrero, ing in Gases, in Proceedings of the Sixth International
J. Santos, I. Tanarro and S. Montero, J. Raman Spectrosc., Conference on Raman Spectroscopy, eds E. Schmid,
25, 589 (1994). R. Krishnan, W. Kiefer and H.W. Schrotter, Heyden,
114. D. Bermejo, J. Santos, P. Cancio, J.M. Fernandez-Sanchez London, 466467, Vol. 1 (1978).
and S. Montero, J. Chem. Phys., 97, 7055 (1992). 136. J.J. Barrett and M.J. Berry, Appl. Phys. Lett., 34, 144 (1979).
115. J. Plva, M. Terki-Hasseine, B. Lavorel, R. Saint-Loup, 137. J.J. Barrett, Photoacoustic Raman Spectroscopy of Gases,
J. Santos, H.W. Schrotter and H. Berger, J. Mol. Spectrosc., in Chemical Applications of Nonlinear Raman Spec-
133, 157 (1989). troscopy, ed. A.B. Harvey, Academic Press, New York,
116. P. Esherick, A. Owyoung and J. Plva, J. Chem. Phys., 83, 89169 (1981).
3311 (1985). 138. J.J. Barrett, Theory of Photoacoustic Raman Spectroscopy
117. P. Esherick and A. Owyoung, Double-resonance Tech- and Applications of Photoacoustic Raman Spectroscopy,
niques for High-resolution Raman Spectroscopy, in Pro- in Non-linear Raman Spectroscopy and its Chemical Appli-
ceedings of the Tenth International Conference on Raman cations, Proceedings of the NATO Advanced Study Insti-
Spectroscopy, eds W.L. Peticolas and B. Hudson, Univer- tute, eds W. Kiefer and D.A. Long, D. Reidel, Dordrecht,
sity Printing Department, University of Oregon, 15-415-5 563602 (1982).
(1986). 139. M. Rotger, B. Lavorel and R. Chaux, J. Raman Spectrosc.,
118. J. Plva, P. Esherick and A. Owyoung, J. Mol. Spectrosc., 23, 303 (1992).
125, 393 (1987). 140. L. Galatry, Phys. Rev., 122, 1218 (1961).
Hyper-Raman Spectroscopy
L.D. Ziegler
Boston University, Boston, MA, USA
1 INTRODUCTION thresholds, HRS cross-sections are typically 105 107 times

weaker than linear RS. However, subsequent improve-
When a high-powered pulsed laser of frequency n0 is ments in the repetition rate of Q-switched and mode-locked
focused in a material, not only is inelastically scattered laser systems and the development of efficient multichan-
radiation displaced from n0 detected [spontaneous Raman nel array detectors have made the observation of this
scattering (RS)], but a much weaker spontaneous emis- relatively weak phenomenon possible with much greater
sion corresponding to scattered radiation displaced from ease since the 1980s.5,7 11 Additionally, the development
twice the incident laser frequency, 2n0 , may also be of HRS techniques such as electronic resonance HRS12
observed. Analogous to the well-known phenomenon of and surface-enhanced hyper-Raman scattering (SEHRS),7
RS, the differences between the observed inelastically scat- which increase the inherent cross-sections of this nonlinear
tered frequencies, ns and 2n0 , correspond to vibrational light-scattering effect, has resulted in greatly improved HRS
(or rotational/rovibrational) energy levels of the ground signal sensitivity and served to promote additional interest
electronic state (see Figure 1). This overall three-photon, in this phenomenon.
higher-order Raman effect is known as hyper-Raman scat- The most commonly exploited feature of HRS is derived
tering (HRS).1 6 It may be observed in both spontaneous from the three-photon nature of this light-scattering pro-
or stimulated spectroscopies. The observed signals due to cess. The corresponding symmetry selection rules permit
spontaneous HRS are proportional to the square of the inci- the observation of vibrational modes in HRS spectra that
dent laser intensity, hence this phenomenon is sometimes may not be seen in other vibrational spectroscopies such as
termed nonlinear RS. infrared (IR) absorption, (linear) RS or inelastic neutron
The inherent cross-sections of this higher-order light- scattering. In particular, for centrosymmetric molecules,
scattering effect are considerably smaller than those asso- strict mutual exclusion rules dictate that gerade and unger-
ciated with the usual, i.e. linear, RS. Consequently, the ade modes are allowed in RS and HRS spectra, respec-
ability to observe this nonlinear spontaneous Raman effect tively. Thus, these two scattering spectroscopies access
has been very much dependent on technological advances complementary vibrational information. A second potential
with respect to both excitation sources and detection sensi- advantage of HRS is that the observed spontaneous scat-
tivity. The advent of high peak powered Q-switched pulse tered frequencies occur in a wavelength range that is well
lasers allowed the first experimental observation of hyper- separated from the incident laser frequency. The weakness
Raman by Terhune et al. in 1965.1 These observed signals, of the hyper-Rayleigh line, particularly for materials with
first observed with a ruby laser operating at 0.03 Hz and inversion centers, allows the observation of low-frequency
detected with a gated photomultiplier tube, corresponded modes without strong quasielastic or scattered light contam-
to about 1013 of the incident energy. When an excita- ination via HRS. This feature has proven to be especially
tion laser source is focused to a flux on the order of valuable in studies of phase transitions in solids, charac-
1010 W cm2 , which is close to typical material breakdown terization of polaritons in solids, glasses and liquids, and
investigations of transverse and longitudinal optic (LO)
John Wiley & Sons Ltd, 2002. phonons in crystals. Third, electronic resonance effects can
m by the sample at frequency !k , [Pn !k ], is described by

0
series expansion in powers of all active radiation fields:
s
n Pn !k D cn !k D !1 C C !n E!1 E!n 1
0
where cn is the electrical susceptibility that describes the
material response at frequency !k due to the interaction
with n electric fields. In this macroscopic susceptibility
f
g approach, both incident and scattered radiation fields are
Figure 1. Energy level diagram corresponding to nonresonant explicitly considered among the n active interacting fields.
g ! f HRS transition. Hence, both spontaneous and stimulated Raman and hyper-
Raman scattering are described by terms that contribute
be used to probe the structure and dynamics of electronic at third order [c3 ] and fifth order [c5 ], respectively,
states in the far-ultraviolet (far-UV) or vacuum-ultraviolet in the expansion of the electric susceptibility.19 When
(VUV) region. Narrow-band tunable radiation in this region the incident laser frequencies are far from any material
of the spectrum is difficult to generate with sufficient inten- electronic resonances, the integrated g ! f HRS transition
sity to observe one-photon resonant RS. Finally, metal differential cross-section (cm4 s erg1 sr1 ) is given by19,20
surface enhancement effects allow HRS to be used as a
ds 24 p2 a3 n4s Pg 2
sensitive probe of surfaceadsorbate interactions. D 2
blmn 2
Several reviews of the experimental and theoretical d Nhc
developments and various applications of HRS have been where ns is the scattered frequency, a is the fine structure
given previously.2 6 This article is intended to summarize constant, Pg is the population of the initial molecular state,
the essential phenomenological characteristics of this light- g, and N is the number of scatterers per unit volume. The
scattering effect, to outline briefly the most salient theo- electronically nonresonant molecular transition hyperpolar-
retical aspects of this higher-order vibrational spectroscopy izability defined with respect to space-fixed polarization
that, in particular, account for its unique properties, and to directions of the incident and scattered radiation fields, blmn ,
summarize recent applications of this technique to problems is given by the following sum of three terms:13 16,19 21
hfj r jmi hmj rm jni hnj rn jgi hfj rn jmi hmj r jni hnj rm jgi hfj rm jmi hmj rn jni hnj r jgi
blmn D l C l C l
3
m,n
n ng n0 nmg 2n 0 nng n0 n mf C n0 nnf C 2n0 nmf C n0
of chemical and physical interest. A few specific HRS appli- The molecular states g, m, n and f are defined by refer-
cations from the diverse areas of surface science, solid-state ence to Figure 1; nng , etc. are the corresponding energy
materials characterization and gas-phase ultrafast chemistry gaps between these levels (cm1 ). Equation (3) has been
are highlighted. derived both by direct polarizability expansions to second
order in the incident radiation fields and by density matrix
diagrammatic techniques. The hyperpolarizability elements,
2 THEORETICAL DESCRIPTION blmn , defined above are formally given for degenerate HRS,
i.e. when the frequencies of the two incident excitation pho-
Treatments of spontaneous RS and HRS were traditionally tons, n0 , are identical. For the more general phenomenon of
derived within the framework of induced dipole responses nondegenerate, i.e. two-color excited, HRS blmn is given by
to the incident radiation electric fields (Em , En ) via the the sum of six terms, which reduce to equation (3) in the
linear polarizability, a, and the first-order hyperpolariz- degenerate limit.16 However, with the exception of a single
ability, b, respectively.2 5,13 16 However, a more gener- experimental study,22 all spontaneous HRS measurements
ally useful approach for the description of spectroscopies, have been performed with identical photons in terms of
particularly for condensed-phase materials, is given by a both energy and polarization (l D m), i.e. degenerate HRS.
susceptibility treatment of the electrical polarization.6,17,18 The molecular transition hyperpolarizability, blmn , can
Such an approach captures both stimulated and sponta- be defined in either Cartesian or vector polar coordi-
neous (Raman) spectroscopies, and provides a natural basis nates or an irreducible tensor framework.23 Correspond-
for considerations of competing nonlinear optical effects. ingly, HRS selection rules can also be developed in either
In this framework, the signal polarization field generated of these representations.2,14,16,23 25 However, HRS (and
Hyper-Raman Spectroscopy 3
RS) is often most conveniently described in terms of and, under conditions of electronic resonance, r is a
an irreducible tensor framework.19,20,23,24 The irreducible sensitive measure of excited vibronic state lifetimes for
HRS tensor elements, b,Q m , are related to polar vector scatterers in the gas phase.27
19,20
components, blmn , by An estimate of the relative intensities of spontaneous
Raman (IRS ) and hyper-Raman (IHRS ) scattering is given
b,Q
m D 2Q C 11/2 2 C 11/2 1mi CQC1Cm in terms of the susceptibility approach by
lmn
miDnCm
IHRS 8pI0 Imc5 4a Ms 2 I0
1 1 Q Q 1 8
blmn 4 IRS c Imc3 hc2 Ee 2
n m mi mi l m
For a nonresonant excitation detuning (Ee D nmg 2n0 )
and of 2 104 cm1 , an electronic transition length (Ms ) of

0
b,Q b,Q D mm0 b,Q
0
,Q 5 109 cm and an incident intensity (I0 ) of 1010 W cm2 ,
m bm 0 5
m,m0 D
the HRS cross-section is of the order of 105 106 that
of linear RS. To put the relative magnitude of this higher-
The polar vector hyperpolarizability components, in turn, order process in perspective, the typical quantum yield of
can be easily transformed back to a Cartesian coordinate spontaneous RS as compared with HRS is equivalent to the
representation. The large parentheses in equation (4) rep- relative quantum yield of fluorescence as compared with
resent the familiar 3j symbols.26 Many of the symmetry (nonresonant) RS.
properties and selection rules governing HRS intensities Detailed discussions of HRS vibrational selection rules
can be derived in terms of the symmetry properties of have been given previously.2 5,14,16,23 25 As already noted,
these angular momentum coupling coefficients. Three scat- owing to the two- and three-photon character of RS and
tering invariants are required in general for the complete HRS, respectively, gerade and ungerade modes are allowed
description of RS. In contrast, the HRS due to linearly in the HRS spectra of centrosymmetric molecules. In the
polarized identical frequency photons requires five scat- context of the irreducible hyperpolarizability tensor ele-
tering invariants.19,20,23 The total intensity of spontaneous ments defined above, we note that all IR-active modes are
HRS in the plane perpendicular to the incident linearly HRS allowed via D 1 hyperpolarizability tensor elements,
polarized electric vector (standard 90 scattering geometry) whereas non-IR-active modes may appear in HRS spectra
in terms of these five invariants is given by19,20 due to D 2 and D 3 irreducible tensor elements.

ds 8pa3 n4s Pg 1 1,2 2 2 1,0 2 1 2,2 2
D b C b C b
d 2
hc N 9 9 15 3 EXPERIMENTAL CONSIDERATIONS

2 3,2 2 29 1,2 1,0
C b C p Re b b 6 The ability to observe this higher-order RS process has
21 5 always been tied to advances in pulsed laser technology.
When HRS emission is excited by elliptically polarized The first reported HRS spectra were excited by a Q-
radiation or by nondegenerate excitation fields, a larger switched ruby operating at a repetition rate of 0.03 Hz.1
number of scattering invariants are required for the com- Subsequent HRS studies have been performed with Q-
plete description of the scattered intensity.23 The HRS switched and or mode-locked Nd : YAG laser systems with
depolarization ratio defined in analogy with that of RS, repetition rates varying from 10 Hz to 82 MHz.7 11 Cu
the intensity of scattered light with its electric vector per- vapor,28 excimer27 and Nd : YAG pumped dye lasers12
pendicular to the incident electric vector as compared with have also been used to excite HRS emission. Advances in
the intensity of scattered radiation with its electric vector laser technology continue to provide improved sources for
parallel to that of the incident excitation, is given in terms observing this weak scattering phenomenon. The ideal HRS
of these scattering invariants by19,20 pulsed laser source is one with a pulse width of 15 ps,
2 2 2 2 p
7 b1,2 C 35 b1,0 C 21 b2,2 C 12 b3,2 14 5 Re b1,2 b1,0
rD 2 2 2 p 7
28 b1,2 C 35 b1,0 C 18 b3,2 C 28 5 Re b1,2 b1,0
Just as found for RS, the polarization characteristics of corresponding to a transform-limited bandwidth of the order
HRS can be used as an indicator of vibrational symmetry24 of typical vibrational line widths. Shorter-duration pulses
result in additional line broadening and radiation dephasing techniques for the study of structural and dynamic
effects under conditions of electronic resonance. Nie et al. information in crystals. HRS has also attracted considerable
demonstrated the effects of pulse duration for a constant attention as a unique and important tool for the study
average laser power on HRS intensities.10 Since sponta- of these materials,5,8,28 56 owing primarily to the HRS
neous HRS is proportional to the square of the incident symmetry selection rules. Ungerade modes are HR active
intensity, I20 (W cm2 ), the observed signal will be propor- in centrosymmetric materials, just as in IR, and in contrast
tional to the product of the pulse energy and the incident to conventional linear Raman spectroscopy. In crystals and
average power for a pulse of a given duration. Dielec- other solid-state materials, it is often low-frequency dipole-
tric breakdown, continuum generation or other competing allowed (ungerade) modes, such as transverse optic (TO),
higher-order processes limit excitation pulse fluences. Thus, LO and mixed materialradiation character oscillations,
once this practical limit has been reached, for a given set of i.e. polaritons, that are of great interest.5 Some of
experimental conditions, i.e. laser excitation geometry and these degrees of freedom, for both centrosymmetric and
focusing, and scattered light collection power, improvement noncentrosymmetric materials, may be silent modes,
of the HRS signal-to-noise ratio (S/N) can only be obtained that is, neither Raman nor IR allowed, but permitted
by the use of higher repetition rate lasers (surface and by the three-photon selection rules that govern HRS.
electronic enhancement effects aside). Pulsed Ti : sapphire Of considerable practical significance, the observation
laser-based systems probably offer the best source of high of low-frequency modes is easier in spontaneous HRS
repetition rate picosecond pulses for HRS, at least when than in RS spectra since the incident excitation radiation
800-nm excitation will suffice.11 Pulse durations of 15 ps is far removed from the scattered frequencies and, in
at 80 MHz and up to 2 W average power are available and centrosymmetric crystals, hyper-Rayleigh scattering is
amplification techniques may result in even greater pulse formally forbidden aside from the effects of defects and
energies at these high repetition rates. quadrupole polarization contributions, further minimizing
Furthermore, just as the quality (S/N and collection time) the interference from broad baseline effects in the low-
of spontaneous linear RS has been dramatically improved frequency regime. Furthermore, since HR excitation is
with the use of charge-coupled device (CCD) array detec- nonresonant at the incident laser frequency, HRS studies
tors, such detectors currently provide similar improvements probe bulk solid-state properties unambiguously, in contrast
for the observation of the inherently weaker nonlinear to IR reflectance measurements or even resonance Raman
spontaneous HRS emission.8 This improvement in HRS studies, which may be affected by crystal surface effects.
signal collection can be especially exploited for excita- Crystals are inherently concentrated samples, which further
tion geometries that match the spectrometer slit height, thus contributes to the high quality of the reported HRS
maximizing the observed scattering volume, and hence take spectra. In addition, the orientational dependence of these
full advantage of the two-dimensional nature of these sen- ordered materials allows the polarization dependence of
sitive detectors.5 Unlike linear RS, such an experimental the observed spontaneous HRS to be an important label
geometry is always possible even under conditions of HRS of vibrational symmetry.
electronic resonance because the incident laser beam is not Denisov et al.5 have given an excellent and extensive
appreciably absorbed (see below). review of many applications of HRS for the study of struc-
It should also be recognized that cascaded linear RS and
ture and dynamics in condensed-phase materials. Drawing
second harmonic generation (SHG) [or sum frequency gen-
on the various advantages that this nonlinear scattering
eration (SFG)] processes, in either order, i.e. RS/SFG or
technique offers, HRS has been used in crystal studies
SHG/RS, can mimic the true HRS signal in noncentrosym-
primarily (1) to learn about the nature and dynamics of
metric materials.5 These cascaded lower-order processes
structural phase transitions via the temperature dependence
have the same incident intensity dependence as HRS. Usually,
of low-frequency soft phonon modes and silent modes,
a comparison of relative vibrational (or rotational) scattering
(2) to describe the electronicnuclear coupling giving rise
intensities can reveal whether the observed inelastic spectrum
to HRS activity and thus probe the electrostatic structure
is due to the cascaded linear or nonlinear Raman transitions.
of crystalline materials, (3) to provide information on the
local microscopic regions of symmetry breaking due to lat-
4 APPLICATIONS OF HYPER-RAMAN tice distortions and (4) to study the wavevector dependence
SCATTERING of vibrational polaritons.
An example of the use of HRS in studies of crystal
4.1 Hyper-Raman scattering of crystals structure and dynamics and the complementary nature of
IR absorption and reflection, Raman and inelastic information derived both by linear and nonlinear sponta-
neutron scattering spectroscopy are widely used vibrational neous Raman measurements is provided by an HRS study
of crystalline quartz.52 Quartz undergoes a phase transition band is attributed to SHG owing to the reduction in local
(D6 hexagonal b, high temperature ! D3 trigonal a, low symmetry caused by local strains and defects in the crys-
temperature) at 573 C. RS measurements reveal an A1 tal. The temperature dependence of the frequency, damping
soft phonon mode that indicates that the phase transi- and intensity of the B1 mode derived from the HRS analy-
tion in the low-temperature regime is well described by sis identifies direct evidence that the a b phase transition
a displacive-type phase transition. However, the nature is of the simple displacive type in the b phase, in contrast
of the phase transition as it was approached from the to an orderdisorder type.
high-temperature side, i.e. in the b phase, had been unclear Another useful application of HRS in crystals is the
following the results of electron and neutron diffraction study of wavevector-dependent vibrational features known
and molecular dynamics calculations. The soft mode asso- as polaritons.5,22,29 31 Polaritons are produced by the inter-
ciated with the a b phase transition in the b phase is action of dipole-allowed TO modes and the accompany-
predicted to be of B1 symmetry, which is a silent mode ing E-field resulting from the dipole oscillation. Thus,
in D6 , but allowed in HRS. Furthermore, SHG is also for- in contrast to RS, polaritons can be observed in crys-
mally symmetry forbidden in D6 . Thus, observations of tals of any symmetry group. Furthermore, small scattering
the temperature dependence of the low-frequency part of angles (<10 ) are required to observe these mixed mechan-
the spontaneous HRS spectrum could provide information icalelectromagnetic features. Owing to the widely sepa-
about the nature of the phase transition at high temper- rated incident and scattered wavelengths and the formal
atures. A 1-kHz YAG laser excited the HRS spectra of forbiddenness of SHG in centrosymmetric crystals in par-
b-phase quartz shown in Figure 2. As can be seen, a strong ticular, HRS inherently allows the observation of these fea-
central feature at low temperatures separates into two peaks tures over a much larger angular range, i.e. at much smaller
at higher temperatures and a weak residual band remains at angles, as compared with corresponding RS observations. In
0 cm1 . The observed low-frequency band shapes are well particular, Denisov and co-workers5,22,29 31 have exploited
fit by a damped harmonic oscillator model over this tem- such observations in HRS for the characterization of these
perature range, except for the weak residual central band. bands in crystals, enabling one to define refractive indices
The line-shape analysis of the temperature dependence of in the IR region and to study their interactions with other
vibrational levels.
this low-frequency feature identifies its origin as the B1
silent soft mode of b-quartz instead of a relaxation mode
resulting from structural fluctuations. The residual central 4.2 Hyper-Raman scattering of glasses,
semiconductors and liquids
590 C
I (w) w2
-Quartz While most of the HRS applications have been found in

z (xxx )y studies of crystalline solids, HRS in other solid materi-

als, such as amorphous glasses and fibers, has also been
30 0 30 observed and provides an additional example of the unique
capabilities of HRS for some samples.57,58 In particular,
1 590 C HRS measurements permit studies of the orientation and
magnitude of the momentum transferred to a vibrational
Hyper-Raman intensity
system, for dipole modes in these isotropic centrosymmet-

(arbitrary units)
4 670 C
ric media. In scattering experiments, one can vary k, the
momentum transferred to the vibrational system, by an
10 740 C appropriate choice of the scattering geometry. Thus, these
observations are not possible in IR reflectance or linear RS
investigations. The HRS spectra of amorphous fused quartz
20 770 C
are shown in Figure 3(a) obtained in the forward scatter-
ing geometry sketched at the top (A D sample, D D sector
50 0 50 diaphragm, L D lens, P D polarizer and S D spectrometer
slit).57 In the HRS of a-SiO2 , without the diaphragm, both
Frequency shift / cm1
LO and polaritons, associated with the TO modes, are
Figure 2. Hyper-Raman spectra of b-quartz as a function of tem- observed. The TO/polariton description of these scattering
perature. Solid circles are observed data. Solid curves are fits to
a damped harmonic oscillator. Residual central peak is attributed features is identified by the scattering angle dependence of
to SHG resulting from defects or impurities. (Reproduced by per- the HRS intensities. As the sector diaphragm is rotated 90 ,
mission from Tezuka et al.52 ) the only difference in the HRS set-up is the angle of the
D L P order in CCl4 , PCl3 and C2 Cl4 via angle-dependent forward

S
A scattering5,65,67,69 are all examples of HRS applications in
liquids that directly result from the unique properties of this
higher-order Raman technique.
LO LO 4.3 Surface-enhanced hyper-Raman scattering
In the 1970s, it was observed that RS cross-sections could

x be enhanced by as much as 106 for molecules adsorbed to
Polariton
roughened metals surfaces, metal colloids and electrodes.71
z
This surface-enhanced Raman scattering (SERS) is largely
(a) attributed to two effects: chemical effects, derived from
Intensity
x ,k metal charge-transfer state resonance, and electromagnetic
k 10z effects, which result from the field enhancement due to
z (b) surface plasmon excitation. Since the inherent HRS cross-
section is so small, it became attractive to investigate the
x
relative enhancement of this nonlinear scattering process
k || 0z
due to these surface effects and, furthermore, to exploit
z,k the complementary information derived from HRS in prob-
(c) ing the nature of the moleculesurface interaction.72 87 In
400 800 1200 contrast to SERS, it has since been established that the
Wavenumber / cm1 SEHRS enhancement factor is typically of the order of
1012 1014 .11,75,83,86 In one study87 an enhancement of 1020
Figure 3. yzzzy scattering geometry and HRS spectra of fused
quartz: (a) without diaphragm; (b) and (c) with diaphragm D; was estimated for single-molecule HRS on colloidal silver
q D 5 . (Reproduced by permission from Podobedov.57 ) and has been attributed to hot spots in the metal cluster
structures. Owing to the larger relative surface enhancement
of HRS, the observed SEHRS spectra are only about a fac-
observed scattering relative to the fixed polarization director of 200 weaker than corresponding SERS spectra.11,83,86
tions of the excitation and the HRS emission. The LO and The first example of SEHRS was reported in 1982 for
polariton/TO modes appear in mutually orthogonal orien- SO3 2 adsorbed on Ag powder.7 In most of the subse-
tations as shown in Figure 3(b) and (c). Hence these HRS quent SEHRS studies, spectra of p-electronic molecules
results reveal that the orientation of the transferred momen- on colloidal silver clusters have been reported. However,
tum is perpendicularly oriented for each of these types of SEHRS due to copper and gold colloids79 and silver-coated
vibrational motions in the glass. Thus polaritons, in addition nanoelectrodes85 have also been observed.
to mechanical vibrations (LO and TO) of glasses, are char- SEHRS studies have been largely concerned with con-
acterized by well-defined wavevectors. trasting the mode enhancement patterns of SERS and
As a final example of the application of HRS of solid- SEHRS spectra for a given adsorbatesurface system and
state materials, the nonlinear RS of semiconductors, most trying to understand the nature of this large enhance-
notably CdS, CuCl and CuBr quantum dots embedded in ment mechanism.72 87 One interesting question that a com-
glass matrices, have been observed.59 64 These spectra have parison of the SERS and SEHRS of a centrosymmetric
been obtained under conditions of two-photon electronic molecule can address is whether the surface significantly
resonance enhancement (see below) and the resulting ana- perturbs the inherent molecular symmetry. If the perturb-
lysis of the HRS spectra reveals the strength of excitonLO ing influence of the surface is minor, then the same mutual
phonon coupling strength as a function of confinement size. exclusion of RS and HRS bands found in the bulk should
These studies will be discussed further in the context of prevail in the surface-enhanced spectra. The SERS and
resonance HRS. SEHRS spectra of pyrazine, a centrosymmetric molecule,
The HRS studies of neat liquids have also been obtained adsorbed on a silver electrode (0.2 V), are shown in
in order to address a number of structural and dynamic Figure 4 and are noticeably different.86 These SERS and
questions in these materials.9,22,65 70 For example, the SEHRS spectra are dominated by gerade and ungerade
frequencies of silent modes in benzene,9 timescales and modes, respectively, indicating that the bulk selection tend
mechanisms of rotational diffusion in a large range of to dominate for these surface-enhanced scattering spec-
organic liquids and water,23,68 and studies of short-range tra. However, several of these modes appear, at least
much more closely resemble the IR absorption spectra.

One conclusion drawn from these studies is that new
vibrational information can clearly be drawn from these
three-photon SEHRS spectra that may be unobtainable via
231
3500 cps
lower-order vibrational spectroscopic methods for some
1594
1021
636 adsorbatesubstrate systems. It has been suggested that
1245
SEHRS is more sensitive to the details of the surface inter-
1225
743
437
action than SERS.86
1620
1516
Finally, although there is general agreement that SEHRS
(a) spectra are enhanced by 1214 orders of magnitude relative
Intensity
to bulk HRS cross-sections, the nature of the mechanism

of this dramatic surface enhancement effect is less clear.
It has been argued that the extra approximately six orders
1593
200 cps of magnitude enhancement of SEHRS as compared with
SERS can be attributed to an electromagnetic enhancement
1624
436
1516
1021
1228
effect derived from resonance with the surface plasmon

1162
absorption.11,83 In the context of the EM enhancement

226
1044
796
model, HRS is relatively more strongly affected because

636
742
it depends nonlinearly on the enhanced excitation fields as

363
compared with RS. Excitation profiles of SERS and SEHRS

(b) appear to support this explanation. However, whatever the
nature of this enhancement factor, there is no doubt that
200 400 600 800 1000 1200 1400 1600 1800
the surface enhancement permits the observation of this
Wavenumber / cm1 inherently weak nonlinear scattering when excited by rel-
Figure 4. (a) SERS spectrum and (b) SEHRS spectrum of pyra- atively modest laser power, bringing the SEHRS signals
zine adsorbed on silver electrode at 0.2 V in a solution of 0.1 M to be typically only two orders of magnitude weaker than
pyrazine and 0.1 M KCl. (Reproduced by permission from Li linear SERS spectra. Thus, SEHRS should continue to be
et al.86 )
a useful probe of surface interactions in a wide variety of
adsorbatemetal substrate systems under the appropriate
weakly, in both spectra. The Raman-active modes are
conditions of laser power, excitation wavelength and mate-
particularly evident in the SEHRS spectra. For the HRS
rials system.
of a larger molecule with an inversion center, trans-1,2-
bis(4-pyridyl)ethylene, on a silver film-coated nanosphere
electrode,85 a careful normal-mode analysis reveals that 4.4 Electronic resonance hyper-Raman
nearly all the modes obey the mutual exclusion rule anti- scattering
cipated for a C2h system. Hence for this system the surface
is even less of a perturber of the adsorbate local sym- 4.4.1 General description
metry under these electrochemical conditions than in the Inspection of the HRS cross-section expression, equa-
pyrazine study. Thus the symmetry perturbation is system tion (3), reveals that the intensity of this nonlinear scattering
dependent. process can be enhanced when the excitation wavelength is
The SERS and SEHRS spectra of noncentrosymmetric coincident with regions of one-photon or two-photon elec-
molecules also exhibit considerably different intensity pat- tronic absorption, i.e. when n0 D nng or 2n0 D nmg . HRS
terns owing to the different sources of scattering activity signals are more effectively maximized by the two-photon
in the two experiments, i.e. polarizability and hyperpolar- resonance, 2n0 D nmg , because the incident radiation is
izability derivatives, respectively. However, the similarity attenuated in the case of the one-photon resonance. HRS
between these two observations for such lower-symmetry observed under these excitation wavelength conditions is
adsorbates has shown considerable variability. For exam- known as resonance hyper-Raman scattering (RHRS) and
ple, the RS and SEHRS spectra of pyridine on silver was first considered by Long and Stanton.15 This excita-
are strikingly similar,75,86 whereas the corresponding spec- tion wavelength-dependent phenomenon is entirely analo-
tra of carbocyanine dyes on colloidal silver are dramati- gous to the well-known resonance Raman effect. The first
cally different.80 The SEHRS of these dyes contains lines unequivocal RHRS spectra were reported by Ziegler and
entirely absent from the RS spectra, either surface enhanced co-workers, who observed the vibrationally and rotation-
or in the bulk, and also the IR spectra, although they ally resolved RHRS of several small molecules, NH3 , ND3 ,
CH3 I, CS2 and (CH3 )2 CO, in the gas phase.12,19,20,88 91 by19,90

Other subsequent examples of electronically enhanced HRS
hfj ri jmi hmj rj jni hnj rk jgi
include the RHRS of crystals (SrTiO3 , TiO2 ),47,48,51,92 bijk D 9
semiconductor quantum dot structures (CdS, CuBr and n
nng n0 nmg 2n0 img /2
CuCl)59 64 and molecular liquids (fluoro-, chloro- and
iodobenzene).93 95 where mg is the dephasing constant associated with the mg
In analogy with resonance Raman scattering (RRS), electronic coherence. Similar expressions have been given
RHRS can provide detailed information about the struc- in theoretical treatments of the RHRS of semiconductors.
ture and dynamics particularly of the resonant electronic When pure dephasing effects, for example, due to solvent
state. The principal advantages of the RHRS technique fluctuations, are on a timescale comparable to the resonant
as compared with nonresonant RHRS and linear RRS are excited electronic state lifetime, both narrow hyper-Raman-
as follows: (1) HRS of lower concentration samples, i.e. like emission and broad (two-photon) excited resonance
solutions not just neat materials, may be observed; (2) the fluorescence will be observed in analogy with the observed
structure and dynamics of electronic states lying in the VUV spontaneous resonance secondary radiation due to one-
region may be accessed via excitation profile techniques photon resonance.96
utilizing readily available tunable near-UV excitation wave- The typical enhancement of the HRS cross-section due
lengths; tunable, narrow-band VUV radiation of sufficient to electronic resonance can be estimated from equation (5)
intensity to excite RRS spectra is difficult to generate; to be of the order of nmg /mg 2 104 106 . Linear RS
and (3) the electronic and nuclear character and dynam- is enhanced by the same factor owing to (one-photon)
ics of two-photon-allowed electronic configurations may be electronically resonant excitation. Since HRS cross-sections
selectively probed as compared with one-photon-allowed are approximately five to six orders of magnitude weaker
states. than RS, electronic resonance increases HRS cross-sections,
The analysis of RHRS spectra is accomplished by the on a per molecule basis, to be about as large as non-
same techniques and methods as used to extract molecular resonant linear RS. Experimental observations confirm this
information from RRS. In the absence of pure dephasing expectation.12,88 Furthermore, owing to the larger num-
effects, due to either material or radiation contributions, ber density, scattering volume and incident intensity of
the molecular transition hyperpolarizability tensor elements the RHRS experiment as compared with the linear RRS
due to two-photon resonance with state m is simply given spectra, observed RHRS and RRS spectra can display
1.0
3 s2 s2 + 1
4 s2 3 s2 + 1
0.8
4 s2 + 1
Relative intensity
0.6 (a)
0.4
0.2
(b)
2000 3000 4000 5000 6000 7000 8000

Wavenumber / cm1
Figure 5. (a) Resonance hyper-Raman spectrum of NH3 excited at 417.6 nm and (b) resonance Raman spectrum of NH3 excited at
208.8 nm. (Reproduced by permission from Ziegler and Roebber.12 )
comparable S/N quality for equal data collection times where Mk 0ge and Mak ge are the nuclear coordinate inde-
despite the higher order nature [c5 vs c3 ] of the pendent (allowed) and the Qa nuclear coordinate linearly
RHRS spectrum.12,88 91 This is demonstrated in Figure 5 dependent (forbidden or vibronic) parts of the g ! e
for RHRS and RRS of NH3 vapor excited at 417.6 and electronic transition moment, respectively. The sum in
208.8 nm, respectively.12 These wavelengths are one- and equation (10) is over all the vibronic levels of the res-
two-photon resonant with the A-state transition in the UV. onant electronic state e and all the nonresonant virtual
Finally we note that care must be taken in order to dis- vibronic levels, sr. The angular brackets correspond to
tinguish between two-photon resonant HRS and two-photon the resulting FranckCondon factors or vibrational over-
fluorescence (TPF). Line widths can be used in favorable lap factors that determine the pattern of vibrational RHRS
cases in order to establish unequivocally the nature of these intensities. These A, B and C terms are grouped accord-
distinct types of resonant secondary emission. The HRS ing to the number of allowed transition moments and
features will be characterized by a width associated with hence their order corresponds to the relative strength of
dephasing on the ground-state surface alone. The TPF-like the corresponding vibrational HRS activity. The ratio of
feature is characterized by a spectral width correspond- an allowed transition moment to a forbidden or vibronic
ing to electronic dephasing timescales. Excitation frequency moment, Mk 0ge /Mak ge Qa , is typically of the order
dependence and polarization can also serve to identify the of 10, thus IRHR (A term) : IRHR (B term) : IRHR (C term) may
nature of these spontaneous emission bands.12,96 be estimated as 1 : 102 : 104 . In descriptions of RHRS in
crystals and semiconductors, the electronicnuclear cou-
4.4.2 Vibronic treatment of resonance hyper-Raman pling required for the appearance of vibrational HRS is
scattering correspondingly ususally derived from electronphonon
coupling of the Frohlich type.47,63,97,98
The role of one- and two-photon-allowed dipole character Many essential spectroscopic properties of vibrational
and molecular symmetry is exposed by a vibronic treatment RHRS, and HRS in general, are captured by a vibronic
of RHRS intensities.21,88 Furthermore, such a treatment analysis of this scattering process [equation (10)]. The
allows the interpretation of vibrational band intensities main conclusions of this approach88 are summarized briefly
in terms of diagonal (FranckCondon effects) and off- below:
diagonal (HerzbergTeller coupling or BornOppenheimer
breakdown) nuclear coupling mechanisms. When the nuc- 1. RHRS can be observed from molecules of all sym-
lear coordinate dependence of the electronic transitions metries. In centrosymmetric molecules, the activity
of a vibronic transition moment is required in order
moments (Ml ) resulting from a zeroth-order BornOppen-
to observe RHRS spectra (B term). Consequently, the
heimer basis description of molecular states is made explicit
RHRS (and also nonresonant HRS) of noncentrosym-
by a series expansion around the ground electronic state
metric molecules should be stronger by approximately
nuclear equilibrium configuration, the leading terms des-
a factor of 100, [Mk 0ge /Mak ge Qa ]2 .
cribing vibrational RHRS are the A, B and C terms defined
2. Only noncentrosymmetric molecules are A term RHRS
below:88
active because a given electronic configuration for-
bijk D A C B C C mally may be both one- and two-photon allowed. An
A term active vibrational RHRS spectrum (excitation
AD Mi 0ge Mj 0es Mk 0sg hfjnihnjrihrj0iFn0
frequency n0 ) will be identical with its correspond-
sr,n
ing (excitation frequency 2n0 ) linear RRS spectrum.
BD Mai 0ge Mj 0es Mk 0sg hfjQa jnihnjrihrj0i The RHRS of methyl iodide88 and monosubstituted
sr,n chlorobenzene and iodobenzene93 are examples of
C Mi 0ge Maj 0es Mk 0sg hfjnihnjQa jrihrj0i A term active RHRS.

3. The leading term describing the RHRS activity of
CMi 0ge Mj 0es Mak 0sg hfjnihnjrihrjQa j0i Fn0 centrosymmetric molecules is the B term. For this term
10 to be important the excitation must be resonant in a

CD a 0 b 0 0
Mi ge Mj es Mk sg hfjQa jnihnjQb jrihrj0i two-photon sense with either a one- or two-photon
sr,n allowed electronic transition. Thus, in contrast to linear
RRS, both one- and two-photon allowed electronic
C Mai 0ge Mj 0es Mbk 0sg

states may be equally effective in promoting RHRS
hfjQa jnihnjrihrjQb j0i Fn0 in centrosymmetric scatterers. A B term active RHRS
1 spectrum will generally exhibit a progression in totally
Fn0 D ngo,sr n0 ngo,ev n0 igo,ev /2 11 symmetric FranckCondon active modes built on one
quantum of the vibronically active ungerade mode structure of these semiconductor nanocrystals, allow the
responsible for breaking down the formal one- or two- observation of different TO and LO phonons than seen in
photon dipole-forbidden character. The RHRS of NH3 , linear RRS and study the effect of electron confinement
CS88
2 and C6 H5 F
95
are examples of B term RHRS on electronphonon interaction strength. Compared with
spectra. one-photon resonant experiments, RHRS studies are more
4. Electronic states that are neither one- nor two-photon generally free from the effects of photodarkening, photo-
allowed can still be sources of RHRS activity but with bleaching, persistent hole burning and other photodegrada-
greatly reduced resonant cross-sections via the C term. tion effects in these materials because the incident radiation
The RHRS spectrum of benzene due to (two-photon) falls in a regime of optical transparency.
resonance with the 1 La (1 B1u ) absorption region is an
example of this weak source of HRS activity.95 4.4.3 Rotational resonance hyper-Raman scattering
In crystals such as TiO2 and SrTiO3 , RHRS has been The RHRS spectra of gas-phase ammonia exhibit both
observed as a function of excitation wavelength and the rotationally and vibrationally resolved features, as seen in
expected large HRS intensity enhancement due to two- Figures 5 and 6. As for any three-photon dipole transition,
photon resonance with absorption edges of band gaps are rotational scattering transitions corresponding to J D 1,
seen.47,48,51,92 The experimental RHRS excitation profiles 2 and 3 are allowed, in general, in HRS.19,20,100 Addi-
are analyzed in terms of the excitonic electronic struc- tional J selection rules are dependent on the symmetry
ture of these materials and Frohlich interaction mecha- and vibronic coupling mechanism involving the two-photon
nisms. Vibrational RHRS of quantum dot structures have resonant vibronic level. An -rank tensor component, b,Qm ,
been observed and the description of intensity enhancement contributes to J D 0, . . . , rotational (rovibrational)
due to electronic resonance has been explicitly theoreti- transitions.19,20,90 Thus, the HRS of a Q-branch transition
cally treated for these semiconductor materials.59 64,97 99 (J D 0) is derived from rank D 1, 2 and 3 scattering
These RHRS observations are used to clarify the electronic invariants. For symmetric-top molecules, the additional K
1.0
0.8
1 + 2s
Relative intensity
0.6
6 5 4 3 0 1 2 3 4
N T
8 7 6 5 4 3 2 0 1 2 3 4 5 6 7
0.4 O S
10 8 6 4 2 02 4 6 8
P R
0.2
(a) (b)
3900 4000 4100 4200 4300 4400 4500 4600 3900 4000 4100 4200 4300 4400 4500 4600 4700
Wavenumber /cm1
Figure 6. (a) Observed rotationally resolved resonance hyper-Raman spectrum of NH3 vapor (2 atm) in the n1 C ns2 band region
excited with 417.8-nm radiation. This wavelength is two-photon resonant with the n02 D 2 vibronic absorption band of the X ! A
electronic transition. Rotational transition assignments are indicated. (b) Calculated RHRS rotational band structure of the corresponding
combination band. The spectrum is calculated for an excited-state lifetime of 85 fs. (Reproduced by permission from Ziegler et al.19 )
RHRS rotational selection rules are determined by the direc- described here. However, the even wider use of this tech-
tions of the molecular fixed-frame transition moments in nique is still hampered by the inherent weakness of this
vector polar coordinates. For example, the RHRS of ammo- higher-order Raman process. Prospects for the more general
nia vapor excited at 417.8 nm, which is resonant with a use of HRS continue to be dependent on the development of
vibronic band of the A state (A002 symmetry in D3h ) exhibits additional high-repetition-rate pulsed laser sources. Appli-
only K D 0 and P, Q, R transitions (an IR-like pattern) cations in the future, aside from the areas where HRS
in the RHRS spectrum (Figure 6).19 This pattern is essen- has already shown itself to be a valuable spectroscopic
tially determined by the symmetry of the electronic state, tool, may include the development of six-wave mixing,
the vibrational mode and the primary route of vibronic cou- coherent HRS experiments with either femtosecond (impul-
pling strength at the two-photon excitation frequency. sive excitation) or picosecond pulses. Such observations
Excitation profiles of these RHRS rovibrational bands are analogous to the well-known four-wave mixing opti-
have been used to determine subpicosecond rovibronic cal heterodyne detected, birefringence measurements that
explicit lifetimes of levels in the resonant electronic state of are proportional to Raman resonances. It remains to be
ammonia.19,20 NH3 ! NH2 C H photodissociation accou- shown whether such stimulated HRS techniques can be
nts for the short lifetimes seen here. The RHRS determined exploited in order to overcome the inherent weakness of
subpicosecond quantum specific lifetimes essentially map this higher-order scattering process. Theoretical treatments
out the shape of the multidimensional, quasibound excited- of this coherent HRS effect have appeared.101,102 An addi-
state surface. In addition, depolarization ratios of RHRS tional intriguing experimental approach that may further
vibrational bands of gas-phase scatterers are also sensitive enhance the observation of this weak spontaneous nonlinear
functions of the resonant electronic state lifetime and have scattering is the use of entangled photonsas an excitation
been used to measure photodissociation rates in high-lying source.103,104 If excitation with such an entangled photon
(VUV) levels of methyl iodide vapor corresponding to pair can drive the intensity dependence towards the linear
subpicosecond lifetimes.90,91 This polarization effect may regime, the corresponding scattering cross-section should
be said to arise from the finite rotation time of the molecular approach that of linear RS, at least for scatterers of lower
nuclear frame over the time period that the resonant HRS symmetry.
amplitude develops, which is just the excited-state lifetime
for these short-lived gas-phase scatterers. These RHRS
polarization measurements reveal photodissociation rates in ABBREVIATIONS AND ACRONYMS
CH3 I that are mode specific for this ultrafast process.
HRS Hyper-Raman Scattering
RHRS Resonance Hyper-Raman Scattering
5 CONCLUSION RRS Resonance Raman Scattering
RS Raman Scattering
Spontaneous HRS is the incoherent nonlinear scattering SEHRS Surface-enhanced Hyper-Raman Scattering
process corresponding to the appearance of an inelastically SHG Second Harmonic Generation
scattered photon of frequency ns which is displaced from TPF Two-photon Fluorescence
twice the incident photon frequency (2n0 ) such that 2n0 ns
is equal to a vibrational (rovibrational) quantum on the
ground electronic state. This technique has been established REFERENCES
as a spectroscopic tool that is complementary to both IR
absorption and normal linear RS primarily owing to its 1. R.W. Terhune, P.D. Maker and C.M. Savage, Phys. Rev.
three-photon character. The ability to observe this higher Lett., 14, 681 (1965).
order nonlinear [c5 ] scattering process has been dependent 2. D.A. Long, Raman Spectroscopy, McGraw-Hill, London
on advances in pulsed laser technology and detectors. (1977).
Experimentalists have also taken advantage of electronic 3. M.J. French, in Chemical Applications of Raman Spec-
resonance and surface effects in order to enhance the troscopy, Academic Press, New York (1981).
inherently weak cross-sections of this higher-order Raman 4. D.A. Long, in Non-linear Raman Spectroscopy and its
effect. Chemical Applications, eds W. Kieffer and D.A. Long,
Problems that can be addressed by exploiting the unique Reidel, Dordercht, 99130, 165182 (1982).
properties of this phenomenon, such as symmetry selec- 5. V.N. Denisov, B.N. Mavrin and V.B. Podobedov, Phys.
tion rules and the separation between incident and scattered Rep., 151, 1 (1987).
frequencies, are ideally suited to HRS investigations as 6. L.D. Ziegler, J. Raman Spectrosc., 21, 769 (1990).
7. D.V. Murphy, K.U. Von Raben and R.K. Chang, Chem. 31. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and J.F. Scott,
Phys. Lett., 85, 43 (1982). J. Raman Spectrosc., 14, 276 (1983).
8. S. Shin, M. Ishigame, K. Deguchi and E. Nakamura, Solid 32. K. Inoue and S. Akimoto, Solid State Commun., 46, 441
State Commun., 65, 749 (1988). (1983).
9. W.P. Acker, D.H. Leach and R.K. Chang, Chem. Phys. 33. K. Inoue, Ferroelectrics, 52, 253 (1983).
Lett., 155, 491 (1989). 34. J.B. Grun, B. Honelage and R. Levy, Solid State Commun.,
10. S.M. Nie, L.A. Lipscomb, S. Feng and N.T. Yu, Chem. 46, 51 (1983).
Phys. Lett., 167, 35 (1990). 35. H. Vogt and H. Uwe, Phys. Rev. B, 29, 1030 (1984).
11. K. Kneipp, H. Kneipp, F. Siefert, Chem. Phys. Lett., 212, 36. G. Kugel, H. Vogt, W. Kress and D. Rytz, Phys. Rev. B, 30,
374 (1993). 985 (1984).
12. L.D. Ziegler and J.L. Roebber, Chem. Phys. Lett., 136, 377 37. H. Vogt, Jpn. Appl. Phys., 24(Suppl.), 112 (1985).
(1987). 38. Y. Morioka and I. Nakagawa, Chem. Phys. Lett., 122, 150
13. L.Y. Yuan, Acta Phys. Sin., 20, 164 (1964). (1985).
14. S.J. Cyvin, J.E. Rauch and J.C. Decius, J. Chem. Phys., 43, 39. M.J. French, R.J.B. Hall, D.A. Long and F.M.F. Al-Rubaiey,
4083 (1965). J. Raman Spectrosc., 16, 398 (1985).
15. D.A. Long and L. Stanton, Proc. R. Soc. London, Ser. A, 40. H. Vogt and H. Presting, Phys. Rev. B, 31, 6731 (1985).
318, 441 (1970). 41. H. Vogt, M.D. Fontana, G.E. Kugel and P. Gunter, Phys.
16. J.H. Christie and D.J. Lockwood, J. Chem. Phys., 54, 1141 Rev. B, 34, 410 (1986).
(1971). 42. H. Vogt, Phys. Rev. B, 36, 5001 (1987).
17. D. Lee and A.C. Albrecht, A Unified View of Raman, 43. S. Shin, A. Ishida, T. Yamakami, T. Fujimura and M. Ishi-
Resonance Raman and Fluorescence Spectroscopy (and their game, Phys. Rev. B, 35, 4455 (1987).
Analogues in Two-photon Absorption), in Advances in 44. Y. Morioka and I. Nakagawa, J. Raman Spectrosc., 18, 533
Infrared and Raman Spectroscopy, eds R.J.H. Clark and (1987).
R.E. Hester, Wiley-Heyden, Chichester, 179213, Vol. 12
(1985). 45. K. Inoue, A. Hasegawa, K. Wanatabe, H. Yamaguchi,
H. Uwe and T. Sakuda, Phys. Rev. B, 38, 6352 (1988).
18. S. Mukamel, Nonlinear Spectroscopy, Oxford University
46. Y. Morioka, M. Wada and A. Sawada, J. Phys. Soc. Jpn.,
Press, Oxford (1995).
57, 3198 (1988).
19. L.D. Ziegler, Y.C. Chung and Y.P. Zhang, J. Chem. Phys.,
47. K. Inoue and K. Watanabe, Phys. Rev. B, 39, 1977 (1989).
87, 4498 (1987).
48. K. Watanabe and K. Inoue, J. Phys. Soc. Jpn., 58, 726
20. L.D. Ziegler, Y.C. Chung, P. Wang and Y.P. Zhang, Res- (1989).
onance Rotational Raman and Hyper-Raman Scattering: A
Probe of Sub-picosecond Dynamics, in Advances in Spec- 49. H. Vogt, Phys. Rev. B, 41, 1184 (1990).
troscopy, Time Resolved Spectroscopy, eds R.J.H. Clark 50. H. Vogt, J. Raman Spectrosc., 21, 813 (1990).
and R.E. Hester, Wiley, Chichester, 55112, Vol. 18 (1989). 51. K. Watanabe and K. Inoue, Phys. Rev. B, 41, 7957 (1990).
21. V.I. Petrov, Opt. Spectrosc., 59, 284 (1985). 52. Y. Tezuka, S. Shin and M. Ishigame, Phys. Rev. Lett., 66,
22. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and Kh.E. 2356 (1991).
Sterin, Opt. Commun., 60, 95 (1986). 53. A. Yamanaka, K. Inoue, T. Furutani and J. Kawamata,
23. P.D. Maker, Phy. Rev. A, 1, 923 (1970). Ferroelectronics, 159, 79 (1994).
24. D.L. Andrew and T. Thirunamachandran, J. Chem. Phys., 54. H. Vogt, J. Phys.: Condens. Matter, 7, 5913 (1995).
68, 2941 (1978). 55. H. Vogt, Phys. Rev. B, 51, 8049 (1995).
25. S. Kielich and T. Bancewicz, J. Raman Spectrosc., 21, 791 56. A. Yamanaka and K. Inoue, Physica B, 219, 620 (1996).
(1990). 57. V.B. Podobedov, J. Raman Spectrosc., 27, 731 (1996).
26. A.R. Edmonds, Angular Momentum in Quantum Mechan- 58. B.G. Varshal, V.N. Denisov, B.N. Mavrin, V.B. Podobedov
ics, Princeton University, Princeton, NJ (1960). and Kh.E. Sterin, Fiz. Chim. Stekla, 8, 115 (1982).
27. D.J. Campbell and L.D. Ziegler, J. Chem. Phys., 150, 98 59. A.V. Baranov, K. Inoue, K. Toba, A. Yamanaka, V.I. Petrov
(1993). and A.V. Fedorov, Phys. Rev. B, 53, R1721 (1996).
28. A.M. Agaltsov, V.S. Gorelick and M.M. Sushchinskii, Opt. 60. K. Inoue, A. Yamanaka, K. Toba, A.V. Baranov, A.A.
Spectrosc., 58, 230 (1985). Onushchenko and A.V. Fedorov, Phys. Rev. B, 54, R8321
29. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and Kh.E. (1996).
Sterin, Zh. Eksp. Teor. Fiz. Pisma Red., 31, 111 (1980); 61. A.V. Fedorov, A.V. Baranov and K. Inoue, Phys. Rev. B,
Fiz. Tverd. Tela, 22, 2829 (1980). 54, 8627 (1996).
30. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and Kh.E. 62. K. Inoue, A.V. Baranov and A. Yamanaka, Physica B, 219,
Sterin, Solid State Commun., 40, 793 (1981). 508 (1996).
63. A.V. Baranov, S. Yamauchi and Y. Masumoto, Phys. Rev. 83. K. Kneipp, H. Kneipp and F. Siefert, Chem. Phys. Lett., 233,
B, 56, 10 332 (1997). 519 (1995).
64. S. Savasta and R. Girlanda, Phys. Rev. B, 59, 15 409 84. S. Schneider, G. Brehm and P. Freunscht, Phys. Status Solidi
(1999). B, 189, 37 (1995).
65. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and Kh.E. 85. W.H. Yang, J. Hulten, G.C. Schatz and R.P. Van Duyne, J.
Sterin, Opt. Commun., 44, 39 (1982). Chem. Phys., 104, 4313 (1996).
66. T.J. Dines, M.J. French, R.J.B. Hall and D.A. Long, J. 86. W.-H. Li, X.-Y. Li and N.-T. Yu, Chem. Phys. Lett., 305,
Raman Spectrosc., 14, 225 (1983). 303 (1999).
67. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and Kh.E.
87. K. Kneipp, H. Kneipp, I. Itzkan, R.R. Dasari and M.S. Field,
Sterin, Sov. Phys. JETP, 57, 733 (1983). Chem. Phys., 247, 155 (1999).
68. V.N. Denisov, B.N. Mavrin and V.B. Podobedov, Opt.
88. Y.C. Chung and L.D. Ziegler, J. Chem. Phys., 88, 7287
Spectrosc., 57, 584 (1984).
(1988).
69. V.N. Denisov, B.N. Mavrin, V.B. Podobedov and Kh.E.
Sterin, J. Raman Spectrosc., 16, 71 (1985). 89. Y.C. Chung and L.D. Ziegler, J. Chem. Phys., 89, 4692
(1988).
70. C. Xu, J.B. Shear and W.W. Webb, Anal. Chem., 69, 1285
(1997). 90. D.J. Campbell and L.D. Ziegler, J. Chem. Phys., 98, 150
(1993).
71. M. Fleishmann, P.J. Hendra and A.J. McQuillan, Chem.
Phys. Lett., 26, 1631 (1974); D.L. Jeanmaire and R.P. Van 91. D.J. Campbell and L.D. Ziegler, Chem. Phys. Lett., 201, 159
Duyne, J. Electroanal. Chem., 84, 1 (1977); R.P. Van Duyne, (1993).
Laser Excitation of Raman Scattering from Adsorbed 92. K. Watanabe, K. Inoue and F. Minami, Phys. Rev. B, 46,
Molecules on Electrode Surfaces, in Chemical and Bio- 2024 (1992).
chemical Applications of Lasers, ed. C.B. Moore, Aca-
demic Press, New York, 101, Vol. 4 (1978). 93. C.C. Bonang and S.M. Cameron, Chem. Phys. Lett., 187,
619 (1991).
72. A.V. Baranov and Ya.S. Bobovich, JETP Lett., 36, 339
(1982). 94. C.C. Bonang and S.M. Cameron, Opt. Commun., 86, 504
73. A.V. Baranov, Ya. Bobovich, V.N. Denisov, V.I. Petrov and (1991).
V.B. Podobedov, Opt. Spectrosc., 56, 355 (1984). 95. C.C. Bonang and S.M. Cameron, Chem. Phys. Lett., 192,
74. A.V. Baranov, Ya.S. Bobovich and V.I. Petrov, Opt. Spec- 303 (1992).
trosc., 61, 490 (1986). 96. L.D. Ziegler, Acc. Chem. Res., 27, 1 (1994).
75. J.T. Golab, J.R. Sprague, K.T. Carron, G.C. Schatz and R.P. 97. A. Garcia-Cristobal, A. Cantarero, C. Trallero-Giner and
Van Duyne, J. Chem. Phys., 88, 7942 (1988). M. Cardona, Phys. Rev. B, 58, 10 443 (1998).
76. R.P. Van Duyne, J.C. Hulteen and D.A. Treichel, J. Chem. 98. E. Menedez-Proupin, C. Trallero-Giner and A. Garcia-
Phys., 99, 2101 (1993). Cristobal, Phys. Rev. B, 60, 5513 (1999).
77. C.K. Johnson and S.A. Soper, J. Phys. Chem., 93, 281
99. A. Garcia-Cristobal, A. Cantarero, C. Trallero-Giner and
(1989).
M. Cardona, Physica B, 263264, 809 (1999).
78. A.V. Baranov, Ya.S. Bobovich and V.I. Petrov, Sov. Phys.
Usp., 33, 812 (1990). 100. K. Altmann and G.J. Strey, J. Raman Spectrsoc., 12, 1
(1982).
79. L.A. Lipscomb, S.M. Nie, S. Feng and N.T. Yu, Chem.
Phys. Lett., 170, 457 (1990). 101. M. Cho, J. Chem. Phys., 106, 7550 (1997).
80. N.Y. Yu, S. Nie and L.A. Lipscomb, J. Raman Spectrosc., 102. M. Yang, J. Kim, Y. Jung and M. Cho, J. Chem. Phys., 108,
21, 797 (1990). 4013 (1998).
81. S.M. Nie, L.A. Lipscomb and N.T. Yu, Appl. Spectrosc., 26, 103. M. Atature, A.V. Sergienko, B.E.A. Saleh and M.C. Teich,
203 (1991). Phys. Rev. Lett., 84, 618 (2000).
82. I. Persaud and W.E.L. Grossman, J. Raman Spectrosc., 24 104. B.E.A. Saleh, B.M. Jost, H. Fei and M.C. Teich, Phys. Rev.
(1993). Lett., 80, 3483 (1998).
Fast Time-resolved Mid-infrared Spectroscopy Using
an Interferometer
Gregory D. Smith and Richard A. Palmer
Duke University, Durham, NC, USA
1 INTRODUCTION maximum time resolution, sensitivity, spectral range, cost,

and ease of use.
Although several laser-based approaches to TRIR, both
The development of fast time-resolved mid-infrared (MIR) single-frequency and continuum-generation methods, offer
spectroscopy since the 1980s has provided an effective tool extremely high time resolution, Fourier transform interfero-
for the study of transient chemical species in dynamic metric time-resolved infrared (FT-TRIR) techniques have
systems.1,2 Interest in this particular spectroscopy stems distinct advantages over both these methods in terms of
from the virtual universality of infrared (IR) absorption, the spectral bandwidth, resolution, and sensitivity. The use of
rich molecular structural information contained in the MIR an interferometer to encode the spectral characteristics also
spectrum, and the fast time scale of vibrational motions brings a wide array of added benefits to TRIR experi-
[compared, for example, with the nuclear magnetic reso- ments. Fourier transform infrared (FT-IR) spectroscopys
nance (NMR) time scale]. The incorporation of an interfero- well-known multiplex (Fellgett), throughput (Jacquinot),
metric spectrometer in these time-resolved studies allows and registration (Connes) advantages, which have led to the
multiple vibrational frequencies to be monitored simulta- almost total replacement of dispersive instruments for static
neously across a wide spectral window. With this tool, measurements, also persist in time-resolved spectroscopy
the processes of ligand binding and dissociation, protein (TRS) applications. The rise time of practical broadband
folding, photocycles, reaction kinetics, and excited-state MIR detectors is currently limited to 1 ns, and therefore,
electronic structures, as well as a host of other dynamic the fastest FT-TRIR technique as currently demonstrated
chemical phenomena, can all be measured and character- has its upper limit in that time regime.3 However, the
ized. Over the years, efforts in this area have centered on pulsed-asynchronous sampling technique (see below and
increasing the speed at which broadband, high-resolution Volume 1, Part 2) promises to extend that limit by several
spectra can be collected.1,3 5 orders of magnitude, thus pressing laser-based techniques
Time-resolved MIR spectroscopic studies of transient even further in their most advantageous area, namely time
species have been conducted over a temporal range stretch- resolution.
ing from minutes to femtoseconds. Figure 1 shows that each Despite the present temporal limitation of FT-IR, the
time regime along that scale has come to be dominated advantages provided by interferometric spectral encoding
by a specific time-resolved infrared (TRIR) technique, that make it superior to dispersive and laser-based methods
a considerable overlap exists at the boundaries, and that for producing fast time-resolved, high spectral resolution
interferometric techniques dominate a large portion of the measurements in which the resulting spectra cover the
dynamic spectrum. The superiority of each technique in a entire MIR region. The chief drawbacks of the latter tech-
niques, and consequently the areas where Fourier transform
particular time regime can be ascribed to that techniques
(FT)-based methods excel, include the limited tunability of
probe lasers (both in range and resolution), the complexity
John Wiley & Sons Ltd, 2002. of multicomponent laser systems, and the expense of the
2 Time-resolved Spectroscopy
lin g
ous
in g
Change in intensity (I )
p
sam nchron
sam
n
in g
scan
g
o p ic
a nn
p lin
sy
id
id sc
bosc
ed-a
arap
Rap
Pu ls
Ultr
Stro
103 100 103 106 109 1012 10-15

Time regime (s)
91 92 93 94 95 96 97 98 99 100
81 82 83 84 85 86 87 88 89 90
Step-scanning Noninterferometric 71 72 73 74 75 76 77 78 79 80
laser-based methods 61 62 63 64 65 66 67 68 69 70
51 52 53 54 55 56 57 58 59 60 )
Figure 1. Time scale of FT-TRIR experiments showing applic- (t 2
41 42 43 44 45 46 47 48 49 50
ta
able range of selected techniques in comparison with laser-based da
31 32 33 34 35 36 37 38 39 40
al
methods. 21 22 23 24 25 26 27 28 29 30 por
11 12 13 14 15 16 17 18 19 20 T em
equipment. For many TRIR experiments, the modest cost 1 2 3 4 5 6 7 8 9 10
and standard availability of time-resolving features on com- Retardation ([t1])

mercial FT-IR spectrometers make their implementation in
time-resolved studies throughout the second to nanosecond Figure 2. Generation of time-resolved interferograms by use of
time scale relatively routine. rapid-scanning FT-IR spectroscopy. Numbered circles below the
interferograms show the order of data sampling. When using
This article deals with each of the FT-TRIR techniques rapid-scanning FT-IR spectroscopy, data collection proceeds in
indicated in Figure 1. The sections cover the techniques rows along the retardation axis.
theory of operation, instrumental requirements, specific
caveats, advantages and limitations, current applications, Volume 1, Part 2). In addition to its use for static spec-
and the state of the art. Although every attempt is made to troscopy, it also represents the simplest and most conve-
keep the discussion of each technique general, occasionally nient FT-TRIR technique.11 In rapid-scanning FT-TRIR,
it is necessary to mention specific instrumentation, and in the moving mirrors velocity is chosen so that on the time
these cases, the authors have chosen to comment on instru- scale of the dynamic event, each scan is effectively instan-
mentation with which they are personally familiar. These taneous. Figure 2 shows the data-collection scheme for gen-
instances should not be taken as an indication of the supe- erating a stack of time-resolved interferograms using rapid-
riority of a particular instrument or piece of equipment, but scanning interferometry. For clarity, the interferograms are
can be considered at least one proven option for collecting shown as the change in the intensity (I) during the course
TRIR data. Specific references are intended to direct the of a time-dependent perturbation, and the numbered circles
reader to more thorough coverage of certain topics in the underneath the stack plot display the order in which data
literature and to applications of each FT-TRIR technique are collected by the spectrometers analog-to-digital con-
to real chemical problems. Although FT-IR instruments verter (ADC). Successive scans of the interferometer collect
are also used for dynamic measurements in the frequency entire interferograms ([t1 ]) instantaneously with respect to
domain (modulation/demodulation method), these applica- the temporal evolution of the dynamic event (t2 ) so that
tions will not be covered in this article, and readers are t1 t2 . The first interferogram (points 110) corresponds
referred to other parts of this work (see Volume 1, Part 2; to the steady state of the sample just before being excited,
Volume 2, Part 7; and Volume 4, Parts 1 and 2). and in the second interferogram (points 1120), the state
just after excitation.
Typical mirror velocities for standard commercial instru-
2 RAPID AND ULTRARAPID SCANNING ments range from a few hundredths to 6 cm s1 , providing
FOURIER TRANSFORM INFRARED optical velocities of twice those values. The better, more
expensive models support a wide range of variability, often
The operation of the classic Michelson interferometer for extending even beyond these two limits, whereas cheaper
continuous, or constant-velocity, scanning has been dealt instruments offer only a limited number of scan speeds,
with fully in the literature6 10 and in this volume (see usually tending toward the middle of the scale where the
Fast Time-resolved Mid-infrared Spectroscopy Using an Interferometer 3
mirror is more easily controlled. Thus, for the generation inertial force occurring at the point of the mirrors rever-
of a 4 cm1 spectrum, the spectrometer requires between sal would jolt the spectrometer enough to disturb the other
1 s and 40 ms, depending on the scan velocity selected, for optical components and contribute instability and a large
the completion of a full pass of the interferometer mirror. amount of mechanical noise to the measurement.
As a general rule, the duration of a scan should be at In spite of the technological difficulties associated with
least an order of magnitude shorter than the half-life of the scanning an interferometer at cycles greater than 25 Hz,
dynamic event being examined. If this criterion is not met the need for an ultrarapid scanning interferometer has
or exceeded, then the intensity of the IR signal will change spurred new efforts in instrument design to increase the
appreciably over the course of the interferometer scan. The rate of acquiring spectra. Manning and co-workers have
result of such a situation is that transient spectral bands will identified several areas in which ultrarapid scanning FT-IR
contribute heavily to the interferogram points collected at spectroscopy would make a significant impact.13 Three of
the beginning of the scan, but will diminish in intensity by the areas ripe for the advent of an ultrarapid interferometer
the end of the scan. Such an unsymmetrical waveform will are reviewed here.
transform into a spectrum laden with noise. First, many nonrepeatable and repeatable chemical and
Therefore, assuming that a maximum scan frequency of physical processes occur on the sub-100 ms time scale and
25 Hz (40 ms per scan) is available, a dynamic event with cannot, therefore, be studied using classical rapid-scann-
a half-life of the order of 500 ms can easily be observed ing FT-IR spectroscopy. Combustion reactions, impulse
and characterized by multiple time-resolved spectra using spectroelectrochemical experiments, and inelastic polymer
rapid-scanning FT-IR spectroscopy and a mercury cadmium deformations are examples of such processes. Remote sens-
telluride (MCT) detector. For much more slowly evolving ing from rapidly moving platforms, e.g. satellites and air-
systems, a lower scan speed and a conventional pyroelectric planes, is another example of a quickly changing observable
triglycine sulfate (TGS) detector would suffice. However, that is often impossible to repeat, and therefore requires
with the previously mentioned faster scan rate and high- ultrarapid data collection.
frequency detector, one could benefit from coadding a num- Second, in photoacoustic depth profiling, where the ther-
ber of mirror scans for each time-resolved spectrum in order mal diffusion depth varies as the square root of the fre-
to increase the signal-to-noise ratio (S/N) while realizing quency of modulation of the incident radiation, current
the advantages of the MCT detectors greater sensitivity. modulation frequencies achieved via step-scan FT-IR (see
Beyond the aforementioned time scale of several hundred below) or rapid-scanning FT-IR limit the probe depth to val-
milliseconds, the duty cycle of the conventional rapid ues greater than about 1 m, far too deep for many thin-film
(constant-velocity) scanning interferometer is too low to studies. A kilohertz scanning interferometer could achieve
collect an interferogram at a rate significantly faster than Fourier frequencies between 100 kHz and 1 MHz. At these
the completion of the dynamic event itself. The reason modulation frequencies, one would be capable of probing
for this limitation is the manner in which the moving layers as thin as 100 nm, assuming that the available detec-
mirror is translated along the variable-pathlength arm of tors (microphones) are sensitive enough to overcome the
the interferometer. This linear motion is often achieved in inescapable 1/f photoacoustic signal reduction that accom-
commercial instruments through the use of a voice coil panies increasing source modulation frequency.14
linear drive motor acting on a mirror that is floating on As a final argument for ultrarapid scanning, the recent
an air bearing, or air cushion, of some sort. surge of progress in FT-IR imaging could receive another
The impedance to faster scanning of conventional mir- boost through the advent of such an interferometer. The
rors, say to 1 kHz, originates in the force necessary to state of the art uses a step-scanning FT-IR spectrometer
quickly move the mirror and reverse its direction after com- that allows all of the pixels of an MCT focal plane array
pletion of a scan. It has been shown that the force necessary (FPA) detector to be read out sequentially at each mir-
to drive a 30-g mirror using conventional voice coil mirror ror position.15 This recording scheme can be inefficient
translation in order to measure a 4 cm1 resolution spec- since the light delivered to the unsampled array elements
trum in 1 ms (now defined as ultrarapid scanning) is equal is wasted if the readout is slow. With numerous fast ADCs
to 375 N.12 This value is well above the typical 12 N force operating in parallel, an improvement not too far on the
generally achieved in commercial instrument voice coils. horizon, one could sample each pixel of the FPA simul-
Although it might be possible to approach the hundreds of taneously while scanning an ultrarapid interferometer at
newtons required for ultrarapid scanning by some modifica- 1000 scans per second. Such an instrument would have the
tion of the linear drive motor, the mirror and the bearings added benefit of maximizing the specific detectivity (D )
required for smooth translation would likely not respond of the MCT detector elements, which increases by approx-
well to such rigorous physical conditions. Furthermore, the imately one order of magnitude as the modulation of the
incident light is increased from direct current (DC) (as in Reference laser
the step-scan case) to 1 kHz. D remains relatively con- signal from
HeNe detected Dual-
stant from 1 kHz to 1 MHz, which is the typical range of by Si channel
Fourier frequencies for an ultrarapid scanning interferom- photodiode MCT 14-bit,
eter operating in the MIR region.13 Using this detection 5-MHz
Collimated ADC
scheme, an IR camera with over 16 384 units of image res- source
olution (128 128 array) and an image capture speed of
1000 frames s1 would be possible. Also, such a camera
might possibly outperform current instruments in sensitiv- Beamsplitter
ity owing to the previously stated advantage of raising the Rotating wedge
mirror
Fourier frequencies of the IR light to the range of maximum
detectivity for the detector. From this example and those
previously mentioned, the need for an ultrarapid scanning
interferometer can be justified. Cube corner Personal computer
Various interferometer modifications have been sug-
gested for increasing scanning speeds by switching from Stationary
translational motion of the mirror to rotational movement. mirror
A recent report has outlined the history of these instru- Figure 3. Schematic of an ultrarapid scanning IR spectrometer
ments in which an optical path difference is generated via designed by Manning and co-workers.
a rotary motion.12 Many of the initial designs incorporate
a tilt-table interferometer in which only a small rotation will generate 1000 spectra per second, making Manning
angle is tolerated by the optical layout, thus leading to a and co-workers design the first demonstration of an ultra-
rockingreciprocating motion. This scheme for encoding rapid scanning interferometer as previously defined. The
the spectral data suffers from the same mechanical limita- effects of air resistance suggest that vacuum operation will
tions as do linear drive motors. More recent designs have be required for full implementation of this design.
utilized a full rotation optical element to generate optical The detection scheme for the ultrarapid scanning instru-
interference. Although this improvement involves a comment described above includes a dual-channel 14-bit 5-MHz
plete revolution, and therefore is capable of being spun with ADC to record the two detector signals. The HeNe laser
high angular velocity without the negative side effects of reference intensity detected using a silicon photodiode and
a reciprocating movement, the fact that the rotating optical the IR interferogram detected by a photoconductive MCT
element is the planar salt plate beamsplitter has meant that are digitized simultaneously at the constant sampling fre-
the maximum scan speeds outside an evacuated bench are quency of the ADC. Since the zero crossings of the laser
limited owing to air resistance. Furthermore, deformations interferogram are not used to space (in terms of retarda-
of the rapidly rotating plate are intolerable, and the design tion) the MCT readouts, the laser and IR interferograms
still suffers from inefficiency since only a portion of the must both be interpolated to reconstruct the values of the
rotational motion allows for optical interference. IR interferogram at the calculated zero crossings of the
A design introduced by Manning and co-workers in 1999 laser signal. This data processing produces IR interfero-
is shown in Figure 3.13 This design includes a rotating, gram points at equal intervals of optical path difference,
highly polished aluminum disk as the rotating optical ele- a prerequisite for fast Fourier transform (FFT) mathemat-
ment. The disk serves as the moving mirror and is cut ics, from what are otherwise detector data points at equal
so that the reflecting surface is angled with respect to temporal spacing.
the rotation axis, producing a maximum path difference of The ultrarapid scanning interferometer described above
0.25 cm over the course of half a revolution. This retarda- has been successfully demonstrated in time resolving the
tion distance yields a fixed spectral resolution of 4 cm1 . A spectral changes associated with the spark combustion of
cube corner retroreflector ensures that the nonplanarity of three fuel sources in air: a blackpowder rocket igniter,
the incident radiations wavefront, after leaving the wedge ethane, and methane. In each instance, the ignition was
wheel, is tilt compensated by being redirected along the begun slightly after data collection was initiated, and the
same return pathway. Further rotation of the mirror wheel dual-channel ADC was then allowed to sample at its
retraces the 0.25 cm retardation path, thus producing two normal frequency until the on-board memory was filled.
interferograms per revolution, similar to bidirectional scan- The resulting interferograms were collected at 4 cm1
ning in a reciprocating interferometer. The element can be resolution throughout the entire MIR region at rates of
spun at a maximum of 500 Hz. This mode of operation 150166 scans s1 . At the fastest rate of data collection,
the time-resolved spectra are separated by 6 ms, a factor of 1 2 3 4 etc.

seven times faster than that claimed for any conventional
rapid-scanning instrument. The spectra reveal both excited
emission of the fuel material and the evolution of H2 O and
CO2 from its combustion. The nominal S/N in each instance
is 15 : 1. These demonstrations, however, do not show the
Change in intensity (I)

maximum time resolution achievable (1 ms, and perhaps Data
(a) point 1
as low as 0.5 ms in the near future), nor the theoretical S/N 80 s
70 s
60 s
50 s
determined to be possible with this type of instrument. 1 2 3 4 etc. 40 s
30 s
20 s
10 s
0 s
10 s
3 STROBOSCOPIC INTERFEROMETRY
The next faster time regime in FT-TRIR spectroscopy is 9

10
Data
accessible by the stroboscopic, also called interleaved, FT- point 2 7
8
)
(b) 6
(t 1
IR technique. We do not say it dominates this regime 5 ta
4 da
because this particular sampling method has only become 1 2 3 4 etc. 3
o ral
mp
2
a standard in a few laboratories where the dynamic events 1
Te
of interest are confined to a time scale from hundreds of Retardation ([t ])
1
milliseconds to several tens of microseconds. In the event
that a wider range of time resolution is required, most
spectroscopists prefer to invest in the less complicated, Data
(c) point 3
albeit more expensive, mode of step-scan operation (see
below). However, the inclusion of step-scan FT-TRIR in Figure 4. Generation of time-resolved interferograms by use of
the spectroscopists arsenal does not preclude the use of the stroboscopic sampling. The data collection scheme is explained
in the text. Numbered circles below the interferograms show
stroboscopic technique on the same instrument, and in fact
the order of data sampling. When using stroboscopic sampling,
many commercial FT-IR research-grade software and hard- data collection proceeds diagonally between the retardation and
ware packages include an interleaved data collection routine temporal axes.
in addition to the step-scan feature as standard components.
Unlike rapid and ultrarapid scanning FT-IR, the strobo- indicate where data points fall chronologically (shown in
scopic method of data collection does not record an entire the numbered circles) as well as spatially along the retar-
interferogram in a single pass of the interferometer mirror. dation axis as they are read off the IR detector. It should
Rather, the data collected on a single scan of the mov- be noted that the time scale of the retardation dependence
ing mirror span a time-resolved set of interferograms that and the time resolution of the measurement are of the same
are gradually built up with successive passes of the inter- magnitude.
ferometer mirror.16,17 The reconstruction of the individual The experiment begins with the interferometer mirror
time-separated interferograms requires the use of a clever positioned at ZPD (Figure 4a). The detector signal at this
data sorting routine. Figure 4 shows a schematic view of point corresponds to a preflash interferogram point that will
how the results of an impulse perturbation are time resolved be used to calculate the background spectrum, shown in
using the stroboscopic method. the stack plot at t1 D 10 s. The computer controlling
The schematic depicts how data collection would be the FT-IR spectrometer is programmed to trigger the short,
conducted for a flash-initiated reaction that can be repro- initiating flash 10 s into the experiment. This perturbation
ducibly repeated many times. The cosinusoid represents the occurs at the second laser fringe (Figure 4b), and the data
interferogram of the HeNe reference laser as the moving point collected here belongs to the interferogram when the
mirror shown below it translates along the interferometer dynamic event is first initiated; therefore, the I interfero-
arm. For this explanation, the laser fringes, or zero cross- gram has its maximum value. By the third laser fringe,
ings, are numbered sequentially from 1, the point of zero the transient has had 10 s to decay when the data point is
path difference (ZPD). In the example used here, the mir- taken (Figure 4c). This data point will contribute to the third
rors constant velocity is assumed to be such that each laser interferogram in the time course. The detector sampling
fringe occurs 10 s after the previous one, corresponding to continues at a rate of one datum every 10 s, with each
a scan speed of 3.16 cm s1 for the translating mirror. The recorded point belonging to the next interferogram in the
stack plots of interferograms, again shown as I for clarity, series.
After the appropriate time decay of the system has samples. However, in the stroboscopic sampling method,
been recorded in the first pass, the interferometer mirror the exact position of the interferogram sampling is also con-
is returned to its initial position, and the scan is repeated. volved with the exact time of the transient samples decay.
However, in order to collect the second point in each of the Therefore, variations in scanning velocity of the moving
interferograms along the time-resolved series, the initiation mirror will translate into improperly spaced detector read-
flash is delayed by one laser reference fringe with respect to ings in time, thus imparting an error into the temporal
the previous scan, and data collection begins at the HeNe information recorded at that point in the interferogram.
fringe just before excitation in order to collect another Often this inaccuracy is found to be a random Gaussian
background data point. In the schematic, this corresponds to distribution of ADC samples around the expected value,
data collection at the second fringe and sample perturbation thus leading to noise in the transformed spectrum. Many
on the third fringe. The strobed excitation and sampling commercial instruments exhibit a 12% variation in mirror
continue until enough data have been collected to produce velocity around the actual value selected in the software.
an interferogram for each time delay. Weidner and Peale have developed a software package
Several caveats of stroboscopic sampling must be further (and modest, optional hardware modification) to be added
elaborated on at this point. First, the timing of the pertur- to commercial instruments in order to decouple the ADC
bation event must be strictly controlled. Most commercial sampling from the temporal HeNe fringes normally used
FT-IR instruments with this sampling feature provide for to trigger data readout.20,21 Their event-locked method
a TTL pulse to trigger external equipment responsible for samples the interferogram at evenly spaced times following
initiating the dynamic event. On subsequent interferometer the generation of the transient species without concern for
scans, the software for interleaved scanning incorporates the absolute spacing of the samples along the interferometer
the moving trigger delay. path. The exact, yet uneven, path-length differences of the
Second, the reproducibility of the dynamic event is interferogram data points are determined by continually
important since the collection of data for the time-resolved measuring the precise speed of the mirror throughout its
spectra requires multiple interferometer scans and multi- movement. The resulting unevenly sampled interferograms
ple repetitions of the time-dependent sample perturbation. cannot be transformed using the conventional FFT, and so
This restriction means that the intensity of the excitation a least-squares fitting routine is necessary. The complete
source must be nearly identical for every repetition. Simi- data-handling algorithm and its comparison to standard FFT
larly, the sample must revert to its original state following computations are described completely in the literature.21,22
the dynamic event without suffering any degradation or The end result is that the event-locked stroboscopic method
attenuation, or alternatively, a new sample must be intro- decreases artifacts and noise associated with mirror velocity
duced before the next transient is initiated. Early work variations at the cost of a small increase in expense, data
using stroboscopic sampling was beleaguered by spectral processing time, and computer memory.
artifacts that were not discovered as such for some time.18 Stroboscopic sampling maximizes data collection effi-
Later, other researchers recognized that fluctuations in the ciency not only by multiplexing spectral data via the
number of transients generated per excitation event due to interferometer, but also by multiplexing the temporal data
shot-to-shot variations in the excitation source had led to through the simultaneous collection of multiple temporally
spectral intensity variations that were in-phase with the data resolved interferograms. This feature leads to a tremen-
collection.19 The result was a series of artificial emission dous decrease in data collection time versus individually
bands that were unfortunately unrecognized as such and recording an entire interferogram for a single time delay
subsequently interpreted as originating from the samples in a pseudo-pulse/probe manner. For instance, in a typical
excited state. The discovery of these insidious artifacts has interferometer scan, over 1000 data points must be taken
plagued the acceptance of stroboscopic sampling ever since, to achieve reasonable spectral resolution over the entire
and the experimental details of any stroboscopic measure- MIR region. To collect a single time-resolved spectrum,
ment should be considered with caution. the reaction of interest would need to be repeated 1000
Third, the scanning velocity of the moving mirror must times with the appropriate excitation delay synchronized
be held at a constant value to guarantee that the temporal to the ADC sampling at each HeNe fringe. To collect a
acquisition of data is consistent. This latter point is espe- time-resolved group of 20 spectra, a total of 20 000 flash-
cially important. Triggering the ADC sampling using the initiated events would have to be reproduced exactly. This
HeNe laser zero crossings removes uncertainties concern- requirement places an unacceptably large obligation on the
ing the position of the sampled point along the interfero- instrument, excitation, and sample stability.
gram, and so it is more reliable than measuring the signal The duty cycle of the stroboscopic method can be further
at even intervals of time in rapid-scanning FT-IR of static increased if the event of interest is relatively short lived in
relation to the time required for a single scan. For instance, 1 2 3 4 5 6 7

t1
if we assume that the spectral range and resolution in the
example used for Figure 3 require a 1000-point interfero-
gram (therefore taking 10 000 s to complete), and that the HeNe
reference
samples dynamic changes are fully completed in 200 s, fringes
then we can potentially repeat the experiment 50 times over
the course of one sweep of the interferometer mirror. In
effect, 50 data points will be added to each time-resolved
spectrum with each pass of the mirror. Hence, a total of ADC
only 20 mirror scans and 1000 laser shots will be necessary sampling
to generate 20 time-resolved spectra of the dynamic event, t2
an increase of over 20-fold in data collection efficiency
Figure 5. The limiting case of stroboscopic sampling requires the
compared with the individual spectrum method described rapid triggering of equally spaced ADC samples following each
above. zero crossing of the HeNe reference laser. t2 must be much
Stroboscopic sampling is a relatively inexpensive means smaller than t1 . Other stringent restrictions apply to this form
of performing TRIR on transient samples with a maximum of FT-TRIR and are discussed in the text.
time resolution of 1 s. The only instrumentation required
At the end of the scan, a number of interferograms equal
is a standard FT-IR spectrometer capable of synchronizing
to the number of ADC samples taken at each laser zero
an external excitation source. The rest of the techniques
crossing will have been generated. The time delay between
parameters are software controlled. The upper limit in time
subsequent interferograms can be as short as 10 ns using
resolution is enforced by the maximum mirror velocity that a fast photovoltaic (PV) MCT detector and a 100-MHz
the spectrometer can achieve, and there is a large degree of transient digitizer. Moreover, because the accuracy of the
variability in mirror velocities between different makes and time resolution is solely a function of the degree of control
models of spectrometers. Many top-of-the-line commercial between the excitation pulse and detector gate, i.e. it is
instruments provide a wide range of mirror velocities rang- decoupled from the scanning velocity of the interferometer
ing from about 0.03 to 6 cm s1 . Less expensive models mirror, this technique of FT-TRIR is less prone to temporal
may only allow a limited number of choices, and normally artifacts than the typical stroboscopic method.
these options are restricted to the intermediate range of the It is important to note that serious restrictions apply to
velocity spectrum. this simplified stroboscopic method. First, the minimum rate
A final, yet limited, application of stroboscopic sam- of digitization is limited by the slowest mirror velocity
pling serves as a good segue to the description of one of available, usually relegating the transient sampling to less
the fastest growing interferometer-based TRIR techniques, than 5 s per time slice. This requirement exists because
step-scan FT-IR time-resolved spectroscopy. In the event a complete set of temporal data points must be collected
that the transient of interest has a lifetime that is very short before the mirror has translated a significant portion of the
with respect to the delay between HeNe laser fringes, a distance toward the next HeNe laser fringe. If the temporal
limiting case of stroboscopic sampling exists.1 In this dis- recording of data extends to greater than a few percent of
tinctive case, the transient can be generated by a quick the time delay between HeNe fringes, then the sampling
perturbation at each HeNe fringe. Following the pertur- position inaccuracies for each time-resolved interferogram
bation, a fast gated detector is used to sample the transient become so great as to introduce a substantial amount of
IR signal at equal time intervals throughout the decay of the noise into the transformed spectra.
dynamic event. Figure 5 shows the data collection scheme A second, but related, restriction is that the transient
for this stroboscopic variant. species that are appropriate for this time-resolved technique
Although the mirror continues to move during the detec- must have lifetimes shorter than 100 s. If the lifetime of
tors successive readouts, the relatively small position the transient is appreciably greater than this value, a signifi-
deviations along the time series of measurements are incon- cant population of molecules will still be in the excited state
sequential if the phase shift introduced is small with respect by the onset of the next perturbation event, even at the mod-
to the spacing of HeNe fringes (<1%). If the phase shift is est interferometer scanning rate previously mentioned. If
larger, then it can be removed by software prior to Fourier this situation were to occur, then the experiment would vio-
analysis. This positional error can be minimized by decreas- late the paramount requirement of stroboscopic sampling,
ing the mirror velocity so that the sampling rate is several namely reproducible generation of the transient species.
orders of magnitude faster than the Fourier frequency of Third, the need to repeat the excitation at every reference
the HeNe laser. laser fringe requires sample excitation at a minimum rate
of 1 kHz, provided that a suitably slow mirror scan veloc- jumps, or steps. At each retardation point, the moving
ity is available. This repetition rate is a fairly stringent mirror is held still while data are collected. Time-resolved
demand, for example, when the experiment necessitates data at each position in the interferogram can be recorded
high-powered laser excitation. Even with significant under- following an excitation event, and multiple data points at
sampling of the interferogram, this requirement can be that constant retardation can be successively coadded if nec-
a challenge. Overall, the limiting case of time-resolved essary to achieve a higher S/N. At each discrete point in
stroboscopic FT-IR spectroscopy has particularly restrictive the interferogram, the spectral multiplexing is effectively
provisions, and there exist few applications that can meet decoupled from the time domain, i.e. there are no temporal
them all. Fourier frequencies, so any degree of time resolution with
Stroboscopic sampling has developed a devoted, yet any S/N and any spectral resolution is theoretically possible.
modest following among IR spectroscopists. Much of the Figure 6 shows the data-recording scheme used in the
recent work using this technique has been generated by step-scan mode of FT-TRIR. The data are collected in time-
Leone and co-workers using IR emission and the clas- resolved columns (temporal data) at each point along the
sic stroboscopic method23 26 and by Sloan and co-workers row of interferogram retardation points (, interferogram
using IR absorption and the limiting case of stroboscopic data). At each position in the interferogram row, the tran-
sampling.27 29 Their numerous investigations into gas- sient is created and the IR detector is sampled at a set time
phase kinetics, including reaction dynamics, intermediates, interval along the temporal axis until the sample has fully
and energy transfer rates, give an interesting and up-to-date relaxed. The first sequence of recorded data is shown by
account of the applications and potential of the stroboscopic the numbered circles in the first column (points 110). At
technique. However, for the study of many shorter-lived as this point, the sample excitation can be identically repeated
well as longer-lived transients, a more recently commer- and the data averaged along the same temporal column to
cialized FT-TRIR technique, step-scan FT-IR time-resolved improve the S/N (again, 110), or, if an appropriate S/N
spectroscopy, has become vital and widely applied. has already been achieved, the mirror can be moved to the
next point in the interferogram and the excitation repeated
to collect the second column of temporally resolved data
4 STEP-SCAN INTERFEROMETRY (points 1120).
Although the ultrarapid scanning and stroboscopic methods

extend the capabilities of continuous-scan FT-IR to include
Change in intensity (I)
relatively fast dynamic measurements, routine continuous-

scan FT-IR is ultimately limited to analyzing processes
whose duration extends a few hundred microseconds, at
least as long as a continuous-wave (CW) source such as a
globar is used. This limitation is imposed because the time
dependence of the dynamic measurement is interwoven
with the time response of the sample. In essence, the
recording of the interferogram using a CW source and
a constant-velocity FT-IR instrument is itself always a
function of time. This complication has given rise to a
10 20 30 40 50 60 70 80 90 100
renewed interest in step-scanning interferometry since the 9 19 29 39 49 59 69 79 89 99
1980s and its application to time-domain spectroscopy, 8 18 28 38 48 58
68 78 88 98
7 17 27 37 47 57 67
particularly on the submicrosecond time scale where few 77 87 97
)
6 16 26 36 46 56
66 76 86 96 (t 1
other techniques can compete (see Figure 1). 5 15 25 35 45 55 ta
65 75 85 95
da
The step-scan method, originally developed in its modern 4 14 24 34 44 54 64 74
84 94
o ral
form in the early 1960s, saw only limited application until
3 13 23 33 43 53 63 73
83 93 mp
2 12 22 32 42 52 62 72 82
92
Te
the early 1990s.4,30 33 The novelty of the technique lies in 1 11 21 31 41 51 61 71 81
91
its ability to decouple the time dependence of the IR mea- Retardation ()

surement from the time dependence of the dynamic event
being investigated. In effect, it is analogous to the limiting Figure 6. Generation of time-resolved interferograms by use of
step-scanning FT-IR spectroscopy. Numbered circles below the
case of the stroboscopic method described above except that interferograms show the order of data sampling. When using step-
the mirror velocity is now infinitely slow. In step-scanning scanning FT-IR spectroscopy, data collection proceeds in columns
interferometry, the mirror retardation is achieved in discrete along the temporal axis.
This cycle is repeated until each interferogram data point stroboscopic method without the concern for spectral arti-
has been collected for the required spectral range and facts arising from mirror velocity uncertainties.
resolution of the experiment. The three-dimensional array Step-scan FT-TRIR, however, is currently limited to time
of data is then sorted into time-resolved interferograms and resolution of the order of 1 ns owing to the state-of-the-art
subjected to FFT to generate a stacked plot of time-resolved detector and electronics technology available, and therefore
spectra. In order to improve the S/N, the results of two or falls below the temporal resolution achievable with tunable
more sequential step-scan data sets can be coadded. The laser and laser-mixing techniques. Despite this shortcoming,
dynamic event, like that in stroboscopic sampling, must the use of the step-scan mode for nanosecond-regime appli-
be infinitely repeatable and exactly reproducible over the cations excels in terms of the spectral range and resolution
length of the experiment or else the S/N will be severely that can be realized with relatively short data collection
degraded. This requirement extends to the integrity of the times. The limited range and tedious tuning of the laser
sample over time as well as to the intensity of the excitation sources used in many ultrafast TRIR techniques make the
event with each cycle. collection of high-resolution, broadband spectra particularly
Several experimental safeguards have been used to guar- frustrating. As a result, applications of step-scan FT-IR for
antee that the these requirements are met. When study- time-resolved spectroscopy have surged (see below).
ing irreversible reactions, flow cells for liquid or gas The instrumentation necessary for step-scan FT-TRIR is
reagents34,35 and rotating or translating sample stages for less costly than that required for the ultrafast laser methods,
other samples36 have been used to assure fresh sample but more expensive than that demanded for stroboscopic
is introduced into the path of the IR radiation with each sampling. The addition of the step-scan mode of interfero-
cycle of the dynamic event. This strategy is also use- meter movement adds considerably to the expense of the
ful when monitoring the early stages in a dynamic event FT-IR instrument and generally pushes the user into the
with an inconveniently long lifetime.37 It is also important realm of research-grade spectrometers. Furthermore, three
that the population of excited molecules is nearly identi- areas of instrumentation and electronics must be considered
cal with each cycle of the dynamic event. As an example, in adapting step-scan interferometry to time-resolved stud-
high-power laser sources, which often vary in shot-to-shot ies. First, a means to synchronize the sample excitation,
intensity by as much as 10%, present a challenge for gener- data collection, and interferometer stepping must be con-
ating comparable numbers of excited-state molecules with sidered. Second, the appropriate detector and digitizer must
each pulse. In order to ensure reproducible excitation, the be selected for the time regime and budget of the exper-
results of numerous laser shots can be coadded at each inter- iment. Finally, in some step-scan interferometers, e.g. the
ferogram position to achieve a reproducible average value, BioRad FTS 60A/896, the stepping mirrors dither, the
or the laser intensity with each shot can be detected with a rapid modulation used in positioning the mirror, must be
photodiode and made to fall within a specified range before demodulated from the transient signal of interest.38,39
accepting the temporal data for that interferogram position. Synchronization of the experiments excitation source,
Alternatively, the excitation intensity can be monitored and data collection, and interferometer movements can be
used to normalize the spectral response to each laser pulse. arranged in two ways. The most convenient manner of tim-
Clearly, the requisite reproducibility and repeatability for ing is to use an external digital pulse generator to serve
step-scan data collection can be met for nearly any dynamic as the master experimental clock. This arrangement ben-
event if suitable precautions are taken. efits from the timing accuracy of the pulse generator, its
Dynamic step-scan FT-IR spectroscopy offers several low phase jitter (normally <0.5 ns), and its ability to man-
advantages over continuous-scan techniques (ultrarapid age complex triggering patterns. In a typical experiment,
scanning and stroboscopic sampling) as well as IR laser, the pulse generator will send a trigger signal to the FT-IR
dispersive, frequency mixing, and upconversion methods of bench or transient digitizer to mark the beginning of data
TRIR. In contrast to continuous-scan instruments, the step- collection. The digitizer will then initiate a stream of detec-
scan mode of data collection is only restricted in terms of tor samples, evenly spaced according to the preset time
time resolution by the speed and sensitivity of available IR resolution of sampling. A second trigger signal from the
detectors, amplifiers, and digitizers, as well as the tempo- master clock will then be sent to the excitation source after
ral profile of the excitation source. Since the mirror is held a preprogrammed delay. This trigger will initiate the sam-
constant during data collection, there is no fundamental lim- ple perturbation (via laser pulse, rheometer motion, electric
itation imposed by the technique to the speed, whether slow field pulse, reagent mixing, ignition, etc.).
or fast, at which temporal data can be recorded. Further- Software controlling the digitization will allow data
more, the step-scan mode of data collection offers a sim- recording to proceed to a preset interval along the decay
pler alternative to the complicated and calculation-intensive of the sample. If the software is programmed to coadd
numerous laser pulses at each step in the interferometer, Unlike most IR detectors, the photoconductive MCT in
then the mirror will remain fixed while awaiting the next this instance is DC-coupled to prevent any low-frequency
excitation cycle. The next transient event will occur after a filtering. This configuration can be contrasted with the typ-
sufficiently long delay to allow the sample to recover fully, ical AC-coupled detector where an RC circuit is used to
or in the event of an irreversible process, for new reagents remove frequencies below 10 Hz. AC coupling is normally
or sample to enter the path of the IR beam. Then, the pulse preferred as a convenient means of eliminating slow
generator triggering cycle is repeated as above. If, how- changes in instrument operation and thus limiting spectral
ever, the number of coadditions has been met, the FT-IR noise in a conventional rapid-scan measurement while max-
bench steps the mirror to the next position along the retar- imizing the digitization resolution of the ADC. However,
dation path and prepares for the next triggering sequence. when measuring the decay of transient species with life-
It is normally imperative that the short triggering pulses times of the order of several milliseconds, the lower filtering
be transmitted via 50-
cabling/BNC connector combina- limit must be overridden via DC coupling to prevent los-
tions since the digitizer circuits are sensitive to impedance ing portions of the transient signal. Following detection of
mismatches that could lead to signal attenuation or ringing the dynamic signal, the acquisition processor (AQP) stores
along the coaxial cable. the data in on-board memory until the end of the experi-
The second option for experiment synchronization is ment, after which they are subjected to FFT and saved to
to use the FT-IR benchs software for sequence timing. hard-disk memory.
Most commercial software and bench electronics allow for To take advantage of the highest time resolution capable
internal experiment triggering and preset data collection in step-scan operation, a specialized transient digitizer must
delays. It is possible to use these features to synchronize the be used, in addition to a high-bandwidth PV MCT. Most of
sample excitation, transient detection, and mirror stepping. the major FT-IR instrument manufacturers that offer a step-
For instance, the spectrometer can be set to step at a scan feature can supply at least a 100- or 200-MHz transient
predetermined rate, say 10 Hz. At each mirror step, the recorder that is programmed to work with their instrument
software triggers the external sample excitation source software to collect fast time-resolved spectra. However,
following a preset delay to allow for mirror stabilization because of the operation speed of the transient recorder,
and the initiation of data collection before the perturbation. these boards often come with only 812-bit digitization.
Using the stepping frequency of the interferometer mirror An example of this type of data recorder is the Spectrum
as the master clock precludes the coaddition of data at each PAD82a. The PAD82a is housed on the motherboard of
position in the interferogram. However, multiple passes the personal computer responsible for controlling the FT-IR
of the interferometer can be signal averaged to increase bench. It accepts a trigger input, two channels of data, and
the S/N of the resulting spectra. This scheme for signal an external sampling trigger. The board typically operates
averaging can, in the end, be beneficial since it removes the at 100 MHz when simultaneously using both channels to
influence of slow drift of electronic and optical components. record two signals, but it can digitize at 200 MHz if only
Its major drawback is that the stepping cannot be much one signal is being recorded.
faster than 10 Hz and still allow the tens of milliseconds To realize the sampling rate of the transient digitizer
necessary for mirror stabilization, and so the repetition of and maximize its sensitivity, an AC/DC dual-output PV
the transient event is also limited to this low duty cycle. MCT with a fast preamplifier is required. The speed of the
The second consideration for adapting step-scan FT-IR detector must be at least as great as the digitization rate
to time-resolved measurements is the proper selection of of the transient recorded, otherwise the time-resolved data
detectors, amplifiers, and digitizers. Again, the modes of will be convolved with the rise time of the detector. With
recording data are generally divided into two categories a dual-channel detector, the AC output is used to carry
that can be identified by the detector coupling that is used, the transient changes in the interferogram at the position
namely DC or alternating current (AC) detection. The DC being monitored. The switch from a DC signal, as in the
method, the least expensive and less complicated means of previous case, to an AC signal is employed in order to
collecting the data, utilizes the internal ADC of the FT-IR maximize the smaller dynamic range of the 100-MHz digi-
bench as the data digitizer. Most on-board ADCs operate tizers. However, a 50-MHz preamplifier has been described
at a maximum rate of 200 kHz with 16-bit amplitude that combines the optimum utilization of dynamic range
resolution, thus limiting time resolution to a maximum of offered by AC coupling with the broadband sensitivity of
5 s between temporal data points. If this is acceptable, DC coupling. The quasi-DC-coupled preamplifier removes
a common photoconductive MCT and amplifier (typically the lower frequency limit of standard AC-coupled detectors
1 MHz, 250 ns rise time) can be used to sample the (1 kHz) so that dynamic events with interesting kinetics
transient IR signal. over a wide time scale (nanoseconds to milliseconds) can
be monitored in a single experiment with maximum digiti- spectrometer, digital pulse generator, AC/DC dual chan-
zation sensitivity.40 nel PV MCT, and transient digitizer.5 The time-resolved
It is important to note, however, that the collection of measurement begins with the stepping mirror in the ZPD
an AC interferogram introduces a new complication to the position. The pulse generator initiates the experiment by
postcollection data manipulation. Being an AC signal, these triggering the laser flashlamps to commence population
data can contain both positive and negative intensities. The inversion within the laser cavity; a Pockels cell, or Q-switch,
final Mertz or Forman phase correction routines of the FFT prevents lasing at this point. After an appropriate delay, the
are unable to handle interferogram data that contain nega- pulse generator sends a second trigger to the transient dig-
tive signed components. Therefore, a DC signal is collected itizer to activate data collection. The AC signal from the
separately, but simultaneously, at each mirror position and PV MCT is recorded at a constant time interval preselected
the resulting DC interferogram is later used to phase cor- using the computer software. A single DC data point is also
rect the AC interferograms. Although a stored phase from recorded from the second, DC-coupled detector preampli-
a previous DC spectrum could be collected ahead of time fier. A third trigger activates the sample excitation source,
and used later for phase correction of the AC time-resolved and the laser pulse generates the transient species instan-
interferograms, the concurrent collection of a DC interfero- taneously on the time scale of the experiment. Throughout
gram with the AC interferograms eliminates any phase this process, data digitization of the AC signal continues at
inaccuracies that might occur between the two experiments. the software-selected rate, thus yielding temporal data both
Kolmar Technologies provides a dual-channel PV MCT and before and after the samples excitation.
fast preamplifier (100 MHz, Model KV104) with an 11-s The software recognizes if multiple coadditions were
cutoff and theoretical rise time of 3.5 ns. With the increasing requested, and allows multiple triggering events to occur at
miniaturization of electronics, one can expect to see even the current mirror position. The PAD82a transient digitizer
faster detectors in the near future that benefit from the sum- averages these repetitions on the fly, and following the
ming of multiple independent MCT elements, each using last excitation cycle, the temporal data are transferred to
its own preamplifier. However, for the time being, using the the AQP where they are stored in on-board memory. The
maximum sampling rate of the transient digitizer described PAD82a, unlike some other digitizers, has a zero latency
above, and recording both the AC and DC channels of period for data transfer, and so the shuffling of data during
the KV104, this detection and recording arrangement can the course of the measurement does not diminish the normal
achieve a maximum time resolution of 10 ns. duty cycle of the experiment. After the AQP has received
Figure 7 shows the experimental setup employed by the data, the mirror is stepped to the next mirror position,
the present authors for collecting nanosecond regime flash and the PAD82a board is signaled to begin collecting the
photolysis time-resolved spectra with a step-scan FT-IR next DC data point and temporal AC data column following
FT-IR spectrometer
Computer with AQP
and transient recorder Globar
source
Sample
AC/DC
PV MCT
2. Trigger transient
recorder
Pulse generator 355 nm
532 nm
1064 nm
3. Trigger laser
1. Trigger laser Q-switch
flashlamps
Harmonic Pellin-Broca
generator prism
Nd : YAG laser
Figure 7. Experimental setup for collecting nanosecond regime, broadband MIR time-resolved spectra of a laser-induced transient
species using a step-scan FT-IR spectrometer.
the subsequent trigger pulse from the signal generator. This of the experiment was limited by the rise time of the
process continues until all the data have been recorded, or photomultiplier detector. However, with the high sampling
until the on-board memory of the AQP has been filled. rate of the DSO, the resolution could theoretically be as
Once the full data set has been recorded and passed to the high as 500 ps. This value can ostensibly be decreased as
AQP, each of the AC interferograms is sequentially phase newer DSO and detector technology is introduced. As yet,
corrected using the DC interferogram and subjected to FFT subnanosecond data collected in the MIR region have not
to generate a time-resolved array of intensity difference been reported in the literature, although a Bruker Optics
(I) vs wavenumber spectra. These data can then be product note reports true 10-ns time resolution for the
processed into an array of absorption difference (A) output of a pulsing MIR (6450 cm1 ) diode.
spectra for subsequent interpretation. As previously mentioned, the advantages of the DSO-
Several variations for recording nanosecond FT-TRIR based method include the higher time resolution of the
data have been demonstrated, some of which seek to reduce oscilloscope data collection versus currently available PC-
costs whereas others attempt to increase the time reso- based digitizers and the ability to use hardware already
lution of the step-scan technique. The added expense of available in many laboratories. Moreover, several other
a dual-channel PV MCT, required for collecting the DC benefits can be realized. First, the DSO method demon-
interferogram for successful phase correction, can be cir- strates a fivefold increase over ISA-bus transient digitizer
cumvented by using a clever double-angle phase-correction boards in the S/N of resulting spectra. This increase has
algorithm described by Hutson and Braiman.41 Their cor- been attributed to the isolation of the oscilloscope digitizer
rection routine overcomes the obstacle of positive and from the PC electronics, and also the increased accuracy of
negative signed intensities in the AC interferograms by the DSO trigger response. Furthermore, the ability to store
convolving a double-sided transient AC interferogram with raw data on the PCs hard disk before sending it through
itself. The convolution doubles the phase angles, thereby the final software algorithms allows the results of various
removing phase-angle discontinuities present in the data apodization, zero-filling, and phase-correction manipula-
due to sign changes. The double-angle phase correction tions to be examined without the loss of the original data.
technique represents a significant advance over the Mertz The data can also be scrutinized to be certain that they meet
Signed phase-correction routine and reduces the cost of the certain boundary conditions, such as an excitation laser
PV MCT by allowing a single-output detector to be used. intensity threshold as measured by a fast photodiode. How-
However, this direct phase-correction technique requires ever, experiment-based specifications such as frequency
complete, double-sided interferograms to be measured, and range, spectral and phase resolution, and double- or single-
therefore the time spent collecting data using this method sided nature of the interferogram are not alterable.
is twice that expended with a two-channel detector capable One drawback of the DSO technique is that the duty
of simultaneously measuring a DC-coupled interferogram. cycle of the data collection is limited by the data-transfer
For faster time resolution, Johnson and co-workers rate of the DSO to the transfer module, TM1000. The
reported a variation on the step-scan transient data col- data recorded at each mirror position in the interferometer
lection routine using an 8-bit, 2 Gsample s1 digital stor- must be moved ultimately to computer storage before the
age oscilloscope (DSO) acting as the transient digitizer transient signal at the next position is recorded. This latency
rather than a PC-based digitizer like the PAD82a.3 The period is larger than that of PC-based digitizer boards.
DSO works in conjunction with the standard instrument- The current transfer limit stands at about 30 Hz, which is
control software and an additional, commercially available, actually well beyond the repetition rates of many of the
data-transfer module, the McLaren Research TM1000, and currently popular high-energy laser excitation experiments
its associated software. This upgrade can be implemented that operate at 1020 Hz. However, since the limit is
fairly easily since many laboratories already possess a fast not intrinsic to the equipment, but rather to firmware and
DSO for laser and electronic diagnostics, leaving only the software, one can expect the data-transfer rates to improve
data transfer module to be purchased for $10 000. The with new technology.
time-resolved signals are recorded by the DSO digitizer, The final consideration for using conventional step-scan
transferred to PC memory, reconstructed as a slower wave- instruments for fast time-resolved spectroscopy involves
form via a 20-s digital-to-analog converter (DAC), and only those instruments that use a phase jitter on the step-
then fed back through the FT-IR instruments ADC at its ping mirror for position control.38,39 The phase dither,
normal digitization rate, usually around 200 kHz. although allowing for constant dynamic alignment of the
The DSO-sampled data that are demonstrated in the mirrors position, adds an unfortunate complication to the
original report are from the visible emission of a dye measurement of time-resolved spectra. The constant phase
laser-pumped NO2 gas cell. The 2.5-ns time resolution modulation of the mirror generates its own transient signal,
and this signals amplitude is normally far larger than that of mode. In either case, the repetition rate for creating the
the transient event being measured. Therefore, techniques transient event must conform to the Nyquist requirement,
must be used to demodulate the mirrors movement from i.e. it must be 2 the highest Fourier frequency in the
the signal of interest. spectral range. A repetition rate of 10-fold greater than the
The most successful procedure for eliminating the instru- highest Fourier frequency is often used for good measure.
ments transient signal has been to synchronize closely the The essence of the asynchronous sampling technique is
phase of data collection and sample excitation with the that the sample perturbation is not timed with respect to
phase of the mirror dither.42 In BioRads FTS 60A/896, the scanning and data acquisition of the spectrometer, i.e.
the mirror dither is a constant 16 kHz. By detecting with the two are asynchronous. It is also worth noting that the
a gated boxcar integrator set 62.5 s before the sample is standard asynchronous sampling technique gives a single
excited and a second integrator coincident with the sam- time-delayed spectrum with each experiment. However, for
ples excitation, the effect of the mirrors dither can be the CW mode, a multichannel gated device is available from
removed. The instrument-derived signal is nulled because Sopac Corporation and S. T. Japan, Inc. that allows the col-
the matched gates are exactly in-phase with the mirrors lection of up to 16 time-resolved spectra in one experiment.
dither, and therefore, when the light off (62.5 s before In either the CW or pulsed mode, the fundamentals
excitation) and light on signals are ratioed such that of the signal processing are identical for the collection
A D loglight/dark , the absorption difference spectrum of a single time-delayed spectrum. The signal generation
no longer contains the transient from the mirror modula- and processing for a single IR wavelength is shown in
tion. This calculation, however, only yields a single excited Figure 8. This simple example is illustrated for the pur-
sample spectrum rather than a time-resolved group of spec- pose of explaining the principles of the technique, but in
tra. Although further modifications can be used to generate an actual asynchronous sampling experiment, the signal,
the required data set, they are not without added difficulty data collection, and processing involve the entire inter-
and cost. Clearly, for the fast time-domain experiment, the ferogram and are thus the sum of all frequencies of the
benefits of the positional accuracy allowed by the mirror IR source. The constant-velocity interferometer modulates
modulation are arguably dispensable. each frequency of the source at its characteristic Fourier fre-
The time regime in which step-scan FT-IR dominates quency (Figure 8a). The sample perturbation induces spikes
is particularly rich in applications. The literature is laden in the sinusoidally modulated intensities corresponding to
with examples of step-scan FT-IR being used to col- each wavelength in the spectrum, at intervals of t, where
lect broadband MIR spectra of dynamic systems. A few t is the spacing between perturbation events (Figure 8b).
examples include studies of biological photocycles,43,44 Sampling the decay of these spikes at a delay of t fol-
organic reactions in zeolites,45 small ligand binding to bio- lowing the perturbation (Figure 8c) using a gated detector
logical proteins,46 impulse photoacoustic measurements,47 (or a probe pulse) produces a discrete high-frequency
mechanisms of pericyclic ring openings,35 hydrosylation interferogram carried by the normal analog interferogram
bond activation,34 time-domain polymer rheology,48,49 pro- (Figure 8d). Low-pass filtering of the signal recovers the
tein folding dynamics,50 and inorganic complex excited analog interferogram, but now with the intensity of each of
states.51 54 As step-scan FT-IR becomes more widely its frequencies altered to reflect the spectral band intensities
known in the scientific community, and as the time res- at time t (Figure 8e). This interferogram is then sampled
olution of the technique creeps closer to the picosecond at the ADCs normal digitization rate, which according to
regime, further applications and more widespread use are a the Nyquist theorem should be at least twice the highest
certainty. Fourier frequency.
As mentioned, signal recording in the CW mode is
performed with a narrow gate applied to the detector at
5 ASYNCHRONOUS SAMPLING t. In the pulsed probe mode, the gate is replaced by a
probe pulse delayed with respect to the pump pulse by
The final technique in FT-TRIR spectroscopy, asynchronous t. The gate width is similarly replaced by the width of
sampling, has been dealt with previously in terms of instru- the probe pulse. In both instances, the limit to the time
mentation. Therefore, the instrumentation and theory will resolution comes from the pulse width of the excitation
be dealt with in only a cursory fashion here. First, it should source, the temporal accuracy of t, the narrowness of the
be emphasized that the asynchronous technique is applied gate or probe pulse, and (in the CW mode) the rise time
with a standard rapid-scan interferometer. Step-scanning is of the detector. The final data collection results from the
not involved. With asynchronous sampling, two possibil- ADC sampling of the analog interferogram after low-pass
ities exist: a CW probe mode (globar) or a pulsed probe filtering.
the Fourier frequencies, means that only relatively rapid

processes can be studied since the sample must fully relax
before the next perturbation event. For example, a 10-kHz
perturbation rate can easily be a 10-fold increase over the
(a)
highest Fourier frequency in the MIR range at moderately

slow scan speeds, but such a repetition rate requires that
the samples excited-state half-life be less than 30 s.

