Академический Документы
Профессиональный Документы
Культура Документы
Michael Haschke
Laboratory
Micro-X-Ray
Fluorescence
Spectroscopy
Instrumentation and Applications
Springer Series in Surface Sciences
Volume 55
Series editors
Gerhard Ertl, Berlin, Germany
Hans Lüth, Jülich, Germany
Douglas L. Mills, Irvine, USA
Laboratory Micro-X-Ray
Fluorescence Spectroscopy
Instrumentation and Applications
123
Michael Haschke
Eggersdorf
Germany
ISSN 0931-5195
ISBN 978-3-319-04863-5 ISBN 978-3-319-04864-2 (eBook)
DOI 10.1007/978-3-319-04864-2
Springer Cham Heidelberg New York Dordrecht London
Micro X-ray fluorescence spectroscopy has enlarged the application fields of X-ray
fluorescence as an analytical method considerably. Nowadays, the possibility to
investigate inhomogeneous and non-flat samples in the micrometer regime facil-
itates elemental analysis of such different objects like technical layers, pharma-
ceutics, catalysts and art objects.
This monograph demonstrates this abundance in applications by concentrating
on the applicative aspects of Micro X-ray fluorescence analysis with X-ray tubes.
With its introduction to the basics of X-ray interactions important for X-ray
fluorescence analysis and to the main components of X-ray spectrometers, it
provides a solid and comprehensive entry into the method. The main merit is the
coverage of all practical analytical aspects which distinguishes the present
monograph from other, more research that related ones. Hence, not only quanti-
fication methods are described but also, e.g., sample positioning and preparation.
The large amount of application examples illustrates various analytical strategies
with Micro X-ray fluorescence spectroscopy. Especially, this detailed description
turns the book into a valuable tool for new users of the method as well as for
experts.
Such a rich, actual description is only possible with a long experience in the
field. Michael Haschke works since over thirty years for various companies in the
sector of analytical equipment, especially for research and development of energy
dispersive X-ray spectrometers, in the last twenty years his core area was Micro-X-
ray-fluorescence where he was involved in the development of the most sophis-
ticated instruments and the introduction of these instruments into the market. In
summarizing his own experiences and actual developments he donates the ana-
lytical community assistance for an improved praxis.
I like to thank Michael Haschke very much for his ‘‘weekend pleasure’’ writing
down important aspects of his professional life which will be valuable not only for
educating young academics. Beyond that my gratitude refers also to his former and
ongoing commitment for the analytical community.
v
Preface
vii
viii Preface
The author had the privilege to be associated with the development of l-XRF
instruments during the last 20 years in different companies and was involved in the
introduction of new instruments into the market. This enabled to accumulate
experiences on the instrumentation as well as about the application.
There are already several monographs available that give a detailed description
of the interaction of X-rays with matter. Therefore, here these topics are described
briefly and summarize only the basic facts required for the understanding of the
main functions of both the different components of a micro-X-ray fluorescence
spectrometer and the quantification procedures. In this way, the described content
should be easily understandable even for beginners but gives instrument users the
necessary knowledge for an efficient use of the method.
The Chap. 1 of the book describes the interaction of high energetic electro-
magnetic radiation with matter and the basic design of X-ray spectrometers, while
Chap. 2 describes in detail the different components of the spectrometer, in par-
ticular, the components of micro-X-ray spectrometers like different X-ray optics
and their applicability for l-XRF instruments, the different possibilities and
requirements for sample positioning, and the energy-dispersive detectors mostly
used in l-XRF instruments. In Chap. 3 the different geometric arrangements for
l-XRF, the corresponding instrument types, and the measurement modes as well
as the possibilities for the presentation and interpretation of elemental distributions
are discussed.
Quantification for small areas can be different from large areas because sample
homogeneity and also the environment of the analyzed area influence the ana-
lytical result. An important application of l-XRF is the examination of layer
systems. Thickness and composition are of interest. The corresponding quantifi-
cation models are discussed together with general considerations about analytical
errors in Chap. 4.
This is followed by a discussion of different sample preparation methods that
can be used for l-XRF. Also if sample preparation is easy or even not necessary or
possible, it is necessary to consider the analyzed volume and its relation to the
material that should be characterized.
In Chap. 6, a comparison is given with other analytical methods with spatial
resolution including a discussion of possibilities for the combination of l-XRF
with these methods.