Also, certain low repetition rate lasers (e.g. a high-power

Q-switched Nd : YAG laser) are unsuitable as excitation
sources. Thus the CW mode of asynchronous sampling
overlaps the faster time regime of step-scanning FT-IR
(b) spectroscopy (100 s to 1 ns), but, unlike step scanning, it
cannot be applied to longer time scales.
The pulsed probe asynchronous sampling technique, on
the other hand, has the potential to open up the sub-
nanosecond and possibly subpicosecond time regimes to
interferometric techniques. The possibility of the pulsed
probe experiment was first suggested by Masutani et al.
in their original publication,55 but since then it has been
(c) demonstrated only in time-resolved resonance FT-Raman
(FT-TR3 )58,59 and in a closely related experiment in the
far-IR region using a pulsed synchrotron source.60 The
important factor here is that the time resolution in such a
pulsed pumpprobe technique depends only on the widths
of the pump and probe pulses and the accuracy of their
(d) relative timing, not on the rise time of the detector.
In the pulsed asynchronous sampling mode, the fast rep-
etition rate of the pump (and now probe) pulse is still
required. Its application in the MIR region requires a high
repetition rate pulsed source that is broadband and tem-
porally narrow, features not found in most conventional
IR sources. As pointed out by the originators of the tech-
(e)
nique, the most promising source is probably synchrotron
Figure 8. Data collection scheme for asynchronous sampling. radiation,61 and a number of research groups are work-
For simplicity, only the Fourier modulation of one wavelength ing to demonstrate such an experiment in the MIR region
of light is shown (a). This wavelength is further modulated by at synchrotron storage rings around the world.5,60,62 The
perturbation pulses at intervals of t (b). Each perturbation is
pulsed output of a synchrotron is normally in the range
followed by a probe at t after the perturbation (c). The detector
gate or probe pulses generate a discrete interferogram (d) that of megahertz, and this feature may restrict its application
is converted back to an analog interferogram for time t by to even shorter-lived transients than previously mentioned
the low-pass filter where the interferogram is sampled by the unless appropriate steps are taken to reject a portion of
spectrometers ADC at its normal rate (e). the pulses from being recorded by the detector. As more
progress is made in developing this method of FT-TRIR,
The CW asynchronous sampling method, first proposed one can expect interferometric techniques to begin to chal-
and demonstrated by Masutani et al. in 1992,55 has been lenge laser-based methods in the ultrafast time regime.
described in detail elsewhere in this volume. It has been
shown to be particularly valuable in the study of the
microsecond reorientation dynamics of ferroelectric liquid 6 SUMMARY
crystals in response to pulsed electric fields.56,57 Although
nanosecond range time resolution is possible, the same In this article, we have systematically reviewed the ways
detector rise time limitation applies here as in the step-scan in which FT-IR spectroscopy, with all its associated advan-
method, which also uses a CW source. It is also worth tages, has been used in the measurements of time-dependent
noting that the requirement of relatively high repetition phenomena and transient species in the MIR region. These
rates, to move the perturbation frequency well outside methods span time-resolution capability from >1 s to <1 ns
and are capable of measuring transient absorptions as small 12. C.J. Manning, in Fourier Transform Spectroscopy: 11th
as 104 absorbance units. The upper time resolution limit International Conference, ed. J.A. de Haseth, American
Institute of Physics, Woodbury, NY, 84, Vol. 430 (1998).
will continue to decrease for FT-TRIR techniques as new
detectors, digitizers, and broadband IR sources become 13. P.R. Griffiths, B.L. Hirsche and C.J. Manning, Vib. Spec-
trosc., 19, 165 (1999).
available.
14. A. Rosencwaig, Photoacoustic and Photothermal Spectro-
scopy, John Wiley & Sons, New York, Vol. 57 (1980).
15. E.N. Lewis, P.J. Treado, R.C. Reeder, G.M. Story, A.E.
ABBREVIATIONS AND ACRONYMS Dowrey, C. Marcott and I.W. Levin, Anal. Chem., 67, 3377
(1995).
AC Alternating Current 16. A.W. Mantz, Appl. Spectrosc., 30, 459 (1976).
ADC Analog-to-digital Converter
17. W.M. Grim, J.A. Graham, R.M. Hammaker and W.G. Fate-
AQP Acquisition Processor ley, Am. Lab., 16, 22 (1984).
CW Continuous-wave
18. A.W. Mantz, Appl. Opt., 17, 1347 (1978).
DAC Digital-to-analog Converter
19. A.A. Garrison, R.A. Crocombe, G. Mamantov and J.A.
DC Direct Current
deHaseth, Appl. Spectrosc., 34, 399 (1980).
DSO Digital Storage Oscilloscope
20. H. Weidner and R.E. Peale, Appl. Opt., 35, 2849 (1996).
FFT Fast Fourier Transform
FPA Focal Plane Array 21. H. Weidner and R.E. Peale, Appl. Spectrosc., 51, 1106 (1997).
FT-TR3 Time-resolved Resonance Fourier 22. H. Weidner and R.E. Peale, Appl. Spectrosc., 52, 587 (1998).
Transform-Raman 23. S.A. Rogers and S.R. Leone, Appl. Spectrosc., 47, 1430
FT-TRIR Fourier Transform Interferometric (1993).
Time-resolved Infrared 24. R.A. Loomis, J.P. Reid and S.R. Leone, J. Chem. Phys., 112,
TRIR Time-resolved Infrared 658 (2000).
TRS Time-resolved Spectroscopy 25. P.W. Seakins, E.L. Woodbridge and S.R. Leone, J. Phys.
Chem., 97, 5633 (1993).
26. J. Lindner, R.A. Loonis, J.J. Klaassen and S.R. Leone, J.
REFERENCES Chem. Phys., 108, 1944 (1998).
27. P.A. Berg and J.J. Sloan, Rev. Sci. Instrum., 64, 2508 (1993).
1. J.J. Sloan and E.J. Kruus, in Time Resolved Spectroscopy, 28. R.V. Kasza, J.G. Shapter, K. Griffiths, P.R. Norton and J.J.
eds R.J.H. Clark and R.E. Hester, John Wiley & Sons, Chich- Sloan, Surf. Sci., 321, L239 (1994).
ester, 219, Vol. 18 (1989). 29. C.A. Carere, W.S. Neil and J.J. Sloan, Appl. Opt., 35, 2857
2. A. Laubereau and M. Stockburger (eds), Time-resolved (1996).
Vibrational Spectroscopy, Springer, Berlin, Vol. 4 (1985). 30. H. Sakai and R.E. Murphy, Appl. Opt., 17, 1342 (1978).
3. L.T. Letendre, H.-L. Dai, I.A. McLaren and T.J. Johnson, 31. R.A. Palmer, C.J. Manning, J.A. Rzepiela, J.M. Widder and
Rev. Sci. Instrum., 70, 18 (1999). J.L. Chao, Appl. Spectrosc., 43, 193 (1989).
4. T.J. Johnson, A. Simon, J.M. Weil and G.W. Harris, Appl. 32. R.A. Palmer, Spectroscopy, 8, 26 (1993).
Spectrosc., 47, 1376 (1993). 33. W. Uhmann, A. Becker, C. Taran and F. Siebert, Appl. Spec-
5. R.A. Palmer, G.D. Smith and P. Chen, Vib. Spectrosc., 19, trosc., 45, 390 (1991).
131 (1999). 34. H. Yang, M.C. Asplund, K.T. Kotz, M.J. Wilkens, H. Frei
6. T. Hirschfeld, in Fourier Transform Infrared Spectroscopy: and C.B. Harris, J. Am. Chem. Soc., 120, 10 154 (1998).
Applications to Chemical Systems, eds J.R. Ferraro and 35. E. Kaufmann, H. Frei and R.A. Mathies, Chem. Phys. Lett.,
L.J. Basile, Academic Press, New York, 193, Vol. 2 266, 554 (1997).
(1979).
36. R. Rammelsberg, S. Boulas, H. Chorongiewski and K. Ge-
7. R. Williams, Spectroscopy and the Fourier Transform: an wert, Vib. Spectrosc., 19, 143 (1999).
Interactive Tutorial, VCH Publishers, New York (1996).
37. C. Rodig and F. Siebert, Vib. Spectrosc., 19, 271 (1999).
8. P.R. Griffiths and J.A. deHaseth, Fourier Transform Infrared
38. R.A. Crocombe, R. Curbelo, J. Leonardi and D.B. Johnson,
Spectroscopy, John Wiley & Sons, New York, Vol. 83
SPIE Proc., 1575, 189 (1991).
(1986).
39. R.A. Crocombe and S.V. Compton, The Design, Perfor-
9. P.R. Griffiths, Chemical Infrared Fourier Transform Spec-
mance and Applications of a Dynamically-aligned Step-scan
troscopy, John Wiley & Sons, New York, Vol. 43 (1975). Interferometer, Bio-Rad Application Note, Digilab Division,
10. S.F. Johnston, Fourier Transform Infrared: a Constantly Richmond, CA (1991).
Evolving Technology, Ellis Horwood, New York (1991). 40. C. Rodig and F. Siebert, in Fourier Transform Spec-
11. M.P. Miller, Appl. Spectrosc. Rev., 23, 329 (1987). troscopy: 11th International Conference, ed. J.A. de Haseth,
American Institute of Physics, Woodbury, NY, 388, Vol. 430 51. M.W. George and J.J. Turner, Coord. Chem. Rev., 177, 201
(1998). (1998).
41. M.S. Hutson and M.S. Braiman, Appl. Spectrosc., 52, 974 52. J.R. Schoonover, C.A. Bignozzi and T.J. Meyer, Coord.
(1998). Chem. Rev., 165, 239 (1997).
42. K.M. Omberg, J.R. Schoonover, J.A. Treadway, R.M. 53. C.A. Bignozzi, J.R. Schoonover and R.B. Dyer, Comments
Leasure, R.B. Dyer and T.J. Meyer, J. Am. Chem. Soc., 119, Inorg. Chem., 18, 77 (1996).
7013 (1997). 54. J.R. Schoonover, G.F. Strouse, K.M. Omberg and R.B. Dyer,
43. A.K. Dioumaev and M.S. Braiman, Photochem. Photobiol., Comments Inorg. Chem., 18, 165 (1995).
66, 755 (1997). 55. K. Masutani, H. Sugisawa, A. Yokota, Y. Furukawa and
44. R. Rammelsberg, B. Hessling, H. Chorongiewski and M. Tasumi, Appl. Spectrosc., 46, 560 (1992).
K. Gewert, Appl. Spectrosc., 51, 558 (1997). 56. K. Masutani, A. Yolota, Y. Furakawa, M. Tasumi and
45. H. Frei, in Fourier Transform Spectroscopy: 11th Interna- A. Yoshizawa, Appl. Spectrosc., 47, 1370 (1993).
tional Conference, ed. J.A. de Haseth, American Institute of 57. M.A. Czarnecki, N. Katayama, Y. Ozaki, M. Satoh,
Physics, Woodbury, NY, 28, Vol. 430 (1998). K. Yoshio, T. Watanabe and T. Yanagi, Appl. Spectrosc., 47,
46. R.A. Palmer, S.E. Plunkett, P. Chen, J.L. Chao and T.J. 1382 (1993).
Tague, Mikrochim. Acta, 14(Suppl.), 603 (1997). 58. A. Sakamoto, Y. Furukawa, M. Tasumi and K. Masutani,
47. B.O. Budevska and C.J. Manning, Appl. Spectrosc., 50, 939 Appl. Spectrosc., 47, 1457 (1993).
(1996). 59. A. Sakamoto, H. Okamoto and M. Tasumi, in Fourier
48. H. Wang, R.A. Palmer and C.J. Manning, Appl. Spectrosc., Transform Spectroscopy: 11th International Conference, ed.
51, 1245 (1997). J.A. de Haseth, American Institute of Physics, Woodbury,
NY, 439, Vol. 430 (1998).
49. H. Wang, R.A. Palmer, C.J. Manning and J.R. Schoonover,
in Fourier Transform Spectroscopy: 11th International Con- 60. G.L. Carr, Vib. Spectrosc., 19, 53 (1999).
ference, ed. J.A. de Haseth, American Institute of Physics, 61. C.J. Hirschmugl and G.P. Williams, Synchrotron Radiat.
Woodbury, NY, 555, Vol. 430 (1998). News, 8, 10 (1995).
50. J.P. Wang and M.A. El-Sayed, Biophys. J., 76, 2777 62. R.A. Palmer, G.D. Smith and P. Chen, Vib. Spectrosc., 19,
(1999). 131 (1999).
Instrumental Aspects of Time-resolved Spectra
Generated Using Step-scan Interferometers
Christoph Rodig and Friedrich Siebert
Albert-Ludwigs-Universitat, Freiburg, Germany
1 INTRODUCTION than the time needed for one scan) this separation of tem-
poral processes is easily performed and thus the rapid-scan
Time-resolved vibrational spectroscopy has developed into method is an effective technique of time-resolved FT-IR
an important research tool for the investigation of the prop- spectroscopy.5 7 Its extension to the so-called stroboscopic
erties of molecular and solid-state systems (see articles on or interleaved FT-IR technique8,9 requires precise timing
time-resolved infrared (IR) and Raman spectroscopy in this between the movement of the mirror and the triggering of
handbook, and also in the Proceedings of the conference the process to be studied. Since only a short part of the
series Time-resolved Vibrational Spectroscopy1 4 ). Nev- interferogram and of the time slice is captured during one
ertheless, owing to either the low radiation power of the scan, it is not very effective. In addition, owing to varia-
source and the relatively low detectivity of detectors [as tions in the speed of the mirror, this method is often prone to
in time-resolved IR spectroscopy] or to weak interaction artifacts,10 although time-resolved experiments with a time
cross-sections (as in time-resolved Raman spectroscopy), resolution of 50 s have been described.8,9 On the other
the total measuring time for covering a broad spectral range extreme of the separation techniques is the situation where
with high time resolution has been often very long, prompt- the measuring process of the interferogram is halted during
ing great efforts to improve the sensitivity. Therefore, it the temporal evolution of the system to be studied. This is
realized by the so-called step-scan method. The interferom-
was obvious to think of means to exploit the great advan-
eter must be capable of a precise stepwise movement of the
tages of Fourier transform (FT) spectroscopy, which already
mirror from one sampling position of the interferogram to
had revolutionized static IR spectroscopy, for time-resolved
the next, keeping its position constant over a period long
spectroscopy. The various realizations of time-resolved FT
enough to capture the process. Some basic principles of this
spectroscopy are described at greater detail in another arti-
technique, that are important for the understanding of this
cle in this volume see Fast Time-resolved Mid-infrared
article, will be recapitulated below.
Spectroscopy Using an Interferometer.
Since the first demonstration of its feasibility with com-
The application of the established method of rapid-scan
mercial step-scan instruments11,12 and the realization that the
Fourier transform infrared (FT-IR) spectroscopy to time-
main advantages of FT spectroscopy are retained,13 the num-
resolved studies involves a basic problem: one has to
ber of publications involving the time-resolved step-scan
disentangle the time dependence of the measuring process
technique has been increasing steadily. This trend has been
of the interferogram caused by the rapid movement of the
especially influenced by the availability of commercial soft-
interferometer mirror (the shortest time needed for one scan
ware supplementing the step-scan interferometers developed
being one to several milliseconds for commercial instru-
earlier. The investigations, which are briefly summarized
ments) from the time dependence of the process to be stud- below, cover a broad range in molecular and solid-state
ied. For slow processes (time constants considerably longer spectroscopy. Therefore, it appears appropriate and timely
to consider the influence of instrumental aspects of time-
John Wiley & Sons Ltd, 2002. resolved step-scan experiments on the quality of the spectra.
The applications range from the study of photoreac- emission studies, where the changes are measured essen-
tions of metal carbonyls,14,15 of organic compounds,16,17 tially against a zero background. However, spectral resolu-
and of the biological systems bacteriorhodopsin,13,18 22 tion, spectral range, and time resolution also determine the
halorhodopsin,23 photosynthetic reaction center,24 myo- experimental procedures and the requirements on the instru-
globin,25,26 hemoglobin,27 and cytochrome P-450.28 The ment. Furthermore, it is obvious that the precision of mirror
very small spectral changes in the biological samples positioning and stability will affect the results. Finally, the
(absorbance changes of the order of 104 and smaller dynamic range of digitization of the interferogram (change)
have to be resolved) require a very high sensitivity of and detector nonlinearity are, as in conventional FT spec-
the method. Further applications encompass measurements troscopy, additional factors that influence the quality of
of the switching behavior of ferroelectric crystals,29,30 of time-resolved step-scan FT spectra. In this article we dis-
impulse rheo-optics of polymer films,31 and of processes cuss, among others, these aspects in greater detail and
at the electrode surface of redox reactions.32 In addition describe, when possible, means to correct for failures or
to these time-resolved absorbance studies, time-resolved to circumvent some of the limitations of this technique.
emission studies have also used step-scan FT spectroscopy:
FT-Raman experiments,33,34 studies monitoring the fluo-
rescence and/or IR emission of excited gaseous molecules
2 PRINCIPLE OF TIME-RESOLVED
to obtain information on gas dynamics,35,36 temporal and STEP-SCAN FT MEASUREMENTS AND
spectral studies of the building up of laser pulses providing INSTRUMENTAL CONSIDERATIONS
information on very weak intracavity absorbers,37 photo-
luminescence experiments at defects of semiconductors,38 As mentioned in the Introduction, in time-resolved step-
time-resolved photothermal FT-IR experiments,39 and time- scan measurements the interferometer mirror moves step-
resolved photoacoustic FT-IR spectroscopic studies.40 The wise from one sampling position to the next. At each
last two techniques are mainly used for the study of the position, the process to be studied has to be triggered and
molecular composition of layered structures. the temporal evolution of the change of the interferogram
It is not surprising that the broad variety of experiments caused by the process is recorded (in the following, this
imposes specific requirements on the triggering of the pro- quantity is referred to as signal). Rearrangement of the
cesses and also on the performance of the step-scan instru- data of a complete scan provides a difference interferogram
ment itself. Time-resolved absorption measurements, in for each time slice covered by the signal acquisition (time
which small intensity changes have to be measured against t minus time zero). A 3D plot using the photolysis reaction
a large background, differ considerably from time-resolved of COmyoglobin25,26 as an example is shown in Figure 1.
30
20
10
Intensity change
0
10
20
30
40
50
60
1E2
1E3 600
1E4 500
400
1E5 300 ts
Tim
e (s 1E6 200 poin
am
) 100 rogr
1E7 0
Interfe
Figure 1. Three-dimensional representation of the interferogram changes induced by the photoreaction of the photolysis of
COmyoglobin. Interferogram points represent the various mirror positions, point zero being 35 positions before the zero optical
path difference (ZPD) position.
Time-resolved Spectra Generated Using Step-scan Interferometers 3
The Fourier transformation yields the corresponding spec- Another way to obtain the zero crossings without the
tral changes. From this outline it is clear that the process superimposition of oscillations on the mirror is to subtract
under study has to be repeated at least once at each sam- the average laser intensity from the direct current (DC)
pling position of the interferogram. Since often the signals coupled laser interferogram. There are different practical
are weak and noisy, signal averaging is needed, resulting solutions depending on the manufacturer. The influence
in a total number of repetitions of multiples of sampling of the various methods and the long-term stability of
points. This requires that the processes can be triggered as the HeNe laser on the precision (not stability) of the
many times. In addition, the time between successive sig- positioning has not been studied systematically.
nals should not be too long to avoid excessive measuring When the mirror is fixed, the laser signal provides
times. Hence, there are some limitations on the systems a convenient way to control precisely for its stability.
which can be studied with the step-scan technique. How- This is shown in Figure 2, where a typical laser signal
ever, at the end, we will present some means to overcome from a Bruker IFS-88 interferometer incorporated into our
these limitations. Nevertheless, since the processes have to laboratory-built vacuum optical bench is shown.42 The sig-
be repeated many times, it appears important to investigate nal fluctuations point to some residual variations in the
the consequences of fluctuations of the corresponding detec- optical path difference. Because of the vacuum bench, fluc-
tor signals. This will be an important issue described later. tuations of the refractive index of air can be neglected,
It is important to emphasize that the decoupling of record- which have been shown to be potentially a source of path-
difference variations in nonvacuum instruments.43 There-
ing of the time-dependent interferogram changes from the
fore, the signal fluctuations arise from mirror instabilities.
movement of the interferometer mirror has a dramatic effect
From the amplitudes of the signal (150 mVpp ) and the
on the achievable time resolution of the measurement. No
laser interferogram (12 Vpp ), the mirror fluctuations can be
longer does the scan speed of the interferometer limit the
estimated to 1.3 nmpp in this case. The influence of such
time resolution, but only the rise time of the detector and
fluctuations on the quality of time-resolved spectra will be
the bandwidth of the electronics.
discussed below.
In all instruments, the stepwise movement of the interfer-
In time-resolved emission spectroscopy, the static light
ometer mirror is controlled electronically by monitoring the source (globar, halogen lamp, etc.) is replaced by a pulsed
sinusoidal interferogram of an HeNe laser. The mirror is source, and at each sampling position the pulsed emission is
usually positioned at the corresponding zero crossings. The triggered. The corresponding signal generated in the detec-
number of positions and their spacing are, as in rapid-scan tor/preamplifier is measured with a transient recorder and
FT spectroscopy, determined by the required spectral range transferred to a computer. In order to determine the proper
and resolution41 (see also Continuous Scanning Inter- scan range of the mirror, a short rapid-scan or step-scan
ferometers for Mid-infrared Spectrometry). Here, it is
important to note that at the zero crossings the intensity 0.30
is not zero. Therefore, one has to provide means to sepa-
rate the sinusoidal modulation from the average intensity. 0.25
In rapid-scan interferometers this is easily accomplished
by alternating current (AC) coupling. However, this is not 0.20
Amplitude (V)
feasable for step-scan interferometers, since an infinitely

long AC coupling time constant would be needed. 0.15
One way to circumvent this problem is to superimpose
a small oscillation of constant frequency on the mirror
0.10
movement, which causes a corresponding oscillation of the
HeNe laser interferogram. Its amplitude is maximum at
0.05
the zero crossings and nearly vanishes at the extremes.
Since the frequency is fixed, the signal can now be AC
coupled and used for mirror control. We shall show below 0.00
100 120 140 160 180 200
that these mirror oscillations have negative consequences
Time (ms)
for time-resolved measurements, even though they are used
for static methods of step-scan FT-IR spectroscopy which Figure 2. Typical intensity fluctuations of the HeNe laser signal
used for positioning the movable mirror in the step-scan mode at
are generally described as phase-modulation techniques a fixed position. [Reproduced with permission from the Society
(see Instrumentation for Step-scan FT-IR Modulation for Applied Spectroscopy from C. Rodig and F. Siebert, Appl.
Spectrometry). Spectrosc., 53, 893901 (1999).]
measurement (depending on the software) with a DC source that either a second digitization channel has to be present or
is performed before the actual time-resolved measurements, the interferogram before the trigger and its time-dependent
providing the necessary information on the center burst changes after the trigger have to be measured by the same
(i.e. ZPD) position. After reshuffling of the data in the analog-to-digital (AD) converter. Usually, the changes are
computer into interferograms for each time slice of inter- much smaller, and the dynamic range of digitization will
est, the FT provides the corresponding emission spectrum. become an important issue (see below).
Since intensities are measured, the usual procedures for However, there may be situations where the simultaneous
phase correction can be applied. Therefore, also essentially digitization is difficult to realize. If so, the interferogram
single-sided interferograms can be evaluated.41 This may changes have to be recorded in a second scan and the step-
be necessary for high-resolution investigations (requiring scan interferometer has to be able to repeat the scan at
maximum optical path difference) or for the study of sys- exactly the same sampling positions. This capability of the
tems requiring a short total measuring time. However, it interferometer is also required when the results of several
should be emphasized that by using a double-sided inter- scans should be averaged before the FT is applied as in
ferogram, the signal-to-noise ratio (S/N) is increased by a rapid-scan FT spectroscopy.
1
factor of 2 2 . A method for direct phase determination from a double-
The measuring principle differs for time-resolved absorp- sided difference interferogram has been described.46 How-
tion measurements. Again, the temporal development has ever, this method appears to be susceptible to noise, which
to be measured of interferogram changes that are now is much larger in the difference interferogram compared
caused by the absorbance changes of the sample. The FT with that of the single-beam spectrum.
yields intensity changes which can be positive or nega- The number of sampling points and, hence, the number
tive, depending on the sign of the absorbance. Hence the of sample excitations depend on the spectral resolution and
usual Mertz method for phase correction41,44 cannot be the spectral frequency range. With a spectral range from 0
applied. A straightforward way to solve this problem has to 1975 cm1 and a spectral resolution of 7.7 cm1 , 512
been derived from the related problem in polarization mod- sampling points are needed for a single-sided interferogram
ulation FT-IR spectroscopy, for studies of both linear and (50 more sampling points beyond the ZPD are required
circular dichroism.45 Also here, the dichroic difference, i.e. for phase correction).
the FT of the corresponding interferogram, can have a pos- In order to avoid errors due to folding back of intensity,
itive or negative sign. In those experiments, a true intensity an optical low-pass filter has to be inserted which blocks
is obtained by FT of the interferogram from a single-beam intensity beyond the upper frequency limit determined by
spectrum of the sample, which is needed anyway for fur- the Nyquist theorem. If the low-frequency part of the
ther treatment of the data. Then, the Mertz phase correction spectrum is not of interest (in our example below 1975/2 D
method can be applied. The resulting phase spectrum can 987.5 cm1 ), the number of sampling points can be further
be used to perform the FT of the interferogram, yielding the reduced by undersampling of the interferogram (in this case
dichroic intensity changes with the correct signs. In order by a factor of 2). Again, the intensity has to be blocked by
to be able to use the same phase spectrum, it is necessary to an optical (high-pass) filter. Inserting optical filters has two
record the interferograms corresponding to the single-beam advantages for the accuracy of the spectra: (1) the dynamic
spectrum and the dichroic difference at exactly the same range of digitization of the difference interferogram is
sampling positions. increased, which is especially important if small spectral
The extension to time-resolved absorption spectroscopy changes, as in biological systems, have to be detected (see
is now clear.13 Also here, the single beam spectrum S0 n section on signal acquisition); (2) since the intensity on
of the sample before the process has been started has to be the detector is reduced, problems arising from nonlinear
measured, since the desired absorbance change At, n is behavior of the detector are diminished (see Section 4).
calculated from the intensity changes St, n and from the
single beam spectrum S0 n by the relation 3 QUANTITIES TO BE MEASURED

S0 n C St, n
At, n D log 1 In all time-resolved step-scan FT measurements, the quan-
S0 n
tity to be evaluated is the time course of the change
Again, the interferogram of the single-beam spectrum of of the interferogram. However, as in conventional FT
the sample has to be measured in the step-scan mode at spectroscopy where the detector sees both the intensity
exactly the same sampling positions as the interferogram modulated by the interferometer (corresponding to the inter-
changes are recorded. A convenient way to achieve this is ferogram) and the integrated intensity, the signal generated
to record both quantities within a single scan.13 This means at the detector in time-resolved experiments comprises
the actual change of the interferogram and the change compensate for the static integrated intensity. This approach
of the integrated intensity. Therefore, the question of the has two disadvantages: (1) the 16-bit AD converter limits
role of the dynamic range of digitization of the signal the time resolution (digitization rate 200 kHz1 MHz);
becomes important. In continuous-scanning FT-IR spec- (2) the time-dependent signal itself is digitized with a low
troscopy, gain switching is commonly used to increase resolution since the 16 bits are used for both the static value
the dynamic range. Since the integrated intensity, being a and the time-dependent changes. This limitation is espe-
constant quantity, can be removed from the actual inter- cially severe if the the time-dependent changes are very
ferogram by AC coupling, those parts beyond a certain small, as is the case for the photoreaction of biological sys-
selectable optical path difference showing smaller inten- tems, where the change of a single molecular group (amino
sity variations can be additionally amplified. This is not acid side chain) has to be detected against the background
possible in time-resolved step-scan experiments with com- of hundreds of other groups.
mercial instruments. The only experiment taking advantage In order to circumvent both disadvantages, the time-
of this procedure described in the literature uses a specially dependent signal is separated electronically from the DC
designed Connes-type interferometer and two detectors, one value. This allows the additional amplification of the signal
in each arm of the interferometer.36 Since ideally the mod- to exploit the full digitization range of the AD converter,
ulated intensity in the two arms is 180 out-of-phase, the which can, in addition, have a low resolution. Even if
subtraction of the signals from the two detectors provides it is only 8 bits, the effective resolution is still a factor
the modulated signal, the change of the integrated inten- of 2550 higher than with the 16-bit digitization of the
sity being cancelled. This modulated signal can be further static value plus the time-dependent change. The tolera-
amplified using gain switching to exploit the full range ble lower resolution allows much higher digitization rates
of digitization. There may be some deviations from this (200 MHz1 GHz for transient recorder boards fitting into
theoretical behavior, which, however, can be compensated a PC). However, it has been reported that increasing the
by formation of a weighted difference of the two detec- resolution to 12 bits increases the S/N in the time-resolved
tor signals. As in static FT-IR spectroscopy, the problem spectra.47
of limited dynamic range is especially critical in high- The easiest way to separate the signal from the DC output
resolution applications for which this special instrument has of the preamplifier is by AC coupling. This is the common
been developed. way chosen by the manufacturers. However, the roll-off of
In all commercial instruments, only one detector is AC coupling limits the size of the covered time range. For
installed, and therefore this method of canceling the inte- many photochemical applications this may not be critical.
grated intensity change cannot be applied. Depending on its However, the reactions of many photobiological systems
size and on the size of the measured signals, this makes it extend into the 100-ms time range. Since these reactions
even more mandatory to use the highest digitization reso- also require high time resolution, the input resistance of all
lution compatible with the time resolution. In addition, for the electronic components has to be low, usually matched
light-induced reactions the signal seen by the detector con- to the 50
of the coaxial cable. In turn, such a low resis-
tains a thermal contribution not discussed so far, the size of tance makes it difficult to realize an AC coupling with a
which depends on the special experiment. Its contribution time constant even of only 1 ms because the capacitor can-
further reduces the dynamic range available for the differ- not be of electrolyte type and must have a low inductance.
ence interferogram. Since both the thermal and integrated Consequently, the measured signals are distorted. In order
intensity signals are not modulated by the mirror position, to overcome this limitation we developed a special ampli-
they do not contribute to the time-resolved spectrum. How- fier that can be switched from AC coupling to quasi-DC
ever, they have drastic effects on the noise performance, as coupling with compensation of the DC value.13 This is real-
will be shown below. ized by an integrating feedback stage causing an effective
The intensity at the detector is the sum of the static AC coupling (see Figure 3). If the input to the integrat-
intensity and of the time-dependent signal. Only the last ing stage is disconnected by a computer-controlled switch,
quantity contains the desired information on the change of the voltage across the integrating capacitor stays essentially
the interferogram and, therefore, has to be separated. For constant over a period of several seconds (depending on the
this, two strategies have been developed. In one implemen- electronic components), providing the compensation of the
tation, the high accuracy of the 16-bit AD converter which DC value of the detector signal. The resulting signal, reflect-
digitizes the interferogram in rapid-scan experiments is used ing only the change of the interferogram, can be amplified
to record simultaneously the static value of the interfero- and the time range is limited only by the droop of the
gram and the time-dependent changes. The offset of the capacitor voltage. We have shown that this method can be
DC-coupled preamplifier has to be adjusted in order to also combined with a 100-MHz amplifier48 and the effect
further manipulations such as signal averaging. With mod-

ern transient recorder boards, Pentium III computers, and
U IN
PCI-bus architecture, this should not impose any difficul-
+ U OUT
+ ties. However, further processing of the data (such as FT,
C
visualization of time traces, and curve fitting) could become
difficult with such large data files. Therefore, in most cases
AC/DC
R it is desirable to compress the data without causing an

essential loss of information. For this purpose, one can
+ Offset take advantage of the fact that the time evolution of most
reactions is homogeneous, i.e. the time dependence slows
Figure 3. Schematic diagram of the compensating quasi-DC down as the reaction proceeds. This allows the implemen-
amplifier. [Reproduced with permission from the American Insti-
tute of Physics from C. Rodig and F. Siebert, in Fourier Trans- tation of a logarithmic time base accounting exactly for this
form Spectroscopy, 11th International Conference, ed. J.A. de behavior. For this purpose, a hard-wired logarithmic tran-
Haseth, 388391 (1998).] sient recorder was first developed.49 Nowadays it is much
easier to implement the logarithmic time base by software,
and usually a quasi-logarithmic time base is generated. The
0.000
time range is divided into several sectors, each containing x
linearly spaced data points (x in the range 812). The first
1201 cm1 sector contains the data as they have been obtained by the
0.005 transient recorder. In the second sector, two successive data
points are averaged, in the third sector four successive data
points are averaged, and so on. In addition to the extensive
Absorbance change
compression of the data, the averaging also provides an

0.010
inherent smoothing filter, adapting the effective electronic
bandwidth to the time dependence of the process, i.e. by
1529 cm1 this treatment of the data the S/N increases towards longer
0.015 times. Nevertheless, the time resolution at the beginning of
the process is, in contrast to a pure logarithmic timescale,
fully retained.
AC coupling
Sometimes, depending on the hardware components
0.020
DC coupling (computer and transient recorder), it may be advantageous
to operate two transient recorder boards simultaneously to
cover a broader time range.22,42 The second board then
0.025 operates at a reduced digitization rate. In order to cope with
1 10 100 this reduced rate, the electronic bandwidth of the signal has
Time (s) to be reduced accordingly with an electronic filter.
Figure 4. Distortions of the time courses caused by AC coupling
of the process-induced signal. The traces represent the absorbance
changes at 1529 and 1201 cm1 of the bacteriorhodopsin photo- 4 ASPECTS OF DETECTOR
cycle (see Figure 10). The roll-off time constant is 1 ms.
PERFORMANCE
of the roll-off has been demonstrated in the case of the In addition to the signal acquisition electronics, the detector
photobiological system bacteriorhodopsin (Figure 4). in combination with its preamplifier determines the attain-
Another aspect of data acquisition relates to the size able time resolution. Whereas photoconductive mercury
of computer memory and disk space required for a mea- cadmium telluride (MCT) (HgCdTe) detectors typically
surement. Covering a broad time range with high time have an intrinsic rise time of 500 ns, photovoltaic MCT
resolution leads to large data files: data collection in the or HgMnTe detectors are much faster. The effective time
spectral range of 1000 cm1 with a spectral resolution of constant is determined by the capacity of the detector ele-
8 cm1 and a time range of 10 ms digitized with 200 MHz ment together with the value of the transimpedance resistor
results in 5 108 double-precision data. Consequently, the of the preamplifier. Thus, reducing the detector area by a
transient recorder has to provide enough memory and the factor of two lowers the rise time by the same factor. A typ-
speed of the data transfer to the computer should not slow ical value for a 1-mm2 detector element is 30 ns. All these
detectors operate at liquid N2 temperature and their detec- also reduced. Otherwise, the S/N becomes worse, because
tivity is between 1012 and 1013 Jones. In addition to the the noise of the detector element is proportional only to
faster response, the photovoltaic detectors have the advan- the linear dimension, whereas the signal is proportional
tage of a lower frequency limit where the 1/f noise starts to the area. This adds to the increased noise due to the
to dominate and of considerably higher tolerable radiation larger electronic bandwidth, and both contributions make
intensity before detector nonlinearities become perceivable. measurements at higher time resolution more difficult.
The consequences of detector nonlinearity for time- Basically, there are two solutions to overcome these lim-
resolved step-scan FT-IR spectroscopy have been demon- itations. Either a brighter source is used (the extreme being
strated using a photoconductive MCT detector.48 As shown a synchrotron source) or the detector area is subdivided into
in Figure 5, the nonlinearity causes a time-dependent several smaller elements, each being connected to its own
change of the baseline. In addition, band intensities are preamplifier and the respective outputs being added elec-
affected at spectral positions where the transmittance is low. tronically. Although the latter solution is not yet available
This was shown by the implementation of an extended soft- commercially, it would have the advantage of not overload-
ware correction method for detector nonlinearity48 based on ing the detector. If a brighter source is used instead, even
a procedure developed for rapid-scan FT-IR spectroscopy.50 the photovoltaic detector may exhibit appreciable nonlinear
For photovoltaic detectors such a correction is hardly behavior, and the application of software correction meth-
needed. ods may become necessary.
The size of the sensitive area of the detector is an
important quantity influencing both the attainable S/N and
the rise time of the detector. Because the capacity of the 5 SIGNAL FLUCTUATIONS
detector element is proportional to its area, the only way
to reduce the response time is to reduce the area. Usually As explained in the section on the principles of time-
the image of the IR source is larger than the area of the resolved step-scan FT spectroscopy, the process has to
detector element, and therefore it is possible to decrease the be triggered at least once at each sampling point of the
size of the detector only if the image of the light source is interferogram. Here, it is tacitly assumed that the process
0.004
can be repeated with high precision, since the FT detects
the correlation between the signals obtained at the different
A mirror positions. However, there are many possible causes
B for variations in the extent of the triggered processes. In the
C case of photoinduced reactions, the energy of the exciting
0.002
laser may vary or the sample may decompose. In stopped-
flow experiments the concentration may vary from shot to
Absorption change
shot. Hence it is interesting to ask how this violation of

correlation will influence the spectra.
0.000 For this, it is first necessary to analyze more carefully
the composition of the measured signals. As mentioned
above, they contain the changes of the integrated inten-
sity and of the intensity modulated by the interferometer.
0.002 In emission studies, the first part represents the total inten-
sity emitted, whereas in absorption studies it represents the
intensity change caused by the absorption changes, inte-
grated over the covered spectral range. Therefore, if no
0.004
correction is applied, the difference interferogram does not
1800 1600 1400 1200 1000 oscillate around zero but around the integrated intensity
Wavenumber / cm1 change. However, in time-resolved IR absorption studies
Figure 5. Effect of detector nonlinearity on the time-resolved an additional contribution may be present, depending on
difference spectrum of the photolysis of COmyoglobin in 2 H2 O. the sample excitation. If some energy is deposited such
Shown is a spectrum obtained 1.4 s after the flash. A, spectrum as in photoreactions, part of this energy is converted into
without; B, spectrum with nonlinearity correction; C, difference. heat, and the corresponding change in sample temperature
[Reproduced with permission from the American Institute of
Physics from C. Rodig and F. Siebert, in Fourier Transform is seen by the detector. With the reaction of the photo-
Spectroscopy, 11th International Conference, ed. J.A. de Haseth, biological system bacteriorhodopsin this effect has been
388391 (1998).] demonstrated.42 The heat signal has been measured by
140 200
120 A
150
Detector signal (arbitrary units)

100
B
100
80
60
50
40 A
B
C 0
20
0.1 1 10 100 1000

0.1 1 10 100 1000
(a) Time (s)
Time (s)
Figure 6. Composition of the time-dependent detector signal: 200
A, signal of the bacteriorhodopsin photocycle at fixed mirror
position; B, thermal signal with blocked IR beam; C, difference
between the signals. [Reproduced with permission from the Soci-
ety for Applied Spectroscopy from C. Rodig and F. Siebert, Appl.
Spectrosc., 53, 893901 (1999).] 150
blocking the IR-measuring beam (Figure 6). By monitoring B

the temperature dependence of the water band between
3700 and 3200 cm1 , a temperature increase of about 1 C 100
could be deduced. It is remarkable that in the early time A
range the heat signal is of similar size to the measured
signal. As shown in Figure 6, the spectral signal, which
is obtained by subtraction of the thermal signal from the 50
measured signal, is first negative. We could show that its
time course can be explained by the evolution of the inte-
grated intensity change: at the beginning, it is very small
and the signal has a considerable contribution from the 0
modulated intensity, whereas at later times the integrated 0 100 200 300 400 500 600
intensity change increases, now dominating the spectral (b) Data point
signal. Figure 7 clearly confirms that the difference inter- Figure 7. Demonstration that the difference interferogram does
ferogram is only part of the measured signal. In Figure 7(a) not oscillate around zero. (a) Time-dependent detector signals dur-
two signals are shown, trace A near the ZPD and trace B ing a step-scan measurement of the bacteriorhodopsin photocycle:
at a large optical path difference. This shows that the A, sampling position near ZPD; B, sampling position at large path
modulated intensity is only a fraction of the total signal. difference. (b) Difference interferogram obtained at (A) 1.4 and
(B) 165 s after photolysis, obtained from the complete data set
In Figure 7(b), two difference interferorgams without off- of which two signals are shown in (a). [Reproduced with permis-
set correction are shown for time slices of (A) 1.4 and sion from the Society for Applied Spectroscopy from C. Rodig
(B) 165 s after the flash. In both cases the interferograms and F. Siebert, Appl. Spectrosc., 53, 893901 (1999).]
do not oscillate around zero. In time-resolved emission
studies from the gas phase, the heat signal is essentially Figure 8, trace A, shows the spectrum as obtained directly
absent. from the measurements at 70 s after the flash. In the
We have simulated the effect of signal fluctuations for first simulation (Figure 8, trace B), additional fluctuations
measurements of the photoreaction of bacteriorhodopsin.42 of the amplitudes of the difference interferogram were
proportional to its energy which fluctuates in the same way

0.010 as the laser pulse, the noise in the spectrum will rather
increase.
The only way to compensate for signal fluctuations
0.005 A
is to monitor the extent of the triggered processes. In
photochemical/photobiological reactions this can be done
Absorbance change
0.000 by measuring the laser pulse energy simultaneously and

B correcting the measured signals accordingly. This cor-
rection method has been implemented13 and is to some
0.005 extent also available in commercial step-scan software.
Note that this correction procedure can only be applied
C if the extent of excitation depends linearly on the laser
0.010
energy.
The extent of the reaction can also be monitored by
0.015 other means, e.g. by a signal in the ultraviolet (UV)visible
range if the process is connected to a corresponding spectral
0.020
change. In time-resolved emission studies, the extent of
1800 1600 1400 1200 1000 the triggered process has been measured using the Connes
Wavenumber /cm1 type of interferometer (see above): the sum of the two
detectors represents the integrated intensity change, and
Figure 8. Simulation of the effect of signal fluctuations. A, time-
resolved step-scan difference spectrum of the bacteriorhodopsin thus provides a corresponding normalization factor for the
photocycle; B, difference spectrum obtained after adding 5% modulated signal.36 Since this modulated signal is directly
fluctuation noise to the difference interferogram corresponding obtained as the difference between the two detector signals,
to spectrum A; C, difference spectrum obtained by adding 5% the spectra are less susceptible to fluctuations (see above).
fluctuation noise to the detector signals of the same measure- Nevertheless, in this case the correction proved to be
ment. [Reproduced with permission from the Society for Applied
Spectroscopy from C. Rodig and F. Siebert, Appl. Spectrosc., 53, necessary.
893901 (1999).] Irrespective of the underlying correction principle, the
measurement has to be precise enough to clearly resolve
introduced by multiplying by a factor [1 C q 1rand the fluctuations. In our studies of bacteriorhodopsin, a
rand/max], where rand is a random number chosen by frequency-doubled Nd : YAG laser was used for excitation.
the software and max is the maximum of the possible Its pulse to pulse energy variation is 9%. In the measure-
random numbers. With q D 0.05, there are hardly any ments shown in Figure 8, eight signals have been averaged
distortions observable (Figure 8, trace B). However, in a per sampling position, reducing the effective fluctuation of
real experiment this situation would apply only if there was the averaged signal to about 3.2%. In this case, a resolution
no integrated intensity change and no heat signal. A more of 810 bits would be sufficient.
realistic picture emerges if not the difference interferogram, As mentioned in the Introduction, time-resolved step-
but the measured signal is multiplied by this factor. The scan FT-IR experiments are also used to study reorienta-
corresponding spectrum is shown in Figure 8 (trace C), tional effects induced by an electric field. Here, it can be
where large distortions are present. Since at mirror positions assumed that the process is highly reproducible. In addition,
far from the ZPD the fluctuations are much larger than thermal effects are negligible. This kind of experiment can
the amplitudes of the difference interferogram, they cause therefore be performed with high precision.29,30
more noise in the spectrum. This noise is equivalent to the
multiplicative noise encountered in normal FT spectroscopy
if the detector is shot-noise limited or if the source shows 6 MIRROR FLUCTUATIONS AND
fluctuations.41 REFRACTIVE INDEX FLUCTUATIONS
The performance of time-resolved step-scan FT-IR exper-
iments has to account for the composition and fluctuations In time-resolved step-scan FT experiments it is assumed
of the measured signals. One could think of compen- that the mirror position is precise and stays constant during
sating laser fluctuations by saturating the photoreaction the acquisition of the signals at each sampling position.
by an increased laser pulse energy. The spectral sig- Since neither requirement is met perfectly in practice, the
nals then would be almost independent of such fluctua- consequences for the quality of the time-resolved spectra
tions. However, since the laser pulse causes a heat signal will be discussed in this section.
In general, it would be more appropriate to discuss 475

optical path difference stability instead of mirror position
stability. In nonvacuum instruments, fluctuations of the 450
#1
refractive index of air in the interferometer, due to ther-

mal inhomogeneities (caused by the heated source and the 425
# 30
flow of the purge gas), can contribute considerably to opti-
cal path-difference variations.43 However, when studying 400
the effect of optical path-difference fluctuations in a vac-
uum instrument, only mirror instabilities have to be taken # 35 (ZPD)
375
into account. In addition, from our experience, the mirror
stability is better in vacuum. As has been mentioned above, 350
these fluctuations amount to 1.3 nmpp . # 40
Mirror fluctuations have two basic effects: (1) they cause 325
errors in the spacing of the sampling positions, resulting in # 100
increased spectral noise,41 and (2) they cause fluctuations 300
of the radiation intensity at the detector, and therefore time- # 500
varying fluctuations of the signal, superposing the spectral
275
signal induced by the studied process. 0.01 0.1 1 10 100
The size of the signal fluctuations depends on the slope
Time (ms)
of the static interferogram and is largest close to the ZPD
Figure 9. Time-dependent detector signals at different mirror
position. This is shown in Figure 9: IR signals at different positions during a step-scan measurement. At each mirror position
mirror positions were recorded without triggering a process. eight signals were averaged without exciting the sample. [Repro-
Eight signals were averaged, which is typical for our time- duced with permission from the Society for Applied Spectroscopy
resolved measurements of photobiological systems. The from C. Rodig and F. Siebert, Appl. Spectrosc., 53, 893901
periodic feature far from ZPD position is independent of the (1999).]
presence of IR radiation and is due to line frequency pick-
up. Hence neither this contribution nor the fluctuations due the photoreaction of bacteriorhodopsin, the protein sam-
to mirror instability are averaged out. The spectral conse- ple has been inserted into the step-scan instrument. This
quences of these fluctuations are shown in Figure 10. Here, causes very low transmission around 1650 cm1 (amide I)
in order to compare the results with real measurements of and 1550 cm1 (amide II). The time-varying fluctuations
3
10
1.0
Transient absorption
0.5
0.0
0.5
1.0
1.5
2.0
0.20
0.15
1000
0.10 1200
Tim
e (s 0.05 1400
) 1
1600
er / cm
0.00 1800 numb
Wave
Figure 10. Effect of residual mirror fluctuations: baseline spectra of a time-resolved step-scan measurement with 100-ns time base. At
each mirror position eight signals were averaged without exciting the sample placed into the IR beam. [Reproduced with permission
from the Society for Applied Spectroscopy from C. Rodig and F. Siebert, Appl. Spectrosc., 53, 893901 (1999).]
are clearly visible and, as expected, they are larger in where the process has almost decayed to zero and where
regions of low transmission. It has been shown that the the noise is strongly reduced owing to the quasi-logarithmic
effect of line frequency pick-up has similar consequences time base (see above). However, they may influence a
and periodic distortions are clearly visible. precise kinetic evaluation of the data.
These fluctuations in the time course are also present In order to demonstrate more clearly the effect of mir-
in the real measurements of bacteriorhodopsin (Figure 11), ror instability, a mirror oscillation of 8 nmpp at a fre-
although they are much smaller than the time-dependent quency of 50 Hz was superimposed. The effect on the
spectral changes and become visible only in a time range time-resolved spectra is shown in Figure 12. In the time
5
Absorption change (103)
10
15
20
1
1E2 1000
1200
1E4
Tim 1400
e (s 1
) 1E6 1600
er / cm
1800 numb
Wave
Figure 11. Time-resolved step-scan measurements of the bacteriorhodopsin photocycle at pH 7 and 298 K sampled with 100-ns time
base. At each mirror position eight signals were averaged. [Reproduced with permission from the Society for Applied Spectroscopy
from C. Rodig and F. Siebert, Appl. Spectrosc., 53, 893901 (1999).]
103
5
Transient absorption
10
15
20
1
1E2 1000
1200
1E4
Tim 1400
e (s 1
) 1E6 1600
e r / cm
1800 numb
Wave
Figure 12. Effect of interferometer mirror modulation on time-resolved step-scan measurement of the bacteriorhodopsin photocycle at
pH 7 and 298 K sampled with 100-ns time base. At each mirror position eight signals were averaged. An oscillation of the moving
mirror around its position by 4 nm with a frequency of 50 Hz during the measurement was artificially introduced. [Reproduced with
permission from the Society for Applied Spectroscopy from C. Rodig and F. Siebert, Appl. Spectrosc., 53, 893901 (1999).]
range beyond 1 ms, distortions can clearly be distinguished. For the study of slow-cycling systems, a sample wheel
Hence increasing the fluctuation amplitude of the mirror by can be used that allows one to monitor the fast phase of
a factor of 6 causes intolerable distortions in the time traces. such reactions. Before the reaction cycle is completed a
It is remarkable that again the distortions are not cancelled new sample is brought into the IR beam by turning the
by signal averaging. wheel. Thereby the repetition rate of sample excitation can
When analyzing the spectra at different times, one be increased.52,53 If the spectral changes are small, the
observes that the effect of mirror instability causes spec- absorbance changes can be linearized (A I/I0 ) and
tral fluctuations with a weak wavenumber dependence (of signal averaging can be performed on the interferogram
course, they are increased by the low transmission at level. In this case, the time-resolved difference spectrum
the amide I/II bands). This weak spectral dependence is has not to be calculated separately for each sample on the
explained by the fact that the sampling positions around wheel and the amount of data is reduced considerably.
the ZPD position are mainly affected. These investigations For irreversible reactions, different approaches must be
clearly demonstrate that introducing mirror oscillations in applied. For the study of the photoreaction of liquid sys-
order to position the mirror causes artifacts in the time tems, a flow system has been used.17 The flow rate has to be
traces. If the spectral changes are large (of the same order adjusted in order not to distort the time range of interest but
as the absorption bands themselves), they can be in some
allowing a reasonable excitation frequency. If only a small
cases neglected.15,29,30,51 For other applications, a method
fraction of the solute is excited by each flash, the solu-
has been implemented to evaluate data only at the zero
tion can be recycled to the reservoir. If a large percentage
crossings of the oscillations. This, however, limits the appli-
undergoes a photoreaction, the solution exiting the cuvette
cability of the method.39
has to be discarded. In any case, enough sample material
We could show that variations in the spacing of the
must be available to allow for acquisition of the difference
sampling positions caused by the mirror fluctuations of less
than 8 nmpp do not cause noise in the difference spectra. interferogram including signal averaging. This amount may
This could be deduced from the early microsecond part of be reduced if microilluminators and microscopes are used,
the data shown in Figure 12 which are very similar to the reducing the sample area to less than 1 mm2 and the vol-
data obtained without mirror oscillation (Figure 10). ume correspondingly. If also relatively slow reactions are
Hence, provided that the mirror oscillation amplitude to be measured, a method of pulsed flow may be more
is kept small and the frequency is relatively low, time- appropriate. Here, the excited sample is just replaced by
resolved spectra can be obtained with little distortion for a a new one and during the signal acquisition no flow is
time range considerably smaller than the oscillation period. present. Another approach uses a very small fraction (less
The same applies to time-resolved emission experiments than 0.1%) of sample excitation, allowing several thousand
from a gas phase. Since the spectral changes are essentially flashes to be applied.16 This method requires the trans-
measured against a zero background, the mirror oscilla- mission of the sample to be essentially constant, i.e. the
tions only influence the signal caused by the emission and transmission is only little determined by the components of
not the intensity due to the background radiation. Hence the reaction.
such measurements essentially correspond to time-resolved Finally, for systems which cannot be passed through
absorption measurements with large absorption changes. the cuvette but from which relatively large areas (around
However, in time-resolved emission studies under the influ- 4 cm2 ) can be produced, a scanning IR microscope has
ence of background radiation, corrections for the mirror been used.54 In principle, the microscope steps to a new
oscillations have to be applied.39 position after each flash where a fresh sample spot can
be excited. Of course, the sample may cause some light
scattering, increasing the bleaching beyond the microscope
7 STUDIES OF SLOW-CYCLING focusing area. Therefore, the step size should be larger
SYSTEMS AND IRREVERSIBLE than the diameter of the focus. If the area is reduced to
REACTIONS 0.1 mm diameter by the microscope and the stepping size
is chosen to be 0.2 mm, enough flashes can still be applied
As we explained in the section on the principles of if the total sample area amounts to 4 cm2 . Of course, if
the method, the process under study has to be triggered microscopic visualization is not necessary, instead of a
many times. Therefore, slow-cycling reactions (cycling time complete microscope, a microilluminator consisting of two
beyond 1 s), which limit the frequency of sample excita- microscope objectives and a scanning device may be more
tion, and irreversible reactions are difficult to study with convenient. It should be noted that the S/N is considerably
the step-scan method. reduced by the use of microilluminators or microscopes.
In all experiments where a new sample is brought into ABBREVIATIONS AND ACRONYMS
the monitoring beam, signal fluctuations caused by con-
centration variations are critical. Especially for nonliquid AC Alternating Current
systems such as biological membranes they will be fairly AD Analog-to-digital
large and correction procedures have to be developed. For DC Direct Current
the scanning microscope system, it should be possible to FT Fourier Transform
scan the sample area first at the positions where it will be MCT Mercury Cadmium Telluride
excited to record the absorption spectra. These can be used ZPD Zero Optical Path Difference
to determine the amount of excited material and to correct
the flash-induced signals.
REFERENCES
8 SUMMARY 1. A.W. Parker (ed.), Proceedings of the 8th Conference on

Time-resolved Vibrational Spectroscopy, April 711, 1997,
Despite the many instrumental causes of spectral distor- Laser Chem., 19(14) (1999).
tions discussed in this article, it is important to mention 2. R.B. Dyer, M.A. Martinez, A.P. Shreve and W.H. Woodruff
that for practical applications none of them represents a (eds), Proceedings of the 7th International Conference on
Time-resolved Vibrational Spectroscopy, June 1116 1995,
major limitation. By carefully designing the experiment,
Los Alamos National Laboratory, Los Alamos, NM
the quality of the spectra in terms of S/N and spectral (1997).
accuracy is increased, causing shorter measuring times.
3. A. Lau, F. Siebert and W. Werncke (eds), Time-resolved
The high time resolution, the great sensitivity, and the Vibrational Spectroscopy VI, Springer, Berlin (1994).
coverage of a broad spectral range make it a very effec-
4. H. Takahashi (ed.), Time-resolved Vibrational Spectroscopy
tive method to study molecular processes, and a broader V, Springer, Berlin (1992).
range of applications therefore becomes accessible. Our
5. M.S. Braiman, P.L. Ahl and K.J. Rothschild, Proc. Natl.
discussions indicate that in order to exploit all the capa- Acad. Sci. USA, 84, 5221 (1987).
bilities of time-resolved step-scan FT spectroscopy, deeper
6. K. Gerwert, G. Souvignier and B. Hess, Proc. Natl. Acad. Sci.
insights into the principles of FT spectroscopy are nec- USA, 87, 9774 (1990).
essary. Also, some basic knowledge of signal acquisition
7. R. Krautle, W. Gartner, U.M. Ganter, C. Longstaff, R.R.
(amplifier design, AD resolution, etc.) are useful. We have Rando and F. Siebert, Biochemistry, 29, 3915 (1990).
not discussed time-resolved FT-Raman experiments with
8. M.S. Braiman, O. Bousche and K.J. Rothschild, Proc. Natl.
step-scan interferometers. In principle, they represent time- Acad. Sci. USA, 88, 2388 (1991).
resolved emission studies. However, compared with IR or
9. G. Souvignier and K. Gerwert, Biophys. J., 63, 1393
UVvisible spectroscopy, the low sensitivity makes it dif- (1992).
ficult to apply this technique and only a few investigations
10. I.W.M. Grim, J.A. Graham, R.M. Hammaker and W.G.
have been published.33,34 Fateley, Int. Lab., 67 (1984).
From the experience with commercial implementations
11. A. Becker, Ch. Taran, W. Uhmann and F. Siebert, Proc.
together with the corresponding software, the follow-
SPIE, 1145, 512 (1989).
ing comments appear necessary. They should be flexible
12. R.A. Palmer, C.J. Manning, J.A. Rzepiela, J.M. Widder and
enough to adapt the experimental setup to the requirements
J.L. Chao, Appl. Spectrosc., 43, 193 (1989).
of the experiment. Some of the shortcomings of the soft-
13. W. Uhmann, A. Becker, Ch. Taran and F. Siebert, Appl.
ware have been discussed. Furthermore, if one observes
Spectrosc., 45, 390 (1991).
some distortions in the spectra, they are not so easy to
14. P. Chen and R.A. Palmer, Appl. Spectrosc., 51, 580
correlate with the experimental causes, because they are
(1997).
connected through the FT. For this, it would be helpful if
15. J.R. Schoonover, G.F. Strouse, K.M. Omberg and R.B. Dyer,
the software allows monitoring of the acquired signals and
Comments Inorg. Chem., 18, 165 (1996).
difference interferograms together with the time-resolved
16. H. Frei, Nanosecond Step-scan FT-infrared Absorption Spec-
spectra. In order to adjust the signal gain to the digitiza-
troscopy in Photochemistry and Catalysis, in Fourier-
tion range of the AD converter, and to obtain information transform Spectroscopy: Eleventh International Conference,
on the contribution of the heat signal, the software should ed. J.A. de Haseth, American Institute of Physics, Woodbury,
also allow the recording of a signal at just one mirror NY, 2839 (1998).
position, including averaging to check for measurement 17. E. Kauffmann, H. Frei and R.A. Mathies, Chem. Phys. Lett.,
conditions. 266, 554 (1997).
18. C. Zscherp and J. Heberle, J. Phys. Chem. B, 101, 10 542 38. K. Thonke, T. Baier, J. Hamann, O. Scheerer and R. Sauer,
(1997). Appl. Spectrosc., 51, 548 (1997).
19. A.K. Dioumaev and M.S. Braiman, J. Phys. Chem. B, 101, 39. R.E. Imhof, A.D. McKendrick and P. Xiao, Rev. Sci. Ins-
1655 (1997). trum., 66, 5203 (1995).
20. W. Hage, M. Kim, H. Frei and R.A. Mathies, J. Phys. Chem., 40. B.O. Budevska and C.J. Manning, Appl. Spectrosc., 50, 939
100, 16 026 (1996). (1996).
21. B. Hessling, J. Herbst, R. Rammelsberg and K. Gerwert, Bio- 41. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared
phys. J., 73, 2071 (1997). Spectrometry, WileyInterscience, New York (1986).
22. C. Rodig, I.V. Chizhov, O. Weidlich and F. Siebert, Biophys. 42. C. Rodig and F. Siebert, Appl. Spectrosc., 53, 893
J., 76, 2687 (1999). (1999).
23. A.K. Dioumaev and M.S. Braiman, Photochem. Photobiol., 43. C.J. Manning and P.R. Griffiths, Appl. Spectrosc., 51, 1092
66, 755 (1997). (1997).
24. J.-R. Burie, W. Leibl, E. Nabedryk and J. Breton, Appl. Spec- 44. L. Mertz, Infrared Phys., 7, 17 (1967).
trosc., 47, 1401 (1993). 45. L.A. Nafie and D.W. Vidrine, Double Modulation Fourier
25. S.E. Plunkett, J.L. Chao, T.J. Tague and R.A. Palmer, Appl. Transform Spectroscopy, in Fourier Transform Infrared
Spectrosc., 49, 702 (1995). Spectroscopy, Academic Press, New York, 83123, Vol. 3
26. C. Rodig and F. Siebert, Temperature and Viscosity Effects (1982).
on Ligand Binding to Myoglobin Studied by Time-resolved 46. M.S. Hutson and M.S. Braiman, Appl. Spectrosc., 52, 974
Step-scan FT-IR Spectroscopy, in Fourier Transform Spec- (1998).
troscopy, eds K. Itoh and M. Tasumi, Waseda University 47. M.W. George, X.-Z. Sun, P.Y. Chen and M.J. Smith, Kinetic
Press, Tokyo, 219220 (1999). and Spectral Studies of Photochemical Reactions Using Step-
27. X. Hu, H. Frei and T.G. Spiro, Biochemistry, 35, 13 001 scan FT-IR Time-resolved Spectroscopy, in 3rd Interna-
(1996). tional Symposium on Advanced Infrared and Raman Spec-
28. J. Contzen and C. Jung, Biochemistry, 37, 4317 (1998). troscopy, Book of Abstracts, Vienna (1998).
29. S.V. Shilov, H. Skupin, F. Kremer, T. Wittig and R. Zentel, 48. C. Rodig and F. Siebert, Improvements in Signal Acqui-
Phys. Rev. Lett., 79, 1686 (1997). sition and Processing for Time-resolved Step-scan FT-IR
Spectroscopy, in Fourier Transform Spectroscopy: Eleventh
30. S.V. Shilov, E. Gebhard, H. Skupin, R. Zentel and International Conference, ed. J.A. de Haseth, American
F. Kremer, Macromolecules, 32, 1570 (1999). Institute of Physics, Woodbury, NY, 388391 (1998).
31. H. Wang, R.A. Palmer and C.J. Manning, Appl. Spectrosc., 49. R.H. Austin, K.W. Beeson, S.S. Chan, P.G. Debrunner,
51, 1245 (1997). R. Downing, L. Eisenstein, H. Frauenfelder and T.M. Nord-
32. M. Osawa, K. Yoshii, Y. Hibino, T. Nakano and I. Noda, J. lund, Rev. Sci. Instrum., 47, 445 (1976).
Electroanal. Chem., 426, 11 (1997). 50. A. Keens and A. Simon, Correction of Non-linearities in
33. T. Ishioka, T. Uchida and N. Teramae, Chem. Lett., 771 Detectors in Fourier Transform Spectroscopy, Bruker Ana-
(1998). lytische Messtechnik, US Patent 4 927 269 (1989).
34. G.S. Jas, C. Wan and C.K. Johnson, Appl. Spectrosc., 49, 645 51. T. Nakano, T. Yokoyama and H. Toriumi, Appl. Spectrosc.,
(1995). 47, 1354 (1993).
35. H.-L. Dai, Mikrochim. Acta, 14(Suppl.), 149 (1997). 52. C. Rodig and F. Siebert, Vib. Spectrosc., 19, 271 (1999).
36. G. Durry and G. Guelachvili, Appl. Opt., 34, 1971 (1995). 53. C. Rodig and F. Siebert, FEBS Lett., 445, 14 (1999).
37. K. Strong, T.J. Johnson and G.W. Harris, Appl. Opt., 36, 54. R. Rammelsberg, S. Boulas, H. Chorongiewski and K. Ger-
8533 (1997). wert, Vib. Spectrosc., 19, 143 (1999).
Time-resolved Mid-infrared Spectrometry Using an
Asynchronous Fourier Transform Infrared
Spectrometer
Koji Masutani
Micro Science Inc., Tokyo, Japan
1 INTRODUCTION In asynchronous time-resolved spectroscopy, the sim-

ple addition of a boxcar integrator to any commercial
In order to time-resolve repetitive dynamic phenomena, FT-IR spectrophotometer enables time-resolved measure-
asynchronous time-resolved Fourier transform infrared ments to be made without any further modifications to
(FT-IR) spectroscopy was developed on the basis of a either the hardware or software of a typical rapid-scanning
continuous-scan interferometer in 1991.1 In contrast to instrument. While the time resolution is limited by the rise
conventional methods,2 11 this method does not require time of the detector used and the repetition rate of the
synchronization between timing for the sampling of the stimuli is determined according to the Nyquist sampling
analog-to-digital converter (ADC) at equidistant intervals theorem, there is, in principle, no time limit in time reso-
based on the sampling theorem and that for time resolution. lution in this method. Furthermore, it does not require any
The term asynchronous time resolution comes from this special mechanical performance such as a strictly constant
feature. speed or a step-scanning interferometer.
Conventional synchronous time-resolved FT-IR spec- In general, time-resolved mid-infrared (IR) measure-
troscopy can be classified into two types. One, often known ments require a long measurement time, and various unex-
pected changes may happen to the samples, the interfero-
as the stroboscopic method, is based on the continuous-
meter, the system components or the other circumstances
scan interferometer4 7 and the other uses a step-scan
during the course of lengthy data collection. These changes
interferometer.8 11 In the former approach, the time resolu-
can lead to unreliability of the spectral data. In the mean-
tion is effected by sampling with an ADC, i.e. the ADC is
time, all the information of the dynamic phenomena in the
triggered at equally spaced time intervals and is sampled at
time domain is buried in the dynamic interferogram that is
equal intervals of optical path difference. This arrangement
created by repetitive stimuli.
imposes a tough restriction on the mechanical aspect that
A multi-channel asynchronous time-resolving system that
the interferometer always has to scan the moving mirror at
allows time-resolved data or time slices to be obtained,
a strictly constant speed in order to synchronize timing for
simultaneously brings further features such as a great reduc-
the sampling with that for the time resolution. The step-scan
tion in the measurement time and the minimization of possi-
approach has been successful in decreasing the time resolu-
ble damage to the sample.12 Furthermore, the simultaneous
tion below 1 s since it is no longer necessary to synchro-
measurement of time slices by the multi-channel system
nize the timing for the sampling and that for the time reso-
significantly improves their reliability. This is because the
lution. At present, the step-scan interferometer is more pop-
influence on the signal in each channel of a multichannel
ular than any other, even though it is still not widely used. system that comes from unexpected changes in the sample,
the interferometer and the circumstances during the mea-
John Wiley & Sons Ltd, 2002. surement is the same. Therefore, it is possible to remove
Continuous-scan FT-IR
Laser interferometer output Trigger signal for sampling
h ()
a d g i
Pre- Main Low-pass Computer
Interferometer amplifier amplifier filter ADC system
c
Infrared source Sample Detector f
b ()
Stimuli Main Gate
generator amplifier circuit
()
Delay circuit e ()
Synchronous for trigger signal Delayed trigger signals
pulse signal
Boxcar integrator
Figure 1. A typical setup for a single-channel asynchronous time-resolving measurement.
the common influence from each datum easily by a proper The modulated analog interferogram, which is measured by
calculation for the channels. the detector and converted into electric signal, is passed to
the gate circuit through the preamplifier and the main ampli-
fier. (e) A trigger pulse from each excitation pulse with a
2 ASYNCHRONOUS TIME-RESOLVED variable-delay time t is generated from the delay circuit.
FT-IR SPECTROMETRY (f) The signal arriving at the gate circuit is converted into
a discrete signal, which constitutes a discrete interferogram
A typical setup for this time-resolved method is shown that is time resolved with the stroboscopic process at the
schematically in Figure 1. The signal-processing assembly time delay t from the excitation set for each trigger pulse.
required for time-resolved measurements is shown at the (g) The low-pass filter converts the gate circuit output back
bottom. This assembly, which is attached to a conventional to a time-resolved analog interferogram for time delay t.
continuous-scan FT-IR spectrometer, consists of a stimulus (h) The laser interferometer in the FT-IR instrument gives
generator and a boxcar integrator including a main ampli- the ADC the sampling timing of the intervals of an optical
fier, a gate circuit and a delay circuit. path difference x. (i) The output from the low-pass filter
The main feature of this method is the use of a low-pass is sampled and reconverted into a discrete interferogram at
filter. The low-pass filter is placed between the gate circuit intervals of the constant optical path difference x by the
and the ADC. It removes the high-frequency components ADC, and is sent to the computer. Inverse Fourier transfor-
from discrete signals time-resolved by the gate circuit and mation of this interferogram gives a time-resolved spectrum
converts them to analog signals that the ADC can acquire for time delay t from the excitation.
at its own sampling timing without relation to the time- In Figure 3 is shown the relationship between the tim-
resolving timing of the gate circuit. ing for time resolving by the gate circuit and that for data
The excitation, the response of the sample and the output acquisition by the ADC during two typical scans of the
signals from various units in this configuration are shown moving mirror. In this diagram, the optical path difference
in Figure 2, where each trace from (a) to (i) shows the sig- of the interferometer caused by the movement of the mov-
nal at the process of the corresponding number in Figure 1. ing mirror and the corresponding ordinary interferogram are
The work of the asynchronous time-resolving system in shown on the x-axis and the lapse of time and the sample
each process is described as follows: (a) The interferome- responses to the excitations on the t-axis. The moving mir-
ter generates a conventional interferogram. (b) The stim- ror scans at a constant speed, corresponding to the slope
ulus generator excites the sample at a period of time of the bold oblique lines in Figure 3, in the same manner
t. (c) After each excitation, the sample relaxes from an as in the conventional continuous-scan FT-IR spectrome-
excited state to the original state. Excitation in this case ter. Excitations are given to the sample at intervals of an
means not only an electronic excitation but also stimulation equal time t without any synchronization with the move-
of the sample in a more generalized sense. (d) The excita- ment of the mirror. The dynamic interferogram modulated
tion pulses given to the sample also modulate the ordinary by the excitations is time resolved by the gate circuit at the
interferogram through the absorption change of the sample. timing (indicated with squares in Figure 3) delayed by a
Time-resolved Mid-IR Spectrometry Using an Asynchronous FT-IR Spectrometer 3
Ordinary interferogram before excitation Ordinary interferogram and sampling timing

at equal optical path difference interval
t
(a)
x (tm )
Excitation pulses

t
(b)
Response of sample
Sample response to excitation at equal time interval

t
(c)
Modulated interferogram by excitation
and gate timing for time resolution

t
(d)
Gate timing for time resolution
t
(e)
Time-resolved discrete interferogram

t
Figure 3. Relationship between the gate timing for time resolving

t by the gate circuit (horizontal broken lines and squares) and the
(f)
sampling timing for data acquisition by the ADC (vertical lines
and circles) during two typical scans of the moving mirror.
Time-resolved analog interferogram
intervals of a constant optical path difference h (indicated

t with circles in Figure 3) by the ADC.
(g) With this system, the interferogram corresponding to
the spectrum of the sample in a certain transient state
Sampling timing of ADC
can be measured in a single scan of the interferometer.
Discrete interferograms sampled by the ADC can be simply
x (tm) co-added to improve the signal-to-noise ratio (S/N). This
(h) is because each analog interferogram obtained from the
low-pass filter keeps the same phase even though the
Sampled time-resolved interferogram
corresponding discrete time-resolved interferogram obtai-
ned from the gate circuit is not synchronous with the others.
x (tm) The function of the low-pass filter is explained in more
(i) detail in Figure 4. When an analog interferogram (a) is
Figure 2. Signals in the processes in Figure 1(a)(i). inversely Fourier transformed, the corresponding spectrum
(b) is obtained. When a discrete interferogram (c) obtained
constant time t from each excitation (indicated with stars by sampling this analog interferogram by the gate cir-
in Figure 3). It is then converted back to time-resolved ana- cuit is inversely Fourier transformed, the corresponding
log signals by the low-pass filter. The output signal from spectrum (d) consists of the original component (b) and
the low-pass filter is sampled at the ordinary sampling posi- high-frequency components in which the original inter-
tions of the conventional FT-IR spectrophotometer having ferogram appears at positions of the sampling frequency
Analog interferogram Spectrum of analog interferogram
FT
(a) t (b) 0 fm f
Discrete interferogram Spectrum of discrete interferogram
FT
(c) t (d) 0 fm 1/ f
Frequency response of low-pass filter
(e) 0 fm f c 1/ f
Reproduced analog interferogram Output spectrum from low-pass filter
FT
(g) t (f) 0 fm f
Figure 4. Role of the low-pass filter.
Continuous-scan FT-IR
Laser interferometer output Trigger signal for sampling
Pre- Main Low-pass Computer

Interferometer ADC
amplifier amplifier filter system
Infrared source Sample Detector
Stimuli
generator
Synchronous
pulse
Time resolving
Main module
signal
amplifier
Delayed
trigger Gate Low-pass
signals circuit 1 filter 1 ADC 1
1
2
Gate Low-pass
Delay ADC 2
circuit 2 filter 2
circuit
for 3
Gate Low-pass
Computer
trigger
ADC 3 system
signal circuit 3 filter 3
32
Gate Low-pass
circuit 32 filter 32 ADC 32
Figure 5. A typical multichannel asynchronous time-resolving system. [Reproduced from K. Masutani et al., Appl. Spectrosc., 53, 588
(1999) by permission of the Society for Applied Spectroscopy.]
1/t multiplied by integers according to the sampling theo- The modulated interferogram Fx, t corresponding to
rem. Therefore, when this discrete interferogram is passed Figure 2(d) is expressed as
through the low-pass filter having a frequency response (e) C1
with an appropriate cutoff frequency fc , the high-frequency Fx, t D Ts, tBs cos 2px s ds
components are not found in the resultant spectrum (f), 1
C1
which becomes identical with the original spectrum (b)
D F0 x C [Ts, t* t t]Bs cos 2px s ds
9
obtained from the analog interferogram (a) before sampling. 1
Consequently, the analog interferogram (g) is obtained from 3
C1
the low-pass filter. F0 x D T0 sBs cos 2px s ds 4
The asynchronous time-resolving system using one box- 1
car integrator is a single-channel system that cannot acquire where F0 x is the constant component of the interferogram
a set of several time-resolved data in a single measurement. unchanged by the excitation and Bs is the spectrum of
However, all the information about the time domain of the radiation from the light source.
dynamic phenomenon is buried in the dynamic interfero- The relation between variables t and x is now consid-
grams that are created by repetitive stimuli. ered. Variable t is used to specify the time relating to the
A multichannel asynchronous time-resolving system can excitation, and another time variable tm is needed to define
solve the problem. A typical setup with 32 channels is the time relating to the position of moving mirror, since the
shown schematically in Figure 5. The time-resolving mod- excitations are applied to the sample asynchronously with
ule is shown at the bottom and has 32 time-resolving units, the movement of the mirror in this approach. For a standard
each of which consists of a gate circuit, a low-pass filter Michelson interferometer, variable x is related to tm by the
and an ADC. Time resolution is carried out by setting a equation x D 2Vtm during each scan of the moving mirror,
different time delay t to each gate circuit. The analog where V refers to the velocity of the moving mirror. If the
time-resolved interferograms passed by each low-pass filter origin of t is set at the excitation pulse of the timing near-
are sampled simultaneously at each zero crossing point of est to tm D 0 during one scan of the moving mirror, a time
the laser interferogram. difference t D tm t exists during that scan. Since t
changes for each scan (Figure 3), the second term in equa-
tion (3) also changes for each scan of the moving mirror.
3 FORMULATION OF ASYNCHRONOUS The modulated analog interferogram input to the gate
TIME-RESOLVED METHOD circuit is converted into discrete signals at times delayed
by t from the excitation pulse (Figure 2e). The discrete
The time-resolving process in this method may be for- signal F0 x, t shown in Figure 2(f) is given as
mulated in the following way. When a single excitation
pulse is applied to the sample, the transmittance spectrum F0 x, t D t t t F0 x
9
is expressed as
C1
C [Ts, t* t t]Bs cos 2px s ds 5
9
Ts, t D T0 s C Ts, t 1

1
C1
where T0 s is the component unchanged by the excitation
D t t t Ts, tBs cos 2px s ds 6
9
and Ts, t is that changed by the excitation. Variables 1

s and t are, respectively, the spectral wavenumber and
where t t t is used to express the gating by the
9
the time variable relating to the excitation. When the

trigger pulse as shown in Figure 2(e).1 For simplicity, the
sample is stimulated repetitively by the external pulses at
gate width is assumed to be infinitesimal. The derivation
time intervals t as shown in Figure 2(b), the transmittance
of equation (6) from equation (5) is given in the Appendix
spectrum is expressed as
to this section. Ts, t is the time-resolved transmittance
Ts, t D T0 s C Ts, t* t t 2 spectrum of the sample in the transient state at time delay
9
t from the excitation pulse. Thus, the time resolution of

This response of the sample is shown in Figure 2(c). t t the transmittance spectrum has been achieved, even though
9
is a Dirac delta comb corresponding to a repetitive operation the interferogram changes its discrete positions in each scan
with a period t, as shown in Figure 2(b).13,14 The symbol of the moving mirror.
* represents the self-convolution operator. The period t The discrete interferogram from the gate circuit is
should be longer than the relaxation time from the transient inversely Fourier transformed for the purpose of under-
state of the sample back to the original state. standing the function of the low-pass filter. Inverse Fourier
transformation is performed with respect to time tm , because as shown in Figure 4(g). It is independent of variable t,
inverse Fourier transformation with respect to the optical indicating that identical interferograms can be obtained in
path difference x relating to tm is carried out to obtain the different scans of the moving mirror. Once a time-resolved
spectrum in this method in the same manner as in conven- analog interferogram has been obtained from the low-
tional FT-IR spectrometry. Inverse Fourier transformation pass filter, the corresponding spectrum can be calculated
of F0 x, t is expressed by a convolution of t t t by the same procedure used for the conventional FT-IR
9
and the integral part in equation (6) after inverse Fourier spectrophotometer.
transformation. Spectrum Tf, tBf/2n, which is the
function of the modulation frequency f D 2Vs, is obtained
by inversely Fourier transforming the integral part. Inverse 3.1 Appendix
Fourier transformation of t t t for each scan of the
9
moving mirror is expressed as Equation (6) may be derived from equation (5) in the
C1 following way. Since t t t is independent of s, it can
9
be included in the integrand of the second term of equa-
t t t expi2pftm dtm
9
tion (5). We then focus on t t t[Ts, t* t t],
9
1
which is a function of t.
1
D 1/t f exp[i2pt C tf] 7
9
t The Dirac delta comb may be expressed as

1 1 t t D t mt 10
9
D f C f
t t m

i2pt C t where t is the Dirac delta function and m refers to inte-
exp C
t gers. By the use of equation (10), the following equations
are obtained:
1 i2pt C t
C f C exp C 8
t t t t t[Ts, t* t t]
9
9

The right-hand side of equation (7) represents a Dirac C1
delta comb having a phase factor exp[i2pt C tf]. D t t mt Ts, t t0

m 1
As shown in Figure 4(d), the spectrum of F0 x, t contains

the original spectrum Tf, tBf and higher frequency
0 0
t nt dt
components. The latter components are equivalent to the
n
original spectrum and appear at positions determined by

multiplying the sampling frequency 1/t by integers. The

D t t mt Ts, t nt 11
spectral component at the origin (f D 0) has no phase fac- m n
tor, and the corresponding interferogram never changes its
phase in each scan of the moving mirror. Since none of the where n also refers to integers. Since t is set longer than
spectral components should overlap with each other, the the lifetime of the transient state under study, Ts, t D 0
sampling rate 1/t must be higher than twice the maximum for t < 0 and t D t. Thus, the signal expressed in equa-
modulation frequency fM of the original spectrum, accord- tion (11) exists only when m D n. The right-hand side of
ing to the Nyquist sampling theorem. The above analysis equation (11) can then be changed into the following forms:

shows that the discrete interferogram F0 x, t has complete D ft t mtTs, t mtg
information on the original spectrum Tf, tBf, which m
can be obtained by eliminating the higher frequency com-
ponents by the low-pass filter with the functions described D ft t mtTs, tg 12
m
above. Then, the following equation is obtained for the out-
put from the low-pass filter: Since Ts, t in equation (12) can be treated as a
C1 constant, the following equations are obtained:
1
F0 x, t D Ts, tBs cos 2px s ds 9
t tt[Ts, t* t t] D Ts, t
9
t 1

tt mt D Ts, t t tt 13
9
where 1/t comes from the first term in equation (8) and the
m
integral part is the same as the original one in equation (6).
Equation (9) corresponds to the analog interferogram for By substituting equations (13) and (2) for t D t into equa-
the transient state at time delay t from the excitation tion (5), equation (6) is derived.
4 PERFORMANCE OF ASYNCHRONOUS 1.2
Ratio
1.0
TIME-RESOLVED FT-IR
0.8
SPECTROMETERS (b)
The most significant features of the asynchronous time-

2.0
resolving method are that it can be applied to the
conventional continuous-scan FT-IR spectrophotometer and 1.5
Intensity
does not require synchronization between the timing for the
time resolution and that for the sampling by the ADC. 1.0
These features yield superior performance including
high time resolution and high spectral accuracy without 0.5
any distortions and/or artifacts. Moreover, while the time
resolution is limited by the rise time of the detector used, 3600 2800 2000 1600 1200 800
there is in principle no shortest limit since the time res- (a) Wavenumber /cm1
olution is determined by the time slice of the time delay Figure 6. Accuracy test for the asynchronous time-resolving
t from each excitation to the sample. On the other hand, measurement. [Reproduced from K. Masutani et al., Appl. Spec-
a restriction on the length of the relaxation time from the trosc., 46, 560 (1992) by permission of the Society for Applied
Spectroscopy.]
excited state is imposed because of the difficulty involved
in scanning the moving mirror very slowly. As is apparent
measurements of both the broad bands, such as the entire
from Figure 4, the sampling frequency 1/t at the gate cir-
profile of the background spectrum, and the sharp absorp-
cuit, that is, the repetition rate of the stimulus to the sample,
tion bands, such as those due to water vapor and carbon
must be not less than twice the highest Fourier frequency
dioxide.
fm of the spectrum. As the chopping frequency of the spec-
The resultant of the accuracy test using the multichan-
trum is determined by the scanning speed of the moving
nel asynchronous time-resolving system is displayed in
mirror, time-resolved measurements of samples having a
Figure 7. These are the difference spectra between the
long relaxation require an interferometer capable of very
slow scanning. transmission spectrum of a polystyrene film measured by
The multichannel asynchronous time-resolving system the first channel and ones by the others. The measuring
yields further features. The measurement time is reduced, conditions were as follows; resolution, 4 cm1 ; number of
the S/N of the data is increased substantially and possible scans, 100; sampling frequency, 5 kHz; cutoff frequency of
damage to a sample by the stimulus is decreased. Moreover, low-pass filter, 2.2 kHz; gate interval t, 5 kHz; gate width,
simultaneous time resolution by the multichannel system 50 ns; and time delay t, 10 s. All the spectra are flat
improves the reliability of the time-resolved data signifi- below the noise level over the entire wavenumber range and
cantly, since the influence of unexpected changes of the do not have any spurious peaks. Even bands due to water
sample and the interferometer contribute to the signal in vapor and carbon dioxide are almost completely eliminated.
each channel in the multichannel system equally.
5
The observed spectra from one test of accuracy are shown
in Figure 6. For this purpose, only the radiation from the 4
Transmittance (%)
light source was measured. Spectrum (a) was measured by

3
using the asynchronous time-resolving system consisting
2 15
of an ordinary continuous-scan FT-IR spectrophotometer,
a mercury cadmium telluride (MCT) detector and a box- 1
car integrator. Spectrum (b) is the ratio of spectrum (a) to
el
0 8
the spectrum measured with the same FT-IR spectropho-
n
an
Ch
tometer without the boxcar integrator. The measurement 1

conditions were as follows; resolution, 4 cm1 ; number of
2 1
scans, 10; rise time of detector, about 2 s; gate interval t, 4000 3500 3000 2500 2000 1500 1000
50 s; and gate width, 0.5 s. Only the atmospheric bands Wavenumber /cm1
due to absorption by water vapor and carbon dioxide can be
Figure 7. Accuracy test using the multichannel asynchronous
seen in this spectrum. From the fact that the ratio spectrum time-resolving system. [Reproduced from K. Masutani et al.,
is flat over the entire wavenumber region, it is clear that the Appl. Spectrosc., 53, 588 (1999) by permission of the Society
asynchronous time-resolving system allows fairly accurate for Applied Spectroscopy.]
These observations prove that an accurate spectrum can be was 30 : 70 by weight. The mixing ratio of base liquid
obtained for each channel. Moreover, it is apparent that an crystals is described in Figure 8. A cell made by EHC
unexpected spectrum arising from disturbances in the inter- (Tokyo, Japan) was used for this test. It consists of a pair of
ferometer, the sample compartment, the detector and the CaF2 windows with a space of 2.0 m between them. The
circumstances is almost equal on all the channels of the inner surface of the windows was coated with indium tin
time-resolving system and can be removed from the resul- oxide serving as an electrode and with polyimide rubbed
tant data. Therefore, the time-resolved data are improved in one direction. After the sample had been heated to the
in reliability substantially. region of the isotropic-liquid phase (90 C), it was injected
into the cell placed in the same heating unit. The sample in
the cell was cooled to 40 C slowly at a rate of 1 C min1
5 APPLICATION TO STUDY OF to obtain a homogeneously oriented liquid crystal in the
REORIENTATION DYNAMICS IN region of Sc*.
LIQUID CRYSTALS A JEOL JIR-6500 FT-IR spectrometer equipped with
an IR microscope, an MCT detector and an IR polarizer
In one of the typical applications of the asynchronous time- was used. This FT-IR spectrometer is capable of acquiring
resolving system, there is a study of reorientation dynamics interferograms in the bidirectional mode. The IR polarizer
in a ferroelectric liquid crystal induced by the reversal of was set at an angle of 45 to the orientation direction of the
the external electric field.15 24 One of the studies to which liquid crystal to obtain the maximum absorption change.
the multichannel system was applied is presented here. The time-resolving measurements were carried out for
The sample employed was a ferroelectric liquid crystal the following three cases. In the first case (case A), a
consisting of two kinds of chiral dopants and five kinds single boxcar integrator was used. For the other two cases
of base liquid crystals (a generous gift from Kashima (cases B and C), a 16-channel asynchronous time-resolving
Oil). The structure and phase sequence of the sample are system made by Sopac (Tokyo, Japan) was employed.
shown in Figure 8, where Cryst, S, N and Iso refer, to the Cases B and C differ in the number of scans, 200 and
crystalline, smectic, nematic and isotropic-liquid phases, 3200, respectively. Spectra were collected under almost the
respectively. Chiral dopants named KLS-372S and KLS- same conditions for three cases except for the number of
340S in Figure 8 were mixed in a ratio of 46 : 54 by weight, scans. Rectangular electric pulses with 20 V peak voltages
and the mixing ratio of the dopant and base liquid crystals and a 200 s period were applied to the liquid crystal cell.
N
C6H13O C9H19 17.5%
N
C6H13O C O OC6H13 N
S S C8H17O C8H17 17.5%
O O
S
CF3 N
KLS-372S N
C9H19O C7H15 17.5%
C6H13O C O OC6H13 N
S R N
O O
S C10H21O C8H17 17.5%
F3C
N
KLS-340S
N 30%
Chiral dopant
C5H11 C7H15
(KLS-372S/KLS-340S = 46 : 54)
N
KB-05
Base liquid crystals
Liquid crystal mixture (chiral dopant/base liquid crystal = 30 : 70)
65 C 73 C 82 C
Cryst. Sc Sa N Iso
59 C 68 C 80 C
Figure 8. Structures of chiral dopants and base liquid crystals, and the phase transition employed in the general performance test.
[Reproduced from K. Masutani et al., Appl. Spectrosc., 53, 588 (1999) by permission of the Society for Applied Spectroscopy.]
The wavenumber resolution was 4 cm1 , the mirror speed rapid-scan interferometer than that for a step-scan inter-
of the interferometer was 1.6 mm s1 and the frequency of ferometer. Rapid sampling of each interferogram ought to
sampling was about 5 kHz. lower the probability of the distortions and artifacts in the
Figure 9 shows the time-resolved polarized spectra of spectrum caused by unexpected changes.
the liquid crystal for the three cases. The ordinate repre- In further exploring repetitive fast phenomena of a ferro-
sents the absorption change from the spectrum measured electric liquid crystal on switching the electric field, it has
with the shortest time delay. For the spectra in Figure 9(a), already been found that the CDO groups respond first and
obtained using our existing (single-channel) asynchronous then the alkyl chain and others follow.24
time-resolving system, the number of scans of the inter-
ferometer for each time-resolved spectrum was 200, but
the number of accumulations of interferograms was 400 6 PRESENT LIMITATIONS AND FUTURE
because the sampling was carried out at both forward and DEVELOPMENTS
backward scans of the interferometer. It took over 6 min to
measure each time-resolved spectrum, and the total mea- The asynchronous time-resolving system is best suited for
suring time for taking 16 time-resolved spectra was 1.5 h, observing repetitive transient phenomena with lifetimes in
excluding the data processing time for eliminating the water the range of approximately 107 103 s using a conven-
vapor spectrum. For the spectra in Figure 9(b) and (c), mea- tional continuous-scan FT-IR spectrometer. The shortest
sured by use of the multichannel system, the numbers of time for the measurement (tmin ) is determined by the
scans and accumulations were 200 and 3200, respectively. response of the MCT detector and the preamplifier. The
Since the function of the sampling at both forward and longest limit (tmax ) is determined by the requirement for
backward scans had not yet been developed for the multi- avoiding the overlap of components in the inversely Fourier
channel system, the numbers of the scan and accumulation transformed spectrum, which is expressed by the Nyquist
were the same. It took only about 6 min to measure the theorem as t 1/2fM D 1/4nsM , where sM is the
16 spectra in Figure 9(b), that is, the same as the measur- maximum wavenumber of the IR region to be measured.
ing time for each spectrum in Figure 9(a). In the case of The limitations imposed on the measurable time scale
Figure 9(c), the total measuring time was 1.5 h, the same may potentially be improved. For example, tmin can be
as that in Figure 9(a). made shorter by employing a detector having a faster
As can be seen from Figure 9(a), each transient spectrum response. If a Cu-doped Ge detector having a time constant
clearly has an S/N sufficient to allow discussion of the of about 5 ns is used, tmin will decrease to the nanosecond
behavior of the functional groups. However, the spectra range. Decreasing tmax to time ranges longer than 1 ms may
in Figure 9(a) show significant unexpected changes as the be carried out in several ways. (1) The most straightforward
spectra do not change smoothly with time. On the other way is to make the scan speed slower, since a smaller V
hand, the spectra in Figure 9(b) show an S/N comparable to leads to a longer t in the Nyquist relation. Some FT-IR
those in Figure 9(a), although the number of accumulations spectrometers already allow a very slow-scanning function.
is half. Furthermore, the changes in transients are very (2) In many cases, it is not necessary to measure the spec-
smooth. In the spectra in Figure 9(c), the advantages of the trum region over the wavenumber range 40002000 cm1 .
multichannel system can be seen more clearly. Although When sM can be set at 2000 cm1 using a proper optical
the spectral regions of the bands of the water vapor and filter, tmax can be doubled.
the carbon dioxide are not shown in Figure 9, the bands It is worth pointing out that the asynchronous time-
are almost eliminated without any further data processing resolving method can apply to an FT-IR spectrophotometer
in Figure 9(b) and (c). with a pulsed IR light source, in which case the pulsed
These spectra give useful insights into the time-resolving light source itself plays the role of the gate circuit. Light
measurement. During the course of lengthy data collec- pulses delayed by t from the excitations to the sample
tion, various unexpected disturbances can affect the sample create a time-resolved discrete interferogram that includes
and the instrument, and can cause unexpected changes the information on the sample in the transient state. This
in them. It is natural that these changes reduce the reli- discrete interferogram is converted back to an analog inter-
ability of the spectral data. Therefore, the measurement ferogram by the low-pass filter, which is processed in
should be made by keeping the measurement time for the same manner as described before. In this case, the
time-resolving interferograms as short as possible, while extremely fast response of the detector that can respond
simultaneously measuring as many interferograms as possi- to the dynamic phenomena of the sample is not required
ble. Furthermore, rapid sampling of each interferogram may because the incident signal into the detector is already time
be required, more similar to the sampling for a conventional resolved. Therefore, only the regular response that reacts to
0.8
Absorbance
0.6
0.4
0.2 70.0
)
s
50.0
e(
tim
30.0
ay
0.0 10.0
el
D
1700 1600 1500 1400 1300 1200 1100 1000
(a) Wavenumber /cm1
0.6
Absorbance
0.4
0.2
78.0
)
s
57.2
e(
36.4 tim
ay
0.0 15.6
el
D
1700 1600 1500 1400 1300 1200 1100 1000

(b) Wavenumber /cm1
0.6
Absorbance
0.4
0.2
78.0
)
s
57.2
e(
tim
36.4
ay
0.0 15.6
el
D
1700 1600 1500 1400 1300 1200 1100 1000

(c) Wavenumber /cm1
Figure 9. Time-resolved polarized IR spectra for the reorientation dynamics in the ferroelectric liquid crystal by the reversal of the
external electric field. [Reproduced from K. Masutani et al., Appl. Spectrosc., 53, 588 (1999) by permission of the Society for Applied
Spectroscopy.]
maximum chopping frequency of the time-resolved spec- 10. R.A. Palmer, C.J. Manning, J.A. Rzepiela, J.M. Widder and
trum is required for detection. A slowly responding detector J.L. Chao, Appl. Spectrosc., 43, 193 (1989).
can be regarded as one having the function of the low-pass 11. R.A. Crocombe and S.V. Compton, FTS/IR Notes of Bio-
filter. This means that the restriction on the time resolution Rad Digilab Division, No. 82, Bio-Rad, Cambridge, MA
(1991).
has been removed in this light-probe asynchronous sam-
pling mode. 12. K. Masutani, K. Numahata, K. Nishimura, S. Ochiai,
Y. Nagasaki, N. Katayama and Y. Ozaki, Appl. Spectrosc.,
In the IR region, synchrotron radiation seems most fea-
53, 588 (1999).
sible as a pulsed light source to realize such a measuring
13. R.N. Bracewell, The Fourier Transform and Its Applica-
system. An attempt has already been started with the goal
tions, 2nd edition, McGraw-Hill, New York, Ch. 5, 77
of 10 ps time resolution at 3 cm1 spectral resolution.25 (1965).
14. G.A. Vanesse and H. Sakai, Fourier Spectroscopy, in
Progress in Optics, ed. E. Wolf, North-Holland, Amster-
ABBREVIATIONS AND ACRONYMS dam, Ch. 7, 275, Vol. 6 (1967).
15. K. Masutani, A. Yokota, Y. Furukawa, M. Tasumi and
ADC Analog-to-digital Converter A. Yoshizawa, Appl. Spectrosc., 47, 1370 (1993).
16. M.A. Czarnecki, N. Katayama, Y. Ozaki, M. Satoh, K. Yo-
shio, T. Watanabe and T. Yanagi, Appl. Spectrosc., 47, 1382
REFERENCES (1993).
17. M.A. Czarnecki, N. Katayama, M. Satoh, T. Watanabe and
1. K. Masutani, H. Sugisawa, A. Yokota, Y. Furukawa and Y. Ozaki, J. Phys. Chem., 99, 14 101 (1995).
M. Tasumi, Appl. Spectrosc., 46, 560 (1992).
18. K.H. Kim, K. Ishikawa, H. Takezoe and A. Fukuda, Phys.
2. P.R. Griffiths and J.A. de Haseth, Fourier Transform Infrared Rev. E, 51, 2166 (1995).
Spectrometry, John Wiley & Sons, New York (1986).
19. B. Jin, Z. Ling, Y. Takahashi, K. Ishikawa, H. Takezoe,
3. J.J. Sloan and E.J. Kruus, Time-resolved Fourier Trans- A. Fukuda, M. Kakimoto and T. Kitazume, Phys. Rev. E, 53,
form Spectroscopy, in Advances in Spectroscopy, eds R4295 (1996).
R.J.H. Clark and R.E. Hester, John Wiley & Sons, Chich-
20. N. Katayama, M.A. Czarnecki, M. Satoh, T. Watanabe and
ester, Ch. 5, Vol. 18 (1989).
Y. Ozaki, Appl. Spectrosc., 51, 487 (1997).
4. A.W. Mantz, Appl. Spectrosc., 30, 459 (1976).
21. L. Verma, B. Zhao, S.M. Jiang, J.C. Shen and Y. Ozaki,
5. A.W. Mantz, Appl. Opt., 17, 1347 (1978). Phys. Rev. E, 56, 3053 (1997).
6. J.F. Durana and A.W. Mantz, Laboratory Studies of React- 22. L. Verma, B. Zhao, H. Terauchi and Y. Ozaki, Phys. Rev. E,
ing and Transient System, in Fourier Transform Infrared 7, 1868 (1999).
Spectroscopy, eds J.R. Ferraro and L.J. Basile, Academic
23. Y. Nagasaki, K. Masutani, N. Katayama, T. Yoshihara,
Press, New York, Ch. 1, Vol. 2 (1979).
K. Numahata, K. Nishimura, S. Ochiai and Y. Ozaki, Int.
7. A.A. Garrison, R.A. Crocombe, G. Mamantov and J.A. J. Vib. Spectrosc., 3, 5 (1999).
de Haseth, Appl. Spectrosc., 34, 399 (1980).
24. Y. Nagasaki, K. Masutani, T. Yoshihara and Y. Ozaki, J.
8. R.E. Murphy, F.H. Cook and H. Sakai, J. Opt. Soc. Am., 65, Phys. Chem. B, 104, 7881 (2000).
600 (1975).
25. R.A. Palmer, G.D. Smith and P. Chen, Vib. Spectrosc., 19,
9. H. Sakai and R.E. Murphy, Appl. Opt., 17, 1342 (1978). 131 (1999).
Grating Spectrometers
Mamoru Hashimoto1 , Teturo Yuzawa2, Chihiro Kato3 , Koichi Iwata2 and

Hiro-o Hamaguchi2
1
Osaka University, Toyonaka, Japan
2
The University of Tokyo, Tokyo, Japan
3
Kanagawa Industrial Technology Research Institute, Ebina, Japan
1 INTRODUCTION current-coupled time-resolved infrared spectrometer in

1989, using a modified Perkin-Elmer (Beaconsfield, UK)
The history of time-resolved infrared spectroscopy extends Model 180 infrared grating spectrometer.3 Microsec-
for more than half a century, beginning when rapid scan ond time resolution was achieved with sensitivity of
infrared spectroscopy was developed in the 1940s by Baker A 104 . We further improved the system using another
and Robb.1 Since then, intensive efforts have been made commercial infrared spectrometer, a Hitachi (Tokyo, Japan)
by many researchers to achieve higher sensitivity, higher Model I-3000. This system utilized a high temperature
time resolution, and higher versatility. Today, AC-coupled MoSi2 source and a high-speed photovoltaic (pv) mercury
dispersive spectrometers have a high enough sensitivity to cadmium telluride (MCT) detector with a low-noise wide-
detect an absorbance change smaller than one part in a band pre-amplifier.4 An improvement by two orders of
million. Sub-picosecond time resolution has been achieved magnitude was accomplished both for time resolution (to
with the use of pulsed infrared lasers of various kinds. 50 ns) and sensitivity (to A D 106 ). A third system con-
Versatile dispersive and Fourier-transform time-resolved sisting of similar components is now commercially avail-
infrared spectrometers are now commercially available. able from Jasco (Tokyo, Japan) as the Model TRIR-1000.5
Even so, high sensitivity and high time resolution are Alternating current-coupled dispersive measurements
still rarely compatible with each other. For example, the provide high sensitivity and excellent versatility, although
ultrafast infrared laser method hardly detects an absorbance the time resolution is moderate. These characteristics are
change (A) of one-thousandth, while versatile commercial shared with the step-scan Fourier-transform method. There
spectrometers are lacking in picosecond time resolution. A has been discussion over the past 5 years with respect to the
universal method for time-resolved infrared spectroscopy is relative time resolution and sensitivity of AC-coupled dis-
yet to be developed. persive spectrometers and AC-coupled step-scan Fourier-
In this article, we describe time-resolved infrared spec- transform spectrometers. Theoretically, both are as good
troscopy based on alternating current-coupled dispersive as each other. In reality, however, many experimental fac-
spectrometers. The method was first used by Siebert et al. tors have to be taken into account. We do not go into this
for point-by-point measurements of the transient infrared discussion here and await more experimental data to be
spectra of the bacteriorhodopsin photointermediates.2 The accumulated to settle the issue. As far as the existing data on
present authors developed the first scanning alternating photochemical applications are concerned, the performance
of the alternating current-coupled dispersive spectrometer
John Wiley & Sons Ltd, 2002. exceeds that of the step-scan Fourier-transform approach in
that it can detect a smaller absorbance change, as small as where At and It are the time-dependent absorbance and
one part in a million. optical power with the stimulation applied to the sample,
and A and I are those without the stimulation.
Figure 1 shows a two-dimensional representation of
2 INSTRUMENTATION a set of nanosecond time-resolved infrared difference
spectra of photoexcited 4-N,N-dimethylaminobenzonitrile
2.1 Transient infrared absorption (DMABN) in butanol.5 The vertical and horizontal axes
correspond to time and wavenumber, respectively. The
absorbance change, At, is indicated by the rainbow col-
Time-resolved infrared spectroscopy detects a time-
ors. The red color, which represents a positive absorbance
dependent infrared absorbance change, At, which is
change, indicates absorbance increases due to the genera-
induced by external stimulation of the sample by light,
tion of new molecular species. The blue color, representing
electric field, etc. The absorbance change is defined as:
a negative absorbance change, indicates the absorbance

It decrease due to the depletion of the parent molecule.
At D At A D log 1
I The green color corresponds to null absorbance change.
Figure 1. Two-dimensional image of 50 ns time-resolved infrared spectra of 4-N,N-dimethylaminobenzonitrile in butanol.5 Vertical and
horizontal axes indicate time and wavenumber, respectively.
Fast Time-resolved Mid-infrared Spectroscopy Using Grating Spectrometers 3
Two positive transient infrared bands showing different 2.3 System configuration
time dependence are observed at 2096 and 2040 cm1 .
They are assigned to the singlet and triplet charge transfer A typical system for time-resolved infrared spectrometry
(CT) states of DMABN. The negative band at 2216 cm1 using the AC-coupled dispersive approach consists of an
is ascribed to depletion of the ground state. The two- infrared light source, a stimulation source to induce the
dimensional representation in Figure 1 clearly demonstrates transient absorption, an infrared grating spectrometer, an
that time-resolved infrared spectroscopy detects not only infrared detector, an alternating current-coupled amplifier,
the transient species generated but also the parent species and signal processing electronics. Figure 2 shows the time-
depleted by photoexcitation. This contrasts with time- resolved infrared system we developed for photochemical
resolved resonance Raman spectroscopy in which only the applications. The infrared light passing through the sample
transient species in resonance with the probing laser are is dispersed by the grating monochromator and is detected
detected by virtue of resonance enhancement. The two- with the infrared detector. The transient infrared inten-
dimensional representation is also helpful as a basis for sity It is amplified with an alternating current-coupled,
the singular value decomposition (SVD, see the following highly sensitive, fast-response and high-gain amplifier, and
section) analysis in which the time-resolved infrared spectra is digitized and averaged by the signal-processing electron-
are represented by a two-dimensional matrix. ics. The optical chopper is used for the measurement of the
steady infrared intensity I without stimulation. A shutter is
used to remove the Q-switch noise (see below). This sys-
2.2 AC-coupling detection tem covers a wide spectral region of 4000700 cm1 and
detects an absorbance change of 3 107 with microsec-
Typical absorbance changes A/A to be detected in pho- ond time resolution or 105 absorbance change with 50 ns
tochemical applications are 104 106 or even less. It time resolution.4 In the following sections of this article,
is prohibitively difficult to detect such small changes in each component of this time-resolved infrared system is
absorbance by simply measuring At and A separately and described in more detail.
taking the difference. The difficulty is similar to that in
observing a star in the daytime. A star can be seen at night
when the sunlight is off but it is hidden in the sunlight 2.4 Infrared light source
during daylight hours. The difficulty can be bypassed by
alternating current-coupling. The time-dependent (transient) A high intensity and highly stabilized infrared light
infrared intensity, It, consists of a time-dependent part, source is required for time-resolved infrared spectroscopy.
It, and a time-independent part, I:
It D I C It 2 Laser
Typically, I is 104 106 times larger than It. A dynamic LS
range of more than four to six orders of magnitude would

be required for the amplifying and the data processing SC SH
electronics if It were to be observed directly. Using
alternating current-coupled detection, small values of It
CP PD
that are usually buried in I can be detected by using a
high-pass filter that selectively passes the time-dependent
It and blocks the time-independent I. As a result, the
D
transient infrared intensity is observed separately and hence G
the problem of dynamic range is solved. PA
Substituting equation (2) into equation (1), the absor- MC
MA
bance change may be rewritten as
PC
It DSO
At D log 1 C 3
I
The absorbance change is given by the transient intensity Figure 2. Time-resolved infrared spectroscopic system for pho-
It and the steady-state infrared intensity, I. Therefore, tochemical experiments. LS, infrared light source; SH, shutter;
SC, sample cell; CP, optical chopper; MC, monochromator; G,
two separate scans, one for measuring It and the other grating; D, infrared detector; PA, AC-coupled pre-amplifier; MA,
for measuring I, are necessary to obtain the transient main amplifier; DSO, digital sampling oscilloscope; PC, personal
absorption spectrum. computer; PD, PIN photodiode.
Assuming that It is much smaller than I, equation (3) high spectral resolution, while the disadvantage lies in the
is approximated by: limitation of the wavelength range over which they can
be tuned. A liquid nitrogen cooled CO laser is suitable
It
At 0.43429 4 for studies of metal carbonyls, because the CO stretching
I frequencies of metal carbonyls fall in the wavenumber
For a constant A, the signal It is proportional to I, range 22001600 cm1 , which is well covered by a CO
which is proportional to the intensity of the light source. laser. A CO laser is tunable with a step of 4 cm1
Therefore, the stability of the light source seriously affects and is operated in a single frequency mode so that a
the signal-to-noise ratio (S/N). From equations (2) and (4), monochromator is not needed for mode selection. Infrared
the infrared intensity It with stimulation is expressed as: laser diodes are also suitable for time-resolved infrared
spectroscopy, particularly for gaseous samples that require
It D [1 2.3026At]I 5 high spectral resolution. A multi-pass cell is often used
to increase the pathlength. An infrared diode laser is
If the infrared intensity I has a noise nt, the observed
continuously tunable by changing its temperature and
intensity I t may be rewritten as:
operating current. These lasers cover the wavenumber
I t D [1 2.3026At][I C nt] 6 region 4003300 cm1 with a typical tuning range of a
single diode being 100150 cm1 .
and the apparent transient signal I t and A , both of
which contain noise, are given by
2.5 Stimulation source
I t D It C nt 2.3026Atnt 7
nt nt
A t D At 0.43429 C A 8 A pulsed laser is most commonly used for exciting sample
I I molecules in photochemical studies. A high repetition-rate
With the condition that A 1, the second term dominates laser is suitable for time-resolved infrared spectroscopy,
the noise. If the noise frequency is higher than the cut- provided that the pulse energy is high enough to ensure
off frequency of alternating current-coupling, this term is efficient photoexcitation and that signal processing can
not filtered out and remains to limit the detectivity. For follow the repetition rate (see Section 2.9). Repetition rates
example, even if the light source has a high-frequency in the range of a few hundred hertz are most often adopted.
fluctuation of only 0.01%, a transient absorbance of 105 We use a Spectra-Physics (Mountain View, CA, USA)
is completely masked. On the other hand, the effect of low TRF CW diode-pumped Q-switched neodymium-doped
frequency fluctuations is negligible because it is filtered out yttrium lithium fluoride (Nd : YLF) laser (pulse duration
by AC-coupling in the measurement of It. A 1% low- 7 ns, repetition rate 010 kHz) for high repetition-rate mea-
frequency fluctuation still causes a 1% fluctuation in the surements. Either the second harmonic (523 nm, 200 J),
absorbance change because of the fluctuation in I which is third harmonic (349 nm, 30 J), or fourth harmonic (262 nm,
measured by low frequency modulation by a chopper. 30 J) is used, depending on the sample absorption wave-
Blackbody radiators are widely used as light sources length. A Spectron Laser Systems (Rugby, UK) SL801
for infrared spectroscopy. They are also suitable for time- low repetition-rate pulsed Q-switched neodymium-doped
resolved mid-infrared spectroscopy with the alternating yttrium aluminum garnet (Nd : YAG) laser (pulse duration
current-coupled detection scheme. Incandescent sources 15 ns, repetition rate 50 Hz) has also been used. The har-
cover a wide wavelength region (from 4000 to 400 cm1 ). monics of this Nd : YAG laser pump a dye laser (Spectron
The time-stability of blackbody sources for time-resolved Laser Systems, SL4000G) and a Raman shifter to generate
measurements is very good because the temporal response more extensive wavelength for photoexcitation. A home-
of their temperature, which determines the emitted inten- made Raman shifter (15 atm of H2 or D2 gas enclosed
sity, is slow. We have tested a high-intensity blackbody in a stainless-steel cell, 600 mm length with 10 mm thick
light source made of MoSi2 (Jasco)6 whose operating tem- fused-silica windows) generates wavelengths shifted by the
perature is about 1850 K. This temperature is much higher multiples of the vibrational frequencies of H2 (4160 cm1 )
than that of ordinary Globar light sources, whose typical or D2 (2944 cm1 ).
operating temperature is 1300 K. An excimer laser and a dye laser pumped by an excimer
Continuous-wave (CW) infrared lasers have also been laser can also be used for time-resolved infrared spec-
used as the probe light source for time-resolved infrared troscopy. These lasers provide intense (>100 mJ per pulse)
spectroscopy.7 The advantage of infrared lasers as the and short duration (2030 ns) pulsed outputs at various
probe light source comes from their high intensity and ultraviolet wavelengths (ArF 193 nm, KrF 248 nm, XeCl
308 nm). The possible drawback of the excimer based oxygen often quenches molecules in the triplet excited state.
system is its high discharge noise which may affect seri- Bubbling with an inert gas, such as nitrogen or argon,
ously the ultra sensitive detecting system. prevents the evaporation of volatile solvents and may be
useful for removing the effect of oxygen.
2.6 Flowing sample cell

2.7 Infrared monochromator
In photochemical studies, the sample is often degraded by
A commercial infrared grating spectrometer can be con-
prolonged irradiation of the excitation laser light. Note
veniently modified for time-resolved infrared spectroscopy.
that it typically takes a few hours to take one set of
Operations such as mechanically controlling the slit widths
time-resolved infrared spectra of photochemical samples.
to keep a constant resolution, exchanging gratings, and
In the case of liquid and solution samples, a flow-cell is
exchanging the order-sorting filters are already effectively
conveniently used to prevent sample degradation. Figure 3
programmed in many commercial spectrometers. A practi-
shows a flow-cell (Figure 3a) and a sample supply system
cal problem here is the difficulty in finding a suitable spec-
(Figure 3b) used in our laboratory. The cell can be easily
trometer that is not too old, because most manufacturers
dismantled to clean up the windows where photochemical
have already shifted from manufacturing the grating instru-
reaction products tend to deposit. The thickness of the cell is
ments to Fourier transform infrared (FT-IR) spectrometers.
adjustable from 25 m to 1 mm with lead spacers. Calcium
When a commercial grating infrared spectrometer is
fluoride or barium fluoride windows are used, since they modified for time-resolved spectroscopy, the light source,
transmit visible and ultraviolet light for photoexcitation. sample compartment, spectrometer, detector, and data pro-
The sample can be treated by bubbling an appropriate cessing electronics all need to be redesigned.
gas through the solution in the reservoir. For example, A Globar light source could be used as it is. However,
it is better to use a light source with higher intensity, if
available. We use a MoSi2 infrared light source operated at
1873 K, which is 1.5 to 3 times more intense than the typical
Globar source. The sample compartment is removed and
a beam condenser is fitted for photochemical experiments.
The infrared probe beam and the photoexcitation laser beam
are overlapped on the sample cell in these experiments. In
order to increase the efficiency of photoexcitation, the laser
beam is loosely focused into a spot of a few millimeters
A W1 S W2 diameter on the sample and the infrared beam must be
B1 B2 focused to a smaller size by a beam condenser.
The double-beam configuration of the spectrometer
should be converted to a single-beam configuration. The
sector mirror is removed or fixed so that either the sample or
(a)
the reference beam is always directed to the spectrometer.
Since the element size of a fast response detector is small
Gas out (typically 1 1 mm2 or smaller), the output beam from the
Gas in monochromator has to be focused by an ellipsoidal mirror.
A stepping motor is most often used to rotate the grating.
P
A periodic pulsed current is supplied to a stepping motor
to increase its torque, even if it is stopped. This pulsed
C
current induces electromagnetic noise that is mixed into
the signal due to the transient infrared intensity. The pulsed
current supplied to the stepping motor should be switched
off during the measurement of the transient signal.
(b) R S
Figure 3. (a) Flow-cell for photochemical applications. A, Lead
adapter made of lead; B1 and B2 , stainless-steel cell body; S, 2.8 Detector and pre-amplifier
lead spacer; W1 , window with two holes; W2 , window without a
hole. (b) The sample circulation system. R, Reservoir gas bubbler; Selection of the detector and preamplifier is crucially
S, sample reservoir; P, peristaltic pump; C, sample cell. important because these components largely determine
the performance of time-resolved infrared spectrometers. with an ultra-low noise alternating current pre-amplifier
A highly sensitive and fast-response infrared detector is (NF Electric Instruments, Yokohama, Japan; NF-218K,
required to detect extremely small and high-speed changes 50 Hz100 MHz bandwidth, 55 dB gain, 230 pV Hz1 input
in the infrared intensity. A MCT semiconductor detector is equivalent noise density), and the bias current was applied
most often used. with low noise power supply (NF Electric Instruments,
Currently, two types of MCT detectors are commercially SA-912S1). By using this pc detector and the amplifier,
available, photoconductive (pc) and pv. Generally speaking, a change in absorbance of 3 107 has been detected with
the sensitivity of a pc detector is higher than that of a pv sub-microsecond time resolution.5 The pv detector (Kol-
detector. On the other hand, the response of a pc detector mar Technologies, Inc., KV103-1-A-1-SMA) was cascaded
is slower than that of a pv detector. A typical D* value with an alternating current-coupled preamplifier (Kol-
and response time of a pc detector are 5 1010 cm Hz1/2 W mar KA020-A1, 100 Hz20 MHz bandwidth, 10 kV A1 ,
(peak, 1 kHz) and 0.5 s, respectively. The corresponding 1 nV Hz1/2 input equivalent noise density) and the bias
values for a pv detector are 3 1010 cm Hz1/2 W (peak, current was supplied with a home-made low-noise power
1 kHz) and 0.5 ns. supply. The full width at half maximum of the temporal
Optimization of the bias voltage is important for obtain- response is 60 ns. Considering the 7 ns duration of the laser
ing the fastest temporal response and the highest S/N. The pulse, the response time of the pv detector is estimated to
temporal response becomes faster and the detector noise be about 50 ns. A 1 105 absorbance change is detectable
increases with increasing the bias voltage. Generally, the with a time-resolution of 50 ns. Both detectors are cooled
detector and a bias register are connected in series and a to 77 K with liquid nitrogen.
constant voltage is applied across them. The resistance of The nonlinearity of an MCT detector often causes a
the detector changes with the change in the input infrared serious problem in FT-IR spectroscopy because the inte-
intensity, and therefore the bias voltage applied to the detec- grated intensity of the infrared light emitted by the source
tor changes with infrared intensity. Users need to optimize gives rise to a large centerburst in the interferogram. This
the bias register for the range of the infrared light inten- nonlinearity is not problematic in dispersive infrared spec-
sities actually used in the experiment. Needless to say, an troscopy, in which only dispersed infrared light with much
ultra-low noise power supply is required. lower intensity is detected.
Figure 4 shows the impulse responses of a pc detec- For the wavenumber region above 1800 cm1 , an InSb
tor (Figure 4a) and a pv MCT detector (Figure 4b) mea- detector has significantly higher performance than MCT
sured using a pulsed Nd : YLF laser with 7 ns duration and can be useful for time-resolved infrared spectroscopy.
at 1.047 nm. The pc detector (Kolmar Technologies, Inc., A typical value of the D* of an InSb detector is 1
Newburyport, MA, USA; MPC-14-2-B1) was cascaded 1011 cm Hz1/2 W (peak, 1 kHz).
2.9 Signal processing
A DSO is often used for measuring the fast time-profiles

of transient infrared intensity changes. Since a DSO detects
(a) the whole time-profile at once, it is very suitable for time-
Intensity
resolved spectroscopy. It is also useful to see the transient

signal on an oscilloscope screen in order to optimize
the overlap of the laser and infrared beams. The signal
is first accumulated and averaged on a DSO (typically
(b)
5121024 times) and then transferred to a host computer.
The required DSO sampling rate depends on the time
resolution of the experiment; a 250 MHz sampling rate is
high enough for time resolution of 50 ns. The maximum
possible repetition rate of the experiment is limited by the
0 500 1000 1500 maximum averaging rate of a DSO, which is determined
Time /ns not only by the duty time (active data acquisition time)
Figure 4. Temporal responses of MCT detectors to a pulse of but also by the dead time (during which the DSO prepares
Nd : YLF laser (1.047 m) of 7 ns duration. (a) Photo-conductive for the next duty cycle). For time-resolved measurements
type MCT detector; (b) photo-voltaic type MCT detector. in the nanosecond and microsecond regime, the duty time
is much shorter than the dead time and in practice the the Q-switch noise is eliminated and only the real tran-
maximum repetition rate is determined by the dead time sient signal is extracted, as shown in Figure 5(c). In the
of the DSO. The dead time is not always shown in case of stimulation by an electric field, a similar noise-
manufacturers literature for DSOs; it is important to check elimination procedure is used in which, instead of the
the averaging rate before purchase. The fastest averaging excitation laser, the probing infrared light is turned on
rate currently attainable with a commercial DSO is 700 Hz. and off.
The maximum number of spectra that can be averaged
on a DSO is limited by its memory size. If the required
number of scans is larger than this maximum number, 2.10 Commercial time-resolved infrared
the DSO and the host computer have to communicate spectrometer
more than once. The total observation time increases
with increasing number of times of communication, since Recently, a dispersive time-resolved infrared spectrom-
the DSO cannot accumulate the transient signal during eter became commercially available from Jasco (Model
communication. A fast sampling rate, fast averaging rate, TRIR1000). The spectrometer consists of a high inten-
and large memory size are required for the ideal DSO to sity MoSi2 infrared light source, a beam condenser (a
be used for time-resolved infrared spectroscopy. We use pair of ellipsoidal mirrors with focal lengths, f1 D 73 mm
two different DSOs, a Tektronix (Beaverton, OR, USA) and f2 D 188 mm and diameter, j D 75 mm), a grating
DSA602 and a Tektronix TDS520A. The highest repetition monochromator (Littrow-mounted grating monochroma-
rate is about 200 Hz for both DSOs. tor, f D 200 mm; 300 lines mm1 , 120 lines mm1 , 60 lines
A boxcar integrator is also used for time-resolved mm1 ), and a collection mirror for the detector (ellip-
soidal mirror f1 D 52 mm, f2 D 220 mm, j D 90 mm).
infrared spectroscopy. The maximum possible repetition
The detector, amplifiers, and data processing electronics
rate of a boxcar integrator is much higher than that of
are custom installed. A compact infrared spectrometer for
a DSO, typically 10 kHz or even more. However, only a
testing liquid crystal displays (LCDs) is also commercially
time-resolved spectrum at a particular delay time or a time-
available from NF Electronics (As-401). The system uses
profile of a transient signal at a particular wavenumber is
an infrared filter monochromator or a set of filters to achieve
obtainable with a boxcar integrator. A DSO is more effi-
a high throughput at predetermined infrared wavelengths.
cient than a boxcar integrator because, after one scan of
It probes the electric-field-induced reorientation of liquid
the spectrometer, it provides the whole two-dimensional
crystals with high sensitivity by monitoring a small change
(time and wavenumber) spectral information. We use a
of infrared transmittance. The trace analysis of impurities
Stanford Research (Sunnyvale, CA, USA) SR250 boxcar mixed into LCDs has also been demonstrated.
integrator.
A serious noise problem is often caused by a Q-switched
laser in which a pulsed high voltage of a few kilovolts
is applied to the Q-switch device. This high voltage gen- (a)
erates a strong electromagnetic field and seriously affects
the highly sensitive and fast-response amplifier, though the
laser head and the power supply are shielded with metal
Intensity
covers. Therefore, the laser should be electrically isolated

from the main spectroscopic system as much as possible. (b)
The DSO or boxcar integrator should be optically trig-
gered by the signal from a high-speed PIN photodiode that
monitors the timing of the excitation laser pulses. Even if
these precautions are made, the Q-switch noise can contam-
inate the transient signal. We can eliminate this Q-switch (c)
noise by post data processing as shown in Figure 5, where
Figure 5(a) shows an oscilloscope trace with the excitation
light on and Figure 5(b) shows the trace with the excita- 0 500 1000 1500
tion light off. Note that, although the laser used in this Time / ns
experiment was placed in an adjacent room separated by a Figure 5. Q-switch noise elimination. (a) The signal with laser
steel wall, a large Q-switch noise was still observed. By irradiation. (b) The signal without the laser irradiation (Q-switch
subtracting the features in Figure 5(b) from Figure 5(a), noise only). (c) (a) (b).
2.11 Spectral data analysis with SVD where B, T, and S are (k k), (m k), and (n k)
matrices, respectively. The ith column of the matrix T
Infrared spectra usually contain numerous vibrational fea- is the temporal profile of the ith spectrum which is the
tures that are characteristic of the molecular species ith column of the matrix S. The transformation matrix B
in the sample (molecular fingerprint). In time-resolved is obtained by least-squares fittings of the columns of U
infrared spectroscopy, fingerprints of many different tran- to kinetic model functions (see an example in Section 3).
sient species make varying (time-dependent) contributions. Once B is obtained, the spectral components (columns of
They are overlapped with one another in both the time and matrix S) are automatically determined by equation (11).
wavenumber domains. The observed time-resolved infrared This procedure corresponds to the multi-wavenumber fitting
spectra are, therefore, much more complicated than ordi- of all the observed temporal profiles with a set of common
nary static infrared spectra and often need mathematical kinetic parameters.
post-treatments before individual fingerprints are retrieved
for structural analysis. The SVD method can be quite useful
in analyzing complicated time-resolved infrared spectra.8,9 3 APPLICATIONS
An arbitrary real matrix A with m rows and n columns
(m n; m > n) can be decomposed into three compo- 3.1 Photoexcited states of DMABN
nents as:
A D UWVt 9 DMABN is well known for its extraordinary emission prop-
where U, W, and V are a real column-orthogonal matrix erties; it exhibits two distinct fluorescence bands (dual
m n, a real diagonal matrix n n, and a real orthog- fluorescence) in polar solvents but only one in nonpolar
onal matrix n n, respectively. The diagonal elements solvents. Numerous spectroscopic investigations have been
of W are called singular values and these values are larger reported for DMABN and related compounds, since the
than zero. A set of time-resolved infrared absorption spec- twisted intramolecular charge transfer (TICT) hypothesis
tra consists of two-dimensional intensity data with time and was proposed in order to account for the dual fluorescence
wavenumber axes and is represented as a matrix, in which phenomena.10 Nevertheless, structural information on pho-
m corresponds to the number of points in the time axis and toexcited DMABN has been scant, because time-resolved
n to those in the wavenumber axis. In other words, a col- fluorescence and/or ultravioletvisible absorption spectro-
umn of the matrix U corresponds to a temporal intensity scopies, which were used in these previous studies, give
profile at a particular wavenumber and a column of V to a little information on molecular structure. Structural infor-
spectrum at a particular delay time. mation on photoexcited DMABN should be obtainable by
By SVD of the two-dimensional infrared data, we know time-resolved vibrational spectroscopy, however.
how many spectral components contribute to the observed In 1995, we observed the first time-resolved infrared
spectra by referring to the number of singular values k that spectrum of photoexcited DMABN using the alternat-
are meaningfully larger than the noise. We then disregard ing current-coupled dispersive system described above.5
the n k singular values that correspond to the noise. Figure 6 shows the time-resolved infrared spectra of
Equation (9) is rewritten as: DMABN in n-butanol with and without nitrogen bubbling.
These spectra were obtained from the two-dimensional tran-
A A D U W Vt 10 sient infrared data similar to those shown in Figure 1 by
collecting the spectral data at two different delay times.
where U , W , and V are (m k), (k k), and (n k)
Two transient infrared absorption bands are observed at
matrices, respectively, and the matrix A is the noise-
2096 and 2040 cm1 . The band at 2096 cm1 has a short
eliminated form of A. The temporal matrix U and the
lifetime and does not show any effect of oxygen bub-
spectral matrix V thus obtained are not directly meaning-
bling. The band at 2040 cm1 has a longer lifetime and is
ful, because no physical constraints are involved in the SVD
strongly quenched by oxygen. These bands are assigned to
analysis. These matrices are, however, related to the phys-
the CN stretch bands of the CT singlet state (2096 cm1 )
ically meaningful temporal matrix T and spectral matrix
S by a transformation matrix B. Equation (10) is written and the CT triplet state (2040 cm1 ), respectively. The fre-
using these matrices as quency 2096 cm1 of the CT singlet state is close to the
CN stretching frequency of the benzonitrile anion radi-
A D TSt D U B1 BW Vt cal (2093 cm1 ). This observation is consistent with a full
CT from the dimethylamino group to the benzonitrile moi-
T D U B 1
ety. The large downshift of 120 cm1 on going from the
St D BW Vt 11 ground state to the CT singlet state (2216 ! 2096 cm1 )
is utilized in order to enhance the Raman signal of a

1 particular transient species by tuning the excitation laser
(a) wavelength to its absorption maximum. As a consequence,
0 only the spectral information on this transient species is
obtainable by Raman and CARS measurements. On the
1 contrary, time-resolved infrared spectroscopy looks at all
(b) the transient species, the reactant, intermediate, and prod-
Absorbance
2 uct molecules, without any differentiation. We studied the

photoisomerization of all-trans-retinal using time-resolved
3 dispersive infrared spectroscopy.13 SVD analysis played a
key role in analyzing the complicated time-resolved infrared
4 spectra of retinal.
(c) Figure 7 shows the time-resolved infrared difference
5 spectra of photoexcited all-trans-retinal in cyclohexane.
The SVD analysis indicates that these complicated time-
(d)
6 resolved spectra are composed of two independent spectral
components (A and B) as shown in Figure 8(a). These
7 103
spectra have the temporal profiles (A and B) shown in
2000 2100 2200 2300 Figure 8(b). The temporal profile A shows a fast rise
Wavenumber /cm1 (<50 ns) and a single exponential decay with a time
Figure 6. Time-resolved infrared difference spectra of DMABN constant of 5 s, while the temporal profile B has a step
in butanol.1 (a) 0100 ns after photoexcitation, nitrogen bub- response (fast rise <50 ns and then constant).
bling; (b) 100200 ns, nitrogen bubbling; (c) 100200 ns, oxygen Spectrum A has negative absorption bands at 1665, 1575,
bubbling.
and 965 cm1 , and positive absorption bands at 1600, 1552,
indicates that the transferred electron delocalizes into the 1182, and 940 cm1 . The negative bands correspond to
anti-bonding orbital of the CN moiety. It is notewor- the depletion of the ground-state all-trans-retinal, while
thy that the CT singlet band with 2.2 ns lifetime is still
clearly seen in the time-resolved infrared spectra measured
with 50 ns time resolution. A short-lived species with a few (a)
nanoseconds lifetime can be detected by the 50 ns infrared
1.5 103
system provided that its infrared intensity is high. In fact, (b)
the estimated absorptivity of the 2096 cm1 band is 8.5
times larger than that of the 2216 cm1 band in the ground (c)
state. The large absorptivity means a large dipole moment
change with the change in the CN bond distance. It seems (d)
1.0
Absorbance
that the electron transferred into the benzonitrile moiety

(e)
moves back and forth between the phenyl ring and CN
group in accordance with the CN vibration.
(f)
0.5
(g)
3.2 Photoisomerization of all-trans-retinal
Retinal has four mono-cis isomers, the 13-, 11-, 9-, and (h)
7-cis isomers, in addition to the all-trans isomer. Photoex-

(i)
citation of any of the five isomers results in photoisomer- 0.0
ization to form other isomers with varying distributions.
The mechanism of the retinal photoisomerization is one 1800 1500 1200 900
of the central problems of contemporary photophysical Wavenumber /cm1
chemistry. Time-resolved Raman and coherent anti-Stokes
Figure 7. Time-resolved infrared difference spectra of photoex-
Raman spectroscopy (CARS) have so far been used for cited all-trans-retinal in cyclohexene:13 (a) 01 s; (b) 12 s;
studying the isomerization pathway in the excited electronic (c) 23 s; (d) 34 s; (e) 45 s; (f) 1015 s; (g) 2025 s;
manifolds.11,12 In these studies, the resonance Raman effect (h) 3035 s; (i) 4045 s.
2 4
Absorbance
(a)
x105
0
A
0
4
4
(b)
2 104
105
0
Absorbance
4
Absorbance
4
B
4
0
(c)
105
0
4 105 4
1800 1500 1200 900 4
(a) Wavenumber /cm1 (d)
0
4 105
1.0
1800 1600 1400 1200 1000
Wavenumber / cm1
Normalized amplitude
0.5 A
Figure 9. (a) Observed SVD spectral component A; (b) the syn-
thesized spectrum, (c) C (d) 0.32; (c) the difference spectrum
0.0 (13-cis S0 ) (all-trans S0 ); (d) the difference spectrum (9-cis S0 )
(all-trans S0 ).
1.0
3.3 Electric-field-induced reorientation dynamics

0.5 B of liquid crystals
0.0
Time-resolved infrared spectroscopy is also useful for
0 10 20 30 40 studying the dynamics of the electric-field-induced reori-
(b) Time / s entation of liquid crystal (LC) molecules. The signals to be
detected in the LC experiments are generally much larger
Figure 8. SVD spectral (a) and temporal (b) components of
the time-resolved infrared spectra of all-trans-retinal shown in than those in the photochemical experiments and therefore
Figure 7. step-scan FT-IR spectroscopy is most often used. The dis-
persive method is powerful when very small absorption
changes associated with the primary process of reorientation
the positive bands correspond to the lowest excited triplet are studied14,15 or when many temporal response curves
state of all-trans-retinal. They are assigned to the CDO of a small number of bands are measured under varying
stretching (1600 cm1 ), CDC stretching (1552 cm1 ), CC conditions.16 18
stretching (1182 cm1 ), and CH out-of-plane bending
(940 cm1 ) vibrations. As shown in Figure 9, spectrum
(b) is very well reproduced by the sum of two difference 3.4 Future perspective
spectra; (13-cis all-trans) C (9-cis all-trans) 0.32.
This fact means that the photoexcitation of all-trans-retinal At present, it takes a long time (typically 23 h or even
eventually results in the formation of the 13-cis and 9-cis more) to obtain a two-dimensional time-resolved infrared
isomers in the ratio of 1 : 0.32. Since the rise of spectrum (b) spectrum at high S/N using the alternating current-coupled
is much faster than the decay of the all-trans triplet state, the dispersive spectrometer described here. Use of a ten-times
13-cis and 9-cis isomers are most likely generated via the higher repetition rate will reduce the experiment time by a
singlet excited state. In fact, the most recent femtosecond factor of ten. Use of a ten-element multichannel detector
time-resolved visibleultraviolet absorption spectroscopy9 will also result in a ten-fold reduction of the time of the
has proved that the all-trans to 13-cis/9-cis isomerization experiment. Future developments in optoelectronics tech-
proceeds via the second excited singlet state (S2 ). nology will certainly make these experiments feasible. If
a ten-element multichannel detector is used at a ten-fold 4. T. Yuzawa, C. Kato, M.W. George and H. Hamaguchi, Appl.
higher repetition rate, the experiment time will be reduced Spectrosc., 48, 684 (1994).
to one hundredth. In this case, it will take only a few 5. M. Hashimoto and H. Hamaguchi, J. Phys. Chem., 99, 7875
minutes to obtain a set of full two-dimensional infrared (1995).
data! At that point, time-resolved dispersive infrared spec- 6. T. Kawaguchi, K. Kawasaki and T. Nakagawa, J. Ceram.
troscopy will be much more extensively used in a variety Soc. Jpn., 101, 224 (1993).
of applications. 7. M.W. George, M. Poliakoff and J.J. Turner, Analyst, 119, 551
(1994).
8. W.G. Chen and M. Braiman, Photochem. Photobiol., 54, 905
9. S. Yamaguchi and H. Hamaguchi, J. Chem. Phys., 109, 1397
CT Charge Transfer (1998).
CW Continuous-wave 10. Z.R. Grabowski and J. Dobkowski, Pure Appl. Chem., 55,
DMABN 4-N,N-Dimethylaminobenzonitrile 245 (1983).
DSO Digital Sampling Oscilloscope 11. H. Hamaguchi, H. Okamoto, M. Tasumi, Y. Mukai and
LC Liquid Crystal Y. Koyama, Chem. Phys. Lett., 107, 355 (1984).
LCDs Liquid Crystal Displays 12. T. Tahara, B.N. Toleutaev and H. Hamaguchi, J. Chem.
Nd : AG Neodymium-doped Yttrium Aluminum Phys., 100, 786 (1994).
Garnet 13. T. Yuzawa and H. Hamaguchi, J. Mol. Struct., 352/353, 489
Nd : YLF Neodymium-doped Yttrium Lithium (1995).
Fluoride 14. T. Urano and H. Hamaguchi, Chem. Phys. Lett., 195, 287
SVD Singular Value Decomposition (1992).
TICT Twisted Intramolecular Charge Transfer 15. T.I. Urano and H. Hamaguchi, Appl. Spectrosc., 47, 2108
(1993).
16. N. Katayama, T. Sato, Y. Ozaki, Y. Murashiro, M. Kikuchi,
REFERENCES S. Saito, D. Demus, T. Yuzawa and H. Hamaguchi, Appl.
Spectrosc., 49, 977 (1995).
1. E.B. Baker and C.D. Robb, Rev. Sci. Instrum., 14, 362 (1943). 17. T.I. Urano, S. Machida and K. Sano, Chem. Phys. Lett., 242,
2. F. Siebert, S. Maentele and W. Kreutz, Can. J. Spectrosc., 26, 471 (1995).
199 (1981). 18. S. Machida, T.I. Urano, K. Sano, Y. Kawata, K. Sunohara,
3. K. Iwata and H. Hamaguchi, Appl. Spectrosc., 44, 1431 H. Sasaki, M. Yoshiki and Y. Mori, Langmuir, 11, 4838
(1990). (1995).
Fast and Ultrafast Time-resolved Mid-infrared
Spectrometry Using Lasers
D.C. Grills and M.W. George
1 INTRODUCTION far down as the femtosecond timescale, and their high

power increases the concentration of transients, permitting
Fast time-resolved spectroscopic measurements using lasers easier detection.
are based on the principle of flash photolysis. Since its The use of mid-infrared spectroscopy as a detection
development1 in the late 1940s by Norrish and Porter, this method in flash photolysis experiments is thought to be
technique has played a crucial role in the study of chemical a recent development. However, time-resolved IR (TRIR)
reaction dynamics, allowing the characterization of reactive spectroscopy, a combination of UV flash photolysis and
intermediates and transient species on timescales inacces- fast infrared (IR) detection also has a long history. In 1958
sible to conventional spectroscopy. Flash photolysis is a Tanner and King developed2 a sophisticated IR rapid-scan
pumpprobe technique (see Scheme 1), in which a photo- technique. They used a flash lamp to produce transient
lysis source (the pump) initiates a photochemical process species in gas phase samples. A dispersive IR spectrometer
with an extremely short flash of UVvisible radiation, gen- equipped with a continuously rotating Littrow mirror was
erating a high concentration of transient species. A second used to obtain fast IR spectra on the millisecond timescale.
beam (the probe) passing through the sample monitors any Pimentel and Herr improved3,4 this technique to detect
changes that occur after the flash. Early flash photolysis intermediates such as CF2 and ClCOOH. Their spectrome-
instruments, which operated on the microsecond timescale, ter had scan rates of 1000 cm1 per 100 s and high sensitiv-
employed a gas-discharge flashlamp as the pump and a sec- ity over a wide spectral range (240 m) (Figure 1). Further
ond, smaller flashlamp as the probe, with photographic film improvements5,6 to their set-up (1 cm1 per microsecond
as the detector. at 1 cm1 resolution) allowed the gas phase IR spectra of
CF3 , CH3 and CD3 to be recorded. However, the design of
Pump such a spectrometer proved to be impractical for widespread
applications. It should be noted that although rapid-scan
dispersive spectrometers have not found extensive use,
Probe Sample Detector rapid-scan Fourier transform infrared (FT-IR) is now com-
monplace and very fast rapid-scan FT-IR7 is currently being
Scheme 1. A schematic diagram of the pumpprobe technique. developed, that may allow FT-IR spectra to be recorded
with a time resolution approaching 100 s.
With the advent of modern laser technology, the flash-
lamps were replaced by intense pulsed lasers and the pho-
tographic film by fast UVvisible detectors. Pulsed lasers 2 FAST TRIR SPECTROSCOPY
have many advantages over flashlamps. Their short pulse
lengths permit the detection of shorter-lived transients, as The application of TRIR spectroscopy to the study of
condensed phase reaction mechanisms and transient inter-
John Wiley & Sons Ltd, 2002. mediates on the millisecond (and subsequently micro- and
SM
S poor signal-to-noise ratio and resolution problems. Atmo-
spheric water absorbs strongly below ca. 1950 cm1 and
even modest absorption of the IR beam by the solvent
used can cause a significant decrease in signal-to-noise.
SO M
However, recent developments by Hamaguchi et al. (see
Grating Spectrometers) have shown that these problems
L can be overcome by using a computer controlled scanning
LO dispersive spectrometer, coupled to a MCT detector with
a low noise preamplifier. This allowed the detection of
LR
very small transients (OD 106 , where OD is optical
Figure 1. Detailed diagram of the scanning elements of density) on the nanosecond timescale. It has been used to
Pimentels rapid-scan dispersive IR spectrometer. L, Littrow
mirror; M, exit mirror; S, exit slit. Four methods of scanning study a number of fast photochemical processes including
were available: SM, slit movement; SO, mirror oscillation; LO, the photoexcited states of 4-dimethylaminobenzonitrile12
Littrow oscillation; and LR, Littrow rotation. Pimentel used the and the photocyclization of N,N0 -diphenyl-1-hydroxy-9,10-
latter method. [Figure adapted from Pimentel and Herr (1964).3 ] anthraquinone.13,14 A similar spectrometer has also been
applied to a wide range of organic systems.15 17
Flash tube
Chopper Filter 2
Detector
3 FAST TRIR SPECTROSCOPY USING
wheel
IR LASERS
Globar Sample
Filter 1 Monochromator
infrared Continuous wave (CW) IR lasers provide tunability across
source
the mid-infrared, with a higher photon flux than globars,
Reference Boxcar and their narrow linewidths can reduce considerably the
Amplifier
signal integrator and filter problem of atmospheric water absorption. In the 1980s,
Transient three different laboratories independently built IR laser-
recorder based TRIR spectrometers18 20 to follow transition metal
Figure 2. Schematic diagram of the globar point-by-point TRIR carbonyl photochemistry both in the gas phase and in solu-
spectrometer developed by Siebert and co-workers for condensed tion. Metal carbonyl complexes have very intense n(CO)
phase measurements on the millisecond timescale. [Figure adapted absorptions, which greatly helped these TRIR experiments.
from Siebert et al. (1980).8 ]
nanosecond) timescale followed the construction of an

instrument, by Siebert et al.,8 which combined UV flash 3.1 CW CO lasers
photolysis and a dispersive IR grating spectrometer with a
fast-risetime mercury cadmium telluride (MCT) IR detector The CW CO laser is a good IR source for such TRIR
(Figure 2). In this spectrometer the change in IR transmis- measurements since a large part of the metal carbonyl
sion at one particular IR frequency was measured following n(CO) region can be probed. However, the use of CW
UV excitation. The grating was then used to change the IR CO lasers in pumpprobe type experiments dates back to
probing frequency and the measurement repeated. IR spec- the 1970s, when they were used to monitor the vibrational
tra were built up point-by-point by plotting the change population distributions of CO molecules in gas phase
in IR intensity against frequency for any required time electronic-to-vibrational energy transfer reactions21 24 and
delay. Sieberts original spectrometer was used to monitor following the UV photolysis of small molecules in the gas
the recombination of photodissociated CO from carboxy- phase, e.g. formaldehyde.25
myoglobin on the millisecond timescale. Schaffner and The CO laser is line tunable and does not necessarily
co-workers used9 11 this approach to detect the interme- require a monochromator. The laser cavity must be cooled
diates formed upon the UV photolysis of a number of to provide the desired IR coverage, which depends on the
transition metal carbonyl complexes in alkane solution, with extent of cooling. A temperature of ca. 100 C provides26
a time resolution of ca. 5 s. a range of 20001550 cm1 but a slightly wider tuning
TRIR spectrometers based around a globar IR source, range (up to 2100 cm1 ) can be obtained if liquid nitrogen
such as those discussed above, suffer from the major prob- cooling is used.27 The laser lines of the CO laser are
lem of low IR photon flux at any IR frequency, causing separated by ca. 4 cm1 giving a nominal resolution of
Fast and Ultrafast Time-resolved Mid-infrared Spectrometry Using Lasers 3
8 cm1 . Higher resolution can be obtained20 if 13 CO is also

S
used since the resulting lines fall between those observed
with 12 CO. The main advantage of the CO laser over a
globar is its much higher power (50500 mW for a given MC G
line). The lower resolution of the CO laser-based TRIR
spectrometer is not a major problem for studying metal MI
carbonyl photochemistry in solution since the n(CO) bands H
I
of organometallics are broad compared to the linewidth C CO
UV
of the CO laser (<100 MHz). Thus, overlap of the laser
B
emission and the n(CO) absorption bands usually occurs.
D
Figure 3(a) shows a schematic diagram of the Notting-
ham University CO laser TRIR spectrometer. Due to the
high power of the CO laser, plane mirrors were used
throughout, making alignment of the system simple. A
motorized iris was placed before the sample and was used PG TD O
to control the overall intensity of the IR beam, compensat-

ing for the different powers on different laser lines and also
PC
for solvent absorptions. This spectrometer was used exten- (a)
sively in Nottingham University to study the photochem-
istry of metal carbonyls in solution. One example is shown
in Figure 4 for the UV flash photolysis of [CpFe(CO)2 ]2 UV beam
Iris
in cyclohexane solution, (Cp D h5 C5 H5 ).20 The TRIR
spectrum obtained 5 s after photolysis shows that irra- F3
diation of [CpFe(CO)2 ]2 forms both CpFe(CO)2 (1938
and 2005 cm1 ) and CpFe(m-CO)3 FeCp (1824 cm1 ). The
n(CO) bands due to CpFe(CO)2 decay rapidly such that
the spectrum obtained 25 s following the flash shows Diode Cell
laser focus
that CpFe(CO)2 has decayed fully to regenerate the parent
whilst the band due to CpFe(m-CO)3 FeCp has not altered in
intensity. CpFe(m-CO)3FeCp reacts with photoejected CO G
on a much longer timescale (k 4.5 104 M1 s1 ).
Beam stop
Laser Mono- Detection
3.2 CW CO2 lasers head chromator unit
(b)
CO2 lasers emit in the mid-infrared, being line tunable over
the range, 1090920 cm1 . The narrow tunability of the Figure 3. (a) Schematic diagram of the CO laser TRIR spectro-
meter at Nottingham University. B, beam stop; C, IR solution cell;
CO2 laser and the spectral region of the emission have CO, CW CO IR laser; D, MCT IR detector; G, Ge beamsplit-
limited its application in TRIR spectroscopy to the gas ter; H, HeNe alignment laser; I, iris diaphragm; MC, motorized
phase measurements of small molecules. For example, it chopper; MI, motorized iris diaphragm to attenuate the IR beam;
has been used to monitor28 the photodissociation of N2 F4 , O, oscilloscope; PC, microcomputer for data storage; PG, pulse
the kinetics of pyrolysis of furan,29 and the vibrational generator; S, IR spectrometer for laser wavelength calibration;
TD, transient digitizer; UV, XeCl excimer, Nd : YAG laser or
relaxation of 12 CF3 I/13 CF3 I and O3 .30,31 excimer-pumped dye laser. [Adapted from George et al. (1994).26 ]
The main disadvantage of using CO and CO2 lasers for (b) Optical layout of the diode laser TRIR spectrometer at Not-
TRIR applications is their limited range of tunability. Many tingham University. The laser head contains four complementary
chemical systems have strong vibrations in other regions of lasers covering the range 22501750 cm1 . D1 and D2 , MCT IR
the mid-infrared, which cannot be probed by these lasers. detectors; F1 , F2 and F3 , flip mirrors; G, diffraction grating. F1
Therefore an alternative IR laser source is required. and F2 are used to bypass the monochromator during alignment.
TRIR spectrometers overcomes the problem of spectral

3.3 Tunable IR diode lasers coverage since diodes are available throughout the mid-
infrared region from 3 to 30 m. Their use in high resolution
The use of tunable lead salt (e.g. PbSnSe) semicon- gas phase TRIR measurements is well established. For
ductor IR diode lasers instead of a CO or CO2 laser in example, in 1979 Kim and co-workers reported32 the IR
(a) B
When higher resolution TRIR measurements are required,
0.1 C e.g. in the gas phase, etalons44,45 or other forms of external
B calibration26 may be used.
0.0
0.1
3.4 Color center lasers
A A
(b) C The color or F center laser has been used as an IR
Absorbance
0.05
source in high resolution gas phase TRIR measurements.
In a color or F center laser, electrons are trapped in the
0.00 defects of an insulating crystal, e.g. an alkali halide. When
pumped by another laser, e.g. Kr ion, they can produce
0.05 A
A
A
continuously tunable laser output ranging from 0.8 to 4 m.
0.04 (c) C
Despite their limited mid-infrared coverage, color center
lasers have found extensive use46 49 in the gas phase TRIR
0.02 measurements of small molecules, partly because they can
be tuned to extremely precise frequencies.
0.00
0.02 A A
A 3.5 Difference frequency generation
2000 1950 1900 1850 1800 1750
Another source of CW IR laser radiation for gas phase
Wavenumber / cm1
TRIR measurements is that of difference frequency genera-
Figure 4. TRIR spectra obtained with the Nottingham CO laser
tion (DFG) in nonlinear optical crystals (see later section on
TRIR apparatus, (a) 5 s, (b) 25 s and (c) 1.25 ms after the UV
flash photolysis of [CpFe(CO)2 ]2 in cyclohexane under 1 atm of ultrafast TRIR for a detailed description of DFG). Briefly,
CO. The bands are assigned as follows: A, [CpFe(CO)2 ]2 ; B, in DFG the beams from two visible lasers at frequencies !1
CpFe(CO)2 ; and C, CpFe(m-CO)3 FeCp. Points marked were and !2 are mixed in a nonlinear crystal with the concomi-
recorded with a 12 CO laser and those marked C with a 13 CO tant generation of the difference frequency (!3 D !1 !2 )
laser. [Figure adapted from Moore et al. (1984).20 ]
in the IR region. The IR radiation produced can have an
extremely narrow linewidth (<0.001 cm1 ) and is therefore
spectrum of the UF5 radical in the 600 cm1 region follow- suitable for extremely high resolution TRIR measurements.
ing excimer laser photolysis of UF6 in a 10 m pathlength A range of nonlinear crystals is available, providing com-
cell. Another early example of the use of gas phase diode plete tunability out to 18 m.50 Petek et al.51 reported the
laser TRIR was the detection of methyl radicals at 608 cm1 use of DFG IR lasers in TRIR experiments when they
following excimer laser photolysis of CH3 I.33 IR diode recorded the IR spectra of singlet CH2 after excimer laser
lasers have been used to study a wide range of photo- photolysis of CH2 CO. DFG laser-based TRIR has since
chemical reactions of small molecules in the gas phase.34 38 become an extremely important method of small molecule
This is partially due to their continuous tunability across the gas phase TRIR.52 55
mid-infrared and the ability to use these lasers to perform Other IR lasers have been used for fast TRIR measure-
ultra-high resolution measurements (<0.001 cm1 ). ments. For example, a pulsed HBr TEA (transversely excited
Although the power of a diode laser (0.110 mW) is atmosphere) laser (with three emission lines between 2470
much less than that of the CO or CO2 lasers, there has and 2300 cm1 ) has been used56 to excite a liquid N2 CO
been more extensive use26,39 43 of this IR source in con- mixture, with the resulting CO IR fluorescence being moni-
densed phase TRIR spectrometers (see Figure 3b). IR diode tored with an InSb photovoltaic detector.
lasers operate at cryogenic temperatures. Early systems The limitation of the point-by-point approach for obtain-
required closed cycle He refrigerators (<60 K); however, ing TRIR spectra is that it is extremely labor intensive
many modern versions can now operate using liquid nitro- and the coverage of a large IR range (>200 cm1 ) is an
gen cooling (80120 K). Each diode has a typical tuning arduous task. An alternative approach to IR laser-based
range of 50200 cm1 , which is attained by adjusting the TRIR spectroscopy on the nanosecond timescale is time-
temperature and electrical current through the diode. The resolved step-scan Fourier transform infrared (s2 -FT-IR).
output of the diode laser is highly divergent compared to the (For a detailed description of the s2 -FT-IR technique, see
CO laser and is often multimode, thus requiring both collec- Fast Time-resolved Mid-infrared Spectroscopy Using an
tion optics and a monochromator in the TRIR spectrometer. Interferometer). Figure 5 compares the TRIR spectra of
(150200 cm1 ) is a whole days work. However, the com-

(c) plete excited state IR spectrum of ClRe(CO)3 (4,40 -bpy)2
was obtained59 in less than 1 h using time-resolved s2 -FT-
IR (Figure 5c). Furthermore the s2 -FT-IR spectrum extends
Absorbance
well beyond the displayed wavenumber limits, across the

entire mid-infrared region. There are clear advantages for
using s2 -FT-IR rather than IR lasers for fast TRIR mea-
surements. These include: (i) TRIR data are obtained over
(b) the entire mid-infrared region in one experiment; (ii) high
quality FT-IR spectra can be produced with a high spectral
resolution; and (iii) the experiment is completely computer
controlled and automated, with no need for tuning between
UV shots.
However s2 -FT-IR is not always the ideal detection
method for certain TRIR experiments. For example there
Absorbance
(a)
is considerable debate over which is the ideal IR source for
temperature-jump TRIR measurements of protein folding
kinetics. IR spectroscopy is a particularly important tech-
nique for studying protein structure.60 The position of the
amide I stretching vibration (16101680 cm1 ) has been
2050 2000 1950 1900 shown to be strongly correlated with protein secondary
Wavenumber / cm1 structure because of its sensitivity to hydrogen bonding,
dipoledipole interactions and the geometry of the peptide
Figure 5. (a) Ground state FT-IR spectrum of ClRe(CO)3 (4,40 -
bpy). TRIR spectra of its 3 MLCT excited state obtained in backbone. Dyer and Woodruff have pioneered the applica-
Nottingham, (b) 240 ns after photolysis using a CO laser TRIR tion of fast TRIR spectroscopy together with laser induced
spectrometer () and 100 ns after photolysis using a diode laser temperature-jumps to probe protein folding kinetics. They
TRIR spectrometer () and (c) 100 ns after photolysis using an s2 - studied a wide range of protein folding events including
FT-IR based TRIR spectrometer. [The data in (b) are adapted from
the helix melting of small peptides61 and the fast relax-
the figures in Glyn et al. (1989)57 and Gamelin et al. (1994)58 and
the data in (c) are from Sun et al. (submitted).59 ] ation dynamics of native apomyoglobin62 and the major
cold shock protein of Escherichia coli (CspA).63 We have
the 3 MLCT (metal-to-ligand charge transfer) excited state used a similar experimental approach to monitor the pro-
of ClRe(CO)3 (4,4-bpy)2 where (4,40 -bpy is 4,40 -bipyridyl) tein folding dynamics of ubiquitin.64 In these experiments
obtained in Nottingham University with s2 -FT-IR and a short pulse from a laser whose wavelength corresponds
point-by-point TRIR. The spectrum shown in Figure 5(b) to a near-infrared absorption of the solvent (H2 O or D2 O)
is presented as a combination of both CO and diode laser- causes a rapid temperature jump in the solution. This ini-
based TRIR measurements. The original TRIR spectrum57 tiates the unfolding of the protein which is then monitored
of ClRe(CO)3 (4,40 -bpy)2 was obtained using the Notting- using an IR diode laser in a point-by-point fashion (see
ham University CO laser-based TRIR spectrometer. This Figure 6). In principle s2 -FT-IR would be an ideal probe
measurement showed that the two low frequency n(CO) in such experiments. However, in practice the rapid heat-
bands of ClRe(CO)3 (4,40 -bpy)2 shifted up in energy (ca. ing of the solvent can introduce artifacts during the Fourier
5565 cm1 ) on promotion to the MLCT excited state. transformation. To our knowledge there has only been one
This is because in the MLCT excited state, the Re(I) metal group that has successfully applied s2 -FT-IR to the study
center is formally oxidized to Re(II), causing less p back- of protein folding.65 The question of whether s2 -FT-IR or
bonding into the CO antibonding orbitals and hence an IR laser-based TRIR is the best method for studying laser-
increase in the CO bond strengths. The high frequency induced temperature-jump kinetics depends on the removal
n(CO) band of the ClRe(CO)3 (4,40 -bpy)2 excited state could of artefacts introduced by the initiation process. How-
not be probed, thus demonstrating the limitation of the ever, there are other applications in which IR diode lasers
restricted range (15502000 cm1 ) of the CO laser for may have distinct advantages for TRIR measurements. For
TRIR spectroscopy. A diode laser-based TRIR spectrom- example, laser-based instruments have a much higher IR
eter was used to obtain58 the high frequency n(CO) excited photon flux at any wavelength compared with FT-IR instru-
state band. In Nottingham University the acquisition of ments. Although excellent kinetics can be obtained using
laser-based TRIR measurements over such a spectral region s2 -FT-IR for transients with strong IR bands, the laser-based
C increases in power as well as arrays of lasers covering

L L D
I several spectral regions may be possible. Such develop-
ments may allow CW IR lasers to be used in more broad-
L band applications, thus providing a viable alternative to
R P s2 -FT-IR and the scanning dispersive approach of Ham-
aguchi et al. (see Fast Time-resolved Mid-infrared Spec-
PC
troscopy Using Grating Spectrometers). The potential
Y of using quantum cascade lasers for time-resolved spec-
O
troscopy has been shown by the recent application for flow
Figure 6. Schematic diagram of the apparatus we have used to injection analysis.74
monitor temperature jump initiated protein folding kinetics with
an IR diode laser. C, temperature controlled IR solution cell; D,
MCT IR detector; I, IR diode laser system; L, CaF2 lens; O, digital
storage oscilloscope; P, pre-amplifier; PC, personal computer; R, 4 ULTRAFAST TRIR STUDIES
H2 filled Raman-shifting tube which generates the 1.9 m pump
beam; Y, pulsed Nd : YAG laser. Many photochemical and photophysical events take place
on the sub-nanosecond timescale. In order to study such
approach is more suitable for studying the kinetics of very ultrafast reaction dynamics, CW IR sources are not suitable
weakly absorbing transients or for measurements in strongly and the experiment requires that both the pump and probe
absorbing solvents (e.g. water). Furthermore, the advantage beams are ultrashort laser pulses. Ultrafast spectroscopy is a
of using IR lasers in high resolution measurements is evi- variant of the pumpprobe technique; however, time reso-
dent from the large number of gas phase diode laser TRIR lution is achieved by spatially delaying the probe pulse with
studies in the literature.34 38 respect to the pump pulse (1 ps D 0.30 mm). By repeating
A method of broadband nanosecond laser-based TRIR, the experiment at a series of time delays accurate kinetic
which has been less frequently used, is time-resolved IR information can be obtained. The pulsewidths of the pump
spectral photography (TRISP). This employs the nonlinear and probe pulses should be of the same magnitude or shorter
phenomenon of stimulated electronic Raman scattering in than the desired time resolution.
which broadband pulsed visible lasers are downconverted The IR pulses required for sub-nanosecond TRIR are nor-
into broadband IR pulses in alkali metal vapors such as mally generated from the outputs of commonly available
Cs, Rb or K. After probing the sample, these broadband ultrashort pulsed laser oscillators (typically in the visible or
pulses can be upconverted back to the visible by mixing near-infrared). There are two common methods of ultrashort
with a second narrowband visible pulse in another metal IR pulse generation: stimulated electronic Raman scattering
vapor, permitting their detection photographically66,67 or in gaseous metal atoms75 and frequency conversion tech-
with an optical multichannel analyzer (OMA).68,69 Alterna- niques in nonlinear optical crystals. The latter approach,
tively, the broadband IR probe pulses can be monitored being somewhat simpler and involving only solid-state
directly by dispersing them onto a multichannel IR lin- media, has become the method of choice for ultrafast IR
ear array detector. Pimentel and co-workers70 72 installed a pulse generation.
120-element MCT IR array detector in such a TRIR spec- Two nonlinear frequency conversion techniques are gen-
trometer and used it to record the broadband transient IR erally used to produce these mid-infrared pulses. Optical
spectra of gas phase radicals such as CF3 and C2 F3 fol- parametric generation (OPG) splits a single pump photon
lowing the UV photolysis of precursor molecules. TRIR at frequency, !1 into two photons at frequencies, !2 and
experiments using IR pulses differ from those described so !3 , where !1 D !2 C !3 . !2 and !3 are referred to as the
far, which have all used CW IR lasers. Kinetic measure- signal and idler frequencies, respectively. The phase match-
ments are built up on a point-by-point basis by repeating ing of this process determines which pair of signal and idler
the experiment at a series of time delays between the pump frequencies are produced. Crystals typically used for OPG
and probe pulses. in the mid-infrared include LiNbO3 and KTiOPO4 (KTP).
Recent developments in laser technology73 promise to If the crystal is seeded by injecting a small amount of light
revolutionize the use of CW IR lasers for nanosecond TRIR at one of the desired output frequencies, along with the
applications. Quantum cascade lasers operate at room tem- pump frequency at !1 , the process is termed optical para-
perature and each laser covers a modest spectral region metric amplification (OPA). Two crystals are often used in
(ca. 100 cm1 ) with powers of up to tens of milliwatts. conjunction, the first being used for OPG and the second
The technology behind quantum cascade lasers contin- for OPA. The second type of nonlinear process that is com-
ues to develop and there is every indication that further monly used for the generation of pulses extending further
into the mid-infrared is that of DFG. In DFG, two input ML QS Nd : YAG Diode laser
photons at frequencies !1 and !2 (with !1 > !2 ) combine SHG
to give a photon of frequency !3 , where !3 D !1 !2 .
Typical mixing crystals for the DFG production of mid- IR
infrared pulses include LiIO3 and AgGaS2 . With LiIO3 , IR
pump
pulses can be generated down to ca. 1700 cm1 .76 However, Dye laser 2
the use of AgGaS2 extends this range to <1000 cm1 .77
The use of picosecond IR laser pulses in spectroscopic
Sample
applications has a long history. Early experiments in the Delay line
1970s used the IR pulses as a source of excitation in order to
determine the vibrational relaxation dynamics of molecules probe
Dye laser 1
in solution. A variety of spectroscopic probes (e.g. fluo-
rescence and anti-Stokes Raman scattering) were used to UC crystal
probe the vibrational relaxation.78 81 Picosecond IR pulses
were later used to both excite and probe vibrational relax- UC
ation. For example, Heilweil et al.82 used picosecond IR
pumppicosecond IR probe experiments to directly moni-
Monochromator
tor the vibrational relaxation of the chemisorbed hydroxyl PM tube
groups bound to colloidal SiO2 dispersed in CCl4 . The
TRIR spectrometer used to obtain these data employed Figure 7. Schematic diagram of the ultrafast TRIR apparatus
developed by Hochstrasser and co-workers. ML QS Nd : YAG,
an OPA in which 1.06 m pulses from a mode-locked mode-locked Q-switched Nd : YAG laser; SHG, second harmonic
Nd : YAG laser pumped a chain of three nonlinear LiNbO3 generation; UC crystal, upconverting LiIO3 crystal; !IR , IR probe
crystals, generating pulses at 3678 cm1 . Since these early frequency; !pump , visible pump frequency; !probe , visible gating
experiments many researchers have extended the technique pulse frequency; !uc , upconverted frequency; PM, photomulti-
by using a second, independently tunable IR pulse to per- plier. [Figure adapted from Moore et al. (1987).95 ]
form two color IR pumpIR probe experiments. This has
allowed the vibrational relaxation dynamics of a wide that undergo chemical change upon UV excitation. In their
range of systems to be studied on the pico- and fem- experimental set-up (see Figure 7), a CW IR diode laser
tosecond timescales. These include surface adsorbates,83 was used to continuously probe the sample at one IR
metal carbonyls,84 86 inorganic anions,87 hydrogen-bonded frequency. A UV pulse from a Nd : YAG pumped visible
complexes,88,89 water,90 solutes dissolved in supercritical dye laser initiated the reaction, causing a temporal change
fluids91 and proteins.92 in the IR transmission. A second, delayed visible gate pulse
Picosecond TRIR absorption measurements were first was then combined with the transmitted IR beam in a LiIO3
performed by Elsaesser et al.,93,94 who generated pico- crystal, generating light at the sum of the frequencies of the
second-IR pulses in the range 30003500 cm1 in a two beams. This upconverted visible pulse was detected
LiNbO3 OPA using 1.06 m pulses from a Nd : YAG using a photomultiplier tube at various time delays. The
laser system. However, picosecond-IR pulses in the system was used to probe the photodissociation of CO
range 14003000 cm1 were generated by the difference from hemin and myoglobin carbonyls95 and to study the
frequency mixing in a nonlinear AgGaS2 crystal, of a photochemistry of [CpFe(CO)2 ]2 .96 Hochstrasser and co-
tunable near-infrared dye laser pulse and part of the 1.06 m workers soon extended their technique into the femtosecond
pulse. After probing the sample and a reference cell, regime.97 99 The time resolution of their experimental
the IR pulses were detected by two IR photodetectors. method is determined by the cross-correlation of the pump
Picosecond transient IR spectra of dye molecules were and gating pulses, but since both pulses were derived from
recorded in the n(NH) and n(CO) stretching regions the same laser, the time resolution was limited only by the
following excitation (355 or 532 nm) into their first optical pulse duration. The pulses from their visible dye
electronic excited states. For example, the appearance94 laser were shortened to ca. 200 fs in an optical fiber-grating
of transient n(NH) and n(CO) stretches in the spectra of compressor and amplified to ca. 5 J at 1 kHz in a multipass
2-(2-hydroxyphenyl)benzothiazole provided evidence for dye amplifier. They also began to employ99 a CW CO laser
intramolecular proton transfer in the excited state. as the IR source since its higher power improved the signal-
A different approach was used by Hochstrasser and co- to-noise level. Due to the extremely narrow linewidths of
workers95 to perform the first high repetition rate (700 Hz) CW IR CO and diode lasers, they were now able to perform,
picosecond UV pumpIR probe experiments on systems for the first time, sub-picosecond TRIR investigations with
no loss of spectral resolution. This TRIR spectrometer to be simultaneously monitored over a wide spectral
was used to study a number of ultrafast events including window, enabling much faster data acquisition times.
metal carbonyl photochemistry,99 the photoexcitation of In 1989, Heilweil reported a picosecond IR spectrometer
photosynthetic reaction centers (RCs),100 the vibrational that used multichannel detection (ca. 100 cm1 ).114 Broad-
relaxation101 of N3 , energy flow between solutes and band picosecond IR pulses were generated at 10 Hz in a
solvents102 and the third-order response of liquids following LiIO3 crystal by the difference frequency mixing of 532 nm
UV excitation.103 pulses from a Nd : YAG laser and broadband pulses from a
The technique of CW IR probing with ultrashort visible synchronously pumped R6G dye laser centered at 598 nm.
pulse gated detection for ultrafast TRIR spectroscopy has Previous to this, ultrashort broadband IR pulses had gener-
not so far found widespread applications. Nevertheless ally been obtained by the relatively inconvenient stimulated
other groups104 have used this spectrometer design to electronic Raman scattering of gaseous caesium or barium
monitor the reductive elimination of H2 from a metal atoms in heat pipes.75,115,116 The transmitted IR pulses were
carbonyl complex105 and to probe the MLCT excited states then upconverted to the visible by mixing with 532 nm
of binuclear rhenium(I) polypyridyl complexes.106 pulses in a second LiIO3 crystal, allowing their detection
Since the late 1980s many groups have built pico- and using an OMA.
femtosecond TRIR spectrometers in order to study a wide With subsequent advances in technology they were able
range of ultrafast events.107 113 These spectrometers have to make a number of improvements to their system.117
all been based on variations of the same principle, with most Sub-picosecond broadband IR pulses (100 cm1 , 400 fs)
using the ultrashort pulsed IR probe technique. However, were now produced at 20 Hz by the difference fre-
all of these ultrafast TRIR experiments are narrowband, quency mixing of DCM (4-(dicyanomethylene)-2-methyl-
i.e. they used IR light covering only a very narrow spectral 6-(p-dimethylaminostyryl)-4H-pyran) dye laser pulses at
region, in order to selectively pump a particular transition 658 nm and compressed broadband R6G dye laser pulses
and/or probe a particular IR frequency using single element centered at 580 nm. These IR pulses were then split into
IR detectors. The use of tunable narrowband lasers, to signal and reference and after passing through the sample
obtain complete transient IR spectra over a wide spectral they were both upconverted back to the visible by mixing
range and at multiple pumpprobe time delays, is not with the 658 nm DCM pulses. Detection of these visi-
only an extremely time consuming task but it also suffers ble broadband pulses was then achieved along two tracks
from the possibility of spectral distortions due to long term of a 384 578 pixel focal-plane charge-coupled device
drift. Broadband ultrafast IR spectroscopy is therefore an (CCD) system (see Figure 8). Shot-to-shot normalization
extremely attractive proposition since this allows transients of the signal and reference pulses significantly improved
UV pump
L L
S Mono-
BS C2 F L chromator
BS
BS
and CCD
IR probe (To
computer)
BS F DL
BS
DL
C1 R6G
L
BS
DCM
Figure 8. Schematic diagram of Heilweils femtosecond broadband TRIR spectrometer with CCD detection. The broadband IR probe
pulse is represented as a dashed line. BS, beam splitter/combiner; C, nonlinear LiIO3 crystal; DCM, Nd : YAG pumped DCM dye laser;
DL, optical delay line; F, filter; L, lens; R6G, Nd : YAG pumped R6G dye laser; S, sample cell. [Figure adapted from Dougherty and
Heilweil (1994).117 ]
the signal-to-noise ratio compared with their original setup, P* 3.5 ps

which used only one IR pulse to probe the sample. Heil- P+BA
weil and co-workers went on to use this system to monitor
1 ps
the UV photochemistry117 121 and vibrational relaxation
dynamics85,86 of a number of metal carbonyls in hydro- h
P+HA
carbon solution.
200 ps
Upconversion of IR pulses to the visible is not the
most efficient way of monitoring their transmission through P+QA
a sample and in 1994 Hamm et al.122 performed ultra-
P
fast broadband TRIR spectroscopy using a linear IR array Ground state
detection system. The direct detection of the IR pulses
Scheme 2. A simplified reaction scheme showing the chain of
without the need for an upconversion step provides a signif- electron transfer reactions studied by Zinth and co-workers124
icant increase in signal-to-noise ratio. They had previously which occur, upon photoexcitation, through the chromophoric
described123 a method for the generation of tunable subpi- molecules present in the purple photosynthetic RC, Rhodobacter
cosecond mid-infrared pulses between 2222 and 870 cm1 sphaeroides.
using a Ti : sapphire (Ti : S) laser-based system. Their exper-
imental set-up is shown in Figure 9. The femtosecond Ti : S spectroscopy. As shown in Scheme 2, the absorption of
regenerative amplifier produced 120 fs pulses at 815 nm at light by the RC of purple photosynthetic bacteria initiates
a 1 kHz repetition rate. These pulses were then used to gen- electron transfer reactions from the special pair of bacteri-
erate tunable near-infrared pulses in a dye laser amplifier ochlorophyll molecules (P) along a chain of chromophoric
(8701000 nm) which were mixed with residual 815 nm molecules within the RC. Figure 10(a), (b) and (c) shows
pulses in type I AgGaS2 , generating tunable broadband the TRIR difference spectra obtained 1, 10 and 1000 ps
mid-infrared pulses (400 fs, ca. 65 cm1 FWHM (full width after 870 nm excitation of the RC. The spectra were con-
at half-maximum), 1 kHz). Broadband TRIR spectra were structed by joining together blocks (ca. 65 cm1 ) of TRIR
obtained by splitting the IR pulses into signal and reference spectra, which were collected with a spectral resolution of
and dispersing them onto two 10-element MCT linear IR 38 cm1 using the 10-element MCT array detector. They
array detectors. This system had the advantage that, because are composed of the excited state, P and the intermediates,
it used relatively small linear array detectors, it could be PC HA and PC QA (where HA is bacteriopheophytin-a,
operated at a high repetition rate (1 kHz) compared with the and QA is primary quinone). However, it is not until the
CCD array setup (20 Hz), thus permitting extensive signal double-difference spectra are constructed (Figure 10e and
averaging. In the following years they used this apparatus f) that the differences in band positions between these inter-
to obtain femtosecond TRIR spectra of the bacterial RC, mediates can be appreciated. Understanding these spectral
Rhodobacter sphaeroides,124 the protonated Schiff base of changes involves unraveling a wide range of subtle changes
all-trans retinal125 and of azobenzene.126 We use Zinths in CDO, CC, CN and CH modes over a large spectral
elegant experiments on Rhodobacter sphaeroides to illus- region (1000 to 1800 cm1 ). This ambitious experiment was
trate the power of using an IR array for ultrafast TRIR greatly facilitated by the use of an IR array detector.
Regen. amplifier fs Ti:sapphire laser
TWDL 1 Ge
filter
Reference
Delay
Spectro-
AgGaS2 MCT
graph
IR probe pulse 5-11 m Signal
Chopper 10 nJ
MCT
Variable Spectro-
TWDL 2
delay Sample graph
Excitation pulse
Figure 9. Schematic diagram of the femtosecond broadband TRIR spectrometer developed by Zinth and co-workers, which incorporates
10-element MCT IR array detectors. AgGaS2 , nonlinear crystal; MCT, 10-element HgCdTe IR array detector; TWDL, travelling-wave
dye laser. [Figure adapted from Hamm et al. (1994).122 ]
1666
1270
1240
3
1458
1118
(a)
1532
2
1403
1
1345
1059
1683
1
1 ps (P/P)
1478
1706
3
1550
(b)
1285
1752
1666
1403
2
0 1736
1654
1059
1345
1683
1 10 ps (P+HA/PHA)
3 (c)
D2O
2
Absorbance change
1
1000 ps (P+QA/PQA)
3
1545
1477
1703
(d)
1749
1284
1657
2
1401
1729
0
1739
1346
1059
1686
1
1664
FT-IR (P+QA/PQA)
2
(e)
1706
1478
1297
P+HA
1752
1
1736
1453
1270
1114
1238
1532
1680 1673
2 P
QA HA (f)
1706
1295
1728
0
1228
1376
1667
1633
1
1605
HA QA
1800 1600 1400 1200

Wavenumber / cm1
Figure 10. Broadband TRIR difference spectra obtained (a) 1 ps, (b) 10 ps and (c) 1000 ps after photoexcitation (870 nm) of the
Rhodobacter sphaeroides RC. (d) shows a steady state FT-IR difference spectrum of PC QA /PQA for comparison. The double-difference
spectra, (e) (b a ) and (f) ((c b)), highlight the different absorption features of these intermediates. [Figure adapted from Hamm
et al. (1995).124 ]
The next major development in ultrafast broadband TRIR pulses to ca. 500 cm1 by off-phase matching collinear
spectroscopy came when Arrivo et al.127 replaced their up- difference frequency mixing in a 0.5 cm-thick LiIO3 crystal.
conversion/CCD array detection system117 with a 256 This enabled them to obtain transient spectra over a much
256 element InSb focal-plane IR array detector. They also wider spectral range without tuning the probe laser, albeit
increased considerably the bandwidth of their IR probe still at a low repetition rate (20 Hz). By using different dyes
in the narrowband and broadband dye lasers they tuned the excellent stability of the femtosecond IR pulses (1% root
broadband IR probe pulses into the 3300 cm1 region. A mean square).
2 cm LiIO3 crystal was also used to produce narrowband Although broadband multichannel IR detection methods
IR pump pulses. The system was used to probe the have revolutionized the technique of ultrafast IR spec-
hydrogen-bonding dynamics in a mixture of triethylsilanol troscopy, there have been recent advances in narrowband
and acetonitrile in CCl4 allowing the broadband transient ultrafast TRIR. Most noteworthy is the use by Okamoto
difference spectra of this 1 : 1 acidbase hydrogen-bonded of an optically heterodyned detection method.136 This pro-
complex to be obtained after IR excitation (3550 cm1 ). vided a large increase in sensitivity over conventional nar-
This system has since been used with the InSb array rowband techniques. They have used this method to record
detector and also a 256 256 element MCT array detector the picosecond transient IR spectra of a series of pho-
to probe the ultrafast UVvisible photochemistry of a series toexcited organic molecules, including trans-stilbene,137
of organometallic complexes128 130 and the vibrational 4-(dimethylamino)benzonitrile138 and diphenylacetylene,139
in the region, 10001600 cm1 .
population distributions131 of W(CO)6 following excitation
by linearly chirped picosecond IR pulses.
The difference frequency mixing, in nonlinear crystals, of 5 TWO-DIMENSIONAL IR
broadband visible dye laser pulses with narrowband pulses
SPECTROSCOPY
is a convenient method for the generation of ultrashort mid-
infrared pulses over the 33331000 cm1 region, with spec- One of the most exciting areas in the use of ultrafast IR
tral bandwidths up to several hundred wavenumbers. The lasers is that of multidimensional IR spectroscopy.140,141
dye-based broadband IR pulses are inherently spectrally Multidimensional vibrational techniques that complement
noisy, requiring shot-to-shot normalization and extensive NMR methods are being developed in which multi-
signal averaging. Therefore, there has been a move in recent ple IR pulses probe structural changes that occur on
years away from dye-based lasers towards systems cen- timescales inaccessible by NMR. In a similar fashion to
tered around solid-state Ti : S oscillators/amplifiers, which two-dimensional (2D) NMR techniques such as COSY
provide ultrashort pulses at ca. 800 nm. The advantages (correlated spectroscopy) and NOESY (nuclear Overhauser
of these lasers are that they provide a source of naturally effect spectroscopy), the coupling between different vibra-
short high repetition rate pulses, which are inherently stable tional modes in a molecule can be elucidated from the
and which maintain their near transform-limited broadband IR cross-peaks, with the potential of predicting molecu-
characteristics (>100 cm1 FWHM). lar structures on ultrafast timescales. These experimental
Snee et al. have built such a system,132 which operates methods are still being developed; however, if they reach
at a repetition rate of 30 Hz and uses a 256 256 element their full potential it is envisaged that in the near future
MCT focal-plane IR array detector, similar to the InSb they will be as common as 2D NMR. As an example, we
detector described by Arrivo et al.127 . Their broadband show in Figure 11 Hochstrassers elegant femtosecond 2D
IR pulses (ca. 200 cm1 FWHM) are obtained by the IR spectra142 of the cyclic penta-peptide (cyclo-Mamb-Abu-
difference frequency mixing in a LiIO3 crystal of 810 nm Arg-Gly-Asp). They are plotted as contours which represent
Ti : S pulses and a portion of a white light continuum the transient difference spectra as a function of IR pump
generated from the 810 nm pulses in a sapphire window. frequency. This peptide is small enough that all its amide
These IR pulses are split into signal and reference and after I vibrations are spectrally resolved, which permitted their
transmitting the sample they are focused onto two tracks coupling mechanisms to be examined in detail. Cross-peaks
in the spectra reveal the vibrational couplings which are
(12 200 pixels) of the array detector. This has been used
in good agreement with those calculated from the known
to probe the ultrafast dynamics of SiH activation by metal
structure of the pentapeptide. In the future, such 2D IR
carbonyl complexes132,133 and chlorine atom abstraction by
experiments could be used to directly predict molecular
the Re(CO)5 radical.134 Both Hochstrasser87,92 and Lian135
structures.
also use Ti : S based systems, operating at a 1 kHz repetition
rate, with detection of the generated broadband IR pulses
being achieved with a single, 32-element MCT linear 6 CONCLUSIONS
IR array detector. Their IR pulses are generated by first
obtaining near-infrared signal and idler pulses from an In this article we have described the historical development
800 nm pumped OPA and then by difference frequency of IR laser-based TRIR spectroscopy from the early days
mixing these in a AgGaS2 crystal. No reference detector of CO laser TRIR through to current state of the art broad-
is required in these high repetition rate systems due to the band femtosecond systems. Clearly, on the nanosecond and
Figure 11. (a) Ground state IR spectrum of the pentapeptide studied by Hochstrasser and co-workers (dashed line is an example
spectrum of one of the pump pulses used). (b and c) 2D IR pumpprobe spectra measured at an 800 fs time delay with perpendicular
and parallel pumpprobe polarizations, respectively. (df) Least-squares fits of the experimental data. Colors represent the intensity at
any point. Red is negative and blue is positive. [Figure taken from Hamm et al. (1999).142 ]
slower timescales, CW IR diode laser point-by-point TRIR is currently the method of choice. However, laser and
spectrometers are still extremely popular due to their con- IR detector technology is constantly developing. The use
tinuous tunability and ability to obtain extremely accurate of even larger linear IR array detectors (up to 1024 or
kinetic measurements. With the current rapid developments 2048 pixels long) at high repetition rates, perhaps with
in quantum cascade laser technology, nanosecond TRIR interleaved InSb/HgCdTe detector elements, enabling com-
with CW IR lasers and broadband multichannel detec- plete, high resolution spectral coverage from 1 to 12 m,
tion looks set to be a promising alternative to s2 -FT-IR. can be envisaged.143 Indeed at the Rutherford Appleton
On the sub-nanosecond timescale, TRIR spectrometers Laboratory in the UK we are constructing144 a picosec-
based around ultrashort pulsed solid-state Ti : S oscilla- ond broadband TRIR spectrometer based on a 64-element
tors/amplifiers, employing frequency conversion techniques MCT linear array. The developments in ultrafast laser tech-
in nonlinear crystals to obtain tunable broadband mid- nology that will have an impact on TRIR spectroscopy
infrared probe pulses that are detected by IR array detectors, are harder to predict; however, increased laser stability
will undoubtedly lead to more sensitive measurements and 14. K. Kobayashi, M. Iguchi, T. Imakubo, K. Iwata and
perhaps to broader band applications. Furthermore, new H. Hamaguchi, Chem. Commun., 763 (1998).
laser technology may lead to new experimental approaches 15. S. Srivastava, J.P. Toscano, R.J. Moran and D.E. Falvey,
for ultrafast TRIR measurements. J. Am. Chem. Soc., 119, 11552 (1997).
16. Y.H. Wang, T. Yuzawa, H.O. Hamaguchi and J.P. Toscano,
J. Am. Chem. Soc., 121, 2875 (1999).
ABBREVIATIONS AND ACRONYMS 17. S. Srivastava, P.H. Ruane, J.P. Toscano, M.B. Sullivan, C.J.
Cramer, D. Chiapperino, E.C. Reed and D.E. Falvey, J. Am.
2D Two-dimensional Chem. Soc., 122, 8271 (2000).
4,40 -bpy 4,40 -Bipyridyl 18. T.R. Fletcher and R.N. Rosenfeld, J. Am. Chem. Soc., 105,
COSY Correlated Spectroscopy 6358 (1983).
CW Continuous Wave 19. A.J. Ouderkirk, P. Wermer, N.L. Schultz and E. Weitz,
DCM 4-(Dicyanomethylene)-2-methyl-6- J. Am. Chem. Soc., 105, 3354 (1983).
(p-dimethylaminostyryl)-4H-pyran 20. B.D. Moore, M.B. Simpson, M. Poliakoff and J.J. Turner,
DFG Difference Frequency Generation J. Chem. Soc., Chem. Commun., 972 (1984).
NOESY Nuclear Overhauser Effect Spectroscopy 21. G. Hancock, C. Morley and I.W.M. Smith, Chem. Phys.
OD Optical Density Lett., 12, 193 (1971).
OMA Optical Multichannel Analyzer 22. M.C. Lin and R.G. Shortridge, Chem. Phys. Lett., 29, 42
OPA Optical Parametric Amplification (1974).
OPG Optical Parametric Generation 23. R.G. Shortridge and M.C. Lin, J. Chem. Phys., 64, 4076
RC Reaction Center (1976).
s2 -FT-IR Step-scan Fourier Transform Infrared 24. D.S.Y. Hsu and M.C. Lin, Chem. Phys. Lett., 42, 78 (1976).
TEA Transversely Excited Atmosphere 25. P.L. Houston and C.B. Moore, J. Chem. Phys., 65, 757
TRISP Time-resolved Infrared Spectral Photography (1976).
26. M.W. George, M. Poliakoff and J.J. Turner, Analyst, 119,
551 (1994).
REFERENCES 27. M. Poliakoff and E. Weitz, Adv. Organomet. Chem., 25, 277
(1986).
1. R.G.W. Norrish and G. Porter, Nature (London), 164, 658 28. P. Lavigne, J.L. Lachambre and G. Otis, Opt. Commun., 22,
(1949). 75 (1977).
2. K.N. Tanner and R.L. King, Nature, 181, 963 (1958). 29. P.P. Organ and J.C. Mackie, J. Chem. Soc.-Faraday Trans.,
3. G.C. Pimentel and K.C. Herr, J. Chim. Phys., 61, 1509 87, 815 (1991).
(1964). 30. M. Quack, R. Schwarz and G. Seyfang, J. Chem. Phys., 96,
4. K.C. Herr and G.C. Pimentel, Appl. Opt., 4, 25 (1965). 8727 (1992).
5. A.S. Lefohn and G.C. Pimentel, J. Chem. Phys., 55, 1213 31. P.R. Dahoo, D. Jasmin, P. Brosset, B. Gauthier-Roy and
(1971). L. Abouaf-Marguin, J. Chem. Phys., 108, 8541 (1998).
6. L.Y. Tan, A.M. Winer and G.C. Pimentel, J. Chem. Phys., 32. K.C. Kim, R. Fleming and D. Seitz, Chem. Phys. Lett., 63,
57, 4028 (1972). 471 (1979).
7. C.J. Manning, B.T. Bowie and P.R. Griffiths, Proceedings 33. G.A. Laguna and S.L. Baughcum, Chem. Phys. Lett., 88,
of PITTCON 2000, New Orleans, Oral Presentation, 1010 568 (1982).
(2000). 34. H. Kanamori, J.E. Butler, K. Kawaguchi, C. Yamada and
8. F. Siebert, W. Mantele and W. Kreutz, Biophys. Struct. E. Hirota, J. Mol. Spectrosc., 113, 262 (1985).
Mech., 6, 139 (1980). 35. E. Tiemann, H. Kanamori and E. Hirota, J. Mol. Spectrosc.,
9. H. Hermann, F.W. Grevels, A. Henne and K. Schaffner, 137, 278 (1989).
J. Phys. Chem., 86, 5151 (1982). 36. T. Suzuki, H. Kanamori and E. Hirota, J. Chem. Phys., 94,
10. S.P. Church, F.W. Grevels, H. Hermann and K. Schaffner, 6607 (1991).
Inorg. Chem., 23, 3830 (1984). 37. R.A. Alvarez and C.B. Moore, J. Phys. Chem., 98, 174
11. S.P. Church, H. Hermann, F.W. Grevels and K. Schaffner, (1994).
J. Chem. Soc., Chem. Commun., 785 (1984). 38. J.M. Cronkhite, R.E. Stickel, J.M. Nicovich and P.H. Wine,
12. M. Hashimoto and H. Hamaguchi, J. Phys. Chem., 99, 7875 J. Phys. Chem. A, 102, 6651 (1998).
(1995). 39. R. Hienerwadel, D. Thibodeau, F. Lenz, E. Nabedryk,
13. K. Kobayashi, M. Iguchi, T. Imakubo, K. Iwata and J. Breton, W. Kreutz and W. Mantele, Biochemistry, 31,
H. Hamaguchi, J. Chem. Soc.-Perkin Trans., 2, 1993 (1998). 5799 (1992).
40. T.P. Causgrove and R.B. Dyer, Biochemistry, 32, 11985 64. C.S. Colley, I.P. Clark, S.R. Griffiths-Jones, M.W. George
(1993). and M.S. Searle, Chem. Commun., 1493 (2000).
41. B.D. Wagner, B.R. Arnold, G.S. Brown and J. Lusztyk, 65. J.P. Wang and M.A. El-Sayed, Biophys. J., 76, 2777 (1999).
J. Am. Chem. Soc., 120, 1827 (1998). 66. D.S. Bethune, J.R. Lankard, M.M.T. Loy and P.P. Sorokin,
42. K.L. McFarlane and P.C. Ford, Organometallics, 17, 1166 IBM J. Res. Dev., 23, 556 (1979).
(1998). 67. P. Avouris, D.S. Bethune, J.R. Lankard, J.A. Ors and P.P.
43. R. Paur-Afshari, J. Lin and R.H. Schultz, Organometallics, Sorokin, J. Chem. Phys., 74, 2304 (1981).
19, 1682 (2000). 68. D.S. Bethune, J.R. Lankard, P.P. Sorokin, A.J. Schell-
44. C.B. Dane, R. Bruggermann, R.F. Curl, J.V.V. Kasper and sorokin, R.M. Plecenik and P. Avouris, J. Chem. Phys., 75,
F.K. Tittel, Appl. Opt., 26, 95 (1987). 2231 (1981).
45. C.B. Dane, D.R. Lander, R.F. Curl, F.K. Tittel, Y. Guo, 69. E.W. Kaiser, L. Rimai, E. Schwab and E.C. Lim, J. Phys.
M.I.F. Ochsner and C.B. Moore, J. Chem. Phys., 88, 2121 Chem., 96, 303 (1992).
(1988). 70. M.A. Young and G.C. Pimentel, Appl. Opt., 28, 4270
46. H. Adams, J.L. Hall, L.A. Russell, J.V.V. Kasper, F.K. Tit- (1989).
tel and R.F. Curl, Jr, J. Opt. Soc. Am. B, 2, 776 (1985). 71. M.A. Young and G.C. Pimentel, J. Phys. Chem., 94, 4884
47. J.W. Stephens, W.B. Yan, M.L. Richnow, H. Solka and (1990).
R.F. Curl, J. Mol. Struct., 190, 41 (1988). 72. B.E. Wurfel, N. Pugliano, S.E. Bradforth, R.J. Saykally and
48. B.J. Opansky, P.W. Seakins, J.O.P. Pedersen and S.R. G.C. Pimentel, J. Phys. Chem., 95, 2932 (1991).
Leone, J. Phys. Chem., 97, 8583 (1993). 73. J. Faist, F. Capasso, D.L. Sivco, C. Sirtori, A.L. Hutchinson
49. J.D. DeSain, P.Y. Hung, R.I. Thompson, G.P. Glass, and A.Y. Cho, Science, 264, 553 (1994).
G. Scuseria and R.F. Curl, J. Phys. Chem. A, 104, 3356 74. B. Lendl, J. Frank, R. Schindler, A. Muller, M. Beck and
(2000). J. Faist, Anal. Chem., 72, 1645 (2000).
50. U. Simon, Z. Benko, M.W. Sigrist, R.F. Curl and F.K. Tit- 75. J.H. Glownia, J. Misewich and P.P. Sorokin, Chem. Phys.
tel, Appl. Opt., 32, 6650 (1993). Lett., 139, 491 (1987).
51. H. Petek, D.J. Nesbitt, P.R. Ogilby and C.B. Moore, 76. V.G. Dmitriev, G.G. Gurzadyan and D.N. Nikogosyan,
J. Phys. Chem., 87, 5367 (1983). Handbook of Nonlinear Optical Crystals, Springer, Berlin,
52. C.B. Dane, D.R. Lander, R.F. Curl, F.K. Tittel, Y. Guo, Vol. 64 (1991).
M.I.F. Ochsner and C.B. Moore, J. Chem. Phys., 88, 2121 77. K.L. Vodopyanov, J.P. Maffetone, I. Zwieback and
(1988). W. Ruderman, Appl. Phys. Lett., 75, 1204 (1999).
53. A.H. Hielscher, C.E. Miller, D.C. Bayard, U. Simon, K.P. 78. A. Laubereau, A. Seilmeier and W. Kaiser, Chem. Phys.
Smolka, R.F. Curl and F.K. Tittel, J. Opt. Soc. Am. B, 9, Lett., 36, 232 (1975).
1962 (1992). 79. K. Spanner, A. Laubereau and W. Kaiser, Chem. Phys. Lett.,
54. J.T. Petty and C.B. Moore, J. Chem. Phys., 99, 47 (1993). 44, 88 (1976).
55. F. Sun, A. Kosterev, G. Scott, V. Litosh and R.F. Curl, 80. A. Laubereau and W. Kaiser, Rev. Mod. Phys., 50, 607
J. Chem. Phys., 109, 8851 (1998). (1978).
56. S.R.J. Brueck and R.M. Osgood, Jr, Chem. Phys. Lett., 39, 81. A. Laubereau, S.F. Fischer, K. Spanner and W. Kaiser,
568 (1976). Chem. Phys., 31, 335 (1978).
57. P. Glyn, M.W. George, P.M. Hodges and J.J. Turner, 82. E.J. Heilweil, M.P. Casassa, R.R. Cavanagh and J.C.
J. Chem. Soc., Chem. Commun., 1655 ( 1989). Stephenson, J. Chem. Phys., 81, 2856 (1984).
58. D.R. Gamelin, M.W. George, P. Glyn, F.W. Grevels, 83. R.R. Cavanagh, E.J. Heilweil and J.C. Stephenson, Surf.
F.P.A. Johnson, W. Klotzbucher, S.L. Morrison, G. Russell, Sci., 300, 643 (1994).
K. Schaffner and J.J. Turner, Inorg. Chem., 33, 3246 (1994). 84. M. Iannone, B.R. Cowen, R. Diller, S. Maiti and R.M.
59. X.Z. Sun, S.M. Nikiforov, J. Yang, C.S. Colley and M.W. Hochstrasser, Appl. Opt., 30, 5247 (1991).
George, Appl. Spectrosc., submitted. 85. W.T. Grubbs, T.P. Dougherty and E.J. Heilweil, Chem.
60. J.L.R. Arrondo, A. Muga, J. Castresana and F.M. Goni, Phys. Lett., 227, 480 (1994).
Prog. Biophys. Mol. Biol., 59, 23 (1993). 86. S.M. Arrivo, T.P. Dougherty, W.T. Grubbs and E.J. Heil-
61. S. Williams, T.P. Causgrove, R. Gilmanshin, K.S. Fang, weil, Chem. Phys. Lett., 235, 247 (1995).
R.H. Callender, W.H. Woodruff and R.B. Dyer, Biochem- 87. P. Hamm, M. Lim and R.M. Hochstrasser, J. Chem. Phys.,
istry, 35, 691 (1996). 107, 10523 (1997).
62. R. Gilmanshin, S. Williams, R.H. Callender, W.H. Woodruff 88. H. Graener, T.Q. Yeh and A. Laubereau, J. Chem. Phys., 90,
and R.B. Dyer, Proc. Natl. Acad. Sci. USA, 94, 3709 (1997). 3413 (1989).
63. D.T. Leeson, F. Gai, H.M. Rodriguez, L.M. Gregoret and 89. S.M. Arrivo and E.J. Heilweil, J. Phys. Chem., 100, 11975
R.B. Dyer, Proc. Natl. Acad. Sci. USA, 97, 2527 (2000). (1996).
90. H.J. Bakker, S. Woutersen and H.-K. Nienhuys, Chem. 114. E.J. Heilweil, Opt. Lett., 14, 551 (1989).
Phys., 258, 233 (2000). 115. M. Berg, A.L. Harris, J.K. Brown and C.B. Harris, Opt.
91. D.J. Myers, M. Shigeiwa and M.D. Fayer, J. Phys. Chem. Lett., 9, 50 (1984).
B, 104, 2402 (2000).
116. J.H. Glownia, J. Misewich and P.P. Sorokin, Opt. Lett., 12,
92. P. Hamm, M.H. Lim and R.M. Hochstrasser, J. Phys. Chem. 19 (1987).
B, 102, 6123 (1998).
117. T.P. Dougherty and E.J. Heilweil, Opt. Lett., 19, 129
93. T. Elsaesser, W. Kaiser and W. Luttke, J. Phys. Chem., 90, (1994).
2901 (1986).
118. T.P. Dougherty and E.J. Heilweil, J. Chem. Phys., 100, 4006
94. T. Elsaesser and W. Kaiser, Chem. Phys. Lett., 128, 231
(1994).
(1986).
95. J.N. Moore, P.A. Hansen and R.M. Hochstrasser, Chem. 119. T.P. Dougherty and E.J. Heilweil, Chem. Phys. Lett., 227,
Phys. Lett., 138, 110 (1987). 19 (1994).
96. J.N. Moore, P.A. Hansen and R.M. Hochstrasser, J. Am. 120. T.P. Dougherty, W.T. Grubbs and E.J. Heilweil, J. Phys.
Chem. Soc., 111, 4563 (1989). Chem., 98, 9396 (1994).
97. P. Anfinrud, C. Han, P.A. Hansen, J.N. Moore and R.M. 121. M.W. George, T.P. Dougherty and E.J. Heilweil, J. Phys.
Hochstrasser, Picosecond and Femtosecond Infrared Spec- Chem., 100, 201 (1996).
troscopy with CW Diode Lasers, in Ultrafast Phenomena 122. P. Hamm, S. Wiemann, M. Zurek and W. Zinth, Opt. Lett.,
VI, Springer, Berlin, 442, Vol. 48 (1988). 19, 1642 (1994).
98. P.A. Anfinrud, C. Han and R.M. Hochstrasser, Proc. Natl. 123. P. Hamm, C. Lauterwasser and W. Zinth, Opt. Lett., 18,
Acad. Sci. USA, 86, 8387 (1989). 1943 (1993).
99. P.A. Anfinrud, C.H. Han, T.Q. Lian and R.M. Hochstrasser,
124. P. Hamm, M. Zurek, W. Mantele, M. Mayer, H. Scheer
J. Phys. Chem., 95, 574 (1991).
and W. Zinth, Proc. Natl. Acad. Sci. USA, 92, 1826
100. R. Diller, S. Maiti, G.C. Walker, B.R. Cowen, R. Pippenger, (1995).
R.A. Bogomolni and R.M. Hochstrasser, Chem. Phys. Lett.,
241, 109 (1995). 125. P. Hamm, M. Zurek, T. Roeschinger, H. Patzelt, D. Oester-
helt and W. Zinth, Chem. Phys. Lett., 268, 180 (1997).
101. J.C. Owrutsky, Y.R. Kim, M. Li, M.J. Sarisky and R.M.
Hochstrasser, Chem. Phys. Lett., 184, 368 (1991). 126. P. Hamm, S.M. Ohline and W. Zinth, J. Chem. Phys., 106,
519 (1997).
102. T.Q. Lian, B. Locke, Y. Kholodenko and R.M. Hochstrasser,
J. Phys. Chem., 98, 11648 (1994). 127. S.M. Arrivo, V.D. Kleiman, T.P. Dougherty and E.J. Heil-
103. T.Q. Lian, Y. Kholodenko, B. Locke and R.M. Hochstrasser, weil, Opt. Lett., 22, 1488 (1997).
J. Phys. Chem., 99, 7272 (1995). 128. T.A. Heimer and E.J. Heilweil, J. Phys. Chem. B, 101,
104. T.-G. Ye, C.J. Arnold, D.I. Pattison, C.L. Anderton, 10990 (1997).
D. Dukic, R.N. Perutz, R.E. Hester and J.N. Moore, Appl. 129. T. Jiao, Z. Pang, T.J. Burkey, R.F. Johnston, T.A. Heimer,
Spectrosc., 50, 597 (1996). V.D. Kleiman and E.J. Heilweil, J. Am. Chem. Soc., 121,
105. M. Colombo, M.W. George, J.N. Moore, D.I. Pattison, R.N. 4618 (1999).
Perutz, I.G. Virrels and T.-G. Ye, J. Chem. Soc., Dalton, 17, 130. T.A. Heimer, E.J. Heilweil, C.A. Bignozzi and G.J. Meyer,
2857 (1997). J. Phys. Chem., 104, 4256 (2000).
106. L.C. Abbott, C.J. Arnold, T.-G. Ye, K.C. Gordon, R.N. 131. V.D. Kleiman, S.M. Arrivo, J.S. Melinger and E.J. Heilweil,
Perutz, R.E. Hester and J.N. Moore, J. Phys. Chem. A, 102, Chem. Phys., 233, 207 (1998).
1252 (1998).
132. P.T. Snee, H. Yang, K.T. Kotz, C.K. Payne and C.B. Harris,
107. L. Wang, X. Zhu and K.G. Spears, J. Am. Chem. Soc., 110,
J. Phys. Chem. A, 103, 10426 (1999).
8695 (1988).
108. P.O. Stoutland, R.B. Dyer and W.H. Woodruff, Science, 133. K.T. Kotz, H. Yang, P.T. Snee, C.K. Payne and C.B. Harris,
257, 1913 (1992). J. Organomet. Chem., 596, 183 (2000).
109. T.Q. Lian, S.E. Bromberg, M.C. Asplund, H. Yang and 134. H. Yang, P.T. Snee, K.T. Kotz, C.K. Payne, H. Frei and
C.B. Harris, J. Phys. Chem., 100, 11994 (1996). C.B. Harris, J. Am. Chem. Soc., 121, 9227 (1999).
110. J. Aschenbrucker, U. Steegmuller, N.P. Ernsting, M. Bu- 135. Y. Wang, J.B. Asbury and T. Lian, J. Phys. Chem. A, 104,
back, J. Schroeder and J. Jasny, Appl. Phys. B, 65, 441 4291 (2000).
(1997). 136. H. Okamoto, Chem. Phys. Lett., 283, 33 (1998).
111. R. Laenen and C. Rauscher, Chem. Phys. Lett., 274, 63 137. H. Okamoto, J. Phys. Chem. A, 103, 5852 (1999).
(1997).
138. H. Okamoto, J. Phys. Chem. A, 104, 4182 (2000).
112. J.B. Asbury, H.N. Ghosh, J.S. Yeston, R.G. Bergman and
T. Lian, Organometallics, 17, 3417 (1998). 139. T. Ishibashi, H. Okamoto and H. Hamaguchi, Chem. Phys.
113. C. Chudoba, A. Kummrow, J. Dreyer, J. Stenger, E.T.J. Lett., 325, 212 (2000).
Nibbering, T. Elsaesser and K.A. Zachariasse, Chem. Phys. 140. R.M. Hochstrasser, M.C. Asplund, P. Hamm and N.H. Ge,
Lett., 309, 357 (1999). J. Chin. Chem. Soc., 47, 843 (2000).
141. M.C. Asplund, M.T. Zanni and R.M. Hochstrasser, Proc. Spectroscopy, ed. M.D. Fayer, Marcel Dekker, New York
Natl. Acad. Sci. USA, 97, 8219 (2000). (2001).
142. P. Hamm, M. Lim, W.F. DeGrado and R.M. Hochstrasser, 144. M. Towrie, R. Barton, P. Matousek, A.W. Parker,
Proc. Natl. Acad. Sci. USA, 96, 2036 (1999). A. Stanley, M.W. George and D.C. Grills, Central Laser
143. E.J. Heilweil, Applications of Broadband Transient Facility, Rutherford Appleton Laboratory, UK: Annual
Infrared Spectroscopy, in Ultrafast Infrared and Raman Report 19992000, 204 (2000).
Linear Dichroism in Infrared Spectroscopy
Thierry Buffeteau1 and Michel Pezolet2