Finally, different examples for the application of l-XRF are presented. The
predominant part of measurements were performed with instruments which were
developed in groups under the responsibility of the author. This was for mea-
surements with collimators, mainly the M1 Ora and M1 Mistral from Bruker Nano
GmbH, but also instruments of the series Compact of Roentgenanalytik GmbH, for
measurements with polycap optics, mainly the M4 Tornado from Bruker Nano
GmbH and also the Eagle from EDAX. Because the measurements were per-
formed over a long period of time the instrumental configuration was different for
most of the measurements. Only few applications were performed with other
instruments, though this will be separately mentioned.
Preface ix
Herewith I would like to acknowledge all colleagues and companies that allowed
me to work in the interesting field of l-XRF and gave me the necessary freedom for
a creative and targeted working. At first U. Theis from Röntgenanalytik Mess-
technik GmbH has to be mentioned who allowed me to start with this interesting
topic and supported me for a long time, but also the support from A. Devenish,
EDAX Inc. and T. Schülein, Bruker Nano GmbH was very important—many
thanks to all of them.
Further I would like to thank all the co-workers who were very important
partners for discussions, for the conjoint instrument development, and the creation
and testing of new applications during the introduction of l-XRF. There were a lot
of them but here I like to mention at first K. Erler who was involved in the design
of most of the developed instruments, but also P. Pfannekuch, A. Seaman, B.
Scruggs, and J. Nicolosi from Röntgenanalytik/EDAX; N. Kempf from IfG; U.
Waldschläger and R. Tagle from Bruker which was essential for the development
of new applications and for their testing with a lot of measurements and for the
discussions required for an optimization.
I will not forget important users like D. Musale, G. Havrilla, B. Warner and N.
Zahler, S. Hanning, J. Dik, N. Mantouvalou, and many others who not only
allowed me the presentation of some of their very interesting application results in
this paper, but were primarily also important partners during the development and
optimizing of the equipment. They had the time to use the instruments and could
give a lot of hints to make their handling more convenient, to improve the effi-
ciency of data collection and evaluation, and most important, to give hints for
further requirements. I remember gratefully a lot of discussions with these partners
where we generated a lot of new ideas for the improvement of the analytical
performance of the instrumentation and also thought of new applications. Many
thanks for this creative cooperation!
Further, I think appreciatively for many discussions with M. Haller, D. Gibson
or Ning Gao, and S. Bejeoumikhova and also others where I learned a lot about the
main part of l-XRF instruments—X-ray optics, in particular capillary lenses.
xi
xii Acknowledgments
Further I would like to thank people for discussions about the content of the
book and their helpful hints for the different chapters, mainly W. Malzer, B.
Kanngießer, H. Legall, and S. Noack. Last but not least, I will mention that a lot of
work from J. Metz, L. Follett and A. Seaman was necessary for editing and
correction of my limited English.
Contents
1 XRF-Basics . . . . . . . . . . . . . . . . . . . . . . . . . . ............... 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . ............... 1
1.2 Interaction of X-rays with Matter Used
for Material Characterization . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Emission of Fluorescence Radiation . . . . . . . . . . . . . . . 6
1.2.3 Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.4 Scattering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.5 Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3 General Design of X-ray Spectrometers . . . . . . . . . . . . . . . . . . 14
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
xiii
xiv Contents
4 Quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.2 Different Types of Quantification . . . . . . . . . . . . . . . . . . . . . . 160
4.2.1 Qualitative and Semi-quantitative Methods . . . . . . . . . . 160
4.2.2 Quantification Methods . . . . . . . . . . . . . . . . . . . . . . . . 160
4.3 Quantification for l-XRF . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
4.3.1 Special Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
4.3.2 Quantification with the Fundamental
Parameter Model. . . . . . . . . . . . . . . . . . . .......... 168
4.3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . .......... 176
Contents xv
7 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.1 Point Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.1.1 Analysis of Precious Metal Alloys . . . . . . . . . . . . . . . . 230
7.1.2 Coating Thickness Analysis . . . . . . . . . . . . . . . . . . . . . 236
7.1.3 Analysis of Particles and Inclusions . . . . . . . . . . . . . . . 249
7.1.4 Analysis of Restricted Elements in Consumer Goods . . . 256
xvi Contents
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Symbols
B Distribution of brightness
BG Background
c Velocity of light
Cdet Detector capacitance
d d-spacing of the scattering lattice
dworking Working distance
Dsource Source diameter
D Multidimensional data cube for mapping data
e Charge of an electron
E Energy
EC Energy of the absorption edge
f Functional relation
F Focal distance
G Gain
h Planck constant
i Current
I Intensity
Io Primary intensity
Id Leakage current