1
Universite de Bordeaux I, Talence, France
2
Universite Laval, Quebec, Canada
1 INTRODUCTION reference direction can be given by the stretching direc-

tion of a polymer, the director of a liquid crystal or of
The absorption of infrared (IR) radiation is caused by a monolayer film, the polarization direction of a laser in
the interaction of the electric field vector of the incident photoinduced orientation studies or, more generally, by the
radiation, E, with the transition moment vector, M, of a direction of the external constraint which has produced the
particular vibrational mode (variation of the dipole moment anisotropy in the sample. The anisotropic optical absorption
with the change of the normal coordinates). If linearly of oriented samples, so-called dichroism, can be analyzed
polarized radiation is used to obtain the IR spectrum, the by measuring spectra using light polarized parallel and
absorbance A for the band associated with the vibration perpendicular to the fixed reference direction. This type
is proportional to the square of the scalar product of M of experiment is called infrared linear dichroism (IRLD).
and E: Parameters commonly used to characterize the degree of
optical anisotropy in oriented samples are the dichroic
A/ M E2 / (ME)2 cos2 g 1 ratio, R:
Ajj
where g is the angle between the transition moment and the RD 2
A?
electric field vectors. The summation is over all transition
moments assuming that the absorption from the whole and the dichroic difference A:
sample is just the sum of the absorption of the independent
A D Ajj A? 3
chemical groups (gas phase approximation). Equation (1)
clearly shows that the absorbance is maximum when M and where Ajj and A? are the integrated absorbances of the
E are parallel and is zero when M and E are perpendicular. investigated band measured with the radiation polarized
For isotropic samples, the absorbance is independent of the parallel and perpendicular to the reference direction, respec-
polarization of the incident radiation since the summation tively.
is over all randomly oriented absorbing molecules. In the Since orientation in polymers is a phenomenon of great
case of a sample showing an anisotropy of the orientation technical and theoretical importance,1 several physical tech-
of the transition moments (stretched polymer films, liquid niques have been used for the determination of molecular
crystals, monolayer films, etc.) the absorbance of certain orientation. Following the pioneering work of Ambrose
bands depends on the polarization of the incident radiation. et al.2 and Fraser3,4 on the use of polarized IR spec-
This directional property makes IR spectroscopy very useful troscopy to determine the orientation of fibrous proteins
to study molecular orientation. and polypetides, numerous studies have demonstrated the
The absorption of IR radiation by an oriented sample will high potential of IRLD to characterize oriented polymers
thus be sensitive to the state of polarization of the inci- at the molecular level.5 10 A major advantage of the tech-
dent radiation with respect to a reference direction. This nique is that it can be used to selectively determine the
orientation of specific chemical groups, a given phase or
John Wiley & Sons Ltd, 2002. component in multiphase or multicomponent systems. In
2 Dichroism and Optical Activity in Vibrational Spectroscopy
addition to polymers, IRLD has also been used extensively 2.2 Linear polarizers
for the determination of molecular orientation in liquid crys-
talline materials11 13 and model biological systems.14 24 The function of a linear polarizer is to select the component
Although most measurements of orientation by IRLD have of radiation whose electric field vector lies in a specific
been done in transmission, attenuated total reflection (ATR) direction and to reject all other components. There are
spectroscopy has also been used extensively to measure the several types of linear polarizers for the IR spectral range,35
state of surface orientation.25 30 but the most commonly used are the wire-grid polarizer36
In this article, the experimental details concerning the and the Brewster angle polarizer.37
polarization of IR radiation and the measurement of IRLD
using either a fixed polarizer or a photoelastic modula-
2.2.1 Wire-grid polarizers
tor (PEM) will be discussed. This will be followed by a
comprehensive review of the determination of molecular This device consists of a set of parallel, equally spaced
orientation by IR spectroscopy for systems with uniaxial strips of metal deposited onto a substrate transparent to
and biaxial symmetry. For the sake of simplicity, the dis- the IR radiation, as illustrated in Figure 1. If the sep-
cussion will be limited to the noninteracting molecular gas aration of the strips is less than about one quarter of
phase approximation. For highly optically anisotropic poly- the wavelength, a high degree of polarization of the IR
mers, this approximation is not always sufficient and more radiation which passes through the grid is produced, and
detailed treatment of the theory that takes into account the electric field vector of the transmitted beam is nor-
surface reflection and internal field effects should be mal to the strips. Indeed, being metallic, the wires pro-
used.9,31,32 Finally, typical examples of static and dynamic vide high conductivity for the electric field parallel to
IRLD will be presented. the wires and, consequently, absorption of the electric
field in this direction. On the other hand, because of
the nonconducting spaces between the wires, the elec-
2 POLARIZATION tric field perpendicular to the wires is totally transmitted.
Various wire-grid polarizers are commercially available.
Light waves are electromagnetic in nature, and require at They consist of holographically produced aluminum or gold
least two field vectors for their description: the electric field grids deposited on various substrates (KRS5, polyethy-
vector, E, and the magnetic field vector, H.33,34 The electric lene, germanium, calcium fluoride, barium fluoride, zinc
and magnetic fields are mutually perpendicular and they are selenide, etc.) depending on the desired spectral range.
both perpendicular to the direction of propagation, z. The These polarizers are often used in optical setup because
two field vectors E and H vary in space and in time, and of their small size (35 mm in diameter, 8 mm in thick-
are given by the equations: ness) but their efficiencies are rarely better than 99% in
the mid-IR range.
E D E0 exp ikz !t 4
H D H0 exp ikz !t 5
2.2.2 Brewster angle polarizers
where k is the wave vector and ! D 2pn is the angular For radiation polarized with its electric field lying in
frequency. the plane of incidence, the reflectivity of a dielectric
2.1 Linear polarization
If the planes of vibration of E0 and H0 are fixed in space,

the wave is said to be linearly polarized or plane polarized.
In general, the electric field E is chosen to define the state
of polarization of light waves. This choice is due to the fact
that the perturbation produced by the electric field on the
matter is much greater than that exerted by the magnetic
field.
In the case of natural light, so-called unpolarized light,
the instantaneous polarization fluctuates rapidly in a random
manner. A device for producing linearly polarized light
from unpolarized light is called a linear polarizer. Figure 1. Wire-grid polarizer.
Linear Dichroism in Infrared Spectroscopy 3
Reference direction
z
(a) y
x
E||
E
(b)
Figure 2. Brewster angle polarizer. Figure 3. Measurements at normal incidence. Geometry of in-
coming IR beam at the sample and definition of the reference
direction and x, y, z coordinate system.
surface vanishes at the Brewster angle qB , which is defined
by the equation tanqB D n, where n is the refractive
index of the dielectric medium. On the other hand, the 3.1.1 Measurement at normal incidence
reflectivity is finite for the radiation polarized perpendic- For samples showing uniaxial symmetry with respect to
ular to the plane of incidence. This property has been a reference axis, the recording of the polarized spectra at
used to construct polarizers of either the reflection or normal incidence is the most straightforward way to mea-
the transmission type. However, in the transmission type, sure the IRLD. As shown in Figure 3, the measurements
as the transmitted beam is not totally polarized parallel are made with the incoming light beam (x direction) polar-
to the plane of incidence, it is necessary to use several ized either parallel (z axis) or perpendicular (y axis) to
(six, for example) plates mounted at the Brewster angle the reference direction. The two spectra can be obtained
to eliminate completely the component perpendicular to by a 90 rotation of either the polarizer or the sample. In
the plane of incidence. Such a transmission polarizer is order to avoid any environmental perturbation in the sam-
shown in Figure 2(a). This type of polarizer has the dis- ple compartment of the spectrometer, the rotation of the
advantage of displacing the beam laterally by refraction polarizer or of the sample can be motorized and controlled
in the plates. This can be compensated with a chevron directly by the spectrometer computer. For the first case
arrangement of the plates, as shown in Figure 2(b). Plates (rotation of the polarizer), the same part of the sample
of different materials (silver chloride, zinc selenide, sil- is illuminated by the IR beam to record the parallel and
icon, germanium, etc.) can be used. These polarizers perpendicular spectra. This method, however, requires the
present very high efficiency (better than 99.5% for four recording of two reference spectra, one for each polariza-
plates) and very good transmission, but they are rather tion, since the beamsplitter of the Fourier transform infrared
bulky. (FT-IR) instruments is polarization sensitive. In the second
case (rotation of the sample), only one reference spectrum
is required but it is important to make sure that the sam-
3 MEASUREMENT OF THE LINEAR ple is perfectly homogeneous, since different areas could
DICHROISM be illuminated as the sample is rotated. For both cases, the
polarizer should be the last optical element before the sam-
As discussed in Section 1, the measurement of the IRLD ple or the first one after the sample to avoid depolarization
for a selected absorption band of oriented samples involves by other elements.
the determination of either the dichroic ratio or the dichroic
difference. This can be achieved using static polarization or
3.1.2 Measurement at oblique incidence
modulated polarization.
In some cases (anisotropy in the thickness of thin films,
biaxial samples, etc.), it is necessary to measure the
3.1 Static polarization absorbance of the sample along the direction of propaga-
tion, x, of the incident beam. The absorbance Ax can be
In this case, the direction of the electric field vector is obtained either from measurements using z polarized inci-
selected by a linear polarizer. Two independent spectra dent light at normal and oblique incidence (rotation of the
are necessary to calculate the dichroic ratio or the dichroic sample around the y axis)38 or from measurements using
difference. y polarized incident light at normal and oblique incidence
(rotation around the z axis).39 For the first case, Schmidt40 linear dichroism of oriented samples.44 49 The sensitivity
has shown that the absorbance Ax is given by: of this method comes from the fact that the dichroic
1/2 difference spectrum is measured directly by modulating the
sin2 a polarization of the incident IR beam at high frequency with
Aa 1 Az
n2 a PEM. Thus, dichroic difference spectra can be obtained in
Ax D C Az 6 a few seconds, allowing kinetic measurements over a short
sin2 a
timescale.50,51
n2
where Aa is the spectrum recorded at oblique incidence, a
3.2.1 Optical setup and electronic processing
is the angle between the incident electric vector and the
sample plane and n is the refractive index of the sample. The optical setup and the two-channel processing used in
Although the values of Aa , Az and a are easily obtained, the polarization modulation experiments are shown in Figure 4.
determination of the refractive index of the sample is more At the output of the interferometer, the IR beam presents
difficult.41,42 In practice, a rotation about the z axis allows an intensity modulation at frequency fi D 2VnQ i (typically
one to measure nz , the refractive index in the z direction, between 0.5 and 5 kHz), depending on the wavenumber of
using the relation: the radiation (ni ) and on the velocity (V) of the interferome-
1/2 ter moving mirror. The beam is then polarized with a linear
polarizer and passes through a PEM. This device, consist-
sin2 a
ing of an IR transparent material (zinc selenide, calcium
nz D 2 7
Az fluoride, silicon) coupled to piezoelastic drivers, modulates
1 at a fixed frequency 2fm (between 60 and 100 kHz) the
Aa
polarization of the IR beam between the linear parallel and
and a rotation about the y axis gives ny . In uniaxially perpendicular polarization states with respect to the refer-
oriented samples, the refractive index in the x direction ence direction. After transmission through the sample, the
is equal to that along the y direction. Otherwise, it is double-modulated IR beam is focused with a lens on an IR
necessary to calculate nx by a trial and error procedure.41 detector.
Detailed experimental conditions and procedures have been If the electronic detection bandwidth is up-limited to
described by Koenig et al.41 and Sibilia.43 2fm , the intensity at the output of the detector preamplifier
It is noteworthy that another method to determine Ax is is given by:52
to subtract the absorbances Az and Ay , measured at normal
incidence, from the absorbance of the isotropic sample, A0 , Id D CI0 !i Tjj C T? C Tjj T? J0 f0
given by IDC
Ax C Ay C Az
A0 D 8 C 2Tjj T? J2 f0 cos2!m t 9
3
IAC
Even though this method is simple, in some cases, it is
difficult to obtain perfectly isotropic samples. where C is a constant accounting for the transmittance
of the optical setup and for the detection yield, I0 !i is
the light intensity after the polarizer, Tjj and T? are the
3.2 Modulated polarization parallel and perpendicular transmittances of the sample,
respectively, Jn are Bessel functions of order n, f0 is the
Dichroism measurements using static polarization require maximum dephasing introduced by the PEM and !m is
the sequential recording of two different spectra to obtain equal to 2pfm .
the polarized absorptions parallel and perpendicular to the The detected signal is thus the sum of two signals, IDC
reference direction. However, Ajj and A? are subjected and IAC ; IDC carries only the intensity modulation induced
to instrumental and sample fluctuations occurring between by the moving mirror of the interferometer whereas IAC
their measurements. This static method of measuring IRLD contains in addition the polarization modulation induced
lacks sensibility for the determination of small dichroic by the PEM. As only IAC has a double modulation, the two
effects. Moreover, the time necessary for the separate mea- signals can be easily separated into two different channels
surement of the Ajj and A? spectra can limit the time scale by means of an electronic filter with high- and low-pass
of the measurements to several minutes. outputs. The signal IAC is demodulated with a lock-in
To overcome these problems, the polarization modulation amplifier, using the reference signal of 2fm coming from
technique can be used to measure with high sensitivity the the power supply of the PEM. Then, the two signals are
Two-channel LP filter
ADC
electronics 10 kHz
and computer
LP filter Lock-in HP filter
ADC
10 kHz amplifier 40 kHz
FT-IR Photoelastic modulator HP filter

spectrometer power supply 300 Hz
MCT/InSb
Reference detector
Michelson direction
interferometer
Lens
Parabolic
mirror
Sample Preamplifier
Polarizer
Photoelastic
modulator
Figure 4. Schematic diagram of the optical setup and two-channel electronic processing for polarization modulation infrared linear
dichroism (PM-IRLD) measurements. ADC, analog-to-digital converter; LP, low-pass filter; HP, high-pass filter; MCT, mercury cadmium
telluride.
treated as usual (high- and low-pass filtering, amplification) By combining equations (10), (11) and (12), the dichroic
and are simultaneously sampled and converted by two difference spectrum can be expressed as a function of the
ADCs. The two interferograms are Fourier-transformed and experimental spectrum S and the two calibration spectra:
ratioed to give the experimental spectrum S:
G
C C
jj,? G0 jj,jj S
2GTjj T? J2 f0
SD 10 A D log

13
[Tjj C T? C Tjj T? J0 f0 ] G
Cjj,jj C jj,? C S
where G is an overall constant factor due to the differ- G0
ent amplification and filtering on the two channels. This
experimental spectrum contains Jn f0 factors which can
3.3 Sources of errors in the calculation of the
be determined by proper calibration measurements.47 49
dichroic ratio and the dichroic difference
3.2.2 Calibration measurements A rigorous determination of the dichroic ratio (or the
dichroic difference) of an absorption band is given by
The calibration measurements are performed by replacing
the ratio (or the difference) of its integrated intensities
the sample by a second linear polarizer oriented either
measured with electric field vector polarized parallel and
parallel or perpendicular to the polarizer positioned in front
perpendicular to the reference direction. For most practical
of the PEM. Thus, the polarizer represents a sample that
cases, the ratio (or the difference) of the absorption maxi-
completely transmits one polarization and totally absorbs
mum is measured. This approximation is justified when the
the other one. The respective calibration spectra, derived
studied band is isolated from other bands and has the same
from equation (10) by suppressing Tjj or T? , are given by
shape for the two polarizations. However, when several
the following relations:
bands are overlapped, a band fitting procedure is neces-
2G0 J2 f0 sary to determine the integrated intensity of each individual
Cjj,jj D 11 component. Moreover, in many cases a baseline correc-
[1 C J0 f0 ]
tion is required to eliminate environmental and instrumental
and perturbations on polarized spectra. All these parameters
2G0 J2 f0
Cjj,? D 12 can affect the accuracy of the calculated dichroic ratio or
[1 J0 f0 ]
dichroic difference.
where G0 is the gain associated with the calibration mea- There are several other instrumental effects which can
surements. affect dichroic ratio measurements. First, if the resolution
of the spectrometer is insufficient, the observed integrated (z, y) and (x, y) planes, all the odd-order terms are equal
(or peak) intensity of an absorption band is smaller than to zero and equation (15) becomes:
the true one. As a consequence, the observed dichroic ratio
is lower than the true one. The difference is very small
1
C
jmj
even even
Nq, f D Qm cos q cosmf 18
when the dichroic ratio is close to 1, but can be impor- D0 mD
tant for very high or very low values of the dichroic ratio.
Second, the polarization efficiency of wire-grid polarizers where the summations are over even and even m. The
results in polarization leakage which makes the measured expansion coefficients Qm are obtained by averaging over
dichroic ratio closer to unity than the true one. A quanti- the orientation distribution of all transition moments in the
tative analysis of the effect of polarization leakage on the sample:
difference between the observed dichroic ratio and the true 2p p
1
dichroic ratio has been made by Zbinden.5 More recently, Qm D Nq, fm cos q cosmf sin qdqdf
2p 0 0
Myers and Cooper53 have calculated the effect of the polar- 19
izer efficiency on the orientation functions of polymers.
Finally, the linearity of the IR detector is also important
when the polarized components of an absorption band are 4.1.1 Uniaxial orientation
very different. The above mathematical treatment can often be simplified
by the presence of transverse isotropy of the distribution
of the molecular axis in the samples, i.e. isotropy in the
4 MOLECULAR ORIENTATION (x, y) plane (uniaxial symmetry). For distributions of such
4.1 Characterization of the molecular orientation a uniaxial symmetry, the distribution Nq can be char-
by a series of Legendre polynomials acterized by using a series of Legendre polynomials,57,58
P cos q D P0 cos q:
The orientation of a vector, such as a molecular axis c, can 1
2 C 1
be described with respect to a Cartesian coordinate system Nq D hP iP cos q 20
(x, y, z) by using the Euler angles q and f. The angle q D0
2
describes the inclination of c with respect to z (polar angle) where the normalization term 2 C 1/2 ensures the con-
and f describes the position of its projection in the (x, vergence of the series, while the coefficients hP i are
y) plane with respect to the y axis (azimuthal angle). The the orientation factors determined experimentally. The first
orientation distribution function of the transition moments even-order Legendre polynomials are given by:
in samples showing orientation anisotropy with respect to
the angles q and f, Nq, f, should satisfy the following P0 cos q D 1 21a
normalization condition, 3 cos q 12
2p p P2 cos q D 21b
2
Nq, f sin qdqdf D 1 14
0 0 35 cos4 q 30 cos2 q C 3
P4 cos q D 21c
and can be expressed as a series expansion into spherical 8
harmonics:54 56 All spectroscopic methods used for the characterization

1
C of molecular orientation allow the calculation of the hP2 i
Nq, f D Qm m
cos q expimf 15 coefficient. Polarized Raman scattering and fluorescence
D0 mD spectrocopies also give a hP4 i value, while coefficients up
where m to hP8 i can be obtained by NMR spectroscopy. Only wide-
are the normalized Legendre polynomials defined
by: angle X-ray diffraction allows the calculation of all hP i
coefficients, because the shape of the orientation distribu-
2 C 1 m! 1/2 m
m D P cos q 16 tion of crystal planes is measured directly by this technique.
2 C m!
Therefore, IRLD allows only the determination of the hP2 i
and Qm are the coefficients of the expanded series given by: coefficient, that is often called the order parameter or the
orientation function. For an isotropic sample, hP2 i is equal
1 2 C 1 m! 1/2
Qm D Fm 17 to zero. For an oriented sample, hP2 i D 1 if all transition
2p 2 C m! moments are perfectly oriented parallel to the reference
where Fm are the generalized orientation factors. If we direction, z, while hP2 i D 0 and 0.5 for a perfect ori-
assume that Nq, f is symmetric with respect to the (z, x), entation at q D 54.7 and 90 , respectively.
z equation (24) is only apparent and depends strongly on

the shape and width of the orientation distribution, which
c
unfortunately cannot be determined by IR spectroscopy.59,60
Calculations of molecular tilt angles using IR spectroscopy
M
thus require a precise knowledge of the nature of the inves-
tigated systems.
E|| y According to equation (1), the integrated absorbance
measured with the radiation polarized parallel to the ref-
E erence direction, Ajj D Az , is proportional to hcos2 gi, while
x that measured with the radiation polarized perpendicular
to the reference direction, A? D AX D AY , is proportional
Figure 5. Geometric representation for the nested axially sym-
metric distributions of orientation of the transition moments M to 12 hsin2 gi. Consequently, the orientation function for the
about the molecular axis c (b angle) which is distributed around transition moments is given by:
the reference direction z (q angle). The distribution of M about
the z axis is characterized by the g angle. The geometry of the 1 2 R1
P2 hcos gi D 3 cos g 1 D
incoming IR beam is also represented. 2 RC2
Ajj A? A
One important advantage of using Legendre polynomi- D D 25
Ajj C 2A? 3A0
als to define orientation averages is that it is easy to relate
nested distributions of orientations. For a system composed where the dichroic ratio R, the dichroic difference A,
of several nested orientation distributions, each of these and the average or isotropic absorbance A0 are defined
showing cylindrical symmetry, the Legendre addition the- in equations (2), (3) and (8), respectively. As seen in
orem can be used to calculate the relation between the equation (25), the second-order orientation function for the
global hP i coefficients obtained experimentally and the distribution of transition moments is thus simply defined as
coefficients characterizing each distribution. Let us con- the normalized dichroic difference. The orientation function
sider the case depicted in Figure 5 where the transition for the molecular axis can be calculated by combining
moment, M, makes an angle b with the molecular axis c, equations (23) and (25):
which itself makes an angle q with the z reference axis. If
1 2
both the distributions of M with respect to c and of c with P2 hcos qi D 3 cos q 1
2
respect to z exhibit cylindrical symmetry, then the distribu-
tion of M with respect to z, characterized by the angle g, R1 2
D 26
also possesses this symmetry. Therefore, from the Legendre RC2 3 cos2 b 1
addition theorem, For samples showing anisotropy of orientation in the
thickness rather than in the plane of the film, such as thin
P2 hcos gi D P2 hcos biP2 hcos qi 22
films of fatty acids, phospholipids and model biological
Since the angle b is generally considered as fixed, i.e. the membranes, it is necessary to perform dichroic measure-
orientation distribution of M about c is infinitely narrow, ments at oblique incidence in order to determine the orien-
P2 hcos bi D P2 cos b and the molecular axis can be char- tation function. For such cases, the dichroic ratio measured
acterized using the following equation: at an angle of incidence a between the incident electric vec-
tor and the sample plane, Ra , is related to the orientation
P2 hcos gi function of the transition moments as follows:15,18,20
P2 hcos qi D 23
P2 cos b
sin2 a 3P2 hcos gi
For an infinitely narrow distribution of orientation of the Ra D 1 C 27
n2 1 P2 hcos gi
molecular axis with respect to the reference direction z,
i.e. when all molecules display the same orientation with where n is the refractive index of the sample and P2 hcos gi
respect to z, it is possible to calculate the molecular tilt is the orientation function of the transition moments with
angle q from P2 hcos qi: respect to the normal of the film. In order to improve the
precision of the orientation determination, the dichroic ratio
1 2P2 hcos qi C 1
q D cos 24 is generally measured at several angles of incidence and
3 the orientation function is calculated from the slope of
It is important to stress that if the distribution of orienta- the straight line obtained by plotting Ra as a function of
tion is not infinitely narrow, the angle q calculated using sin2 a.
4.1.2 Biaxial orientation depending on the orientation anisotropy of the studied

material. In the second part, examples of the dynamic mea-
In some cases, the projection of the transition moments
surement of IRLD using either a rotating polarizer or a
is not evenly distributed in the (x, y) plane. For such a
PEM will be presented. This section is not meant to be a
biaxial distribution of orientation, the generalized orienta-
comprehensive review of the literature on IRLD, but rather
tion factors Fm have been reported by Nomura et al.56 for
to present to the reader some examples that illustrate the
normalized Legendre polynomials and by Ward et al.9,61
potential of IRLD for orientation measurements.
for nonnormalized Legendre polynomials. The two sets of
coefficients are slightly different, except the F0 orientation
factors which are those defined for uniaxially symmetric
5.1 Static IRLD
systems. More recently, Kaito et al.39 have shown that
biaxially symmetric systems can be easily characterized
5.1.1 At normal incidence
considering the two orientation functions fq and ff , which
are given by: During the last three decades, the most widely spread appli-
cation of static IRLD has been the determination of molecu-
1 2
fq D 3 cos q 1 28 lar orientation in polymeric systems; for further details, the
2 reader should refer to the references cited in the follow-
hcos 2f sin2 qi ing books and review articles.1,5 7,62 64 Due to its intrinsic
ff D 29
hsin2 qi nature, IR spectroscopy is particularly well suited to study
For an isotropic sample, both fq and ff are equal to multicomponent systems such has polymer blends, copoly-
zero. If all molecules are oriented with their molecular axis mers and semicrystalline polymers because the orientation
oriented parallel to the z direction, fq D 1 and ff D 0. On of the different components of these systems can be deter-
the other hand, the complete orientation along the y and x mined independently. In addition, the high sensitivity and
axes leads to fq D 0.5 and ff D 1, and fq D 0.5 and photometric accuracy of modern FT-IR spectrometers has
ff D 1, respectively. revitalized the utilization of IRLD in polymer research by
In order to determine fq and ff , it is necessary to decreasing significantly the measurement time.
determine the three components of the absorbance, Ax , An example of IRLD measurement is shown in Figure 6,
Ay and Az . While the absorbances Ay and Az can be where are reported the IR dichroic spectra of uniaxially
measured at normal incidence, the determination of Ax oriented atactic polystyrene (PS) stretched at 100 C to a
requires measurements at oblique incidence as described draw ratio l (ratio of the stretched length to the initial
in Section 3.1.2. The absorbances Ax , Ay and Az are related one) of 5.7.65 As can be seen, several bands due to either
to fq and ff as follows: the phenyl ring or the backbone of the polymer show a
strong dichroism. This is the case, for example, for the
Ax / hsin2 f sin2 qi D A0 1 fq 1 ff 30 bands at 1028, 906 and 540 cm1 , associated with the n18a
in-plane CH bending of the aromatic ring and the n17b and
Ay / hcos2 f sin2 qi D A0 1 fq 1 C ff 31
n16b out-of-plane ring modes, respectively, and the band at
Az / hcos2 qi D A0 2fq C 1 32 2850 cm1 due to the CH2 symmetric stretching vibration
of the main chain. The variation of the dichroic ratio of
where A0 is the isotropic absorbance defined in equa- these bands was studied as a function of temperature and
tion (8). From equations (3032), the orientation functions draw ratio. The results obtained at 110 C, which is above
fq and ff can be calculated as: the glass transition temperature (Tg ) of the polymer, are
Az /A0 1 shown in Figure 7. As expected, the dichroic ratio of the
fq D 33 bands either increases or decreases steadily with the draw
2
ratio depending on the value of the angle b between the
and
Ay Ax transition moments of the vibrations and the chain axis.
ff D 34 To determine the orientation function of polymer chains
Ax C Ay
using equations (25) and (26), conformationally insensitive
bands with known b-angle must be used. Good results were
5 APPLICATIONS obtained from both the 1028 and 2850 cm1 PS bands using
values of b of 90 and 70 , respectively. Comparison of
In this section, typical examples of the use of IRLD will the results obtained for the 1028 cm1 band with those
be presented. The first part deals with the static measure- of the 540 cm1 band has revealed that the latter could
ment of IRLD using either normal or oblique incidence, not be used to determine the orientation of the PS chains
1.0
Absorbance
0.5
0 3000 2800 1500 1000 500

Wavenumber / cm1
Figure 6. IR spectra of uniaxially oriented atactic PS stretched at 100 C to a draw ratio of 5.7 recorded with the electric field of the
radiation polarized parallel (- - - ) and perpendicular ( ) to the stretching direction. [Reproduced by permission of John Wiley &
Sons, Ltd. from B. Jasse and J.L. Koenig, J. Polym. Sci., Polym. Phys. Ed., 17, 799 (1979).]
1.4
1.3
1.2
1.1
1.0
R
0.9
0.8
0.7
1 2 3 4 5 6 7 8 9 10 11

Figure 7. Effect of the draw ratio on the dichroic ratio R for atactic PS stretched at 110 C for the 540 (), 906 (), 1028 () and 2850
() cm1 bands. [Reproduced by permission of John Wiley & Sons, Ltd. from B. Jasse and J.L. Koenig, J. Polym. Sci., Polym. Phys.
Ed., 17, 799 (1979).]
since it is sensitive to the amount of trans bonds, which since it allows the determination of the angle b for a given
increases with draw ratio. Finally, the orientation of the vibration when it is not known.66,67
phenyl rings with respect to the chain axis was determined
from the 906 cm1 out-of-plane mode, which is conform-
5.1.2 At oblique incidence
ationally insensitive and possesses a transition moment
that is perpendicular to the plane of the ring. As shown Linear dichroism at oblique incidence has been used mostly
in Figure 8, an angle of about 34 was determined from to determine the orientation of phospholipids, peptides and
equations (25) and (26) by plotting the orientation function proteins in model and natural biological membranes. Since
calculated for both the 1028 and 2850 cm1 bands as a the phospholipid molecules in membranes are arranged in
function of the ratio R 1/R C 2 obtained experiment- a bilayer structure in which the acyl chains of the lipids are
ally from the 906 cm1 band. This procedure is quite useful oriented nearly perpendicular to the plane of the membrane,
Amide I Amide II
= 34 O C O C
C O CH3 P O P O C
0.25 A
H
B
0.20
C
P2(cos )
0.15 D
E
0.10
H
2000 1800 1600 1400 1200 1000 800 600

(cm1)
0.05
Figure 9. Polarized absorption spectra of dried purple membrane
from 600 to 2200 cm1 . A polarizer was placed before the sample
in either vertical (Ajj ) or horizontal (A? ; denoted H) polarization
and the sample tilted at an angle a relative to the vertical
0 polarization (rotation around the x axis). Spectra A to E were
0.05 0.10 0.15 obtained for a D 0, 15, 30, 45 and 60 , respectively. [Reproduced
R 1 by permission of the Biophysical Society from K.J. Rothschild
R + 2(906) and N.A. Clark, Biophys. J., 25, 473487 (1979).]
Figure 8. Determination of the angle b (a in the figure) between
the normal to the plane of the phenyl ring and the chain axis of and Clark.18 Membranes were oriented preferentially in
atactic PS. IR measurements were done on samples stretched at the plane of zinc selenide slides by drying a suspension
100 C for the 1028 () and 2850 () cm1 bands. [Reproduced of membrane fragments. The polarized spectra (Ajj and
by permission of John Wiley & Sons, Ltd. from B. Jasse and A? ) obtained between 600 and 2200 cm1 for angle of
J.L. Koenig, J. Polym. Sci., Polym. Phys. Ed., 17, 799 (1979).] incidence a of 0, 15, 30, 45 and 60 are reproduced in
Figure 9. The two major peaks at 1662 and 1545 cm1
determination of the tilt angle of phospholipid acyl chains
are assigned to the amide I (CDO stretch) and amide
or of transmembrane proteins or peptides by transmission
II (in-plane NH bend) vibrations of the highly helical
IR spectroscopy is only possible at oblique incidence.
polypeptide backbone of the protein.68 As seen in Figure 9,
Akutsu et al.15,17 have pioneered the use of IRLD at oblique
the dichroic ratio Ra of the amide I band depends strongly
incidence to study the conformation and orientation of
on the angle of incidence indicating that the transition
phospholipid multilayers. Between 125 and 150 layers of
moments in the purple membrane have an anisotropic
L-a-dipalmitoylphosphatidylethanolamine were transferred
orientation. Figure 10 shows that the linear relationship
on solid substrates by the LangmuirBlodgett technique
between Ra and sin2 a predicted from equation (27) is
and studied by transmission IR spectroscopy. The results
well followed for the amide I, II and A bands. Least
obtained have shown that the hydrocarbon chains are tilted square fits of the results of Figure 10 yielded values of
from the bilayer normal by approximately 15 and that the the orientation function P2 hcos gi of the transition moments
polar groups are oriented parallel to the plane of the built-up with respect to the bilayer normal of 0.570, 0.083 and
film. More recently, IRLD at different angles of incidence 0.488 for the amide A, II and I vibrations, respectively.
has also been used to study the orientation of phospholipid In order to calculate the orientation of the a-helices of
molecules with unsaturated hydrocarbon chains in multi- bacteriorhodopsin, the Legendre addition theorem was used
layered films obtained by spontaneous alignment of aqueous assuming that the protein is entirely a-helical and that the
dispersions between calcium fluoride windows.24 orientation distribution function is determined by a set of
The orientation of the a-helices of bacteriorhodopsin, the three nested axially symmetric distributions:
single protein of the purple membrane from Halobacterium
halobium, has been thoroughly investigated by Rothschild P2 hcos gi D P2 hcos aiP2 hcos biP2 hcos qi 35
time-resolved techniques (TRS) to follow polymer deforma-

tion under oscillatory strain.72,73 Finally, the dynamics of
3 polymer orientation and relaxation have also been studied
either by recording sequentially parallel and perpendicular
polarized spectra7,8 or by using the polarization modula-
tion technique.49,50 Since the DIRLD and TRS techniques
are covered elsewhere in this volume, only the two last
2 techniques will be discussed here.
R
5.2.1 Rheo-optical infrared spectroscopy of

polymers
1
Rheo-optical techniques combine the simultaneous mea-
surement of stress, strain and an optical parameter (bi-
refringence, light scattering, X-ray diffraction or IR spec-
troscopy) as a function of time. Siesler has elegantly
demonstrated that rheo-optical FT-IR spectroscopy is a
0 0.1 0.2 0.3 powerful technique to study the molecular mechanism of
sin2 polymer deformation and relaxation.7,8,74,75 This technique
Figure 10. Effect of the angle of incidence, a, on dichroic ratio, is based on the simultaneous measurement of the dichroic
R, for the amide I (), II (r) and A () vibrations of dried spectra and stressstrain diagrams during the deformation,
purple membrane showing the expected linear dependence on recovery and relaxation of the polymer using a miniatur-
sin2 a. [Reproduced by permission of the Biophysical Society
ized computer-controlled stretching device. Spectra with
from K.J. Rothschild and N.A. Clark, Biophys. J., 25, 473 (1979).]
incident radiation polarized parallel and perpendicular to
where q is the angle between the helix axis and the bilayer the stretching direction are recorded sequentially using a
normal, b is the angle between the transition moment and pneumatically rotatable polarizer.
the helix axis and a is the angle between the bilayer plane Although rheo-optical FT-IR spectroscopy has been
and the supporting crystal (membrane mosaic spread that applied to several polymeric systems such as thermoplas-
takes into account that the membranes are not perfectly tics, elastomers74,75 and liquid-crystalline polymers,76 the
parallel to the crystal surface). Using 0.9 for the orientation usefulness of this technique can best be illustrated from the
function due to the mosaic spread, the average orientation study of the segmental orientation of a polyurethane elas-
of the a-helices of bacteriorhodopsin was found to be in tomer with alternating soft and hard segments. In the solid
a range less than 26 away from the bilayer normal, in polymer, the hard segments tend to be associated with each
agreement with electron microscopy results. Such a trans- other through hydrogen bonding and form domains in the
membrane orientation of the a-helices was also determined mobile soft segment matrix, resulting in a two-phase sys-
by IRLD at oblique incidence for the photoreceptor mem- tem with fibrillar and lamellar hard segments. The separate
brane protein rhodopsin19 and for the more heterogeneous hard segment domains crosslink the linear polymer primary
proteins of photosynthetic membranes.20 chains and produce a network that gives the elastic proper-
ties to the polymer. The dichroic spectra of such a polymer
recorded during the elongation of the film up to 600% strain
5.2 Dynamic infrared linear dichroism are shown in Figure 11 along with the assignment of the
most relevant absorption bands. The applied strain rate was
Different approaches have been used to follow the time 115% min1 and 12-scan spectra were recorded at 300 K in
evolution of IRLD. Noda et al.69 71 have developed the 9-s intervals with light alternatingly polarized parallel and
dynamic infrared linear dichroism (DIRLD) technique, perpendicular to the stretching direction.
where a small-amplitude oscillatory strain (ca. 0.1% of Films subjected to successive elongationrecovery cycles
the sample dimension) with acoustic-range frequency is increasing by 50% strain intervals up to 500% were studied
applied to a thin polymer film. The submolecular in-phase by IR spectroscopy. For the region of the mechanical treat-
and out-of-phase response of individual chemical groups ment up to 150%, almost no hysteresis was observed on the
induced by the applied dynamic strain is then monitored stressstrain diagrams. However, substantial irreversible
by using polarized IR spectroscopy. Hsu et al. have used effects were observed at higher elongation, resulting in
12
2 5 II
II
4
3.0 1 3
Strain (%) 1.5 Strain (%)
0 0
2.0 1.0
200
200
1.0
Absorbance
0.5 400
Absorbance
400
600
600
3400 3000 2600 1800 1600 1400 1200 1000

Wavenumber /cm1 Wavenumber /cm1

Dichroism and Optical Activity in Vibrational Spectroscopy

3.0 1.5
Strain (%) Strain (%)
2.0 0 0
1.0
200 200
1.0
Absorbance
0.5
Absorbance
400 400
600 600
3400 3000 2600 1800 1600 1400 1200 1000

Figure 11. FT-IR dichroic spectra of a polyurethane elastomer recorded during the elongation of the film up to 600% strain in (a) the 35002500 cm1 and (b) 1800900 cm1
regions. Assignments of the most relevant bands: 1, n(NH); 2, n(CH2 ); 3, free n(CDO)-urethane; 4, associated n(CDO)-urethane; 5, n(CDO)-urea. [Reproduced by permission
of John Wiley & Sons, Ltd. from H.W. Siesler, Makromol. Chem. Macromol. Symp., 53, 89 (1992).]
Orientation function
0.50 (NH) (CH2) (C=O)Urea
0.25
0.00
0.25
0 0 0 0 0 0 0 0
0.50
50 100 150 200 250 300 350 400 500
Strain (%)
Figure 12. Orientation function calculated from different IR bands of a polyurethane elastomer as a function of the applied strain.
[Reproduced by permission of John Wiley & Sons, Ltd. from H.W. Siesler, Makromol. Chem. Macromol. Symp., 53, 89 (1992).]
significant hysteresis. In Figure 12, the orientation func- 5.2.2 Polarization modulation infrared linear
tions calculated from the n(CH2 ), n(NH) and n(CDO)-urea dichroism
bands, using 90 between the direction of the transition
moment of these vibrations and the chain axis, are shown Marcott46 first demonstrated that the polarization modula-
as a function of the strain. For all elongationrecovery tion approach is a sensitive way to measure dichroic dif-
cycles, the orientation function of the soft segments, as fol- ferences in polymer films. The polarization modulation and
the normal linear dichroic difference spectra of a 135 m
lowed from the methylene groups of the polyether chain,
thick film of oriented isotactic polypropylene are presented
is always positive and increases almost linearly with the
in Figure 13(a) and (b), respectively. As can be seen, the
elongation and decreases reversibly upon recovery. This
polarization modulation linear dichroic difference spectrum
result indicates that the soft segments orient reversibly
has a much higher signal-to-noise ratio (S/N). In addition
along the stretching direction during the deformation cycles.
to the S/N advantage, the polarization modulation approach
The behavior of the hard segments, that can be moni-
eliminates the need for rotating either the sample or the
tored from the n(CDO)-urea band, is completely different.
polarizer when recording dichroic spectra.
Up to 150% elongation, the orientation function is always
PM-IRLD has been recently used to investigate molec-
negative, indicating a preferential perpendicular orienta-
ular orientation in uniaxially stretched polymers.49,77,78
tion of the hard segments, and it follows reversibly the
Since the structural absorbance of a uniaxially oriented
elongationrecovery cycles. Above 150% elongation, the
filmpdecreases due to sample thinning during elongation as
orientation function of the hard segment reaches a mini-
A0 / l (where l is the draw ratio), the orientation function
mum before the maximum elongation of each cycle and
for the chain axis is given by:
becomes positive at the end of the recovery period. This
p
effect, that becomes more important as the strain applied 2 l A
on the sample increases, has been associated with the par- P2 hcos qi D 36
3 cos2 b 1 3A0
tial disruption of the lamellar domains of hard segments
into smaller fragmental fibrillar domains. Since the long In model poly(dimethylsiloxane) (PDMS) networks,67,68
axis of the fibrillar hard segments is parallel to the poly- PM-IRLD has allowed the determination of dichroic differ-
mer chain axis, as opposed to the lamellar hard segments ences lower than 103 . Since the bands associated with
where it is perpendicular, the minimum of the orientation the fundamental modes are very intense, thin films (of
function is reached when the perpendicular orientation of about 100 m in thickness) were used to follow the orienta-
the lamellar hard segments is compensated by the parallel tion of PDMS networks. Figure 14(a) shows the dichroic
alignment of the smaller fragmental fibrillar domains result- behavior of the band located at 2500 cm1 , assigned to
ing from the disruption of the lamellar hard segments. Since the overtone of the symmetric bending vibration of the
this process is not completely reversible, the minimum of methyl groups. It can be seen that the absolute value of
the perpendicular orientation becomes less important with the dichroic difference, and thus the anisotropy of the sam-
the increase of the number of strain cycles. This example ple, increases with the draw ratio. This example clearly
demonstrates that rheo-optical FT-IR spectroscopy based shows the very high sensitivity of PM-IRLD since dichroic
on the simultaneous measurement of polymer mechanical differences as low as 2 104 were detected. Figure 14(b)
properties and dynamic dichroic spectra is a powerful tech- shows that similar information can be obtained from near-
nique that provides a deeper insight into the morphology infrared (NIR) spectroscopy. Indeed, the bands at 4164 and
and the segmental orientation and relaxation mechanisms 2500 cm1 , which have the same symmetry, give identical
of polymers. P2 hcos qi values at each draw ratio. However, the advantage
5 0.0003
4 0.0000 =1
3 0.0003
A = A||A
A 102
2 0.0006
1 0.0009
0 0.0012
= 1.8
0.0015
2000 1800 1600 1400 1200 1000 800 2600 2550 2500 2450 2400
(a) Wavenumber /cm1
(a) Wavenumber /cm1
0.0035
0.0030
3
0.0025
2 0.0020
A = A||A
A 102
0.0015
1
0.0010
0 0.0005
0.0000
1
0.0005
2 0.0010
4500 4400 4300 4200 4100 4000
2000 1800 1600 1400 1200 1000 800
(b) Wavenumber /cm1
(b) Wavenumber /cm1
Figure 14. Polarization modulation linear dichroic difference
Figure 13. (a) Polarization modulation linear dichroic difference spectra of a 100 m thick film of PDMS for draw ratios vary-
spectrum of a 135 m thick film of oriented isotactic polypropy- ing from 1 to 1.8 (a) in the 26002400 cm1 spectral range and
lene. (b) Normal linear dichroic difference spectrum. Each spec- (b) in the 45004000 cm1 spectral range. [Reproduced by per-
trum is the average of 100 scans recorded at 4 cm1 resolution. mission of Elsevier Science from T. Buffeteau, B. Desbat and
[Reproduced by permission of the Society for Applied Spec- L. Bokobza, Polymer, 36, 43394343 (1995).]
troscopy from C. Marcott, Appl. Spectrosc., 38, 442443 (1984).]
at constant strain during the relaxation period. Figure 15
of NIR spectroscopy is the possibility of analyzing thick shows the behavior of the orientation function calculated
samples because of the reduced intensity of the absorp- from the PS band at 1028 cm1 during this period for PS
tions in this spectral range. As shown in Figure 14(b), contents between 65 and 100%. A fast decay is observed
another advantage of PM-IRLD is the possibility of separat- for the short times, while a slow process due to the chain
ing bands because of their different dichroic behavior (see reorientation occurs at longer times. It can be seen that the
bands at 4164 and 4216 cm1 ). In a conventional spectrum, relaxation rate is faster for pure PS than for the different
these bands overlap, and the determination of the dichroic blends and that the increase of the PVME fraction leads
ratio requires the use of a band fitting procedure and an to longer relaxation times. This result can be interpreted in
appropriate baseline correction. terms of an increase of the friction coefficient between the
Due to the high sensitivity of PM-IRLD, only a few dissimilar polymer chains.
scans are necessary to obtain dichroic difference spectra It has recently been shown that PM-IRLD is a pow-
with good S/N, allowing kinetic measurements over a short erful technique to study in situ the dynamics of photo-
timescale. A direct determination of the relaxation curves induced orientation in polymers containing azobenzene
in polymer blends has been obtained using the polarization goups.50,51,79 82 It is well known that azobenzene groups
modulation technique.51 In this study, PS and poly(vinyl bound or dissolved in amorphous, semicrystalline or liquid
methyl ether) (PVME) blends were quickly stretched at a crystalline polymers can be oriented in a preferred direction
draw ratio of 2 and then PM-IRLD spectra were recorded using polarized laser light. When such polymer films are
0.14 1.0
0.12
0.8
0.10
Order parameter
65%
Absorbance
0.6
0.08
70%
DR1
0.06 0.4
0.04 80% 0.2

90%
0.02
100% PS
0.0
0.00 1800 1700 1600 1500 1400 1300 1200
0 500 1000 1500 2000 2500 3000 3500 4000
(a) Wavenumber / cm1
Time (s)
Figure 15. Relaxation of the PS chain order parameter P2 hcos qi 0.005
calculated from the 1028 cm1 band for different PSPVME
0.000
blend compositions. The films were stretched to a draw ratio of 2
at Tg C 8. The spectra used to calculate the order parameters were 0.005
obtained from 4, 30 and 75 scans at a resolution of 8 cm1 (0.4 s
A = A|| A
scan1 ) for short, intermediate and long times after the deforma- 0.010
tion, respectively. [Reproduced by permission of Elsevier Science
0.015
from M. Pezolet, C. Pellerin, R.E. Prudhomme and T. Buffeteau,
Vib. Spectrosc., 18, 103110 (1998).] 0.020
addressed with linearly polarized visible light, the azoben- 0.025

zene groups orient perpendicularly to the radiation polariza- 0.030
tion due to their photochemically induced transcistrans
0.035
isomerization. An example that illustrates the strength pro-
1800 1700 1600 1500 1400 1300 1200
vided by the differential nature of PM-IRLD is shown in
(b) Wavenumber /cm1
Figure 16 for poly(methyl methacrylate) (PMMA) doped
with the azobenzene dye DR1. Figure 16(a) shows the IR Figure 16. (a) Absorbance spectra between 1200 and 1800 cm1
spectra in the 12001800 cm1 region of a pure PMMA of a 5% DR1 doped PMMA film recorded before irradiation ( )
and of pure PMMA (- - - ) and (b) the PM-IRLD spectrum in
film and of a 5% DR1 doped PMMA film recorded before the same spectral range of the doped PMMA film after 1 h of
irradiation.82 Due to the low dye concentration, the bands irradiation at 514.5 nm. [Reproduced by permission of the Amer-
associated with the DR1 molecules are very weak and, ican Chemical Society from T. Buffeteau, F. Lagugne Labarthet,
in some cases, are masked by the PMMA absorptions. M. Pezolet and C. Sourisseau, Macromolecules, 31, 73127320
The PM-IRLD spectrum recorded on the 5% DR1 doped (1998).]
PMMA film after 1 h irradiation at 514.5 nm is shown in
Figure 16(b). As seen on this figure, the S/N of this dif- appear at 1339 and 1519 cm1 , respectively. Since the
ference spectrum is remarkable, even though the A for vibrations associated with these bands have their transi-
the strongest band is only 0.03. Except for the band at tion dipole moments nearly parallel to the long molecular
1731 cm1 due to the carbonyl vibration, all the absorption axis of the dye molecule, their negative sign indicates that
bands associated with the PMMA matrix have completely irradiation of the film produces a preferential orientation of
dissappeared in the PM-IRLD spectrum, indicating that the DR1 molecules perpendicular to the writing polarization
the polymer matrix remains unoriented. As a result, all direction.
the bands due to the azobenzene molecules are empha- The real time kinetics of photoinduced orientation in
sized in the PM-IRLD spectrum. The bands due to the n8a azopolymers was easily followed by PM-IRLD. As an
and n8b (CDC) stretching vibrations of the para-substituted example, the time dependent behavior of the PM-IRLD
phenyl rings of the DR1 molecules are located at 1601 spectra of the amorphous azopolymer pDR1A50 during the
and 1588 cm1 , respectively. The band due to the (NDN) orientation (laser on) and relaxation (laser off) periods is
stretching vibration appears at 1385 cm1 , whereas the presented in Figure 17. For clarity, the scale of the A
bands associated with the symmetric ns (NO2 ) and anti- axis has been inverted. It is important to stress that the
symmetric na (NO2 ) stretching vibrations of the NO2 group measurements were done at room temperature, which is
0.00
0.06
0.05 0.02
C = O (1736 cm1)
0.04 C = C (1601 cm1)
A
0.03 0.04 C = C (1588 cm1)
N = N (1389 cm1)
P2 cos
0.02
sNO2 (1339 cm1)
0.01 0.06
30
0.08
20
)
in
(m
1800 10
1700 1600 0.10
e
0
m
1500 1400
Wavenum
Ti
ber (cm1 1300 1200
(a) )
0.12
0 20 40 60 80 100 120
0.06 Time (min)
0.05 Figure 18. Time dependence of the orientation function for sev-
0.04 eral vibrations of pDR1A. [Reproduced by permission of the Soci-
0.03 ety for Applied Spectroscopy from T. Buffeteau and M. Pezolet,
Appl. Spectrosc., 50, 948955 (1996).]
0.02
0.01 results of Figure 18, it is possible to compare the degree of
30 orientation of several chemical groups within the polymer.
20 The orientation of the azobenzene groups induced by the
)
in
1800 laser beam is accompanied by a small orientation of the

(m
1700 1600 10
e
1500 1400 polymer main chain, as revealed by the small dichroism

m
Wavenum 0
ber (cm1 1300 1200
Ti
(b) ) of the band due to the ester group CDO stretching vibra-
Figure 17. Time dependence of the dichroic difference spectra tion at 1736 cm1 . The negative value of the orientation
of (a) photoinduced orientation and (b) relaxation processes of function indicates a preferential alignment of the carbonyl
a pDR1A film. [Reproduced by permission of the Society for groups perpendicular to the direction of the laser beam
Applied Spectroscopy from T. Buffeteau and M. Pezolet, Appl. polarization. The time dependent behavior of P2 hcos gi for
Spectrosc., 50, 948955 (1996).] the bands assigned to the azobenzene side chains is quite
similar. However, the limiting value of the orientation func-
approximately 70 C below the Tg of the polymer. As can tion is not the same for all bands since the direction of the
be seen, photoinduced orientation is readily induced in this transition dipole moment of each vibration with respect to
polymer and the kinetics of orientation and relaxation can the azobenzene molecular axis is different.
be easily followed from several absorption bands associated
with different chemical groups of the polymer. When the
writing laser is turned off, the orientation decreases slightly ABBREVIATIONS AND ACRONYMS
for the first few minutes showing that the induced optical
anisotropy is not completely preserved, but a significant ADC Analog-to-digital Converter
residual orientation remains. This decrease is most likely ATR Attenuated Total Reflection
associated with both thermal relaxation and angular reori- DIRLD Dynamic Infrared Linear Dichroism
entation of some azobenzene groups. In order to allow the FT-IR Fourier Transform Infrared
quantitative comparison of the time dependent behavior of HP High-pass Filter
different bands, the orientation function P2 hcos gi for a uni- IR Infrared
axial orientation distribution with respect to the laser polar- IRLD Infrared Linear Dichroism
ization was calculated for the transition dipole moments of LP Low-pass Filter
given vibrations using equation (25). In Figure 18, values MCT Mercury Cadmium Telluride
of the orientation function for several vibrations are repre- NIR Near-infrared
sented as a function of time for the irradiation period and PDMS Poly(dimethylsiloxane)
after turning off the writing laser. The dispersion of the PEM Photoelastic Modulator
data points on these curves is quite small due to the high PM-IRLD Polarization Modulation Infrared Linear
quality of the spectra used to calculate P2 hcos gi. From the Dichroism
PMMA Poly(Methyl Methacrylate) 25. U.P. Fringeli and H.H. Gunthard, Infrared Membrane
PS Polystyrene Spectroscopy, in Membrane Spectroscopy, ed. E. Grell,
Springer-Verlag, Berlin, 270332 (1981).
PVME Poly(Vinyl Methyl Ether)
TRS Time-resolved Techniques 26. F.M. Mirabella, Appl. Spectrosc., 42, 1258 (1988).
27. E. Goormaghtigh and J.M. Ruysschaert, Polarized Attenu-
ated Total Reflection Spectroscopy as a Tool to Investigate the
REFERENCES Conformation and Orientation of Membrane Components, in
Molecular Description of Biological Membranes by Com-
puter Aided Conformational Analysis, ed. R. Brasseur, CRC
1. I.M. Ward (ed.), Structure and Properties of Oriented Poly- Press, Boca Raton, FL, 285329, Vol. 1 (1990).
mers, Chapman & Hall, London (1997).
28. D.J. Ahn and E.I. Franses, J. Phys. Chem., 96, 9952 (1992).
2. E.J. Ambrose, A. Elliot and R.B. Temple, Nature, 163, 859
(1949). 29. P.H. Axelsen and M.J. Citra, Prog. Biophys. Mol. Biol., 66,
227 (1997).
3. R.D.B. Fraser, J. Chem Phys., 21, 1511 (1953).
30. N.J. Everall and A. Bibby, Appl. Spectrosc., 51, 1083
4. R.D.B. Fraser, J. Chem. Phys., 28, 1113 (1958). (1997).
5. R. Zbinden, Infrared Spectroscopy of High Polymers, Aca- 31. A. Cunningham, G.R. Davies and I.M. Ward, Polymer, 15,
demic Press, New York (1964). 743 (1974).
6. B. Jasse and J.L. Koenig, J. Macromol. Sci. Rev. Macromol. 32. A. Cunningham, I.M. Ward, H.A. Willis and V. Zichy, Poly-
Chem., C17, 61 (1979). mer, 15, 749 (1974).
7. H.W. Siesler and K. Holland-Moritz, Infrared and Raman 33. M. Born and E. Wolf, Principles of Optics, North Holland,
Spectroscopy of Polymers, Marcel Dekker, New York (1980). Amsterdam (1980).
8. H.W. Siesler, Adv. Polym. Sci., 65, 1 (1984). 34. R.M.A. Azzam and N.M. Bashara, Ellipsometry and Polar-
9. I.M. Ward, Adv. Polym. Sci., 66, 81 (1985). ized Light, North-Holland, Amsterdam (1977).
10. D.J. Bower, Infrared Dichroism, Polarized Fluorescence and 35. W.A. Shurcliff and S.S. Ballard, Polarized Light, D. Van
Raman Spectroscopy in Structure and Properties of Ori- Nostrand Company, New York (1964).
ented Polymers, ed. I.M. Ward, Chapman & Hall, London, 36. J.B. Young, H.A. Graham and E.W. Peterson, Appl. Opt., 4,
181231, Chapter 4 (1997). 1023 (1965).
11. U. Wiesner, N. Reynolds, C. Boeffel and H.W. Spiess, Liq- 37. A. Elliott, J. Polym. Sci., Part C, 7, 37 (1963).
uid Crystals, 11, 251 (1992).
38. L.J. Fina and J.L. Koenig, J. Polym. Sci., Part B, 24, 2509
12. F. Hide, N.A. Clark, K. Nito, A. Yasuda and D.M. Walba, (1986).
Phys. Rev. Lett., 75, 2344 (1995).
39. A. Kaito, K. Nakayama and H. Kanetsuna, J. Macromol. Sci.
13. Kulinna, S. Hvilsted, C. Hendann, H.W. Siesler and P.S. Phys., B26, 281 (1987).
Ramanujam, Macromolecules, 31, 2141 (1998).
40. P.G. Schmidt, J. Polym. Sci., Part A, 1, 1271 (1963).
14. M. Tsuboi, J. Polym. Sci., 59, 139 (1962).
41. J.L. Koenig, S.W. Cornell and D.E. Witenhafer, J. Polym.
15. H. Akutsu, Y. Kyogoku, H. Nakahara and K. Fukuda, Chem. Sci., Part A2, 5, 301 (1967).
Phys. Lipids, 15, 222 (1975).
42. L.E. Wolfram, J.G. Grasselli and J.L. Koenig, Appl. Spec-
16. Y. Koyama, M. Yanagishita, S. Toda and T. Matsuo, J. Col- trosc., 24, 263 (1970).
loid Interface Sci., 61, 438 (1977).
43. J.P. Sibilia, J. Polym. Sci., Part A2, 9, 27 (1971).
17. H. Nakahara, K. Fukuda, H. Akutsu and Y. Kyogoku, J. Col-
44. L.A. Nafie and M. Diem, Appl. Spectrosc., 33, 130 (1979).
loid Interface Sci., 65, 517 (1978).
45. L.A. Nafie and D.W. Vidrine, Double Modulation Fourier
18. K.J. Rothschild and N.A. Clark, Biophys. J., 25, 473
Transform Spectroscopy in Fourier Transform Infrared
(1979). Spectroscopy, Techniques using Fourier Transform Interfer-
19. K.J. Rothschild, R. Sanches, T. Hsiao and N.A. Clark, Bio- ometry, eds J.R. Ferraro and L.J. Basile, Academic Press,
phys. J., 31, 53 (1980). New York, 83123, Vol. 3 (1982).
20. E. Nabedryk and J. Breton, Biochim. Biophys. Acta, 635, 515 46. C. Marcott, Appl. Spectrosc., 38, 442 (1984).
(1981). 47. I. Noda, A.E. Dowrey and C. Marcott, Appl. Spectrosc., 42,
21. E. Nabedryk, M.P. Gingold and J. Breton, Biophys. J., 38, 203 (1988).
243 (1982). 48. V. Abetz, G.G. Fuller and R. Stadler, Makromol. Chem.,
22. P.A. Chollet and J. Messier, Chem. Phys., 73, 235 (1982). Macromol. Symp., 52, 23 (1991).
23. P.A. Chollet and J. Messier, Thin Solid Films, 99, 197 49. T. Buffeteau, B. Desbat, M. Pezolet and J.-M. Turlet, J. Chim.
(1983). Phys., 90, 1467 (1993).
24. A. Holmgren, L.B.-A. Johansson and G. Lindblom, J. Phys. 50. T. Buffeteau and M. Pezolet, Appl. Spectrosc., 50, 948
Chem., 91, 5298 (1987). (1996).
51. M. Pezolet, C. Pellerin, R.E. Prudhomme and T. Buffeteau, 69. I. Noda, A.E. Dowrey and C. Marcott, J. Polym. Sci., Polym.
Vib. Spectrosc., 18, 103 (1998). Lett. Ed., 21, 99 (1983).
52. K.W. Hipps and G.A. Crosby, J. Phys. Chem., 83, 555 70. I. Noda, A.E. Dowrey and C. Marcott, Appl. Spectrosc., 42,
(1979). 203 (1988).
53. C.W. Myers and S.L. Cooper, Appl. Spectrosc., 48, 72 (1994). 71. R.A. Palmer, C.J. Manning, J.L. Chao, I. Noda, A.E. Dow-
54. R.J. Roe and W.R. Krigbaum, J. Chem. Phys., 40, 2608 rey and C. Marcott, Appl. Spectrosc., 45, 12 (1991).
(1964). 72. J.E. Lash, D.J. Burchell, T. Masoaka and S.L. Hsu, Appl.
Spectrosc., 38, 351 (1984).
55. R.J. Roe, J. Appl. Phys., 36, 2024 (1965).
73. S.E. Molis, W.L. MacKnight and S.L. Hsu, Appl. Spectrosc.,
56. S. Nomura, H. Kawai, I. Kimura and M. Kagiyama, J. Polym.
38, 529 (1984).
Sci., Part A2, 8, 383 (1970).
74. H.W. Siesler, Rheo-optical FT-IR Spectroscopy of Macro-
57. R.S. Stein, J. Polym. Sci., 31, 327 (1958).
molecules, in Advances in Applied Fourier Transform
58. R.J. Samuels, Structured Polymer Properties, Interscience, Infrared Spectroscopy, ed. M.W. McKenzie, Wiley & Sons,
New York (1974). Chichester, 189246, Chapter 5 (1988).
59. D.J. Ahn and E.J. Franses, Thin Solid Films, 224, 971 (1994). 75. H.W. Siesler, Makromol. Chem. Macromol. Symp., 53, 89
60. C.-P. Lafrance, A. Nabet, R.E. Prudhomme and M. Pezolet, (1992).
Can. J. Chem., 73, 1497 (1995). 76. H.W. Siesler, I. Zebger, C. Kullinna, S. Okretic, S. Shilov
61. D.A. Jarvis, I.J. Hutchinson, D.I. Bower and I.M. Ward, and U. Hoffmann, Segmental Mobility of Liquid Crystals
Polymer, 21, 41 (1980). and Liquid-crystalline Polymers Under External Fields: Char-
acterization by Fourier Transform Infrared Polarization Spec-
62. P.C. Painter, M.M. Coleman and J.L. Koenig, The Theory troscopy, in Modern Polymer Spectroscopy, ed. G. Zerbi,
of Vibrational Spectroscopy and its Application to Polymeric Wiley-VCH, Weinheim, 3385, Chapter 2 (1999).
Materials, Wiley, New York (1982).
77. T. Buffeteau, B. Desbat, S. Besbes, M. Nafati and L. Bo-
63. H. Ishida (ed.), Fourier Transform Infrared Characterization kobza, Polymer, 35, 2538 (1994).
of Polymers, Plenum Publishing, New York (1984).
78. T. Buffeteau, B. Desbat and L. Bokobza, Polymer, 36, 4339
64. G. Zerbi (ed.), Modern Polymer Spectroscopy, Wiley-VCH, (1995).
Weinheim (1999).
79. T. Buffeteau, A. Natansohn, P. Rochon and M. Pezolet,
65. B. Jasse and J.L. Koenig, J. Polym. Sci., Polym. Phys. Edn., Macromolecules, 29, 8783 (1996).
17, 799 (1979).
80. T. Buffeteau and M. Pezolet, Macromolecules, 31, 2631
66. J.F. Tassin, L. Monnerie and L.J. Fetters, Polym. Bull., 15, (1998).
165 (1986). 81. A. Natansohn, P. Rochon, X. Meng, C. Barrett, T. Buffeteau,
67. C. Bouton, V. Arrondel, V. Rey, Ph. Sergot, J.L. Manguin, S. Bonenfant and M. Pezolet, Macromolecules, 31, 1155
B. Jasse and L. Monnerie, Polymer, 30, 1414 (1989). (1998).
68. T. Miyazawa and E.R. Blout, J. Am. Chem. Soc., 83, 712 82. T. Buffeteau, F. Lagugne Labarthet, M. Pezolet and C. Souri-
(1961). sseau, Macromolecules, 31, 7312 (1998).
Trichroic Infrared Measurements
of Orientation
Jack L. Koenig
Case Western Reserve University, Cleveland, OH, USA
1 INTRODUCTION 1. stretching a material while it is hot enough

to be drawn and then quenching (e.g., fiber
Orientation can be defined as the preferential spinning);
alignment (either uniaxial or biaxial) of polymer 2. stretching a material which cold draws;
chains or segments when subjected to an external 3. cold-rolling of a ductile material by passing the
force. The orientation induces significant changes polymer through a roll mill (calendering).
in properties, which can be either beneficial or Orientation can also be induced during the
detrimental. solidification of a thin film from a polymer solu-
A high degree of alignment is the basis of tion or melt. Oriented polymers are generally
improved mechanical, optical, and electrical prop- anisotropic; the properties along a particular direc-
erties in almost all polymers. Many factors tion are affected according to the conditions of
influence the orientation behavior of polymer stretching along that direction. To a first approx-
chains, including crystallinity, phase separation, imation, stretching in one direction produces
free volume, chain rigidity, molecular friction, uniaxial materials and stretching in two mutually
interchain interactions, and chain entanglements. perpendicular directions produces biaxial mate-
Consequently, it is necessary to understand the rials. However, even uniaxially stretched samples
correlations between molecular order, material can develop biaxial orientation if the sample is
properties, and fabrication procedures. Experi- not free to relax in the transverse direction during
mental characterization of the degree and direc- stretching.
tion of alignment is required in order to under-
stand how molecular design and processing strate-
gies lead to the ultimate state of alignment. 3 TYPES OF ORIENTATION
There are different types of orientation:
2 METHODS OF ORIENTING random: no preferred direction of the chains;
POLYMERS planar: a crystal or chain axis is oriented
parallel to a reference plane (usually the plane
An oriented polymer may be prepared by: of a film).
In polymer structures, orientation will not be torsional rotations of chain segments around the
complete and distributions of directions of planes chain backbone. There is an uncoiling of the chain
or axes will be observed. It is also possible for as extended conformers are formed as well as an
different overall orientation types to coexist as alignment of the segments with the direction of
is frequently found in metallurgical specimens. the mechanical load. As the alignment continues,
In addition, one often finds that amorphous and it reaches a point where the mechanical load is
crystalline phases, or different chain conformers, transferred to the deformation of covalent bonds
exhibit different degrees of orientation in polymer and angles.
systems.
5 TRICHROIC ORIENTATION
4 MECHANISMS OF MEASUREMENTS
ORIENTATION Infrared (IR) dichroic measurements can be used
to measure the orientation in films (see Linear
The orientation process can involve three compe- Dichroism in Infrared Spectroscopy). The
ting but interrelated deformations: second moment of the distribution function is
obtained from equation (1):
1. elastic or Hookean deformation resulting from
valence-angle deformation; 1
P2 (cos ) = (3cos2 1)
2. viscous flow deformation resulting from slip- 2
(1)
page of molecules. (This deformation is nonre- {Ro 2} (R 1)
coverable.) =
{Ro 1} (R + 2)
3. deformation resulting from uncoiling of poly-
mer chains. in which Ro = 2 cot2 , being the angle between
the dipole moment vector of the vibration under
The ideal orientation process leads to maximum consideration and the chain axis, and being the
uncoiling and chain alignment and to minimum angle between the chain axis and the stretching
viscous flow or chain slippage. direction. R is the dichroic ratio defined as
Orientation is a highly time-dependent process A|| /A , where A|| and A are the integrated
since polymers are viscoelastic materials; their intensities of the vibrational band in the parallel
molecular relaxation is also highly rate and and perpendicular direction, respectively, to the
temperature dependent. When the temperature is film stretching direction. This method can only
too low, i.e. when the polymer is in the glassy be used for uniaxial films, but most commer-
state, stretching requires excessively high stresses cial films have three-dimensional (3D) (biaxial)
and can lead to rupture. At high temperatures, orientation, so their absorptivities can be different
i.e. in the purely viscous state, deformations can in the machine, transverse, and thickness (TH)
easily be applied, but may not result in orienta- directions. Chains can also exhibit nonrandom
tion because rapid relaxation can exceed the rate orientations about their extended axes.
of deformation. In reality, many polymers have orientation with
Polymer drawing involves the extension of a 3D character. It is therefore necessary to deter-
an entangled molecular network. The macro- mine the relative amounts of the structural compo-
molecules form a transient network with entan- nents in the sample in the presence of orientation.
glements acting as nonlocalized junctions. The To do this, the average absorbance in the three
tensile behavior of the polymer can be affected directions x, y and z (see Figure 1), often called
by the extension and the alignment of its chain the structural absorbance, must first be calculated.
segments. For low degrees of alignment, defor- 1
mation at the molecular level is mostly related to A0 = Ax + Ay + Az (2)
3
Trichroic IR Measurements of Orientation of Polymers 3
y The intensity in any direction is given by equa-

tion (3)
=2 =/2
Mabc Ay = G( , )My2 d d (3)
a =0 =0
b where My is the absorbing transition moment in

the y direction and G( , ) is the distribution
x
c function describing the transition moments.2 In
Figure 1, = and is the angle between the z
axis and the projection of the moment vector on
z the x z plane. In general form for any transition
Figure 1. Location of sample in a Cartesian coordi- moment in the axis system, we have, equation (4)
nate system. [Reproduced with permission from Fina and
Koenig,6 Figure 1. Copyright 1986 John Wiley & Sons.] =2 =/2
A,, = 2
G( , )M,, d d
The 3D orientation function can be determined =0 =0
by using the ratios Ax /A0 , Ay /A0 and Az /A0 , (4)
where Ax and Ay correspond to the parameters where , , and are the angles that the moment
A|| and A discussed above; Az can be measured makes with each principal axis of the coordinate
using an IR technique based on an IR absorp- system. As a general rule, the distribution function
tion intensity ellipsoid approximation coupled is symmetric about the principal axes.3 The inte-
with a sample tilting procedure. The method of gral of the distribution function G over the total
performing such measurements in transmission absorption intensity is equal to unity. Under these
has been described in some detail.1,2 Similar conditions, equation (4) reduces to equation (5)
measurements can be made using attenuated total
reflection (ATR) methods,3,4 as well as spec- A,, = Ay cos2 + Ax cos2 + Az cos2
ular reflection, but only the simpler principles (5)
for transmission measurements will be reviewed This is the equation for an ellipsoid with axes
here. For thicker films, a method of microtoming of length (1/Ax )1/2 , (1/Ay )1/2 , and (1/Az )1/2 .
the film has been described which allows use of When the absorption intensities in two principal
polarized IR microscopy to make the transmission axes (x, y) and any other angle are known, the
measurements.5 absorption in the thickness (z) direction can be
found.4
The bulk sample is placed in the axis system
6 BACKGROUND with the preferred orientation direction parallel
The absorption of IR radiation for solids is depen- to y, the perpendicular direction is x, and the
dent on the mass of material and the transition film thickness is z (Figure 1). By imposing a
moment orientation.1 The maximum absorption specific tilt on the bulk sample and correcting
of plane-polarized light is found when the elec- for refraction of the incident light and path-length
tric vector of the incident radiation is parallel changes, equation (1) is reduced to equation (6)
to the transition moment; when the polarization
is perpendicular to the transition moment of a A,, (1 sin2 /n2 )1/2 Ay
Az = + Ay (6)
particular vibrational mode, it will not absorb at sin2 /n2
all. Orientation of the transition moment between
these extremes in any given direction has contri- where Az is the thickness-direction spectrum,
butions from each of the component absorptions. A,, is the tilted absorption spectrum, is the
angle that the incident electric vector subtends 8 ANALYSIS OF THE TILTING
with the sample, and n is the refractive index. METHOD
This equation assumes that the refractive index of
the surrounding medium is n = 1.0 (air). The determination of Az is described by equa-
The calculation of the thickness-direction spec- tion (4) and is straightforward and well establi-
trum requires a minimum of two polarized shed.7 The calculation of a thickness-direction
spectra, collected at normal and tilted incidence spectrum Az involves the subtraction of a spec-
of light. The accuracy of the prediction can be trum collected with normal incident radiation
improved by collecting spectra at several angles from one collected at an angle of tilt. A number of
of tilt. The data points are fitted to an ellipse by a issues affect the accuracy of the Az measurement,
nonlinear least-squares procedure, and the z axis such as reflective losses, the use of a frequency-
point on the ellipse is predicted. The disadvantage dependent versus a constant refractive index, and
of this method is a decrease in the signal-to-noise errors in n , where n is the refractive index at
ratio (S/N) of the predicted spectrum when the infinitely high frequency.
raw data are measured at from two to four tilt The IR spectra of thin films at any angle of
angles. The S/N improves with more than four tilt can be simulated using generalized optical
points. Another method, which increases the accu- theory. The transmittance and reflectance of each
racy of the prediction, is to calculate the point phase of a multiphase system can be calcu-
lated using Fresnel coefficients as described by
along the z axis from two orientations of the
Heavens.8 The algorithm assumes that each phase
sample turned around the z axis (see Figure 1),
is isotropic and homogeneous and that the inter-
provided birefringence is not a problem. It should
faces are plane-parallel. Optical constant spectra
be noted that an analogous problem exists for
(n and k) for films of poly(vinylacetate) (PVAc),
Raman spectroscopy, in that when we analyze an
were generated using the KramersKronig trans-
oriented polymer article the Raman intensities are formation and used as input for the tilting film
distorted as well. Correcting the Raman intensities simulation. The baselines in regions of nonabsorp-
for birefringence is no small problem. tion are a function of the amount of reflectance,
which is approximately 8% higher for s polar-
ization (where p and s are defined with the
7 APPARATUS AND electric vector of the incident polarized light
parallel (p) or perpendicular (s) to the plane
SPECTROSCOPY of incidence) than for p at an incident angle
of 45 .9 The dispersion in the refractive index
Angle-dependent absorption information is obtai- through an absorption band affects the band
ned from a tilting apparatus. The sample is moun- shape. The refractive index dispersion causes
ted in a preferred direction on the tilting stage. band shape distortions characterized by a peak
The device is constructed such that the sample intensity shift, a half-width increase, and small
can be rotated about a horizontal or vertical increases in the high-frequency wing of the band.
axis, both of which are orthogonal to the direc- These purely optical distortions are translated into
tion of propagation. The apparatus is interfaced the z (the thickness-direction) spectra. However,
with a computer for automated and precise tilting under actual experimental conditions of thin-film
control. A wire grid polarizer is used to polarize analysis, the band distortions are detectable only
the incident light beam. Three-dimensional infor- in strongly absorbing vibrational modes and at
mation is obtained by rotation of the bulk sample highly reflecting interfaces.
on the axis perpendicular to the incident electric The relative error involved as a function of tilt
vector. The data were processed using computer angle can also be estimated through the simula-
programs developed in our laboratories. tion. As the angle of tilt increases, the rms (root
mean square) S/N increases linearly, particularly A = 1.25

above 30 . At the lower angles of tilt (<30 ), 874
the subtractive nature of the calculation tends to 973-T

Draw 848-T
degrade the signal. Larger angles of tilt increase
ratio 794
the accuracy of the orientation-independent (Ao )
spectrum. For example, a tilt of an angle of 30 5.0
Absorbance
4.0
reduces the calculation error to below 10%. 3.0
2.5
2.0
9 APPLICATION OF THE 1.6
875
TRICHROIC METHOD TO 1.0 843
UNIAXIALLY DRAWN 974 896-G 794
POLY(ETHYLENE 1050 950 850 750

TEREPHTHALATE) Wavenumber (cm1)
Figure 3. Calculated isotropic spectra of uniaxially
In Figure 2, the trichroic spectra of uniaxially drawn PET as a function of percentage elongation. A
drawn poly(ethylene terephthalate) (PET) illus- is the absorbance range for each spectrum. (G, gauche;
trate the differences in the spectra for the direc- T, trans). [Reproduced with permission from Fina and
tions parallel, perpendicular, and through the Koenig,6 Figure 2. Copyright 1986 John Wiley & Sons.]
thickness direction of a one-way drawn PET film.6
The isotropic IR spectra of the PET films of percentage elongation in Figure 3 for the
are obtained from a linear combination of these frequency region 8001000 cm1 .
three spectra. The calculated isotropic spectra for By using the spectra of the pure gauche and
one-way drawn PET are shown as a function trans isomers and the least-squares curve-fitting
technique, the conformer composition can be
determined as a function of draw ratio. The results
are shown in Figure 4.
II
70
60
Absorbance
50

Relative amount
40 trans Amorphous
Total trans
30
TH
20
trans Crystal
10
1000 950 900 850 800
Wavenumber (cm1) 0
Figure 2. Spectra of a one-way drawn poly(ethylene 1.0 2.0 3.0 4.0 5.0
Draw ratio
terephthalate) (PET) film in the directions parallel (||) and
perpendicular () to the direction of draw and through the Figure 4. Conformer composition of uniaxially drawn
thickness (TH) direction of the film. [Reproduced with PET as a function of draw ratio. (G, gauche; T, trans).
permission from Fina and Koenig,6 Figure 1. Copyright [Reproduced with permission from Fina and Koenig,6
1986 John Wiley & Sons.] Figure 5. Copyright 1986 John Wiley & Sons.]
A /A o 891 A = 1.1
2.0 881
973 cm1
1.8
845
1.6 Parallel TH 882
Orientation parameter
842
Absorbance
1.4
1.2
Random

1.0
0.8 Perpendicular
Thickness II
0.6
940 920 900 880 860 840 820 800
0.4
0 50 100 150 200 250 300 350 Wavenumber (cm1)
Percent elongation Figure 6. Spectra of oriented phase-I PVDF annealed
under tension. Parallel, perpendicular, and calculated
Figure 5. Trichroic orientational parameters for one-way
thickness-direction spectra are shown. [Reproduced with
drawn PET as a function of percentage elongation. [Repro-
permission from Fina and Koenig,6 Figure 9. Copyright
duced with permission from Fina and Koenig.6 Copyright
1986 John Wiley & Sons.]
1986 John Wiley & Sons.]
The orientation parameters for the three differ- 890 A = 1.0

ent directions can also be calculated as a func-
tion of the percentage elongation, as shown in
Figure 5. 846
882
Thus, using the trichroic method, it is possible TH
to quantify the conformers in PET as well as
Absorbance
the orientation in the three mutually perpendicular 842

directions.

10 APPLICATION OF THE
TRICHROIC METHOD TO II
POLY(VINYLIDENE
940 920 900 880 860 840 820 800
FLUORIDE)
Wavenumber (cm1)
When structural changes occur, the trichroic Figure 7. Spectra of oriented phase-I PVDF annealed
measurements will reveal frequency shifts as well under strain relaxation. Parallel, perpendicular, and calcu-
as changes in relative intensities. This is the lated thickness-direction spectra are shown. [Reproduced
with permission from Fina and Koenig,6 Figure 10. Copy-
case when phase I of poly(vinylidene fluoride)
right 1986 John Wiley & Sons.]
(PVDF) is oriented and annealed both under stress
(Figure 6) and without stress10 (Figure 7).
Frequency differences can be observed at there is a difference in the frequencies at 891 cm1
842 cm1 in the perpendicular direction and at in the thickness direction and at 882 cm1 in
845 cm1 in the thickness direction in the samples the perpendicular direction. The origin of these
annealed under stress and then relaxed. Similarly, shifts has been attributed to defect structures
in the phase-I lattice. Defects in the crystal REFERENCES

lattice will occur through the introduction of
gauche (+) or gauche () conformations. Nucle- 1. J.L. Koenig, S.W. Cornell and D.E. Witenhafer, J.
ation of a transforming chain onto a pre-existing Polym. Sci., A2, 301 (1967).
defect-containing chain can propagate a kink in 2. L.J. Fina and J.L. Koenig, J. Polym. Sci., Part B:
a direction-specific manner. Molecular modeling Polym. Phys., 24, 2509 (1986).
studies suggest a steric feasibility of phase-I crys- 3. P.H. Axelsen and M.J. Citra, Prog. Biophys. Mol.
tals with defects aligned approximately perpendic- Biol., 66, 227 (1997).
ular to the trans lattice. In essence, the vibrations 4. N.J. Everall and A. Bibby, Appl. Spectrosc., 51,
of the defect regions are in register with and at 1083 (1997).
a fixed and defined angle from the defect-free 5. J. Chalmers, L. Croot, J. Eaves, N. Everall,
crystal regions. Hence, the orientation-dependent W. Gaskin, J. Lumsdon and N. Moore, Spectrosc.
frequencies of these defect bands are observed.6 Int. J., 8, 13 (1990).
6. L.J. Fina and J.L. Koenig, J. Polym. Sci., Part B:
Polym. Phys., 24, 2525 (1986).
ABBREVIATIONS AND 7. P.G. Schmidt, J. Polym. Sci. A, 1, 1271 (1963).
ACRONYMS 8. O.S. Heavens, Optical Properties of Thin Solid
Films, Dover, New York, Chapter 4 (1965).
ATR Attenuated Total Reflection
IR Infrared 9. N.J. Harrick, Internal Reflection Spectroscopy,
Wiley Interscience, New York (1967).
PET Poly(Ethylene Terephthalate)
PVAc Poly(Vinylacetate) 10. L.J. Fina and J.L. Koenig, J. Polym. Sci., Part B:
Polym. Phys., 24, 2541 (1986).
PVDF Poly(Vinylidene Fluoride)
TH Thickness
S/N Signal-to-Noise Ratio
Basics of Orientation Measurements
in Infrared and Raman Spectroscopy
Richard Ikeda1, Bruce Chase2 and Neil J. Everall3

1
University of Delaware, Newark, DE, USA
2
DuPont Experimental Station, Wilmington, DE, USA
3
ICI Measurement Science Group, Wilton, UK
1 INTRODUCTION of polarized measurements in infrared (IR) and

Raman spectroscopy from the point of view
Molecular orientation plays a critical role in of measurements of chain orientation in poly-
many aspects of materials science. Polymeric meric systems. These concepts are by no means
fibers, liquid crystal displays, polarizing filters, new. Orientation measurements have been made
and ferroelectric switches are just a few exam- using vibrational spectroscopy for more than
ples of materials where orientation of molecular 50 years. Our purpose was to bring together the
units plays a significant role in the material prop- basic concepts of orientation distributions and
erties. At a more fundamental level, measurement include the derivations of some of the funda-
of molecular orientation is invaluable for studying mental expressions relating measured intensities
deformation mechanisms in drawn polymers (see to those orientation distributions.
Molecular Orientation Studies of Polymers by
Infrared Spectroscopy). The vibrational spec- 2 POLARIZED IR
troscopies are some of the few molecular level
probes that are inherently orientation sensitive. SPECTROSCOPY
This arises from the directionality inherent in Polarized IR spectroscopy can be used to deter-
photons, which have both a propagation direc- mine the orientation of polymeric chains since the
tion and can be polarized in the plane normal to dipole change associated with a normal mode or
the propagation direction. In order to extract the vibration acts as a vector marker for the chain
sample orientation information from the spectro- orientation. The absorption is proportional to the
scopic measurements, one needs a clear under- square of the dot product of the dipole vector and
standing of how polarized light interacts with the electric vector of the incoming radiation. Since
the sample for both absorptive and scattering the dipole vector is fixed in the sample and we
processes. This chapter reviews the basic concepts can vary the direction of the electric vector of the
z P
= x + y + z (2)
E = E0 (i sin cos + j sin sin + k cos )
(3)
E m E = Ex + E y + E z (4)
hw where 0 and E0 are the scalar magnitudes of the

vectors and E, and i, j, and k are the unit vectors
y along the directions x, y and z, respectively. The
angles are defined in Figure 2. We choose a
polar coordinate system for convenience since
x the IR transition moment is a vector, which
is very easily represented in a polar coordinate
Figure 1. Axis system for an oriented polymer chain.
system.
In any real sample, there will not be one
incoming radiation, we can effect a change in the unique direction for all the dipoles but there
intensity of the radiation absorbed, which will be will be a distribution of these dipole directions
a function of the cosine of the angle between the in space. In order to calculate the absorption
vectors (i.e. the direction cosine). If one knows intensity we need to sum over all the individual
the orientation of the dipole change with respect dipoles, i .
to the chain axis then, by determining the orienta-

tion of the dipole vectors, one can determine the A= E2 (5)
i
orientation of the chains. In Figure 1, the polymer i
chain axis is depicted by the vector, P, and the
dipole change associated with a vibration by the Such a summation is very difficult to deal with
vector, . The laboratory axis system is defined analytically so we convert from a summation to an
by the cartesian coordinates (xyz ) and the elec- integral using a distribution function. One defines
tric vector associated with the incoming radiation that distribution by the polar coordinate function
(h) is represented by the vector, E. The vectors g(, ), which is proportional to the number of
and E can be defined in a polar coordinate dipoles oriented at polar angles (, ). If we did
system (where x , y , z are the projections onto have a single unique direction for all dipoles then
the Cartesian axes): g(, ) = 1 for a specific (, ) and 0 for all other
angles. The distribution function is defined such
= 0 (i sin cos + j sin sin + k cos ) that g(, ) sin d d (Appendix 1) is the frac-
(1) tion of dipoles within the solid angle, sin d d ,
z m z E
q a
y y
f b
x x
Figure 2. Angular definitions for Cartesian to polar transformation.
Basics of Orientation Measurements of Polymers 3
and: Now from equations (810) and the definitions

2
of Ex , Ey and Ez equations (3) and (4) we have:
g(, ) 2 sin d d d
0

0
2
0
= 1 or A = Axx sin2 cos2 + Ayy sin2 sin2
2
sin d d d
0 0 0 + Azz cos2 (11)
2
1
g(, ) sin d d = 1 (6) where
4 0 0
The total absorbance, A, is then defined. 2
2 Axx = E02 20 g(, ) sin3 cos2 d d

0 0
A= g(, )| E|2 sin d d (7) (12)
0 0 2
Ayy = E02 20 g(, ) sin3 sin2 d d
The | E|2 term can be expanded: 0 0
(13)

| E| =
2
2x E2x + 2y E2y + 2z E2z + 2x y Ex Ey 2
Azz = E02 20 g(, ) sin cos2 d d
+ 2x z Ex Ez + 2y z Ey Ez 0 0
(14)
and equation (7) written as a sum of six integrals,
which we will call Mxx , Myy , Mzz , Mxy , Mxz , and Equation (11) is the equation for an ellip-
Myz . Though this expansion is written in the soid with major axes (1/Axx )1/2 , (1/Ayy )1/2 , and
Cartesian coordinate system, before integration (1/Azz )1/2 . The average of A over all orientations
we will convert to polar coordinates. The first is found to be A0 (Appendix 3):
three terms in the expansion of equation (7) are
defined as:
Axx + Ayy + Azz
2 A0 = (15)
Mxx = Ex2 20 g(, ) sin3 cos2 d d 3
0 0
(8)
2 A0 is defined as the structural absorbance and
Myy = Ey2 20 g(, ) sin3 sin2 d d is orientation independent and proportional only
0 0
(9) to the number of oscillators being observed.
2 If the specimen has fiber symmetry (the distri-
Mzz = Ez 0
2 2 2
g(, ) sin cos d d bution function, g( ), is independent of ), two of
0 0
(10) the axes in the absorption ellipsoid will be equal
and (Appendix 4):
and the second three integrate to zero if g(, )
is symmetrical about the three Cartesian planes A|| + 2A
(Appendix 2). This symmetry restriction may not A0 = (16)
3
be met by some self-assembled films but it is
common for most commercial polymeric mate-
rials. The absorbance can thus be written The absorbances are related to Hermans orien-
tation function:
A = Mxx + Myy + Mzz
A|| A
P2 (cos ) = (17)
with the next three terms being zero. 3A0
This is often expressed in terms of the dichroic tilting a film sample, such that the z axis and the
ratio D(D = A|| /A ). optical propagation axis are no longer coincident,
usually leads to progressively increasing reflec-
D1 tivity losses. To minimize these losses, the films
P2 (cos ) = (18)
D+2 can be placed between KBr hemispheres with the
Equation (18) yields the orientation of the interface filled with some liquid that does not
vibrational dipole. Appendix 5 shows how this absorb in the regions of interest. (Fluorocarbon
can be related to the chain orientation. and hydrocarbon oils can be used as the wetting
liquids depending on the region of the IR spec-
trum under investigation.) Tilting also leads to
3 EXPERIMENTAL an effective thickness change and a need for the
determination of the three normal elements of the
DETERMINATIONS
refractive index ellipsoid as well. In Figure 3 a
3.1 Tilted film method rotation experiment is depicted in which only Ayy
of the absorption tensor is measured. Any changes
For most polymer film samples, the film formation in the band absorbance must be due to thickness
process, whatever it is, usually results in two of changes and these data can be used to compute
the major axes (i.e. Axx and Ayy ) lying in the an effective refractive index (ny ) using Snells
plane of the film. In this case, the film sample law. Turning the films so that the x-axis is in the
can be held normal to the direction of the IR vertical position and performing the same rotation
radiation and absorption spectra measured as the experiment will lead to the determination of the
film is rotated with respect to the polarization second refractive index (nx ). In the tilt experiment
direction. If the film is unbalanced, the minimum the absorption will be due to the interaction of E
and maximum absorbances define the two axes with both Ayy and Azz and the effective refractive
of the absorption ellipsoid. In a balanced film the index, which is a combination of ny and nz . Since
spectra will be invariant with rotation and any two there are two unknowns, Azz and nz , a second
axes 90 degrees to each other will suffice as the tilt experiment must be performed starting with
axes of the ellipsoid. the x-axis in the vertical position. One thus has
One is then left to determine the absorbance two experiments and two unknowns and Azz can
in the third direction (Azz ). As long as the film be determined. For more details on this proce-
is normal to the radiation direction, the elec- dure see Trichroic Infrared Measurements of
tric vector, E, and Azz will be normal and their Orientation.
dot product zero. As shown in Figure 3, tilting
the film with respect to the radiation direction
will bring an interaction between E and Azz , but 3.2 Thick samples
If the polymer sample is too thick or is an
y y inherently strong absorber, the tilted film method
is often not applicable because the key absorp-
tion bands can saturate. For example, studies
E E of aromatic polyesters typically require films
x x thinner than 10 m to avoid saturation. A typical
polyester bottle wall thickness might range from
z
z 100 to 1000 m, and as such cannot be studied
(a) (b) directly in transmission. Fortunately, a number
Figure 3. Definitions of (a) rotation and (b) tilt required of alternative techniques have been developed to
for determination of refractive indices. allow the study of strongly absorbing polymers or
x IR
Axx
microscope
Bottle
y z
Axx
Azz
Cut II(x ) Azz
Ayy
Cut II(y ) Azz Ayy
Azz
IR
microscope
Figure 4. Experimental design for orientation determination using IR absorption on thick samples.
thick films. These approaches have been reviewed 4 POLARIZED RAMAN

in detail elsewhere1 and are only briefly summa-
SPECTROSCOPY
rized below.
In some cases it is possible to microtome thin
sections that can then be analyzed in transmission When compared to IR spectroscopy, Raman
using an IR microscope fitted with a polarizer. In experiments have the advantage of accepting
this mode, assuming that the in-plane macroscopic a wider variety of sample shapes. This is a
optical axes are denoted x and y, two thin sections distinct advantage in an industrial setting where
are cut parallel to the x and y axes, respectively. the shape and size of the samples are often not
One then records Axx and Azz spectra from the options. On the other hand, additional compli-
x cut, and Ayy and Azz spectra from the y cut cations arise because the Raman measurements
sections. The required dichroic ratios (Axx /Azz are scattering experiments. This is illustrated in
and Ayy /Azz ) can then be computed. This is Figure 5. Here we depict a cylindrically shaped
illustrated in Figure 4 using the example of a sample containing a polymer chain. The incoming
bottle wall. radiation (h) induces an oscillating dipole in
If microtoming a section is not an option the electronic cloud surrounding the polymer
(for instance, with a soft material it can some- chain and the molecular vibration modulates this
times induce significant orientation), a number of induced dipole at the frequency of the vibration.
approaches based on polarized reflection spec- It is this modulation of the induced dipole that
troscopy have been developed. These involve permits the interaction of the incoming radiation
attenuated total reflection or specular reflection. with the vibration that leads to the transfer of
Both approaches have the disadvantage that they energy and the resulting frequency shifts in the
only probe the properties of the surface (top scattered radiation. As with the absorption of IR
few micrometers) of the sample, but they are radiation, the amount of radiation absorbed or
readily applicable to optically thick samples and radiated is proportional to the square of the dot
can be powerful tools for studying polymer arti- product of the induced dipole vector, , and the
cles as received. Again, detailed examples and electric vector of the radiation. The magnitude
references to these, and other approaches for of the induced dipole, in turn, is related to the
measuring molecular orientation, have been given intensity of the incoming radiation, E, via the
elsewhere.1 polarizability tensor, .
E
h(w w0)
a2
a1
hw
a3
E
x
Figure 5. Axis system for an oriented polymer chain and polarized Raman measurements.
In matrix notation:
= E (19) f
It can be shown that a unique set of axes exists q

in the molecule that diagonalizes the polarizability
matrix leaving three terms (1 , 2 , and 3 ) called
the principal values of the polarizability.
In any given Raman experiment, the intensity j
of the polarized, scattered radiation, IR , is thus
given as a sum of products of the polarizability
tensor elements:2 Figure 6. Euler angles for coordinate transformation
2 used in polarized Raman scattering.

IR = I0 l(I)m l(s)n mn (20)
mn
oriented within the solid angle, sin d d d.
where l(I)m are the direction cosines for the Here and are the polar and azimuthal angles
electric vector of the incoming radiation and l(s)n which define the orientation of some particular
are those for the scattered radiation. mn are the axis of a structural unit, and is the rotation
elements of the polarizability tensor in the labora- of the entire unit about that axis (see Figure 6).
tory coordinate system. These experimental polar- These are the Euler angles and define the coor-
izability elements can be related to the principal dinate transformation used to go from the labo-
values of the polarizability and the orientation ratory Cartesian system to the molecular frame
distribution of the scattering units via a procedure Cartesian system. This is analogous to the Carte-
first outlined by Bower.2 sian to polar coordinate transform used in the IR
Bower assumed some general distribution of analysis. Changing = cos and assuming that
oscillator units given by N (, , ), where N (, , ) is some smooth function of , and
N (, , ) sin d d d is the fraction of units we can expand N (, , ) as a polynomial series
in generalized spherical harmonics: 1 dl 2

Zlmn ( ) = ( 1)l
2l l! d l

l
l
Zlmn ( ) = (1)l Zlmn ( ) (22)
N (, , ) = lmn Zlmn ( )eim ein
l=0 m=l n=l This expansion is possible because the Leg-
(21)
endre polynomials form a complete orthogonal set
of basis functions. It is very convenient because
Here the terms, Zlmn ( ), are generalized Leg- many of the average values of the polynomials
endre polynomials that are real and defined by the are readily determined using polarized optical
Rodrigues formula: spectroscopy. For convenience, the explicit forms
of the Zlmn that are relevant to IR and Raman
studies of molecular orientation are summarized
Z000 ( ) = 1 in Table 1.
Table 1. Legendre polynomials Zlmn of relevance to polarized IR and Raman spectroscopy.3 These are non-normalized
forms, and are often written as Plmn in the literature. The average values, Zlmn , depend on the orientation distribution
function; these are the parameters which are determined using IR and Raman analysis. Confusingly, they are also often
summarized simply as Plmn in the literature.
Case 1 Fiber symmetry uniaxial alignment of polymer chain chains, random rotation
about chain director (both and are random)
Z000 = 1
Z200 = 12 (3 cos2 1)
Z400 = 18 (3 30 cos2 + 35 cos4 )
Case 2 General uniaxial symmetry (only is random)

Z202 = 14 (1 cos2 ) cos 2
Z402 = 1
24
(1 + 8 cos2 7 cos4 ) cos 2
Z404 = 1
16
(1 2 cos2 + cos4 ) cos 4
Case 3 Biaxial symmetry (only random)

Z220 = 14 (1 cos2 ) cos 2
Z420 = 1
24
(1 + 8 cos2 7 cos4 ) cos 2
Z440 = 1
16
(1 2 cos2 + cos4 ) cos 4
Case 4 General biaxial symmetry (no random angles)

Z222 = 14 [(1 + cos2 ) cos 2 cos 2 2 cos sin 2 sin 2]
Z422 = 14 [(1 6 cos2 + 7 cos4 ) cos 2 cos 2 + (5 cos 7 cos3 ) sin 2 sin 2]
Z442 = 1
16
[(1 cos4 ) cos 4 cos 2 2(cos cos3 ) sin 4 sin 2]
Z424 = 1
16
[(1 cos4 ) cos 2 cos 4 2(cos cos3 ) sin 2 sin 4]
Note that it is assumed, in common with Bower,2 that the polar and azimuthal angles are and , respectively, and
is the rotation about the extended chain director. The equations appear differently in Jarvis et al.,3 where is still the
polar angle but is the azimuth and denotes rotation about the director.
These are quite complex functions which are making use of equation (24) and the assump-
difficult to visualize; Jarvis et al.3 have shown tion of fiber symmetry. Fiber symmetry stip-
how they can be related to direction cosines ulates that the distribution is independent of
between the structural unit axes and the macro- the angles, and (Case 1, Table 1). Thus
scopic sample axes. Roe and Krigbaum4 have substituting N ( ) for N (, , ), equation (24)
shown that the terms, Zlmn ( ), have the following becomes
symmetry properties: 1
1
lmn = 2 N ( )Zlmn ( ) d
Zlmn ( ) = (1)l+m Zlmn ( ) = (1)m+n Zlmn ( ) 4 1
2 2
= (1)l+n Zlmn ( ) (23) eim ein d d (27)
0 0
where m = m and n = n. Since the Legendre When m and n = 0, each of the second two
polynomial series forms an orthogonal set in the integrals become 2. For any non-zero values of
-interval from 1 to +1, the coefficients are m and n one finds:
defined: 2 2
2 2 1 eim d = (cos m + i sin m) d = 0
1
vlmn = N (, , )Zlmn ( ) 0 0
(28)
4 2 0 0 1
and one only needs to consider:
eim ein d d d (24)
1
l00 = N ( )Zl00 ( ) d Zl00 ( ) (29)
Note, these coefficients are simply the average 1
values of the spherical harmonics, taken over the
distribution N (, , ). Depending on whether When working with non-normalized Legendre
the Legendre functions are specified in normal- polynomials as defined in Table 1 1, this relation
ized or non-normalized forms, the coefficients is often written as Pl00 = 1 N ( )Pl00 ( ) d .
lmn used by different authors can differ by simple The Pl00 are the parameters which are most
numerical factors. Equation (24) assumes normal- often quoted in studies of uniaxial orientation
ized forms, but in one particularly common repre- (specifically Case 1 in Table 1); they often are
sentation, non-normalized forms are used (these abbreviated simply to Pl (or Plmn in the general
are the forms which are given in Table 1). In this biaxial case). Since N ( ) is symmetrical around
case, the coefficients are usually written as Plmn , = 0 and Zl00 ( ) = Zl00 ( ) for odd values
and the following relation holds:2 of l, the integral in equation (27) will vanish for
odd values of l and thus one need consider only
4 2 even values of l.
Plmn = (25) Equation (20) indicates that the intensity of the
Nlmn lmn
polarized scattered radiation is the sum of prod-
ucts of the polarizability tensor elements. If one
where
regards the tensors as orientable units, choosing
(2l + 1) (l + m)!(l n)! 1 their principal axes fixed within the units one can
2
Nlmn = write for the aggregate:
2 (l m)!(l + n)! [(m n)!]2
(26)
ij pq =
2 2 1
The expression for the distribution function
N0 N ( )ij pq d d d
[equation (21)] can be greatly simplified by 0 0 1

ij pq = The product terms ij pq in equation (31) may
be considered as the elements of a fourth rank
2 2 1 tensor with components, ij pq = dijpq . Birss5
N0 l00 Zl00 ( )ij pq d d d has shown that for a fourth rank tensor describing
0 0 1 l
the properties of a material with D2h symmetry
(30) the coefficients, dij kl must satisfy the relation:
or dij kl = ip j q kr ls dpqrs (35)

ijpq
ij pq = 4 2 N0 l00 Al00 (31) where the s are the elements of the generating
l matrices for the various applicable symmetry
operations. For D2h symmetry one has:
where
1 0 0 1 0 0
ijpq
Al00 = (1)
= 0 1 0 (2)
= 0 1 0
0 0 1 0 0 1
1 2 2 1
Zl00 ( )ij pq d d d 1 0 0
4 2 0 0 1 (3) = 0 1 0 (36)
(32) 0 0 1
First one notes that the above are all diagonal
The elements, ij , of the laboratory frame
matrices so the following must hold: i = p, j =
polarizability tensor, (lab), can be obtained from
q, k = r and l = s (all other elements vanish).
the principal polarizabilities, (princ), using the
If (2) is used in equation (35), the product,
transformation matrix:

cos cos cos sin sin cos cos sin sin cos sin cos
T= cos sin cos cos sin cos sin sin cos cos sin sin (33)
sin cos sin sin cos
ip j q kr ls = 1, if the subscript 2 appears an

In matrix notation:
odd number of times this would imply dij kl =
dij kl . All the coefficients, dij kl , where 2 appears
(lab) = T1 (princ)T an odd number of times must thus be zero. Like-
wise, using (3) , all coefficients, dij kl , must also
or writing out the individual terms: vanish when 3 appears an odd number of times.
The remaining coefficients are therefore:

3
ij = tik tj k k (34) d1111 d1122 d1133
k=1 d2211 d2222 d2233
d3311 d3322 d3333
where the ts are the elements of the transforma- d2323 d2332 d3223 d3232
tion matrix. d3131 d3113 d1331 d1313
Equations (32) to (34) thus permit the deter- d1212 d1221 d2112 d2121
ijpq
mination of the elements, Al00 , in terms of the
principal polarizabilities (1 , 2 , 3 ) plus the coef- Since the polarizability matrix possesses fiber
ficients of the Legendre expansion. symmetry (in this case the fiber axis corresponds

to the subscript 3), the subscript 1 is the same a12 + a22 2 + 4a1 a2 + 2a1 + 2a2
as 2. Since the overall polarizability tensor is + P2
21
also symmetric (i.e. ij = j i ), the order of the

numbers in the pairs of subscripts can be inter- 3a12 + 3a22 + 8 + 2a1 a2 8a1 8a2
+P4
changed. This reduces the above set to five: 280
(40)
d2222 = (22 )2
I0 2
23 =
d3333 = (33 )2
d2323 = (32 )2 2
(37) a1 + a22 + 1 a1 a2 a1 a2
d1212 = (12 )2 b
15
d2233 = 22 33 2
a + a2 2 + 4a1 a2 + 2a1 + 2a2

2
P2 1
For a sample with fiber symmetry the polar- 42
ized Raman experiment is thus capable of deter- 2

3a1 + 3a2 + 8 + 2a1 a2 8a1 8a2
2
mining only five independent quantities. Since the P4
three principal polarizabilities are unknown, the 70
Legendre expansion of the orientation function (41)
must be truncated above l = 4 to keep the number
I0 22 33 =
of unknowns equal to the experimental determi-
2
nations. Returning to equations (30) to (34) the a1 + a22 + 1 4a1 a2 4a1 4a2
five, non-zero, and independent quantities can be b
15
written: 2
a1 + a2 2 + 10a1 a2 5a1 5a2

2
P2
I0 2
22 = 42
2
2

3a1 + 3a22 + 3 + 2a1 a2 + 2a1 + 2a2 3a1 + 3a2 + 8 + 2a1 a2 8a1 8a2
2
b P4
15 70
2
3a1 + 3a22 6 + 2a1 a2 a1 a2 (42)

+ P2
21
2
where b = I0 N0 32 , a1 = 1 /3 and a2 = 2 /3
3a1 + 3a22 + 8 + 2a1 a2 8a1 8a2
+3P4 and P2 and P4 represent the second and fourth
280
coefficient of the orientation distribution function
(38) [equation (21)] Z200 (cos ) and Z400 (cos ),

I0 33 =
2
respectively. The angle is the angle between
2
the 0x3 axis of the structural unit and the
3a1 + 3a22 + 3 + 2a1 a2 + 2a1 + 2a2
b unique axis of the sample. For non-fiber (i.e.
15
2
biaxial) symmetry the analysis is more complex
3a1 + 3a2 6 + 2a1 a2 a1 a2
2
and the reader is referred elsewhere2,3 for further
2P2
21 details.
2

3a1 + 3a22 + 8 + 2a1 a2 8a1 8a2
+P4
35
(39) 5 EXPERIMENTAL

I0 21 =
2 DETERMINATIONS
2
a1 + a22 + 1 a1 a2 a1 a2 For the experimental set-up depicted in Figure 7,

b
15 0xyz is the laboratory axis system and 0x y z
z
z
z
y,y
Source
x
z f
x
x
y
Detector
Figure 7. Axis system for laboratory measurements of polarized Raman scattering.

is the fiber symmetry axis system. Eight inde- IZY [90]() = I0 (12 sin + 13 cos )2
pendent experiments can be performed, four
employing 180 degree backscattering and four (48)
with 90 degree scattering. Returning to equa-
IXZ [90]() = I0 22 sin 2
tion (20), the expressions for these eight exper- 2
iments are shown below where the first subscript 2
indicates the polarization of the source and the + 23 cos 2 + 33 sin 2
2
second, the polarization of the detector.
(49)

IZZ [180]() = I0 (22 sin2 IXY [90]() = I0 (12 cos + 13 sin )2
23 sin 2 + 33 cos2 )2 (43) (50)

IZX [180]() = I0 22 sin 2 When = 0 (fiber held in the vertical position),
2 equations (43) to (50) reduce to:
2
+ 23 cos 2 + 33 sin 2
2 IZZ [180](0) = I0 (33 )2 (51)
(44)
IZX [180](0) = I0 (23 )2 (52)
IXZ [180]() = I0 22
sin 2
2
2 IXZ [180](0) = I0 (23 )2 (53)
+23 cos 2 + 33 sin 2
2 IXX [180](0) = I0 (22 )2 (54)
(45)
IZZ [90](0) = I0 (33 )2 (55)
IXX [180]() = I0 (22 cos2 + 23 sin 2
IZY [90](0) = I0 (13 )2 (56)
+ 33 sin2 )2 (46)

IZZ [90]() = I0 (22 sin2 23 sin 2 IXZ [90](0) = I0 (23 )2 (57)

+ 33 cos2 )2 (47) IXY [90](0) = I0 (12 )2 (58)
When the fiber is rotated (/4) the expressions by noting that all the cross-polarization measure-
for the backscattering experiments become: ments should be the same as well as expressions
(51) and (55). Finally, to relate the experiments
1 with the rotated fiber to the other two sets one
IZZ [180] = I0 (22 223 + 33 )2
4 4 notes that:
(59)
IZZ [180](0) + IXX [180](0) + 2IZX [180](0)
1
IZX [180] = I0 (22 + 33 )2
4 4 = IZZ [180] + IXX [180]
4 4
(60)
+ 2IZX [180] (63)
1 4
IXZ [180] = I0 (22 + 33 )2
4 4 and this identity can be used to normalize the
(61) experiments with the rotated fiber to those with
the vertical fiber.
1
IXX [180] = I0 (22 + 223 + 33 )2 Following normalization one will have deter-
4 4
mined absolute values for four of the five needed
(62) parameters: equations (38), (54), equations (39),
(51), equations (40), (58) and equations (41),
A perusal of the expressions (51) to (58) and (52). For the term, I0 22 33 , we note that:
(59) to (62) shows that all the elements in the
five equations (38) to (42) are determined by these 1
12 experiments but, some experimental problems I0 22 33 = IZZ [180](0) + IXX [180](0)
2
must be overcome that are specific to the exper-
iments. In the experiments depicted above the 4IZX [180] (64)
4
incoming radiation is brought to the probe head
via an optical cable. The polarization scrambling and thus the five experimental quantities [equa-
during the transmission of the radiation is not tions (38) to (42)] can be determined. Solutions
controlled and thus some intensity calibration to these equations can be obtained using commer-
must be made when the polarization is changed cial mathematical programs such as Mathematica
from the z- to the x-axis. The second experi- (Wolfram Research) but the computer hardware
mental problem is that the focus of the analyzers required for reasonable computational times is not
in the backscattering and 90 scattering exper- always available. A solution to equations (38) to
iments are not the same so these two sets of (42) is presented in Appendix 6 that is amenable
results must be normalized. Finally, the volume to simple spreadsheet computations.
of sample exposed changes when the sample is With samples exhibiting significant birefrin-
rotated and those experimental results must be gence, polarization rotation of the incident and
normalized to the other two. Fortunately, there are scattered radiation can invalidate the experiments
a number of redundancies in the 12 experimental with the (/4) tilted samples. To circumvent this
measurements that allow one to normalize all the problem one can eliminate Equation (42) and
experimental measurements. substitute a depolarization ratio (DR) measure-
First, the cross-polarization terms in each set ment made on an isotropic sample. There are
of experiments should be the same and thus the several possible approaches to making such a
possible differences in the laser intensities in sample. A cast film which has been completely
the two polarization directions can be normal- relaxed will often produce an isotropic sample.
ized. Normalization between the backscattering Cryogrinding the polymer will sometimes produce
and the 90 scattering experiments can be made isotropic material, but the grinding process must
reduce the particle size to below 1 micron or less. where the transformation Jacobian is defined by a
Since the isotropic sample will have both P2 and determinant:
P4 equal to zero, we can express the depolariza- X X X

tion ratio in terms of the remaining constants in
u v w
Equations (39) and (41). We find:
(x, y, z) Y Y Y
=

(u, v, w) u v w (70)
a12 + a22 + 1 a1 a2 a1 a2
DR =
3a12 + 3a22 + 3 + 2a1 a2 a1 a2d Z Z Z

(65) u v w
This expression along with Equations (38) to
For the transformation from Cartesian (x, y, z)
(41) can be used to determine the elements of the
to polar coordinates (, , ) we have:
polarizability tensor and P2 and P4 .6
x = sin cos
APPENDICES y = sin sin
APPENDIX 1. TRANSFORMATION z = cos (71)

JACOBIAN
and the transformation Jacobian becomes:
Assume one has an orthogonal transformation of
(x, y, z)
coordinates defined by:
(, , )
x = X(u, v, w)
sin cos cos cos sin sin

y = Y (u, v, w)
= sin sin cos sin sin cos

z = Z(u, v, w) (66) cos sin 0
and one wishes to transform the integral: = 2 sin (72)

The units of volume in Cartesian coordinates
f (x, y, z) dx dy dz (67) are dx dy dz and are thus given in polar coordi-
xyz
nates through the Jacobian transform as:
into the new coordinate system. Defining:
vol = dx dy dz = 2 sin d d d
F (u, v, w)
4 3
= f [x(u, v, w), y(u, v, w), z(u, v, w)] (68) = (73)
3
the integral becomes: In the case of the vibrational dipole moment
vectors, we assume that the magnitude of the
dipole moment vector is constant. This is saying
f (x, y, z) dx dy dz
xyz that the intensity of the inherent absorption does
not change from oscillator to oscillator and is
(x, y, z)
=
F (u, v, w) du dv dw dependent only upon orientation with respect to
uvw (u, v, w) the incoming electric field, and one is only inter-
(69) ested in their orientation distribution. Therefore,
in the polar coordinate representation the distribu- 1

tion function depends only on and and not ,
and can be written, g(, ). One further defines a g (f,q)
normalization:

g(, ) 2 sin d d d
0
=1 (74)
2 sin d d d
sin 2f
since the integration over all space must count
all dipoles once it is normalized to the volume
1
integrated. If we now take = 1 as an arbitrary 0 1 2
definition of the intensity of a single oscillator, q(p)
we can simplify equation (74) to:
Figure 8. Integration behavior for even and odd
functions.
2
1
g(, ) sin d d = 1 (75)
4 0 0 In Figure 8, sin cos = sin 2/2 is plotted
in the interval between 0 to 2 as well as a
The normalizing (1/4 ) term is sometimes left representation of g(, ). At this point, one does
out. not know the shape or magnitude of g(, )
but the dashed line assumes that g(, ) is of
even symmetry around the rotation axis. The
APPENDIX 2. ABSORPTION assumption is made that orientation distributions
must be invariant to C2 rotations around any
INTEGRALS: RESULT OF of the three Cartesian axes. Since one function
SYMMETRY
The first cross-term integral is given as:
q

sin2f
Mxy = 2 g(, )Ex Ey x y sin d d

= 2Ex Ey 20 g(, ) sin3 sin
cos d d (76)
g(q,f)
Mxy = 2Ex Ey 20 sin3 g(, ) sin
sin3f

cos d d (77) f
0 2
We can now separate the two integrals as Angle(p)
follows. Figure 9. Integration behavior for sin2 and sin3 functions.
is even and the other is odd, the product must Since from equation (11) we have:
be an odd function in the interval 0 to 2.
As can be seen, the product of the two func- A =Axx sin2 cos2 + Ayy sin2 sin2
tions would integrate to zero over the 0 to 2
interval and hence, Axy must be equal to zero. + Azz cos2 (80)
By similar arguments Axz and Ayz can be shown
We have to evaluate:
2
(Axx sin2 cos2 + Ayy sin2 sin2 + Azz cos2 ) sin d d
A0 = A = 0 0
2 (81)
sin d d
0 0
to vanish because of the symmetry properties of

g(, ).
We find that:
Using similar arguments one can show that the
first three integrals do not vanish. For example: 2
2
sin d d = 2
0 0
Myy = Ey2 20 3
g(, ) sin sin d d 2
2 2
0 0
cos d =
2
sin2 d =
(78) 0 0

2
Now we are dealing with an even function, (1 cos2 ) sin d =
3
g(, ) and another even function, sin2 ( ). These
0

functions are plotted in Figure 9 which illustrates 1
cos2 sin d = (82)
that sin3 is always positive in the -interval 0 3
from 0 to as is sin2 in the -interval from
0 to 2. Ayy thus does not integrate to zero And thus:
if g(, ) has the shown symmetry properties.
Similar arguments can be made for Axx and 2 2 2
Azz . Axx + Ayy + A
A0 = 3 3 3 zz
2
Axx + Ayy + Azz
APPENDIX 3. ABSORPTION = (83)
3
INTEGRALS: AVERAGE
ABSORBANCE
One wishes to obtain the average of the absorption APPENDIX 4. ABSORPTION
over all angles. This is simply defined: INTEGRALS: RELATIONSHIP TO
2 P2 (COS )
A sin d d
A0 = A = 0 0 If the distribution has fiber symmetry then g(, )
2
(79)
is independent of and one finds from equa-
sin d d
0 0
tions (12) to (14) that:

2 APPENDIX 5. RELATING THE
Axx = E02 20 g( ) sin3 cos2 d d
0 ORIENTATION OF A DIPOLE OR

= E02 20 g( ) sin3 d
TENSOR AXIS TO THAT OF A
0 STRUCTURAL UNIT, AND THE
2
Ayy = E02 20 g( ) sin3 sin2 d d RELATIONSHIP BETWEEN
0
ORDER PARAMETERS AND THE

= E02 20 g( ) sin3 d ACTUAL SHAPE OF THE
0
DISTRIBUTION
2
Azz = E02 20 g( ) cos2 sin d d Neither IR nor Raman spectroscopy directly deter-
0

mines the orientation of a structural unit. Rather,
= 2E02 20 g( ) cos2 sin d (84) they measure the orientation of a transition dipole,
0 or a tensor element, associated with a vibrating
species. The alignment of the scattering tensor
Since:

element or the absorption dipole relative to the
Axx = E02 20 g( ) sin (1 cos ) d
2 structural sub-unit is vibration specific. Fortu-
0 nately, provided one knows the angle between
(85) the dipole or tensor axis and the structural
axis of interest, one can calculate the structural
Axx = E02 20 g( ) sin d orientation from the measured dipole or tensor
0 orientation.

2
g( ) cos sin d (86)
0 Uniaxial orientation (Case 1, Table 1)
Axx = Ayy = a b
Figure 10 illustrates the situation for IR absorp-
Azz = 2b (87) tion (the same analysis holds when relating the
orientation of a Raman tensor axis to a struc-
it follows that: tural unit axis). We measure the orientation of the
Azz Axx 2b a + b 3b a transition dipole (Pl (cos )) but wish to calcu-
= =
3A0 2a 2b + 2b 2a late the orientation of the chain backbone director
r(Pl (cos)).
3b 1
= (88) The relationship between these factors can be
2a 2 derived from the Legendre Addition Theorem.
By definition of the average value of cos2 , Derivation of this theorem is beyond the scope
which is the distribution function multiplied by of this article (it can be found in standard texts
the parameter integrated over all space normalized of mathematical physics), but in this situation it
to the integrated distribution function we have: reduces to:

Pl (cos ) = pl (cos )P1 (cos ) (91)
g( ) cos2 sin d
b
= 0
= cos2 (89) where pl (cos) is the l th Legendre polynomial
a
g( ) sin d in cos . (This relationship can also be derived
0 without recourse to the properties of Legendre
3 1 A Axx polynomials, by using a series of coordinate
P2 (cos ) = cos2 = zz (90)
2 2 3A0 transformations).5
These equations can easily be expressed in

terms of the observed dichroic ratios Axx /Azz
r c
and Ayy /Azz . The P2mn are identically equal
to the Z2mn summarized in Table 1, and they
q
explicitly describe the chain, rather than dipole,
b m
orientation.
Note that there are two equations but four
d c
unknowns (the P2mn ), so for each sample we
actually have to measure dichroic ratios for two
bands with different, known values of . This
provides four independent expressions that can be
c
solved simultaneously to yield the P2mn .
Obviously, without knowledge of we can
state nothing about the orientation of the chain
Figure 10. Axis system for relating molecular axis to
chain axis. axis, we have only determined the orientation of
the dipole moment for the vibration. The value of
(or the tensor axis angle) can be estimated in
In other words, the observed average orien- several ways:7
tation is given by the product of the backbone
average orientation with the Legendre polyno- measuring samples of predetermined molecular
mial in cos . If the dipole moment lies along orientation, and calculating values of consis-
the chain director (parallel band), = 0 and tent with the known data;
so Pl (cos ) = Pl (cos). For a perpendicular computation via molecular modelling or normal
band, = 90 and so we have co-ordinate analysis;
measurement of very highly oriented samples
P2 (cos ) = 12 P2 (cos ) and P4 (cos )
(for which the chain orientation is assumed to
= 38 P4 (cos ) (92) be near perfect);
estimates based on Group Theory analysis of
the main symmetry co-ordinate contributing to
IR measurement of biaxial orientation the normal mode.
The biaxial orientation distribution is defined, The latter method only provides a very crude
to fourth order, by the averages summarized in estimate, since mode mixing can make the tran-
Table 1. Jarvis et al.3 have given the relationships sition dipole angle difficult to calculate without
between the measured polarized absorbances, a full normal co-ordinate or molecular modelling
p2 (cos), and the Plmn up to second order for treatment.
c
the extended chain directors (P2mn ). These are
reproduced below.
2Azz Axx Ayy

= 2p200 (cos )P200
c
+ 4p200 (cos )P202
c
4P202
c
(93)
Azz + Axx + Ayy
Axx Ayy 4 4 c
= 4p200 (cos )P220
c
+ p200 (cos )P222
c
P222 (94)
Azz + Axx + Ayy 3 3
Relating the order parameters to the P4 = a12 + a22 + 1 a1 a2 a1 a2 (103)

shape of the orientation distribution
P5 = a12 + a22 2 4a1 a2 + 2a1 + 2a2 (104)
c
It is important to realize that thePlmn , which can P6 = a12 + a22 + 1 + 4a1 a2 + 4a1 + 4a2
be calculated from vibrational spectroscopy do not
unambiguously specify the orientation distribution = P1 2P4 (105)
function. For example, for fiber symmetry we can P7 = a12 + a22 2 + 10a1 a2 5a1 5a2
c c
only determine P200 and P400 , which only give
information on the average 2nd and 4th powers = P2 2P5 (106)
of direction cosines. IR measurements alone can
c A1 to A5 are constants determined by the exper-
only yield P200 , and there are any number of
different shapes of distribution (N ( )) which iments and defined in equations (38) to (42). , ,
c
can yield the same P200 . Lafrance et al.8 have and are numerical constants (1/15), (1/42), and
illustrated this point by showing how a range of (1/280), respectively, and a, b, c, a1 , and a2 are
Gaussian and Lorentzian distributions give rise to the unknowns. Here a is the same as the earlier
the same P200c
value. Thus we cannot calculate b(b = I0 N0 32 ), the new b is bP2 ( ) in our
without first assuming a shape for the distribution. earlier notation, c is bP4 ( ) in our earlier nota-
c
If P400 data are also available the range of possible tion while a1 (a1 = 1 /3 ) and a2 (2 = 2 /3 )
distributions can be constrained more tightly, but retain their earlier definitions.
still cannot be uniquely specified.9 To solve equations (95) to (99) one first elim-
inates b from equations (95) and (96) by adding
A2 to 2 A1 :
APPENDIX 6. SOLUTION TO
EQUATIONS 3aP1 + 14 cP3 = 2A1 + A2 (107)
Citra (M. Citra, unpublished work) derived the Similarly, adding 2 A4 to A3 yields:
following solution to the five equations (38) to
(42) with their five unknowns. Equations (38) to 3aP4 7 cP3 = A3 + 2A4 (108)
(42) can be rewritten:
A pair of equations from which a has been
aP1 + 2bP2 + 3 cP3 = A1 (95) eliminated may be obtained from equations (95)
aP1 4bP2 + 8 cP3 = A2 (96) to (98) by subtracting A2 from A1 and A4 from
A3 . The results are:
aP4 + 2bP5 + cP3 = A3 (97)
aP4 bP5 4 cP3 = A4 (98) 6bP2 5 cP3 = A1 A2 (109)
aP6 bP7 4 cP3 = A5 (99) 3bP5 + 5 cP3 = A3 A4 (110)
where Eliminating c from equations (107) and (108)

yields
P1 = 3a12 + 3a22 + 3 + 2a1 a2 + 2a1 + 2a2
3a(P1 + 2P4 ) = 2A1 + A2 + 2A3 + 4A4
(100)
(111)
P2 = 3a12 + 3a22 6 + 2a1 a2 a1 a2 (101) Eliminating c from equations (109) and (110)
yields
P3 = 3a12 + 3a22 + 8 + 2a1 a2 8a1 8a2
(102) 3b(2P2 + P5 ) = A1 A2 + A3 A4 (112)
To obtain an equation analogous to equa- Equations (118), (117), (115), (116), and (119)
tions (111) and (112) that involves c, one substi- become, respectively:
tutes equations (105) and (106) into equation (99)
to obtain: 15aP1 = B1 (125)
21bP2 = B2 (126)
aP1 2aP4 bP2 + 2bP5 4 cP3 = A5
(113) 35 cP3 = B3 (127)
Multiplying equation (112) by 6 and re-
15aP4 = B4 (128)
arranging, one obtains:
21bP5 = B5 (129)
2(3aP1 )4(3aP4 ) (6bP2 ) + 4(3bP5 )
The general method for solving the above
24 cP3 = 6A5 (114) equations would be to eliminate a from equa-
Eliminating the quantities in parentheses on tions (125) and (128) and b from equations (126)
the left-hand side of equation (114) using equa- and (129) and solving first for a1 and a2 . It is
tions (107) to (110), rearranging, and dividing more convenient to work with the variables:
by 3, one obtains:
w = a1 a2 (130)
35 cP3 = A1 + A2 4A4 2A5 (115) x = a1 + a2 (131)
Equation (115) can be used to eliminate c from In terms of these new variables, equations (100)
equation (108). The result is: to (103) become:
15aP4 = A1 + A2 + 5A3 + 6A4 2A5 (116) P1 = w2 + 2x 2 + 2x + 3 (132)
Similarly, one can use equation (115) to elimi- P2 = w + 2x x 6

2 2
(133)
nate c from equation (109) to obtain: P3 = w + 2x 8x + 8
2 2
(134)
21bP2 = 4A1 3A2 2A4 A5 (117) 3w2 + x 2 4x + 4
P4 = (135)
4
One can use equation (116) to eliminate P4 3w x + 4x 4
2 2
from equation (111) and equation (117) to elimi- P5 = (136)
nate P2 from equation (112) yielding: 2
One first considers equations (125) to (129) for
15aP1 = 8A1 + 3A2 + 8A4 + 4A5 (118) the general case where none of the constants, B1
21bP5 = A1 A2 + 7A3 3A4 + 2A5 to B5 , vanish.
Eliminating a from equations (125) and (128)
(119) and substituting equations (132) and (135), yields
the following:
To simplify, one defines the constants:
B4 3B1 B1
B1 = 8A1 + 3A2 + 8A4 + 4A5 (120) w =
2
2B4 x 2
4 4
B2 = 4A1 3A2 2A4 A5 (121) (2B4 + B1 )x + (B1 3B4 )
B3 = A1 + A2 4A4 2A5 (122) (137)
B4 = A1 + A2 + 5A3 + 6A4 2A5 (123)
Likewise, eliminating b from equations (126)
B5 = A1 A2 + 7A3 3A4 + 2A5 (124) and (129) and substituting equations (133) and
(136) yields: Either (B4 3B1 /4) or (B5 3B2 /2)

vanishes but not both
3B2 B2
B5 w = 2B5 +
2
x2
2 2 In this case equation (143) is still the solution
+ (B5 + 22 )x + (6B5 2B2 ) for x and the non-vanishing equation [either
equation (137) or (138)] is used to solve for w2 .
(138)
Both (B4 3B1 /4) and (B5 3B2 /2)

(B4 3B1 /4) and (B5 3B2 /2) both vanish
non-vanishing
If the right-hand sides of equations (137) and
Eliminating w2 from equations (137) and (138) (138) differ only by some multiplicative factor,
yields: then either can be used to solve for x, but w2 is
0 = C1 x 2 2C2 x + C3 (139)
where

3B2 B1 3B1 B2
C1 = B5 2B4 + B4 2B5 + (140)
2 4 4 2

3B2 3B1
B5 (2B4 + B1 ) + B4 (B5 + 2B2 )
2 4
C2 = (141)
2

3B2 3B1
C3 = B5 (B1 3B4 ) B4 (6B5 2B2 ) (142)
2 4
indeterminate. Otherwise the equations have no

The solution for equation (138) is:
solution.

C2 + C2 C1 C3
x= (143)
C1 ABBREVIATIONS AND
ACRONYMS
Substituting equation (143) back into either
equation (137) or (138) will result in two solu- DR Depolarization Ratio
tions for w2 . The resulting pair of solutions IR Infrared
for (x, w2 ) can then be substituted into equa-
tions (132) to (134) to obtain P1 , P2 , and P3 , and
then the solutions for a, b, and c follow from REFERENCES
equations (125), (126), and (127). Inverting equa-
1. J. Chalmers and N. Everall, Analusis, 28, 53 (2000).
tions (130) and (131) yields
2. D.I. Bower, J. Polym. Sci., Polym. Phys. Ed., 10,
2135 (1972).
x w2
a1 = (144) 3. D.A. Jarvis, I.J. Hutchinson, D.I. Bower and I.M.
2 Ward, Polymer, 21, 41 (1980).

x w2 4. R.J. Roe and W.R. Krigbaum, J. Chem. Phys., 40,
a2 = (145) 2608 (1964).
2
5. R.R. Birss, Symmetry and Magnetism, North- 8. C.-P. Lafrance, A. Nabet, R.E. Prudhomme and
Holland Publication Co, Amsterdam, 5051 (1964). M. Pezolet, Can. J. Chem., 73, 1497 (1995).
6. S. Frisk, R.M. Ikeda, D.B. Chase and J.F. Rabolt, 9. D.I. Bower, J. Polym. Sci., Polym. Phys. Ed., 19, 93
Appl. Spectrosc., 57, 1053 (2003). (1981).
7. N. Everall, Spectroscopy, 15, 38 (2000).
Vibrational Circular Dichroism
Laurence A. Nafie1,3, Rina K. Dukor2,3 and Teresa B. Freedman1

1
Syracuse University, Syracuse, NY, USA
2
Vysis, Inc., Downers Grove, IL, USA
3
BioTools, Inc., Elmhurst, IL, USA
1 INTRODUCTION 18 years later by the first dedicated commercial VCD

spectrometer,21 one based on the FT-IR measurement
1.1 Brief history of vibrational optical design. For ROA, the key advances to modern instruments
activity were the introduction of multi-channel detection,22,23 most
importantly, backscattering geometry,22,24 charge-coupled
Vibrational circular dichroism (VCD)1 4 is one of two prin- device solid-state detectors,23,25 and multiple modulation
cipal forms of vibrational optical activity (VOA).5 7 The schemes involving the manipulation of the circular polar-
other form is vibrational Raman optical activity (ROA).8 12 ization states of both the incident and scattered radiation
Both ROA13,14 and VCD15,16 were discovered in the early beams as provided in the definitions below.26 29 ROA is not
1970s after a number of key technological developments. yet commercially available, but that achievement appears to
For VCD, these were solid-state electronics and lock-in be close at hand.11,30
amplifiers, liquid-nitrogen-cooled semiconductor detectors, Dramatic progress in the theoretical interpretation and
and infrared (IR) polarization modulators.17 These pro- simulation of VOA spectra has been achieved in recent
vided stable electronic control of experiments, background- years. Early methods involved reliance on various models
limited detection of IR radiation, and high-frequency polar- of VCD and ROA intensity that provided only marginally
ization modulation sufficiently far into the IR region. For reliable predictive power for understanding VOA spectra.
ROA, the key developments were the argon-ion laser, In the 1980s, the theoretical formalism and methodology
cooled-photomultiplier detectors and photon-counting elec- were developed for calculating VCD31 34 and ROA35,36
tronics. With these advances the necessary sensitivity from first principles without reliance on approximate mod-
became available to measure accurately circular polariza- els. With the advent of more powerful computers and the
tion difference spectra that are smaller in intensity by availability of software for carrying out ab initio quan-
several orders of magnitude compared to their parent IR tum calculations, VCD and ROA can now be simulated
or Raman spectra. The first samples for which VCD and at a high level of accuracy for medium-sized molecules,
ROA were measured were neat liquids of simple, chiral including those with as many as 70 heavy atoms in
organic molecules. Over the next several years, measure- the case of VCD.37 Software for reliable calculations of
ments of VCD and ROA were carried out for most classes VCD is now commercially available,38 making it possi-
of biological molecules including amino acids, peptides, ble for nonspecialists for the first time to measure and
carbohydrates, polypeptides, proteins, and nucleic acids. calculate VCD spectra for wide varieties of stereochemi-
Key instrumental developments for the modern-day mea- cal applications. ROA of large biological molecules must
surement of VCD included the advent of Fourier trans- still be interpreted empirically, at least for the foresee-
form infrared (FT-IR) VCD in 197918 20 and followed able future. The stage is now set for a new era of
widespread application of VCD spectroscopy with ROA
John Wiley & Sons Ltd, 2002. soon to follow.
1.2 Definition of VCD and ROA arising from electronic transitions in the ultraviolet (UV)
and visible regions of the spectrum to the IR and near-
VCD and ROA are both chiroptical spectroscopies that infrared (NIR) regions where predominantly vibrational
probe the stereochemical structure of chiral molecules transitions occur. Figure 1 shows an energy level diagram
through their vibrational transitions. Both are defined as dif- to illustrate the definitions of VCD and ROA. Here we see
ferences in the vibrational spectrum of a molecule for left for the transition, say, from the zeroth to the first vibrational
versus right circularly polarized (RCP) radiation.5 The def- sublevel of the ground electronic state that VCD is defined
inition of VCD is simpler than that needed for ROA. VCD as the difference in the IR absorbance, A, for left circularly
is simply the extension of traditional circular dichroism polarized (LCP) minus that for RCP radiation given by the
Figure 1. Energy level diagram illustrating the definitions of VCD and four forms of ROA for a molecular transition from the zeroth
vibrational level of the ground electronic state, g0, to the first vibrational level of the ground electronic state, g1.
Vibrational Circular Dichroism 3
equation and why it is used. VCD combines the stereo-sensitivity

AnN D AL nN AR nN 1 inherent in all forms of optical activity, where mirror-image
molecules exhibit oppositely signed spectral intensities,
Here nN represents the spectral frequency in wavenumbers. with the rich structural specificity of IR spectroscopy.
The definition of ROA is more complex owing to the fact This structural sensitivity relative to UVvisible absorp-
that a Raman transition involves principally the incident tion spectroscopy is due to the large number of vibrational
laser radiation and its polarization, the scattered Raman transitions representing all the internal oscillatory degrees
radiation and its polarization, and the angle of scattering of freedom of the molecule. Contrary to electronic circu-
f. The scattering angle can vary from forward scattering lar dichroism, there is no need to be concerned whether
to right-angle scattering to backscattering, and in principle, the sample molecule has a chromophore since every vibra-
all angles in between. The original form of ROA is called tional transition represents an IR chromophore, and all
incident circular polarization (ICP) ROA. In this form of molecules display many vibrational transitions throughout
ROA, the polarization state of the incident laser radiation the IR region. In addition, VCD possesses the advan-
only is switched between RCP and LCP states according to tage associated with IR spectroscopy wherein all molecules
the definition,
can be measured directly without the need for chemical
Ia f, nN D IRa f, nN ILa f, nN 2 modification, diastereomeric interaction, crystallization or
enantiomeric separation in order to obtain the desired chi-
where the superscript refers to the polarization state of roptical information.
incident radiation and the subscript refers to the polarization The simplest application of VCD is the determination
state of the scattered radiation, in this case the fixed, of the enantiomeric purity of a sample relative to a known
nonelliptical state, a. A second class of ROA spectra is standard.21 Since VCD scales linearly with the enantiomeric
called scattered circular polarization (SCP) ROA. Here, purity of a sample, the relative enantiomeric purity of two
the incident laser radiation is held in a fixed, noncircular samples can be determined directly by the measurement of
polarization state, and the ROA is measured by isolating their VCD spectra. This can be an advantage when enan-
and subtracting the RCP and LCP components of the tiomeric purity measurements are needed, either quickly
scattered Raman radiation with the definition, to some desired level of accuracy, as in real-time process
monitoring of a stereo-controlled reaction involving chiral
Ia f, nN D IaR f, nN IaL f, nN 3 molecules, or if the enantiomers are difficult to separate by
Finally, there are two forms of ROA in which the polariza- column chromatography.
tion states of both the incident and scattered radiation are A second, important area of application is the determina-
modulated between RCP and LCP states, either in-phase tion of absolute configuration. This is a relatively new use
with respect to polarization state or out-of-phase. These two of VCD, one that has risen to prominence with the arrival
forms are referred to as in-phase dual circular polarization of accurate ab initio calculations of VCD spectra.37,39 In
(DCPI ) ROA and out-of-phase dual circular polarization principle, the method is quite simple. One measures the
(DCPII ) ROA and have the definitions, VCD spectrum of the chiral molecule of interest for which
the absolute configuration, its exact three-dimensional rep-
II f, nN D IRR f, nN ILL f, nN 4 resentation in space with respect to its mirror image, is
unknown. The VCD spectrum of the same molecule is sim-
III f, nN D IRL f, nN ILR f, nN 5
ulated, as described later in this article, using an ab initio,
In the remainder of the article, we focus our attention on quantum-chemistry program. A particular absolute config-
VCD. Many excellent recent reviews are available for those uration is chosen for the quantum calculation. One then
interested in obtaining more information about ROA, as compares the experimentally measured VCD spectrum to
well as the next article in this Handbook, Raman Optical the theoretically calculated VCD spectrum. If the bands
Activity Spectroscopy.8 10 Recent reviews of VCD are correlate and agree in sign, the theoretical absolute config-
also available,1,3,4 some of which cover both VCD and uration corresponds to that of the unknown sample. If they
ROA in the same review.5 7 are opposite, the absolute configuration of the unknown
molecule is opposite to that which was calculated. If the
calculated and measured VCD spectra can be correlated
1.3 Areas of application of VCD in this way, the absolute configuration of the molecule
is unambiguously determined without having to crystal-
VCD spectra can be viewed from several perspectives. lize it and obtain its absolute structure from phase-specific
These help to gain an appreciation for where VCD is used X-ray diffraction. To date, hundreds of determinations
of absolute configuration have been successfully carried quantum calculations of VCD spectra and show how these
out using VCD, making this a well-documented nearly calculated spectra are compared to experimental VCD spec-
routine approach to the determination of absolute molecular tra. The theoretical calculations are now the basis for the
stereochemistry. detailed interpretation of VCD as well as their application to
A third, more long-standing, use of VCD is the deter- the determination of the absolute stereochemical structure
mination of the solution conformations of large and small of chiral molecules, primarily in solution. In this article, we
biological molecules. In the case of large molecules, the will also describe some recent advances in VCD instrumen-
conformations are typically the backbone conformations of tation and indicate directions where VCD may fruitfully be
proteins or nucleic acids. The procedure for these kinds applied in the future. Since many reviews are readily avail-
of conformational studies, as mentioned above, must be able among the references at the end of this article, we will
primarily empirical4,40 43 due to the current limitations limit the number of examples presented to those needed to
of molecular size for accurate quantum calculations. For illustrate the principles of the measurement, calculation and
smaller molecules, including oligopeptides of up to five spectral interpretation of VCD.
or six residues, ab initio calculations can be used to sort
out the populations of conformers present at a particular
temperature.1,7 Here VCD is making a unique contribution 2 THEORETICAL BACKGROUND
to our understanding of the structure of smaller biologi-
cal molecules since these conformers typically inter-convert 2.1 Connection between theory and experiment
faster than the NMR timescale and hence cannot be stud-
ied in detail by this technique. VCD spectra, and all other The intensity for the absorption of IR radiation during
optical spectra, correspond to a linear superposition of the vibrational excitation is expressed as absorbance, A D
spectra associated with each of the conformers populated to logI/I0 or molar absorptivity e D A/c , where c is the
a significant degree in solution. By contrast, the NMR spec- molar concentration and is the pathlength of the sam-
trum represents the average of the spectra of the species ple. VCD, the difference in absorption of LCP and RCP
present, and the conformations of individual populations IR radiation, is expressed as differential absorbance, A D
cannot be recovered. AL AR , or differential molar absorptivity e D eL eR .
The largest single area of application of VCD is the study A chiral molecule with N atoms will have 3N 6 normal
of the structure and conformation of biologically significant modes of vibration. From quantum mechanics, a fundamen-
molecules. These are the traditional biological molecules tal vibrational transition for a normal mode a corresponds to
where nearly all classes have been studied extensively quantum excitation for a molecule in the ground electronic
except membranes and lipids, but there are no intrinsic bar- state and ground vibrational state, ya0 j0a i, to the first
riers to studying these molecules, many of which possess excited vibrational state of mode a in the ground electronic
one or more chiral centers. An important subset of biolog- state, ya1 j1a i. The absorption intensity is proportional to
ical molecules is that of pharmaceutical molecules. Large the dipole strength, the absolute square of the electric dipole
numbers of new chiral pharmaceutical molecules are syn- transition moment,5,6
thesized or isolated each year. For many of these molecules 2
2
the absolute configuration is unknown. VCD is a technique a
D01 D y0 my1 dt 0a mO 1a
a O a
6
that can be used to determine the absolute configuration of
these molecules and address a wide variety of stereochem-
where the electric dipole moment operator is mO D i qi Eri
ical problems of pharmaceutical and biological interest.
for all charged particles in the molecule with charge qi
(Ze for a nucleus with atomic number Z, and e for
an electron) and position Eri . Experimentally, the dipole
1.4 Goals of this article strength is obtained from the frequency-weighted integrated
area of the absorption band for mode a,
The goal of this article is to present the principles and appli-
band
cations of VCD. A chapter devoted to the related form of a en
D01 D 9.184 1039 dn
VOA, called ROA, is given in Raman Optical Activity n
Spectroscopy. In the present article, we will cover both the
9.2 1039 band
theoretical and instrumental methodology used to measure en dn 7
n0
VCD spectra and relate these spectra to theoretical quan-
tities that emerge from the theory of VCD. We will also where in practice the wavenumber frequency n can be con-
explain the rudimentary aspects of carrying out ab initio sidered constant over the absorption band, replaced by the
frequency n0 at the band maximum and removed from strength involves the change in the magnetic moment of the
the integration as shown in equation (7). If we assume a molecule, m,
E with the conjugate normal mode momentum
Lorentzian band shape with half-width g at one-half the Pa D QP a D i!a Qa (!a D 2pnc),
band maximum e0 , the dipole strength is further approxi-
mated as a h @mE 2

D01 D 13
2!a @Qa 0
g
D 9.2 1039 e0 p 8
n0
a h @m E @m
E
In these expressions, the dipole strength is in Gaussian R01 D 14
2 @Qa 0 @Pa 0
units of esu2 cm2 , where esu represents electrostatic units.
In SI units, absorption intensity is given in km mol1 , which VCD is thus seen to arise from simultaneous linear oscil-
is converted to esu2 cm2 by multiplication by 3.987 lation of charge and angular oscillation of charge produced
1037 /n. by the vibrational motion. In achiral molecules the nonzero
VCD intensity is proportional to the rotational or rotatory electric and magnetic dipole moment derivatives for a given
strength, the scalar product of the electric and magnetic vibrational mode are always orthogonal, but for chiral
dipole transition moments, molecules these derivatives can have parallel or anti-parallel
components, leading, respectively, to positive or negative
a a O a a a rotational strength.
R01 D Im y0 my1 dt y1 m O y0 dt
a a a These components of the electric and magnetic dipole
Im 0a m O 1 1 m O0 9 derivatives, given in equations (13) and (14), can be further
broken down into atomic contributions,
where m O D i qi /2mi cEri p
E i is the magnetic dipole
@mb
@mb
moment operator, a pure imaginary operator proportional D SAa,a 15
to the angular momentum operator Eri p E i ihEri @/@Eri @Qa 0 A,a
@RA,a 0
for the ith particle with mass mi and linear momentum
@m b
@mb
pi . The imaginary part of the scalar product is taken in D SAa,a 16
order to obtain the rotational strength as a real quantity. @Pa 0 A,a
@RP A,a 0
The rotational strength is obtained experimentally from the
frequency-weighted area of the VCD band, where a and b represent Cartesian x, y or z directions, RAa
is the displacement of atom A in the a-direction, RP Aa is
band
a en the velocity of atom A in the a-direction, and the s-vector
R01 D 2.296 1039 dn
n SAa,a for atom A in mode a can be expressed in terms of
either nuclear displacement or nuclear velocity,
2.3 1039 band
en dn 10
n0 @RAa @RP Aa
SAa,a D D 17
and can be approximated for a Lorentzian band from the @Qa 0 @Pa
0
half-width at half-height and the frequency and intensity at
the band maximum, From equations (13), (14) and (15), we see that both the
A
dipole and rotational strengths involve components Pab of
g
R 2.3 1039 e0 p 11 the atomic polar tensor (APT) for atom A,
n0
A @mb
Expressing both D and R in esu2 cm2 allows definition of Pab D D EA A
ab C Nab 18
@RAa 0
the dimensionless parameter g, the anisotropy ratio, as
e 4R with nuclear contribution
gD D 12
e D NA
ab D ZA e ab 19
which connects experiment to theory.
and electronic contribution

@ E
2.2 Further theoretical formulation EAab D y g O
m b y g 20
@RAa RD0
In the harmonic approximation, the dipole strength can where mO Eb is the electronic contribution to the electric dipole
be related to the change in the electric dipole moment moment operator.
of the molecule, mE , with the mass-weighted nuclear dis- The rotational strength (see equations (14) and (16)) also
placement for the normal mode Qa . Similarly, the rotational involves components MA ab of the atomic axial tensor (AAT)
for atom A, the use of dispersive VCD is relatively limited, we will

consider only the performance of FT-VCD spectrometers
@mb
MA
ab D D IA A
ab C Jab 21 in the remainder of this section.
@RP Aa 0
with nuclear contribution

ZA e 3.2 Double modulation FT-VCD measurements
JAab D
0
eabg RA g 22
gDx,y,z
2c
Double modulation FT-IR VCD was first measured over
where eabg D 1 for even permutations of the order x, y, z, 20 years ago.18,19 The basic idea is to introduce a source
eabg D 1 for odd permutations of the order x, y, z and of high-frequency polarization modulation into the sample
eabg D 0 if any two of a, b, g are the same, and electronic chamber just before the sample.44 This can be achieved
contribution with a linear polarizer and an IR photoelastic modulator
(PEM). For spectral coverage in the mid-infrared the mod-
A @ Q E Q
Iab D yg m O b yg 23 ulator material is ZnSe and the modulation frequency is
@RP Aa P
RD0,RD0 close to 37 kHz. This frequency is much higher than the
band of Fourier frequencies, and the information carried at
O bE is the electronic contribution to the
In equation (23), m
this frequency is the interferogram of the raw VCD inten-
magnetic dipole moment operator and the tilde denotes a
sity of the sample. Hence there are two kinds of modulation
complex quantity.
present at the detector, Fourier modulation by the interfer-
ometer in the range of 1001000 Hz, and the polarization
3 EXPERIMENTAL MEASUREMENT OF modulation frequency at 37 kHz. This leads to the simulta-
neous presence at the detector of two interferograms. One
VCD is the ordinary IR transmission interferogram of the sam-
3.1 Fourier transform versus dispersive VCD ple, the so-called direct current interferogram, and the other
instrumentation is the VCD interferogram centered at the PEM modulation
frequency, called the alternating current interferogram. The
Dispersive VCD instruments predate Fourier transform (FT) alternating current interferogram can be reduced to the nor-
VCD instruments by approximately 5 years.19 With the mal Fourier range of frequencies by demodulating it from
advent of FT-VCD measurements, all the advantages that the PEM carrier frequency by a lock-in amplifier (LIA).
FT-IR possesses compared to dispersive IR were realized The LIA must be set with a microsecond time constant to
for VCD measurements. Due to spectral multiplexing, wide allow the Fourier frequencies to pass to its output without
spectral ranges could be measured simultaneously, resolu- significant attenuation. Fourier transformation of these two
tion could be increased without additional loss of radiation interferograms leads to the isolation of the VCD and IR
and blocks of spectra could be accurately co-added without absorbance spectra as defined above and explained in more
the worry of mechanical drift of the frequency scale. When detail below.
these advantages are coupled with the fact that nearly all IR The block diagram of the optical-electronic set-up is
instruments on the market over the past decade have been shown in Figure 2. An IR beam taken from an auxiliary port
FT-IR instruments, FT-VCD is the overwhelming choice of an FT-IR spectrometer is directed to a linear polarizer P,
as an instrumental approach to VCD measurement. The followed by the PEM that creates alternately LCP and RCP
only serious drawback that FT-VCD instruments have rel- IR radiation. The sample, X, is placed immediately after the
ative to dispersive instruments is detector saturation. If one PEM where the sample imposes an intensity modulation
is interested in a narrow range of vibrational frequencies, on the IR beam at the PEM frequency when VCD is
dispersive instruments can hold an advantage compared to present in the sample. For VCD to be present, the sample
broad band FT-VCD measurements where the amount of molecule must be chiral such that its mirror image is not
IR radiation reaching the detector must be controlled to superimposable on itself. The IR beam is then focused on
avoid detector saturation. Saturation is essentially never the detector D, where the signal is first pre-amplified and
encountered in dispersive IR measurements, where only one processed further electronically. The detector signal can
narrow band of frequencies, corresponding to the spectral be represented as the sum of two terms by the following
resolution, is focused on the detector. Hence for a narrow equation,
spectral range, it is currently possible to achieve marginally
ID nN D IDC nN C IAC nN D I0 nN 10ANn
higher signal-to-noise ratios, albeit at a lower spectral res-
olution, with dispersive VCD compared to FT-VCD. Since f1 C 1.1513AnN sin[M nN ]g 24
ratio taken, the final circular dichroism spectrum, AnN , is

obtained from equation (27) through the relationship

0 1 IAC n N
AnN D [2J1 [M nN ]1.1513] 28
IDC nN
FT-IR P PEM X D ID
where the quantity in square brackets can be obtained
REF from a calibration spectrum using a multiple waveplate
IAC as described previously.44 The first-order Bessel function
J1 [0M nN ] describes the efficiency of the PEM modulation
A LIA at each wavenumber frequency in the spectrum.
IDC
It is interesting to compare the normalization of the
absorbance spectrum, AnN , to the VCD spectrum. The
DIV absorbance is obtained from the ratio of the direct current
Figure 2. Block diagram showing the basic optical and electronic intensity spectra with and without the sample. This ratio
components needed to measure an FT-VCD spectrum with a removes all spectral contributions due to the characteristics
single PEM. See text for the definitions of the abbreviations used.
of the FT-IR spectrometer leaving a spectrum of just the
The first term in this expression, IDC nN , represents that sample. Similarly the alternating current intensity spectrum
part of the detector signal that is constant with respect to is ratioed to the direct current transmission spectrum to
the PEM modulation. IDC nN represents the intensity (i.e. remove the particular spectral dependence of the FT-VCD
single-beam) spectrum associated with the sample that has instrument.
an absorbance spectrum, AnN , and I0 nN is the intensity
at the detector in the absence of a sample. From these
intensities the expression for the absorbance spectrum is 3.3 Commercial VCD instrumentation
given by
IDC nN Until 1997, all VCD spectrometers were constructed by
AnN D log10 25
I0 nN the user, in some cases with assistance from an FT-IR
manufacturer, as an accessory to a stand-alone dispersive
The alternating current term in equation (24), IAC nN , rep- IR or FT-IR spectrometer. Since then it has been possible to
resents the intensity spectrum associated with the circular purchase a pre-assembled, pre-aligned dedicated FT-VCD
dichroism oscillating at the PEM frequency. The sinusoidal spectrometer, the Chiralir, from Bomem-BioTools (Quebec,
PEM term can be written as an expansion of odd-order Canada). Also in recent years, most other manufacturers of
Bessel functions as FT-IR instruments, including Bruker Instruments, Nicolet
sin[M nN ] D sinf[0M nN ] sin !M tg Instruments and Bio-Rad, offer to varying degrees, VCD

accessory benches that are optimized in the laboratory of
D 2Jn [0M nN ] sinn!M t 26 the user after delivery. The Chiralir is unique to date, in
n,odd that it is constructed on one optical platform that includes
The alternating current term is detected by an LIA tuned to the interferometer and all the VCD optical components. The
the modulation frequency of the PEM, !M , and the higher- PEM is mounted at the time of assembly and the VCD zero-
order harmonics are rejected by the lock-in which measures baseline and all performance characteristics are determined
the magnitude of the sin !M t term. This magnitude is the prior to shipment. Thus, VCD capability is available at the
raw circular dichroism intensity term and we can then time of installation. This relieves the user of the need to
rewrite equation (24) as become an expert in VCD technology and to focus instead
on using the dedicated VCD software for stereochemical
ID nN D IDC nN C IAC nN D I0 nN 10ANn applications of interest.
[1 C 2J1 [0M nN ]1.1513AnN ] 27
This is the basic equation for the spectral intensity at the 3.4 Intensity standards for VCD spectra
detector for a VCD measurement using a PEM. The alter-
nating and direct current terms in this equation are shown in There are two principal intensity standards for VCD spectra
Figure 2 where the alternating current term emerges from in the mid-infrared regions. One is neat ()-(1S)-a-pinene
the LIA. If the alternating and direct current interferograms measured with a pathlength of 75 m. This spectrum is
are Fourier transformed, as indicated in Figure 2, and their relatively easy to measure in a time period from 5 to 20 min
8 depending on the sensitivity of the VCD spectrometer.

The absorbance and VCD spectra, together with the VCD
6
noise spectrum, are presented in Figure 3. The signs of the
4 bands for this absolute configuration are known definitely
A 10 4
and have been confirmed by the results of ab initio VCD

2
calculations. The intensities have been confirmed to be the
0 values shown to within an accuracy of approximately 10%
by comparison of the results of this measurement using
2
different instruments in several locations.
4 A second standard is the VCD spectrum of ()-(1S,4S)-
camphor as a 0.6 M solution in CCl4 . Here again the
1.0 signs have been confirmed with ab initio techniques, and
the relative magnitudes agree for measurements at dif-
0.8
ferent locations within 10%. Any mid-infrared VCD can
Absorbance
0.6 be calibrated by reference to these standard VCD spec-

tra. The absorbance and VCD spectra for ()-(1S,4S)-
0.4 camphor are presented in Figure 4 under two different
sampling conditions, a solution in CCl4 , and also as
0.2
a mull in nujol oil.45 It is clear that these two sam-
pling methods yield nearly the same VCD spectrum.
0.0
1300 1200 1100 1000 900 This can be achieved for solid samples when the con-
Wavenumber / cm1
stituent sample molecules do not interact strongly in
the solid and where the particle size of the solid is
Figure 3. The unpolarized absorbance, A, and VCD, A, for reduced to below the wavelength of the IR radiation.
()-(1S)-(a)-pinene as a neat liquid. The pathlength used was For larger particle sizes, the IR and VCD spectra con-
69 m and the spectral collection time for the VCD is 20 min for
each enantiomer at 4 cm1 on the Chiralir with baseline correc- tain contributions from the scattering of the IR radiation
tion from the opposite enantiomer under the same experimental by the individual particles in the sample. VCD spectra
conditions. from crystalline samples containing large particles thus
Camphor in mull Camphor in CCl4 solution

4
3
2
A 10 4
1
0
1
2
0.8
Absorbance
0.6
0.4
0.2
0.0
1500 1400 1300 1200 1100 1000 900 1400 1300 1200 1100 1000 900
Wavenumber / cm1 Wavenumber / cm1
Figure 4. Comparison of the absorbance and VCD spectra of ()-(1S,4S)-camphor as a 0.6 M solution in CCl4 at a pathlength of
150 m and as a small-particle-size mull of the corresponding solid sample. The two sets of spectra are nearly the same.
differ significantly from the corresponding solution-phase calculated frequencies and intensities to Lorentzian bands
VCD spectra. of the desired half-widths to produce a calculated spectrum
for display with graphing software.
Although commercial software for ab initio calculation
3.5 Optimal measurement procedures of IR intensities has been available for many years, imple-
mentation of methods to calculate VCD spectra has only
The optimal measurement of a VCD spectrum starts with recently been made widely available, through the release
the optimal measurement of the underlying IR spectrum. of Gaussian 98.38 The method for calculating the electronic
The absorbance values for the bands of interest should contributions to the APTs, equation (20), involves taking
lie in the range of 0.1 and 1.0. For a background-noise derivatives of the electronic energy of the molecule with
limited detector, as are most IR detectors, the optimum respect to electric field E and nuclear position,
trade-off between transmission and absorption corresponds 2
to an absorbance value of 0.434. To optimize the signal-to- @ Eel
EA ab D 29
noise ratio for a given VCD spectrum, one should aim for an @Eb @RAa 0
average bulk IR absorbance, not necessarily average peak
absorbance, across the spectrum of between 0.4 and 0.5. For These derivatives are calculated with coupled-perturbed-
VCD measurements, the maximum efficiency of the PEM, HartreeFock (CPHF) methods in Gaussian 98 and other
the maximum of the Bessel function J1 [0M nN ], should be molecular orbital packages. The electronic contribution to
centered in the region of interest. If the baseline of the the AATs (equation 23) is calculated from
VCD instrument is not sufficiently close to zero for a single 2
A @ Eel
measurement, some method of baseline correction needs to Iab D 30
@Hb @RP Aa 0
be applied. The most accurate baseline measurement is the
VCD spectrum of the sample of the opposite enantiomer where the derivative in equation (30) is taken with respect
of the chiral sample molecule measured under the same to magnetic field H and nuclear velocity. The deriva-
conditions of concentration and pathlength using the same tive with respect to nuclear velocity requires a non-
cell. This is the method used in Figure 2. Subtraction of the BornOppenheimer formulation. Two of the main for-
two opposite-enantiomer VCD spectra with division by two malisms for calculating the AAT are the vibronic coupling
gives the corrected VCD spectrum of a single chiral species. theory (VCT) method, where IA ab is calculated as a sum
Alternatively, the measurement of racemic mixture under over excited electronic states:
identical sampling conditions provides a direct measure of eh2

the VCD baseline but without the benefit of additional IA

ab D ebg
mc g e6Dg
VCD spectral intensity, and hence is less desirable. If

neither the opposite enantiomer nor the racemic mixture

@ye @yg
is available, which is the case for complex biological yg rjg ye E1 31
molecules such as DNA or proteins, then the VCD baseline @rj @RAa eg

j
RD0
of the solvent or an achiral sample with similar absorption
characteristics provides, in many cases, a suitable correction and the magnetic field perturbation (MFP) method, where
to the VCD baseline. IA
ab is calculated as the overlap of derivatives of the elec-
tronic wavefunction with respect to magnetic field and
nuclear displacement:

Q g @y
4 SPECTRAL CALCULATION OF VCD Qg
A @y
Iab D 2ih
4.1 Ab initio formulations of VCD @Hb @RAa
HD0,RD0

@yQ g @y Q
Calculation of VCD spectra first involves selecting a D 2ih g 32
@RAa @Hb
method, most often either an ab initio HartreeFock or HD0,RD0
a density functional theory (DFT) calculation, and a basis
set, most commonly 6-31G* or 6-31G(d). Next, one cal-
culates the optimized geometry, the vibrational force field, 4.2 Implementation of theoretical calculations
and normal modes. Next the APTs and AATs are calcu-
lated, which are then used with the atomic displacements The VCT method has been implemented at the University
from the normal modes, via equations (13) to (14) to obtain of Calgary and at Syracuse University as a program utilizing
the IR and VCD intensities.6 Finally, one converts the the output of Gaussian 94 or 98. The MFP method was first
implemented with the CADPAC molecular orbital package, size of the molecule. Recently, Kubelka et al.47 have trans-
and is now commercially available with Gaussian 98 for ferred ab initio parameters from peptide subunits to calcu-
gauge invariant atomic orbitals, which eliminates the ori- late the VCD of larger peptide molecules, opening the way
gin dependence of the rotational strength calculation. We to calculating VCD of much larger structures. The current
describe briefly the procedure used in our laboratory to cal- heavy atom limit for the calculation of individual molecules
culate VCD intensities, although other commercial software will no doubt increase steadily in future years as the com-
is available to carry out some of the steps. First, a confor- putational capacity and speed of computers increases.
mation of the molecule is selected and a three-dimensional As an example of a measured and calculated VCD
structure is generated with standard bond lengths and angles spectrum, we present in Figure 5 the absorbance and VCD
with molecular modeling software such as Hyperchem.46 spectra of (C)-trans-pinane. The experimental spectrum
The structure is saved in Protein Data Bank (pdb) format was obtained from a single measurement on the Chiralir
and exported to the platform on which Gaussian 98 is run- FT-IR VCD spectrometer, and the calculated spectrum was
ning. The pdb file is converted to a Gaussian 98 input file carried out using DFT and the VCD intensity subroutine of
with a utility (newzmat) provided in the Gaussian pack- Gaussian 98 as described above. It is clear by comparing
age, and the desired method and basis set are selected. The these two sets of spectra that the DFT-calculated absorbance
Gaussian 98 input file command lines specify saving the and VCD spectra agree quite closely with the corresponding
Gaussian checkpoint file, and designate that a geometry measured spectra. This can only be achieved if the correct
optimization and a frequency calculation with VCD inten- equilibrium structure and normal vibrational mode motions
sities is to be carried out. One can also specify isotopic are calculated. If one were interested in determining the
masses, and the geometry optimization can be carried out absolute configuration of this molecule, the results of this
in the presence of a solvent field. The optimized geome- comparison leave no doubt as to the absolute configuration.
try, normal modes and IR spectrum are converted into a As a second example of a comparison of measured and
form readable by Hyperchem with a Gaussian utility (fre- calculated IR and VCD spectra, we present in Figure 6 the
qchk) which requires the Gaussian 98 checkpoint file. We absorbance and VCD spectra of ()-R-mirtazepine.48 The
extract the calculated frequencies, dipole strengths and rota- experimental spectra were measured using the Chiralir, and
tional strengths from the Gaussian 98 output file into a
program written in Mathcad to obtain a calculated spec- Calculated
trum from equations (8) and (11) and the expression for 20
Lorentzian bands
10
103
g2
en D e0 33
n n0 2 C g2
0
4.3 Examples of calculated VCD spectra 10 H3C Observed

H3C
In this section, we provide an example of a comparison of
Calculated
experimentally measured to theoretically calculated VCD
Molar absorptivity,
30
spectra. To date, hundreds of examples of such comparisons
have been published, and it is now established that VCD
can provide unambiguous determinations of the solution 20
Observed
state structure of chiral molecules when accurate measure-
ments and calculations can be carried out. Virtually all 10
chiral molecules for which a solution state IR spectrum of
sufficient absorbance intensity can be obtained exhibit mea- 0
surable VCD spectra. When a molecule possesses multiple
1400 1300 1200 1100 1000 900 800
competing solution-state conformations, the observed VCD
Wavenumber / cm1
is generally weaker than a molecule with a single predom-
inant conformation. The only cases where the VCD is too Figure 5. Comparison of measured to calculated IR and VCD
small to measure are those where there are multiple confor- spectra of (C)-trans-pinane under conditions of neat liquid and
sample pathlength of 55 m and a spectral resolution of 4 cm1 .
mations with nearly canceling VCD, but this is a relatively The VCD spectra were measured with the Chiralir and the
rare occurrence. On the theoretical side, there is currently calculations were performed using the Gaussian 98 program
an upper limit of approximately 70 heavy atoms37,39 for the package.
1.0
Observed
0.5
FT-IR P1 PEM1 X LB PEM2 P2 D
0.0 ID
REF 1
A 104
103
20
0.5
10 IAC1
0 1.0 IAC1IAC2 LIA1 REF 2

IAC2
10
20 A IDC SUB LIA2
30 Calculated
DIV
Figure 7. Block diagram illustrating the optical-electronic layout
N 1.0
for the measurement of FT-VCD using two PEMs to correct the
VCD baseline in real time across the entire spectrum with each
Molar absorptivity,
H N
600
N scan of the Chiralir spectrometer.
Absorbance
Observed 0.5
mixture is not available. In this case the solvent VCD
400 measurement must be used, and in order for the solvent
VCD baseline to be the same as the racemic mixture, the
0.0
200
deviation of the baseline from zero must be small.
Calculated
Recently, it has been found that adding a second PEM
to the VCD optical setup after the sample can be used to
0 reduce the VCD baseline to very small deviations from
1800 1600 1400 1200 1000 800
zero by adjusting the voltage of the second PEM.49,50
Wavenumber / cm1
In Figure 7, we illustrate the dual PEM optical-electronic
Figure 6. Comparison of the measurement and calculation of layout. The method consists of measuring simultaneously
()-R-mirtazepine. The sampling conditions were 75-m path- the VCD spectrum associated with each PEM (PEM1 and
length and 8 cm1 spectral resolution using the Chiralir and the
calculations were carried out using Gaussian 98. PEM2 ) with one of two LIAs (LIA1 and LIA2 ), each
referenced to one of the PEMs. The signal from PEM1
the calculated spectra were obtained using Gaussian 98. contains the VCD spectrum plus the birefringent artifact
Again, it is clear that the calculated absorbance and VCD signal. The signal from PEM2 contains a nearly negligible
spectra agree quite closely with the corresponding measured VCD spectrum and nearly the same birefringence artifact
spectra. The results of this figure represent an independent, signal as measured by the first PEM. The two PEM signals
nonambiguous determination of the absolute configuration are then subtracted (SUB) in real time as the VCD spectrum
of ()-mirtazepine to the R-enantiomer without reference is being measured, and this results in a VCD spectrum that
to any previous spectral result and without the use of X- is largely free of the artifacts. The VCD baseline can be
ray crystallography. The VCD band near 900 cm1 is an adjusted toward zero by changing the retardation level of
artifact due to CCl4 absorption. PEM2 . Further details regarding the dual PEM method can
be found in recent publications.49,50
In Figure 8 we show the VCD spectra of ()-(R)-
5 DUAL PEM VCD INSTRUMENTATION ibuprofen measured in three different ways. The VCD
spectrum that is furthest from zero was measured with a
In this section, we describe a new technical development single PEM and shows a large offset due to stray opti-
for VCD spectroscopy that dramatically reduces artifacts in cal birefringence. If a second PEM in the optical train
VCD spectra for any optical alignment of the instrument. is turned on and used to correct the baseline, the result
As noted above, due to small levels of stray birefringence yields the VCD spectrum with a baseline that is very close
present in the optics associated with the VCD spectrometer to zero. Notice that the small artifact bands in the VCD
after the first polarizer, it is necessary to establish a VCD spectrum due to water vapor between 1500 and 1700 cm1
baseline by carrying out a VCD measurement either with are essentially eliminated. Finally, we show the result of
the opposite enantiomer, the racemic mixture or the solvent. VCD spectral subtraction of the VCD of the (S)-enantiomer
In many cases, the opposite enantiomer or the racemic from that of the (R)-enantiomer, and then divide this result
applications using the NIR for industrial stereochemical

COOH
H3C applications involving the production of chiral molecules.
H Given the rapid pace of improvement of quantum cal-
culations over the past decades, it is likely that the power
of quantum chemistry to predict accurately the VCD spec-
tra of chiral molecules will improve in the future. Some of
this improvement will come from the increase in memory
size and speed of computers, and the rest from improve-
ments in quantum chemistry software. Calculations of the
VCD spectra of smaller chiral molecules should become
increasingly accurate as effects such as solvent influence
and anharmonicity are taken into account. In the not-too-
distant future, an accurate calculated VCD spectrum of a
small- to medium-sized molecule should be available dur-
ing the same period of time it takes to measure the VCD
spectrum experimentally, thus making the assignment of
absolute configuration completely routine. Concurrently,
Figure 8. VCD spectra obtained for ()-R-ibuprofen with sin- the size limits for current calculations of good accuracy
gle uncorrected PEM measurement, dual PEM measurement, and
should be extended in varieties of ways to include virtually
subtraction of two single PEM measurements of the opposite
enantiomers, showing the capability of dual PEM VCD to elim- all molecules for which VCD spectra can be measured.
inate the birefringent background and the absorption artefacts At some point, with the increasing size of molecules, the
present in the uncorrected single PEM measurement. vibrational spectrum becomes so congested that most indi-
vidual bands are not resolved. Nevertheless, it should still
by two. This is the most accurate measure of the VCD be possible to compare the overall VCD intensity pattern
spectrum, but nevertheless the dual PEM measurement is for calculated and observed spectra, even though the assign-
very close to the true VCD spectrum and represents an ment of bands to particular vibrational motions may be lost.
important improvement to the VCD measurement process. Currently, VCD applications are focused on the deter-
mination of absolute configuration, the determination of
6 FUTURE APPLICATIONS AND enantiomeric excess, and the determination of molecular
solution-state conformations, as discussed above. In the
DEVELOPMENTS future, wider classes of molecules will be investigated
VCD spectroscopy is a relatively new spectroscopic tech- for absolute configuration and conformational distribution.
nique that has required the most advanced levels of instru- This should aid in identifying unambiguously the abso-
mental and theoretical technology to bring to light useful lute stereochemistry of molecules currently too complex
applications. Nevertheless, VCD is a very simple, funda- for VCD analysis, and similarly for solution-state confor-
mental spectroscopic measurement that is extremely sen- mations. Applications involving the real-time determination
sitive to the absolute stereochemical structure of chiral of enantiomeric excess should improve process monitoring
molecules in all phases, including gases, liquids, solutions, of the production of chiral molecules, including chiral phar-
solids, films and powders. VCD spectra can be obtained in maceutical molecules, where chiral chromatography is too
principle whenever an IR spectrum can be obtained with time-consuming or impractical.
absorbance values between 0.1 and 1.0. Future improve- The future of VCD looks bright now that reliable exper-
ments can be envisioned in three major areas: instrumenta- imental and theoretical spectra are available for molecules
tion, calculations and applications. of medium size. Since VCD is a combination of two very
For instrumentation, the goals will be to increase fur- fundamental, structural probes of molecules, namely IR
ther signal-to-noise ratios, improve the quality of VCD absorption and circular dichroism, it should continue to find
baselines, and extend measurement methods to include a increasingly useful and valuable applications in the future.
wider range of IR sampling techniques including ATR.
It will also be of interest to extend VCD measurements
into the NIR where overtone and combination bands occur. ABBREVIATIONS AND ACRONYMS
In recent years, the NIR region has become increasingly
important for industrial applications of quality control and AAT Atomic Axial Tensor
process monitoring. It is likely that VCD will find many APT Atomic Polar Tensor
CPHF Coupled-perturbed-HartreeFock 12. L.A. Nafie, Raman Optical Activity: Theory, in Encyclo-
DCPI In-phase Dual Circular Polarization pedia of Spectroscopy and Spectrometry, eds T. Lindon,
G. Tranter and J. Holmes, Academic Press, London,
DCPII Out-of-phase Dual Circular Polarization
19761985 (1999).
ICP Incident Circular Polarization
LCP Left Circularly Polarized 13. L.D. Barron, M.P. Bogaard and A.D. Buckingham, J. Am.
Chem. Soc., 95, 603 (1973).
LIA Lock-in Amplifier
MFP Magnetic Field Perturbation 14. W. Hug, S. Kint, G.F. Bailey and J.R. Scherer, J. Am. Chem.
Soc., 97, 5589 (1975).
pdb Protein Data Bank
RCP Right Circularly Polarized 15. G. Holzwarth, E.C. Hsu, H.S. Mosher, T.R. Faulkner and
A. Moscowitz, J. Am. Chem. Soc., 96, 251 (1974).
ROA Raman Optical Activity
SCP Scattered Circular Polarization 16. L.A. Nafie, J.C. Cheng and P.J. Stephens, J. Am. Chem. Soc.,
97, 3842 (1975).
VCT Vibronic Coupling Theory
VOA Vibrational Optical Activity 17. S.D. Allen, A.I. Braunstein, J.C. Cheng and L.A. Nafie,
A 10.6 Micron Modulated Light Ellipsometer, in Optical
Properties of Highly Transparent Solids, eds S.S. Mitra and
B. Bendow, Plenum, New York, 503514 (1975).
REFERENCES 18. L.A. Nafie and M. Diem, Appl. Spectrosc., 33, 130 (1979).
19. L.A. Nafie, M. Diem and D.W. Vidrine, J. Am. Chem. Soc.,
1. L.A. Nafie and T.B. Freedman, Enantiomer, 3, 283 (1998). 101, 496 (1979).
2. L.A. Nafie, Vibrational Circular Dichroism: Spectrometer, 20. E.D. Lipp, C.G. Zimba and L.A. Nafie, Chem. Phys. Lett., 90,
in Encyclopedia of Spectroscopy and Spectrometry, eds 1 (1982).
J.C. Lindon, G.E. Tranter and J.L. Holmes, Academic Press,
London, 23912402 (1999). 21. L.A. Nafie, F. Long, T.B. Freedman, H. Buijs, A. Rilling,
J.-R. Roy and R.K. Dukor, The Determination of Enan-
3. R.K. Dukor and L.A. Nafie, Vibrational Optical Activity tiomeric Purity and Absolute Configuration by Vibrational
of Pharmaceuticals and Biomolecules, in Encyclopedia of Circular Dichroism Spectroscopy, in Fourier Transform
Analytical Chemistry: Instrumentation and Applications, ed. Spectroscopy: 11th International Conference, ed. J.A. de
R.A. Meyers, John Wiley & Sons, Chichester, 662676 Haseth, American Institute of Physics, Woodbury, NY,
(2000). 432434, Vol. 430 (1998).
4. T.A. Keiderling, Peptide and Protein Conformational Stud- 22. W. Hug, Instrumental and Theoretical Advances in Raman
ies with Vibrational Circular Dichroism and Related Spec- Optical Activity, in Raman Spectroscopy, ed. J. Lascombe,
troscopies, in Circular Dichroism: Principles and Appli- 312 (1982).
cations, Second Edition, eds N. Berova, K. Nakanishi and
R.W. Woody, Wilely-VCH, New York, 621666 (2000). 23. H. Surbeck, W. Hug and M. Gremaud, Opt. Commun., 38(1),
57 (1981).
5. L.A. Nafie, Ann. Rev. Phys. Chem., 48, 357 (1997).
24. L. Hecht, L.D. Barron and W. Hug, Chem. Phys. Lett., 158,
6. L.A. Nafie and T.B. Freedman, Theory of Vibrational Opti- 341 (1989).
cal Activity, in Circular Dichroism: Principles and Appli-
cations, Second Edition, eds K. Nakanishi, N. Berova and 25. L.D. Barron, L. Hecht, W. Hug and M.J. MacIntosh, J. Am.
R. Woody, Wiley-VCH, New York, 97131 (2000). Chem. Soc., 111, 8731 (1989).
7. L.A. Nafie and T.B. Freedman, Biological and Pharmaceuti- 26. K.M. Spencer, T.B. Freedman and L.A. Nafie, Chem. Phys.
cal Applications of Vibrational Optical Activity, in Infrared Lett., 149, 367 (1988).
and Raman Spectroscopy of Biological Materials, eds H.- 27. L.A. Nafie and T.B. Freedman, Chem. Phys. Lett., 154, 260
U. Gremlich and B. Yan, Marcel Dekker, Inc., New York, (1989).
1554 (2001).
28. D. Che, L. Hecht and L.A. Nafie, Comparison of Incident
8. L.D. Barron, L. Hecht, E.W. Blanch and A.F. Bell, Prog. and Scattered Circular Polarization Raman Optical Activ-
Biophys. Mol. Biol., 73, 1 (2000). ity, in Twelfth International Conference on Raman Spec-
9. L.D. Barron and L. Hecht, Vibrational Raman Optical troscopy, eds J.R. Durig and J.F. Sullivan, John Wiley &
Activity: from Fundamentals to Biological Applications, in Sons, New York, 846847 (1990).
Circular Dichroism: Principles and Applications, Second 29. D. Che, L. Hecht and L.A. Nafie, Chem. Phys. Lett., 180, 182
Edition, eds N. Berova, K. Nakanishi and R.W. Woody, (1991).
Wiley-VCH, New York, 667701 (2000).
30. M. Vargek, T.B. Freedman and L.A. Nafie, J. Raman Spec-
10. W. Hug, Raman Optical Activity: Spectrometers, in trosc., 28, 627 (1997).
Encyclopedia of Spectroscopy and Spectrometry, eds
J. Lindon, G. Tranter and J. Holmes, Academic Press, 31. L.A. Nafie and T.B. Freedman, J. Chem. Phys., 78, 7108
London, 19661976 (1999). (1983).
11. W. Hug and G. Hangartner, J. Raman Spectrosc., 30, 841 32. L.A. Nafie, J. Chem. Phys., 79, 4950 (1983).
(1999). 33. P.J. Stephens, J. Phys. Chem., 89, 748 (1985).
34. A.D. Buckingham, P.W. Fowler and P.A. Galwas, Chem. 40. J. Kubelka, P. Pancoska and T.A. Keiderling, Appl. Spec-
Phys., 112, 1 (1987). trosc., 53, 655 (1999).
35. P.K. Bose, L.D. Barron and P.L. Polavarapu, Chem. Phys. 41. J. Kubelka, P. Pancoska and T.A. Keiderling, Appl. Spec-
Lett., 155(4,5), 423 (1989). trosc., 53, 666 (1999).
36. P.L. Polavarapu, J. Phys. Chem., 94, 8106 (1990). 42. P. Pancoska, V. Janota and T.A. Keiderling, Anal. Biochem.,
37. P.J. Stephens and F.J. Devlin, Chirality, 12, 172 (2000). 267, 72 (1999).
38. M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, 43. B.I. Baello, P. Pancoska and T.A. Keiderling, Anal. Bio-
M.A. Robb, J.R. Cheeseman, V.G. Zakrzewski, J.A. Mont- chem., 280, 46 (2000).
gomery, Jr, R.E. Stratmann, J.C. Burant, S. Dapprich, J.M. 44. L.A. Nafie, Polarization Modulation FTIR Spectroscopy,
Millam, A.D. Daniels, K.N. Kudin, M.C. Strain, O. Farkas, in Advances in Applied FTIR Spectroscopy, ed.
J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, M.W. Mackenzie, John Wiley & Sons, New York, 67104
C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G.A. Peters- (1988).
son, P.Y. Ayala, Q. Cui, K. Morokuma, D.K. Malick, A.D. 45. X. Cao, T.B. Freedman and L.A. Nafie, Unpublished results
Rabuck, K. Raghavachari, J.B. Foresman, J. Cioslowski, J.V. (2000).
Ortiz, B.B. Stefanov, G. Liu, A. Liashenko, P. Piskorz,
I. Komaromi, R. Gomperts, R.L. Martin, D.J. Fox, T. Keith, 46. Hyperchem, Hypercube, Inc., Gainsville, FL.
M.A. Al-Laham, C.Y. Peng, A. Nanayakkara, C. Gonzalez, 47. J. Kubelka, P. Bour, R.A.G.D. Silva, S.M. Decatur and T.A.
M. Challacombe, P.M.W. Gill, B. Johnson, W. Chen, M.W. Keiderling, ACS Symp. Ser. (in press).
Wong, J.L. Andres, C. Gonzalez, M. Head-Gordon, E.S.
48. T.B. Freedman, R.K. Dukor, E. Kellenbach and L.A. Nafie,
Replogle and J.A. Pople, Gaussian 98, A.5 edition, Gaussian,
to be submitted.
Inc., Pittsburgh, PA (1998).
49. L.A. Nafie, Appl. Spectrosc., 54, 1643 (2000).
39. A. Aamouche, F.J. Devlin and P.J. Stephens, J. Am. Chem.
Soc., 122, 7358 (2000). 50. L.A. Nafie and R.K. Dukor, ACS Symp. Ser. (in press).
Raman Optical Activity Spectroscopy
Werner Hug
University of Fribourg, Fribourg, Switzerland
1 INTRODUCTION chiral molecules in his theoretical treatment of light scatter-

ing by the Stokes formalism.7 Gunthard, at the suggestion
A substance is optically active if it shows a difference in of the stereochemist Prelog, tried unsuccessfully in 1954 to
its interaction with right and left circular light. The rota- measure ORD of molecular origin in the infrared.8,9 Other
tion of the plane of polarization of linearly polarized light efforts centered on discovering manifestations of vibrational
by crystalline quartz as described by Arago,1 and by cer- optical activity in crystals, but in view of the more limited
tain organic liquids like turpentine as discovered by Biot,2 practical interest of such observations none of the results
were historically the first observations of such an interaction seems to have been confirmed independently.
difference. The rotation is due to a differing index of refrac- As it happened, the observation of vibrational optical
tion for right and left circular polarized light equal amounts activity of molecular origin was first accomplished in 1973
of which compose linearly polarized light.3 It took eight by Raman optical activity (ROA) scattering by Barron
decades following the discovery of optical activity for the et al.10,11 (with some confusion12,13 about the signs reported
existence of circular dichroism (CD) of molecular origin, in Barron et al.11 ), and about a year later in the infrared
the absorption difference for right and left circular light, through the measurment of vibrational circular dichro-
to be eventually demonstrated by Cotton.4 The observation ism (VCD) by Holzwarth et al.14 These observations were
of both phenomena, optical rotation and its dispersion as limited to individual vibrational bands. They were soon
a function of wavelength (optical rotary dispersion, ORD), complemented by higher-quality VCD data over a more
extended spectral region in the infrared by Nafie et al.,15
and CD, remained at first confined to visible light. Follow-
and by the recording of entire ROA spectra by Hug et al.16
ing the development of polarizing optics in the ultraviolet,
For ROA, the decisive technological advance that made its
and of appropriate light sources and detectors, it was grad-
measurement possible and relatively straightforward was
ually extended towards shorter wavelengths and eventually
the invention of the argon ion laser.
into the vacuum ultraviolet region. With the exception of the
The observation of vibrational optical activity by ROA
analysis of the vibronic structure of electronic absorption
was preceded by a number of theoretical investigations on
bands, the scope of application of optical activity remained,
the importance of magnetic dipole and electric quadrupole
however, limited to the probing of electronic chirality until
contributions in polarized light scattering. Some of this
the last quarter of the 20th century.
work involved cross terms of the magnetic dipole and
The interest of the extension of optical activity to include
the electric quadrupole polarizability tensor with the elec-
molecular vibrational chirality was evident long before mea-
tric dipole polarizability tensor. It was also known that
surements were technically feasible. Kastler in 1930 tried
appropriate combinations of such tensor elements, reflecting
unsuccessfully to measure optically active Raman scattering
molecular chirality, survived isotropic averaging.17,18 It was
of chiral molecules,5 with Bhagavantam and Venkateswaran
only in a 1971 paper by Barron and Buckingham,19 how-
reporting spurious results on the subject at about the same ever, that the understanding was reached that such inter-
time.6 Perrin in 1942 included optically active scattering by ference terms could yield scattering differences for right
and left circular light of sufficient size to make Rayleigh
John Wiley & Sons Ltd, 2002. and Raman optical activity observable effects. It was this
insight that led to the renewed quest for actually measuring has also been true for ROA, though over the years unique
optically active light scattering from isotropic samples of solutions have been found for alleviating some problems.
small chiral molecules. The difficulty in measuring molecular optical activity is
Considerable conceptual and methodical progress has inherent, though, irrespective of how it is done. It stems
been made since on the theory and on the measurement from the fact that optical activity represents the diastereo-
of ROA. The most important developments have certainly meric interaction difference of right and left circular light
been the invention of dual circular polarization ROA by with chiral molecules, and thus a small difference of two in
Nafie et al.20,21 and the revision22,23 of the theoretical foun- general much larger quantities. The ratio of the difference
dations of ROA24 to include the general resonance case, to the average interaction for visible light is typically one
with an explicit formulation of the single electronic state in a thousand or less. It corresponds to the ratio of the
limit.25 There have also been various theoretical formu- interaction of a magnetic dipole moment to that of an
lations of nonlinear ROA26 28 but so far they have not electric dipole moment with the electromagnetic radiation
found practical application. On the computational side one field, and decreases therefore proportionally to the increase
can cite the gauge invariant direct calculation of ROA by of the wavelength of the probing light.
linear response theory29 as a major development. The com- This is an aspect that strongly favors ROA over VCD,
parison of the results of such calculations with measured though much of ROAs inherent advantage is lost on
spectra generally allows the determination of the absolute account of the weakness of spontaneous Raman scattering.
configuration of chiral molecules beyond any doubt. It also shows why it is, a priori, advisable to choose the
Instrumental advances in ROA have mostly followed
exciting wavelength in an ROA experiment at as short a
progress in general Raman instrumentation, though the
wavelength as is compatible with concurrent experimental
opposite happened with solid-state detectors when self-
constraints.
scanned diode arrays were first used in multichannel Raman
It is advantageous to base further discussion on an actual
spectroscopy.30,31 Multichannel detection with backthinned
example. We will use for this purpose the data of the
charge-coupled device (CCD) detectors and high-luminosity
ROA measurement of the CDO stretch vibration of 3-
spectrographs is the current standard in ROA spectroscopy.
methylcyclohexanone (Table 1), a typical chiral molecule
To the original right angle depolarized scattering arrange-
that has generated considerable interest in the field of
ment, polarized right angle scattering,30 backscattering,32,33
and forward scattering34 have been added. optical activity. The stated experimental parameters are rea-
In the subsequent sections some of the major aspects of sonable practical limits on the throughput of the dispersive
ROA scattering will be discussed. It is not feasible to give system, and on the laser power that can be focused into
a detailed account here of all important facets of this varied a sample without causing thermal lensing or damage. The
subject, or of the fairly involved underlying theory. CDO stretch vibration is one of several medium to strong
bands. The result of such a measurement is an ROA spec-
trum of high quality.
2 GENERAL ASPECTS OF RAMAN
OPTICAL ACTIVITY Table 1. Typical parameters for the recording of a polarized
ROA spectrum of an organic molecule, here (R)-(C)-3-methyl-
MEASUREMENTS cyclohexanone. The quantum efficiency of the cooled (70 C)
and back-thinned CCD detector is approximately 75%. Note that
While in principle straightforward, the measurement of sample illumination is cut during read-out of the detector.
vibrational optical activity is subject to a number of practi-
cal limitations. Some of them are common to ROA as well Exciting wavelength 532 nm
as VCD, while others are specific to one of the two methods. Laser power (at sample) 500 mW
The information provided by vibrational optical activity Sample size (pure liquid) 35 L
Speed of light collection optics f/1.1
is, however, unique and cannot in general be obtained by Illumination time before read-out 0.94 s
other means. To the extent that limitations are inherent and Total illumination time 13 min
not technological, stereochemists and spectroscopists will Read-out overhead 3.8 min
therefore have to learn to live with them. Spectral range 01900 cm1
Optical resolution 7 cm1
Spectral width per CCD column 2.4 cm1
Total recorded photons at peak height of band 1010 e
2.1 Signal strengths and signal-to-noise of CDO stretch vibration
Total root mean square read-out noise per CCD 2300 e
Signal-to-noise problems and spurious results have been column
a companion to optical activity since its inception. This
Raman Optical Activity Spectroscopy 3
The instrument used for the example of Table 1 is Spontaneous Raman scattering has been termed a spec-
capable of shot noise limited performance.35 If NR and troscopic method with a low dynamic range, and this
NL are the number of right circular and left circular statement is equally pertinent to ROA. The limitations that
scattered photons, respectively, and with NR NL , root this implies become clear if one considers the recording of
mean
p square shot noise for NR NL and NR C NL will be gas phase and solution spectra. With the single pass back-
NR C NL D 105 photons. Thus, in view of the fact that scattering instrument used as the reference for Table 1 with
the full width of the band due to the CDO stretch vibration a gaseous sample at 1 bar at room temperature, with a laser
at half its height extends over some 25 cm1 , or approxi- power of 1 W at the sample, and for a similarly good scat-
mately 10 elements of resolution of the detector, one will terer as 3-methylcyclohexanone, approximately a full day
be able to identify beyond doubt ROA effects with a ratio of measurement time will be needed for a spectrum with
D NR NL /NR C NL down to the 105 level. This an acceptable signal-to-noise ratio and a modest 7 cm1
is the size of the effect that one actually observes for the resolution as used for the liquid phase.
CDO stretch vibration. The gas phase case demonstrates the influence of dilution
It is illustrative to consider the laser power one would in an experiment that relies on a fixed scattering volume. In
need to measure a similar ROA spectrum with the same solution spectra there is the added problem of solvent bands.
signal-to-noise ratio in the same illumination time of 13 min Their obvious effect is to obscure spectral regions where
with exciting light at other wavelengths. The quantum the solute might have bands of interest. The recording of
efficiency of the detector, and/or the throughput of the solution ROA spectra is therefore limited to solvents with
spectral analyzing system, will in general be different, and small molecules and few vibrational bands.
at present lower at infrared and ultraviolet wavelengths A less obvious problem in a multichannel experiment
than in the visible, but such differences are subject to is saturation of the detector at specific wavelengths by
change and they are disregarded here. The Raman scattered intense solvent bands. This condition can lead to offsets
energy for constant exciting energy depends on the wave- throughout an ROA spectrum and requires a reduction of
length of the scattered light as the fourth inverse power. sample illumination time before detector read-out. The sol-
For a detector yielding a signal proportional to the num- vent best suited to ROA is water, while good scatterers
ber of detected photons, the exciting energy required for with strong polarized bands, like CS2 or CCl4 , are poor
a constant signal therefore varies as the third power of choices in polarized ROA spectroscopy. For a 5% solu-
the wavelength. Thus, we will need 4 W at 1.06 m and tion of 3-methylcyclohexanone in CCl4 for example, time
62 mW at 266 nm, without considering possible resonance before detector read-out will have to be reduced six-fold as
enhancement, for a Raman signal of the same size. In addi- compared with the example of Table 1, due to the strong
tion, in ROA we will have to take the variation of the 460 cm1 band of CCl4 , while total sample illumination
-value with wavelength into consideration. For a band time will have to be increased by 20-fold. The 120-fold
with D 1 105 at 532 nm one expects D 5 106 increase in detector read-out cycles will push recording
at 1.06 m and D 2 105 at 266 nm. Therefore, 16 W time from 17 min to roughly 12 h, and read-out noise to
would be required in the infrared and 16 mW in the ultra- 25 000 e , with current technology. The example shows
violet, which yields the quite impressive ratio of 1000. Of the interest of developing depolarization backscattering to
course, laser power is not the only consideration in ROA. the routine level.
Offset problems, sample decomposition, fluorescence and Fluorescence is the scourge in both ROA and ordi-
so on are others, and they do not all favor the ultraviolet nary Raman spectroscopy. Absence of dust is an absolute
over the visible or infrared. necessity in slow modulation ROA instruments, but the
simultaneous detection of right and left circular scattered
light has drastically improved this aspect.35 Small amounts
2.2 Sample considerations of absorption of the exciting laser beam by the sample cause
thermal lensing, a concomitant reduction in signal strength,
In practice, the recording of ROA of molecules requires and possible offsets.
isotropic samples of good optical quality. While there is
no inherent impossibility to measure molecular ROA of
anisotropic probes, no such experiments have so far been 3 THE THREE CIRCULAR
reported. One reason is a lack of enthusiasm to tackle POLARIZATION VARIANTS
the particularly severe offset problems looming in such
measurements. All published ROA data on molecules to There are three different circular polarization schemes
date have been recorded for liquid samples only. by which spontaneous ROA scattering can be measured,
namely incident circular polarization (ICP),10,11,19 scattered theoretical expressions for an ensemble of molecules with
circular polarization (SCP),22,36 and dual circular polar- a statistical distribution of their spatial orientation are, per
ization (DCP).20,21 In ICP, the original ROA variant, the molecule, for exciting light that is naturally polarized (n), or
incident light is modulated between right and left circular polarized perpendicular (?) or parallel (jj) to the scattering
but the scattered light is not analyzed with respect to its plane
circular content. In SCP the circular content of the scat-
?
tered light is analyzed while the incident light is chosen dsqSCP D ? dsL q C ? dsR q
linearly or naturally polarized. In DCP, the circular content D 2K45a2 C 7b2 d
1a
of the scattered light is analyzed while the incident light
jj
is synchronously circular polarization modulated. In DCPI dsqSCP D jj dsL q C jj dsR q
the circularity of the incident and of the detected light are D 2K[6b2 C cos2 q45a2 C b2 ] d
1b
of the same kind, while in DCPII they are opposite. Far n 1 ? jj
from the resonance limit DCPII vanishes and DCPI yields dsqSCP D 2 [ dsqSCP C dsqSCP ]
the same information as ICP and SCP, but with a higher D K[45a2 C 13b2
ratio of ROA to Raman intensity.
Except where explicitly stated, further discussion and all C cos2 q45a2 C b2 ] d
1c
given formulae correspond to the far from resonance limit ? ?
dsqSCP D dsR q dsL q ?
of spontaneous ROA scattering. The detailed consideration 4K

of other situations37 complicates the discussion of ROA D [45aG0 C 7b2G C b2A
c
out of proportion to their currently negligible practical
importance, but the reader should be aware that resonance, C cos q45aG0 5b2G 3b2A ] d
2a
and nonlinear ROA, might become important spectroscopic 4K
jj dsqSCP D jj dsR q jj dsL q D [6b2G 2b2A
tools over the coming decades. c
C cos q45aG0 5b2G 3b2A
3.1 Scattered circular polarization Raman C cos2 q45aG0 C b2G C 3b2A ] d
2b
optical activity n dsqSCP D 12 [? dsqSCP C jj dsqSCP ]
2K
The measurement of SCP ROA can be considered the basic D [45aG0 C 13b2G b2A
ROA experiment. No modulation is required between right c
and left circular light. SCP ROA can be measured in a C cos q90aG0 10b2G 6b2A
static way by passing the scattered light through a circular
C cos2 q45aG0 C b2G C 3b2A ] d
2c
polarization analyzer and by simultaneously, but separately,
recording the intensity of the right and of the left circular The integral scattering cross-sections for naturally polarized
light scattered by the probe.35 The sum is called the circular incident light are
intensity sum (CIS) and the difference the circular intensity
p 2p
difference (CID). The measurement is analogous to the n n 4pK
measurement of the degree of circularity in ordinary Raman sSCP D dsqSCP D 180a2 C 40b2
0 0 3
spectroscopy except that the net circularity of the exciting 3a
light is kept as small as possible, in order for the CID 8pK
to reflect the chirality of the molecule and not that of the n sSCP D 180aG0 C 40b2G 3b
3c
exciting light.
The molecular quantities underlying the CIS and CID where integration over j corresponds to integration over all
scattering processes are, for light scattered into a solid angle azimuthal orientations of the scattering plane.
of 4p, the circular sum and circular difference scattering These expressions are valid for a temperature T D 0 K.
cross-sections sSCP and sSCP . In an actual measurement For T 6D 0 K they have to be multiplied by the Boltz-
one generally observes only the light scattered into a cone man factor 1/[1 exp100hcnQ p /kT] for Stokes and
the axis of which forms an angle q with the direction 1/[exp100hcnQ p /kT 1] for anti-Stokes scattering.38
of propagation of the exciting light. The molecular quan- If s is defined as the rate at which photons are removed
tities appropriate to this situation are dsSCP q/d
and from the incident beam by scattering, relative to the rate at
dsSCP q/d
, the circular sum and difference differen- which they cross a unit area perpendicular to the direction
tial scattering cross-sections per unit of solid angle
. The of propagation of the incident beam, then for s in units of
m2 the constant K is given by KDP onoff

switch
1 m0 2 3 107 2 2 4 3
K D Kp D !p !0 D p m0 c nQ p nQ 0 4a Beam-
90 4p 9 steering
!p !0 Frequency-doubled + optics
nQ p D D nQ p 4b Nd : YVO4 laser
200pc 200pc Polarization
conditioner
where !0 D 2pn0 is the frequency of the incident and !p
the frequency of the scattered light, nQ p the Raman fre-
quency shift (in cm1 ), c the speed of light (in m s1 ),
and m0 the permeability of the vacuum. a2 is the isotropic Sample cell
and b2 the anisotropic invariant of the molecular elec-
tric dipoledipole transition tensor. aG0 and b2G are the
isotropic and anisotropic invariants due to cross products of Liquid Notch filter
crystal retarder
this tensor with the electric dipolemagnetic dipole tran-
sition tensor, and b2A the anisotropic invariant due to its f-ratio conserving
fiber optics
cross products with the electric dipoleelectric quadrupole
transition tensor. In terms of the complete set of the 13
Raman and ROA invariants defined in Hecht and Nafie23
one has aG0 D aG, b2G D bs G Q 2 , and b2A D bs A
Q 2 . Except
Modified
2 2
for gA D 3bA the invariants used here are identical to the spectrograph
invariants aG0 , g2G and g2A as introduced to ROA by Long.38 CCD detector
It should be noticed that quadrupole contributions are absent
Chopper wheel
in the invariant combination 9aG0 C 2b2G for scattering into
Figure 1. Simplified optical layout of the SCP back-scattering
4p, and that this same invariant combination can also be instrument used as reference for Table 1. Throughput is almost an
obtained by eliminating quadrupole contributions through order of magnitude higher than in conventional ROA backscat-
a combination of polarized and depolarized right angle tering instruments. The right and left circular scattered light is
scattering.30 These results are general and equally valid in simultaneously registered on the two halves of the CCD detector
the resonance case, provided the complete set of Raman chip and digitized in the same read-out cycle. Sample illumina-
tion is switched off by the potassium dideuteriophosphate (KD P)
and ROA invariants is substituted for the set used in equa-
electro-optic onoff switch during read-out. The chopper wheel
tions (1a) to (3b). is only used in the DCP modification of the instrument where
The sign convention in CID scattering is to represent alternating exposure of the two CCD halves is synchronized
measured data as right minus left. This ROA convention is to the circular polarization modulation of the incident light. A
also implied by the way the above formulae are written. The more detailed description of the optics can be found in Hug and
general sign convention for molecular quantities in optical Hangartner.35
activity, on the other hand, is left minus right. Molecular
quantities, like n s D n sL n sR in the present context, The fiber optics of the instrument serves the function of a
or the difference of extinction coefficients e D eL eR of cross-section transformer. This not only helps in maximiz-
importance in resonance ROA, therefore carry a minus sign ing throughput, but also allows the generation of a straight
if used within the ROA convention. slit image on the backthinned slow-scan CCD detector by
Figure 1 shows a schematic of the SCP back-scattering curving the slit-shaped output section of the fiber optics
instrument used as reference in Table 1. The first instru- that forms the input slit of the holographic transmission
ment of its kind,39 it exploits the SCP advantage of the grating spectrograph. The close-to-circular input face leads
simultaneous availability of right and left circular scattered to the collection of an almost regular cone of scattered light,
light. This drastically reduces the deleterious effect of dust which helps in reducing instrumental offsets.32,40
in the sample on the quality of measured ROA data and Sample size is an important consideration in any analyt-
permits the use of pulsed laser sources. Small differences ical method. In contrast to right-angle scattering, where the
in the two circular detection channels are compensated by measurement of high-quality ROA spectra is possible with
interconverting their function through the switching of the sub-microliter volumes,32 the possibility to reduce sam-
liquid crystal retarder from the l/4 to the Cl/4 retarda- ple size in collinear spontaneous SCP ROA scattering is
tion state. The channel of the circular polarization analyzer restricted by the requirement that the collected scattered
that transmitted right circular light will, upon switching, light has to be able to leave the sample volume without
transmit left circular light, and vice versa. being reflected by the walls of the sample cell. With a
specially designed sample cell a volume of 35 L can be SC

achieved, without a reduction in light collection efficiency.
It is worthwhile mentioning that, in view of the reversal L1
of the sense of circularity upon reflection, SCP is the only L1
ROA variant that allows for a multipass arrangement. M1
M2
3.2 Dual circular polarization Raman optical l1
activity M2
S
L2 L3
The differential scattering cross-sections for DCP are
M1
L R 2 2 2 2
dsL qC dsR q D K[45a C13b Ccos q90a 10b
Figure 2. Dual lens light collection system for right angle scatter-
C cos2 q45a2 C b2 ] d
DCPI 5a ing. This optical arrangement provides substantial offset reduction
and is capable of filling the large etendue of modern holographic
L
dsR qC R dsL q D K[45a2 C13b2 cos q90a2 10b2 transmission grating spectrographs without the use of fibre optics.
The two light collection lenses L1 and L01 form a joint intermedi-
C cos2 q45a2 C b2 ] d
DCPII 5b ate image I1 , which is projected by lens L3 onto the entrance slit
S of the spectrograph. L2 , field lens; SC, scattering cell; M1 , M01 ,
dsqDCPI D 2n dsqSCP 6a 22.5 mirrors; M2 , M02 , 45 mirrors.
dsqDCPII D 0 6b
This yields the differential and integral scattering cross-
Upon integration over 4p the terms linear in cos q in sections:
equations (5a) and (5b) average to zero, and the same L,R
integral circular sum scattering cross-sections result for ds?,jj q D ?,jj dsL,R q 8a
DCPI and DCPII : L,R L,R L,R
ds q D ds? q C dsjj q
L
sL C R sR D L sR C R sL D n sSCP 7a D 2n dsL,R q 8b
?,jj
sDCPI D 2n sSCP 7b ds?,jj qICP D ds qSCP 9a
From equations (7a) and (7b) it is seen that, while the inte- ds qICP D 2n ds qSCP 9b
L
grated scattered CIS is equally split between DCPI and s CR s D 2n sSCP 10a
DCPII , the CID, which carries the optical activity infor- n
sICP D 2 sSCP 10b
mation, is exclusively contained in DCPI in the far from
resonance limit. The development of routine DCPI ROA Right angle ICP scattering, the oldest ROA variant, is
scattering is therefore of utmost interest. The experimental by far the easiest to perform. A high degree of preci-
DCP backscattering arrangement used by Nafie is of elegant sion can be obtained with the dual lens light collection
simplicity41 but not well suited for routine work. The DCP system40 shown in Figure 2. ICP back-scattering32 is like-
modification of the SCP instrument indicated in Figure 1 wise a relatively easy ROA experiment to do. Except for
should be a step forward in this direction. the use of new detector and spectrograph technology, instru-
mentation has, however, stagnated. Yet, where the poten-
tially higher signal-to-noise ratio of DCPI back-scattering is
3.3 Incident circular polarization Raman optical not required, fast modulation ICP back-scattering,35 using
activity spinning disk or scanning mirror technology for switch-
ing between CCD halves, does have potential for routine
The expressions for ICP sum and difference scattering applications.
cross-sections without analyzer amount to twice those for
the polarization equivalent SCP case for naturally polar-
ized incident light. This does not mean, of course, that 4 THEORETICAL INTERPRETATION OF
scattering is twice as intense in ICP as compared with RAMAN OPTICAL ACTIVITY DATA
SCP. Rather, it expresses the fact that ICP, like DCP, is a
modulation experiment where scattering cross-sections are Early attempts at understanding ROA generation in mole-
defined for the exciting light as being fully right and left cules were based on simplifying models. They have an
circular for the two respective half-periods of modulation. esthetic appeal on account of their computational simplicity
Table 2. Comparison of Raman and ROA differential scattering cross-sections for collinear and
right angle scattering. Note that DCPI values refer to the sum and difference of cross-sections
measured for two half-periods of modulation.
Difference Sum
Scattering angle ( ) Cross-section
4K/c K
0 SCP, unpolarized 90aG0 C 2b2G 2b2A 90a2 C 14b2