IScat Scattered intensity
k Factor of confidence
LOD Limit of detection
m Atomic mass, medium count rate
m0 Mass of an electron
M Matrix interaction
n Refraction index, number of channels, pixels
N Number of photons
P Probability
Q Mass per area
R Intensity ratio
t Thickness of a layer, measurement time
T Temperature, transmission function of optics
U Voltage, uncertainty
v Measured value
xvii
xviii Symbols
w Mass fraction
Z Atomic number
D Difference
g Efficiency
e Energy for generation of a charge carrier
k Wave length
l Mass attenuation coefficient
lCompound Mass attenuation coefficient of a compound
q Density of the absorbing material
0 Scatter angle
0crit Critical angle
w Incident, take off angle
s Shaping time, linear absorption coefficient
x Fluorescence yield
X Captured angle
Chapter 1
XRF-Basics
Abstract This chapter describes the main interactions of X-rays with matter and
how these can be used for the characterization of material. Such interactions are
absorption, emission of fluorescence (or secondary) radiation, refraction, scattering
and diffraction. This discussion will be limited because there are several papers
and even monographs which describe these interactions very detailed. An over-
view of the main components of an X-ray spectrometer with their expected
functionality is provided at the end of the chapter.
1.1 Introduction
The excitation of small sample areas was realized at the beginning only with
collimators [19–26]. But this was limiting the excitation intensity even in case of
using high power X-ray tubes with rotating anodes [27] which was restricting also
the analytical performance. Smaller spots with higher intensities could be realized
with X-Ray optics only. The properties of capillaries for the propagation of
X-Rays were investigated already in the 1970s [28, 29]. With the availability of
X-ray optics, in particular of capillary optics also smaller spots was possible and
then l-XRF could be performed also with laboratory instruments. The first
instruments were prepared with mono-capillaries [30–34]. In this case the spot size
could be already very small, down into the 5 lm range [34, 35] but the measured
intensities was also very low. The availability of poly-capillary optics then offers
the combination of small spots—at the beginning down to 40–50 lm—with high
excitation intensity which allows the use of this technique also for laboratory
instrumentation [36–42].
New components for l-XRF methods as well as new applications often are
tested at first on synchrotron sources and only then they are transferred to labo-
ratory instruments. This is valid for the introduction of X-ray optics but also for
other technologies, quantification methods and a wide range of applications. Also
nowadays synchrotrons are used for more sophisticated investigations, for example
with higher spatial resolution or flexible excitation conditions. But their limited
availability and high effort for the examination requires also the availability of
laboratory instruments for the general daily use.
Therefore this book concentrates only on the description of laboratory instru-
ments because meanwhile they are available in a lot of laboratories and with
different instrumental concepts—both as prototypes in research laboratories but
also as commercially available instruments. This enables their use nowadays for a
wide range of different applications.
A more detailed discussion of l-XRF with synchrotron radiation and a com-
parison of differences in analytical performance will follow in 6.1.1.
Already a short time after the discovery of X-rays there was several attempts to use
this form of radiation for the characterization of material.
The short wavelengths of this radiation which is in the range of the atomic
distances in solid and liquid material and the high penetration into material offers
different possibilities for the examination of material. There are different inter-
actions of X-rays with material that can be used—absorption, scattering or dif-
fraction, refraction and emission [for example 43–48]. These interactions are
schematically displayed in Fig. 1.1.
4 1 XRF-Basics
1.2.1 Absorption
If X-rays incident into material with atomic number Z they interact with the
material and will be attenuated. This interaction is described by the mass
absorption l. The mass absorption is the sum of absorption and scattering which
are described by s and r, respectively. The contribution of scattering to attenuation
is relatively small that means
with
Io primary intensity
l mass attenuation coefficient as function of energy
q density of the absorbing material
t thickness of the absorbing layer
The mass attenuation coefficient depends on the radiation energy i.e. l = l (E). If
the material is composed of different elements the mass attenuation coefficient can
be calculated as the average
X
lcompound ¼ wi li ð1:3Þ
1.2 Interaction of X-rays with Matter Used for Material Characterization 5
with
wi mass fraction of element i
li mass absorption coefficient of element i
This interaction was the first which was seen by Roentgen itself. The image of his
hand with a ring is well known. X-ray absorption is often used for medical
applications. X-rays penetrate for example the human body but will be different
absorbed in the bones and tissue. This allows the generation of images of the bones
and the different tissues, particularly when materials of heavy elements that are
injected into the blood’s circulation system or into the alimentary tract cause
greater contrast between the differing tissues and organs.