DCPI 180aG0 C 4b2G 4b2A 180a2 C 4b2
180 SCP, unpolarized 12b2G C 4b2A 90a2 C 14b2
DCPI 24b2G C 8b2A 24b2
90 SCP, polarized 45aG0 C 7b2G C b2A 90a2 C 14b2
SCP, depolarized 6b2G 2b2A 12b2
SCP, unpolarized 45/2aG0 C 13/2b2G 1/2b2A 45a2 C 13b2
DCPI 45aG0 C 13b2G b2A 45a2 C 13b2
and the apparent immediate insight they provide. An 1

impressive success of the venerable two-group model of Tanis
1
: Tanis
2
D T1mn T2mn C T1mn T2nm
m ,n
2
optical activity, reworked for ROA,42,43 was qualitative
experimentally confirmed34 prediction of the relative size 1
T1mm T2nn 12b
of forward and backward scattering. The interpretation of 3
low-frequency ROA bands with the hindered single-bladed 1
propeller model,44 one of its offshoots, did not meet with Ta1 : Ta2 D T1mn T2mn T1mn T2nm 12c
2 m ,n
the same success.13 The ROA version24 of the bond polar-
izability theory fared better in its application to coupled with cross terms vanishing.
methyl-torsional modes,45 but in essence this remained a Substituting Raman and ROA transition tensors for a
lone success in the realm of small molecules. From this transition between an initial state jii and a final state hfj for
and similar observation one must conclude that, wherever T1 and T2 , and with amn D anm in the far from resonance
vibrational modes are not well understood, or wherever limit, yields the invariants of equations (1a) to (10b):
electron exchange matters, such models, including the atom
1 is 1
dipole interaction model,46 or the equivalent orbital polar- af2 i D a : aisf D amm f i ann f i 13a
izability and similar models,47,48 are prone to failure. The 3 f i i 9 m ,n
present discussion will therefore stress the direct calculation 3 anis 1
of ROA spectra. b2f i D af i : aanis
f i
D [3amn f i amn f i
2 2 m ,n
amm f i ann f i] 13b
4.1 Raman optical activity invariants 1 is 1
aG0f i D a : G0 f D amm f 0
i Gnn f i 13c
3 f i i 9 m ,n
A second-rank tensor T can be written as the sum of a
symmetric Ts and an antisymmetric Ta tensor, with 3 anis 1
b2Gf i D a : G0 f D [3amn f 0
i Gmn f i
the symmetric part further decomposable into an isotropic 2 f i i 2 m ,n
Tis and an anisotropic Tanis part. Contracting the direct
product of two tensors T1 and T2 twice yields amm f i G0nn f i ] 13d
!0 anis
b2Af i D a : Af i
T1 : T2 D T1mn T2mn 2 f i
m ,n !0
D amn f i Amn f i 13e
2 m ,n
D Tis1 : Tis2 C Tanis
1
: Tanis
2
C Ta1 : Ta2 11
where The fact that molecular invariants refer to a specific tran-

sition f i is explicitly indicated here, in contrast to
1 equations (1a) to (10b). amn f i and G0mn f i stand for
Tis1 : Tis2 D T1mm T2nn 12a
3 m ,n the elements of the electric dipoleelectric dipole and the
imaginary part of the electric dipolemagnetic

dipole tran- where i stands for the sum over Cartesian components, and
sition tensor, and Alkn f i in Amn f i D lk emlk Alkn f i derivatives of Te are to be taken at the nuclear equilibrium
is the electric dipoleelectric quadrupole transition ten- position. Each term
sor, with emlk representing the antisymmetric unit tensor of @Te
LeviCivita. Amn f i transforms like G0mn f i but is trace- Ta,p D a Laxi,p 18
less, which explains the absence of an isotropic invariant. i
@x i 0
One recognizes that contracting aa and aG0 twice yields the in equation (17) transforms as Tf i with respect to rota-
combination of isotropic and anisotropic invariants charac- tions and translations of the coordinate system.
teristic for the integral scattering cross-sections s and s: The electronic tensors are the real dynamic property
1 2 tensors50 given by
asf : asf D 9af i C 2b2f i 14a
i i 3 2 !jn
1 aemn D O m jjihjjmO n jni
Rehnjm 19a
asf : G0 f D 9aG0f i C 2b2Gf i 14b h j6Dn !2jn !20
i i 3
e 0 2 !0
This result extends to the resonance case if the appro- Gmn D O m jjihjjm
Imhnjm O n jni 19b
h j6Dn !2jn !20
priate complex transition tensors as defined in Hecht and
Nafie23 are substituted for asf i and G0 f i . The anisotropic 2 !jn
Aem,nl D O nl jni
O m jjihjj
Rehnjm 19c
invariants b2f i and b2Gf i on the right-hand side of equa- h j6Dn !2jn !20
tions (14a) and (14b) are then each replaced by the sum of
the appropriate anisotropic and antisymmetric invariants. where m O n, m O nl stand, respectively, for the elec-
O n , and
tric dipole, magnetic dipole, and the traceless electric
quadrupole operator.
4.2 Placzek polarizability theory expressions With an eye on using polarizability theory type expres-
sions in a subsequent section for visualizing ROA gener-
In the far from resonance limit a transition tensor Tf i ation in molecules, we can write invariant expressions of
can, by using the Placzek polarizability theory approach,49 the form of equation (13) for individual nuclei a D b or
be written as a molecular electronic tensor Te effect- pairs of nuclei a 6D b. For example, for aG0ab,p and b2Gab,p ,
ing transitions between the molecular vibrational states jii the isotropic and anisotropic invariants resulting from the
and hfj: contraction of aa,p and G0 b,p , one has:
e
Tf i hfjTe jii
@T
hfjQp jii 1 is 1
@Qp 0 aG0ab,p D aa,p : G0 b,p D amm a,p G0nn b,p
3 9 m ,n
e
@T h 0e
15 1 @aemm @Gnn
@Qp 0 400pcnQ p D Laxi,p Lbxj,p 20a
9 m,n i,j @xia 0 @xjb
0
with nQ in units of cm1 . In equation (15) it is additionally 3 1
assumed that f i is a fundamental transition, that the b2Gab,p D aaanis
,p
: G0 b,p D [3amn a,p G0mn b,p
2 2 m ,n
vibration is harmonic, and that it is described by the normal
coordinate Qp . If the transition starts from another than the amm a,p G0nn b,p ]
level np D 0, where np is the vibrational quantum number, 0e
1 @aemn @Gmn
then the right-hand side of equation (15) must be multiplied D 3 a
by np C 11/2 for an upward, and by n1/2 2 m,n i,j @xi 0 @xjb
p for a downward 0
transition. The normal coordinate Qp can be expressed in e 0e
@amm @Gnn
terms of the cartesian displacement vectors x a of the Laxi,p Lbxj,p 20b
a
nuclei a: @xi 0 @xjb
0
x D L x Q 16
with

and therefore aG0p D aG0ab,p 21a
e @Te @x a @Te
@T i
a,b
D a D a Laxi,p
@Qp 0 a,i
@x i 0 @Q p 0 a,i
@x i 0 b2Gp D b2Gab,p 21b
17 a,b
4.3 The direct calculation of Raman optical A typical ROA benchmark calculation for a small chiral
activity molecule like methyloxirane would use for the force field a
coupled cluster approach55 with singles, doubles and a per-
A direct ROA calculation requires two distinct steps: the turbational correction for connected triple excitations, with
determination of molecular vibrational modes, and the eval- a cc-pVTZ basis set,56 while the calculation of electronic
uation of the gradients of the electronic tensors appearing in tensors would be based on linear response theory and an
equations (17) to (20b). In a puristic vein one would want to aug-cc-pVDZ basis set.57
accomplish both tasks within the frame of the same ab initio
approach. This is not an efficient way, however, to tackle
4.4 Visualizing Raman optical activity
the problem of the computational interpretation of ROA
generation in molecules
spectra. The criteria for the calculation of precise normal
modes frequency and shape are different from those Finding general links between observed vibrational opti-
underlying the calculation of gradients of higher electronic cal activity and molecular structure is one of the obvious
tensors. goals of the computational interpretation of ROA spectra.
In their pioneering work on the ab initio calculation of The traditional methods of vibrational analysis are indis-
VCD Stephens et al.51 showed that the inclusion of elec- pensable tools to this end, but the inspection of the shape
tron correlation is required to render vibrational modes of vibrational modes, and of their potential energy distri-
sufficiently precise for the description of vibrational opti- bution matrices, gives no clues as to the size or sign of
cal activity. The omission of electron correlation has a far the ROA generated by nuclear motion. This information is
more degrading influence on calculated spectra than the contained, on the other hand, in the N N matrices formed
use of the harmonic approximation, including the neglect by the mono- and dinuclear terms exemplified by equa-
of overtones and Fermi resonances, at least in the region tions (20a) and (20b). The snag is that the large number
below 2000 cm1 . Density functional theory (DFT) with of elements several hundred for a typical medium-sized
the B3LYP functional has turned out to be the most cost- chiral molecule renders the direct inspection of these data
effective approach to include it.51 This has more recently unwieldy at best.
been found to be equally true for the calculation of The problem can be eased by the conversion of the
the vibrational part in ROA. The separate evaluation of numerical information into a pictorial representation of the
the vibrational part permits the use of basis sets opti- contribution to ROA, which is made by the motion of
mized with respect to the molecular Hessian matrix alone, individual nuclei in a vibrational mode. This amounts to
which is a big advantage if going beyond B3LYP DFT is the definition of atomic, or rather quasi-atomic, quantities
required. for ROA generation, and requires a strategy for allocat-
Three different ab initio approaches are currently used ing parts of dinuclear terms to individual contributing
to evaluate the derivatives of electronic ROA tensors: the atoms.
static finite field procedure developed by Amos52 and first The terms in equations (20a) and (20b) are of the form
used in ROA by Polavarapu,53 SCF linear response theory
yielding dynamic tensor values by Helgaker et al.,29 and a Jab,p D L xa,p Vab L xb,p 22
more approximate but faster scheme based on the sum-over-
where Jab,p stands for a mononuclear a D b or dinuclear
states (SOS) formalism by Bour.54 Analytical gradients are a 6D b term of any of the Raman or ROA invariants. For
only available in the SOS formalism. In the two other Jab,p D b2Gab,p , the elements of the dyadic Vab would be
approaches numerical differentiation is necessary through
given by
the calculation of electronic tensors at 6N displaced geome- 0e
tries, where N is the number of nuclei. The inclusion of 1 @aemn @Gmn
electron correlation is not required for qualitatively cor- Vai,bj D 3 a
2 m ,n @xi 0 @xjb
rect results, but basis sets must include diffuse functions. 0
e 0e
Gauge invariance29 can be achieved in variational calcu- @amm @Gnn
lations through the use of perturbation-dependent London 23
@xia 0 @xjb
0
orbitals. This is mainly of interest where basis sets are
marginal but not an issue with large ones. Conversely, the A condition for the weighting coefficients raab,p and
higher precision of dynamic tensor values provided by lin- rbab,p , required for splitting the terms Jab,p between
ear response theory, i.e. values evaluated at the frequency atoms, is rotational invariance. The best choice58 under
of the exciting light, is important only in calculations with this constraint is basing them on the norms of vectors of
large basis sets. the form L a Vab . One can write
jL xa,p Vab j C jL xa,p Vaa j

raab,p D 24a
jL xa,p Vab j C jL xa,p Vaa j C jVab L xb,p j C jVbb L xb,p j
jVba L xa,p j C jVaa L xa,p j
raba,p D 24b
jVba L xa,p j C jVaa L xa,p j C jL xb,p Vba j C jL xb,p Vbb j
with a quasi-atomic term on atom a given by of atom a in vibrational mode p to the integral sum and
difference scattering cross-section
J a p D Jab,p r aab,p C Jba,p r aba,p 25
b n hKp
s a D [180a2 ap C 40b2 ap ] 26a
300cnQ p
These expressions include nuclear excursion weighting and
weighting by the relative size of the tensor gradients on the 2hKp
n sa D [180aG0 ap C 40b2G ap ] 26b
two atoms. 300c2 nQ p
The subdivision of molecular invariants according to
The use of this visualization procedure is shown for
equation (25) provides the basis for breaking down molec-
those two of the six coupled scissoring modes of
ular scattering cross-sections into quasi-atomic terms. With
equations (3a), (3b) and (15) one has for the contribution
1.0 RC()
RC () 0.5
1.0
0.0
0.5
0.5
0.0
1.0
0.5
d()/d
1.0
n/ () n/ () 2.0
2.0 R L
1.5 38
[1017 2/sr]
1.0
1.0
10 7 electrons
0.0
0.5
0.0 1.0 39
0.5 2.0
1.0
1.5 nd()/d
2.0 n/ () + n/ ()
4.0
R L 5.0
1010 electrons
[1017 2/sr]
3.5 4.0
3.0
3.0
2.5
2.0 2.0
1.5 1.0
1.0
0.0
0.5 1800 1600 1400 1200 1000 800 600 400 200 0
0.0 Wavenumber /cm1
1800 1600 1400 1200 1000 800 600 400 200 0
Figure 4. Calculated SCP back-scattering spectrum of s-(M)-
Wavenumber /cm1 ()-[4]helicene. Normal coordinates: B3LYP DFT with 6311G
Figure 3. Experimental SCP back-scattering spectrum of neat basis set (GAUSSIAN9861 ). Electronic tensors: SCF linear res-
s-(M)-()-[4]helicene. Sample size 35 L, exciting wavelength ponse theory with 631CCG basis set (DALTON62 ). Full width
532 nm, average resolution 7 cm1 , laser power at sample 115 mW, at half-maximum height of the Lorentzians used to represent
sample illumination time 42 min, exciting energy 290 J. Indicated vibrational bands is 10 cm1 , with their peak height indicating
charge is the charge per CCD column (2.4 cm1 spectral width). the value of the calculated differential sum and difference scat-
R
Cp is the degree of circularity. The small polarized band at tering cross-sections. Temperature assumed for the population of
732 cm1 is due to an impurity. vibrational states: 25 C. Frequencies are unscaled.
(M)-()-[4]triangulane, which underly the couplet cen- On each achiral half of the molecule the scissoring
tered at 1395 cm1 in the experimental (Figure 3) and at motions of the three CH2 groups couple in the same way for
1435 cm1 in the calculated (Figure 4) ROA spectrum.59 vibrations 38 and 39: in phase for the two CH2 groups of the
The C2 symmetric molecule, recently synthesized in enan- terminal cyclopropane ring, with this combined scissoring
tiomerically pure form by de Meijere et al.60 and probably motion being out of phase to that of the CH2 group of the
best designated as s-(M)-()-[4]helicene, represents the spiro-anelated inner ring. For the vibration with symmetry
archetype of a helical vibrational chromophore in the same A the scissoring motions on one half of the molecule are
way as the aromatic p-[n]helicenes represent the basic in phase with the C2 -symmetry-related scissoring motions
model for helical electronic chromophores. The excellent on the other half, while for the vibration with B symme-
agreement of the theoretical and experimental ROA spec- try they are out of phase. The pertinent dinuclear terms,
tra, inspite of the marginal basis set used for the electronic equations (20a) and (20b), therefore have opposite signs for
tensor calculation, indicates that the ROA of the molecule these vibrations. This in turn leads to opposite calculated
is completely determined by the helical arrangement of the signs for the two ROA bands, in agreement with experi-
atoms. One therefore also expects the atomic contribution ment, even though there are substantial differences in the
pattern of integral ROA difference scattering cross-sections, excursions of individual nuclei in the two vibrations. The
as shown for vibration 38 in Figure 5 (symmetry A, positive energetic splitting between the two vibrations is due to this
band in the right minus left convention) and for vibration differing composition, and not to any sort of exciton-type
39 in Figure 6 (symmetry B, negative band), to be largely interaction.
independent of computational limitations.
5 THE INFLUENCE OF ELECTRONIC

CHIRALITY ON RAMAN OPTICAL
ACTIVITY
5.1 Empirical evidence
Assessing the relative importance of the chirality of the

nuclear motion and of the electron distribution in ROA gen-
eration is of considerable theoretical and practical interest.
Except for the conclusion, obtained in the single electronic
state limit, that the sign of the CD of electronic transi-
(a) (b) tions dominates resonance ROA,25 no theoretical work on
Figure 5. Quasi-atomic distribution of integral (a) Raman (n sSCP ) this subject has so far been done. From empirical evidence
and (b) ROA (n sSCP ) scattering cross-section for vibration on the other hand it is well known that in the far from
38. The surface of the spheres is proportional to the contribution resonance limit, in contrast to the resonance case,63 ROA
of individual atoms. Dark shading indicates a positive and light spectra are composed of close to equal amounts of posi-
shading a negative sign. The combination of invariants for SCP
tive and negative difference intensities, and that molecules
backscattering yields a qualitatively similar pattern.
chiral by isotope substitution, but with an achiral electron
distribution, can show ROA of the same size as molecules
with a chiral mass and electron distribution.32 Both facts
strongly suggest that the chirality of the electron distribu-
tion is far less important than vibrational chirality in ROA
generation.
5.2 A quantitative measure
The influence of electronic chirality on ROA cannot be

(a) (b) specified for individual vibrational modes, but it is possible
Figure 6. Quasi-atomic distribution of the integral (a) Raman to quantify it for the whole of an ROA spectrum. This
(n sSCP ) and (b) ROA (n sSCP ) scattering cross-section for vibra- requires summing expressions of the form of equation (22)
tion 39. Other details as in Figure 5. over all normal modes of vibration p. This sum can be
carried out if the rotranslational coordinates are formally where, as in equation (27), the sum over q is formally
added. As translation does not generate ROA, the rotational over rotational coordinates, and p is limited to vibrational
coordinates q only need to be explicitly considered: modes.
Rotational ROA is determined by the anisotropy of
Jab,p C Jab,q D Vai,bj Lax i,r Lbx j,r purely electronic tensors. We can therefore consider it
p q r i,j an expression of electronic chirality and subtract it from

D Vai,bj ma1 ai,bj equation (28) in order to get a measure of the influence of
i,j electronic chirality on vibrational ROA alone:

D Vai,ai ma1 ab vib D rovib rot

i
9aG0p C 2b2Gp
D Va ma1 ab 27 p
D
j9aG0p C 2b2Gp j
The sum of the rovibrational matrices Jab,r is a diag- p
onal matrix with the elements Va ma1 . Dinuclear terms
therefore do not contribute to summed Raman and ROA Ip nQ p [1 exp100hcnQ p /kT]/nQ 3p
intensity. p
D
In the ROA case the purely electronic factors Va rep- p jIp jnQ p [1 exp100hcnQ p /kT]/nQ 3p
resent invariant combinations of elements of a true and 30
a pseudo second-rank tensor. Their sign and value reflect where Ip are difference intensities measured in terms of
the local electronic chirality at the site of each nucleus. If the number of scattered photons of energy hnp , and where it
nucleus a is located in a reflection plane, then Va D 0, and is assumed that the spectral width of individual vibrational
for two nuclei a and b at enantiotopic sites Va D Vb .64 bands is small enough so that they may be characterized by
The multiplication with ma1 corresponds to a weighting by a discrete frequency nQ p .
the square of the average excursion of nucleus a. The sum The ab initio values for s-(M)-()-[4]helicene are
of the terms Va ma1 can thus be considered the contribution rot D 0.13% and vib D 1.40%. These are typi-
by uncompensated electronic chirality, as locally sampled cal values for molecules with little anisotropy in their
by the moving nuclei, to the generation of ROA difference electron distribution but a strong vibrational ROA spec-
intensities. trum. The values calculated with the same method and
This contribution due to electronic chirality, as seen the same type of basis set (see Figure 4) for (R)-(C)-
from the perspective of the inverse mass distribution of a-phenylethylamine, a molecule with a weak vibrational
the nuclei, is best specified by a fractional, dimensionless ROA spectrum but a strongly anisotropic electron distri-
measure. For integral ROA scattering into 4p one can write bution, are rot D 2.72% and vib D 2.01%. Even in
this latter case electronic chirality contributes negligibly to
9Aa C 2Ba ma1
the vibrational ROA spectrum, which retains an essentially
a
rovib 28 balanced appearance between positive and negative con-
j9aG0p C 2b2Gp j tributing bands.
p
where Aa and Ba stand for Va with a combination of tensor

gradients that corresponds to the isotropic and anisotropic 6 CONCLUDING REMARKS
invariants aG0 and b2G , respectively. In the denominator
rotational terms are omitted: isotropic contributions vanish, In a sense, the birth of vibrational optical activity was late
and the size of the anisotropic ones is negligible in com- timing. When its evident potential for providing stereo-
parison to the other terms. In the numerator they cannot chemical information was, in the late 1970s and early
be neglected, though. In contrast to vibrational modes all 1980s, eventually demonstrated in specific applications,
classical stereochemistry was considered by many, rightly
rotational transitions have the same sign.65 Their fractional
or wrongly, to have reached or passed its zenith. On the
contribution is given by
other hand, using ROA or VCD to solve problems in the
2b2Gq 2Ba ma1 2b2Gp then much favored area of biomolecular chemistry was not
q a p yet feasible. One might think about the impact vibrational
rot D D
j9aG0p C 2b2Gp j j9aG0p C 2b2Gp j optical activity would have had on stereochemistry if it
p p had been available in the 1950s, as envisaged by the late
29 stereochemist Prelog.
However, we believe that ROA and VCD both have an CIS Circular Intensity Sum
important future. This is true even in the face of alter- DCP Dual Circular Polarization
native techniques, such as chiral shift reagents in nuclear ICP Incident Circular Polarization
magnetic resonance (NMR) or enantiospecific columns in ORD Optical Rotary Dispersion
gas chromatography, which have emerged over the past two ROA Raman Optical Activity
decades as physico-chemical methods, and as alternatives to SCP Scattered Circular Polarization
the well-established CD of electronic transitions, for deter-
mining absolute configurations. For one, such methods can
depend upon the support by tedious synthetic chemistry REFERENCES
for the preparation of appropriate derivatives, while ROA
and VCD can be applied directly to any chiral molecule, 1. F. Arago, Mem. Classe Sci. Math. Phys. Inst. Imper. France,
provided the compound is liquid, or soluble in an appro- 12I, 93 (1811).
priate solvent. For another, vibrational optical activity pro- 2. J.B. Biot, Ann. Chim. Phys., II, 4, 90 (1817).
vides not only information on the absolute configuration of 3. A. Fresnel, Ann. Chim., 28, 147 (1825).
molecules but a wealth of additional stereochemical infor- 4. A. Cotton, Ann. Chim. Phys., VII, 8, 347 (1896).
mation as well. With the emergence of a bulk of empirical 5. A. Kastler, C.R. Acad. Sci. Paris, 191, 565 (1930).
material within the realm of ROA and VCD, this informa-
6. S. Bhagavantam and S. Venkateswaran, Nature, 125, 237
tion will become as valuable, or perhaps in many cases even (1930).
more so, than the ability of the two methods to determine
7. F. Perrin, J. Chem. Phys., 10, 415 (1942).
absolute configurations. ROA in particular, where water is
the perfect solvent, shows a high potential for probing the 8. H.J. Hediger and H.H. Gunthard, Helv. Chim. Acta, 37, 1125
(1954).
tertiary structure and the dynamics of biopolymers.66
The lack of appropriate commercially available instru- 9. H.R. Wyss and H.H. Gunthard, Helv. Chim. Acta, 49, 660
(1966).
mentation, an important obstacle to a more general appli-
cation of vibrational optical activity, no longer exists in 10. L.D. Barron, M.P. Bogaard and A.D. Buckingham, J. Am.
Chem. Soc., 95, 603 (1973).
VCD. The somewhat more modest level of optical sophisti-
cation in standard ROA instrumentation, as compared with 11. L.D. Barron, M.P. Bogaard and A.D. Buckingham, Nature,
241, 113 (1973).
VCD, would in the past have made the development of
commercial ROA instruments difficult. Yet, recent progress 12. L.D. Barron and B.P. Clark, J. Chem. Res. (S), 36 (1979).
fosters hope that commercial instrumental ROA alternatives 13. W. Hug, A. Kamatari, K. Srinivasan, H.-J. Hansen and H.-R.
to VCD spectrometers might nevertheless become available Sliwka, Chem. Phys. Lett., 76, 469 (1980).
over the coming years. In spite of their difficult start, ROA 14. G. Holzwarth, E.C. Hsu, H.S. Mosher, T.R. Faulkner and
and VCD will then have assumed the place in the realm of A. Moscowitz, J. Am. Chem. Soc., 96, 251 (1974).
the analytical methods of chiral chemistry merited by their 15. L.A. Nafie, J.C. Cheng and P.J. Stephens, J. Am. Chem. Soc.,
potential. 97, 3842 (1975).
16. W. Hug, S. Kint, G.F. Bailey and J.R. Scherer, J. Am. Chem.
Soc., 97, 5589 (1975).
ACKNOWLEDGMENTS 17. P.W. Atkins and L.D. Barron, Mol. Phys., 16, 453 (1969).
The SCP instrument mentioned was constructed at the Uni- 18. L. Blum and H.L. Frisch, J. Chem. Phys., 52, 4379 (1970).
versity of Zurich, in the frame of a collaborative project 19. L.D. Barron and A.D. Buckingham, Mol. Phys., 20, 1111
with the Institute of Organic Chemistry (Prof. H.-J. Hansen) (1971).
with support by the Legerlotz Foundation (to H.-J. H.). 20. L.A. Nafie and T.B. Freedman, Chem. Phys. Lett., 154, 260
s-(M)-()-[4]helicene was obtained from Prof. A. de Mei- (1989).
jere, Gottingen. Computations were done by G. Zuber, 21. D. Che, L. Hecht and L.A. Nafie, Chem. Phys. Lett., 180, 182
Fribourg. The main financial support of the ROA project (1991).
was through Swiss National Science Foundation Grants 21- 22. L.D. Barron and J.R. Escribano, Chem. Phys., 98, 437 (1985).
50555.97 and 20-56905.99. 23. L. Hecht and L.A. Nafie, Mol. Phys., 72, 441 (1991).
24. L.D. Barron, Molecular Light Scattering and Optical Activ-
ity, Cambridge University Press, London (1982).
25. L.A. Nafie, Chem. Phys., 205, 309 (1996).
CD Circular Dichroism 26. J.O. Bjarnason, H.C. Anderson and B.S. Hudson, J. Chem.
CID Circular Intensity Difference Phys., 72, 4132 (1980).
27. G. Wagniere, Chem. Phys., 54, 411 (1981). 47. P.L. Polavarapu, J. Chem. Phys., 77, 2273 (1982).
28. J.-L. Oudar, C. Minot and B.A. Garetz, J. Chem. Phys., 76, 48. T.B. Freedman and L.A. Nafie, J. Chem. Phys., 78, 27, 1104
2227 (1982). (1983).
29. T. Helgaker, K. Ruud, K.L. Bak, P. Jorgenson and J. Olsen, 49. G. Placzek, Rayleigh-Streuung und Raman-Effekt, in
Faraday Discuss., 99, 165 (1994). Handbuch der Radiologie, Akademische Verlagsgesell-
30. W. Hug and H. Surbeck, Chem. Phys. Lett., 60, 186 (1979). schaft, Leipzig, 205374, Vol. VI (1934).
31. W. Hug and H. Surbeck, J. Raman Spectrosc., 13, 38 (1982). 50. A.D. Buckingham, Adv. Chem. Phys., 12, 107 (1967).
32. W. Hug, Instrumental and Theoretical Advances in Raman 51. P.J. Stephens, F.J. Devlin, C.S. Ashvar, C.F. Chabalowski
Optical Activity, in Raman Spectroscopy, Linear and Non- and M.J. Frisch, Faraday Discuss., 99, 103 (1994).
linear, eds J. Lascomb and P. Huong, Wiley-Heyden, Chich- 52. R.D. Amos, Chem. Phys. Lett., 87, 23 (1982); 124, 376
ester, 312 (1982). (1986).
33. L. Hecht, L.D. Barron and W. Hug, Chem. Phys. Lett., 158, 53. P.L. Polavarapu, J. Phys. Chem., 94, 8106 (1990).
341 (1989).
54. P. Bour, Chem. Phys. Lett., 288, 363 (1998).
34. L.D. Barron, L. Hecht, A.D. Gargaro and W. Hug, J. Raman
Spectrosc., 21, 375 (1990). 55. G.D. Purvis, III and R.J. Bartlett, J. Chem. Phys., 76, 1910
(1982).
35. W. Hug and G. Hangartner, J. Raman Spectrosc., 30, 841
(1999). 56. T.H. Dunning, Jr, J. Chem. Phys., 90, 1007 (1989).
36. K.M. Spencer, T.B. Freedman and L.A. Nafie, Chem. Phys. 57. R.A. Kendall, T.H. Dunning, Jr and R.J. Harrison, J. Chem.
Lett., 149, 367 (1988). Phys., 96, 6796 (1992).
37. L.A. Nafie and D. Che, Theory and Measurement of Raman 58. W. Hug, Chem. Phys., in press.
Optical Activity, in Modern Nonlinear Optics, Part 3, 59. W. Hug, G. Zuber, A. de Meijere, A.F. Khlebnikov and H.-J.
eds M. Evans and S. Kielich, J. Wiley & Sons, New York, Hansen, Helv. Chim. Acta, 84, 1 (2001).
105149, Vol. 85 (1994).
60. A. de Meijere, A.F. Khlebnikov, R.R. Kostikov, S.I. Kozhush-
38. D.A. Long, Raman Spectroscopy, McGraw-Hill, New York kov, P.R. Schreiner, A. Wittkopp and D.S. Yufit, Angew.
(1977). Chem., 111, 3682 (1999).
39. W. Hug, Raman Optical Activity, Spectrometers, in 61. GAUSSIAN98, Gaussian Inc., Pittsburgh, PA.
Encyclopedia of Spectroscopy and Spectrometry, eds
J. Lindon, G. Tranter and J. Holmes, Academic Press, 62. T. Helgaker, H.J.Aa. Jensen, P. Joergensen, J. Olsen,
London, 19661976 (1999). K. Ruud, H. Aagren, T. Andersen, K.L. Bak, V. Bakken,
O. Christiansen, P. Dahle, E.K. Dalskov, T. Enevolsen, B.
40. W. Hug, Appl. Spectrosc., 35, 115 (1981).
Fernandez, H. Heiberg, H. Hettema, D. Jonsson, S. Kirpe-
41. M. Vargek, T.B. Freedman and L.A. Nafie, J. Raman Spec- kar, R. Kobayashi, H. Koch, K.V. Mikkelsen, P. Norman,
trosc., 28, 627 (1997). M.J. Packer, T. Saue, P.R. Taylor and O. Vahtras, DALTON
42. L.D. Barron, Raman Optical Activity, in Advances in release 1.0 (1997), an electronic structure program.
Infrared and Raman Spectroscopy, eds R.J.H. Clark and 63. M. Vargek, T.B. Freedman, E. Lee and L.A. Nafie, Chem.
R.E. Hester, Heyden, London, 271331, Vol. 4 (1978). Phys. Lett., 287, 359 (1998).
43. D.L. Andrews, Faraday Discuss., 99, 375 (1994). 64. B. Cuony and W. Hug, Chem. Phys. Lett., 84, 131 (1981).
44. L.D. Barron and A.D. Buckingham, J. Am. Chem. Soc., 101, 65. L.D. Barron and C.J. Johnston, J. Raman Spectrosc., 16, 208
1979 (1979). (1985).
45. L.D. Barron and J. Vrbancich, Mol. Phys., 48, 833 (1983). 66. L. Hecht, L.D. Barron, E.W. Blanch, A.F. Bell and L.A. Day,
46. P.L. Prasad and L.A. Nafie, J. Chem. Phys., 70, 5582 (1979). J. Raman Spectrosc., 30, 815 (1999).
Electromagnetic Mechanism of Surface-enhanced
Spectroscopy
George C. Schatz and Richard P. Van Duyne
Northwestern University, Evanston, IL, USA
1 INTRODUCTION Let us begin by defining what we mean by the EM. When

an electromagnetic wave interacts with any metal surface,
The purpose of this article is to review the electromag- be it smooth or rough, the field near the surface is different
netic mechanism (EM) of surface-enhanced spectroscopy from that far away from it. If the surface is flat, the appli-
(SES). This mechanism has played an important role in our cation of Maxwells equations to determine the field near
understanding of SES since the original papers on surface- the surface, and the dependence of Raman intensities on the
enhanced Raman spectroscopy (SERS),1 3 and in many incident and scattered angles, was worked out by Greenler
respects it is a mature topic, having been the subject of and Schlager well before the discovery of SERS.15 There
numerous reviews.4 13 However, there are many features of is a small enhancement in Raman intensity compared with
the comparison with experiment that are still rather vague, that in the absence of the surface, on the order of a fac-
and this has served as a source of confusion in assessing tor of 10 or less for metals like Ag, arising primarily from
the role of other mechanisms, such as chemical mecha- coherent superposition of the incident and reflected fields at
nisms, in SES.14 Since it has now been several years since the position of the molecule doing the scattering. Note that
the topic has been discussed in a general way, it seems although collective excitations of the conduction electrons
appropriate to present a new review in which the current known as surface plasmons exist for Ag at frequencies of
roughly h! D 3.5 eV, they cannot be excited by the electro-
status of the EM is considered. Much of this discussion will
magnetic field when irradiating a flat surface, as momentum
center on SERS, as it is still the most extensively studied
cannot be conserved in the excitation process.
of the enhanced surface spectroscopies, but other spectro-
If the surface under consideration is roughened, then
scopic methods will also be brought in where relevant, as
the electrodynamics of the irradiated surface become much
the comparison of theory and experiment for other meth-
more interesting. The key result is that surface plasmons
ods has in many cases provided important new insights
can now be excited by electromagnetic radiation, resulting
into the predictive ability of the theory. The other meth-
in enhanced electromagnetic fields close to the surface.16,17
ods that will be considered include surface-enhanced sec-
In Raman scattering, the intensity depends on the square of
ond harmonic generation (SESHG) and surface-enhanced
the incident field strength, and as a result, the intensity is
hyper-Raman spectroscopy (SEHRS). We will also discuss
enhanced relative to what it would be in the absence of the
recent advances in the electromagnetic description of linear
surface. The Raman emitted field may also be enhanced,
extinction and scattering that are relevant to SES. Much
though generally by a different amount than the incident
of the discussion will focus on isolated nanoparticles, as
field, since the frequency is different. Also the field is
this provides the simplest problem that still exhibits all
spatially different as it arises from an oscillating dipole
the features of SES that are also found in more complex located at the position of the emitting molecule. The over-
systems. all enhancement associated with the incident and emitted
fields is what we shall consider to be the EM contribu-
John Wiley & Sons Ltd, 2002. tion to SERS. An attractive feature of this mechanism
2 Surface-enhanced Vibrational Spectroscopy
is that the enhancement and its wavelength dependence arises in the values of the metal particle dielectric constants.
may be estimated using calculations that require only the Usually these are taken from measurements on bulk metals,
size and shape of the metal particles or surface involved, but for small particles (smaller than the conduction electron
and the relevant dielectric constants. A number of cal- mean free path) this choice is inappropriate, and corrections
culations of the enhancement based on this idea have are necessary.34
appeared in the literature, including work by Kerker and Extensions and additional complications in applications
co-workers,18 21 Gersten and Nitzan,22 Adrian,23 Metiu,11 of the EM to single-metal particles are the subject of later
Barber and co-workers,24 Schatz and co-workers25,26 and sections of this review. One issue is the influence of the
Kottmann and co-workers27 on isolated particles. In addi- adsorbate on the interaction of the electromagnetic field
tion, several groups have modeled SERS on gratings,28 30 with the metal particles or surfaces. (Usually the adsor-
Laor and Schatz31,32 considered assemblies of particles on bate has been neglected in the spheroid calculations noted
surfaces, and Sanchez-Gil and Garcia-Ramos33 and others above.) This is especially important when the adsorbates
(cited in Sanchez-Gil and Garcia-Ramos33 ) have studied absorb light at the incident or scattered frequencies, i.e.
randomly roughened surfaces. The theory predicts large resonant Raman scattering, as then the plasmons can be
enhancements for all of these structures provided that the damped by interaction with the adsorbate,35,36 but it also
metal involved has narrow plasmon resonances at con- plays a role in nonresonant scattering,37 when it contributes
venient frequencies for Raman measurements (i.e. in the to the coverage dependence of the Raman intensities.
visible region of the spectrum). In the next section, we Another important extension to the EM that we dis-
examine what this means for a variety of different mate- cuss later arises in the consideration of nonlinear surface
rials, and we also consider what the model predicts for spectroscopies, including SESHG and SEHRS. As long as
wavelengths that have not been considered in traditional the nonlinear polarization arises from the electromagnetic
experiments, but which are now becoming accessible thanks field gradient near the particle surface (the jury is still
to advances in laser and detector technology. out on this38 41 ), it is straightforward to extend the elec-
To clarify the qualitative definition of the EM just given, tromagnetic model to the description of nonlinear surface
we begin the next section with a simple quasistatic model optical spectroscopy by factoring in the additional pow-
of the EM enhancement. In this model, we consider that ers of the incident field at the appropriate wavelength. This
molecules are adsorbed on the surface of a sphere that model implicitly assumes that SESHG is actually incoherent
is small compared with the wavelength of light, so that hyper-Rayleigh scattering, which for isolated nanoparticle
Maxwells equations may be replaced by the LaPlace equa- systems is probably a reasonable assumption. Comparisons
tion of electrostatics. The frequency dependent dielectric with experiment can then be made, and these provide two
constant is still used in this calculation. This model contains important pieces of information concerning the enhance-
most of the essential issues that are important to SES, even ment that are not available from SERS alone. First, the
for SES on roughened electrode surfaces, but it is not useful enhancement can be measured for materials and wave-
for quantitative estimates of enhancements, as the plasmon lengths that are not generally accessible to SERS. This is
resonances associated with spherical particles lie far enough especially true for SESHG, where even materials like Pt
to the blue that even for the noble metals they are damped have been studied.42 Second, the comparison of electro-
by interband transitions and other excitations, i.e. the metals magnetic model enhancements with experimental enhance-
are absorbers at the plasmon frequencies. More quantitative ments provides an independent determination of the relative
estimates of enhancement factors are obtained from mod- magnitudes of the electromagnetic and nonelectromagnetic
els that allow the metal particles to have spheroidal shapes. contributions to SERS. In addition, one can use the com-
We discuss such particles in the next section, although to parison of SERS and SEHRS to study the local symmetry
keep the mathematics simple, we omit most details, and experienced by the adsorbed molecules, as the selection
emphasize instead the results of calculations and compar- rules for SERS and SEHRS are quite different.
isons with experiment. However, even these calculations Another useful application of electromagnetic theory that
are not quantitative unless a number of additional issues we consider in this review is to the calculation of extinc-
are addressed. For example, the correct treatment involves tion and scattering cross-sections for the same nanoparti-
electrodynamics not electrostatics, and often the differences cles as are considered in calculating SERS spectra. These
are significant even when the particles are only 10% of cross-sections may be obtained from the same solutions
the wavelength of light in longest dimension. Approaches of Maxwells equations as are used to model SERS, but
for treating this issue range from brute force numerical they depend on the asymptotic properties of the solutions
solutions of Maxwells equations24,26,27 to perturbation cor- rather than surface fields. For very small spherical particles,
rections of the electrostatic treatment.25 Another problem there is a close relationship between these cross-sections
Electromagnetic Mechanism of Surface-enhanced Spectroscopy 3
and the SERS excitation profiles (which will be given in the Electromagnetic coupling between the adsorbate and sur-
next section), so that both have plasmon peaks at the same face at optical frequencies arises from the dipole moment
wavelength. However, for nonspherical particles and larger in the adsorbed molecule that is induced by the applied
spherical particles there are significant differences in wave- electromagnetic field. This dipole will polarize the metal
length dependence that provide important information about electrons at optical frequencies, leading to a change in the
the EM. In addition, we can learn much from extinction effective polarizability of the adsorbate. This mechanism,
spectra about the influence of solvent and substrate dielec- which is just another part of the polarization response of
tric properties on EM enhancements. the adsorbate/substrate system that also produces the EM,22
In the last section of this review, we consider applications is often called the image field mechanism,10 as it can be
of the EM to models of the surface other than a single- pictured as arising from interaction of the molecule with its
metal particle. Such models can take many different forms, image in the metal. However, the magnitude of the enhance-
ranging from a small number of interacting spheres or other ment arising from this effect is significant only when the
shapes, to large assemblies of particles, either randomly molecule is too close to the surface to model its interaction
arranged or in regular arrays, and to gratings. accurately using electrodynamic theory with point dipoles,
While the discussion to this point has emphasized how so in most treatments of the EM22 it has ben explicitly
the EM has been developed and extended, it is also impor- excluded from consideration. This is not correct, of course,
tant to consider how the EM fits into the larger picture but the treatment of molecules close enough to the surface
that theorists have attempted to develop of the complete to experience orbital overlap is best done using approaches
enhancement process. A common definition is to assume that consistently combine the electronic wavefunctions of
that any part of the surface enhancement that is not the molecule and metal in describing the electrodynamic
accounted for using the EM involves the chemical mech- response of the system. So far, there have been relatively
anism. With this definition, chemical mechanisms refer few calculations of this type,45,50 52 and all have made
both to enhancements that arise from interactions between severe enough approximations that it is hard to quantify
molecule and surface that require orbital overlap between the true significance of the image effect. Most of the SERS
the molecule and metal wavefunctions, and to those inter- enhancement can be explained using electrodynamics mod-
actions that do not require overlap. The most commonly els that ignore the image effect, so it seems best to ignore it.
considered interaction that requires overlap occurs when
charge transfer between surface and adsorbate leads to
the formation of excited states that serve as resonant 2 ELECTROMAGNETIC MECHANISM
intermediates in Raman scattering.43 Interactions that do OF SURFACE-ENHANCED RAMAN
not require overlap arise from electromagnetic coupling SPECTROSCOPY FOR ISOLATED
between the vibrating molecule and the metal. These inter-
METAL PARTICLES
actions can occur either at the vibrational frequency or
at optical frequencies. The former case involves interac- 2.1 A simple model: quasistatic treatment of an
tion of the oscillating dipole and/or quadrupole field of isolated sphere
the vibrating adsorbed molecule with the surface electrons.
Such fields would be expected to modulate the metal sur- The model that we will develop consists of a single metal
face dielectric properties (especially at frequencies close to sphere, small compared to the wavelength of light, which
the surface plasmon frequency where the metal is highly is irradiated by a laser field. Raman scattering arises from
polarizable), leading to Raman shifted reflection of light molecules that are adsorbed on the surface of this sphere. For
from the surface. This mechanism is sometimes called the now we will ignore the molecules in determining the field
Raman reflectance mechanism,44 48 and it can also arise near the surface, and assume that the sphere is embedded in
from modulated overlap between the molecule and surface a medium (which could be vacuum or solvent) of dielectric
orbitals.47,48 The electrostatic part of Raman reflectance has constant eo . The dielectric constant inside the metal sphere
been explored by Dignam and co-workers,49 who use the is denoted by ei , and for now we take this dielectric constant
term surface phonon-induced Raman scattering (SPIRS) to to be independent of the size of the sphere.
describe this mechanism. This mechanism is most important We take the electric field of the incident electromagnetic
for small metal particles (25 A), where it can be comparable wave to be Eo , a vector that points along the z-axis, and that
in enhancement to the EM. However, its importance drops can be assumed independent of coordinates for distances
off rapidly with particle size. It is unfortunately not clear at least as large as the sphere. Under these conditions,
how important this mechanism is for conditions that model Maxwells equations may be approximated by LaPlaces
real experiments. equations to determine the field both inside and outside the
sphere. The resulting field outside the sphere, Eout , can then in equation (4) determines the magnitude of the induced
be written as: dipole, but there is also the possibility of enhanced emis-

z 3z sion from this dipole. This enhancement is more difficult to
Eout D Eo z aEo 3 5 zz C xx C yy
1
evaluate, as unlike the applied field Eo which is accurately
r r
described as being constant over the volume of the sphere,
where the first term is the applied field, and the second the emitted field is more complex. The proper treatment
is the induced dipole that results from polarization of the of this field has been given by Kerker et al.19,20 To a first
sphere electron density. a is the metal polarizability, x, approximation that is appropriate for small particles (see
y, z, r, x, y, z are the usual Cartesian coordinates, radial Kerker21 ), the enhancement is determined by an expression
distance and Cartesian unit vectors, respectively and Eo is similar to equation (4), but evaluated at the Raman-shifted
the magnitude of Eo . For a metal sphere with the dielectric frequency. If we consider molecules located at the posi-
constants indicated above, the polarizability is tion of maximum enhancement and consider only the limit
a D ga3 2
jgj 1, then the overall enhancement arising from incident
and scattered fields is approximately
where a is the sphere radius and
0
ei eo E2 E 2
gD 3
ER D out 4 out D 16jgj2 jg0 j2 5
ei C 2eo Eo
Note that the polarizability is a number that is on the where the primed symbols refer to fields evaluated at
order of the sphere volume in the limit of zero frequency, the scattered frequency. The corresponding angle averaged
as in that case ei is 1. However, whenever the real enhancement factor would be smaller than this by a factor
part of ei equals 2eo and the imaginary part is small, of 4. Equation (5) is identical to one derived by Kerker
a becomes very large, and as a result the induced field et al.19,20 using more rigorous theory. For small Stokes
becomes large. It is this induced field that is responsible for shifts, jgj and jg0 j maximize at approximately the same
the electromagnetic enhancement. Note that this plasmon wavelength, in which case if jgj and jg0 j are both approx-
resonance condition requires the real part of ei to be large imately 10 in magnitude, then the enhancement will be
in magnitude and negative, and this is always the case for a approximately 105 . If we use the free-electron Drude model
free electron metal at long enough wavelengths. In addition, for the metal dielectric constant:
the formula says that the plasmon resonance wavelength
!2p
will change when the dielectric constant eo of the external ei D 1 6
medium is varied. Indeed, since the metal dielectric constant !! C ig
typically increases (becomes less negative) as wavelength where !p is the plasmon frequency and g is the plasmon
decreases, the resonance condition will be satisfied for width, then the peak magnitude of g will be roughly 10
longer wavelengths, i.e. red-shifts, as solvent dielectric when the ratio of the plasmon frequency to plasmon width
constant increases. is roughly 10. There are several materials such as Ag where
The Raman intensity depends on the absolute square of the peak jgj 10. However the free electron model for
Eout which we denote as E2out , evaluated at the surface of the dielectric constant is rarely sufficiently accurate for
the sphere (i.e. r D a). From equation (1), this is given by this prediction to be quantitative. In fact, electromagnetic
enhancements are generally rather small for spherical parti-
E2out D E2o [j1 gj2 C 3 cos2 q2 Reg
C jgj2
] 4
cles because the surface plasmons for most materials are

where q is the angle between the applied field direction and far enough to the blue that g is large (typically due to
the vector r that locates positions on the sphere surface. interband transitions). Larger enhancements arise for pro-
Note that if jgj is large, equation (4) reduces to E2out D late spheroidal objects, as the plasmon resonances for such
E2o jgj2 1 C 3 cos2 q
. This indicates that the largest field objects occur at frequencies that are red-shifted relative to
intensities are obtained for angles q equal to zero or 180 , spherical particles to frequencies where the plasmon width
i.e. along the polarization direction. Also, the ratio of the is sufficiently small to make jgj 10 or larger. For this reason
largest to smallest field enhancement as a function of q is 4, we turn our attention in the next section to the electrody-
and the angle integrated expression is 1/2 the peak intensity. namics of spheroidal particles.
In Raman scattering the applied field induces an oscil- The results obtained to this point are related closely
lating dipole in the adsorbed molecule. This dipole then to the extinction and scattering cross-sections for spheri-
radiates, and the component of this radiation that has been cal particles, as obtained from the long wavelength limit
shifted by the vibrational frequencies of the molecule deter- (a/l 1) of Mie theory.53 In particular, if we define
mines the Raman scattering intensity. In SERS the field x D 2paeo
1/2 /l, then the extinction cross-section Qext is:
Qext D 4x Img
7a
and
ei eo
gD 12
and the scattering cross-section Qsca is: ei C ceo
8 4 2 The parameter c in equation (12) is given by

Qsca D x jgj 7b
3 1
c D 1 C 13
where g is still given by equation (3). Evidently, if we xo xo C 1

x2o 1
ln 1
leave out the x 4 prefactor (which varies as l4 ), the wave- 2 xo 1
length dependence of equation (7b) is the same as the large Note that g in equation (12) is a generalization of that in
jgj limit of equation (4). In addition, as long as the plas- equation (3). In fact the parameter c equals 2 for a sphere,
mon width is much smaller than the plasmon energy, the but for prolate spheroids, i.e. those with b > axo > 1
,
extinction cross-section has the same wavelength depen- c is larger than 2 and for oblate spheroids (xo < 1) it is less
dence (after leaving out the x factor) as the large jgj limit than 2. When c is greater than 2, the plasmon resonance
of equation (4). Since the extinction cross-section is much condition, Reei C ceo
D 0, is satisfied for a wavelength
larger than the scattering cross-section for a small metal par- that is to the red of that for a sphere (due to the fact that
ticle, it is extinction that is most often compared to SERS for metals, the real part of ei is, according to equation (7),
excitation profiles. What we see therefore is that in the more negative for longer wavelengths). Of course this also
small particle limit, SERS excitation and extinction plas- means that for oblate spheroids, the resonance will be blue-
mon peaks and widths should be the same. shifted relative to a sphere. However, it should be noted that
the resonance described here refers to an incident field with
the electric polarization parallel to the symmetry axis of the
2.2 Electrostatics of isolated spheroids spheroid. There is another plasmon resonance associated
with excitation perpendicular to the symmetry axis. This
The exact analytical solution to the electrodynamics of resonance is identical in frequency to the parallel resonance
spheroidal particles is very complex, so the work done for a sphere, but it shifts in the opposite direction for a
has involved either numerical solutions of Maxwells equa- spheroid, i.e. blue-shifting for prolate spheroids and red-
tions24,26 or analytical electrostatic solutions19,20,23,25 that shifting for oblate spheroids. Indeed, the expression for g
are valid only for small particles. Although the numerical in equation (12) still applies, but the expression for c is:
solutions are more accurate, the electrostatic solutions are
1
more instructive. Fortunately, for particle sizes of signifi- c D 1 14
cant interest to the experiments (<100 nm), the two theories 1 xo C 1

xo x2o 1
ln x2o
give very similar results. 2 xo 1
If we consider a spheroid whose major axis is of length Thus we see that both oblate and prolate spheroids have
2b and minor axis 2a, with a constant field Eo applied plasmons that are red-shifted relative to the sphere case.
along the major axis, then an explicit expression for the Since most SERS-active metals (Ag, Au, Cu) exhibit
Raman enhancement factor for molecules that are randomly free electron-like behavior at long wavelengths, it is the
distributed on the spheroid surface (i.e. averaged over red-shifted plasmon resonance that contributes most to
the surface) has been given by Zeman and Schatz25 as SERS. Thus nonspherical particles, either prolate or oblate,
follows: give larger SERS intensities than spheres. In addition, we
E R D R!
R!
8
see that increasing the dielectric constant of the external
medium, eo , leads to additional red-shifting of the plasmon
where resonances, just as with the sphere case that we described

2 2 Re1 g
g jgj2 earlier.
R!
D j1 gj C C 2 Equation (9) refers to the enhancement averaged over
Q1 x o
Q1 xo
x2o 1
the surface of the spheroid. In some applications it is also

x2o 1
1/2 C x2o sin1 1/xo
of interest to have an expression for the enhancement as
9
x2o 1
1/2 C x2o sin1 1/xo
a function of location on the surface. This expression is
1/2 given by:
a2
xo D 1 2 10
b R!, h
D j1 gj2

1 xo C 1 2 Re[1 g
g ]h2 jgj2 h2
Q1 xo
D xo ln 1 11
C C 2 15
2 xo 1 Q1 xo
x2o h2
Q1 xo
x2o 1
x2o h2
Note that this expression depends on the spheroidal coor- and

dinate h, which is related to the usual spherical polar f D b2 a2
1/2 19
coordinate q via:
Here E is the (enhanced) applied field. Inclusion of elec-
x2o 1 trodynamics corrections leads to the modified expression
h D cos q 16
xo cos2 q
2
mind D ao E C Erad
20
Some specific examples of results based on equa- where

tions (816) will be given later. First, however, we need k2 2i 3
Erad D C k mind 21
to discuss corrections and improvements to them, as these b 3

equations are not yet sufficiently accurate to be useful. and the wavenumber k was defined previously. The term in
Erad proportional to k 2 describes dynamic depolarization,
while that proportional to k 3 describes radiative damping.
2.3 Improvements to the spheroid model Substitution of equation (21) into equation (20) leads to a
modified expression for the induced dipole moment
The treatment up to this point has considered an isolated 1
spheroid with dielectric constant ei interacting with an ao k 2 2ik 3 ao
mind D 1 ao E 22
electromagnetic field that is treated in the static limit. b 3

It turns out that for particle sizes of most relevance to Note that the second and third terms in the denominator in
the experiments (5200 nm) and excitation wavelengths in equation (22) are proportional to the particle polarizability
the range 4001000 nm, there are two corrections to this ao . In the k ! 0 (l ! 1) limit where g ! 1, ao is roughly
development that need to be implemented in order to obtain the particle volume (proportional to f3 in equation 18)
realistic results. These are electrodynamics corrections, so these two terms are proportional to particle volume
which in lowest order are called radiation damping and (or area for the first term) divided by l3 (l2 for the
dynamic depolarization, and corrections to the dielectric first term) and hence are negligible. However, at the
constant of the particle that are important when the particle plasmon resonance frequency, jgj 1, so these terms can
is smaller than the conduction electron mean free path. be significant (k 3 ao 1) even when the particle size is only
10% of the wavelength l. In addition, it should be noted
2.3.1 Electrodynamic corrections that the effect of these terms is both to reduce mind and
to shift where it peaks to longer wavelengths. Of course
If one solves Maxwells equations exactly for a spheroid,
it is important to realize that equation (22) is correct only
and then expands the solution in powers of the photon
through the k 3 term in an expansion of the exact solution
wavenumber kk D !/c
, one finds corrections to the static
of Maxwells equations in powers of k, so errors are likely
(k D 0) limit that can be thought of as coming from
whenever k 3 ao is of order unity. In that case one should
two sources. One is radiation damping, which refers to
use more accurate approaches.24,26,27
reduction in the polarization induced by the applied field
To incorporate the electrodynamic corrections into the
that results from emission of radiation by the induced dipole
field enhancement factor R!
, we have multiplied equa-
(i.e. the dipole emits so fast that its size is diminished).
tion (9) by the absolute square of the term in parentheses
The second is dynamic depolarization, which refers to
in equation (22).
interference between radiation emitted at different points on
the surface of the particle. A simple argument for including
these effects has been given by Meier and Wokaun.54 2.3.2 Size-dependent dielectric constants
Their theory is developed specifically for spheres, but it It is known that the dielectric constant of a metal particle
can be extended approximately to spheroids55 through the becomes size dependent whenever the particle is smaller
following argument. than the conduction electron mean-free path. For noble and
When a particle is irradiated by an external field, the alkali metals, this applies to particles that are roughly 20 nm
induced dipole mind in the spheroid is given in the static or less in smallest dimension. A number of models for this
limit by effect have been given, ranging from classical models56,57
mind D ao E 17
that determine the rate of scattering of the conduction elec-
trons from the particle surface, to quantum models34,58 60
where the spheroid polarizability ao is given by
in which the dielectric constant of the particle is calculated
1 gxo f3 using an approximate electronic structure theory (free elec-
ao D 18
3 Q1 x o
tron in a cavity34,46,48 or jellium61 ). The results of these
studies show that the most significant influence of parti- model. The results to be presented were calculated for
cle size as long as the particles are not too small is in spheroids using equations (8) and (9) with the electrody-
broadening of the plasmon width. Other features such as namic corrections from equation (22) and using dielectric
blue- (for alkalis) or red- (for silver) shifting of the reso- constants that incorporate the size-dependent plasmon width
nance frequency and the appearance of multiple absorption based on equations (23) and (24). Except for the size-
bands have been found in the context of gas-phase metal dependent dielectric constants, the Raman enhancements
cluster spectroscopy61 but have been less prominent in SES obtained from this procedure are nearly identical to those
studies. from the accurate electrodynamic calculations of Barber
To describe the plasmon width broadening associated et al.24 and Yang et al.26 (Barber et al. and Yang et al.
with small particles, it is convenient to consider the Drude assume that dielectric constants are size independent.) The
model dielectric constant (equation 6). The width g can be dielectric constants used in these calculations are either
considered to be composed of two contributions, a bulk width from Palik62 or from Hagemann et al.63 The former are
gb and a (size-dependent) surface contribution gS such that more recent, and probably more accurate, but the latter were
used extensively in earlier work.11,25,55,64 In a few cases, we
g D gb C gS 23
give comparisons between the results to indicate the mag-
For a spherical particle, classical models show that gS is nitude of the differences. Figure 1 shows the field enhance-
inversely proportional to particle radius, so we write ment R and Raman enhancement E R as a function of
wavelength for silver spheroids of varying b/a between 1 : 1
vF
gS D A 24
and 5 : 1. The semimajor axis b has in all cases been fixed
R at b D 25 nm, and the Stokes shift ! is 1500 cm1 . Here
where vF is the Fermi velocity (vF /R being the classical rate we have used dielectric constants from Palik, but equivalent
of scattering from the particle surface) and A is a dimen- results from the Hagemann dielectric constants are given in
sionless constant that equals unity according to classical Zeman and Schatz.25 The surrounding dielectric constant eo
theory. The latest jellium estimate of A is 0.7636 but ear- has been taken to be unity.
lier theories based on free-electron models suggested that Figure 1 shows typical plasmon excitation profiles, with
A D 1.16.34 Fortunately, the value of A is close enough to the plasmon frequency shifting to the red as the spheroid
unity in both these expressions that the classical expression becomes more prolate. The Raman enhancements are
can be used without serious error for typical SERS experi- large, as large as 105 , and they are largest for spheroids
ments. The classical models have also been used to obtain having a plasmon wavelength of about 600 nm. These
formulas like equation (24) for other particle shapes. In this Raman enhancements are a factor of about 10 smaller
case, one sets A to unity, and replaces R by an effective than are obtained if the corrections of the previous section
particle length Leff . For a spheroid, Leff is given by60 are not used. These corrections have been left out in
1/2 1
4 x2o C 1 1 1 2 3/2 1 2 2 1 1 1 1
Leff D b sin C xo 1
ln 1 2 xo xo 1
sin C 2 25
3 x2o 1 xo xo xo xo 1
1/2
Note that this expression reduces to Leff D b for b D a (a some of the past work,18 20,22,27 but it is likely that the
sphere), and to Leff D 16a/3p for b a (a needle). results in Figure 1 are more realistic for typical SERS
The main consequence of the increase in plasmon width measurements.
associated with decreased particle size is suppression of In some past calculations, only the field at the tip of the
the field enhancement without changing its frequency or spheroid has been used in calculating enhancements. Again
shape dependence. Of course the theory presented here is this is probably not representative of typical SERS measure-
not appropriate for very small particles (<2 nm), so caution ments. However, recent single-molecule SERS measure-
should be used in applications. ments65 69 have generated significant interest in what is
the largest EM enhancement near the surface of particles
with tips. A variety of estimates are available18 20,22,27,70
2.4 Current status of theory/experiment with the largest EM SERS enhancements being around 1010 .
comparisons Figure 2 shows how the Raman enhancement varies with
polar angle q as one goes around the spheroid. The figure
Let us begin this section by showing some representa- includes results for all five spheroids from Figure 1, with
tive SERS enhancement factors from the electromagnetic the wavelength chosen to be close to the plasmon peak for
500 200 000

E-field enhancement
400 40
300 50
150 000 30
Raman enhancement
200
60
100
0 100 000
300 400 500 600 700 800 20 70
(a) (nm)
150 000
Raman enhancement
50 000
4 :1
5 :1
10
100 000
3:1
0
500 600 700
50 000
2 :1 (nm)
1:1
0
Figure 3. Raman enhancement R for a 4 : 1 Ag spheroid (using
300 400 500 600 700 800 the same parameters as in Figure 1) as a function l for values of
(b) (nm) the semimajor axis parameter b ranging from 10 to 70 nm.
Figure 1. (a) Field enhancement R for Ag versus photon wave-
length l for spheroids having b D 25 nm. The ratios of major to enhancement should be 102 103 larger than the average
minor axis are: 1 : 1, 2 : 1, 3 : 1, 4 : 1 and 5 : 1. (b) Raman enhance- Raman enhancement, so if the average enhancement is 106 ,
ment for ! D 1500 cm1 versus l. Dielectric constants are from the peak enhancement would be 109 .
Palik.62
The results in Figure 1 refer to a fixed semimajor axis
of 25 nm. Figure 3 shows the Raman enhancements that
are obtained if this axis size is allowed to vary in the
range 1070 nm for 4 : 1 spheroids. This figure shows that
Relative field enhancement
20.0
the enhancement rises to a peak at b D 40 nm, with the
5 :1
peak enhancement being close to double what it was for
b D 25 nm. This optimum arises from the competing effects
4 :1 of the electrodynamics corrections (which favor smaller
objects) and the size-dependent dielectric constants (which
10.0
favor larger objects). Note also that larger spheroids pro-
3 :1
duce red-shifted plasmon resonances due to electrodynamic
effects.
2 :1
Figure 4 shows the peak Raman enhancement factor
1 :1
E R as a function of photon wavelength. In this figure
0.0
0.0 30.0 60.0 90.0 the enhancement has been optimized with respect to size
(deg) and shape for each wavelength. Several metals have been
Figure 2. Ratio of local to average enhancement R for the same considered, including the noble metals (for which SERS
Ag spheroids as in Figure 1 as a function of the polar angle q. has been well characterized), the alkali metals, Al and
(The graph is symmetric with respect to q D 90 so only q 90 In (for which there have been a few reports), and other
is plotted.) metals (Ga, Zn, Cd) for which there are no clear-cut SERS
observations. The dielectric constants in this case are from
each spheroid. The quantity plotted is the ratio of the local Hagemann et al.63 For Ag, we find an enhancement factor
field enhancement factor to the average enhancement factor close to 105 for l D 730 nm, which drops slowly with
for the same spheroid, i.e. R!, h
/R!
from equations (9) decreasing l. The corresponding results for Ag using the
and (15). Note that the sphere result (b/a D 1.0) repro- Palik dielectric constants (derived from Figure 3) show
duces the q dependence described earlier. For highly prolate peak enhancement factors that are about a factor of 2
spheroids, i.e. b/a D 4 : 1 and 5 : 1, the field enhancement larger, and which peak at about 600 nm. The Cu and Au
at the tip is very large, 1025 times larger than the results in Figure 3 show large enhancements (>104 ) for
average enhancement. This means that the peak Raman wavelengths longer than 700 nm, then a dramatic drop-off at
5.50 electrochemical experiments,2,3 and about 105 for ultrahigh

Li
vacuum measurements.71 Although larger enhancements
5.00
Al
Ag (105 ) are obtained from the Palik dielectric constants, there
Cu
4.50 is still a gap between electromagnetic theory and experi-
Log10 (E R )
In
ment, presumably due to the chemical enhancement mech-
4.00 Na anism. For Cu and Au there is a similar gap between theory
3.50 Zn and experiment, although the drop in intensity at 600 nm
Au
3.00 Ga in Figure 4 matches observations accurately.74 Experimen-
tal information concerning the magnitudes and frequency
2.50 Cd
dependence of the enhancements for the alkalis Al and In
2.00 is not extensive. Although the available results are roughly
350 400 450 500 550 600 650 700 750 consistent with Figure 4, a chemical enhancement factor
(nm) may also be present.
Figure 4. Fully optimized Raman enhancement (optimized with One experiment that provides an important test for the
respect to size and shape) versus l. Adapted from Zeman and electromagnetic model is the observation of SERS by
Schatz,25 Figure 22. The dielectric constants are from Hagemann Jennings et al.75 at an excitation wavelength of 1.06 m
et al.63
(1.17 eV) for molecules adsorbed on Au metal island films
and CaF2 roughened Au films. The estimated enhancement
shorter wavelengths corresponding to the onset of interband factor was 50100. This extends the range of SERS
excitation, which greatly increases the imaginary part of the observations to much longer wavelengths than in previous
dielectric constant. measurements (for which l < 650 nm was the typical
Although Figure 4 serves as a useful guide in quantifying limit). Qualitatively one expects a reduction in enhancement
the EM for a variety of metals, the detailed comparison with compared with the visible, since only needle-shaped or
experiment must be done with some caution as the exper- pancake-shaped particles have plasmon resonances in the
iments have almost always been done with an unknown near-infrared (NIR), and most of the molecules on these
or poorly characterized distribution of roughness features. needles or pancakes, are located away from the tips
One exception refers to the well-known experiments by where enhancements are relatively small (as in Figure 2).
Liao and co-workers71 on microlithographically prepared However, there are no quantitative comparisons between
posts. The wavelength dependence of the SERS inten- theory and experiment for this case.
sities for both Ag and Au posts has been successfully The determination of adsorbate geometry and orienta-
modeled71 using a spheroid electromagnetic model that is tion is one of the most important features of the chemical
almost identical to that described above. (The dynamic information content of SERS. An early example demon-
depolarization term was not included.) In this modeling, the strating that relative intensity pattern in SERS was sensitive
shape of the spheroid was adjusted so that the plasmon res- to adsorbate orientation was given by Allen et al.76 for the
onance wavelength matched experiment, a procedure that case of 2-, 3- and 4-cyanopyridine. Moskovits and Suh77
is necessary because the shape and size of the particle, have given a general discussion of this, and Creighton78
as well as the underlying substrate, cannot be controlled has provided an extensive review of surface selection rules
accurately enough to develop a parameter-free comparison for SERS. There are many applications of this sort in the
between theory and experiment. Recent studies using litho- SERS literature for a large number of different molecules,
graphic methods with colloidal crystal masks have produced including phthalizine, substituted benzenes and carboxylic
triangular-shaped particles whose extinction spectra,72 as acids by Moskovits and co-workers,77,79,80 monosubstituted
well as substrate and solvent effects,73 may be explained benzenes by Weaver and co-workers,81 and a variety of
quantitatively using parameter-free electrodynamics calcu- alkanethiols by Pemberton et al.82 The conclusions regard-
lations. Although it is not yet clear how well EM theory ing adsorbate geometry and orientation in these studies, by
will do in describing SERS on these particles (Yang et al.26 and large, make good chemical sense.
describe an initial attempt in this direction), this at least In spite of these successes, a significant number of
provides a benchmark for EM estimates. papers continue to be published in which the residual
One issue concerning Figure 4 that deserves further uncertainty over the relative contributions of the electro-
note is the comparison of absolute enhancement fac- magnetic and chemical enhancement mechanisms in SERS
tors with experiment. For Ag at 515 nm, Figure 3 indi- is interpreted by the authors to preclude the capability
cates an optimized enhancement factor of 4 104 while of determining adsorbate geometry and orientation. This
experimental enhancements are typically 13
106 for is reflected in the wider surface science community by a
certain reluctance or nervousness to embrace such con- 7

clusions from SERS studies. Consequently, a few remarks 6
concerning the surface selection rules of SERS are appro- 5
priate here.
4
Provided that one uses a SERS-active surface for which
Raman intensity (counts s 1 W1)

3
the EM dominates, one can indeed use the relative intensity
pattern to determine adsorbate geometry and orientation. 2 102
To draw such conclusions in a quantitative manner, one 1
needs at make an estimate of the relative magnitudes of 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
the electric fields parallel and perpendicular to the surface, (a)
something that is easily extracted from the spheroid model.
6
In addition, one needs at least qualitative information about
the polarizability derivative tensor of the molecule and 5 106
modes being studied. This information is now readily avail- 4
able from ab initio quantum chemistry calculations. Thus it
3
is possible to predict the relative SERS intensity pattern as a
function of adsorbate geometry and orientation. A detailed 2
example of this for the case of trans-1,2(4-pyridyl)ethylene 1
(BPE) adsorbed onto silver film over nanosphere electrodes
0
has been published by Yang et al.83 For usage in a more 0.0 0.5 1.0 1.5 2.0 2.5 3.0
qualitative sense, one can simply note that since electric (b) Coverage (ML)
field components are largest perpendicular to the surface,
Figure 5. Coverage dependence of the 683-cm1 line of CoPc:
modes involving large changes in the perpendicular polar- (a) CoPc on a thick Cr (unenhancing substrate) using laser power
izability are most strongly enhanced. of 300 mW; (b) CoPc on a roughened Ag substrate using 10 nW
power.
3 EXTENSIONS OF THE The coverage dependence of SERS and SERRS can be

ELECTROMAGNETIC MODEL explained using a simple extension of the spheroid model
presented previously in which the spheroid is assumed
3.1 Effect of adsorbed layers on intensities to be coated with a layer whose dielectric constant is
denoted el . If the layer is assumed to be defined as the
The electromagnetic model as developed to this point con- region between two confocal spheroidal surfaces, then the
siders a metal particle of dielectric constant ei embedded solution of LaPlaces equation is very similar to that in
in a homogeneous medium of dielectric constant eo . Sev- equation (9), but with R!
multiplied by an additional
eral experiments done since 1984 have demonstrated that an layer-dependent factor. Surface coverage in this model may
adsorbed layer can significantly alter enhancements (chang- either be described by varying the layer thickness, or by
ing both magnitude and frequency dependence) even when using an effective medium approximation to express the
present in relatively low coverages (0.1 monolayers). layer dielectric constant el in terms of the adsorbate and
Both theory and experiments concerning the effect of solvent dielectric constants.
adsorbed layers on SERS have been presented by Murray An alternative theory of coverage effects in SERS and
and Bodoff,84,85 Murray,37 and Waseleski et al.36 and a SERRS has been developed by Murray and Bodoff84,85 and
related treatment of surface-enhanced resonance Raman by Zeman et al.35 based on the explicit treatment of the
scattering (SERRS) has been given by Zeman et al.35 In interaction of adsorbed molecules with the applied field,
the Murray and Bodoff experiments, the SERS intensity with the metal particle, and with each other. In this theory,
for CN adsorbed onto silver island films was found to the molecules and the metal particle are replaced by polariz-
vary nonlinearly with coverage, with a peak at about 0.25 able dipoles mi . The magnitudes of these dipoles are deter-
monolayers, and falling intensities at higher coverages. In mined by solving the usual coupled dipole (CD) equations:
the Zeman et al. experiments, a similar dependence on
coverage was found for SERRS of cobalt phthalocyanine
mi D ai Eo C Mij mj 26
(CoPc) on roughened silver films, with a peak at 0.07

j6Di
monolayer. Figure 5 shows the latter results, contrasting
the SERRS intensities with the corresponding spectra on where Eo is the applied field, ai is the polarizability of the
an unenhancing substrate. ith dipole, and Mij is the dipole interaction between the ith
and jth dipoles: Calculations using the layered spheroid model for the
same parameters as in Figure 5 give similar results, indi-
eikrij cating that approximations involved in both models are not
Mij mj D k 2 rij mj
rij 3
rij important to the final results. The layered model has been
used by Murray37 to describe SERS from particles coated
2 1 ik with an overlayer upon which molecules are adsorbed. In
C [3rij rij mj
rij mj ] 5
4 eikrij
rij r ij this case the molecules experience the field induced at the
27
outer surface of the coating. Detailed studies of silver par-
For molecules on a flat surface84,85 only dipole coupling ticles coated with a layer of gold, nickel or insulators were
between molecules is included, and equation (26) may be carried out,37 and these demonstrated that substantial damp-
solved analytically (in the static, k ! 0, limit) as long as the ing and shifting of the plasmon resonance can be induced
molecules occupy sites on a square lattice. For molecules on by the layer.
a metal particle, equation (26) must be solved numerically. Although these results all seem very encouraging, we
Usually the interactions between dipoles are too strong to should point out that layer and dipole models of solvent
allow for its solution by iteration, but as long as the number effects on the extinction spectra of silver nanoparticles often
of molecules is not too large, equation (26) may be solved seriously overestimate the shift of the plasmon wavelength
by direct inversion for assumed locations of the molecules. as solvent dielectric constant is increased.73 This suggests
By averaging over a sufficient number of molecular loca- that chemical interactions between nanoparticle and solvent
tions that are consistent with the desired surface coverage, could cause the electromagnetic environment near the
the realistic simulation of experiments is possible. surface to be significantly different from the simple model.
Figure 6 shows the CD predictions for SERRS intensity This indicates that caution should be applied in interpreting
versus coverage for CoPc on silver. Several choices of the significance of the results in Figures 5 and 6.
spheroid shape are included, as are two choices of Stokes
shift. The figure shows that the more prolate the metal
particle, the lower the coverage where intensity peaks. At 3.2 Nonlinear surface optical processes
the observation wavelength (647.1 nm), Figure 1 indicates
that 4 : 1 and 5 : 1 spheroids should be dominant. Figure 6 As mentioned in the introduction, the electromagnetic
then indicates that intensities should peak at about 0.1 model can be used with no additional computational
monolayer, which is consistent with results in Figure 5. effort to predict enhancement factors for surface-enhanced
nonlinear spectroscopies, including SESHG and SEHRS.
400 This assumes that the nonlinearity is determined by the field
350 (a) 3 :1 or field gradient at the interface, which is a reasonable but
300 not completely correct assumption. At present, the quality
250 of theory/experiment comparisons is not high enough to
4 :1
ICD
200
quantify the error involved, but some theoretical estimates
150
have been provided for SESHG.38 41
100 5:1
2 :1 The relevant enhancement factors (analogous to equa-
50 6:1
0
tion 8) are given by
0.0 0.1 0.2 0.3 0.4 0.5 0.6
SESHG!0 D 2!
: E SHG D R2!
R2 !
28
350
300
(b) SEHRS!0 D 2!
: E HRS D R2!
R2 !
29
250 3 :1
where R denotes the field enhancement factor from equa-
200
ICD
4 :1 tion (9).
150
Now let us consider specific examples in which E SHG has
100
5:1 been compared with experiment. We restrict our discussion
50 6:1 2 :1 to experiments where isolated particle models can be used.
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 The next section provides discussion of other types of
Coverage (ML) surfaces like gratings. Table 1 presents a comparison of the
Figure 6. CD model Raman intensities ICD for Ag spheroids SHG enhancement factors for several metals (Ag, Au, Cu,
having b D 12 nm versus coverage for h! D 1.916 eV: (a) ! D Li, Na, Al, Ga and In) at h! D 1.17 eV. The two theoretical
683 cm1 ; (b) ! D 1544 cm1 . calculations refer to averages over spheroid sizes and shapes
Table 1. SHG enhancement factors.
Secondharmonic enhancement
10 000 8 :1
5 :1
Metal E SHG E SHG (normalized to Ag) 4 :1
Theorya Theorya Theoryb Exptb 1000 3 :1
1 :1
Ag 1.5 106 1 1 1 100
Au 8.4 104 0.06 0.10 0.008
Cu 8.3 104 0.06 0.10 0.01
Li 4.6 104 0.03 10
Na 3.3 105 0.22 0.16 0.36
Al 3.6 105 0.24 0.40 4.6 1
Ga 9.5 106 6.3 4.0 27.0 500 1000 1500 2000 2500 3000 3500
In 1.2 104 0.01 0.04 0.004 (nm)
a Taken from Zeman and Schatz.25 Figure 7. Predicted SESHG enhancement factor S as a function
b Taken from Boyd et al.87 of input photon wavelength l for a Pt spheroid having b D 200 nm
and b/a D 8 : 1, 5 : 1, 4 : 1, 3 : 1 and 1 : 1.
such that the volume is in the range 105 l3 to 104 l3 . The
averages used in the two studies25,86,87 were not the same, determining SEHRS enhancements involves determining
but the resulting enhancements are similar. The results from the terms in the expression88
Zeman and Schatz25 are given both as absolute and relative
enhancements (normalized to Ag), while the other two 1
Isrf
HR Isrf
R Isrf
HR Ibulk
HR
enhancements are only available as relative enhancements. D 30
Ibulk
HR Ibulk
R Isrf
R Ibulk
R
The comparison between theory and experiment is gen-
erally quite good for the relative enhancements, with Ga
The first term on the right-hand side is the well-known 106
showing the largest value for each column in Table 1. Only
SERS enhancement factor. The second term is the ratio of
the result for Al is in serious error, otherwise the variation
hyper-Raman to Raman intensities on the surface, which
with metal is correct. The absolute enhancement factors
is a quantity that can be obtained straightforwardly from
in Table 1 are generally larger than the SERS factors in
experiment. The third term is the corresponding bulk ratio.
Figure 4, but not by much. Ga, however, shows much larger
This is hard to obtain from experiment (although estimates
SHG than SERS.
do exist),88 but can be obtained from molecular orbital
Figure 7 presents the calculated SHG enhancement factor
for Pt spheroids42 over a range of photon energies. Although calculations as it involves isolated molecule properties. The
Pt does not show significant electromagnetic enhancement best estimate of this ratio is from Yang and Schatz,89 who
effects in the visible, at l D 1060 nm enhancements of a used HartreeFock methods with a basis set that includes
few hundred are possible. This result is consistent with polarization and diffuse functions. To calibrate their results,
the observation42 of SESHG from rippled Pt microstruc- they calculated Raman spectra from benzene, and were
tures where the ripple size is comparable to that used in the able to match high-quality gas-phase measurements. In
electromagnetic modeling. This demonstrates the potential addition, their bulk hyper-Raman spectra were consistent
broad applicability of SESHG, as many metals show dielec- with available measurements.
tric behavior similar to Pt (i.e. small SHG enhancements in The SEHRS enhancement obtained by Yang and Schatz89
the visible, but significant enhancements in the NIR). In was in the range 1011 1012 . This is much larger than is
fact, for Pt, Figure 7 indicates that SESHG enhancements predicted by the electromagnetic model (for which the SHG
are even larger for l > 2000 nm. enhancement factor for Ag in Table 1 provides a realistic
There has been relatively little theoretical work done estimate). The likely explanation for the discrepancy is
on SEHRS, but Golab et al.88 and Yang and Schatz89 that chemical enhancement mechanisms are much more
have provided important results concerning the SEHRS important for SEHRS than they are for SERS. This is
enhancement factor for pyridine and BPE adsorbed onto in accord with the strong sensitivity that the molecular
roughened silver. This enhancement factor is difficult to hyperpolarizability shows to the presence of excited states
measure directly because when SEHRS is observable, the that have a significant change in dipole moment relative to
corresponding bulk hyper-Raman scattering measurement the ground state. Such states are not important for gas-phase
requires extreme laser powers that are difficult to quantify. pyridine, but they are easily produced when molecules
(See, however, the work of Johnson and co-workers90,91 are adsorbed on surfaces via charge transfer between the
for one attempted measurement.) An indirect method for molecule and surface.
4 ELECTROMAGNETIC MECHANISM and Schaich and Mendoza103 and references therein.) For
FOR MANY COUPLED PARTICLES Ag gratings, the largest measured enhancement factor
found so far is 7000.104 This result is in good agreement
AND FOR GRATINGS with theoretical predictions. An optimum groove depth
is also found that is similar to that found for SERS.105
The theoretical development to this point has considered
Recently it has been demonstrated both theoretically and
only isolated particles. In most cases, experiments involve
experimentally that SESHG on gratings can be selectively
more complex surfaces so it is important to develop
enhanced by using a grating that has two spatial fre-
electrodynamics theory for more general applications. This
quencies, one that couples the incident light to plasmon
section is concerned with the problem of modeling many
excitation at the fundamental frequency, and the other
coupled particles, or extended structures such as gratings
that couples the second-harmonic plasmon to the emitted
and randomly roughened continuous surfaces. As with
photon.106,107
several earlier parts of this review, substantial work on this
topic was carried out before 1985, and has been reviewed
already. Our discussion of this work will therefore be
representative rather than exhaustive. 5 CONCLUSIONS AND FURTHER WORK
The simplest coupled particle system is just two coupled
spheres. LaPlaces equation was solved for this problem This review has dealt mostly with features of the elec-
by Aravind et al.92 and the results were used to study tromagnetic model that correlates well with experiment,
the field enhancement in the region between the spheres. but it is important to mention features where the relation
This enhancement can become very large under certain between theory and experiment merits further work. We will
circumstances. leave out of this discussion problems such as the SEHRS
The next more complicated problem involves treat- enhancement factor where it appears that the primary dis-
ing many coupled particles, either in a regular array, as crepancy between theory and experiment is associated with
was done elsewhere93 96 for spheres, or in a random chemical contributions to the enhancement.
array, as was done elsewhere31,32 for spheroids. In either Perhaps the key weakness in the comparison between
case, densely packed arrays exhibit substantially different electromagnetic theory calculations and SES measurements
plasmon resonance features that are associated with iso- lies in the mismatch between the surface modeled and that
lated particles, including multiple resonances that can be measured. Most measurements refer to randomly rough-
either red- or blue-shifted relative to the single particle ened surfaces that have important features like pores and
resonances.32,93,94 cracks that are rarely included in calculations. Only a few
There has been substantial recent work concerned with experiments on well-characterized surfaces (periodic arrays
particle coupling effects in the context of extinction spectra of posts, gratings) have been performed. Most calculations
that goes beyond the simple dipole coupling theories.95,96 have modeled isolated particles or perfect gratings, with
This work shows that the dipole approximation is adequate little attention given to experimental reality. Because of
for small particles (10 nm) provided that the particles are this, the comparison between theory and experiment is still
separated by a particle radius or more. very imprecise. This makes it difficult to explain why two
Gratings are another type of extended surface that rough surfaces prepared with slightly different roughening
has received considerable attention in applications of the procedures may show orders of magnitude enhancement
EM. There have been several theoretical studies of SERS differences (as sometimes happens). Also, it is difficult to
on gratings (see Garcia et al.97 and references therein), quantify the fraction of molecules that experience substan-
and the calculated maximum enhancements are similar in tial SERS enhancements.
magnitude to what has been calculated for spheroids that A second problem that is connected closely to the
have comparable size features to the grating corrugation. first refers to the complexity of accurate solutions of
The maximum SERS enhancement occurs for incident Maxwells equations for arbitrarily irregular surfaces that
angles and grating structures where reflectivity minimizes, have roughness comparable in size to the wavelength of
as that leads to the maximum conversion of incident photon light. LaPlaces equation has often been used, but it can be
energy to plasmon energy. inaccurate. Numerical solutions are possible but may not
Gratings and other types of surface roughness have also be worth the significant computational expense for poorly
been considered in studies of SESHG. The literature on characterized surfaces. A better solution would be to have
this topic is quite extensive, including several electrody- more experiments done on structurally simple surfaces.
namics studies.98 101 There has also been extensive work A third problem is that dielectric constants are not always
on calculating nonlinear susceptibilities (see Sipe et al.102 available at frequencies where they are needed. Even for
Ag, Au, and Cu, where dielectric constants are generally 8. M. Kerker, Acct. Chem. Res., 17, 271 (1984).
available, there are significant differences between results 9. M. Kerker (ed.), Selected Papers on Surface-enhanced
generated with different literature values.64 Raman Scattering, SPIE Optical Engineering Press, Belling-
Finally, it should be noted that there is a significant ham, WA (1990).
need for experimental studies in which more than one SES 10. G.C. Schatz, Acct. Chem. Res., 17, 370 (1984).
measurement is done on the same surface. The comparison 11. H. Metiu, Prog. Surf. Sci., 17, 153 (1984).
of SERS and SESHG would be especially interesting 12. S. Efrima, Surface-enhanced Raman Scattering (SERS), in
in testing the EM, as competing (chemical) mechanisms Modern Aspects in Electrochemistry, eds J.OM. Bockris,
influence SERS but not SESHG. A grand challenge in R. White and B.E. Conway, Plenum, New York, 253369,
this area would be to measure the extinction and Rayleigh Vol. 16 (1986).
scattering spectrum, along with the SERS, SESHG and 13. A. Campion and P. Kambhampati, Chem. Soc. Rev., 27, 241
(1998).
SEHRS excitation profiles for the same adsorbate on a well-
characterized nanofabricated surface. 14. E.J. Liang and W. Kiefer, J. Raman Spectrosc., 27, 879
(1996).
15. R.G. Greenler and T.L. Schlager, Spectrochim. Acta, A29,
193 (1973).
ACKNOWLEDGMENTS
16. M. Moskovits, J. Chem. Phys., 69, 4159 (1978).
We would like to acknowledge several present and former 17. S.L. McCall, P.M. Platzman and P.A. Wolff, Phys. Lett.,
77A, 381 (1980).
members of the Schatz and Van Duyne groups who have
contributed to the theory/experiment collaboration that is 18. D.-S. Wang, H. Chew and M. Kerker, Appl. Opt., 19, 2256
(1980).
described in this review. This research was supported
19. D.-S. Wang and M. Kerker, Phys. Rev., B24, 1777 (1981).
by ARO Grant DAAG559710133, and by the MRSEC
program of the NSF (Grant DMR-0076097) at the Materials 20. M. Kerker, D.-S. Wang and H. Chew, Appl. Opt., 19, 4159
(1980).
Research Center of Northwestern University.
21. M. Kerker, J. Coll. Interface Sci., 118, 417 (1987).
22. J. Gersten and A. Nitzan, J. Chem. Phys., 73, 3023 (1980).
ABBREVIATIONS AND ACRONYMS 23. F.J. Adrian, Chem. Phys. Lett., 78, 45 (1981).
24. P.W. Barber, R.K. Chang and H. Massoudi, Phys. Rev.,
BPE trans-1,2(4-pyridyl)Ethylene B27, 7251 (1983).
CD Coupled Dipole 25. E.J. Zeman and G.C. Schatz, J. Phys. Chem., 91, 634 (1987).
CoPc Cobalt Phthalocyanine 26. W.-H. Yang, G.C. Schatz and R.P. Van Duyne, J. Chem.
EM Electromagnetic Mechanism Phys., 103, 869 (1995).
SEHRS Surface-enhanced Hyper-Raman Spectroscopy 27. J.P. Kottmann, O.J.F. Martin, D.R. Smith and S. Schultz,
SES Surface-enhanced Spectroscopy Optics Express, 6, 213 (2000).
SESHG Surface-enhanced Second-harmonic Generation 28. M. Arnold, P. Bussemer, K. Hehl, H. Brabhorn and A. Otto,
SPIRS Surface Phonon-induced Raman Scattering J. Mod. Opt., 39, 2329 (1992).
29. I. Baltog, N. Primeau, R. Reinisch and J.L. Coutaz, Appl.
Phys. Lett., 66, 1187 (1995).
REFERENCES 30. N. Garcia, Opt. Commun., 45, 307 (1983).
31. U. Laor and G.C. Schatz, J. Chem. Phys., 76, 2888 (1982).
1. M. Fleischman, P.J. Hendra and A.J. McQuillan, Chem. 32. U. Laor and G.C. Schatz, Chem. Phys. Lett., 82, 566 (1981).
Phys. Lett., 26, 163 (1974).
33. J.A. Sanchez-Gil and J.V. Garcia-Ramos, J. Chem. Phys.,
2. D.L. Jeanmaire and R.P. Van Duyne, J. Electroanal. Chem., 108, 317 (1998).
84, 1 (1977).
34. W.A. Kraus and G.C. Schatz, J. Chem. Phys., 79, 6130
3. M.G. Albrecht and J.A. Crieghton, J. Amer. Chem. Soc., 99, (1983).
5215 (1977).
35. E.J. Zeman, K.T. Carron, G.C. Schatz and R.P. Van Duyne,
4. M. Moskovits, Rev. Mod. Phys., 57, 783 (1985). J. Chem. Phys., 87, 4189 (1987).
5. A. Otto, Top. Appl. Phys., 54, 289 (1984). 36. S.A. Waseleski, S. Zou and M.J. Weaver, Appl. Spectrosc.,
6. H. Metiu, in Surface Enhanced Raman Scattering, eds 54, 761 (2000).
T.E. Furtak and R.K. Chang, Plenum, New York (1981). 37. C.A. Murray, J. Opt. Soc., B2, 1330 (1985).
7. H. Metiu and P. Das, Annu. Rev. Phys. Chem., 35, 507 38. G.S. Agarwal and S.S. Jha, Solid State Commun., 41, 499
(1984). (1982).
39. X.M. Hua and J.I. Gersten, Phys. Rev. B, 33, 3756 (1986). 69. H. Xu, E.J. Bjerneld, M. Kall and L. Borjesson, Phys. Rev.
40. D. Ostling, P. Stampfli and K.H. Bennemann, Z. Phys. D, Lett., 83, 4357 (1999).
28, 169 (1993). 70. T. Jensen, L. Kelly, A. Lazarides and G.C. Schatz, J. Clus-
41. J.I. Dadap, J. Shan, K.B. Eisenthal and T.F. Heinz, Phys. ter Sci., 10, 295 (1999).
Rev. Lett., 83, 4045 (1999). 71. P.F. Liao and M.B. Stern, Opt. Lett., 7, 483 (1982).
42. K.L. Haller, L.A. Bumm, R.I. Altkorn, E.J. Zeman, G.C. 72. T.R. Jensen, G.C. Schatz and R.P. Van Duyne, J. Phys.
Schatz and R.P. Van Duyne, J. Chem. Phys., 90, 1237 Chem. B, 103, 2394 (1999).
(1989).
73. T.R. Jensen, M.L. Duval, K.L. Kelly, A. Lazarides, G.C.
43. A. Otto, J. Raman Spectrosc., 22, 743 (1991). Schatz and R.P. Van Duyne, J. Phys. Chem., 103, 9846
44. A. Otto, Surf. Sci., 75, L3926 (1978). (1999).
45. P.K.K. Pandey and G.C. Schatz, J. Chem. Phys., 80, 2959 74. C.S. Allen, G.C. Schatz and R.P. Van Duyne, Chem. Phys.
(1984). Lett., 75, 201 (1980).
46. T. Maniv and H. Metiu, Chem. Phys. Lett., 79, 79 (1981). 75. C.A. Jennings, G.J. Kovacs and R. Aroca, J. Phys. Chem.,
47. S.L. McCall and P.M. Platzman, Phys. Rev. B, 22, 1660 96, 1340 (1992).
(1980). 76. C.S. Allen and R.P. Van Duyne, Chem. Phys. Lett., 63, 455
48. S.S. Jha, J.R. Kirtley and J.C. Tsang, Phys. Rev. B, 22, 3973 (1979).
(1980). 77. M. Moskovits and J.S. Suh, J. Phys. Chem., 88, 5526 (1984).
49. S. Mamiche-Afara and M.J. Dignam, J. Chem. Phys., 90, 78. J.A. Creighton, in Spectroscopy of Surfaces, eds R.J.H.
3861 (1989). Clark and R.E. Hesters, John Wiley & Sons, New York,
50. T. Maniv and H. Metiu, Surf. Sci., 101, 399 (1980). 3789, Vol. 16 (1988).
51. P.J. Feibelman, Phys. Rev. B, 22, 3654 (1980). 79. M. Moskovits and J.S. Suh, J. Phys. Chem., 88, 1293 (1984).
52. F.J. Adrian, J. Chem. Phys., 77, 5302 (1982). 80. M. Moskovits, D.P. DiLella and K.J. Maynard, Langmuir,
53. C.R. Bohren and D.F. Huffman, Absorption and Scattering 4, 67 (1988).
of Light by Small Particles, John Wiley & Sons, New York 81. X. Gao, J.P. Davies and M.J. Weaver, J. Phys. Chem., 94,
(1983). 6858 (1990).
54. M. Meier and A. Wokaun, Opt. Lett., 8, 581 (1983). 82. J.E. Pemberton, M.A. Bryant, R.L. Sobocinski and S.L. Joa,
55. E. Zeman and G.C. Schatz, in Dynamics on Surfaces, J. Phys. Chem., 96, 3776 (1992).
Proceedings of the 17th Jerusalem Symposium, eds 83. W.H. Yang, J.C. Hulteen, G.C. Schatz and R.P. Van Duyne,
B. Pullman, J. Jortner, B. Gerber and A. Nitzan, Reidel, J. Chem. Phys., 104, 4313 (1996).
Dordrecht, 413424 (1984).
84. C.A. Murray and S. Bodoff, Phys. Rev. Lett., 52, 2273
56. U. Kreibig and C.v. Fragstein, Z. Phys., 224, 307 (1969). (1984).
57. J. Euler, Z. Phys., 137, 318 (1954). 85. C.A. Murray and S. Bodoff, Phys. Rev. B, 32, 671 (1985).
58. L. Genzel, T.R. Martin and U. Kreibig Z. Phys., 21, 339 86. C.K. Chen, A.R.B. de Castro and Y.R. Shen, Phys. Rev.
(1975). Lett., 46, 145 (1981).
59. M. Xu and M.J. Dignam, J. Chem. Phys., 96, 3370 (1992).
87. G.T. Boyd, T. Rasing, J.R.R. Leite and Y.R. Shen, Phys.
60. W.A. Kraus and G.C. Schatz, Chem. Phys. Lett., 99, 353 Rev. B, 30, 519 (1984).
(1983).
88. J.T. Golab, J.R. Sprague, K.T. Carron, G.C. Schatz and R.P.
61. A. Liebsch and W.L. Schaich, Phys. Rev. B: Condens. Mat- Van Duyne, J. Chem. Phys., 88, 7942 (1988).
ter, 52, 14219 (1995).
89. W.H. Yang and G.C. Schatz, J. Chem. Phys., 97, 3831
62. D.W. Lynch and W.R. Hunter, Comments on the Optical (1992).
Constants of Metals and an Introduction to the Data for Sev-
eral Metals, in Handbook of Optical Constants of Solids, 90. J.P. Nedderson, S.A. Mounter, J.M. Bostick and C.K.
ed E.D. Palik, Academic Press, New York, 350356 (1985). Johnson, J. Chem. Phys., 90, 4719 (1989).
63. H.J. Hagemann, W. Gudat and C. Kunz, DESY Report No. 91. C.K. Johnson and S.A. Soper, J. Phys. Chem., 93, 7281
SR-74/7, May (1984). (1989).
64. M. Kerker, J. Opt. Soc. B, 2, 1327 (1985). 92. P.K. Aravind, A. Nitzan and H. Metiu, Surf. Sci., 110, 189
(1981).
65. S. Nie and S.R. Emory, Science, 275, 1102 (1997).
93. M. Inoue and K. Ohtaka, J. Phys. Soc. J., 52, 145 (1983).
66. K. Kneipp, H. Kneipp, G. Deinum, I. Itzkan, R.R. Dasari
and M.S. Feld, Appl. Spectrosc., 52, 175 (1998). 94. M. Inoue and K. Ohtaka, J. Phys. Soc. J., 52, 3853 (1983).
67. A.M. Michaels, M. Nirmal and L.E. Brus, J. Am. Chem. 95. A.A. Lazarides and G.C. Schatz, J. Phys. Chem., 104, 460
Soc., 121, 9932 (1999). (2000).
68. K. Kneipp, H. Kneipp, I. Itzkan, R.R. Dasari and M.S. Feld, 96. A.A. Lazarides and G.C. Schatz, J. Chem. Phys., 112, 2987
Chem. Rev., 99, 2957 (1999). (2000).
97. N. Garcia, G. Diaz, J.J. Sainz and C. Ocal, Surf. Sci., 143, 103. W.L. Schaich and B.S. Mendoza, Phys. Rev. B, 45, 14 279
342 (1984). (1992).
98. G.S. Agarwal and S.S. Jha, Phys. Rev. B, 26, 482 (1982). 104. J.C. Quail and H.J. Simons, J. Opt. Soc. Am., B5, 325
99. R. Reinisch and M. Neviere, Phys. Rev. B, 28, 1870 (1988).
(1983). 105. J.L. Coutaz, M. Neviere, E. Pic and R. Reinisch, Phys. Rev.
100. G.A. Farias and A.A. Maradudin, Phys. Rev. B, 30, 3002 B, 32, 2227 (1985).
(1984). 106. A.C.R. Pipino and G.C. Schatz, Phys. Rev. B, 49, 8320
101. R.T. Deck and R.K. Grygier, Appl. Opt., 23, 3202 (1984). (1994).
102. J.E. Sipe, V.C.Y. So, M. Fukui and G.I. Stegeman, Phys. 107. A.C.R. Pipino, R.P. Van Duyne and G.C. Schatz, Phys. Rev.
Rev. B, 21, 4389 (1980) B, 53, 4162 (1996).
Surface-enhanced Raman Scattering
W. Ewen Smith and Caroline Rodger