But very soon this interaction was also used for the analysis of inorganic
materials. The first target was a 2-dimensional imaging of the material for the
investigation of its internal structure, for example, the identification of inclusions
like blow holes or cracks in metals. If the X-ray source is a point source it is even
possible to adjust the image size by changing of the distances between X-ray
source, sample and detector. Then the image can be magnified which simplifies the
image post-processing.
But then, with the availability of more computing power it was also possible to
generate 3-dimensional images due to the collection of absorption images in dif-
ferent directions and their reconstruction to a complete model of a specimen—this
is called X-ray tomography. Tomography can be performed for large bodies like
technical design elements by using tube voltages in the range up to 350 kV. It can
be done for the human body with tube voltages in the range up to 150 kV but also
for small grains, particles or films with tube voltages in the range up to 80 kV. The
reconstruction generates a 3-dimensional model of the investigated material with
possibilities to prepare cross sections in different directions or semi-transparent
models to show all interesting details.
For all these absorption applications the energy of X-rays has to be adapted to
the size of the analyzed sample and to their matrix to have a sufficient dynamic
range of the signal. The energy of the X-rays needs to be large enough to penetrate
the sample that should be analyzed but there should be also a sufficient absorption
that a change of the signal can be detected.
One of the latest applications of X-ray absorption is a real X-ray microscopy
[49–53]. In that case the sample is illuminated with X-rays and the absorbed
radiation will be magnified by using X-ray optics with a similar beam path as in
optical microscopes. Typically Fresnel-lenses are used for that purpose. They can
magnify the image of the examined sample. The magnification depends on the
structure of the Fresnel-lenses—in particular from the dimensions of the outer ring
of the optic. So far the spatial resolution is limited to the range of approx. 20 nm.
Due to the high absorption this technique can be used in the moment only for light
materials like organic materials.
6 1 XRF-Basics
The absorption process of X-rays in matter can generate phonons i.e. enhance the
oscillations of the lattice or they excite atoms by the emission of a photo-electron.
Due to the high energy of the incident X-ray this electron can come even from the
inner shells of the atom. If the vacancy in this shell is filled up by outer electrons
the atom is going into the ground state and energy can be emitted as electro-
magnetic radiation. This radiation has typically energies in the range of X-rays.
This process is called X-ray fluorescence because it can be excited by radiation or
it can be called also emission of characteristic radiation because their energies are
characteristic for the involved atom [54–59, 60]. Moseley [61, 62] found in 1913
that there is a relation between the energy of characteristic radiation E and the
atomic number Z of the emitting atom which is called now Moseley’s law:
E ¼ C1 ðZ C2 Þ2 ð1:4Þ
energy emission is possible, the probability for their sum needs to be unit i.e.
pAuger + pX-ray = 1. The probability for an X-ray emission is called fluorescence
yield x and depends on energy and the electron transition as shown in Fig. 1.3.
The emission of characteristic X-rays can be used for the determination of
elemental composition of samples but also for the analysis of coating systems
according to their thickness and composition.
In addition to the characterisitic radiation with discrete energies there is also a
continuous radiation in the energy range of X-rays. This is called bremsstrahlung
according to its generation process (for a more detailed decription see 2.1.1). This
possibility of X-ray fluorescence was used already very early for elemental
analysis. The first instruments were built in the 30th of the last century for the
analysis of homogeneous samples [64, 65]. Typical applications were in metal-
lurgy and in mining.
The special advantage of X-ray fluorescence is the good repeatability and
reproducibility i.e. their high precision. The high precision is result of the small
contribution of statistical errors caused by the high count rate. Another benefit is
the wide range of weight fractions that can be covered with XRF.
These exceptionally useful aspects of XRF mean it is now used in countless
laboratories for a large variety of applications. Several generations of X-ray
fluorescence spectrometers have been developed and used. Nowadays there are
different types of X-ray spectrometers available which can be used for different
analytical tasks:
• The detection of fluorescence radiation allows on the lowest level the deter-
mination of the elements that have emitted this radiation i.e. a qualitative
analysis.
• The peak intensity of every peak depends on the amount of atoms of this
element in the analyzed volume. This allows the calculation of the weight
fractions of existing elements. The weight fraction range covered by XRF
depends on the used method but can be very large—up to 6 orders of magnitude.
It depends also on the examined element. Very light elements cannot be ana-
lyzed due to the very low energy of their characteristic radiation.
8 1 XRF-Basics
• Accuracy requirements are also different—it ranges from the control of levels of
weight fractions to their accurate determination with the help of references or
even with bracketing procedures.
• The sample quality can vary from liquids to powders and solid samples, also the
analysis of material streams are possible.
• In case of layer systems the peak intensity depends on the layer thickness. For
the analysis of layer systems further knowledge about the structure of the layer
system is necessary.
1.2.3 Refraction
The refraction index n in the energy range of X-rays can be written as:
n ¼1 dþ ib ð1:5Þ
with
d deviation from 1 (approx. 10-6–10-7)
b absorption coefficient
This shows the refraction index for X-rays is for all materials very close to 1. It
depends mainly on the density. Therefore differences between refraction indices
for X-rays in different materials are very small and their refraction is negligible.
Further it can be seen that all refraction indices are negative. That means the very
small refraction is in direction of the higher density and according to Snellius’s
law total reflection happens for X-rays only for very small incident angles and
materials with higher density.
X-ray refraction is not used for material characterization but for X-ray optics
refraction is an important process which can be used for beam shaping (see
Sect. 2.2.3).
1.2.4 Scattering
X-rays are scattered on electrons. This scattering can be described by the classical
electromagnetic theory as well as with the particle model for the X-ray photon.
Both descriptions give different behaviors of the scattered X-rays.
The electromagnetic theory describes an elastic scattering i.e. without a loss of
energy of the scattered X-ray [66, 67]. This scattering is also called coherent or
Rayleigh-scattering. For randomly distributed directions of the electromagnetic
field of the X-rays the scattering can be described by the Rayleigh formulae:
ffi 2 2
1 e
Iscat ¼ I0 2 ð1 þ cos2 0Þ ð1:6Þ
r m0 c2
1.2 Interaction of X-rays with Matter Used for Material Characterization 9
with
IScat scattered intensity
I0 primary intensity
r distance to the observation point
e charge of an electron
m0 mass of an electron
0 scatter angle
This formula shows no change of energy but a dependence from the scatter angle.
In Fig. 1.4 this dependence is displayed in polar-coordinates i.e. the scatter
intensity is presented as the distance from the point of origin (incident point of the
radiation) in every direction.
For scattering angles H close to 0 or 180 i.e. in forward or backward direction
the scattered intensities is high but for scattering with angles close to 90 i.e.
scattering perpendicular to the incident beam the scattered intensity has a
minimum.
By using the corpuscular image for the description of the scattering the X-ray
photon hits an electron. Due to that hit the photon transfers energy and momentum
to the electron. For that process both energy and momentum conservation is valid.
Therefore the energy loss of the photon depends on the on its scatter angle.
Because the scattered photon has lost energy it is called inelastic scattering or also
incoherent or Compton-scattering [68]. Due to the loss of energy the wavelength
of the photon is increased. The wavelength of the scattered radiation kscatt can be
described as follows:
with
kc = h/mc Compton wavelength
m mass of the scattering particle, here an electron
k0 wavelength of the incident photon
0 scatter angle
E0
EScatt ¼ ð1:8Þ
1 þ mc1 2 ð1 cos 0Þ
with:
E0 energy of incident radiation.
This formula shows that the energy loss of the photons depends on the energy of
the incident photon and on the scattering angle. The dependence of the energy
change E0 – Escatt on the scattering angle is displayed in Fig. 1.5—again in polar-
coordinates i.e. the energy change is displayed as the distance from the origin in
every scatter direction. It shows that for scatter angles close to 0 i.e. if the photons
don’t hit the electron there is no loss of photon energy. For small scattering angles
only a small fraction of the photon energy is transferred to the electron. The largest
energy loss happens in case of a direct hit of the photon with the electron i.e. for a
scattering angle close to 180. Then a large fraction of the photon energy is
transferred to the electron and the Compton shift has a large value.