University of Strathclyde, Glasgow, UK
1 INTRODUCTION adsorption to specific metal surfaces has limited the devel-

opment of the technique.9 11 Surface photolysis and fluo-
Surface enhanced Raman spectroscopy (SERS) was first rescence have proved to be a problem in some experiments.
demonstrated by Fleischmann et al. in 1974.1 They were Surface selection rules are available. In their simplest
using Raman spectroscopy to study the adsorption of form, and assuming no specific symmetry selection rules,
pyridine at a silver electrode and strong Raman scatter- the most intense bands are predicted as those from vibra-
ing was observed when the surface was electrochemically tions which induce a polarization of the adsorbate electron
roughened. Jeanmaire and Van Duyne2 and Albrecht and cloud perpendicular to the metal surface. Reviews of the
Creighton3 reported in 1977 that the Raman scattering use of surface selection rules are available.12,13 This infor-
from pyridine adsorbed on a roughened metal surface was mation can be applied qualitatively to give an approximate
enhanced by a factor of 106 compared to the equivalent indication of the angle a molecule subtends to the surface.
scattering for the same concentration of pyridine in solution. These advantages enable unique insights to be made into
This enhancement phenomenon was called SERS and it a number of surface problems. However, the technique does
depended upon the surface roughness and effective adsorp- have significant limitations:
tion of the analyte to that surface. Surface enhancement
was mainly observed from silver, gold and copper surfaces, 1. The analyte requires to be in close proximity to a
although aluminum, lithium and sodium also gave enhance- suitably roughened surface, which can result in the
ment to a lesser extent. The surface enhancement from other involvement of complex surface chemistry.
metals has been investigated.4 A book containing the key 2. Disappointing reproducibility is obtained when SERS
papers which report the development of this technique is is used quantitatively, as the signal is extremely sen-
available.5 sitive to a number of factors including any change in
Many of the advantages of normal Raman scattering adsorbate orientation at the metal surface, the extent of
are applicable to this technique. These include molecu- adsorption and the nature of the surface roughness.
larly specific vibrational spectra, simple versatile sampling
3. As SERS gives an enhancement of up to 106 and has
and ready determination of analytes in air, under vacuum
specific selection rules, it can be difficult to relate the
and in water. However, with SERS, increased sensitivity
spectra to the Raman scattering from the parent species
is obtained and much wider concentration ranges can be
before adsorption. It is possible that a contaminant
studied. Detection limits are considerably lower (down to
within a sample or a species formed by surface pho-
109 M) than those for resonance Raman (RR) scattering.6,7
tochemistry may be preferentially enhanced. In these
In addition to providing vibrational information, electronic
cases, the spectrum becomes complex and difficult to
information is available through surface enhanced exci-
interpret.
tation profiles.8 However, in practice the sensitivity of
these experiments to interference and the requirement for Despite these difficulties, SERS has become an increas-
ingly popular analytical tool which has been applied to
John Wiley & Sons Ltd, 2002. numerous fields as diverse as catalysis and forensic science.
2 MECHANISMS OF SURFACE explain why the enhancement factor of the first adsorbate
ENHANCEMENT layer is much greater than subsequent layers.
Basically, the enhancement experienced from charge
The SERS phenomenon arises from an interaction between transfer results when molecules physisorb or chemisorb
an adsorbate and the surface plasmons which can be con- directly onto the roughened surface, forming an adsor-
sidered to be a wave of electrons present on the surface batemetal complex. If chemisorption occurs, the molecu-
of the metal substrate. The importance of adsorption of the lar orbitals of the adsorbate are broadened by an interaction
analyte to the metal surface has been highlighted.14,15 with the conduction bands of the metal surface. The result
The nature of the mechanism(s) which produce the sur- is ready transfer of electrons and excitation from the metal
face enhancement effect is unclear and remains the focus to the adsorbate and back. As the mechanism depends on a
of debate.16,17 Most researchers believe that much of the metaladsorbate bond, the enhancement experienced dra-
enhancement is due to an electromagnetic mechanism. matically reduces as the distance of the adsorbate from
However another mechanism, charge transfer enhancement, the metal surface increases (it is expected to occur with
has been proposed and there is significant evidence that distances up to approximately 20 A). Thus, it effectively
this also contributes. Discussions over the relative contri- operates only on the first layer of adsorbates. Campion
bution to the total enhancement of both these mechanisms et al. reported the first direct experimental evidence link-
are ongoing and no conclusions are imminent in the fore- ing new features in the electronic spectrum of an adsorbate
seeable future. Both electromagnetic and charge transfer to SERS, under conditions where electromagnetic enhance-
enhancements are discussed below. ments are unimportant.21 They noted that it was difficult to
observe charge transfer only because the electromagnetic
effects had to be accounted for and removed. They over-
2.1 Electromagnetic enhancement came this problem by measuring SERS enhancements on
an atomically flat, smooth single crystal surface where the
Surface roughness is an essential requirement for surface electromagnetic effects were small and well understood.
enhancement. On a smooth metal surface, surface plasmons They adsorbed pyromellitic dianhydride (PMDA) on to
exist as waves of electrons bound to the metal surface and copper(III) and observed an enhancement factor of 30. In
are capable of moving only in a direction parallel to that addition, a low-energy band in the electronic spectrum from
surface. On a roughened metal surface the plasmons are the adsorbed PMDA was observed which was absent in the
no longer confined and the resulting electric field can radi- solution PMDA spectrum.
ate in both a parallel and perpendicular direction. When an Hildebrandt and Stockburger carried out an extensive
incident photon falls on the roughened surface, excitation study on surface enhanced resonance Raman scattering
of the plasmon resonance of the metal may occur and on (SERRS) of rhodamine 6G on colloids, to explore the
the roughened surface this permits scattering. The result enhancement mechanisms involved in this technique.22
of the excitation is a large increase in the electric field. It Dyes such as rhodamine give resonance as well as sur-
is the interaction of the electric field with the adsorbate that face enhancement. The SERRS technique is discussed later.
gives the enhanced scattering. Additionally, due to the dif- They reported that two different types of adsorption sites
ference in dielectric constant between the roughened surface on the colloid surface were responsible for the enhancement
and the surrounding media, a concentration of electric field experienced: a nonspecific adsorption site that had high sur-
density occurs at sharp points on the surface. This enhance- face coverage on the colloid surface, and which resulted in
ment mechanism is known as the lightning rod effect. an enhancement factor of 3000 and could be explained by a
Detailed reviews and mathematical derivations of the elec- classical electromagnetic mechanism; and a specific adsorp-
tromagnetic approach have been reported separately.16,17 tion site that was only activated in the presence of certain
Experimental data has indicated that this enhancement may anions (Cl , I , Br , F and SO4 2 ). This specific site had
be as large as 106 .2,3 a low surface coverage (approximately three per colloidal
particle); however, an enhancement of 106 was claimed to
result. This enhancement was believed to be due to a charge
2.2 Charge transfer mechanism transfer mechanism.
Hildebrandt et al. continued the study of rhodamine 6G
This term covers the second mechanism of enhancement into the near-infrared (NIR) region and extended it to
believed to exist by several groups. Numerous studies have include gold colloid and gold and silver colloid supported
been carried out in order to establish or disqualify the on filter papers.23 Once again the enhancement factor expe-
existence of this mechanism.18 20 Its supporters use it to rienced was calculated for all the substrates, both activated
Surface-enhanced Raman Scattering 3
and unactivated. The enhancement experienced from anion used and developments within this field are continuing.27
activation with silver colloid was stronger by a factor of 47 Metal island films and metal films roughened by chem-
in the NIR region compared with the visible region. They ical or mechanical means28,29 were also sensitive. How-
concluded that this phenomenon could be accounted for ever, the production of the films is difficult to control
by the charge transfer transition being shifted towards the and hence it is difficult to obtain reproducible results.30
red for rhodamine 6G, therefore the resonance effect was Colloidal suspensions are widely used. There are many
stronger in the NIR region. It is interesting to note that no other SERS-active substrates. For example, metal colloidal
charge transfer mechanism was observed with gold colloid particles formed upon porous membranes such as filter
and gold and silver colloid on filter papers. papers, gels, beads, polymers etc. have been developed.31,32
In a final study, Hildebrandt and Stockburger concluded Although these adsorbed substrates are prepared easily
that charge transfer enhancement strongly depended upon they are unpopular, probably because they can be expen-
the structural and electronic properties of the analyte. They sive and in some cases irreproducible. They involve more
concluded that from all the dyes studied, charge transfer complicated preparations, and they are more susceptible
was observed from dyes with amino and cationic iminium to contamination. Several papers evaluate different SERS
groups where both nitrogen groups link to the metal surface substrates.33 36
via a delocalized p electron system. Therefore both nitrogen
atoms carried partial positive charge. Where this two-site
adsorption is not possible, for example with fluorescein 3.1 Electrodes
dyes, no charge transfer enhancement was observed.24
Ultimately the understanding and experimental proof The first observation of the surface enhancement phe-
supporting this enhancement mechanism is limited. The nomenon by Fleischmann et al.1 used an electrochemically
problem is made more complex by the fact that electro- roughened electrode as a surface. This group were studying
magnetic enhancement increases as the adsorbatesurface the identification of contaminants absorbed on the sur-
distance decreases and only additional enhancement can faces of working electrodes in electrochemical cells using
be classified as charge transfer. However, the degree of Raman spectroscopy. They reported that by electrochem-
enhancement of the first layer is very large. Thus, many
ically etching a silver electrode for 15 min from 0.2 to
questions remain unanswered and therefore the charge C0.3 V relative to the saturated calomel electrode (SCE)
transfer mechanism is not yet completely accepted.
at 0.5 V s1 , in the presence of 0.05 M pyridine and an
electrolyte (0.1 M KCl), they could achieve a large enhance-
ment in Raman scattering from the absorbed pyridine at
3 SERS-ACTIVE SUBSTRATES the electrode surface. Furthermore, a change in potential at
the silver electrode produced changes in the relative inten-
Surface enhancement is observed from a limited number sity and the Raman shifts of the bands already present,
of roughened metals, i.e. silver, gold, copper, aluminum, together with the appearance of new bands, particularly at
lithium and sodium. Flat surfaces appear incapable of 1025 cm1 . This band was present at 0 V relative to the SCE
enhancing Raman scattering.25 The intensity of scattering and decreased in intensity with increasing negative poten-
from adsorbed analytes is no longer proportional to the tial. It was assigned to pyridine coordinated to a metal atom.
frequency to the fourth power (n4 ); in fact, the intensity A typical electrochemical cell used for SERS is shown in
of the bands is related to the frequency of the surface Figure 1.
plasmon resonance and the laser excitation frequency.1 3 Several different metal electrodes have been used for
The exact dependence is related to the nature of the metal SERS, but the largest surface enhancement is observed
substrate, in particular to the identity of the metal and from those made of silver. When using an electrode as the
its roughness. The SERS spectra collected from adsorbed SERS substrate, the surface roughness can be created and
molecules are different from those of the corresponding the degree of roughness partially controlled by the choice
solution form. Normally large differences in relative inten- of electrolytes and electrochemical cycle. The degree of
sities are observed although differences in band frequency adsorption is also affected by the potential finally set for the
are small and shifts are limited to a few wavenumbers. electrode during the Raman measurement. As the potential
The enhancement decreases sharply by increasing the dis- is increased positive versus SCE the silver ions are oxidized
tance of the analyte from the roughened surface.26 Since and become solubilized, and as the potential is lowered the
SERS was first observed from pyridine on an electrochem- ions are reduced to produce silver metal:37
ically roughened silver electrode,1 numerous SERS active
substrates have been developed. Electrodes were originally AgC (solution) C e ! Ag0
ring-breathing modes of absorbed pyridine are shown to be

sensitive to the electrode potential. Both the overall Raman
Working electrode connector intensity and the relative intensity of the bands are affected.
Secondary electrode connector This is due to the nature of the bonding of the pyridine to
metal surface. The relative energies of the Fermi level in
Reference electrode capillary
the metal and of the p p levels in pyridine are affected
by the electrode potential. The maximum enhancement is at
0.600 V, where the band at 1010 cm1 is twice as intense
as the band observed at 1039 cm1 . The overall relative
intensities alter with increasing positive potential until at
Probe assembly 0.00 V the band at 1039 cm1 is more intense than the
band at 1010 cm1 . At potentials more positive than 0.00 V
the signal-to-noise ratio is drastically reduced. Addition-
ally, the signal is very sensitive to the orientation of the
pyridine. Well below monolayer coverage, the pyridine lies
flat on the surface and the SERS is weak. However, at
higher concentrations, the molecules pack with the ring
4 cm3 glass cell vertical to the surface. This provides a much greater change
in polarization perpendicular to the surface from displace-
ments associated with in-plane ring modes and the intensity
Supporting electrolyte of the SERS increases rapidly.
Electrochemical SERS has been extensively investigated
both to obtain a better understanding of the surface and of
the key processes, such as prevention of copper corrosion,
Capillary port
Compacted working electrode and as a detector for chromatography.38,39
The main disadvantage of the technique is the repro-
Secondary platinum ring electrode
ducibility of the electrode surface. It is incredibly difficult
Figure 1. An electrochemical cell typical of those used for SERS. to ensure that the same degree of surface roughness and
potential is achieved for each experiment. This may result
in different degrees of surface enhancement. Additionally,
0.000 V
+0.400 V
selection rules depend upon the analyte adsorption process
+0.200 V and orientation. Therefore the surface chemistry must be
closely controlled.
Raman intensity
0.600 V
0.400 V
0.800 V
1039 1010
0.200 V 3.2 Colloids
1.000 V
Colloidal suspensions are attractive as a substrate for SERS
as they can be reproducibly prepared in a one pot process
and are relatively inexpensive. Silver and gold colloidal
Figure 2. SERS spectra of the ring-breathing modes of absorbed solutions can be purchased commercially. Reliable SERS
pyridine (1010 and 1039 cm1 ) at various electrode potentials analysis is possible as a fresh reproducible colloidal surface
(l D 514.5 nm, 100 W). is available for each analysis. Numerous metal colloidal
suspensions have been used including gold and copper;
The analyte is adsorbed to the surface and SERS col- however, silver is the most popular.
lected. To clean the electrode surface the potential is
increased, thereby removing the analyte and silver surface.
A repeat of the cycle is required to create a roughened 3.3 Silver colloid
surface that creates a new SERS-active substrate. The
effect of controlling the potential of the electrochemical Several methods are suggested for the preparation of sil-
cell used for SERS is illustrated below for the original ver colloid.40 42 Each produces silver particles of differing
silver electrode and pyridine experiment (Figure 2). The diameters and uniformity. The Carey-Lea preparation is
reported to produce particles with diameters in the range The effectiveness of surface enhancement depends on
1422 nm,40 whereas the Creighton et al. colloid produces efficient adsorption of the analyte to the silver surface.
particles with diameters of 1822 nm;41 however, both Positively charged analytes will adsorb readily to the colloid
substrates were relatively unstable. One process that pro- surface, although problems are experienced with neutral and
duces reproducible and reliable results is a controlled cit- negatively charged analytes. This problem is overcome by
rate reduction of silver nitrate, originally reported by Lee the introduction of surface modifiers. Generally surfactant-
and Meisel42 and modified by Munro et al.43 It produces type compounds such as poly(L-lysine) are used for this
monodispersed colloidal particles with a diameter of 28 nm. purpose.
The difference in colloidal particle size may account for The SERS enhancement is further increased by aggregat-
the varying stability of colloid preparations. There is a ing the colloidal particles.49 The act of aggregation forms
natural tendency for small particles to reduce their sur- clusters of colloid, within which are interstices. The elec-
face area and energy in a bid to increase their stability by tric field established in these regions is predicted to be very
self-aggregating. As a Lee and Meisel colloid has a lower large, hence the surface enhancement experienced from this
surface area and energy than colloids prepared by other is very intense. Many different methods of aggregation have
methods, it is less likely to self-aggregate. been suggested including acids, salts and surfactants even
Numerous studies have been undertaken to character- using the analyte itself to induce aggregation.24,50,51 This
ize this substrate.44 46 A comprehensive study by Munro aggregation process is detected by a change in color of
et al.43 concluded that a stable monodispersed colloid with the colloidal suspension and a bathochromic shift of the
a diameter in the range 2430 nm and a maximum ultra- absorbance maximum.
violetvisible absorption between 402 and 404 nm with a
full width at half height of 6065 nm, could be reproducibly
prepared by a modified Lee and Meisel process. The result- 4 APPLICATIONS
ing colloidal particles are believed to be stabilized by a
polymeric surface layer of silver citrate which renders the 4.1 Surface enhanced Raman scattering
particles negatively charged.43 Figure 3 illustrates the Lee
and Meisel colloid surface. Several books have been devoted to SERS and numer-
The analytical range of the colloid can be extended by ous reviews have been written: early developments within
replacing the aqueous phase by an organic solvent.47,48 the field have been reported by Furtak and Reyez;52 Vo
This permits the application of SERS to molecules which Dinh et al. reviewed the successful application of SERS for
are insoluble in water. If a solvent-based solution of an chemical analysis;53 and Brolo et al. discussed applications
analyte is introduced to an aqueous colloid the analyte could of SERS to study metaladsorbate interactions.54 A review
precipitate and behave in an unpredictable manner at the by Nabiev et al. illustrates the advantages of SERS over
colloidal surface. Therefore the solvent used for the colloid normal Raman spectroscopy and its use in the medical field
should be one in which the analyte is soluble, especially as a diagnostic and analytical tool. The SERS technique has
if quantitative analysis is required. However, when the been used to probe the structure, topology and composition
aqueous phase is replaced, the colloid becomes rather less of biomedical species such as complexed DNA (deoxyri-
stable. The stability is dependent on solvent and ethanol has bonucleic acid), optical lenses and cellvirus interactions.55
been used successfully. Zeiri and Efrima discovered that Kneipp et al.56 and Laserna57 produced informative and
dichloromethane produced the most stable solvent-based objective overviews of the subject. Laserna explored the
colloid.47 possibility of coupling the fingerprinting capabilities and
trace detection powers of SERS in order to develop a pow-
O erful quantitative and qualitative analytical and research
O
C tool.57 He noted that certain chemical groups, such as ionic,
H C H strongly dipolar, highly polarizable heterocyclic groups
H H O
C C containing nitro and amino groups, gave intense SERS,
H probably because they strongly adsorbed to the metal sur-
C C
O O
O face. A detection limit of 109 M was reported. He con-
O
cluded that SERS had enormous potential; however, further
development was required, especially in improving repro-
Ag Ag ducibility. Munro et al. have addressed the problems asso-
Figure 3. The surface of a modified Lee and Meisel silver col- ciated with reproducibility and have focused much attention
loidal particle. on improving and standardizing the production of the silver
colloid generally used to obtain SERS.43 They concluded 1 2

that with careful attention to detail, a relative standard
deviation (RSD) of 5% was routinely obtained. A recent A
Absorbance
overview of SERS which covers the basic experimental
B
facts, mechanisms and development of the techniques has
been published by Campion and Kambhampati.58
The SERS technique has been applied extensively to
probe surfaces and boundaries in situ in polymer chem-
istry, when characterizing the surface of polymers for com-
parison with the bulk material properties.59 It has been (a) Wavenumber / nm
used to determine the molecular geometry, orientation of
polymer side groups adjacent to the metal surface, and 1 2
to obtain information on bonding, for example in poly-
mermetal composites such as adhesives and coatings.60 A
Absorbance
Its selectivity and sensitivity have been exploited in the
successful analysis and identification of seven chemically B
similar sulfur-containing drugs.61 Finally, it has been used
to follow photochemical reactions where the determination
of small quantities of fluorescent material is required62
SERS is perfectly suited for this purposes as fluorescence
is quenched and it is an extremely sensitive technique. (b) Wavenumber / nm
Figure 4. Illustration of the different arrangements for SERRS; in
which A is the molecular absorption and B is the plasmon absorp-
4.2 Surface enhanced resonance Raman tion. (a) The molecular and plasmon absorbances do not coin-
spectroscopy cide; position 1 represents excitation at the molecular maximum
and position 2 represents excitation at the plasmon maximum.
The SERRS technique is derived from SERS. It requires the (b) The molecular and plasmon maxima coincide; position 1 rep-
resents excitation away from molecular and plasmon maxima and
analyte to be a dye, which is adsorbed on the SERS-active position 2 represents excitation at the absorbance and plasmon
substrate. The laser frequency is chosen to be in resonance maximum.
or close to resonance with the molecular chromophore in the
dye. This provides a combination of molecular resonance this arrangement it has been reported that, for azo dyes, the
from the dye and surface enhancement from the metal. The surface selection rules expected for surface enhancement
method turns out to have unique advantages as a selective either do not apply or have a relatively minor effect.8,70,71
and sensitive analytical technique. What appears to occur is a scattering process, enhancing
The SERRS technique was originally reported by Stacy resonance scattering from the molecular chromophore. As is
and Van Duyne in 1983.63 In most studies the efficiency the case in resonance and in absorption, the input polariza-
of the scattering process is improved by three to four tion from the laser beam is altered by the chromophore dur-
orders of magnitude compared to that experienced in SERS, ing the scattering process. This results in a loss of the polar-
therefore an enhancement of up to 1010 may be expected ized information and consequently little or no dependence
and fluorescence is quenched.64,65 As in the simplest case on orientation of the molecule to the surface is observed.
the scattering is from one excited state, electronic as well This arrangement is ideal for quantitative analysis.
as vibrational information can be obtained. The second possible arrangement, also illustrated by
The enhancement obtained is very much greater than Figure 4(a), is where the laser excitation is set off from
either RR or SERS, enabling very sensitive analysis and the frequency of the adsorbate resonance and is at the
low detection limits to be achieved.66 69 It is a unique pro- maximum of the plasmon resonance (position 2). For reso-
cess and different effects can be obtained depending on the nance experiments on the molecule alone, this would be
nature of the chromophore and the choice of laser excita- described as a preresonant condition and often SERRS
tion. Figure 4 illustrates the main choices. In Figure 4(a) the undertaken in this way is written as SE(R)RS. The polar-
molecular chromophore A is chosen not to coincide with the ization information although affected is not lost to the
maximum of the plasmon resonance B. If laser excitation extent found in resonance. Thus, orientation information
at the molecular absorption maximum (position 1) is used, is to be expected. However, in this preresonant condi-
maximum resonance enhancement would be expected. With tion, the selectivity of resonance still applies. Thus, it is
possible to pick out individual molecules in the presence

of a matrix of interferents, but the effect will now be more
dependent on the angle of the adsorbate to the surface.
For many surface studies this is a key point, and con-
sequently this experimental process may be preferred for
surface analysis.
Resonance Raman
Figure 4(b) gives an alternative case in which the molec-
from rhodamine 6G
ular chromophore A coincides with the surface plasmon
maximum B. Similar considerations will apply but a greater
increase in sensitivity is likely at the resonance and plas-
Intensity
mon resonance maximum frequency. Orientation informa-
tion will becomes more explicit the greater the difference
in frequency between the excitation frequency used and the
plasmon resonance maximum.
In summary, SERRS offers improved sensitivity and
SERRS from
selectivity. By careful selection of the lower frequency to rhodamine 6G
maximize the resonance contribution and to minimize its
angular dependence it becomes more suitable to quantita-
tive analysis. However, the angle adopted by an adsorbed
molecule at a surface can be studied using different exci-
tation frequencies. Additionally, fluorescence is readily
quenched with SERRS. In general it is a more robust tech-
nique than SERS, and can offer selectivity, sensitivity and
200 400 600 800 1000 1200 1400 1600 1800
surface chemistry information. However it should be noted
Raman shift /cm1
that SERRS has some of the limitations of SERS and it
requires a chromophore. Figure 5. Raman scattering from a 106 M solution and SERRS
Rhodamine 6G has been used extensively as a model from a 108 M solution of rhodamine 6G (514.5 nm excitation, 1
accumulation of 10 s).
dye to probe the parameters of SERRS and illustrate its
advantages.24,25 Rhodamine 6G is an extremely strong
fluorophore when irradiated by visible excitation. Hence obtained from SERRS has been exploited by sev-
normal Raman is not observed except with NIR excita- eral researchers.24,73,74
tion. However, when SERRS is used the dye strongly As SERRS is sensitive, selective and produces structural
adsorbs to the roughened metal surface and consequently information and characteristic fingerprint spectra, it has
this strong fluorescence is quenched and an extremely been adopted as a detection method for high performance
strong, enhanced, Raman signal is observed.24 This is illus- liquid chromatography (HPLC) and flow injection analysis
trated in Figure 5. (FIA).75 77 Reproducibilities of 1% and a limit of detection
Attomolar levels (1018 M) of detection have been re- of 50 parts per million have been reported from an HPLC
ported for this system, which is approaching single molecule set-up for drugs, dyes etc.75 An optimized quantitative FIA
detection.66,67 Additionally, analysis of rhodamine 6G using system with maximum sensitivity and throughput achieved
silver colloid has identified two different adsorption sites on a RSD of 2.37% and a limit of detection of 1012 M.76
the surface, an unspecific site and a specific site. Each expe- SERRS has even been collected in situ from dyes separated
riences different enhancements, kinetics and physics, but on thin layer chromatography plates.78,79
the mechanisms involved still remain unclear and further The characteristic spectra routinely observed with
research is required.24,25 It should be remembered that rho- SERRS permits the identification of mixtures without
damine 6G appears to be the most efficient dye at adsorbing the need for preseparation. Munro et al. have reported
to the roughened metal surface. However, the detection the analysis and characterization of 20 similar monoazo
of single adsorbates of dopamine or phthalazine on col- dyes, all of which produced unique characteristic spectra
loidal clusters, with a limit of detection at picogram levels, which in turn permitted the simultaneous analysis and
illustrates that ultrasensitivity of this technique for other detection of five dyes presented in a crude mixture.69 This
adsorbates is possible.72 illustrates the potential of SERRS in forensic science, where
The fluorescence quenching properties of surface trace amounts of dyes etc. require to be analyzed and
enhancement coupled with the additional sensitivity identified.
4.3 Advantages of SERRS allowing strong SERRS to be recorded. Changes in the

pH result in changes in the chromophore of the dyes
The advantages reported above have been exploited in which were easily detected by SERRS.88 As the SERRS
numerous research fields. spectrum obtained from the complex was pH sensitive
it was possible to obtain quantitative pH determination.
Another example involves the exploitation of the sensitiv-
4.3.1 Biology ity of the technique to analyse trace amounts of nitrite in
When carefully executed, SERRS can provide a biocom- fresh and sea waters. In this case sulfanilamide was added
patible environment for biological material. The identi- to the water sample and reacted with any nitrite present
fication of water-soluble porphyrins, their photostability to form a diazonium salt, which was then coupled with
and interaction with roughened metal surfaces, have been ethylenediamine to produce an azo dye which was then
reported.80 The probing of copper chlorophyllin resulted detected by SERRS. Enhancement by a factor of 109 , a
in the collection of different spectra by using differ- RSD of 1015%, and a limit of detection of picograms
ent excitation frequencies. This has permitted a better was reported.89 This method was superior to existing colori-
understanding of a complex system.81 The identification metric and chemiluminescence techniques used to analyze
of novel chromophores in eye lenses without presepara- water samples for nitrite. Ultrasensitive metal ion detec-
tion of the crude mixture has been reported.82,83 Finally, tion has been reported. A limit of detection of nanogram
the orientation of cytochrome P450 by SERRS has been levels has been reported. The metal ions, nickel or cobalt
probed, and a fluorescent low-concentration protein solu- and a ligand, a mixture of 2-pyridinecarboxyaldehyde and
tion used to study hemesubstrate interactions and labeled 2-pyridinehydrazone or 1,10-phenolanthroline, form a com-
tyrosines.84,85 plex on the roughened metal surface that is then detected
by SERRS.90
4.3.2 Medicinal chemistry
4.3.5 Polymers
Detection of the antitumor drug mitoxantrone and its
interaction with DNA in situ have been probed by SERRS. Polymerization reactions at the metal surface have been
The adsorption of the drug complex onto a colloidal surface reported by Tsai et al.91 They prepared thin films of 1,4-
did not destroy or interfere with the native structure. dinitrobenzene on silver island metal films and under a
Therefore, bonding information from the complex was nitrogen atmosphere polymerization occurred which was
extractable from the SERRS spectra.86 followed by the appearance, and increases in intensity,
of new bands in the spectra collected. It is interesting
to note that polymerization of 1,3-dinitrobenzene did not
4.3.3 Surface chemistry
occur at the metal surface. Tsai et al. accounted for this
Electrode surfaces have been examined in situ by SERRS to by considering the orientation adopted by the monomer:
extract structural information and to provide quantification. when the aromatic rings of 1,3-dinitrobenzene lie per-
The in situ SERRS detection of compounds such as 2,4,6- pendicular to the silver surface only one nitrogen group
trinitrobenzene sulfonic acid, covalently bound to tin oxide, is in contact with the surface; however, when the aro-
was observed when the chemically modified surface was matic rings of 1,4-dinitrobenzene lie parallel to the surface
coated in silver. The spectra collected provided rapid sen- both nitrogen groups are in contact with the silver sur-
sitive structural information which was semiquantitative.87 face and are available to form a double bond with their
The technique has also been used to follow reactions occur- neighbor.
ring at well below monolayer coverage at roughened metal In summary, surface enhancement results in ultrasen-
surfaces.66 sitive and selective analysis using low excitation powers
which prohibit photodecomposition. The resulting spectra
provide molecular information and are unique to individual
4.3.4 Analysis
molecules. Sample preparation is simple and it is possi-
Several practical examples of SERRS have been devel- ble undertake analysis in situ. Additional advantages can
oped. It has been used to prepare a robust disulfide pH be obtained from using SERRS because this technique is
indicator by coupling pH-sensitive dyes, methyl red, cresol more sensitive and, as the excitation frequency is tuned
violet and 4-pyridinethiol, to cystamine, which adsorbs to coincide with an electronic transition within the chro-
strongly to the roughened metal surface forming mono- mophore, the resulting spectrum is simple and fluorescence
layer coverage of the complex with colloidal silver and is quenched.
5 CONCLUSIONS 16. D.A. Weitz, M. Moskovits and J.A. Creighton, in Chem-

istry and Structure at Interfaces, New Laser and Optical
Techniques, eds R.B. Hall and A.B. Ellis, VCH, Florida,
Both SERS and SERRS provide unique insights into the 197 (1986).
fundamental science of metalparticle interactions, and the
17. A. Otto, I. Mrozek, H. Grabhorn and W. Akemann, J. Phys.:
sensitivity, selectivity and informative nature of the scat-
Condens. Matter, 4, 1143 (1992).
tering make them ideal candidates for development as
18. A. Otto, Appl. Surf. Sci., 6, 309 (1980).
analytical methodologies. However, the techniques are chal-
lenging and require a considerable development in order to 19. A. Otto, J. Raman Spectrosc., 22, 743 (1991).
make them available for routine use. The increased practical 20. D.A. Guzonas, D.E. Irish and G.F. Atkinson, Langmuir, 6,
understanding of the techniques since the mid-1990s, cou- 1102 (1990).
pled to the huge improvements in Raman instrumentation, 21. A. Campion, J.E. Ivanecky, C.M. Child and M. Foster, J.
make the technical challenges more feasible and suggest a Am. Chem. Soc., 117, 11807 (1995).
bright future for both techniques. 22. P. Hildebrandt and M. Stockburger, J. Phys. Chem., 88, 5935
(1984).
23. P. Hildebrandt, S. Keller, A. Hoffmann, F. Vanhecke and
ABBREVIATIONS AND ACRONYMS B. Schrader, J. Raman Spectrosc., 24, 791 (1993).
24. P. Hildebrandt and M. Stockburger, J. Raman Spectrosc., 17,
DNA Deoxyribonucleic Acid 55 (1986).
PMDA Pyromellitic Dianhydride 25. A. Campion and D.R. Mullins, Chem. Phys. Lett., 54, 576
RSD Relative Standard Deviation (1983).
SCE Saturated Calomel Electrode 26. C.A. Murray, in Surface Enhanced Raman Scattering, eds
R.K. Chang and T.E. Furtak, Plenum Press, New York,
62108 (1982).
REFERENCES 27. L. Markwort and P.J. Hendra, J. Electroanal. Chem., 397,
225 (1995).
1. M. Fleischmann, P.J. Hendra and A.J. McQuillan, Chem. 28. K. Suzan, A. Baiker, M. Meier and J. Wokaun, J. Chem.
Phys. Lett., 26, 163 (1974). Soc., Faraday Trans., 80, 1305 (1984).
2. D.L. Jeanmaire and R.P. Van Duyne, J. Electroanal. Chem., 29. G. Xue, J. Dong and M. Zhang, Anal. Chem., 63, 2393
84, 1 (1977). (1991).
3. M.G. Albrecht and J.A. Creighton, J. Am. Chem. Soc., 99, 30. A. Ruperez and J.J. Laserna, Anal. Chim. Acta, 291, 147
5215 (1977). (1994).
4. J.A. Creighton and D.G. Eadon, J. Chem. Soc., Faraday 31. J.J. Laserena, A.D. Campiglia and J.D. Winefordner, Anal.
Trans., 87, 3881 (1991). Chim. Acta, 208, 21 (1988).
5. M. Kerker (ed.), Selected Papers on Surface Enhanced
32. W.S. Sutherland and J.D. Winefordner, J. Colloid Interface
Raman Scattering, SPIE Milestone Series, Washington,
Sci., 148, 129 (1992).
Volume MS 10 (1990).
33. J. Billmann, G. Kovacs and A. Otto, Surf. Sci., 92, 153
6. J.M. Sequaris and E. Koglin, Anal. Chem., 321, 758 (1985).
(1980).
7. E.L. Torres and J.D. Winefordner, Z. Anal. Chem., 59, 1626
(1987). 34. K.L. Norrod, L.M. Sudnik, D. Rousell and K.L. Rowlen,
Appl. Spectrosc., 51, 994 (1997).
8. C.H. Munro, W.E. Smith, D.R. Armstrong and P.C. White,
J. Phys. Chem., 99, 879 (1995). 35. R. Perez, A. Ruperez and J.J. Laserna, Anal. Chim. Acta,
376, 255 (1998).
9. M. Moskovits, J. Chem. Phys., 79, 4408 (1982).
36. W.H. Li, X.Y. Li and N.T. Yu, Chem. Phys. Lett., 312, 28
10. M. Moskovits and J.S. Suh, J. Phys. Chem., 88, 1293 (1984). (1999).
11. X. Gao, J.P. Davies and M.J. Weaver, J. Phys. Chem., 94, 37. T. VoDinh, D.L. Stokes, G.D. Griffen, M. Volkan, U.J. Kim
6858 (1990). and M.I. Simon, J. Raman Spectrosc., 30, 785 (1999).
12. J.A. Creighton, The Selection Rules for Surface Enhanced 38. B.H. Loo, Ann. Rep. INCRA, Project No. 341 (1982).
Raman Spectroscopy, in Spectroscopy of Surfaces, eds
R.H. Clark and R. Hester, John Wiley & Sons (1988). 39. K.T. Carron, G.I. Xue and M.L. Lewis, Langmuir, 1, 2
(1991).
13. M. Moskovits and J.S. Suh, J. Phys. Chem., 88, 5526 (1984).
40. M. Carey-Lea, Am. J. Sci., 37, 476 (1889).
14. P. Hildebrandt and M. Stockburger, J. Raman Spectrosc., 17,
55 (1986). 41. J.A. Creighton, C.G. Blatchford and M.G. Albrecht, Trans.
15. D. Graham, C. McLaughlin, G. McAnally, J. Jones, P.C. Faraday Soc., 75, 790 (1979).
White and W.E. Smith, Chem. Commun., 11, 1187 (1998). 42. P.C. Lee and D.P.J. Meisel, J. Phys. Chem., 86, 3391 (1982).
43. C.H. Munro, W.E. Smith, M. Garner, J. Clarkson and P.C. 69. C.H. Munro, W.E. Smith and P.C. White, Analyst, 120, 993
White, Langmuir, 11, 3712 (1995). (1995).
44. J.J. Laserna, E.L. Torres and D. Winefordner, Anal. Chim. 70. K. Xi, S.K. Sharma and D.W. Muenow, J. Raman Spec-
Acta, 200, 469 (1987). trosc., 23, 621 (1992).
45. S.M. Heard, F. Grieser and C.G. Barraclough, J. Colloid 71. S. Chattopadhyay and S.K. Brahma, Spectrochim. Acta, 49A,
Interface Sci., 93, 545 (1983). 443 (1993).
46. S. Sanchez-Cortes, J.V. Garcia-Ramos and G. Morcillo, J. 72. B. Vlckova, X.J. Gu, D.P. Tsai and M. Moskovits, J. Phys.
Colloid Interface Sci., 167, 428 (1994). Chem., 100, 3169 (1996).
47. L. Zeiri and S. Efrima, J. Phys. Chem., 96, 5908 (1992). 73. A.C. Pineda and D. Ronis, J. Chem. Phys., 83, 5330 (1985).
48. B. Bozlee, B. Lian, J. Kahn, R.L. Garrell, T. Herne, A. 74. I. Persaud and W.E.L. Grossman, J. Raman Spectrosc., 24,
Leiden, P. Palko, N. Hess and G. Exarhos, Chem. Phys. Lett., 107 (1993).
196, 437 (1992). 75. R.D. Freeman, R.M. Hammaker, C.E. Meloan and W.G.
49. O. Siiman and H. Feilchenfeld, J. Phys. Chem., 92, 453 Fateley, Appl. Spectrosc., 42, 456 (1988).
(1988). 76. G.T. Taylor, S.K. Sharma and K. Mohanan, Appl. Spectrosc.,
50. Y. Xu and Y. Zheng, Anal. Chim. Acta, 225, 227 (1989). 44, 635 (1990).
51. S. Fu and P. Zhang, J. Raman Spectrosc., 23, 93 (1992). 77. V.J.P. Gouveia, I.G. Gutz and J.C. Rubim, J. Electroanal.
52. T.E. Furtak and J. Reyez, Surf. Sci., 93, 351 (1980). Chem., 371, 37 (1994).
53. T. Vo Dinh, A. Alak and R.L. Moody, Spectrochim. Acta, 78. C.D. Tran, Anal. Chem., 56, 824 (1984).
43B, 605 (1988). 79. S.A. Soper and T. Kuwona, Appl. Spectrosc., 43, 1180
54. A.G. Brolo, D.E. Irish and B.D. Smith, J. Mol. Struct., 405, (1989).
29 (1997). 80. T.M. Cotton, S.G. Schultz and R.P. Van Duyne, J. Am.
55. I. Nabiev, I. Chourpa and M. Manfait, J. Raman Spectrosc., Chem. Soc., 104, 6528 (1982).
25, 13 (1994). 81. P. Hildebrandt and T.G. Spiro, J. Phys. Chem., 88, 3355
56. K. Kneipp, H. Kneipp, I. Itzkan, R.R. Dasari and M.S. Feld, (1984).
Chem. Rev., 99, 2957 (1999). 82. S. Nie, C.G. Castillo, K.L. Bergbauber, J.F.R. Kuck, I.R.
57. J.J. Laserna, Anal. Chim. Acta, 283, 607 (1993). Nabiev and N. Yu, Appl. Spectrosc., 44, 571 (1990).
58. A. Campion and P. Kambhampati, Chem. Soc. Rev., (1998). 83. K.V. Sokolov, S.V. Lutsenko, I.R. Nabiev, S. Nie and N. Yu,
Appl. Spectrosc., 45, 1143 (1991).
59. R.L. Garrell and K.D. Beer, Spectrochim. Acta, 43B, 617
(1988). 84. I.D.G. Macdonald, W.E. Smith and A.W. Munro, FEBS Lett.,
396, 196 (1996).
60. G. Xue and Y. Lu, Macromol. Rapid Commun., 99, 15
(1994). 85. L. Quaroni, J. Reglinski, R. Wolf and W.E. Smith, Biochim.
Biophys. Acta, 1296, 5 (1996).
61. R. Montes and J.J. Laserna, Analyst, 115, 1601 (1990).
86. I. Nabiev, A. Baranov, I. Chourpa, A. Beljebbar, G.D.
62. D. Franzke and A. Wokaun, J. Phys. Chem., 96, 6377 (1992). Sockalingum and M. Manfait, J. Phys. Chem., 99, 1608
63. A.M. Stacy and R.P. Van Duyne, Chem. Phys. Lett., 102, 365 (1995).
(1983). 87. B.B. Kaul, R.E. Holt, V.L. Schlegel and T.M. Cotton, Anal.
64. O. Siiman, A. Lepp and M. Kerker, J. Phys. Chem., 87, 5319 Chem., 60, 1580 (1988).
(1983). 88. K.I. Mullen, D. Wang, L.G. Crane and K.T. Carron, Anal.
65. E.J. Zeeman, K.T. Carron and G.C. Shatz, J. Chem. Phys., Chem., 64, 930 (1992).
87, 4189 (1987). 89. K. Xi, S.K. Sharma, G.T. Taylor and D.W. Muenow, Appl.
66. C. Rodger, W.E. Smith, G. Dent and M. Edmondson, J. Spectrosc., 46, 819 (1992).
Chem. Soc., Dalton Trans., 5, 791 (1996). 90. M. Hildago, R. Montes, J.J. Laserna and A. Ruperez, Anal.
67. K. Kneipp, Y. Wang, R.R. Dasari and M.S. Feld, Appl. Spec- Chim. Acta, 318, 229 (1996).
trosc., 49, 780 (1995). 91. W.H. Tsai, F.J. Boerio, S.J. Clarson and G. Montaudo, J.
68. Y. Xu and Y. Zheng, Anal. Chim. Acta, 225, 227 (1989). Raman Spectrosc., 21, 311 (1990).
Surface-enhanced Infrared Absorption Spectroscopy
Masatoshi Osawa
Hokkaido University, Sapporo, Japan
1 INTRODUCTION range of picograms to micrograms. Its use for chemical and

biochemical sensing, such as monitoring of waste water
Molecules adsorbed on metal island films or metal colloids and immunoassay, has been proposed. It is obvious that
show 101000 times more intense infrared (IR) absorp- this technique is useful for studies of molecular adsorption
tion than would be expected from conventional measure- and reactions at metal surfaces. The surfaces can be char-
ments without metals. This effect has several similarities to acterized in vacuum, air, gas and solution phases. So far,
surface-enhanced Raman scattering (SERS)1,2 and is now SEIRAS has been applied most extensively to in situ and
referred as surface-enhanced infrared absorption (SEIRA), time-resolved studies of the electrochemical interface. Solid
to stress the analogy with SERS. surfaces other than metals, such as polymers, glass, semi-
The SEIRA effect was discovered in the attenuated conductors, human skin, agricultural products, etc., can also
total reflection (ATR) measurement of aromatic carboxylic be characterized at a higher sensitivity by using the SEIRA
acids adsorbed on vacuum-evaporated Ag and Au films.3 effect.
Although the enhanced bands observed in this measure- In this article, the nature and mechanisms of SEIRA and
ment were of hydrocarbon contaminants,4,5 this study developments in experimental techniques and applications
constituted the starting point for consecutive investiga- are described.
tions of surface-enhanced infrared absorption spectroscopy
(SEIRAS). Subsequent research on the mechanism of the 2 NATURE OF SURFACE-ENHANCED
enhancement2,4 11 led to a general consensus that at least
two different mechanisms, electromagnetic (EM) and chem-
INFRARED ABSORPTION
ical, are likely to contribute to the total enhancement,
Since the discovery of the effect, a number of SEIRA spec-
as in SERS. Coupling of photons with the metal and
tra of organic and inorganic molecules on various metals
dipole interactions between the metal and the adsorbed
have been observed and reported in the literature. An exam-
molecules are believed primarily to enhance the IR absorp-
ple of SEIRA spectra is shown in Figure 1(a), which is a
tion of the molecule. A theoretical EM model taking into
set of transmission spectra of 1.7 nm thick p-nitrobenzoic
account these interactions12,13 explains well many exper-
acid (PNBA) layers cast on Ag films with different thick-
imental results reported so far. There exist also several
ness (dAg ) ranging from 0 to 14 nm.2 The metal films
experimental results that can be explained by assum-
were prepared by evaporating Ag on CaF2 in vacuum.
ing some chemical interactions between the surface and
The dAg is the mass thickness measured with a quartz
molecules,2,9,11,14 although the details are not yet fully crystal microbalanceQuartz Crystal Microbalance (QCM).
understood. The organic layers were formed by dropping acetone solu-
The applications of SEIRAS are rapidly expanding. The tion on top of the metal surfaces and then evaporating the
high sensitivity of this technique allows quantitative and solvent. The absorption bands of the molecule are super-
qualitative analyses of trace amounts of chemicals in the posed on a very broad absorption of the metal ranging
from the visible to IR regions.2 The background spectra
John Wiley & Sons Ltd, 2002. of the metal films that had been measured before forming
the organic layers were subtracted from each spectrum. No enhanced roughly 500-fold on a 10 nm thick Ag film, by
bands are detectable on the bare (dAg D 0) substrate, but taking into account the difference in the thickness of the
several bands increase in intensity as dAg increases up to molecular layer.
10 nm and then decrease. Obviously the absorption of the Vacuum-evaporated thin films that exhibit SEIRA are
organic layer is significantly enhanced in the presence of not continuous but consist of islands smaller than the
the Ag films. wavelength of light (20100 nm depending on evaporation
The enhanced spectra differ considerably from the nor- conditions).15 The islands grow in size with increasing
mal spectrum of a 132.8 nm thick PNBA layer deposited mass thickness and eventually form a continuous film. The
directly on a bare CaF2 plate, shown in Figure 1(b), enhancement factor is very sensitive to the morphology
owing to the chemisorption of PNBA on Ag. The 1390- of the island structure.3,15,16 The largest enhancement is
cm1 band, which is absent from the normal spectrum, is observed when these islands are closely spaced but not
assigned to the symmetric COO stretching mode (ns COO) touching each other. The decrease in the band intensities
of p-nitrobenzoate (PNBA ), indicating that PNBA is at dAg D 14 nm in Figure 1(a) is related to the coagulation
chemisorbed on the surface by releasing the proton of (percolation) of islands.3 It has been found that electro-
the carboxylic group. The strongest band at 1350 cm1 chemically deposited Pt-black16 and metal colloids17 19
is assigned to the symmetric NO2 stretching. The peak also exhibit strong SEIRA. These findings strongly suggest
intensity of this band is about 0.07 (in absorbance units) that the metal islands or particles play an important role in
on an Ag film of dAg D 10 nm, whereas it is only about this effect. The surface areas of island films are larger than
0.01 for a thick PNBA layer on CaF2 . Assuming ten- smooth bulk metal surfaces. The difference in the surface
tatively that absorptivity of this mode is not changed area undoubtedly contributes to the enhancement. In the
upon chemisorption, the band intensity is estimated to be case of vacuum-evaporated Ag and Au island films, how-
ever, the roughness factor (the ratio of the actual surface
area to the geometric area) estimated by several techniques
(b) PNBA (132.8 nm)/CaF2 was only 24.13,20
The SEIRA spectra shown in Figure 1(a) are much sim-
1350
pler that the normal IR spectrum of the potassium salt

1690
0.02
of PNBA dispersed in a KBr pellet. In the latter spec-
trum, the antisymmetric COO and NO2 stretching modes
are observed at 1590 and 1540 cm1 as strong as the corre-
(a) PNBA (1.7 nm)/Ag/CaF2
sponding symmetric modes.13 By contrast, these antisym-
dAg = 14 nm metric modes are completely missing from the enhanced
Absorbance
spectra. The features of the SEIRA spectrum are identical to

1390
1350
those of the spectrum of PNBA adsorbed on a smooth bulk

0.05
Ag surface measured with infrared reflectionabsorption
10 nm spectroscopy (IR-RAS).13 In the case of IR-RAS, the EM
field that excites adsorbed molecules is essentially polar-
ized along the surface normal, and hence vibrations that
8 nm
have dipole moment derivative components perpendicular
6 nm
to the surface are exclusively IR-active.21 On the basis of
4 nm
the surface selection rule, the IR-RAS spectrum can be
0 nm interpreted by PNBA being oriented with its C2 symmetry
axis perpendicular to the surface. Since this molecule will
3000 2600 1800 1400 1000 be adsorbed on metal island films with the same orientation,
Wavenumber /cm1 the similarity of the spectra indicates that the surface selec-
Figure 1. IR spectra of PNBA deposited on (a) Ag-coated CaF2 tion rule in IR-RAS can be applicable also to SEIRAS. The
and (b) metal-free CaF2 .2 The average thickness of the organic presence of the surface selection rule might seem to be very
layers is 1.7 nm for (a) and 132.8 nm for (b). The mass thickness of confusing, because the molecular orientation should be ran-
the Ag films is shown. The down-going bands around 2900 cm1 dom on average with respect to the substrate surface even
are of hydrocarbon contaminants that were removed from the
surface by the adsorption of PNBA. [Reproduced by permission
if molecules are perpendicularly oriented with respect to
of the American Chemical Society from M. Osawa and M. Ikeda, the surfaces of the metal islands. This puzzle can be under-
J. Phys. Chem., 95, 9914 (1991).] stood by assuming that the EM field that excites molecular
Surface-enhanced Infrared Absorption Spectroscopy 3
0.05 3 ENHANCEMENT MECHANISMS
0.04 3.1 Electromagnetic mechanism

Peak intensity/abs.
0.03 It is well established that metal islands or particles play

an important role in SERS, as in the case of SEIRA.
0.02 Metal particles are polarized in the incident photon field
through the excitation of collective electron resonance or
localized plasmon modes and the dipole p induced in the
0.01
particles generates a local EM field around them,24 28 as
illustrated in Figure 3. The local EM field is much stronger
0.00 than the incident EM field in the visible region, and can
0 5 10 15 20 25
strongly excite the Raman scattering of adsorbed molecules,
dPNBA resulting in SERS.
The same EM mechanism can be expected also for
Figure 2. Peak intensity of the 1390-cm1 band of chemisorbed SEIRA.3 5 Since the enhanced EM field is polarized
PNBA (circles) and the 1690-cm1 band of PNBA overlayers
along the surface normal at every point of metal particles
(solid squares) on a 10 nm thick Ag film vacuum evaporated
on CaF2 as a function of the thickness of the organic layer and decays very sharply as the distance from the sur-
(dPNBA ). Open squares are the intensity of the 1690-cm1 band of face increases,24 28 this mechanism is very convenient to
PNBA observed on a metal-free substrate. The curves were drawn explain the surface selection rule and short-range enhance-
by hand. ment. Nevertheless, the EM field enhancement estimated in
the IR region is only 10-fold or less2,29 and is not large
vibrations is perpendicular to the surface at every point of enough to explain the total enhancement of up to 1000.
metal islands.5,13 A more important EM effect is the perturbation of the
In conventional measurements, band intensities are a optical properties of the island film by the presence of sur-
linear function of the thickness of the film (LambertBeer face molecules. Oscillating dipoles in the surface molecule
law). In SEIRA experiments, on the other hand, band can induce dipoles P in the metal particles,24 28 result-
intensities are not a linear function of the film thickness. In ing in the change of the dielectric function of the metal.
Figure 2, the intensities of the ns COO band of PNBA (at Since the perturbation should be larger at frequencies of the
1390 cm1 ) and the nCDO band of PNBA (at 1690 cm1 ) molecular vibrations than at other frequencies, the molecu-
are plotted as functions of the thickness of the organic layer lar vibrations can be observed through the change in the
(dPNBA ) formed on an Ag film of dAg D 10 nm (solid circles transmittance or reflectance of the metal island film. It
and solid squares, respectively). The ns COO band increases should be noted that absorptivities of metals are larger than
sharply in intensity as dPNBA increases and then saturates at those of molecules in the mid-infrared (MIR) region. More-
about dPNBA D 3 nm, suggesting that the metal islands are over, the volume fraction of the metal is much larger than
completely covered with PNBA . After the saturation of that of monolayer adsorbates. Consequently, the change in
the ns COO band, the nCDO band starts to increase owing optical properties of the island film will dominate the spec-
to the multilayer deposition of the acid. The gradient of the trum. In other words, the metal island film can function as
intensity versus dPNBA plot is relatively large at dPNBA D an amplifier of IR absorption.
38 nm and eventually approaches that for the same band of The effective dielectric function of an island film can
PNBA films deposited on the bare substrate (open squares). be connected with the polarizability of islands by using
The enhancement factor, estimated from the gradients with an appropriate effective medium theory.30 Several effec-
and without the metal film, is 2030 at dPNBA D 38, tive medium theories model the island film as a composite
which is much smaller than that for the first monolayer film consisting of metal particles, molecules and voids (or
(about 500). The result indicates that the enhancement is host medium). The transmittance and reflectance of the host
significant for the first layer directly attached to the surface mediumcomposite filmsubstrate three-phase system can
and decays as the distance from the surface increases. be calculated using the Fresnel equation.31 Spectra of a
The effective spatial range of the enhancement has been model molecule on a 8 nm thick Ag island film simulated
investigated in more detail by using LangmuirBlodgett by using the Bruggemann effective medium theory32 are
(LB) films7,10 and other organic films,22,23 and it was shown shown in Figure 4 (solid curves), where metal particles
that the enhancement diminishes within about 5 nm from were modeled as rotating ellipsoids of a uniform size, and
the surface. h represents the aspect ratio of the ellipsoid defined by
IR radiation
Io ()
Adsorbed molecule
Local electric field
p mn p
p dp
Substrate
Metal
I ()
I ( = mn) < I ( nmn)
Figure 3. Schematic representation of the EM mechanism of SEIRA on an island metal film. Incident photon field I0 polarizes the
metal islands and the dipole p generates an enhanced local electric field around the islands, which excites molecular vibration nmn . The
molecular vibration induces an additional dipole p in the metal islands and perturbs the optical properties of the metal island film.
0.07 the observation that densely packed island films give a large
0.06 =5 enhancement.
b The calculated enhancement factor that arises from the
0.05 a
=3 EM mechanism is about 150 for h D 3 (a typical value for
= a /b vacuum-evaporated Ag island films). It should be noted that
Absorbance
0.04
=2 chemisorption of molecules often aligns molecular dipoles
0.03 =1 in a specific direction with respect to the surface. On the
0.02 Without metal basis of the surface selection rule, dipoles oscillating per-
pendicular to the surface should be observed three times
0.01
more intensely than randomly oriented dipoles. Accord-
0.00 ingly, a total enhancement of about 150 3 D 450 can be
estimated for vibrational modes that have dipole derivatives
0.01
1800 1750 1700 1650 1600 along the surface normal. This estimate is in reasonable
Wavenumber /cm1 agreement with the observation.
Figure 4. Transmission spectra of a model molecule adsorbed SERS is significant on free electron metals such as Au,
on an 8 nm thick Ag film simulated with Bruggemann effective Ag and Cu, and is hardly observed on transition met-
medium theory. h is the aspect ratio of the islands assumed als owing to the strong damping of localized plasmon in
to be rotating ellipsoids. See Osawa et al.13 for details of the the visible and near-infrared (NIR) regions.24 26 In con-
simulation.
trast, the EM calculations predict enhancement on transition
metals as strongly as on coinage metals.13 This is a rea-
the ratio of its major and semi-major radii. The details of sonable consequence of the fact that dielectric constants
the simulation can be found in elsewhere.1,13 The dashed of metals do not differ significantly from each other in
curve corresponds to the spectrum of the same molecule the IR region. Although very small or negligible enhance-
without the metal. The simulation clearly demonstrates the ment was reported on vacuum-evaporated transition metals
enhancement of IR absorption in the presence of metal par- in earlier experiments,33,34 several subsequent experiments
ticles. As can be seen, the enhancement factor depends on clearly demonstrated that SEIRA occurs on island films of
the shape of the islands (i.e. h). The simulation revealed many transition metals.14,16,35 42
also that interactions between metal islands play an impor- The SEIRA bands often have very asymmetric
tant role for the enhancement,1,13 which is consistent with or derivative-like shapes. In some cases, molecular
vibrations are observed as negative peaks (i.e. negative ZnS,35 KRS-5,49 sapphire,53 MgO,14 glass,10,43 glassy
absorption).16,37,40,41 These spectral features are well carbon,37,40,41 polymers,43 and metals.16,54 The enhance-
simulated with the EM model.16,43 ment factor depends on the refractive index and the chem-
ical nature of the substrate.15 Low-reflective substrates are
optically favorable.15 However, the chemical nature of the
3.2 Chemical mechanisms substrate affects the enhancement more strongly because it
determines the morphology of the deposited film.15 Hence
In general, chemisorbed molecules exhibit larger enhance- careful cleaning of the substrate before the metal deposi-
ment than physisorbed molecules, suggesting a contribution tion is necessary to obtain reproducible results. Modifica-
of some chemical effects to the total enhancement. A pos- tion of the substrate surface increases the enhancement in
sible origin is the orientation effect as described above. some cases.15,55 A hydrogen-terminated Si(100) surface is
Chemisorption will also change the polarizability (i.e. reported to give a 1.7 times larger enhancement than an
absorptivity) of molecules. Two- to six-fold enlargements oxidized surface.55 KBr and KRS-5 sometimes give spec-
of vibrational polarizability have been estimated for CO tral features different from other substrates,49 the reason of
chemisorbed on metals.44 46 Vibrations of strongly polar- which is not clear.
ized groups within adsorbed molecules generally give larger
enhancements than other groups, which was explained by
donoracceptor interactions with the metal.11 4.2 Preparation of surface-enhanced infrared
Some theoretical considerations suggest that charge oscil- absorption-active metal surfaces
lations between molecular orbitals and the metal increase
the absorption coefficient of the adsorbates.44,45,47 The SEIRA has been observed on various metals such as Ag,3 5
molecular vibrations can gain their intensities by intensity Au,3,56 Cu,11,56,57 In,56 Li,58 Sn,35,59 Pb,36 Fe,14 Pt,16,37 39
borrowing from the charge oscillation (or charge transfer) Pd,37 Rh,41 Ir,40 and PtFe alloys.42 In general, less noble
between the metal and adsorbates. The SEIRA spectrum of metals show smaller enhancements than noble metals if
CO adsorbed on Fe island films evaporated on MgO(001) they are measured in air, because the formation of oxide
is characterized by a very asymmetric bandshape. The layers on the surface reduces the enhancement (note that
bandshape was well simulated by such a charge-oscillation the enhancement is short range).
mechanism.14 Nevertheless, asymmetric band shapes can The most widely used method for preparing SEIRA-
also be explained by the EM mechanism.16 Therefore, active metal surfaces is vacuum evaporation. Figure 5
the interactions between the surface and molecule are shows a typical vacuum-evaporation system, which consists
still a matter for discussion, and further detailed theoret-
ical and experimental studies of chemical mechanisms are Glass bell jar
desirable.
Quartz crystal
microbalance
(QCM) Substrate or
4 EXPERIMENTAL PROCEDURES sample
4.1 Sample preparation

Viton gasket
SEIRA can be observed with both metal-underlayer (sam-
plemetalsubstrate) and metal-overlayer (metalsample
substrate or metalsample) geometries. In the metal-
underlayer experiments, a thin metal layer is formed first
on a suitable substrate, and then the sample molecules are
Cooling water AC voltage
adsorbed on top of the metal surface. In the metal-overlayer
measurements, on the other hand, a metal layer is directly Tungsten
deposited on a sample layer. In general, both geometries basket
give almost equal enhancement. However, thermal deposi- QCM controller
tion of metals leads to the possibility of the decomposition and monitor
or desorption of the sample molecules. Turbo molecular
To rotary pump pump
Many materials have been used as the substrate on
which a metal film or organic film is formed, includ- Figure 5. Schematic drawing of a typical vacuum evaporation
ing Si,3 Ge,4,5 CaF2 ,2,48 BaF2 ,49 51 KBr,35,36,49 ZnSe,50,52 system used for sample preparation.
of a stainless-steel chamber, a glass bell jar, an evaporation enhancement factor owing to the change in the morphology
source (a tungsten basket), a QCM, and vacuum pumps. of metal islands.15,53 It is essential to place the substrate at
To minimize the contamination of the prepared metal sur- least 20 cm away from the evaporation source to reduce
face, an oil-free pumping system is highly recommended. the irradiative thermal heating of the substrate from the
Although ultra-high vacuum is not required in general, evaporation source. Since QCM thickness monitors are very
higher vacuum will give more reproducible results. The sensitive to temperature, this condition is important also for
thickness and deposition rate of the metal are controlled by reliable measurements of the thickness and deposition rate
adjusting the voltage applied to the tungsten basket under of the metal.
the monitoring with the QCM thickness monitor. For more precise control of the morphology of evapo-
Metals with low melting points can be easily evaporated metal films, templates are used. Periodic particle array
rated by resistive thermal heating of a tungsten basket or films have been prepared by nanosphere lithography.60
boat. For metals with high melting points, direct resis- In this method, a hexagonally close-packed monolayer
tive heating of metal wires or sputtering is used. Alloys of polystyrene nanospheres is formed on a substrate by
can be prepared by evaporating the component metals drop coating a suspension of the polystyrene nanospheres,
simultaneously.42 and then a desired metal is vacuum evaporated onto the
The enhancement depends strongly on the morphology film. Following the Ag deposition, the nanosphere mask
of the metal film (shape, size and density of islands). is removed by sonication in methylene chloride, leaving
Therefore, the control of the morphology is the most behind a hexagonal array of uniformly sized Ag particles
essential aspect of SEIRAS measurements. In the case of with triangular cross-sectional shapes. Anodized porous
vacuum evaporation, the morphology is greatly affected surface of Al61 and island films of different metals38 also
by the deposition rate. In Figure 6, the intensity of the can be used as templates.
symmetric NO2 stretching band of PNBA adsorbed on Ag
Electrochemical deposition is an alternative inexpensive
films prepared with three different deposition rates is plotted
method for preparing SEIRA-active metal films. Desired
as a function of the mass thickness of the film, dAg . The
metals can be deposited on conducting substrates in elec-
substrate is BaF2 in this case. It can be seen that a slower
trolyte solutions containing salts of the metals by applying a
deposition gives a larger enhancement and that the optimal
suitable potential or current. The thickness of the deposited
thickness of the metal depends on the deposition rate. Au
metal film can be estimated from the charge passed during
on CaF2 and Si shows a similar trend, but the optimal
the deposition.
conditions are different from system to system.15
Electrochemical deposition of Ag, Pt, Pd, Rh and Ir
Metal evaporation is generally carried out at room tem-
on Au-coated glass, n-type Si, glassy carbon, and Pt has
perature. Heating of the substrate usually reduces the
been reported.16,37,40,41,62,63 The deposition was carried
0.25 out under potentiostatic,16,63 galvanostatic,16,62 or poten-
tial cycling37,40,41 conditions. Additives were used in some
C
cases to control the morphology of the deposited films;
0.20
for example, ethylenediaminetetraacetic acid (EDTA) for
Ag deposition63 and lead acetate for Pt deposition.16
Intensity/abs.
0.15 B On electrochemically deposited Pt-group metals, enhanced

A bands are often observed as negative peaks (i.e. negative
0.10
absorption),16,37,40,41 which can be related to the mor-
phology of the metal film.16 All these electrochemically
deposited metal films were reported to exhibit over 10-fold
0.05 more intense absorption than smooth bulk surfaces mea-
sured with IR-RAS (which corresponds to an enhancement
factor of the order of 100 if compared with transmission
0 5 10 15 20 25
measurements without metals13 ).
dAg / nm
Metal colloids also can be used as enhancers. They are
Figure 6. Peak intensity of the NO2 stretching band of PNBA prepared by reducing metal ions with suitable reductants.17
chemisorbed on vacuum-evaporated Ag island films plotted as Au and Ag colloids are commercially available. Metal
a function of the mass thickness of the metal film (dAg ). The
deposition rates of the Ag films are (A) 0.1, (B) 0.05, and
colloids are used by spreading over substrates17 or by
(C) 0.001 nm s1 . The substrate is BaF2 . The curves were drawn filtering with porous polyethylene membranes (3M IR Card;
by hand. 3M, St Paul, MN).19
Enhanced IR spectra have also been observed on chem-

ically etched rough Ag surfaces for thiourea.64
p-polarization
4.3 Spectral acquisition
Band intensity/a.u.
Enhanced spectra can be measured with transmission, ATR,
and external reflection configurations. Multiple ATR and
s-polarization
external reflection configurations are useful to increase the
band intensities further.3,7,54,62,65 Since the metal island
films strongly absorb IR radiation, however, the multiple
c
reflection configurations can be used only for very thin
metal films.7 For metallic colloids, the diffuse reflection
mode is also available.18 0 30 60 90
A suitable subtraction of the strong background absorp- (a) Incident angle ()
tion of the metal is necessary for detecting weak vibrational
modes. In metal-underlayer measurements, a spectrum of
the metal film acquired before molecules are adsorbed is
used as the reference. In electrochemical systems, the refer- p
ence spectrum is acquired at potentials at which molecules
are totally desorbed from the surface. In metal-overlayer s
experiments, the spectrum acquired before the metal depo- Band intensity/a.u.
sition is used as the reference.
In both ATR and external reflection measurements, the p-polarization
observed band intensity depends strongly on the incident
angle q and polarization of the IR beam.5,8,43 The angle
and polarization dependences are described below for each
configuration. Here, p- and s-polarizations are defined as the s-polarization
electric field being parallel and perpendicular, respectively,
to the plane of incidence.
In Figure 7, the intensity of the symmetric NO2 stretch- 0 30 60 90
ing band of PNBA observed with ATR configuration is (b) Incident angle ()
plotted as a function of q.8 Solid and open circles repre- Figure 7. Peak intensity of the NO2 stretching band of PNBA
sent the data for p- and s-polarization, respectively. The adsorbed on (a) 5.5 and (b) 12 nm thick Ag island films measured
prism used in the measurements is a Ge hemicylinder, the with ATR configuration plotted as a function of incident angle (q)
refractive index (n) of which is 4. On a 5.5 nm thick Ag for p- and s-polarization (solid and open circles, respectively).
film (Figure 7a), the intensity is nearly independent of the The data were taken from Suzuki et al.8 The solid curves are the
q dependence simulated with the (a) MaxwellGarnett and (b)
polarization and decreases as q increases, except around the Bruggemann effective medium approximation. See Osawa et al.13
critical angle of total reflection [qc D sin1 1/n D 14.48 ], for details of the simulations.
where the intensity for p-polarization decays sharply. On
a 12 nm thick Ag film (Figure 7b), on the other hand, the but positive (i.e. normal) absorption is observed at higher
absorption is observed only for p-polarization and increases angles. The dependence of polarization and q has been
as q increases. The polarization and q dependences of the simulated with the EM model mentioned above.43 The
band intensity are well simulated by the theoretical EM simulation reveals that the change of sign of the absorption
calculations,13 as shown in Figure 7 by solid curves. for p-polarization occurs at the Brewster angle [qB D
Figure 8 shows a set of spectra of PNBA adsorbed on tan1 n D 54.46 ]. The intensity reaches a maximum
an Ag-coated CaF2 (n D 1.4) measured with the external around qB for p-polarization. However, since the reflectivity
reflection configuration at different angles of incidence q.43 of the substrate is lowest at qB for p-polarization, the
For s-polarization, all vibrational modes are observed as signal-to-noise ratio (S/N) of the spectrum is lower than
reflection maxima (i.e. negative absorption) independent at larger and smaller angles. In addition, the superposition
of q. The same negative spectral feature is observed also of up-going and down-going bands results in bipolar band
for p-polarization at angles lower than a certain angle, shapes (see the spectrum at q D 45 ).

log(R /R0) 0.1
log(R /R0) 0.1
= 85
75
= 85
65
75
1390
1350
65
55
45
55
45
35
25 35
2200 1800 1400 1000 600 2200 1800 1400 1000 600
Figure 8. External reflection SEIRA spectra of a PNBA layer on a 5 nm thick Ag film evaporated on BaF2 for (a) p- and (b) s-
polarization.43 Incident angles (q) used are shown. [Reproduced by permission of the American Chemical Society from Y. Nishikawa,
K. Fujiwara, K. Ataka and M. Osawa, Anal. Chem., 65, 556 (1993).]
4.4 Determination of molecular orientation I a1 / cos2 q I a1 2

I b1 / cos2 q sin2 c I b1 3
The orientation of molecules adsorbed on island films
can be determined based on the surface selection rule.20 I b2 / sin2 q sin2 c I b2 4
Recalling that the surface electric field, E, effectively
has only a normal component, the intensity of a band is where I represents the intrinsic intensity of the bands.
proportional to the square of the scalar product of the From these three equations, the following two equations
electric field and the dipole moment derivative, dm/dQ, as can be introduced:
2 2
dm dm I b1 I a1 1
I / E D jE j2 cos2 a 1 tan2 q D 5
dQ dQ I a1 I b1 cos2 c
I b2 I b1
where a is the angle between E and dm/dQ and is the tan2 c D 6
surface concentration of the molecule. I b1 I b2
For a further detailed treatment of the surface selection
rule, we use benzenethiol adsorbed on a metal surface as
Z
an example. The coordinate system is defined in Figure 9,
x (b1)
where q is the tilt angle of the z-axis of the molecular system z (a1)
from the surface normal (Z) and c is the twist angle of
c
the molecular plane around the z-axis (which is 0 when
y is parallel to the surface). In-plane a1 and b2 modes
have dipole moment derivatives along the z- and y-axes, S Y y (b2)
respectively. On the other hand, out-of-plane b1 modes have

dipole moment derivatives perpendicular to the phenyl ring X
(along the x-axis). The experimental axis system XYZ and
the molecular-fixed axis system xyz can be connected with
three Eulerian angles q, c, and the rotation angle f of the Figure 9. Definition of experimental (XYZ) and molecular (xyz)
systems for benzenethiolate adsorbed on a metal surface. The two
molecular system around Z.66 By using the three Eulerian axis systems can be connected with three Eulerian angles q, c and
angles, equation (1) can be rewritten for the three symmetry f. Symmetry species a1 , b1 and b2 have dipole derivatives along
species as follows: z, x and y directions, respectively.
By substituting band intensities in the enhanced spectrum The other is that peak positions and relative intensities
I and those in normal spectrum of randomly oriented in enhanced spectra are generally different from those of
molecules I (e.g. a KBr spectrum of the corresponding normal spectra of the same molecules, which is significant
molecule or its metal complex) into these equations, molec- for molecules that strongly chemisorb on metals. Therefore,
ular orientation can be determined. the existing databases of (normal) IR spectra cannot be used
If the low-frequency region where out-of-plane modes directly and new databases must be built.
are located cannot be measured owing to the strong absorp-
tion of the substrate, molecular orientation can be estimated
semiquantitatively by comparing the band intensities with 5.2 Biological applications
the surface concentration determined with independent
techniques.67,68 SEIRAS is promising for biosensing and immunoassays.
Gold island films or colloids modified with antibodies
were used to study interactions with antigens.19,71 The
5 APPLICATIONS localization of bacteria on geological material surfaces
was investigated by evaporating Au island films on the
5.1 Analysis of trace chemicals samples and analyzing them with a microspectrometer.72
The structural differences between the top and bottom of
biomembranes can also be investigated by using metal-
IR spectroscopy has been established as a routine mole-
overlayer and -underlayer configurations.73
cular-specific technique for qualitative and quantitative
The adsorption and orientation of nucleic acid bases,
purposes. However, the relatively low molar absorptivities
such as thymine (5-methyluracil),59 cytosine,74 and uracil,75
in the IR region limit its use in many applications.
on metal surfaces have been investigated by SEIRAS. The
Therefore, SEIRAS, which lowers the detection limit of IR
adsorption of proteins and bacteria at Cu/aqueous solution
spectroscopy, is of great interest for the analysis of trace
interfaces and biochemical corrosion of Cu have also been
molecules.
investigated.76,77
The detection limit depends on the chemical nature of the
analyte, methods of preparation of the metal film, measuring
configurations, etc. The lowest detection limit ever reported 5.3 Characterization of thin organic films and
is 10 pg cm2 (6 1014 mol cm2 ) for PNBA, which solid surfaces
was achieved with a multireflection ATR configuration.65
In transmission measurements, it typically ranges from Since small metal particles show high catalytic activities in
nanograms to micrograms, which is one to two orders many reactions, SEIRAS is very promising as a new tool for
of magnitude lower than in conventional measurements catalytic science. From this point of view, SEIRA studies
without SEIRA-active metals.49,50 If the analyte could be on the adsorption and oxidation of CO, methanol, and some
spotted in a very small area on a SEIRA-active surface, other molecules on Pt-group metal electrodes16,37,39 42 are
detection limits can be reduced to the subpicogram level important and will contribute to, for example, fuel cell
by using an IR microscope.49,50,56 The detection limit for technology. SEIRA studies of molecular adsorption and
molecules that do not chemisorb on metals is much higher, photochemical reactions on metal island films in ultrahigh
but it can be lowered by modifying the metal surface with vacuum have also been reported.14,53
a suitable molecule such as benzenethiol.69 Concentration Self-assembled monolayers (SAMs) of thiols or
of the analyte into the benzenethiol monolayer via strong disulfides on metal surfaces are good targets for ana-
molecular interactions was proposed for this reason. lysis by SEIRAS.20,70,78 80 This technique was used
The SEIRA-based technique was used as a detec- successfully to investigate the acidbase equilibria
tion system for flow-injection analysis and applied to of SAMs in solution.79,80 LB films have also been
the analysis of trace environmentally hazardous chem- characterized.7,10,43,48,81
icals in waste water.52,54,62 The use of SEIRAS as a Considerable attention has been focused on the applica-
gas sensor65 and detection systems for liquid and gas tion of SEIRAS to the surface analysis of nonconducting
chromatography49,50,56,70 have also been proposed. and low-reflecting materials, such as polymers, semicon-
For quantitative analysis, the following two issues should ductors and glasses, because very few surface analytical
be noted. One is that the enhancement is short range and techniques can be applied to these samples. Although
the band intensity is not a linear function of the amount IR-RAS is very useful for surface analysis of highly reflect-
of molecules (Figure 2). SEIRAS is effective only for trace ing metals, its sensitivity is not so high on low-reflecting
amounts of chemicals existing on or near the metal surface. materials. The ATR technique is used most frequently
for polymers. However, conventional ATR is not surface IR-transparent cell window to reduce the strong absorption
sensitive because the penetration depth of the evanescent of the electrolyte solution, as shown in Figure 10(a). The
wave is of the order of the wavelength. The strong signal thickness of the solution layer between the electrode and the
from the bulk prevents the detection of the weak signal window is typically 110 m. Polycrystalline and single-
from the surface. The surface sensitivity of the conven- crystal metals and glassy carbon can be used as the
tional ATR technique can be improved significantly by working electrodes. This technique was used in observing
using metal-underlayer geometry.82 In this measurement, SEIRA of CO and several other molecules adsorbed on
the sample is pressed against a metal-coated ATR ele- electrochemically deposited island films of Pt, Pd, Rh and
ment. This technique was used for the characterization of Ir.16,37,40,41
plasma-modified polymer surfaces.83 Unfortunately, since However, this technique has two serious problems. First,
good contact between the ATR element and the sample is mass transportation between the thin layer and reservoir is
required, this technique is applicable only to soft materials hampered and the electrochemical system does not respond
with smooth surfaces. quickly to externally applied potential changes owing to the
For the characterization of very rough and rounded large solution resistance. Therefore, this technique can be
surfaces of hard materials to which ATR-SEIRAS cannot be used only under static or quasi-static conditions. Second,
applied, an external reflection method with metal-overlayer the solution layer is much thicker than a monolayer on the
geometry was developed.43 Thin organic films on plastics electrode. The signals from the adsorbates are superimposed
and glass have been characterized with this technique. on the absorption bands of the solution, but are about
SEIRAS is so surface sensitive that the signals of both the three orders of magnitude stronger. Although the solution
substrate surface and molecular layer on it are enhanced.43 background can be substantially removed by measuring
To eliminate the signal from the substrate completely, a potential difference spectra,87 the complete subtraction of
SEIRA-based transfer technique was developed, in which the solution background is very difficult.
a metal-coated IR-transparent substrate is softly pressed These problems can be removed by using the ATR
onto the sample surface and the analyte is transferred configuration shown in Figure 10(b). A thin metal film
to the metal surface.51 This technique was used for the deposited on a prism is used as the working electrode.
analysis of trace chemicals on polymers,51 secretions and Since the evanescent wave penetrates into the solution
sweat on human skins,84 and residual agricultural chemicals phase only a few hundred nanometers,76 the solution
on fruits.85 A similar measurement using Ag-coated KBr phase is not required to be thin and further the solution
powder has been reported.86 Trace amounts of chemicals background can be reduced significantly compared with IR-
on a sample solid surface were transferred on to the Ag- RAS. On the other hand, the absorption of species adsorbed
coated KBr powder by abrading the sample surfaces with on the electrode surface is enhanced due to SEIRA.
the powder. The collected powder was analyzed using the
diffuse reflection mode. The sensitivity of these transfer IR beam
methods is reported to be equivalent to that of X-ray Si rubber CaF2
110 m
photoelectron spectroscopy.
Reference Counter
electrode electrode
5.4 In situ study of electrochemical dynamics
Solution
Research in electrochemistry has increasingly involved the (a) Working electrode

use of surface-sensitive analytical techniques in combi-
Counter electrode
nation with conventional electrochemical techniques. The
chief reason for this development is that electrochemical
techniques inevitably measure the sum of all surface pro- Solution
Reference
cesses. The results primarily provide kinetic information electrode
and no direct molecular information on the species involved Thin metal film
in the reactions can be obtained. SEIRAS has been applied O-ring (working electrode)
very successfully to examine molecular events occurring at
IR beam
the electrochemical interface.
The IR-RAS technique has been used most frequently for (b) Prism
in situ IR studies of the electrochemical interface.87 In this Figure 10. In situ monitoring of the electrochemical interface
measurement, the working electrode is pushed against an with (a) IR-RAS and (b) ATR/SEIRAS.
Consequently, the signals from the adsorbates and the as R/R D Rsample Rref /Rref , where Rsample and Rref
solution background are comparable in the ATR/SEIRAS represent the reflectivity of the electrode at the sample
measurements,88 which facilitates the complete subtraction and reference potentials, respectively. On the other hand,
of the solution background. the ATR/SEIRA spectra are presented in the absorbance
To demonstrate the advantages of ATR/SEIRAS, IR spec- units defined as A D log Rsample /Rref . In both measure-
tra of 2,20 -bipyridine adsorbed on an Au(111) single-crystal ments, the reference potential used is 0.75 V versus an
electrode and a vacuum-evaporated 20 nm thick Au island SCE at which the molecule is totally desorbed from the
film measured with IR-RAS89 and ATR/SEIRAS,68 respec- surface. Therefore, the adsorbate on the electrodes gives
tively, are compared in Figure 11. The IR-RAS spectra down-going bands in Figure 11(a) and up-going bands in
are represented in the relative reflectance change defined Figure 11(b). Several adsorbate bands are clearly identified
in both sets of spectra. The frequencies of the bands are
identical in the two measurements. However, the adsor-
0.0005 bate is observed about 10 times more strongly by SEIRAS
E2/mV
than by IR-RAS (note that a reflectance change of 0.0005
350 corresponds to 0.0002 absorbance).
1591 1464
1483 It is important to note that several up-going bands are
250
observed in the IR-RAS spectra. Since the adsorption of the
1570 1305 molecule on the electrode surface reduces its concentration
1437
in the solution and the mass transportation between the thin
R /R
150 layer and the reservoir is restricted, the up-going bands are
1425
assigned to the molecule in the solution. The spectrum of
50 this molecule is fortunately greatly changed on adsorption
on the electrode owing to the conformational change from
0
trans to cis.68,89 Therefore, bands of the adsorbate and
solution species are clearly distinguished in the present
50
case. However, if the spectral changes are small, the overlap
of up-going and down-going bands makes the interpretation
1600 1500 1400 1300 1200 of the spectra obtained difficult. In addition, a strong
(a) Wavenumber /cm1 solution background strongly interferes with the IR-RAS
measurements. To avoid the interference from the solution,
1592 1484 1179 D2 O was used as the solution in the IR-RAS measurements
0.005 1424 1307 * in Figure 11(a). ATR/SEIRAS is free from this problem and
the bands of the molecule in the solution are completely
Absorbance
0.8 missing from Figure 11(b). By the complete subtraction of

the solution background, the bending mode of water that
was removed from the interface by the adsorption of 2,20 -
E
SC
+0.6 bipyridine is clearly observed as the down-going peak at

vs
1620 cm1 , which is much lower than that of bulk water

E/V
(1645 cm1 ).
1620 (HOH) 1564
0.8
ATR/SEIRAS has been applied to several electrochemi-
1800 1600 1400 1200 cal systems.1,9,23,57,67,68,74 80,88,90 99 The details of exper-
imental setup and procedures can be found elsewhere.88,94
(b) Wavenumber /cm1
Si,57,90,94,95 Ge88,92 and ZnSe75 have been used as the
Figure 11. IR spectra of 2,20 -bypyridine adsorbed on (a) Au(111) ATR elements. Ge and ZnSe have wider spectral windows
single-crystal electrode89 and (b) a vacuum-evaporated (111) ori-
ented 20 nm thick Au film electrode68 in 0.1 M HClO4 containing than Si, but Si is the most stable among these materi-
1 mM 2,20 -bypyridine measured with IR-RAS and ATR/SEIRAS, als in electrolyte solutions. Although the electrodes used
respectively. Reference potential is 0.75 V vs saturated calomel in the ATR/SEIRAS measurements are island films, rela-
electrode (SCE) in both measurements. [Reproduced by permis- tively thick (1520 nm) island films have sufficient con-
sion of (a) the American Chemical Society from M. Hoon-Khosla, ductivity for electrochemistry (note that slow deposition
W.R. Fawcett, J.D. Goddard, W.-Q. Tian and J. Lipowski, Lang-
muir, 16, 2356 (2000) and (b) Elsevier Science from H. Noda, is crucial in preparing SEIRA-active thin-film electrodes;
T. Minoha, L.-J. Wan and M. Osawa, J. Electroanal. Chem., 481, see Figure 6). Evaporated metal films are generally poly-
62 (2000).] crystalline, except for Au. The Au surface has a tendency
to recrystallize towards a (111) single-crystal orientation. symmetric COO stretching are seen at 1720 and 1380 cm1 ,
The recrystallization of the Au surface towards (111) can respectively. The presence of the two bands indicates
be accelerated by annealing the film briefly (for about that fumaric acid is adsorbed on the surface via one of
10 s) with a hydrogenoxygen flame without reducing the the two carboxylic groups. The band intensities can be
enhancement.97 Electrodes are used after cleaning their sur- used for kinetic analysis of the adsorption and desorp-
faces in solutions by a repeated oxidationreduction cycle tion, which cannot be readily accessible by conventional
of potential. The oxidationreduction cycle cleaning in electrochemical techniques owing to the flow of double-
solutions containing halide ions is not recommended in layer charging current.96 Electrochemical reactions can also
order to avoid complex electrochemistry. If ultrapure water be investigated.92,93 By applying two-dimensional correla-
is used, the surface can be kept clean for several hours. tion analysis, detailed information on the reactions can be
Since the S/N of a spectrum is proportional to the extracted from time-resolved spectra.93,95
square root of the number of interferograms co-added, Developments in FT-IR instrumentation have realized
one will recognize immediately that the high sensitivity of time-resolved spectral measurements on the timescale of
ATR/SEIRAS facilitates real-time or time-resolved mon- submicroseconds. However, it should be noted that the
itoring of dynamic processes at the interface. The quick response of conventional electrochemical systems is limited
response of the electrochemical ATR cell is also convenient to a few milliseconds owing to the charging of the double
for such measurements. In general, S/Ns good enough for layer.99 Therefore, time-resolved measurements much faster
detailed analyses of monolayer adsorbates can be obtained than the double-layer charging are meaningless in most
by the co-addition of less than 50 interferograms. For cases even if very fast data acquisition systems are avail-
strongly absorbing molecules, high-quality spectra can be able. For kinetic studies of fast reactions that are limited
obtained without interferogram co-addition.23 The short by slower processes, potential modulation tactics are used
acquisition time allows the accumulation of a series of spec- in electrochemistry. In alternating current (ac) impedance
tra under potential modulations simultaneously with electro- spectroscopy, for example, potential is modulated and the
chemical data (e.g. cyclic voltammograms). A comparison phase shift of the corresponding current flow with respect to
of spectral and electrochemical data is important for obtain- the modulation is measured, from which kinetic parameters
ing deeper insights into electrochemistry.23,67,68,74,92 98 can be estimated. The idea of replacing the current measure-
If the system to be investigated is reversible, very fast ments in ac impedance spectroscopy by IR measurements
processes occurring at the interface can be monitored with for obtaining dynamic information of the interface has been
microsecond time resolution by using step-scan Fourier examined by several groups. In the case of FT-IR spec-
transform infrared (FT-IR) spectroscopy.92,96,98 Figure 12 troscopy, the key for this measurement is how to remove
shows a series of SEIRA spectra demonstrating the adsorp- cross talk between Fourier frequencies and the poten-
tion and desorption of fumaric acid at an Au(111) surface tial modulation frequencies. This problem was removed
in HClO4 associated with a double-potential step. The by the use of a very slow-scan FT-IR spectrometer100,101
time resolution used was 100 s. Both CDO stretching and or a step-scan FT-IR102,103 spectrometer. Unfortunately,
s(COO)
(C=O)
25
20
15
s
10
m
e/
m
5
Ti
0
5
1800 1600 1400 1200
Wavenumber /cm1
Figure 12. Series of time-resolved IR spectra showing the adsorption and desorption of fumaric acid on an Au(111) electrode in 0.1 M
HClO4 C 20 mM fumaric acid for a double-potential step from 0.3 to 1.2 V (15 ms) and back to 0.3 V. Time resolution was 100 s.
several other technical problems have prevented its wider resonance Raman spectra,104 different conclusions might be
application. However, this potential-modulated FT-IR spec- reached if the SEIRA spectrum were not available. As is
troscopy has been advanced significantly by coupling with seen from this example, a comparison of SEIRA and SER
ATR/SEIRAS.99 spectra is very helpful for the comprehensive analysis of
Potential-modulated FT-IR spectroscopy coupled with both types of spectra.
ATR/SEIRAS was used to study potential-dependent spec-
tral changes for a 4-mercaptopyridine SAM on an Au
electrode.99 From a mathematical analysis of the IR data ABBREVIATIONS AND ACRONYMS
measured at modulation frequencies ranging from 40 to
100 kHz, it was found that the spectral changes occur at
ac Alternating Current
a rate of 5 105 s1 , which is much larger than the time
EDTA Ethylenediaminetetraacetic Acid
constant of the double-layer charging (1.9 103 s1 ).
EM Electromagnetic
Vibrational frequencies of molecules adsorbed on elec-
LB LangmuirBlodgett
trodes generally shift depending on the applied potential.
PNBA p-Nitrobenzoic Acid
For CO on metal electrodes, Stark tuning and back-donation
PNBA p-Nitrobenzoate
models have been proposed.87 The Stark tuning model
QCM Quartz Crystal Microbalance
predicts that the change of the electric field across the
SAM Self-assembled Monolayer
double-layer affects the vibrational properties of the adsor-
SCE Saturated Calomel Electrode
bate. The back-donation model predicts that the change in
SEIRAS Surface-enhanced Infrared Absorption
the potential (that is, Fermi level of the electrode) changes
Spectroscopy
the degree of the charge transferred from the metal to
the 2p orbital of CO, resulting in spectral changes. If
the Stark tuning effect is dominant, then the rate of the
spectral changes should be the same as that of the double- REFERENCES
layer charging (i.e. the change of actual potential at the
interface). In the case of 4-mercaptopyridine, the rate of 1. M. Osawa, Bull. Chem. Soc. Jpn., 70, 2861 (1997).
the spectral changes is much faster than the double-layer 2. M. Osawa and M. Ikeda, J. Phys. Chem., 95, 9914 (1991).
charging, indicating that the charge-transfer mechanism 3. A. Hartstein, J.R. Kirtley and J.C. Tsang, Phys. Rev. Lett.,
contributes to the spectral changes rather than the Stark 45, 201 (1980).
tuning mechanism. 4. A. Hatta, T. Ohshima and W. Suetaka, J. Appl. Phys. A, 29,
71 (1982).
5. A. Hatta, Y. Suzuki and W. Suetaka, Appl. Phys. A, 35, 135
6 CONCLUSION (1984).
6. M. Osawa, M. Kuramitsu, A. Hatta, W. Suetaka and H. Seki,
So far, SEIRA has received less attention than SERS. Surf. Sci. Lett., 175, L787 (1986).
This is probably due to the smaller enhancement factor 7. T. Kamata, A. Kato, J. Umemura and T. Takenaka, Lang-
of SEIRA. However, IR spectroscopy inherently has a muir, 3, 1150 (1987).
higher sensitivity than Raman spectroscopy, and hence the
8. Y. Suzuki, M. Osawa, A. Hatta and W. Suetaka, Appl. Surf.
sensitivities of SEIRAS and SERS are comparable. Both Sci., 33/34, 875 (1988).
techniques can detect submonolayer amounts of chemicals
9. T. Wadayama, T. Sakurai, S. Ichikawa and W. Suetaka,
on metal surfaces. It should be noted that SEIRAS and Surf. Sci., 198, L359 (1988).
SERS are not competitive but are complementary to each
10. E. Johnson and R. Aroca, J. Phys. Chem., 99, 9325 (1995).
other, as in the case of normal IR and Raman spectroscopy.
For example, the SER spectrum of p-aminothiophenol 11. G.T. Merklin and P.R. Griffiths, Langmuir, 13, 6159 (1997).
adsorbed on Ag is completely different from the normal 12. M. Osawa and K. Ataka, Surf. Sci. Lett., 262, L118 (1992).
Raman spectrum of this molecule in the solid state.104 Some 13. M. Osawa, K. Ataka, K. Yoshii and Y. Nishikawa, Appl.
reactions or significant changes in vibrational properties Spectrosc., 47, 1497 (1993).
due to the adsorption were considered first for this reason. 14. O. Krauth, G. Fahsold and A. Pucci, J. Chem. Phys., 110,
However, the SEIRA spectrum of the same sample revealed 3113 (1999).
that the adsorption on Ag changes the vibrational properties 15. Y. Nishikawa, T. Nagasawa, K. Fujiwara and M. Osawa,
of this molecule only slightly. Although the strange SERS Vib. Spectrosc., 6, 43 (1993).
features were found to arise from a photo-driven charge- 16. A.E. Bjerke, P.R. Griffiths and W. Theiss, Anal. Chem., 71,
transfer mechanism from a comparison with ultraviolet 1967 (1999).
17. S.Y. Kang, I.C. Jeon and K. Kim, Appl. Spectrosc., 52, 278 46. P. Dumas, R.G. Tobin and P. Richards, Surf. Sci., 171, 555
(1998). (1986).
18. S.J. Lee and K. Kim, Vib. Spectrosc., 18, 187 (1998). 47. J.P. Devlin and K. Consani, J. Phys. Chem., 85, 2597
19. J.A. Seelenbinder, C.W. Brown, P. Pivarnik and A.G. Rand, (1981).
Anal. Chem., 71, 1963 (1999). 48. K. Ito, K. Hayashi, Y. Hamanaka, M. Yamamoto, T. Araki
20. L.-J. Wan, M. Terashima, H. Noda and M. Osawa, J. Phys. and K. Iriyama, Langmuir, 8, 140 (1992).
Chem. B, 104, 3563 (2000). 49. Y. Nishikawa, K. Fujiwara and T. Shima, Apl. Spectrosc.,
21. F.M. Hoffmann, Surf. Sci. Rep., 3, 107 (1983). 44, 691 (1990).
22. A. Hatta, N. Suzuki, Y. Suzuki and W. Suetaka, Appl. Surf. 50. Y. Nishikawa, K. Fujiwara and T. Shima, Appl. Spectrosc.,
Sci., 37, 299 (1989). 45, 747 (1991).
23. M. Osawa and K. Yoshii, Appl. Spectrosc., 51, 512 (1997). 51. Y. Nishikawa, Y. Ito, N. Yamakami, K. Fujiwara and
T. Shima, Surf. Interface Anal., 18, 481 (1992).
24. R.K. Chang and T.E. Furutak (eds), Surface Enhanced
Raman Scattering, Plenum, New York (1982), and refer- 52. R. Kellner, B. Mizaikoff, M. Jakusch, H.D. Wanzenebock
ences cited therein. and N. Weissenbacher, Appl. Spectrosc., 51, 495 (1997).
53. S. Sato and T. Suzuki, Appl. Spectrosc., 51, 1170 (1997).
25. H. Metiu, Prog. Surf. Sci., 17, 153 (1984).
54. H.D. Wanzenbock, B. Mizaikoff, N. Weissenbacher and
26. M. Moskovits, Rev. Mod. Phys., 57, 783 (1985).
R. Kellner, Fresenius J. Anal. Chem., 362, 15 (1998).
27. A. Wakaun, Surface-enhanced Electromagnetic Processes,
55. T. Wadayama, O. Suzuki, Y. Suzuki and A. Hatta, Appl.
in Solid State Physics, eds H. Ehrenreich and D. Turnbull,
Phys. A, 64, 501 (1997).
Academic Press, New York, 223294, Vol. 38 (1984).
56. M. Osawa, K. Ataka, M. Ikeda, H. Uchihara and R. Namba,
28. J.I. Gersten and A. Nitzan, Surf. Sci., 158, 165 (1985).
Anal. Sci., 7(Suppl.), 503 (1991).
29. C. Kuhne, G. Steiner, W.B. Fischer and R. Salzer, Frese-
57. A. Hatta, Y. Chiba and W. Suetaka, Surf. Sci., 158, 616
nius J. Anal. Chem., 360, 750 (1998).
(1985).
30. G.A. Niklasson and C.G. Granqvist, J. Appl. Phys., 55, 3382
58. Y. Suzuki, K. Sagisaka and A. Hatta, Appl. Surf. Sci., 84, 1
(1984).
(1995).
31. For example, W.N. Hansen, Infrared Reflection Spec-
59. R. Aroca and R. Bujalski, Vib. Spectrosc., 19, 11 (1999).
troscopy in Electrochemistry, in Advances in Electrochem-
istry and Electrochemical Engineering, eds P. Delahay and 60. T.R. Jensen, R.P.V. Duyne, S.A. Johnson and V.A. Maroni,
C.W. Tobias, John Wiley & Sons, New York, 159, Vol. 9 Appl. Spectrosc., 54, 371 (2000).
(1973). 61. N.A.F. Al-Rawashdeh, M.L. Sandrock, C.J. Seugling and
32. D.A.G. Bruggemann, Ann. Phys., 24, 636 (1935). J.C.A. Foss, Langmuir, 102, 361 (1998).
33. Y. Nakao and H. Yamada, Surf. Sci., 176, 578 (1986). 62. H.D. Wanzenbock, B. Mizaikoff, N. Weissenbacher and
R. Kellner, J. Mol. Struct., 410411, 535 (1997).
34. Y. Nakao and H. Yamada, J. Electron Spectrosc. Relat.
Phenom., 45, 189 (1987). 63. F. Maroun, F. Ozanam, J.-N. Chazalviel and W. Theiss, Vib.
Spectrosc., 19, 193 (1999).
35. R. Aroca and B. Price, J. Phys. Chem. B, 101, 6537 (1997).
64. G. Mattei, Y. Mo, M. Pagannone and V.A. Vladimir, Mikro-
36. T. Yoshidome, T. Inoue and S. Kamata, Chem. Lett., 533
chim. Acta, 14(Suppl.), 653 (1997).
(1997).
65. S.A. Johnson, N.-H. Pham, V.J. Novick and V.A. Maroni,
37. G.-Q. Lu, S.-G. Sun, S.P. Chen and L.-R. Cai, J. Elec- Appl. Spectrosc., 51, 1423 (1997).
troanal. Chem., 421, 19 (1997).
66. E.B. Wilson, Jr, J.C. Decius and P.C. Cross, Molecular
38. S. Sato, K. Kamada and M. Osawa, Chem. Lett., 15 (1999). Vibrations, McGraw-Hill, New York, 285286 (1955).
39. Y. Zhu, H. Uchida and M. Watanabe, Langmuir, 15, 8757 67. W.-B. Cai, L.-J. Wan, H. Noda, Y. Hibino, K. Ataka and
(1999). M. Osawa, Langmuir, 14, 6992 (1998).
40. R. Oritz, A. Cuesta, O.P. Marquez, J. Marquez, J.A. Mendez 68. H. Noda, T. Minoha, L.-J. Wan and M. Osawa, J. Elec-
and C. Gutierrez, J. Electroanal. Chem., 465, 234 (1999). troanal. Chem., 481, 62 (2000).
41. G.-Q. Lu, S.-G. Sun, L.-R. Cai, S.-P. Chen and Z.-W. Tian, 69. J.A. Seelenbinder, C.W. Brown and D.W. Urish, Appl. Spec-
Langmuir, 16, 778 (2000). trosc., 54, 366 (2000).
42. M. Watanabe, Y. Zhu and H. Uchida, J. Phys. Chem., 104, 70. G.T. Merklin, L.-T. He and P.R. Griffiths, Appl. Spectrosc.,
1762 (2000). 53, 1448 (1999).
43. Y. Nishikawa, K. Fujiwara, K. Ataka and M. Osawa, Anal. 71. C.W. Brown, Y. Li, J.A. Seelenbinder, P. Pivarnik, A.G.
Chem., 65, 556 (1993). Rand, S.V. Letcher, O.J. Gregory and M.J. Platek, Anal.
44. B.N.J. Persson and R. Ryberg, Phys. Rev. B, 24, 6954 Chem., 70, 2991 (1998).
(1981). 72. H.-Y.N. Holman, D.L. Perry and J.C. Hunter-Cevera, J.
45. B.N.J. Persson and A. Liebsch, Surf. Sci., 110, 356 (1981). Microbiol. Methods, 34, 59 (1998).
73. C. Kuhne, G. Steiner, W.B. Fischer and R. Salzer, Frese- 89. M. Hoon-Khosla, W.R. Fawcett, J.D. Goddard, W.-Q. Tian
nius J. Anal. Chem., 360, 750 (1998). and J. Lipowski, Langmuir, 16, 2356 (2000).
74. K. Ataka and M. Osawa, J. Electroanal. Chem., 460, 188 90. A. Hatta, Y. Chiba and W. Suetaka, Appl. Surf. Sci., 25, 327
(1999). (1986).
75. M. Futamata and D. Diesing, Vib. Spectrosc., 19, 187 91. T. Wadayama, Y. Momota, A. Hatta and W. Suetaka, J.
(1999). Electroanal. Chem., 289, 29 (1990).
76. K.P. Ishida and P.R. Griffiths, Anal. Chem., 66, 522 (1994). 92. M. Osawa, K. Yoshii, K. Ataka and T. Yotsuyanagi, Lang-
77. K.P. Ishida and P.R. Griffiths, J. Colloid Interface Sci., 213, muir, 10, 640 (1994).
513 (1999).
93. M. Osawa, K. Yoshii, Y. Hibino, T. Nakano and I. Noda, J.
78. N. Matsuda, K. Yoshii, K. Ataka, M. Osawa, T. Matsue and Electroanal. Chem., 426, 11 (1997).
I. Uchida, Chem. Lett., 1385 (1992).
94. K. Ataka, T. Yotsuyanagi and M. Osawa, J. Phys. Chem.,
79. I. Taniguchi, S. Yoshimoto, Y. Sunatani and K. Nishiyama, 100, 10 664 (1996).
Electrochemistry, 67, 1197 (1999).
95. K. Ataka and M. Osawa, Langmuir, 14, 951 (1998).
80. L.-J. Wan, H. Noda, Y. Hara and M. Osawa, J. Electroanal.
Chem., 489, 68 (2000). 96. H. Noda, K. Ataka, L.-J. Wan and M. Osawa, Surf. Sci.,
81. J. Zhao, J. Zhang, H.-X. He, H.-L. Li and Z.-F. Liu, Chem. 427428, 190 (1999).
Phys. Lett., 278, 220 (1997). 97. S.-G. Sun, W.-B. Cai, L.-J. Wan and M. Osawa, J. Phys.
82. Y. Nishikawa, Y. Ito, K. Fujiwara and T. Shima, Appl. Chem., 103, 2460 (1999).
Spectrosc., 45, 752 (1991). 98. K. Ataka, G. Nishina, W.-B. Cai, S.-G. Sun and M. Osawa,
83. S. Geng, J. Freidrich, J. Gahde and L. Guo, J. Appl. Polym. Electrochem. Commun., 2, 417 (2000).
Sci., 71, 1231 (1999). 99. K. Ataka, Y. Hara and M. Osawa, J. Electroanal. Chem.,
84. Y. Nishikawa, K. Fujiwara, M. Osawa and K. Takamura, 473, 34 (1999).
Anal. Sci., 9, 811 (1993).
100. F. Ozanam and J.-N. Chazalviel, Rev. Sci. Instrum., 59, 242
85. Y. Nishikawa, K. Fujiwara and K. Takamura, Bunseki (1988).
Kagaku, 43, 425 (1994).
101. J.-N. Chazalviel, V.M. Dubin, K.C. Mandel and F. Ozanam,
86. N. Makino, K. Mukai and Y. Kataoka, Appl. Spectrosc., 51, Appl. Spectrosc., 64/65, 1411 (1993).
1460 (1997).
102. B.O. Budeveska and P.R. Griffiths, Anal. Chem., 65, 2963
87. R.J. Nichols, IR Spectroscopy of Molecules at the (1993).
SolidSolution Interface, in Adsorption of Molecules at
Metal Electrodes, eds J. Lipowski and P.N. Ross, VCH, 103. C.M. Pharr and P.R. Griffiths, Anal. Chem., 69, 4665
New York, 347389 (1992), and references cited therein. (1997).
88. M. Osawa, K. Ataka, K. Yoshii and T. Yotsuyanagi, J. Elec- 104. M. Osawa, N. Matsuda, K. Yoshii and I. Uchida, J. Phys.
tron Spectrosc. Relat. Phenom., 64/65, 371 (1993). Chem., 98, 12 702 (1994).
Other Instrumental Approaches for Vibrational
Spectroscopy
Sum-frequency Spectroscopy
David A. Beattie and Colin D. Bain
University of Oxford, Oxford, UK
1 INTRODUCTION There are obvious similarities between SFG and Raman

scattering, but there are also a number of important dif-
Sum-frequency spectroscopy (SFS) is a form of surface ferences. First, SFG is a coherent process whereas normal
vibrational spectroscopy based on the second-order non- Raman scattering is incoherent; the electric fields of the
linear optical process of sum-frequency generation (SFG). oscillating dipoles are in phase in SFG and light is emit-
A laser beam in the mid-infrared is overlapped at an inter- ted as a coherent beam in a well-defined direction. In this
face with a second laser beam (usually in the visible or the way, SFG resembles coherent anti-Stokes Raman scattering
near-infrared) and the light generated at the sum of the two (CARS), although SFG is a second-order nonlinear optical
input frequencies is detected. The sum-frequency (SF) sig- effect and CARS is a third-order effect. Second, vibrational
nal from the molecules at the interface is enhanced when the modes have to be both Raman and IR active to be observed
mid-infrared laser is in resonance with a vibrational transi- in SFS. Third, SFG (unlike Raman scattering or IR absorp-
tion. If one scans the wavelength of the IR laser and records tion) is forbidden in centrosymmetric and isotropic media;
the intensity of the SF signal, a vibrational spectrum of the SF light is selectively generated from an interface, where
sample is obtained. The first SF vibrational spectra were the symmetry of the bulk phase is broken. It is this abil-
obtained in 1986 by Zhu et al.1 for Coumarin 504 on fused ity of SFS to discriminate between molecules preferentially
quartz. Subsequently, many different molecules have been oriented at an interface and the same molecules randomly
studied at a variety of different interfaces. Recent reviews oriented in the bulk phase either side of the interface that
cover the application of SFS to surfactant systems, liquid makes SFS such a useful technique.
interfaces, and catalysis.2 4
The basic principles of SFG can be understood in a sim-
ple classical picture. An IR laser at frequency !IR induces 2 INSTRUMENTATION
an oscillating dipole in a molecule when it is resonant with
an IR-active vibrational mode, n. This oscillating dipole There are three main components to any SF spectrometer: a
fixed frequency pulsed laser in the visible or near-infrared, a
modulates the polarizability of the molecule at frequency
tunable laser in the mid-infrared, and a detector for the light
!n provided that the polarizability derivative of the mode
generated at the SF. Figure 1 shows a schematic diagram
is nonzero, i.e. the mode is Raman-active. The polariza-
of a SF spectrometer.
tion induced by a second laser at frequency !p therefore
contains a component with a time dependence of the form
cos!p t cos!n t. This product can be re-expressed as a
2.1 Fixed frequency laser
sum of polarizations oscillating at frequency !df D !p
!n (difference frequency generation (DFG), analogous to Pulsed lasers are essential to generate the high fields nec-
Stokes scattering) and !sf D !p C !n (SFG, analogous to essary for SFG; for a fixed pulse energy, the SF signal is
anti-Stokes scattering). These oscillating dipoles emit light. inversely proportional to the pulse duration. One laser is
normally in the visible or near-infrared so that the SF light
John Wiley & Sons Ltd, 2002. generated is easily detected. DFG occurs simultaneously
2 Other Instrumental Approaches for Vibrational Spectroscopy
Pulsed dye Nd :YAG

emitted. The sum of the two emitted photon frequencies
laser 2 is equal to that of the incident photon. OPO and OPG
lasers are both based around the process of optical para-
metric generation but differ in their method of wavelength
Hydrogen Raman shifter
Control selection: phase matching (OPG) or cavity tuning (OPO).

electronics Both barium borate and lithium borate have tuning ranges
between 400 nm and 2.5 m.6 As with dye lasers, OPO and
OPG lasers rely on being pumped by another laser. The
Computer Delay most common pump source is the 2nd or 3rd harmonic of
line
a Nd : YAG laser. A broad band of frequencies can also
Detector be generated simultaneously if an intense, ultra-short laser
electronics pulse is focused tightly into a suitable liquid, a process
known as continuum generation.7,8
Ge plate
CCD 2.2.2 Wavelength shifting

Filter
None of the three methods for generating tunable light
SF described above reliably produce light in the wavelength
range needed for SFS. There are two common techniques
IR VIS to shift the wavelength of the tunable light into the mid-
infrared: Raman shifting and DFG. In Raman shifting,
Solidsolid the tunable light is directed into a high-pressure multi-
interface pass cell containing a scattering gas such as hydrogen.
Figure 1. Schematic diagram of the SF spectrometer used by The wavelength of the input light is Raman shifted to
the authors. A nanosecond Nd : YAG laser pumps a dye laser higher and lower wavelengths by the process of stimu-
and Raman shifter. The delay line is used to obtain temporal
overlap between the visible pulse and the IR pulse generated in
lated Raman scattering. In this way, successive Stokes
the multipass Raman laser. shifts can convert the incoming visible light into the mid-
infrared. Tunability is inherent in the process as the input
with SFG, however, the detection of difference frequency wavelength is tunable. A photon efficiency of >20% can
(DF) light is hampered by background fluorescence from routinely be achieved with this method for IR wavelengths
the sample. The commonest pump lasers are Nd : YAG or up to 5 m. The other wavelengths generated (due to anti-
Nd : YLF (fundamental or 2nd harmonic) and Ti : sapphire. Stokes and lower order Stokes shifts) can be removed by a
filter.
DFG is normally achieved in a nonlinear crystal such
2.2 IR laser as AgGaS2 . The tunable light from the OPO or dye laser
is mixed with the output of the pump laser. The crystal
There are two aspects to generating IR light for SFS: generates light at the difference of the two frequencies.
tunability and wavelength shifting. With this method it is possible to extend the tuning range
of OPO type systems up to l 10 m.
2.2.1 Tunability
2.2.3 Free electron lasers
There are two common ways to generate pulsed tunable
light: dye lasers and optical parametric oscillators (OPOs) There is a third way to generate the tunable IR light needed
or optical parametric generators (OPGs). Dye lasers are for SFS that does not involve frequency shifting. Free
very common and can be tuned across the visible fre- electron lasers (FELs) are based around the generation of
quency range.5 The discovery of nonlinear optical crystals coherent light by the interaction of relativistic electrons with
with large nonlinearities and high damage thresholds, such a periodic transverse magnetic field structure.9,10 The equip-
as b-barium borate and lithium triborate, has opened up ment necessary to generate light in this way is expensive
optical parametric generation as a way to generate tunable and as a result there are only a few of these sources spread
light. Optical parametric generation can be viewed as an around the world, the majority of which operate as multi-
inelastic scattering process whereby a photon is absorbed user facilities. Two FELs have been used for SFG: CLIO in
in a nonlinear crystal and two photons are simultaneously France and FELIX in the Netherlands. The tuning range of
Sum-frequency Spectroscopy 3
FELIX is 4 to 200 m, the pulse length is 0.52 ps and the sum vis
bandwidth is variable from <1% to >5% of the wavelength.
IR
2.3 Detectors
Photomultiplier tubes (PMTs) with gated averaging, and

charge-coupled devices (CCDs) are the detectors normally
used for SFS. CCDs are generally more efficient (as high
as 80%) than PMTs. The nature of a CCD (an array
(a)
of pixels) allows for spatial discrimination of SF and
scattered light, allowing for better signal-to-noise ratios. sum
If broad bandwidth lasers are used for SFS then the use
of CCD detectors enables spectra to be recorded along the vis
length of the CCD chip. In this approach, spectra can be
recorded from every laser shot, which improves the quality
of spectra as fluctuations in the input laser power affect all
IR
wavelengths equally.
2.4 Methods of data acquisition
The properties of the different laser sources lead to a num-

(b)
ber of different ways of recording SF spectra. With narrow-
band lasers, a point-by-point approach is best whereby the Figure 2. Incident and emitted light for SFG for (a) co-propa-
gating and (b) counter-propagating incident beams. The visible
SF signal is collected at discrete IR wavelengths. If broad-
(dashed line) and IR (dotted line) lasers are incident at 40 and
band IR sources are used (combined with narrowband vis- 60 , respectively. The range of SF angles (solid arrow) is for
ible pulses to give spectral resolution) then self-dispersive incident IR light with a range of wavenumbers between 2000 and
detection can be adopted. Self-dispersive detection relies 3000 cm1 .
on the fact that the angle at which the SF light is emitted
is determined by conservation of momentum parallel to the
surface and is therefore dependent on the frequency of the 3 THEORY
IR laser:
jj jj jj
ksum D kvis C kIR 1 3.1 Nonlinear optics
where k jj is the component of the wavevector (2p/l)
parallel to the surface. The SF frequencies that result from In an applied electric field, E, the dipole moment of a
a broadband IR pulse are dispersed across the detector.11,12 molecule can be written as a power series in E:
The SF frequencies generated from broad-band IR sources .
can also be separated with a monochromator.13 A recently m D m0 C a E C b : EE C g .. EEE C 2
introduced method for obtaining SF spectra employs two
IR pulses with a variable time delay, plus a visible pulse. where m0 is the static dipole moment and the term a
Varying the delay generates an interferogram that can be E, where a is the polarizability, characterizes the linear
Fourier transformed to give a SF spectrum.14 response to the incident electromagnetic (EM) field. b and
The incident lasers in a SF experiment can be either g are known as the first and second hyperpolarizabilities.
counter or co-propagating (see Figure 2). With femtosecond The higher order terms are normally only important in the
laser pulses, co-propagating beams must be used as tempo- intense fields generated by laser beams. a, b, and g are
ral overlap is poor with counter-propagating beams. With tensors of rank 2, 3, and 4, respectively.
The short hand
longer pulses (nanosecond, or picosecond) the counter- notation b : EE is a contraction of jk bijk Ej Ek . It is this
propagating geometry is advantageous as there is a larger term that is responsible for SFG.
angle between the SF light and the reflected pump laser For bulk materials, it is convenient to express the
beams, which aids the collection of the SF light, and the response in terms of the dipole moment per unit volume,
degree of dispersion in self-dispersive SFG is greater. or polarization, P:
P D P0 C P1 C P2 C P3 C For a molecular material, c2 is determined by the
. molecular hyperpolarizability, b (see equation 4). Within
D P0 C e0 c1 E C c2 : EE C c3 .. EEE C 3 the electric dipole approximation, the expression for b
derived from second-order perturbation theory15 contains
where P0 is the static polarization and ci is the ith-
eight terms. For the case where !vis and !sum are non-
order susceptibility. The expression can be modified if there
resonant with electronic transitions in the molecule, two of
is more than one incident electric field. The second-order
the eight terms dominate:
susceptibility, c2 , is related to the orientational average of
the hyperpolarizability: blmn !sum ; !vis , !IR

Nf 1 hgjml jsihsjmm jni hgjmm jsihsjml jni
c2 D hbi 4 D
e0 2h s h!sum !sg h!vis C !sg

where N is the number density of molecules sampled and hnjmn jgi
f is a local field correction to account for the additional 6
!IR !ng C ing
electric field arising from the polarization of neighboring
molecules. The first expression in brackets can be identified with the
The higher-order terms in the polarization expansion are Raman transition moment Mlm for a transition between
responsible for the field of physics known as nonlinear some ground state jgi and an excited vibrational state jni
optics. To see how SFG arises, we consider two incident via a virtual state jsi. The matrix element hnjmn jgi is the
EM fields at frequencies !1 and !2 (E1 D E1 r cos !1 t, IR transition dipole for the transition between jgi and jni
E2 D E2 r cos !2 t). and ng 1 is the transverse relaxation time of the excited
vibrational state. From equation (6) it is evident that a
P2 D e0 c2 : E1 E1 C e0 c2 : E2 E2 C 2e0 c2 : E1 E2 vibration has to be both Raman and IR active to give rise
to SFG. Also, as the incident IR wavenumber (!IR ) scans
D e0 c2 : E1 rE1 r cos2 !1 t
over the wavenumber of the vibrational transition (!ng ), the
C e0 c2 : E2 rE2 r cos2 !2 t hyperpolarizability goes through a maximum at !IR D !ng .
In general, several different components of the c2 tensor
C 2e0 c2 E1 rE2 r cos !1 t cos !2 t
may add coherently to give the SF signal.
D e0 c2 : E1 rE1 r1 cos 2!1 t 2

2
C e0 c2 : E2 rE2 r1 cos 2!2 t Isum D Lsum,i Kvis,j KIR,k cijk Ivis IIR 7
ijk
C 2e0 c2 E1 rE2 rcos!1 C !2 t
C cos!1 !2 t 5 where Ivis and IIR are the intensities of the two incident
lasers and Isum is the intensity of the emitted SF light.
The first two terms represent an oscillating polarization at The Fresnel coefficients (K) relate the electric fields in the
frequencies 2!1 and 2!2 together with a static polarization incident lasers to the fields at the interface.16 They depend
of the sample. The oscillating polarizations emit light with on the angles of incidence at the two input wavelengths,
an intensity proportional to P2 2 a process known as on the refractive indices of the two bulk media, and on the
second harmonic generation. P2 also contains terms oscil- dielectric properties of the molecules at the interface. L is
lating at the sum and difference of the incident frequencies, a parameter, similar to K, that relates the electric field in
which give rise to SFG and DFG, respectively. the emitted SF light to the polarization of the sample at
the SF.17
3.2 Sum-frequency generation

3.3 Determination of c.2/ from SF spectra
SFG depends on the magnitude of the second-order suscep-
tibility, c2 . c2 is a third-rank tensor, which changes sign To deduce quantitative information on coverage/orientation
under inversion; centrosymmetric and isotropic materials it is necessary first to determine the individual components
are invariant under inversion and thus, for these materials, of the c2 tensor from experimental SF spectra. The task
c2 is zero. Only noncentrosymmetric systems give rise to of relating c2 to SF intensity is simplified when sym-
SFG by electric dipole processes. Interfaces are intrinsically metry reduces the number of independent elements of the
noncentrosymmetric and thus SFG is surface selective for tensor. For a sample that is isotropic in the plane of the
interfaces between centrosymmetric media. surface there are only four independent nonzero elements:
czzz , cxxz D cyyz , cxzx D cyzy and czxx D czyy (for a coordi-
nate system where the z-axis is perpendicular to the surface
SF signal / a.u.
and the xz plane is the plane of incidence). A further simpli-
fication occurs when the SF and visible wavelengths are far
from electronic resonances of the sample, cxzx D czxx . This
identity arises from the symmetry properties of the Raman
tensor, a0 .18 The three independent nonzero elements can
be obtained from three SF experiments with plane-polarized 2800 2850 2900 2950 3000
light: ssp, sps, and ppp, where the three letters denote the (a) IR wavenumber / cm1
polarization of the SF, visible, and IR light, respectively.
2

Issp D Fyyz c2
yyz Ivis IIR 8
SF signal / a.u.
2

Isps D Fyzy c2
yzy Ivis IIR 9

Ippp D Fxxz c2 2 2
xxz C Fxzx cxzx C Fzxx czxx
2
CFzzz c2
zzz Ivis IIR 10
2800 2850 2900 2950 3000
where Fijk is the combination of the three Fresnel coef-
(b) IR wavenumber /cm1
ficients for each c2 element (see equation 7). Thus, cyyz
and cyzy can be obtained directly from experiments with ssp
and sps-polarized light, respectively, while czzz is usually SF signal /a.u.
obtained only in combination with other elements.
In general, the second-order susceptibility consists of a
nonresonant contribution from the interface, c2
NR , and a res-
onant contribution, c2
R,n , which arises when the frequency
of the IR electric field is resonant with a vibrational mode,
n, of the molecules at the interface: 2800 2850 2900 2950 3000
2
c2 D c2 cR,n 11 (c) IR wavenumber /cm1
NR C
n
Figure 3. SF spectra of a monolayer of (a) octadecanethiol on
2
cNR and 2
cR,n are both complex quantities. The phase, , gold (e D p/2), (b) octadecanethiol on silver (e D p/4), and
2 (c) dodecanethiol on platinum. The incident laser beams were p-
of cR,n varies with IR frequency in much the same way as polarized. [Reproduced from C.D. Bain, J. Chem. Soc. Faraday
the phase of the linear susceptibility, c1 ; passes from Trans., 91, 1281 (1995) by permission of The Royal Society of
0 to p either side of the resonance, with a value of p/2 Chemistry.]
at the peak of the resonance. The intensity and appearance
of peaks in SF spectra are modified by the interference susceptibility on the appearance of SF spectra. Figure 3(a)
between the resonant and nonresonant contributions to c2 . shows a SF spectrum of a self-assembled monolayer of
In the simplest case where c2 NR has a frequency-independent octadecanethiol on gold. The peak shapes appear entirely
phase, e, the Fresnel coefficients are the same for c2 NR and normal for a vibrational spectrum as, in this instance,
2 2
cR,n , and the vibrational modes are nonoverlapping, the cNR is in phase with the SF signal at the peak of the
intensity of the SF output can be expressed as vibrational resonance (e D D p/2). The cross term in
2 equation (12) is symmetrical about !n and falls off rapidly
2
Isfg / c2NR C cR,n when j!n !IR j > ng . Figure 3(b) shows the SF spectrum