This behavior is demonstrated in Fig. 1.6 which shows the scattered tube
radiation on a Plexiglas sample (PMMA) for two scatter angles in backward
direction. The blue spectrum had an angle between tube and detector of 78 i.e. a
scatter angle of 102, for the red spectrum the scatter angle is 128.
The spectra show that both elastic and inelastic scattering are responsible for
the spectral background. The elastic scattered fluorescence radiation of the tube
(Rh) can be seen in the spectrum as narrow peaks both for K- and L-radiation. The
large peak in front of the Rh-K-line is the Compton-scattered K-line. This
Compton-peak is less shifted towards lower energies for the smaller scattering
angle (blue spectrum). But this dependence is valid not only for the fluorescence
radiation of the tube but also for their bremsstrahlung. This can be seen by a
comparison of the slope of the bremsstrahlung—also this shows a larger shift for
the larger scatter angle. Further the energy dependence of the Compton shift can be
seen—for the Rh-Ka line at 20.2 keV it is smaller than for the Rh-Kb-line at
22.8 keV.
The Rh-L-radiation is not Compton scattered because the energy is too small
and the scattering probability is reduced (see Fig. 1.8).
That these shifts are only valid for scattered radiation can be seen on the Zr-peak
which is excited on the detector collimator and is therefore not Compton shifted.
1.2 Interaction of X-rays with Matter Used for Material Characterization 11
Fig. 1.6 Scattered tube spectra with scatter angles of 102 (blue) and 128 (red)
For the energy range that is interesting for X-ray spectroscopy i.e. from 1 to
40 keV absorption is dominating. For low energies the Rayleigh scattering is 3
orders of magnitude smaller than absorption but Compton scattering even 5 orders
of magnitude. For higher energies—at approximately 30 keV the scattering cross
section are comparable and at approximately 60 keV they have an even higher
probability than absorption.
Scattered X-rays are also used for material characterization. So it is possible to use
the scattering at small angles for the determination of short range order structures.
1.2.5 Diffraction
Immediately after the discovery of X-rays scientists starts to examine this new type
of radiation. Laue and co-workers found 1912 that X-rays are scattered on crystals
[70]. This could be interpreted as an interference of scattered X-rays on the
periodically structure of the crystal. Further investigations of this scattering
behavior brought father and son Bragg the cognition of their scattering law [71]:
n k ¼ 2d sin # ð1:9Þ
1.2 Interaction of X-rays with Matter Used for Material Characterization 13
with
k wavelength of the scattered radiation
d d-spacing of the scattering lattice
0 scatter angle
This relation can be understood from Fig. 1.9. The incident radiation is coherent
scattered on the electrons of the lattice atoms. The scattered waves of the different
atoms interfere and can amplify or annihilate each other in dependence of their
path length differences. If the path length is a multiple n of the wavelength k
the overlapping of the waves generates high intensity reflections at the scatter
angle 0.
This diffraction of X-rays finally explained the nature of X-rays but opened also
the possibility for investigations of the structure of material [72–76].
The Bragg-relation offers different possibilities for material investigation. It has
three parameters. If one of them is fixed and another is measured, the third one can
be calculated. The different possibilities for using that relation for material analysis
are summarized in Table 1.1.
With XRD it is possible to examine the symmetry of single crystals and explain
their structure and it is possible to study the structure of powders and poly-
crystalline materials to understand their mechanical behavior [73–75]. With ED-
XRD it is possible to look for individual diffraction peaks and control fast changes
of their intensity due to structural changes and with WDX the spectroscopic
investigation of spectra is possible allowing, for example, the analysis of char-
acteristic radiation of material and in this way the determination of its elemental
composition.
14 1 XRF-Basics
All X-ray spectrometers have a similar design with the same main components.
The intended analytical task defines the use of the different components and their
requested parameters. A general design with the main components is presented in
Fig. 1.10.
The following main components are required:
• Excitation source
In laboratory instruments the excitation is performed mostly with an X-ray tube.
But also the excitation with radioactive sources in transportable instruments or
with synchrotrons for highly sophisticated investigations is common. Direct
excitation with electrons is mostly used in electron microscopes.
• Primary optic
Primary optics are used to change the energy distribution of the radiation or
perform a beam shaping. Changes of the energy distribution are possible with
filters, secondary targets or monochromators, the shape of the excitation beam can
be influenced by collimators or X-ray optics. Often with X-ray optics both shape
and energy distribution of the beam are influenced.