2 of the same monolayer on silver, for which e D p/4.
D c2NR exp ie C cR,n exp i!
There is now constructive interference between c2 NR and

2 2
cR,n on the low frequency side of the peak and destructive
c2 NR expie C cR,n exp[i!]
interference on the high frequency side, giving rise to
2
2 2 2 2 highly asymmetric peak shapes. Figure 3(c) shows an SF
D c2
NR C c R,n C2 c NR c R,n cos[e !] 12
spectrum of a system in which the magnitude of the non-
The spectra shown in Figure 3 illustrate the influence resonant contribution is small: dodecanethiol on platinum.
of the magnitude and phase of the surface nonresonant The peak shapes are little affected by the nonresonant
contribution, although c2

NR does have to be included in
b if molecular symmetry reduces the number of nonzero,
a quantitative analysis. Overlapping vibrational peaks also independent elements of b. Group tables list the nonva-
interfere coherently; counter-intuitively, if c2 has the same nishing elements of the IR and Raman transition dipole
sign for both vibrations then they interfere destructively, for each irreducible representation of the point group to
and vice versa. which a molecule belongs. The combination of the indices
of these elements (aij mk ) gives the nonzero elements of b,
e.g. a representation that has a Raman tensor element bc
3.4 Analysis of SF spectra and an IR transition dipole element b will give rise to bbcb
(and bcbb due to the symmetry of the Raman tensor). The
SF spectroscopy can be used to obtain a number of molec- result of this simplification depends on the symmetry of
ular parameters, such as the identity of adsorbed molecules the molecular system. Two examples from common sym-
and the conformational order of the hydrocarbon chains in metry groups will be given: an A1 vibration in a molecule
surfactants or lipids. In order to extract this information, it of C3n symmetry, and a B1 vibration in a molecule of C2n
is necessary to relate the susceptibility tensor, c2 , which symmetry.
is expressed in the coordinates of the surface, to the hyper-
polarizability tensor, b, which is expressed in molecular
coordinates. Details of the coordinate transformation can 3.4.1 C3 n symmetry
be found in a paper by Hirose et al.17 The transformation C3n symmetry is frequently used to describe the termi-
can be expressed in terms of the Euler angles, q, y, and f. nal methyl groups of hydrocarbon chains in surfactant
If the surface axes (x, y, and z) and the molecular axes (a, molecules. For C3n symmetry there are 11 nonzero elements
b, and c) are initially coincident then the molecular frame of b of which only four are independent: baac (D bbbc ), baaa ,
can be obtained by rotating the molecular axes around z (D bbba D bbab D babb ), baca (D bcaa D bbcb D bcbb ),
by an angle f, tilting the molecular axes an angle q away and bccc . The vibrations of a molecule are split into groups
from the z axes towards the a axis, and finally rotating the depending upon the irreducible representation to which they
molecular axes by an angle y around the c axis (Figure 4). belong; in C3n there are the A1 , A2 and the doubly degener-
It is possible to simplify the expression relating c2 R,n and ate E representations. Only A1 and E modes are SF active
zc
as the A2 representation corresponds to modes that are nei-
ther IR nor Raman active. The nonzero elements of bn for
vibrations of A1 symmetry are baac (D bbbc ) and bccc . If
the molecules at an interface have random azimuthal ori-
entation (f) then the equations for the nonzero elements of
c2 are17
fNbccc

c2
zzz D rhcos qi r 1hcos3 qi 13
xa yb e0
fNbccc

c2
xxz D r C 1hcos qi C r 1hcos3 qi 14
2e0
fNbccc

z c2
xzx D 1 rhcos qi C r 1hcos3 qi 15
c 2e0

where r D baac /bccc is the ratio of the two independent
components of the molecular hyperpolarizability. Algebraic
manipulation of equations (13), (14), and (15) allow us to
obtain an expression for r in terms of the measurable c2
a elements:
baac c2 2
xxz cxzx
rD D 2 2
16
bccc czzz C 2cxzx
y
x
It is also possible to obtain an expression for the molecular
b orientation parameter, D:

Figure 4. Coordinate transformation from the lab frame to the hcos3 qi c2 2 2
zzz cxxz C cxzx
DD D 2 2 2
17
molecular frame illustrating the Euler angles, q, j, and y. hcos qi czzz cxxz C 3cxzx
3.4.2 C2 n symmetry have been studied as well as amorphous polymer blends,20

where SFS is sensitive to preferred orientation and surface
For a molecule with C2n symmetry there are seven nonzero
segregation.
elements of bn of which four are independent: baac (D bbbc ),
baca (D bcaa ), bbcb (D bcbb ), and bccc . For a vibration that
belongs to the B1 representation only two of these elements
4.2 Surfactants at liquid interfaces
are nonzero: baca and bcaa . The expressions relating b to
2
cR,n are as follows: SFS is well-suited for the study of surfactants adsorbed
fNbaca at the airliquid, liquidliquid, and solidliquid interfaces
c2 2
zzz D cxxz D hcos qi hcos3 qi 18 because SF light is only generated from the molecules
2e0
adsorbed at the interface and not from surfactant in the
fNbaca bulk solution. Furthermore, SFS is unusually sensitive to
c2
xzx D hcos3 qi 19
2e0 conformational order in hydrocarbon chains, and hence to
The orientation parameter, D, is given by packing in surfactant films.
Figure 6 shows the SF spectra of dodecanol on D2 O,
hcos3 qi c2 recorded with three different polarization combinations
DD D 2 xzx 2 20
hcos qi cxzx cxxz of incident and emitted light.21 The spectra are dom-
inated by the vibrations of the terminal methyl group
of the hydrocarbon chains (symmetric stretch, r C , at
C
4 APPLICATIONS 2880 cm1 , and its Fermi resonance, rFR , at 2935 cm1 ,
and antisymmetric stretch, r , at 2960 cm1 ), while the

4.1 Neat liquid interfaces methylene (CH2 ) stretching modes (symmetric stretch, dC ,
expected at 2850 cm1 , and its Fermi resonance, dC FR , at
The structure of the surface of water is a matter of long-
standing debate. Since the bulk liquid is isotropic, SFS is (a)
sensitive only to those molecules at the interface. The SF
spectrum of water (Figure 5) shows peaks due to both free
OH (3680 cm1 ) and hydrogen-bonded OH (broad reso-
nances between 3100 cm1 and 3500 cm1 ).19 The phase
of the emitted SF light indicates that the free OH is
directed away from the water surface. The structure of
the surface of water has also been studied in contact with
SF signal / a.u.
(b)
oil, solid surfaces and in the presence of a surfactant
monolayer. Other neat liquids and concentrated solutions
SF signal / a.u.
(c)
2800 2850 2900 2950 3000

IR wavenumber /cm1
3000 3200 3400 3600 Figure 6. SF spectra in the CH stretching region of a mono-
layer of dodecanol on D2 O at 298 K: (a) ssp polarization; (b) ppp
IR wavenumber /cm1
polarization; and (c) sps polarization. The use of D2 O rather than
Figure 5. SF spectrum for the neat water/vapor interface at H2 O reduces the size of the nonresonant susceptibility. [Repro-
40 C recorded with ssp polarization. [Reproduced from Q. Du, duced from B.D. Casson, R. Braun and C.D. Bain, Faraday Dis-
R. Superfine, E. Freysz and Y.R. Shen, Phys. Rev. Lett., 70, 2313 cuss., 104, 209 (1996) by permission of The Royal Society of
(1993).] Chemistry.]
28902930 cm1 ), are absent. These spectra are indicative 4

(a)
of a well ordered monolayer with all-trans chains: in an all-
trans chain of CH2 groups there is a local center of inversion
in the middle of each CC bond and hence the methy-
lene modes are SF-inactive. Gauche defects lower the local 2
SF signal / photons pulse1 mJ2

symmetry and consequently the intensity of the methylene
modes in SF spectra is a sensitive indicator of conforma-
tional disorder. Figure 7 shows spectra of four different
surfactants, each forming a monolayer with a progressively
0
larger area per molecule, A. The spectra have been normal-
ized to A2 to compensate for the effect of surface coverage (b)
on the appearance of the spectra. As A increases there is an
increasing number of gauche defects, and hence an increase 1.0
in the strength of the CH2 modes in the SF spectra. At the
same time, the strength of the r C mode decreases as the
average tilt, q, of the methyl group increases. 0.5
Selective deuteration allows different components in
mixed monolayers to be distinguished. Figure 8 shows
SF spectra of a mixed monolayer of dodecanol and an 0.0
alkyltrimethylammonium bromide surfactant, C16 TAB, at 2800 2850 2900 2950 3000
the D2 Oair interface.22 The spectra were recorded at two IR wavenumber /cm1
Figure 8. SF spectrum of a mixed monolayer adsorbed at the
airD2 O interface from a solution containing 0.6 mM C16 TAB
and 10 M dodecanol, recorded with ssp polarization: (a) C16 TAB
deuterated, dodecanol protonated; (b) C16 TAB protonated, dode-
canol deuterated. The different traces correspond to temperatures
below (solid line) and above (dotted line) the phase transition for
the monolayer. [Reprinted with permission from B.D. Casson and
C.D. Bain, J. Phys. Chem. B, 103, 4678 (1999). Copyright 1999
American Chemical Society.]
SF signal / a.u.
different temperatures, above and below the liquidsolid

phase transition of the monolayer. Selective deuteration of
(a)
1 the surfactant molecules makes one or other component of
(b)
the mixture invisible to SFS in the CH or CD stretch-
2.6 ing region. If the mean orientation of the terminal methyl
groups is the same for the two components of the mono-
(c) layer, the composition of the monolayer can be calculated
4.2 from the relative intensities of the r C modes. One inter-
(d)
5.7
esting feature in Figure 8 is the lack of intensity in the
methylene modes for the dodecanol in the monolayer, above
and below the phase transition temperature, indicating that
2800 2850 2900 2950 3000 the dodecanol retains an ordered all-trans structure in the
IR wavenumber /cm1 liquid phase of the monolayer.
Figure 7. SF spectra of four different soluble surfactants at the
airD2 O interface, recorded with ssp polarization: (a) dodecanol
(area per molecule, A D 21 A2 ); (b) didodecyldimethylammonium 4.3 Solidgas interface
bromide (A D 34 A2 ); (c) dodecyl maltoside (A D 43 A2 ); and
(d) dodecyl betaine (A D 50 A2 ). The spectra have been normal- SFS is just one of many forms of vibrational spectroscopy
ized by A2 , and offset, to highlight the effect of changes in the that have been used to study adsorbates on solid surfaces
conformation of the hydrocarbon chains. [See C.D. Bain, Non-
linear Optical Techniques, in Modern Characterization Methods under ultrahigh vacuum (UHV) conditions. Practical catal-
of Surfactant Systems, ed. B.P. Binks, Marcel Dekker Inc., New ysis occurs at pressures of several bar, however, not in
York, 335373 (1999).] a vacuum. Techniques such as high-resolution electron
energy loss spectroscopy (HREELS) and infrared reflec-

tionabsorption spectroscopy (IRRAS) cannot easily be H
H H
used at high pressures due to the short mean-free path
for the electrons (HREELS) or to absorption by gas-phase
species (IRRAS). There is growing evidence that the struc-
ture and identity of reaction intermediates in catalytic
SF signal / a.u.
processes under turnover conditions are different from
those detected in UHV and it is thus essential to character-
ize surfaces under realistic conditions of temperature and
H H
pressure.4 SFS provides a bridge between the two pres- H H
sure regimes. It can be applied in UHV, along with other
H H
spectroscopic techniques, and at ambient pressure where the H H
surface selectivity of SFS comes to the fore. Figure 9 shows
the CDO stretching region for CO adsorbed on Pt (111) at
two different pressures; 107 Torr, and 400 Torr at 300 K.23
Figure 10 shows SF spectra recorded during the hydrogena-
tion of ethylene on Pt (111) at a total pressure of 750 Torr
at 295 K.24 Comparison of the ethylene SF spectrum with 2850 2900 2950 3000
those obtained under UHV conditions allowed the authors IR wavenumber /cm1
to identify p-bonded ethylene as the active species during Figure 10. SF spectrum of the Pt (111) surface during ethy-
hydrogenation. lene hydrogenation at 295 K (H2 , 100 Torr; C2 H4 , 35 Torr; He,
615 Torr). [Reprinted in part with permission from P.S. Cremer,
X. Su, Y.R. Shen and G.A. Somorjai, J. Am. Chem. Soc., 118,
2942 (1996). Copyright 1996 American Chemical Society.]
4.4 Buried interfaces
Buried interfaces always provide challenges for vibrational

SF signal / photons pulse 1
spectroscopy. For the solidsolid interface, the usual prob- 15

lems of accessing the interface are compounded by the
small area of contact between hard solids. In lubrication
research, the determination of the structure of boundary 10
lubricants under conditions of pressure and shear is a key Increasing
pressure
goal. SFS can be used to study lubricants in a frictional
contact provided that at least one of the solid surfaces is 5
transparent to visible and IR wavelengths. Figure 11 shows
0
12
2800 2850 2900 2950 3000
IR wavenumber / cm 1
Normalized SF signal / a.u.
Figure 11. SF spectra of a monolayer of Zn arachidate at the

400 Torr sapphiresilica interface, recorded with ssp polarization and pres-
8
sures between 33 and 45 MPa. The dotted trace is the spectrum
of the same monolayer at the sapphireair interface.
4
a set of spectra of a boundary lubricant (Zn arachidate)
at the interface between a sapphire prism and a fused sil-
ica lens, recorded as a function of increasing pressure.
107 Torr
Also shown in Figure 11 is the SF spectrum of the same
0 monolayer at the sapphireair interface, prior to mak-
2000 2100 ing contact. Total internal reflection of the laser beams
IR wavenumber /cm 1 in the prism was used to maximize the electric fields at
Figure 9. SF spectra of CO adsorbed on Pt (111) at 300 K. [Data the interface.25 The strength of the methylene modes did
supplied by Prof. Y.R. Shen. For more detail see Kung et al.23 ] not increase under pressure, indicating that the monolayer
20 ABBREVIATIONS AND ACRONYMS

SF signal / photons pulse 1
a.u. Arbitrary Unit

15 DF Difference Frequency
DFG Difference Frequency Generation
EM Electromagnetic
10
OPG Optical Parametric Generator
OPO Optical Parametric Oscillator
SF Sum-frequency
5
SFS Sum-frequency Spectroscopy
UHV Ultrahigh Vacuum
0
2800 2850 2900 2950 3000
REFERENCES
IR wavenumber /cm1
Figure 12. SF spectra of a monolayer of Zn arachidate at the 1. X.D. Zhu, H. Suhr and Y.R. Shen, Phys. Rev. B, 35, 3047
sapphiresilica interface, recorded with ppp polarization under (1987).
static pressure (solid line), and shear (dotted line). The shear speed 2. C.D. Bain, Non-linear Optical Techniques, in Modern
was 200 m s1 . Characterization Methods of Surfactant Systems, ed. B.P.
Binks, Marcel Dekker Inc., New York, 335373 (1999).
retained a high degree of conformational order. Figure 12 3. P.B. Miranda and Y.R. Shen, J. Phys. Chem. B, 103, 3292
shows SF spectra of the same monolayer under shear; the (1999).
well-ordered structure of the monolayer is retained. The 4. G.A. Somorjai and G. Rupprechter, J. Phys. Chem. B, 103,
spectrum after shear was identical to that recorded before 1623 (1999).
shear. 5. F.J. Duarte and L.W. Hillman, Dye Laser Principles and
Applications, Academic Press, Boston (1990).
6. J.Y. Zhang, J.Y. Huang, Y.R. Shen and C. Chen, J. Opt. Soc.
5 SUMMARY Am. B, 10, 1758 (1993).
7. R.L. Fork, C.V. Shank, C. Hirlimann, R. Yen and W.J. Tom-
SFS is a form of nonlinear vibrational spectroscopy. The linson, Opt. Lett., 8, 1 (1983).
technique is experimentally more complex than Raman 8. A. Migus, A. Antonetti, J. Etchepare, D. Hulin and A. Orszag,
scattering or IR spectroscopy, however, SFS has certain J. Opt. Soc. Am. B, 2, 584 (1985).
advantages over both of these techniques. The foremost 9. T.C. Marshall, Free Electron Lasers, Macmillan, New York
(1985).
advantage is that SFG does not occur in centrosymmetric
or isotropic media and SFS can therefore discriminate 10. D. Luchini and H. Motz, Undulators and Free Electron
Lasers, Oxford University Press, Oxford (1990).
between molecules at an interface and the same species
in the bulk phases on either side. This selectivity makes 11. E.W.M. van der Ham, Q.H.F. Vrehen and E.R. Eliel, Opt.
Lett., 21, 1448 (1996).
SFS an excellent probe of molecules at interfaces and of
12. E.W.M. van der Ham, Q.H.F. Vrehen and E.R. Eliel, Surf.
related interfacial processes. The coherent nature of SFG
Sci., 368, 96 (1996).
confers two other benefits on the technique of SFS. First,
13. L.T. Richter, T.P. Petralli-Mallow and J.C. Stephenson, Opt.
the phase of the emitted light contains information on the Lett., 23, 1594 (1998).
polar orientation of molecules at interfaces. Second, the
14. J.A. McGuire, W. Beck, X. Wei and Y.R. Shen, Opt. Lett.,
SF light is emitted as a collimated laser beam, allowing 24, 1877 (1999).
high collection efficiency and good discrimination of the
15. Y.R. Shen, Principles of Nonlinear Optics, Wiley, New
SF signal from background light. The advantages of SFS York (1984).
over other vibrational spectroscopic techniques do come at 16. M. Born and E. Wolf, Principles of Optics, 6th edition,
a price. The interpretation of SF spectra requires knowledge Pergamon, Oxford (1980).
of both the IR absorption and Raman scattering properties 17. C. Hirose, N. Akamatsu and K. Domen, Appl. Spectrosc., 46,
of a molecule and is therefore more complex than either 1051 (1992).
of these two techniques individually. However, this has not 18. D.A. Long, Raman Spectroscopy, McGraw-Hill, London
stopped SFS being applied successfully to a wide range of (1977).
different interfaces and over the whole of the vibrational 19. Q. Du, R. Superfine, E. Freysz and Y.R. Shen, Phys. Rev.
spectrum (2.550 m). Lett., 70, 2313 (1993).
20. D. Zhang, R.S. Ward, Y.R. Shen and G.A. Somorjai, J. Phys. 23. K.Y. Kung, P. Chen, F. Wei, Y.R. Shen and G.A. Somorjai,
Chem. B, 101, 9060 (1997). Surf. Sci., 463, L627 (2000).
21. B.D. Casson, R. Braun and C.D. Bain, Faraday Discuss., 24. P.S. Cremer, X. Su, Y.R. Shen and G.A. Somorjai, J. Am.
104, 209 (1996). Chem. Soc., 118, 2942 (1996).
22. B.D. Casson and C.D. Bain, J. Phys. Chem. B, 103, 4678 25. D.A. Beattie, S. Haydock and C.D. Bain, Vib. Spectrosc., 24,
(2000). 109 (2000).
Inelastic Electron Tunneling Spectroscopy
K.W. Hipps and U. Mazur
Washington State University, Pullman, WA, USA
1 INTRODUCTION resolution can be better than 5 cm1 ; broad spectral range

(50 to above 19 000 cm1 ); and similar tunneling, Raman,
1.1 Why study inelastic electron tunneling
IR, and near-infrared band positions and widths are obser-
spectroscopy?
ved for similar samples.1,6,12,20,22,23,25 27,35 38,54 63
Inelastic electron tunneling spectroscopy (IETS) is an
all electronic spectroscopy that has been extensively
reviewed.1 7 Walmsley and Tomlin3 have published a very 1.2 The focus of this article
useful compendium of IETS spectra. By measuring currents
The focus of this article is on the applications of IETS to
and voltages across a metalinsulatoradsorbatemetal
vibrational spectroscopy. Even in this context, however, we
(MIAM0 ) device in a special way (see below),
will need to briefly describe other types of transitions that
one is able to extract vibrational and electronic
are active in the tunneling spectrum so that we may distin-
spectroscopic information about the metals (magnons and
guish vibrational features from electronic excitations, metal
phonons), the insulator, and the adsorbate. It has been
phonons, magnons, and orbital mediated elastic tunneling
successfully applied to problems in surface chemistry and
events. In order to appreciate the origins of the variety of
catalysis,1 16 adhesion and corrosion,1 7,11,17,18 molecular
bands that may be observed, the basic concepts of elas-
vibrational1,6,19 32 and electronic6,33 44 spectroscopy, and
tic and inelastic electron tunneling will be introduced. The
electronic orbital mediated tunneling (OMT).45 48 The
basic features of the experiment, both instrumental and sam-
motivation for tunneling studies is based on several unusual
ple preparation, will be described. Comparison of IETS,
properties of the electron-molecule scattering process. The
IR, Raman, and other optical spectra will be used to exem-
advantages of IETS include:
plify the notion that IETS is a technique complementary
1. ultra-high sensitivity. Less than 1013 molecules are to the more common photon spectroscopies. IETS is an all
required to provide a spectrum;1,49,50 electronic technique for performing molecular and ordered
2. overtone and combination bands are exceptionally layer spectroscopy. As we shall see, the interaction between
weak.1,6,19,22,36,51 54 Thus, it is easier to identify fun- the molecule of interest and the IETS environment some-
damentals in IETS than in IR or Raman; times leads to chemical changes in the molecule. Thus, any
3. optically forbidden transitions may be observed as careful IETS study must include a component that clearly
strong bands;20 22,28 41 identifies the chemical form of the trapped species. The sig-
4. oxide bands are generally less intense than adsorbate nificance of this will become clear in the following sections.
bands. It is possible to obtain adsorbate spectra in the
IR opaque regions of the oxide spectrum.1 6
2 THEORY OF ELECTRON TUNNELING
In addition to these advantages, IETS also has important IN MIAM DEVICES
features which are necessary for any viable spectroscopy:
By electron tunneling, we mean the motion of electrons
John Wiley & Sons Ltd, 2002. from one classically allowed region to another through
a region where the electron is classically forbidden to of the STM experiments merits separate treatment, we will
exist (see Figure 1a). The forbidden or barrier region is focus on MIM0 diodes here. In a typical MIM diode
one in which the potential energy, U, is greater than the used for tunneling studies, 0.1 m of metal, M, is vapor
total classical energy, E. If the particle moves from one deposited on a clean microscope slide to form a film 1 mm
allowed region to another, it must tunnel through the wide by 3 cm long. An insulating film of about 2 nm in
potential barrier. Of course, there is no such prohibition thickness is either grown or deposited on that metal film.
in quantum mechanics. An electron of energy E impinging At this point in the conventional fabrication the material
on the potential barrier from the left in Figure 1 has of interest, A, is adsorbed onto the oxide at a coverage
an exponentially damped probability of penetrating the ranging between 0.01 and 1.0 monolayers. The device is
barrier a distance z. If U and E are both large (several completed by depositing a second metal, M0 , strip (typically
electron volts) and d is very small (of the order of 1 mm 10 mm 0.2 m) at right angles to the first and
1 nm), there is a finite probability of the incident electron crossing over it. Thus, a MIAM0 sandwich is formed
emerging into the classically allowed region III. Since wherein the insulator and adsorbate act as a fixed thickness
the incident and emergent energies are the same, this barrier to conduction.
is an elastic process. Quantum mechanics predicts1,19,64 The MIAM0 diode can be related to the simple
that the transmission probability will be proportional to model shown in Figure 1(b). In the crude model shown
1
exp[AdU E1/2 ], where A is about 1.0 (eV)1/2 A . in Figure 2(a), the free electron theory is used to represent
This elastic tunneling process is easily realized in the form the metals and the insulator plus adsorbate is treated as a
of a metalinsulatormetal (MIM) tunnel diode. For vacuum space. The electron states are filled up to the Fermi
such a device, U E is typically about 4 eV and d 2 nm. energy, Ef , and the work function of the metal is given by
Tunneling spectroscopy studies are carried out on f. The hatched regions in Figure 2(a) represent completely
MIM0 tunnel diodes, shown schematically in Figure 1(b), filled metal orbitals.
or (very seldom to date) in the MIM0 structure formed Figure 2 shows a typical tunnel junction, such as an
by a scanning tunneling microscope. Because IETS and the AlAl2 O3 anthracenePb diode. The HOMO and LUMO
related orbital mediated tunneling spectroscopy (OMTS) in of anthracene are also shown in Figure 2, as are a few
the scanning tunneling microscopy (STM) environment are schematic ground vibrational levels. The weak interaction
in their infancy,65,66 almost all the spectroscopy has been between the top metal and anthracene causes a shift in
performed in tunnel diodes. Also, because the description anthracenes energy levels but does not produce signifi-
cant changes in its level spacing, i.e. its electronic and
d 20eAd (U E ) /2 vibrational spectra hardly change. This is not true for
every molecule/metal combination, but the body of IETS
Region I Region III literature1 7 and recent Raman and multiple reflection-
d
absorption infrared (MRAIR) experiments1,29,56 60 indicate
that it is usually true.
U
In a more realistic model, one must take into account that
Region II each metal will have an associated work function, fM . As
E the two metals and insulator are brought together, electrons
z =0 z =d flow via tunneling until the Fermi surfaces are matched.
(a) If the left-hand metal electrode is taken as the reference,
the right-hand electrode develops a net charge, of sign and
I magnitude determined by the difference in work functions.
Once the Fermi energies are matched, no net current flows
and the approximate barrier height is f fM C fM0 /2.
V In the case of metals with widely different work functions
I there is an internal potential generated that can amount to
M
M millions of volts per centimeter. It is a fortunate coincidence
that the work functions for Al and Pb are almost the same.
Pyrex
(b)
Figure 1. Schematic representation of the elastic tunneling pro-
2.1 Elastic tunneling
cess (a), and of a typical tunnel diode (b). [Reproduced by
permission of the American Chemical Society from K.W. Hipps Suppose, as in Figure 2, that a small positive voltage, V, is
and U. Mazur, J. Phys. Chem., 97, 7803 (1993).] applied to the top (Pb) electrode. Because the electric field
Inelastic Electron Tunneling Spectroscopy 3
eV
Equation (1) approximates the elastic tunneling current at
0 K for small bias voltage, V. The average barrier height
is f fM C fM0 eV/2. For a typical IETS diode, f

4 eV and d 2 nm. For values of V less than about 0.5 V,

the current is given approximately by equation (1). Note
M (Al) M (Pb)
Elastic that the more familiar cm1 units can be simply obtained
from the bias voltage through the conversion factor of
Inelastic 8066 cm1 V1 .
xx
In actuality, the IV curve is more complex than
indicated by equation (1). The barrier shape is affected
by the adsorbate layer,21 image charge effects round the
edges of the barrier making the simple square barrier model
(a) Positive bias: eV = h
inaccurate,67 and the barrier potential itself is a function
of the applied voltage.21,67,68 Thus, the real IV curve
observed over finite voltages is nonlinear and asymmetric
dl /dV
0.1%
about zero bias.
I
This elastic tunneling current is the ever-present source
of the background signal with which all tunneling spectro-
h /e scopists must deal. Since IETS depends on derivatives of I
Bias, V with respect to V (see below) it would first appear (equa-
tion 1) that the elastic current would have little effect on
Current, I
the IETS. Unfortunately, the number of electrons tunnel-

d2l /dV 2
ing elastically is usually orders of magnitude larger than

the number utilizing inelastic channels. Thus, relatively
Inelastic
tunneling small deviations from equation (1) produce large back-
h /e h/e ground changes. This background is depicted as linear in
Bias, V Bias, V
Figure 2, but becomes nonlinear above about 0.5 V, and is
(b)
even nonmonotonic around 1 V.
Figure 2. Energy diagram for an AlAl2 O3 anthracenePb tun-
nel diode showing elastic and inelastic tunneling processes (a).
Equation (1) assumes that there is no energetic inter-
The hatched region represents the filled states of the top and action between the tunneling electron and the barrier. In
bottom metal electrodes. The area in the center represents the the last few years it has been found that resonance or
anthracene coated alumina. The HOMO (highest occupied molec- near-resonance processes, often referred to as OMT, can
ular orbital) (p) and LUMO (lowest unoccupied molecular orbital) occur and also provide spectroscopic information about the
(p ) orbital energies and a few vibrational levels are indicated.
The case shown is where the bias energy (eV) is just sufficient
absolute energies (as opposed to energy differences) of
to allow inelastic tunneling with excitation of the first vibrational electronic levels in the sample.45 48 Figure 3 shows this
level, eV D hn. Energy loss (equilibration) for the tunneling elec- condition schematically and also provides a real spectrum
tron occurs through a cascade process in the M0 electrode. (b) The that contains the strong asymmetric feature associated with
IV curve, conductanceV curve, and the IETS spectrum that such OMT bands. Because the energy levels of the sample
would result from both elastic processes and the first inelastic
channel as shown in the barrier diagram. [Reproduced by per-
(tetracene here) are physically close to the M0 electrode,
mission of the American Chemical Society from K.W. Hipps and the sample levels move like the M0 Fermi surface under
U. Mazur, J. Phys. Chem., 97, 7803 (1993).] bias. Thus, so long as the sample layer is thin compared
to the oxide layer, LUMO mediated tunneling will always
be seen in forward bias (M negative relative to M0 ) while
inside a metal must vanish, all the potential drop occurs HOMO mediated tunneling will appear in reverse bias. If
within the barrier. Because the molecule, X, is closer to the entire barrier width is composed of electroactive mate-
Pb than to Al, its potential stays very close to that of Pb. rial, the situation is more complex and a symmetric bias
Because of the applied bias, a band of electrons of width response will be observed. While some questions yet remain
eV have sufficient energy to tunnel into empty states of the concerning the timescale of these OMT processes (whether
M0 (Pb) electrode. If they do so without loss of energy, it is they are a true resonance process or actually sequential
elastic tunneling. near resonance processes), the energetics of the observed
bands appear to correlate well with a simple model based
I D CV expAd f 1 on electrochemical redox potentials.46 48 Thus, OMTS can
is an excellent complementary vibrational spectroscopy for

(y -axis offset for viewing) high symmetry molecules where many of the modes are
optically forbidden.

Adsorbate (S) Another interesting facet of the vibrational IETS is the
Intensity
weakness of overtone and combination bands. There are
sound theoretical reasons to expect that overtone bands
Oxide
OMT via
should be very weak in IETS.19,51 To our knowledge,
Al+ Pb+ there has been no theoretical investigation of the intensi-
Positive bias = 1.0 V ties of combination bands in tunneling spectra. However,
12 000 0 12 000
there are experimental papers that contain tunneling band
(a) (b) Energy /cm1
assignments that include assignments as combination and
Figure 3. Schematic energy diagram (a) and actual tunneling overtone bands. Most of these are unreliable in that they
spectrum (b) obtained from an AlAl2 O3 tetracenePb tunnel
diode. The energy of the LUMO in tetracene is such that at were either made by simply choosing the energetically near-
a forward bias of about 1 V (Pb positive) elastically tunneling est equivalent IR or Raman band assignment, or were used
electrons can occupy the LUMO producing a transient reduction. when the real issue was chemical composition. Consider
Such OMT can be distinguished from IETS in that these bands two examples. The best known overtone band in all of
are not energetically symmetric about zero bias, while the IETS IETS was the overtone of the AlO motion that occurs
bands occur at the same bias values irrespective of bias direction.
Note also that the OMT band is not scan direction dependent. near 1900 cm1 . In the mid 1980s, however, Igalson and
[Reproduced by permission of the American Chemical Society Adler52 and Gauthier et al.53 showed that this band was
from K.W. Hipps and U. Mazur, J. Phys. Chem., 97, 7803 actually the AlH fundamental stretching motion! Another
(1993).] example comes from the work of Yang and White.72 In
their study of thiourea, they suggest that unspecified com-
be used to determine electron affinity levels and oxidation bination bands may be responsible for the extra bands
potentials for thin films of materials. A useful corollary seen in the NH stretching region of the tunneling spec-
of this work for the vibrational spectroscopist is that a trum. In the NH stretching region no individual bands are
species A with reduction potential more positive than about resolved and it is difficult to determine how many transi-
0 .52 (SCE, saturated calomel electrode) will sponta- tions occur near 3300 cm1 . A more likely explanation for
neously reduce in the Al-Al2 O3 -A-Pb tunnel diode structure. the observed lumpy band is that the NH region is complex
Thus, species such as ferricyanide and tetracyanoethylene because of inhomogeneity in adsorption sites. The observed
(TCNE) are present in the tunnel diode as ferrocyanide and texture might also be due to coupling between lattice and
TCNE or reaction products of TCNE .47,69,70 internal motions.
Based on empirical observation, our statement concern-
ing overtones and combinations is: Overtones do occur,
2.2 Inelastic tunneling
but they are very weak. Combination bands are seldom,
In addition to elastic tunneling, there are other tunneling if ever, observed. Kirtley, for example, says that over-
mechanisms which may contribute to the current. While tones are about a factor of 200 weaker than fundamentals in
OMTS can be used to determine absolute energy levels, the case of the benzoate ion.2,51 Ramsier et al.54 identify a
IETS gives the energy level differences familiar to the single clear overtone in the tunneling spectrum of the phos-
conventional photon spectroscopist. IETS is based upon phite ion (HPO3 2 ). The fundamental associated with this
inelastic scattering as shown in Figure 2. The moving overtone is a very strong PH bending band and occurs at
electronic charge interacts with the time-varying molecular 1034 cm1 . This overtone is about 1/50th the height of the
dipoles (electronic or vibrational) to induce excitation fundamental. All the other bands in the spectrum appear to
of the molecule in the barrier with concomitant loss of be fundamentals. In our own work, we have seen little clear
energy by the electron. This interaction has both a long- evidence for combination and overtone bands. An excep-
and short-range interaction component and so selection tion is the intense CH stretch in the tetraethylammonium
rules are more relaxed than in IR or electron energy ion (2971 cm1 ).37 This band has a very weak overtone at
loss spectroscopy (EELS).19,51,71 As we will show later, 5940 cm1 . The integrated intensity of the overtone is about
there are very few (if any) hard selection rules in IETS. 1/60 of the fundamental band.
Instead, we often speak of selection preferences. In IETS There have been several attempts to calculate the IETS
we see some bands that appear in IR or Raman spectra and intensities for real molecules.19,51,71,73 These calculations
others that are totally optically forbidden. Electronic IETS require significant computational power and produce results
transitions with S 6D 0 are often observed. Thus, IETS that are qualitatively correct but quantitatively wrong.
Perhaps the most useful result of these studies was the procedure. Typically, a high vacuum chamber is used (often
observation of an orientational preference. For dipoles close a glass bell jar system having base pressure of about
to one metal surface, maximum intensity was obtained 4 107 Torr). Most researchers have used liquid nitrogen
when the dipole was normal to the metal surface. For trapped diffusion pumps for establishing the vacuum, but a
dipoles near the center of the barrier, the reverse is true. cryopumped system would be suitable. The substrates are
Let us consider the IETS spectrum on an energetic basis. typically glass microscope slides that have been thermally
If the applied voltage is less than hn/e, the inelastic channel tested by rapid immersion in liquid nitrogen (and discarding
is closed because the final states are already filled (Pauli all that crack) and then thoroughly cleaned in HNO3 /H2 O2 .
principle). At V D hn/e the inelastic channel opens. Further A thin strip of metal, most often Al or Mg, about 1 mm
increases in V result in additional possible final states with wide is deposited on the substrate to form the base metal
an associated increase in current due to this channel. As is (M) strip described in the previous section (Figure 1b). This
depicted in Figure 2(b), there is a break in the IV curve is often accomplished through resistive heating of metal
at V D hn/e. If one measures the conductance, dI/dV, the filaments or appropriate crucibles. The work of Maissel and
opening of the inelastic channel is signaled by a step. Glang83 can be consulted for matching deposition fixtures
Plotting d2 I/dV2 versus V produces a peak at V D hn/e. to the metal to be deposited. The insulator is then either
Both vibrational and electronic transitions may be observed deposited on the metal strip, or grown on it. The best
as peaks in the d2 I/dV2 versus V plots. One cannot only insulators are dense native oxides grown by exposure to
probe states of the barrier region, but also of the metal air or to an oxygen plasma (0.08 Torr pressure and 500
electrodes as well. Thus, phonons and magnons of metal to 1000 ac or dc volt discharge). At this point in the
and semiconductor films have been studied by IETS.74 78 preparation, the molecule of interest is usually adsorbed on
The width of the peaks in IETS depends upon the sharp- the insulator. This can be done by direct vapor deposition
ness of the onset of the inelastic process, which in turn for molecular species, or by adsorption from solution for
depends upon the thermal distribution of electron energies either molecular or ionic species. In either case, the junction
about Ef . Thus, the IETS line width depends on temperature is said to be doped with the molecule of interest, A. It is
and is about 3.5T cm1 K1 .82 Because of this, vibrational also possible to coat the surface with a LangmuirBlodgett
IETS is most often performed below 5 K. Electronic transi- film.42,43,84,85 Because of the exponential dependence of
tions are usually much broader than vibrational transitions;
the device resistance on barrier width, the barrier including
therefore, electronic IETS is usually performed at liquid
adsorbate is typically less than about 6 nm wide. In order
nitrogen temperature and above (77 K).
to satisfy this requirement, the molecules studied must not
In its simplest form, an IET spectrum is a plot of d2 I/dV2
be too large and the method of doping must not leave
versus V. It turns out that using d2 I/dV2 /dI/dV as the
a high density of crystallites or molecular aggregates on
y-axis provides spectra having flatter baselines and is most
the surface. In the case of solution phase doping, it has
appropriate for high bias work.36,79 81 These are called
been found that spin doping can produce uniform adsorbed
normalized tunneling intensities (NTI) or constant mod-
layers of the appropriate thickness provided the solution
ulation spectroscopy. Simple tunneling spectra are mea-
concentration is adjusted appropriately. As a starting point,
sured by applying both a variable bias, V, and a small
1 g L1 is often used.
modulation component, Vf , at frequency, f. A lock-in
amplifier (LIA) is used to detect the 2f signal which Once the doping process is complete, the top electrode
is proportional to d2 I/dV2 . The instrumentation required (most often Pb, Tl, Sn, or Ag) is deposited. Thus, the
for obtaining normalized intensities (NTI) is a bit more adsorbate must be able to survive exposure to high vacuum.
complex.79 81 In general, the bias voltage may be converted In practice, a number of M0 metal strips are laid across the
to the more conventional wavenumbers through the factor MIA strip to form four or more tunnel junctions on the
of 8066 cm1 V1 . The amplitude of the modulation always same substrate.
determines the observed signal strength and resolution. The Another interesting method of introducing the analyte
signal increases as Vf 2 but the experimental line width is into the junction, called infusion doping, relies on the
proportional to Vf .63,82 production of a porous top metal electrode. Thus the
clean MIM0 diode is first produced, and later exposed
to an analyte that can diffuse through the pores and
3 EXPERIMENTAL TECHNIQUES be incorporated into the already assembled junction.86 90
3.1 Sample preparation There is still much to be learned about the pore size
distribution in these metal films and how changes in the
Methods for making MIXM0 tunnel diodes have film fabrication process affect that distribution. There are
been extensively described.1 7 We will only outline the also questions remaining about the long term stability of the
pores. Nevertheless, given the IR-like selectivity of IETS, case of chemisorption reactions, IETS results may be a
there is a great potential here for applying these devices as superposition of the spectrum of the reacted species and
chemical sensors. of crystallite edges from unreacted adsorbate.
There are two interesting side issues concerning the The second issue concerning the barrier region has to
barrier region that come from the literature. The first is do with multilayer barriers. Nolen and Ruggiero32 have
the morphology of spin doped layers. When spin doping measured the vibrational IETS of C60 and C70 using a
compounds are not volatile, a bimodal distribution of multilayer barrier of Cn interleaved with Ge. These barriers
surface species tends to form. Figure 4 shows atomic are much thicker than any previously measured by IETS and
force microscopy (AFM) images of an atomically flat therefore have surprisingly low barrier heights. The origin
Al(111)/AlOx surface after spin doping with 0.2 and of these tiny barrier heights is unclear.
0.5 g L1 tetracyanoquinodimethane (TCNQ) in CH3 CN. Once the tunnel diode is completed, one generally makes
These images have been flattened and are shown on a a four terminal electrical measurement. One of the M arms
gray scale of 100 nm. The images are 10 m on a side. and one of the M0 arms (see Figures 1 and 5) carry the bias
Below the images are height histograms. These histograms and modulation currents, while the other two arms are used
indicate that the surface roughness is bimodal with a high for a true potentiometric bias voltage measurement. The
density of small amplitude adsorption induced roughening. easiest and most direct method for making electrical contact
[The small particle distribution width (7 nm) in Figure 4 to the completed tunnel diode is through direct soldering
is about a factor of three larger than the value obtained with pure indium metal. Indium is very sticky (it can be
from the solvent only treated alumina surface.] The obvious used to solder glass to glass) and noncorrosive. Most other
microcrystals seen in the AFM image are the origin of the solders will require a flux to adhere to the metal electrodes
broad peak extending to above 50 nm. For a typical organic and the flux invariably eats through these thin metal films.
compound, the area coated with microcrystals would not Spring loaded pin contacts have also been successfully used
contribute to the tunnel current because the distance through to make solder-free low temperature contacts. Others have
the crystals is too large. The edges of the microcrystals, used printed circuit board edge connectors.91,92
however, might contribute to the current. Thus, in the
3.2 Spectrometer design

(a) (c)
There are no commercial tunneling spectrometers, so

these studies must be performed with home made
instruments. A variety of designs are described in the
literature.21,79 81,93 96 The simplest design appropriate for
the vibrational region of the spectrum (0 to 500 mV or 0
to 4000 cm1 ) is shown in Figure 5. This instrument21 is
relatively easy to build provided that an ultra-low distortion
(b) (d)
50.0
Height (nm) 50.0
Oscillator Lock-in (2f )

Height (nm)
R1
Ref. f Input Output Ref.
7.75 nm 6.28 nm
Recorder Ramp
L L
0
(bias)
0
y-axis x-axis
0 1.00 2.00 0 1.00 2.00 3.00

Hist (%) Hist (%)
Figure 4. AFM top view (10 m on a side and 100 nm gray DVM
scale) and bearing histogram of TCNQ in CH3 CN spin doped
onto atomically flat Al(111)-Al2 O3 . The solutions were 0.5 g L1 Substrate
(a and b) and 0.2 g L1 (c and d) and a 5-s spin time was used. Figure 5. Block diagram of a simple d2 V/dI2 IETS spectrometer
[Reproduced by permission of The Chemical Society of Japan suitable for most vibrational studies. [Reproduced by permis-
from M. Higo, X. Lu, U. Mazur and K.W. Hipps, Chem. Lett., sion of the American Chemical Society from K.W. Hipps and
679 (1999).] U. Mazur, J. Phys. Chem., 97, 7803 (1993).]
oscillator operating at frequency, f, a high input impedance is the experimental equivalent of the slit width in an optical
dc digital voltmeter (DVM), and an LIA are available. spectrometer. Reducing the modulation voltage by two
The resistor, R1 , is set to about 100 times the junction increases the resolution by a factor of two and decreases
impedance. The inductors, L, must provide high impedance the signal by a factor of four.
at f and 2f. The capacitors prevent the LIA and oscilla- If neither metal is superconducting, the half width at 1/e
tor from providing a low impedance dc path to ground. height, , of an IETS band is given (in cm1 ) by:82
The spectrum obtained is a plot of root mean square (rms)
voltage amplitude measured at the LIA at frequency 2f D [2.29T2 C 7.07Vf 2 C 20 ]1/2 7
plotted versus the energy as determined by the DVM. In
analogy with conventional optical spectroscopy, the oscil- where temperature T is in Kelvin, Vf is the rms modulation
lator and ramp together act as the source whose slit in millivolts, and 0 is the intrinsic half width of the line
width is determined by the modulation amplitude. The LIA in cm1 . When one or both metals is superconducting, the
measures signal intensity, and the DVM provides precise line narrows and shifts slightly to higher energy.25,63,97
values for the energy (cm1 ) axis through the conversion, At the greater bias voltages typical of electronic tunnel-
1 mV D 8.066 cm1 . This particular spectrometer actually ing, the spectrometer shown in Figure 5 is not appropriate
measures d2 V/dI2 . To see this consider a Taylor series since the conductance increases rapidly above about 0.7 V
expansion of the voltage across the device about the bias and the LIA output is scaled as (1/s3 ). The lower trace of
current, Ib . Figure 6 demonstrates how rapidly intensity is lost at high
voltages in a normal d2 V/dI2 spectrum. A much improved
dV spectrometer results when an ac voltmeter is added to mon-
VI D VIb C I Ib
dI itor the rms modulation voltage, Vf , across the tunnel
2
1 d V junction and a feedback loop is inserted to maintain constant
C I Ib 2 C 2 modulation voltage and, therefore, constant resolution.79 81
2 dI2
In the more complex designs, the magnitude of the 2f sig-
The total current through the device is the sum of the bias nal is determined in order to correct for phase changes that
current provided by the ramp and the modulation current occur at very high bias.79 The constant resolution spectrum
amplitude, i. Thus, the total current is I D Ib C i cos2pft. of the same tunnel junction as used to obtain the d2 V/dI2
Substituting into equation (2), we find spectrum is also shown (offset slightly) in Figure 6. Con-

dV stant resolution spectra offer several experimental advan-
VI D VIb C i cos2pft tages. The most important benefit of the constant resolution
dI
2 spectrometer is made clear by comparing the two types of
1 dV
C [i cos2pft]2 C 3
2 dI2
Expanding the expression in cos2pft2 , we find a term
in cos[2p2ft], namely the rms amplitude detected by the
LIA tuned to 2f is:
2 2
1 dV i
V2f D p 4
4 dI2 2
Intensity
The above quantity may be converted to d2 I/dV2 through

the relationship:
2 3 2
d2 I d V dI d V
D D s3 5 (a)
dV2 dI2 dV dI2
where the conductance of the device at a particular bias is s, (b)
and s D dI/dV. Since the conductance is nearly constant
throughout the vibrational region of the spectrum, there are 0 8000 16 000
usually no qualitative differences between d2 I/dV2 versus Energy /cm1
V and d2 V/dI2 versus V vibrational spectra. Figure 6. Comparison of constant resolution (a) [(ds /dV/s]
The rms modulation voltage across the junction: and (b) d2 VdI2 spectra obtained from the same Al-Al2 O3 -
(Et4 N)2 CoCl4 -Pb junction at 4 K. [Reproduced by permission of
i the American Chemical Society from K.W. Hipps and U. Mazur,
Vf D 6 J. Phys. Chem., 97, 7803 (1993).]
1.414s
spectra presented in Figure 6. The signal detected by the intrinsic width of electronic bands is generally much greater
LIA in constant resolution mode is: than vibrational bands and the improvement in resolution
with bias voltage predicted by equation (6) has no practical
Vf 2 ds/dV
V2f D 8 value. Worse, since the signal level varies as the square
1.68 s
of Vf , the spectrometer of Figure 5 actually introduces
This output is directly related to ds/dV/s, or NTI. It a significant decrease in signal to noise at higher bias
is a much more useful quantity than d2 V/dI2 at high bias voltages.
because the elastic background increases monotonically and While acquiring phase independent NTI spectra over
almost linearly instead of maximizing and then decreasing large voltage ranges requires two LIAs and computer
sharply. In Figure 6, for example, the very strong transition control,79 a good approximation to NTI can be obtained
near 12 000 cm1 is not visible in the d2 V/dI2 spectrum. at intermediate bias with an inexpensive and apt solution
The second benefit of constant resolution spectra is recently demonstrated by Seman and Mallik.81 Their circuit
the relative ease of reporting a quantitative measure of is shown in Figure 7. It is placed between the oscillator and
transition intensity, s/s, the overall change in conduc- the junction and amplifies the modulation current as nec-
tance associated with an inelastic channel scaled by the essary to maintain constant modulation voltage across the
overall conductance. If the total change in conductance, s, junction.
is small, then the area under an inelastic band is:

ds/dV s
dV 9 3.3 Cryogenics, bias asymmetry, and
band s s superconductivity
where s D se C si , the sum of elastic and inelastic tun-
neling terms. In order to determine s/s from d2 V/dI2 As indicated earlier, the ultimate line width in IETS
or d2 I/dV2 spectra, the conductance must be separately is directly proportional to the sample temperature. In
measured and recorded. Empirical observation suggests that vibrational IETS this generally means immersing the sam-
s/s 0.8% for most vibrational bands and s/s 15% ple in liquid helium. This can be done in a fairly straight-
for most electronic transitions. forward and inexpensive (save for He cost) way. A sample
At first it may seem a disadvantage that in NTI the rod about 122 cm long having sufficient leads to connect
resolution remains constant throughout the spectral range to all the junctions on a given substrate is used. The IETS
studied rather than progressively decreasing. However, the substrate, sample rod, and wiring can all be made small
Rv1
Rv2
Junction OutS R2
R4
R6
1
signal C2
monitor In R3
C3 D1 R8
U1 U3
C1 R 1 + U2 + U4
+ +
R5 R7 + +15V
C4
Rv3
D2 R9 Rv6 R13
Rv5
+ R11
1 2 3 R12
C5 U5
U6
6 5 4 + U7
Oscillator In
C6 +
S2 Rv7
input
Out C7
R10 In
Rv4 S3 Signal
Out output
Figure 7. Modulation feedback amplifier for use in NTI measurements. C1 D 0.22, C2 D 0.0047, C3 D 0.56, C4 D 3.3 F, C5 D 10.0
and C6 D 0.0022 F. R1 D R3 D R9 D R11 D 10, R2 D 470, R4 D 330, R5 D 4.7, R6 D R12 D 1, R7 D R10 D 100, R8 D panel mounted
1.5, R13 D 15, Rv1 D 10 k, Rv2 D panel mounted 100 k, Rv4 D Rv5 D Rv7 D 100 k, Rv6 D 1 000 k. D1 D D2 D 1N4004, U1 D
U2 D U6 D U7 D LF357N, U3 D U4 D LF411 CN, and U5 D NTE3085 photon coupled bidirectional analog field effect transistor
(FET). S1 , S2 , and S3 are selected by a ganged rotary switch. [Reproduced by permission of the American Institute of Physics from
T.R. Seman and R.R. Mallik, Rev. Sci. Instrum., 70, 2808 (1999).]
enough to fit down the neck of most helium storage dewars. barrier height and width.51,53,98,99 An electron tunneling
Thus, the canister in which the helium is delivered becomes from the Al electrode to the Pb electrode (normal bias)
the temperature bath. For electronic IETS and OMTS, liq- does not lose energy by inelastic scattering with the
uid nitrogen temperature more than suffices and a standard adsorbate until it is near the Pb electrode. An electron
liquid nitrogen storage dewar can be used. At this point tunneling in the opposite direction, however, will lose
it might be wise to note that Pb reacts very rapidly with energy while still facing a significant width of barrier.
humid air. Most practitioners of IETS wire the junctions According to the transition probability shown in Figure 1,
and store them in liquid nitrogen until they are ready to the probability of inelastically tunneling in the reverse
make measurements. direction
(from
p Pb to Al)is less by a factor of roughly
p
Inelastic structure should and does appear in both bias exp d f f eV . Until recently, it was thought
directions. Typically, the spectrum of an MIAM0 that this bias direction dependence was universal and most
junction is only measured with M0 relatively positive, i.e. tunneling spectra were only recorded in the Al, or normal,
the normal bias direction. Figure 8, for example, shows bias direction. For electroactive species where OMT bands
the tunneling spectra of two differently doped Al-Al2 O3 - are possible, it is essential to have both polarity spectra to
A-Pb junctions recorded with the Pb electrode positive aid in distinguishing OMT bands from inelastic electronic
(Al) and in reverse bias.35 (The CH stretching mode transitions. Thus, we now strongly advise that spectra be
in this compound is the strongest IETS vibrational band recorded in both bias directions.
we have ever measured; s/s D 1.8%.) Note that all the It is also of interest to note the intense structure due
vibrational bands on the normal bias (Al electrode negative) to Pb phonons in Figure 8. These intense features actually
side are more intense. Figure 8(b) includes a strong dd appear smaller than they are due to the coarse (4 mV) data
electronic transition.35 Even for this strong and broad acquisition interval. These spectra were taken at 4 K where
electronic feature, the reverse bias (AlC) spectrum is Pb is superconducting. The very strong electronphonon
weaker than the normal bias side. This bias dependence is coupling completely dominates the region from about
a natural consequence of the asymmetry in the barrier and 260 to C260 cm1 . When it is desirable to study dopant
the exponential dependence of tunneling probability upon bands in this spectral region, it is possible to quench the
superconductivity, and therefore most of the Pb phonon
IETS, by applying a magnetic field with an inexpensive
2A 2
1g B 2g CoSm magnet.
Pb phonon
(b)
2A 2 4 REPRESENTATIVE SPECTRA WITH

B 2g
(CH)
1g
COMPARISON TO IR AND RAMAN
(CH)
4.1 Organic compounds

Intensity
Early in the development of IETS, Simonsen et al.25

(a)
reported that the tunneling spectrum of anthracene shows
both IR only and Raman only active bands with equal
intensity while a small number of strong IR and Raman
active bands do not appear. They also compared IR
and Raman spectra of benzoic acid adsorbed on bulk
Al+ Al alumina with their IETS of benzaldehyde on alumina
(both chemisorb to form benzoates). They found that,
10 000 0 10 000 the agreement in peak positions is amazingly good the
Energy /cm1 agreement between the tunneling peak positions and those
Figure 8. Magnitude of the constant modulation tunneling spec- from IR and Raman data is as close as between IR and
tra (NTI) of two different junctions as a function of bias polarity. Raman positions. Cass et al.26 studied long chain fatty
Spectrum a was obtained from an Al-Al2 O3 -(en)silane-Pb junc- acids by IETS. They found that IR and Raman intensities
tion. Spectrum b was obtained from a similar device but with
Co2C complexed to the (en)silane. [Reproduced by permission of of bands characterized by different wave vector, k, values
the American Chemical Society from K.W. Hipps and U. Mazur, decrease as k increases, but that the tunneling intensities of
J. Phys. Chem., 97, 7803 (1993).] these modes are similar. They attributed this to the short
wavelength of the tunneling electron relative to an IR or

Raman photon.
The first focused experimental study of selection
preferences was the high resolution tunneling study of
anthracene in Al-Al2 O3 -Pb junctions.61 Anthracene has
66 fundamentals and 37 of these were observed in the
IETS below 1600 cm1 . The normal mode analysis for (a)
anthracene gives 12A1g C 5A1u C 11B1g C 6B1u C 4B2g C
11B2u C 6B3g C 11B3u modes. Of these, 33 are Raman
active (A1g , B1g , B2g , B3g ), 28 are IR active (B1 , B2u , B3u ),
Intensity
and 5 are optically inactive out of plane motions (A1u ).
Kirtley and Hansma61 concluded that band positions in
IETS are in good agreement with Raman and IR, and that (b)
Raman and IR active modes appear with roughly the same N
intensity, no conclusions could be drawn about orientation C
selection rules, and anthracenes optically inactive modes N
N C C C
have very small intensity. C C
Hall and Hansma55 measured and assigned the IETS C C
spectra of methanesulfonate (CH3 SO3 ) and trifluoro-
N N (c)
methanesulfonate. Assuming C3v symmetry for the ad-
sorbed CH3 SO3 , vib D 5A1 C A2 C 6E. Both A1 and E
0 1000 2000
modes are IR and Raman active while the A2 mode is
optically inactive (forbidden). The A2 mode could not be Energy /cm1
observed since it occurs below the spectral window used Figure 9. (a) IETS, (b) Raman (powder), and (c) IR (in KBr)
in the reported study. All the other fundamentals were spectra obtained from the cesium salt of pentacyanopropenide,
CsPCP. [Reproduced by permission of the American Chemical
observed. The tunneling intensities were not a simple lin- Society from K.W. Hipps and U. Mazur, J. Phys. Chem., 97, 7803
ear combination of IR and Raman spectra. The authors say, (1993).]
In fact, no measured tunneling intensities have ever been
reproduced as a simple linear combination of infrared and particular spectrum because the PCP tunneling spectrum
Raman intensities. is a weak one.
A graphical comparison of the results obtained from IR, The tricyanomethanide ion, C(CN)3 , is a planar D3h
Raman, and IETS studies of the same species is shown in symmetry ion that behaves chemically like a halide. The
Figure 9. IETS, Raman, and Fourier transform infrared (FT- vibrational analysis gives, vib D 2A01 C A02 C 4E0 C 2A002 C
IR) of the pentacyanopropenide ion (PCP ) all compare E00 . Of these, the A01 and E00 modes are only Raman active,
favorably in terms of peak positions but have significantly the E0 modes are both IR and Raman active, the A002 modes
different intensity distributions.57 PCP has C2v symmetry are IR active only, and the A02 mode is inactive in both IR
and is shown as an insert in Figure 9. The IETS has a and Raman.22 Since the inactive mode is a bending motion,
similar number of bands to that of the Raman, but the IETS it is expected to occur in the region below 800 cm1 . This
bands are more uniform in intensity. The IR spectrum, on region of the IR, Raman, and tunneling spectrum is shown
the other hand, has fewer strong bands than the tunneling in Figure 10.22 Note that the low frequency A002 mode is
spectrum. The line widths in IETS taken at 4 K are very observed in the tunneling spectrum. This band could be
similar to those in the IR, and broader than the Raman lines. seen because the superconductivity of the Pb electrode was
When narrower lines are needed, a reduction in line width quenched by a static magnetic field providing an appropriate
(by about a factor of 2) of the IETS bands can often be window.
achieved by simply attaching a mechanical vacuum pump In Figure 10 the IETS lines are clearly broader than the
to the helium dewar and using the heat of vaporization of IR or Raman bands, but this is an instrumental artifact.
the helium to cool it to below 2 K. The IETS spectrum was recorded at 4.2 K and 2 mV (rms)
Another point to note in Figure 9 is the strong broad modulation. The line width is therefore almost completely
feature near 900 cm1 in the tunneling spectrum. This determined by the modulation voltage (equation 7). The
band is due to AlO motion in the oxide barrier and is assignments in Figure 10 were made based upon data
present in every tunneling spectrum obtained from an Al- acquired by all three spectroscopic methods. Moreover,
Al2 O3 -M0 tunnel diode. It is especially prominent in this the peak positions from three different isotopomers (the
While the spin-doped junctions provided only 24 of the

E 1
N 46 modes, these only partially overlapped the 24 modes
C seen in inelastic neutron scattering (INS) and the 4 IR and
C A1
10 Raman modes. Thus, all but 9 of the 46 fundamental
C C bands could be assigned through a combination of all
N N
(c) four data sets. Recently, Nolen and Ruggiero32 have used
A2 an interesting composite barrier design to make tunnel
Intensity
Al phonon junctions that provided IETS wherein 26 of the silent

A2 (b)
modes of C60 were identified. Thus, all but 6 of the
E 46 fundamentals have been directly observed through the
E concerted use of IR, Raman, INS, and IETS.
A2 The above comparisons indicate that IETS data can be
(a)
usefully combined with Raman and IR spectra taken from
0 400 800 conventional samples, e.g. KBr pellets. Nonetheless, IETS
Wavenumber /cm1 is a matrix isolation technique and the presence of an
Figure 10. IETS (a), FT-IR (b), and solution phase Raman (c)
oxide substrate and a metal overlayer might produce severe
spectra of the C(CN)3 -ion. [Reproduced by permission of the chemical changes in the molecule of interest (for an extreme
American Chemical Society from K.W. Hipps and U. Mazur, J. case, see Mazur et al.100 ). We know, for example, that
Phys. Chem., 97, 7803 (1993).] organic acids generally adsorb on alumina as carboxylate
ions. The quest for a better understanding of organic
natural abundance, 13 C enriched, and 15 N enriched species) chemisorption has included a number of IETS studies,
were measured and an 11 parameter empirical force field highlighting the fact that the insulators typically used in
was fitted to the data. This procedure provided a 2 cm1 IETS do react with many compounds. Moreover, the optical
standard deviation fit to all the normal modes of the ion.22 selection rules might be sufficiently modified by the simple
Note that the inactive A02 mode appears as a medium reduction in symmetry associated with the diode structure to
intensity band in the tunneling spectrum, making it easy account for all the intensity variations seen between optical
to identify and to evaluate isotopic shifts. One further spectra of conventional samples and tunneling spectra.
point of interest is the band near 300 cm1 in the IETS. The top metal electrode may also introduce steric effects
This is a vibrational mode of the aluminum electrode (a that change the conformation of the adsorbed species. To
phonon) and can be seen in Figure 10 only because the address these issues requires several different experimental
tricyanomethanide ion adsorbs weakly. approaches.
Graves and White28 analyzed the IETS of phenazine, The first step in any tunneling study must be a care-
a planar D2h symmetry ion having vib D 11Ag C 5Au C ful determination of what species is actually present in the
5B1g C 10B1u C 4B2g C 10B2u C 10B3g C 5B3u . Of these diode. To this end, tunneling spectra obtained from isotopi-
modes, 30 are only Raman active (11Ag , 5B1g , 4B2g , cally substituted species and thorough comparison of IR
10B3g ); 25 are only IR active (10B1u , 10B2u , 5B3u ); and 5 and Raman spectra with tunneling spectra are invaluable.
are optically inactive out of plane motions of Au symmetry. We have also found it very helpful to determine what, if
Graves and White report that Raman and IR active modes any, reactions occur between conventional high surface area
are seen with similar intensities in the IETS. Moreover, two oxides and the molecular species of interest.
of the strong peaks (456, 858 cm1 ) and one strong shoul- The effect of the tunneling environment on band posi-
der (949 cm1 ) are assigned to optically forbidden modes. tions and optical selection rules has been studied by mea-
A weak shoulder at 731 cm1 is assigned as a fourth Au suring IETS and Raman or IR spectra in environments
mode. similar to, and in some cases identical with, that of a tunnel
An unusual class of organic compounds are the bucky diode. Surface enhanced resonance Raman scattering was
balls, of which buckminsterfullerene (C60 ) is the most used to study 4-pyridine-carboxylic acid chemisorbed in an
widely known. Early attempts to measure IETS through Al-Al2 O3 -Ag junction.19,56 The intensity patterns for IETS
conventional spin doping yielded encouraging results.31 and resonant Raman spectra were quite different, indicating
C60 has 176 possible vibrational eigenvalues, but the high that the local environment of the ion does not account for
symmetry and associated degeneracy reduce the number of the difference in IETS and optical selection rules. Both IR
unique frequencies to 46. These modes have one-, three-, and Raman active modes are observed in the IETS with no
four-, and five-fold degeneracy. Four of these modes (the significant peak shifts but the IETS spectrum has signifi-
F1u ) are IR active, and 10 are Raman active (2Ag C 8Hg ). cantly different intensities. While not all the IETS lines are
seen in the resonance Raman spectrum, all the Raman lines modes, and significant differences in intensity of symmetric
have counterparts in the IETS spectrum. and asymmetric carboxylate motions were observed in IETS
Hipps and Keder58 used the intrinsic electronic resonance relative to MRAIRS. They argue that spectral analysis may
enhancement of the Raman spectrum of PCP to probe the be enhanced by recording both IET and MRAIR spectra.
tunneling environment. Raman spectra of a single mono- Since these volumes are dedicated to vibrational spec-
layer of the pentacyanopropenide anion (PCP ) adsorbed troscopy, we will resist the temptation to discuss the recent
on the alumina surface of an unfinished Al-Al2 O3 -Pb tun- advances in the IETS and OMTS of electronic state spec-
nel diode were compared to the IETS obtained from the troscopy. References given in the theory section should act
finished junction. While both A1 and B2 in-plane motions as a good starting point for the interested reader.
appear in the Raman spectrum of the ion in solution, only
the A1 modes appear strongly in the monolayer film Raman
spectra. In contrast, the IETS spectrum contains A1 and B2 4.2 Inorganic compounds
in-plane modes as well as out of plane motions at 453 and
468 cm1 .101 Comparison of IR and Raman spectra taken The relationship between IETS, IR, and Raman spectro-
from pure solid CsPCP with IETS data shows that the out- scopy for an inorganic compound, potassium ferrocyanide,
of-plane motions are weak in both photon techniques but is illustrated in Figure 11. The ferrocyanide ion is octa-
strong in IETS.58 hedral in solution but is slightly distorted in the potassium
The resonance enhanced Raman scattering of metal half-hydrate crystal20,27 so the Raman spectrum shows more
phthalocyanines (MPcs) also was used to probe completed bands than expected. It is clear from Figure 11 that the
tunnel diodes.23 Raman spectra of Al-Al2 O3 -MPc-M0 tunnel IETS contains different information than either the Raman
diodes, where M D Cu or H2 and M0 D Ag or Pb, were or IR spectrum and is not simply a linear combination of
reported. No evidence was found for chemical modification IR and Raman results. Let us consider further the IETS
of either CuPc or H2 Pc at any step in the fabrication of hexacyanides, showing that some optically forbidden
process. No shifts in band positions were observed that transitions are strongly allowed in the tunneling spectrum.
amounted to more than the experimental uncertainty of The octahedral M(CN)6 4 ion has 13 fundamentals, vib D
2 cm1 .16 In this study Raman spectra were taken of 2A1g C 2Eg C T1g C 4T1u C 2T2g C 2T2u , six of which are
working tunnel diodes while they were under bias. By Raman active (2A1g , 2Eg , 2T2g ), four are IR active (4T1u ),
observing the Raman spectra of functioning tunnel diodes, and three are totally inactive, or forbidden (T1g , 2T2u ).20,21
it was demonstrated that no modification of the MPc Raman Thus, a large fraction of the ions spectrum, and the corre-
spectrum occurs in the bias range 1.3 to C1.3 V. Thus, sponding information about the molecular forces present in
despite a local field of about 3 106 V cm1 , Stark shifts the ion, is not available to the optical spectroscopist.
of less then 2 cm1 occurred. The Raman spectra provide Table 1 identifies the 13 normal coordinates of the
solid evidence that CuPc and H2 Pc are not significantly ferrocyanide ion in terms of both mode number and
modified by being deposited onto the alumina layer, by symmetry. Also given in Table 1 is the method by which
being coated with Ag or Pb, or by application of a static a particular fundamental was observed. The Raman and
potential of up to 1.3 V across the device. Even the Stark IR data were obtained from solid K4 Fe(CN)6 3H2 O, a
shift is small. crystalline environment that differs only slightly from
Mallik and Henriksen have pioneered the use of multiple octahedral. This reduction in symmetry can be seen in
reflection-absorption infrared spectroscopy (MRAIRS) as the presence of a weak n7 band (IR active) in the Raman
a tool for comparative studies with IETS.29,30,62,102 Vibra- spectrum.
tional spectra were obtained using IETS and MRAIRS to Vibrational spectra (IR, Raman, and IETS) of 13 C and
study monolayers of pyruvic acid adsorbed on alumina 15
N isotopomers of the ferrocyanide ion were measured
incorporated in thin film Al/alumina/pyruvic acid/Pb sam- and analyzed.20,21 Although the CN stretching motions
ples. Spectra obtained using both techniques do not exhibit in both the IR and IETS appear as a poorly resolved
a carbonyl peak, however MRAIR spectra for similar sam- band, well resolved metalcyanide bands are found in all
ples without a Pb cover film do show the carbonyl stretch. three methods. Four prominent tunneling transitions in the
This behavior may be attributed to interactions between the region between 300 and 700 cm1 are assigned to bending
adsorbed molecules and the cover film. Their study cor- motions; two of these (n5 and n12 ) are optically forbidden
roborates the above listed Raman studies in that the effect modes. Using the peak positions and isotopic shifts reported
of the internal field of the diode is negligible on the posi- in Table 1, a 10-parameter valence-force potential was
tions and intensities of typical IR bands. MRAIRS seems to fit to the data. The band positions and isotopic shifts
be intrinsically less sensitive than IETS to CH stretching calculated from this potential are also reported in Table 1.
assigned.20 Here again the four fundamentals seen in the

K4Fe(CN)6 3H2O metalligand region of the IETS correspond to one Raman
active (n10 ), one IR active (n7 ) and two inactive (n5 and
8 in KBr n12 ) modes. They are all bending modes. While the internal
CN stretches are observed as a weak broad band in the
IETS, the metalCN stretches were not observed. As is the
(a) 10 case for Fe(CN)6 4 , the tunneling line widths are similar to
5 those in the IR. The solution Raman bands are wider than in
12 1, 3, 6 the solid state and are comparable to the tunneling widths.
Intensity
7 K4Fe(CN)6
An interesting consequence of the change in bonding upon
going from iron to ruthenium is the order reversal of n10 and
(b)
2 n12 . In the osmium salt, n12 is almost intermediate between
n10 and n7 . These changes are paralleled by a collapse in
4
the n2 to n10 spacing.
K4Fe(CN)6 1/2 H2O
7
Layered compounds, such as GaSe, InSe, FePS3 ,
8 10 Ga2/3 PS3 , and MnPS3 are especially good candidates for
(c) single crystal IETS studies.74,77,78 One first deposits a base
metal electrode on the single crystal, then the crystal is
300 600 1900 2200
successively cleaved (usually with Scotch tape) until it
Energy /cm1
is only a few nanometers thick, and then the top metal
Figure 11. Comparison of (a) IR, (b) IETS and (c) Raman
electrode is deposited. In this way the phonon and magnon
spectra of the ferrocyanide ion obtained under different conditions.
[Reproduced by permission of the American Chemical Society spectra of these interesting materials may be determined
from K.W. Hipps and U. Mazur, J. Phys. Chem., 97, 7803 by IETS.
(1993).] Three aluminum alloys, each with different major alloy-
ing components, have been studied using IETS and X-ray
The data presented in Table 1 allow the four force constants photoelectron spectroscopy (XPS).103 As may be seen in
associated with the forbidden modes to be determined with Figure 12, the oxides grown on alloys containing magne-
precision. sium and copper have spectra similar to that grown on
The vibrational IETS, Raman, and IR spectra of pure aluminum, and their spectra when doped with benzoic
Ru(CN)6 4 and Os(CN)6 4 have also been measured and acid also show no differences. The alloy containing 7%
Table 1. Observeda and calculated positions of vibrational fundamentals for potassium ferrocyanide
trihydrate based on a 10-parameter potential functionb .
Mode Obsd Calc 13 C 15 N Observed in

Obsd Calc Obsd Calc
A1g n1 2090 2088 45 47 29 30 Raman

n2 385 386 7 7 7 8 Raman
Eg n3 2059 2055 44 47 29 29 Raman, IETS
n4 392 393 8 7 8 8 Raman
T1g n5 354 355 9 10 1 3 IETS
T1u n6 2044 2049 47 45 35 30 IR, IETS
n7 586 588 11 13 1 3 IR, IETS
n8 415 413 7 6 6 5 IR
n9 83 1 2
T2g n10 511 511 18 17 1 2 Raman, IETS
n11 113 1 3
T2u n12 474 475 16 15 0 3 IETS
n13 61 1 2
a Based upon IR, Raman, and IETS data. Mean percent error D 0.25%.
b 13 C D n13 C n12 C.
[Reproduced by permission from Hipps and Mazur (1993). Copyright 1993 American Chemical Society.]
Wavenumber /cm1 junctions at this temperature. Full IETS spectra reported are
500 1000 1500 2000 2500 3000 3500 4000 essentially identical to those previously recorded for similar
SiO2 films on aluminum. Atomic force and scanning tun-
ALOX
neling microscopic images of the SiO2 and underlying gold
films were also presented; these data confirm that the SiO2
(1/dI /dV0)(d2I /dV 2)V 1 (arbitrary units)
H30 films are sufficiently uniform and continuous for IETS. The
authors provide a useful AFM/STM evaluation of the bar-
FC1 rier and junction surface. However, they have been unable
to make these SiO2 films thin enough to allow IETS of
adsorbed species to be observed.
The structure of MoS2 /Al2 O3 catalyst and the initial
step of the hydrodesulfurization (HDS) reaction have been
studied by in-situ Raman, infrared emission (IRE), and
IETS spectroscopies.12 In tandem, thermal desorption mea-
100 200 300 400 surements were performed and accompanied by molec-
LM25 ular modeling and EH-MO (extended Huckel molecu-
lar orbital) calculations. The results show that the cata-
lyst consists of MoS2 -like crystallites which are linked
via MoOAl bonds almost perpendicular (but slightly
(Not to same scale)
inclined) to the surface of g-Al2 O3 . The orientation of
100 200 300 400 500 the MoOAl bonds was determined by IETS and sub-
Bias potential (mV) stantiated by calculation. A pretreatment of the surface
Figure 12. IETS of oxides grown on pure aluminum (ALOX), Al of the MoS2 crystallites with hydrogen is mandatory to
with 1%Si, 1% Mg, and 1% Mn (H30), Al with 5.5% Cu and 0.7% remove all kinds of S2 groups and to form active (for-
Fe (FC1), and Al with 7% Si and 0.5% Fe (LM25). [Reproduced mally) Mo(II) centers at the edge positions. Thiophene,
by permission of John Wiley & Sons from R.G. Pritchard and acting as a Lewis acid only towards such centers, is destabi-
C.R. Werrett, Surf. Interface. Anal., 23, 705 (1995).]
lized upon coordination and becomes susceptible for further
reaction with hydrogen. This is an interesting combined use
silicon gives a different spectrum that has band positions of IRE, Raman, and IETS, with IETS playing an essen-
similar to those of aluminosilicates. When doped with ben- tial role.
zoic acid, the spectrum of the latter alloy also shows major Vibrational spectra of thin (0.21.0 nm) films of ther-
differences to that of benzoic acid on aluminum (oxide), mally evaporated Ge on Al2 O3 surfaces were measured
both in band intensities and particularly in the carboxy- by IETS and compared with those of the Si films.105 The
late stretching modes. XPS analysis of the oxides on the tunneling spectra were also compared with the vibrational
alloy thin films and metal tokens confirms the results sug- spectra of surface species on crystalline Ge formed from
gested by the spectra magnesium and copper do not reactions with H, H2 O, and O measured by high-resolution
appear in the oxides of the first two alloys but silicon is EELS and multiple internal reflection infrared spectroscopy.
present in the oxide of the third and in quantities greater Monohydride is formed predominantly in films prepared
than that of the bulk alloy. This paper is interesting both both in high vacuum (106 Torr) and in an atmosphere con-
as an analytical application, and because it provides a taining H2 O vapor (105 Torr). The tunneling spectra of the
potentially useful way to make a variety of new barrier heated (135 C) Ge films showed that the monohydride is
materials. stable. The vibrational frequencies for the hydride species
It is also possible to make good quality barriers from in the films compare more closely to those on crystalline
amorphous SiO2 . Films can be radio frequency sputter- Ge surfaces than to those in amorphous Ge. The hydride is
deposited onto thin film gold electrodes and incorporated formed from the reaction with residual H2 O molecules in
as the insulating barrier in Au/SiO2 /Pb junctions.104 For the vacuum and is considered to be present mainly on the
many years it was thought impossible to make such tunnel surfaces and grain boundaries of the film. Its role is to ter-
junctions because the presence of small pin holes in the minate dangling bonds. The peak intensity of the hydride
oxide layer would allow AuPb contacts to form. However, species in the tunneling spectra of the Ge films is much
in this study IETS spectra recorded at 4.2 K reveal charac- weaker than that in the tunneling spectra of the Si films,
teristic zero bias anomalies which show unequivocally that indicating a very small number of the hydride species in
tunneling is the primary conduction mechanism through the the Ge films.
4.3 Biological compounds Optically forbidden (IR and/or Raman) fundamental

transitions may be observed. In addition, spin and
A number of papers concerning biological compounds can dipole forbidden electronic transitions are also allowed.
be found in the book by Hansma.1 We will illustrate this Overtone and combination bands are very weak or
research area using two examples from the literature of the absent. Thus, virtually every band is due to a funda-
last five years. We note first that a long standing concern mental transition.
about the application of IETS to large biomolecules (such Very small amounts of compound (1014 molecules)
as hemoglobin) is that they are so thick that tunneling is can provide a complete spectrum.
only effective at the periphery. The success with which The practical resolution limit (4 cm) has proven more
Nolen and Ruggiero32 were able to measure the relatively than sufficient for accurately measuring isotopic shifts.
large C70 via a laminated barrier structure suggests that IETS offers the opportunity of studying molecules in
there yet may be hope for IETS studies of moderately thick buried interfaces thereby providing unique insights into
biomolecules. the effects on molecular states produced by intimate
Brousseau and co-workers106 have studied UV-vis (ultra- contact with metallic and insulating layers.
violet-visible) and vibrational spectra of the brevetoxins IETS provides a high sensitivity method for studying
PbTx-2, PbTx-3, PbTx-9, and 2,3,27,28,41,43-hexahydro- chemisorption and catalysis on alumina, magnesia, and
(reduced) PbTx-3. The absorption at 194 nm of PbTx-2 silica. Model supported metal catalysts have also been
and PbTx-3 is assigned to the p p transition of the studied by IETS.
lactone conjugated double bonds. The n p transition of It is an all solid-state spectroscopy. In principle the
the lactone is seen at 239 nm. The absorption band at sample and spectrometer could be built on a single
208 nm for PbTx-2 is assigned to the conjugated double chip. When this feature is combined with the fur-
bonds of the aldehyde function located on the K-ring ther development of porous electrodes, IETS becomes
side chain. The inelastic electron tunneling spectra of uniquely situated for sensor applications.
PbTx-2, PbTx-3, PbTx-9, and fully reduced PbTx-3 are
presented and the vibrational bands are assigned. The
authors show that IETS can be used to monitor and measure ACKNOWLEDGMENT
the vibrational fingerprint of the brevetoxins at a level as
low as 1 ng. Acknowledgment is made to the donors of the Petroleum
IETS has also been used for the detection and character- Research Fund, administered by the ACS and to Research
ization of biological compounds in human sweat adsorbed Corporation for partial support of this work.
on alumina surfaces.107 The sweat was collected from stu-
dents and distributed to aqueous and benzene phases by
solvent extraction. Both solutions were doped onto alu- ABBREVIATIONS AND ACRONYMS
mina surfaces and then coated with cross Pb electrodes
to form tunneling junctions. The vibrational spectra of the AFM Atomic Force Microscopy
adsorbed species were then measured by IETS. Analy- DVM Digital Voltmeter
sis of these tunneling spectra showed that lactic acid in EELS Electron Energy Loss
the sweat is extracted with water and selectively adsorbed Spectroscopy
on alumina. Adsorbed fatty acids were detected from the EH-MO Extended Huckel Molecular
benzene phase. These biological compounds in the sweat Orbital
solutions were found to react with the alumina to give FT-IR Fourier Transform Infrared
about a monolayer of adsorbed carboxylate anions on the HDS Hydrodesulfurization
surfaces. HOMO Highest Occupied Molecular
Orbital
IETS Inelastic Electron Tunneling
5 SUMMARY OF BENEFICIAL Spectroscopy
FEATURES INS Inelastic Neutron Scattering
IRE Infrared Emission
Although IETS requires building a spectrometer, develop- LIA Lock-in Amplifier
ing vacuum deposition skills, and the use of cryogenics, it LUMO Lowest Unoccupied Molecular
offers a number of benefits to the spectroscopist. The most Orbital
important of these are: MIAM0 MetalInsulatorAdsorbateMetal
MIM MetalInsulatorMetal 17. J. Comyn, C. Horley, R. Pritchard and R.R. Malik, Adhe-
MPcs Metal Phthalocyanines sion (London), 11, 38 (1987).
MRAIR Multiple Reflection-absorption 18. H.W. White, L.M. Godwin and R. Ellialtioglu, Adhesion,
Infrared 13, 177 (1981).
MRAIRS Multiple Reflection-absorption 19. J. Kirtley, ACS Symp. Ser., 137, 217 (1980).
Infrared Spectroscopy 20. K.W. Hipps, S.D. Williams and U. Mazur, Inorg. Chem., 23,
NTI Normalized Tunneling Intensities 3500 (1984).
OMT Orbital Mediated Tunneling 21. K.W. Hipps and U. Mazur, J. Phys. Chem., 84, 3162
OMTS Orbital Mediated Tunneling (1980).
Spectroscopy 22. K.W. Hipps and A.T. Aplin, J. Phys. Chem., 89, 5459
rms Root Mean Square (1985).
SCE Saturated Calomel Electrode 23. J.J. Hoagland, J. Dowdy and K.W. Hipps, J. Phys. Chem.,
STM Scanning Tunneling Microscopy 95, 2246 (1991).
TCNQ Tetracyanoquinodimethane 24. J. Kirtley and J.T. Soren, Phys. Rev. B, 22, 848 (1980).
UV-vis Ultraviolet-visible
25. M.G. Simonsen, R.C. Coleman and P.K. Hansma, J. Chem.
XPS X-ray Photoelectron Spectroscopy Phys., 61, 3789 (1974).
26. D.A. Cass, H.L. Strauss and P.K. Hansma, Science, 192,
1128 (1976).
REFERENCES
27. K.W. Hipps, U. Mazur and M.S. Pearce, Chem. Phys. Lett.,
68, 433 (1979).
1. P.K. Hansma Tunneling Spectroscopy, Plenum Press, New
York (1982). 28. R.J. Graves and H.W. White, Spectrochim. Acta, 43A, 107
(1987).
2. W.H. Weinberg, Vib. Spectra Struct., 11, 1 (1982).
3. D.G. Walmsley and J.L. Tomlin, Prog. Surf. Sci., 18, 247 29. S. Devdas, R.R. Mallik, R. Coast and P.N. Henriksen, Surf.
(1985). Sci., 326, 327 (1995).
4. J.T. Yates and T.E. Madey (eds), Vibrational Spectroscopy 30. R. Coast, M. Pikus, P.N. Henriksen and G.A. Nitowski, J.
of Molecules on Surfaces, Plenum Press, New York (1987). Phys. Chem., 100, 15 011 (1996).
5. J.C. Reviere, Surface Analytical Techniques, Clarendon 31. J.L. Brousseau, K. Tian, S. Gauvin, R.M. Leblanc and
Press, Oxford, Chapter 20 (1990). P. Delhaes, Chem. Phys. Lett., 202, 521 (1993).
6. K.W. Hipps and U. Mazur, J. Phys. Chem., 97, 7803 (1993). 32. S. Nolen and S.T. Ruggiero, Chem. Phys. Lett., 300, 656
(1999).
7. S. Reynolds and D.P. Oxley, Inelastic Electron Tunnelling
Spectroscopy (IETS): Principles, Applications and Future 33. S. de Cheveigne, J. Klein and A. Leger, Electronic State
Directions, in Proceedings of the International School Tunneling, in Tunneling Spectroscopy, ed. P.K. Hansma,
on Condensed Matter Physics, ed. J.M. Marshall, World Plenum Press, New York, Chapter 4 (1982).
Scientific, Singapore, 209217 (1997). 34. K.W. Hipps, J. Phys. Chem., 93, 5958 (1989).
8. A. Sleigh, C. Adkins and W.A. Phillips, Vacuum, 38, 301 35. K.W. Hipps and U. Mazur, Surf. Sci., 207, 385 (1989).
(1988).
36. K.W. Hipps and U. Mazur, J. Phys. Chem., 91, 5218 (1987).
9. S. de Cheveigne, S. Gauthier, C. Guinet, M. Lebrun, J. Klein
and M. Belin, J. Chem. Soc., Faraday Trans. 2, 81, 1375 37. K.W. Hipps and U. Mazur, J. Am. Chem. Soc., 109, 3861
(1985). (1987).
10. M. Higo, H. Hayashi and S. Kamata, Appl. Surf. Sci., 32, 38. S. de Cheveigne, J. Klein, A. Leger, M. Belin and
338 (1988). D. Defourneau, Phys. Rev. B, 15, 750 (1977).
11. M.J. Line, R.G. Pritchard, R. Pynenburg, R.J. Latham and 39. A. Adane, J. Klein, A. Leger, M. Belin and D. Defourneau,
R.G. Linford, Surf. Interface Anal., 23, 565 (1995). Phys. Rev. B, 29, 4443 (1984).
12. E. Diemann, Th. Weber and A. Mueller, J. Catal., 148, 288 40. H. Luth, U. Roll and S. Ewert, Phys. Rev. B, 18, 4241
(1994). (1978).
13. E.S. Yamaguchi and P.R. Ryason, Tribol. Trans., 36, 367 41. U. Roll, S. Ewert and H. Luth, Chem. Phys. Lett., 58, 91
(1993). (1978).
14. E.S. Yamaguchi, P.R. Ryason and E.Q. Labrador, Tribol. 42. M. Wada, T. Kubota and M. Iwamoto, Jpn. J. Appl. Phys.,
Trans., 38, 243 (1995). Part 1, 32, 3868 (1993).
15. S. Kamata, A. Hida, M. Higo and H. Inadome, J. Phys. 43. M. Iwamoto and T. Kubota, Proceedings of the Interna-
Chem., 98, 1648 (1994). tional Conference on the Properties of Applied Dielectric
16. S. Kamata, A. Hida, M. Higo and M. Yazaki, Microchem. Materials, Institute of Electrical and Electronics Engineers,
J., 49, 194 (1994). New York, 197200, Vol. 1 (1994).
44. C. Garrigou-Lagrange, P. Delhaes, V. Yartsev, J. Brousseau International Conference on the Physics of Semiconduc-
and R.M. Leblanc, Chem. Phys. Lett., 225, 297 (1994). tors, eds P. Jiang and H.-Z. Zheng, World Scientific, Sin-
45. U. Mazur and K.W. Hipps, J. Phys. Chem., 98, 5824 gapore, 661664, Vol. 1 (1992).
(1994). 75. Y. Ando, J. Murai and T. Miyazaki, J. Magn. Magn. Mater.,
46. U. Mazur and K.W. Hipps, J. Phys. Chem., 98, 8169 198199, 161 (1999).
(1994). 76. R.J. van de Veerdonk, J.S. Moodera and W.J de Jonge, J.
47. U. Mazur and K.W. Hipps, J. Phys. Chem., 99, 6684 Magn. Magn. Mater., 198199, 152 (1999).
(1995). 77. K. Yamaguchi, Phys. Status Solidi B, 179, K11 (1993).
48. U. Mazur and K.W. Hipps, J. Phys. Chem. B, 103, 9721 78. K. Yamaguchi, Phys. Status Solidi B, 190, 409 (1995).
(1999). 79. K.W. Hipps and U. Mazur, Rev. Sci. Instrum., 59, 1903
49. J.D. Langan and P.K. Hansma, Surf. Sci., 52, 211 (1975). (1988).
50. A.A. Cederberg, Surf. Sci., 103, 148 (1981). 80. K.W. Hipps and U. Mazur, Rev. Sci. Instrum., 58, 265
51. J. Kirtley and P. Soven, Phys. Rev. B, 19, 1812 (1979). (1987).
52. J. Igalson and J.G. Adler, Phys. Rev. B, 28, 4970 (1983). 81. T.R. Seman and R.R. Mallik, Rev. Sci. Instrum., 70, 2808
(1999).
53. S. Gauthier, S. de Cheveigne, J. Klein and M. Belin, Phys.
Rev. B, 29, 1748 (1984). 82. K.W. Hipps and S.L. Peter, J. Phys. Chem., 93, 5717 (1989).
54. R.D. Ramsier, P.N. Henriksen and A.N. Gent, Surf. Sci., 83. L.I. Maissel and R. Glang, Handbook of Thin Film Tech-
203, 72 (1988). nology, McGraw-Hill, New York, Chapter 1 (1970).
55. J.T. Hall and P.K. Hansma, Surface Sci., 71, 1 (1971). 84. M. Iwamoto and T. Kubota, Synth. Met., 71, 1981 (1995).
56. J.C. Tsang, J.R. Kirtley, T.N. Theis and S.S. Jha, Phys. Rev. 85. J.L. Brousseau, S. Vidon and R.M. Leblanc, J. Chem. Phys.,
B, 25, 5070 (1982). 108, 7391 (1998).
57. U. Mazur and K.W. Hipps, J. Phys. Chem., 86, 2854 86. W.J. Nelson, D.G. Walmsley and J.M. Bell, Thin Solid Films,
(1982). 79, 229 (1981).
58. K.W. Hipps and J.W. Keder, J. Phys. Chem., 87, 3186 87. M. Higo, X. Lu, U. Mazur and K.W. Hipps, Chem. Lett.,
(1983). 679 (1999).
59. K.W. Hipps, J. Dowdy and J.J. Hoagland, Langmuir, 7, 5 88. J.R. Bellingham, C.J. Adkins and W.A. Phillips, Thin Solid
(1991). Films, 198, 85 (1991).
60. K.W. Hipps and J.J. Hoagland, Langmuir, 7, 2180 (1991). 89. K.W. Hipps and U. Mazur, J. Phys. Chem., 96, 1160 (1992).
61. J. Kirtley and P.K. Hansma, Surf. Sci., 66, 125 (1977). 90. U. Mazur, X.D. Wang and K.W. Hipps, Anal. Chem., 64,
1845 (1992).
62. R. Coast, M. Pikus, P.N. Henriksen and G.A. Nitowski, J.
Phys. Chem., 100, 15 011 (1996). 91. D.P. Oxley, A.J. Bowles, C.C. Horley, A.J. Langley, R.G.
Prithchard and D.L. Tunnicliffe, Surf. Interface Anal., 2, 31
63. S. Reynolds and D.P. Oxley, Surf. Sci., 368, 324 (1996). (1980).
64. K.H. Gundlach, Festkorperprobleme, XI, 237 (1971). 92. Y. Wang, R.R. Mallik and P.N. Henriksen, Rev. Sci. Ins-
65. D. Barlow, U. Mazur and K.W. Hipps, J. Phys. Chem., 104, trum., 64, 890 (1993).
2444 (2000). 93. A.B. Dargis, Rev. Sci. Instrum., 52, 46 (1981).
66. B.C. Stipe, M.A. Rezaei and W. Ho, Rev. Sci. Instrum., 70, 94. S. Colley and P. Hansma, Rev. Sci. Instrum., 48, 1192
137 (1999). (1977).
67. J.G. Simons, J. Appl. Phys., 34, 2581 (1963). 95. G.J. Gajda and W.H. Weinberg, Rev. Sci. Instrum., 56, 700
68. C.S. Korman, J.C. Lau, A.M. Johnson and R.V. Coleman, (1985).
Phys. Rev. B, 19, 994 (1979). 96. T. Huriuchi, F. Ebisawa and H. Tabei, Rev. Sci. Instrum.,
69. K.W. Hipps and U. Mazur, J. Phys. Chem., 86, 5105 60, 993 (1989).
(1982). 97. J.R. Kirtley and P.K. Hansma, Phys. Rev. B, 13, 2910
70. K.W. Hipps and U. Mazur, Rev. Sci. Instrum., 55, 1120 (1976).
(1984). 98. J. Kirtley, D.J. Scalpino and P.K. Hansma, Phys. Rev. B, 14,
71. R. Knockenmuss and K.W. Hipps, J. Phys. Chem., 86, 4477 3177 (1976).
(1982). 99. M.F. Muldoon, R.A. Dragoset and R.V. Coleman, Phys.
72. W.J. Yang and H.W. White, Surf. Sci., 118, 303 (1982). Rev. B, 20, 416 (1979).
73. L.M. Godwin, H.W. White and R. Ellialtioglu, Phys. Rev. 100. U. Mazur, S.D. Williams and K.W. Hipps, J. Phys. Chem.,
B, 23, 5688 (1981). 85, 2305 (1981).
74. K. Yamaguchi, Inelastic Electron Tunneling through Anti- 101. K.W. Hipps, U. Geiser, U. Mazur and R.D. Willett, J. Phys.
ferromagnetic Semiconductors MPS3 (M D Fe, Mn), in Chem., 88, 2498 (1984).
102. S. Devdas and R.R. Mallik, Int. J. Adhesion Adhesives, in 105. M. Higo, M. Isobata and S. Kamata, J. Phys. Chem., 97,
press. 4491 (1993).
103. R.G. Pritchard and C.R. Werrett, Surf. Interface Anal., 23, 106. J.L. Brousseau, N. Vibet, N.R.M. Leblanc, K.S. Rein and
705 (1995). D.G. Baden, Appl. Spectrosc., 52, 523 (1998).
104. R.R. Mallik, T. Butler, W.J. Kulnis and B. DeVier, J. Appl. 107. M. Higo and S. Kamata, Anal. Chem., 66, 818
Phys., 73, 2347 (1993). (1994).
High-resolution Electron Energy Loss Spectroscopy
Michael A. Chesters
1 INTRODUCTION spectroscopy was not widely appreciated until after that

time), and Raman spectroscopy remains difficult to apply
In the field of vibrational spectroscopy applied to chemi- to monolayers of small molecules unless under the spe-
cal analysis the contribution of electron energy loss spec- cial conditions of surface-enhanced Raman spectroscopy
troscopy (EELS) has been in applications to solid surfaces, (SERS).6 EELS currently remains a widely used method
monolayers and thin films, under ultra-high vacuum (UHV) in surface science and its applications have continued to
conditions. The electron energy loss technique has been be driven by improvements in spectrometer performance.
applied to atoms and molecules in the gas phase since While the advantages of the technique are fully realized
the 1920s,1 where it was mainly applied to the study of only in experiments on single crystal surfaces and ordered
electronic excitations, although with sufficient resolution to monolayers, there are increasing numbers of examples of
observe vibrational fine structure. Different acronyms have its application to surfaces of a more practical nature, such
been applied to this technique with the aim of distinguish- as polymer films.
ing the high resolution (210 meV) measurement of purely
vibrational excitations from other applications of excitation
by electron energy loss. Hence HREELS, (high-resolution 2 INSTRUMENTATION
electron energy loss spectroscopy) and VEELS (vibrational
electron energy loss spectroscopy) have both been used, The basic elements of an electron energy loss spectrometer,
and indeed the latter is the IUPAC recommended acronym, which must be mounted in a UHV chamber, are
but the simpler EELS remains the most common acronym.
an electron gun, in which electrons are thermionically
The first reported application to solid surfaces, in 1967,2
emitted from a hot cathode, accelerated/decelerated and
demonstrated that it was possible to measure vibrational focused through a series of electrodes;
spectra of monolayers of atoms or small molecules with a monochromator, which takes the beam from the
high sensitivity but relatively poor resolution (100 cm1 ). electron gun with an energy spread of 300 meV
Improvements in instrument design and the appearance of (1 meV D 8.065 cm1 ) and selects a narrow energy
commercial UHV instruments that could routinely achieve distribution (110 meV);
5 meV (40 cm1 ) resolution, led to rapidly growing interest a zoom lens, which projects the beam across a
during the 1970s and 1980s in EELS as a surface sci- field-free scattering chamber and focuses it onto a
ence technique.3,4 The considerable impact of EELS owed sample surface with kinetic energies adjustable in the
much to its lack of competitors. Infrared (IR) spectroscopy 1100 eV range;
did not achieve the routine sensitivity over the full mid- a second zoom lens, which decelerates the scattered
infrared range necessary for application to a wide variety beam and focuses it onto the entrance slit of an
of monolayer materials until the late 1980s5 (or at least analyzer;
the full potential of Fourier transform infrared (FT-IR) an analyzer, which selects an energy spread of scattered
electrons similar to that selected by the monochroma-
John Wiley & Sons Ltd, 2002. tor and scans downwards in mean energy from the
Second analyzer Sample
Second monochromator
Cathode
Channeltron First analyzer Lenses
Emission
First monochromator system
Figure 1. Schematic diagram of an EELS spectrometer. [Reproduced from H. Ibach, M. Balden, D. Bruchmann, S. Lehwald, Electron
Energy Loss Spectroscopy: Recent Advances in Technology and Application, Surface Science, 269/270, 94 (1992) with permission
from Elsevier Science.]
impact kinetic energy in order to measure energy loss a coefficient that expresses the space charge effect, both
processes; of which may be optimized by spectrometer design. The
a channeltron electron multiplier detector. effect of space charge is to limit success in passing high
electron fluxes through narrow slits due to the spreading
A recent design is shown in Figure 1, which serves to of the beam arising from electronelectron repulsion. This
illustrate some of the improvements introduced over the consideration originally led to the belief that 35 meV was
past two decades.7 The initially most successful designs
the lower limit of energy resolution likely to be achieved,
employed a double monochromator and analyzer using
which at 2540 cm1 was considerably poorer than typical
127 cylindrical sector devices for all dispersion elements.
linewidths of IR or Raman spectra measured of solids or
The first monochromator typically operated at 2030 meV
surfaces.
resolution (defined as, E1/2 , the half-width of the elec-
The 127 cylindrical sector analyzer is a very simple
tron energy distribution) and was followed by a second
device, which disperses in only one plane. More efficient
monochromator typically operating at 510 meV resolu-
designs, such as the spherical section analyzer3,4 have been
tion. The use of the first, low-resolution, monochromator
used very successfully for EELS but have not generally
improves the level of current extracted from the electron
achieved higher energy resolution than cylindrical sector
gun and the use of double elements for both monochroma-
tor and analyzer reduces the incidence of ghost loss peaks devices. Improvements in EEL spectrometer performance
arising from electrons that have scattered from a surface to give practical currents at 1 meV (8 cm1 ) resolution have
inside the spectrometer. been achieved with the 127 cylindrical section device7
The resolving power, Ep /E1/2 , of electrostatic dis- through:
persing elements is typically a few hundred so that pass design studies using numerical simulation of electron
energies, Ep , need to be generally less than 1 eV to produce trajectories, which led to introduction of repeller elec-
resolution, E1/2 , in the millielectron volt range. Such low trodes above and below the plane of the electron tra-
kinetic energy electrons are very susceptible to small elec- jectory to compensate for spacecharge repulsion;
tric and magnetic fields which are reduced by coating all
the use of a higher pass energy in the first dispersing
electrode surfaces with graphite, to produce uniform work
element of the double monochromator, necessitating
functions, avoiding all possibilities for surface charging and
the introduction of lenses between the first and second
shielding from magnetic fields. The fact that electrons are
elements to decelerate and refocus the be

Vibrational Spectros PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Vibrational Spectros PDF

Загружено:

Авторское право:

Доступные форматы

Volume 1

Theory and Instrumentation

INTRODUCTION TO THE THEORY AND PRACTICE OF VIBRATIONAL

The Historical Development of Experimental Techniques in Vibrational Spectroscopy

INSTRUMENTATION FOR MID- AND FAR-INFRARED SPECTROSCOPY

Resolution and Instrument Line Shape Function

INSTRUMENTATION FOR NEAR-INFRARED SPECTROSCOPY

Monochromators for Near-Infrared Spectroscopy

INSTRUMENTATION FOR RAMAN SPECTROSCOPY

Raman Monochromators and Polychromators

Fast Time-Resolved Mid-Infrared Spectroscopy Using an Interferometer

DICHROISM AND OPTICAL ACTIVITY SPECTROSCOPY IN VIBRATIONAL

Linear Dichroism in Infrared Spectroscopy

SURFACE-ENHANCED VIBRATIONAL SPECTROSCOPY

Electromagnetic Mechanism of Surface Enhanced Spectroscopy

OTHER INSTRUMENTAL APPROACHES FOR VIBRATIONAL SPECTROSCOPY

CALIBRATION PROCEDURES AND STANDARDS FOR VIBRATIONAL SPECTROSCOPY

Applications of Vibrational Spectroscopy in Life, Pharmaceutical and

Vibrational Spectroscopy and Pathology

Infrared Spectroscopy of Proteins

Polymorphs, Solvates and Hydrates

Vibrational Spectroscopy of Food and Food Products

Vibrational Near-infrared SpectroscopyApplication to Grains and Derived Commodities

Glossary of Terms Used in Biomedical, Biochemical and Pharmaceutical Vibrational

Sampling Techniques for Vibrational Spectroscopy

MID- AND NEAR-INFRARED TRANSMISSION SPECTROSCOPY

Standard Sampling Techniques for Infrared Spectroscopy

MID-INFRARED EXTERNAL REFLECTION SPECTROSCOPY

External Reflection Spectroscopy

MID-INFRARED INTERNAL REFLECTION SPECTROSCOPY

Principles, Theory and Practice of Internal Reflection Spectroscopy

DIFFUSE REFLECTION SPECTROSCOPY

Continuum Theories of Diffuse Reflection

OTHER IR SAMPLING TECHNIQUES

Infrared Emission Spectroscopy

The Optical Problem in Raman Sampling

LOW TEMPERATURE AND HIGH PRESSURE SAMPLING TECHNIQUES

Liquefied Gases as Solvents for Vibrational Spectroscopy

DEPTH PROFILING BY VIBRATIONAL SPECTROSCOPY

Depth Profiling by Microspectroscopy

OPTICAL CONDUITS FOR VIBRATIONAL SPECTROSCOPY

Introduction to Fiber Optics for Spectroscopists

Gas Chromatography/Fourier Transform Infrared Spectroscopy

Applying Extractive FT-IR Spectrometry to Industrial and Environmental Gas Analysis

Sample Characterization and Spectral Data Processing

Spectra-Structure Correlations in the Mid And Far Infrared

GROUP THEORETICAL AND NUMERICAL APPROACHES TO THE CALCULATION OF

Group Theory and Effects of Symmetry Lowering on Vibrational Spectra

TWO-DIMENSIONAL (2D) ANALYSIS

General Theory of Two-dimensional (2D) Analysis

SPECTRAL ENHANCEMENT AND BAND RESOLUTION TECHNIQUES

Smoothing and Derivatives in Spectroscopy

ANOMALIES, ARTIFACTS AND COMMON ERRORS IN USING VIBRATIONAL

Mid-infared Spectroscopy: Anomalies, Artifacts and Common Errors

Glossary of Terms used in Chemometrics

Applications of Vibrational Spectroscopy in Industry, Materials and the

ANALYSIS AND CHARACTERISATION OF POLYMERS AND RUBBERS

Qualitative and Quantitative Analysis of Polymers and Rubbers by Vibrational

RHEO-OPTICAL MEASUREMENTS OF POLYMERS AND RUBBERS

Rheo-optical Fourier Transform Infrared Spectroscopy of Polymers

Vibrational Spectroscopy of Carbon and Silicon Materials

PROCESS VIBRATIONAL SPECTROSCOPY

ATMOSPHERIC AND ASTRONOMICAL VIBRATIONAL SPECTROSCOPY

nQ C nQ dnQ where R represents an internal coordinate of the molecule,

d12 mH C mO cos2amO sin2a

S1 mO [1 C cos2a] C mH 0 21/2 sin2a

C6 E C6 C3 C2 C23 C56 e D exp2pi/6

C2v E C2 sv xz sv yz

C3h E C3 C23 sh S3 S35 e D exp2pi/3

D2 V E C2 z C2 y C2 x

D2h Vh E C2 z C2 y C2 x i sxy sxz syz

D2d Vd E 2S4 C2 2C02 2sd

S8 E S8 C4 S83 C2 S85 C34 S87 e D exp2pi/8

D1h E 2Cf 1s v i 2Sf 1C02