• Sample positioning system
The sample needs to be positioned in the beam. For spectrometers that analyze
large sample areas this is often only a sample cup that will be positioned with a
sample tray. For position-sensitive analysis manual or motorized sample posi-
tioning in X and Y is possible. Then the sample position often is controlled by an
optical microscope.
• Secondary optic
The secondary optic can be very complex. It can be required as a beam shaper
which improves resolution or peak-to-background ratio but it can be also a dis-
persive optic that is used as (variable) monochromator.
• Detector
The detector measures the photons coming from the sample. For a WDX instru-
ment the detector needs only to count the photons because the dispersion is per-
formed in the secondary optic. In the case of energy dispersive detectors the
detector makes the dispersion as well as counting. Therefore there are different
detectors that have also different counting behaviors.
Additional to the mentioned components some further components are required for
a complete spectrometer—X-rays are ionizing radiation and can damage the
human tissue, therefore the complete spectrometer should be shielded against
accidental radiation of the environment, in particular of humans. Further the dif-
ferent component settings need to be operated and the data acquisition needs to be
controlled. For that purpose nowadays, all instruments are operated by processors.
Often the measurement has to be performed in a special medium: vacuum, He or
air which has to be supplied by pump, flushing systems etc. All these components
are important for the analytical performance but don’t are main components of the
1.3 General Design of X-ray Spectrometers 15
instrument. Last, but not least, the acquired data has to be evaluated. This requires
often complex calculations. Also for that purpose microprocessors are used.
The main components can be very different. Their parameters determine the
performance of the spectrometer and their design depends on the analytical task.
References
1. F.R. Elder, A.M. Gurewitsch, R.V. Langmuir, H.C. Pollock, Phys. Rev. 71, 829 (1947)
2. E.E. Koch, D.E. Eastman, Y. Farge, Handbook of Synchrotron Radiation, vol. 1A (North
Holland Publishing Company, Amsterdam, 1983), p. 1
3. P. Horowitz, J. Howell, Science 178, 608 (1972)
4. C.J. Sparks, S. Raman, H.L. Yakel, R.V. Gentry, M.O. Krause, Phys. Rev. Lett. 38, 208
(1977)
5. K.W. Jones, in Handbook of X-Ray Spectrometry, ed. by R.E. van Grieken, A.A. Markowitz
(Marcel Dekker, New York, 1992), p. 411
6. B. Gordon, K.W. Jones, in Microscopic and Spectroscopic Imaging of the Chemical State, ed.
by M.D. Morris (Marcel Dekker, New York, 1993)
7. K.W. Jones, B. Gordon, Anal. Chem. 61, 341 (1989)
8. S. Hayakawa, A. Iida, Y. Goshi, Rev. Sci. Instrum. 60, 24 (1989)
9. R.D. Vis, F. van Langefelde, Nucl. Instr. Meth. B54, 417 (1991)
10. M.L. Rivers, S.R. Sutton, Synchrotron Radiat. News 4, 223 (1991)
11. A.C. Thomson, K.L. Chapman, C.E. Sparks, W. Yun, D. Lai, D. Legini, P.J. Vicaro, M.L.
Rivers, D.H. Bilderback, D.J. Thiel, Nucl Instr. Meth A319, 320 (1992)
12. F. Lechtenberg, S. Garbe, J. Bauch, D.B. Dingwell, J. Freitag, M. Haller, T.H. Hansteen, P.
Ippach, A. Knöchel, M. Radtke, C. Romano, P.M. Sachs, H.U. Schmincke, H.J. Ullrich, J.
Trace Microprobe Tech. 14–3, 561 (1996)
13. P. Chevallier, P. Dhez, F. Legrand, A. Erko, Y. Agafonov, L.A. Panchanko, A. Yakshin, J.
Trace Microprobe Tech. 14–3, 517 (1996)
16 1 XRF-Basics
14. K. Janssens, L. Vincze, B. Vekemans, F. Adams, M. Haller, A. Knöchel, J. Anal. At. Spectr.
13, 339 (1998)
15. G. von Hevesy, Chemical Analysis by X-Rays and Its Application (McGraw Hill, New York,
1932)
16. R. Glocker, H. Schrieber, Ann. Phys. 85, 130 (1928)
17. M.A. Blokhin, Methods of X-Ray Spectroscopic Research (Pergamon press, New York, 1965)
18. R. Jenkins, J.L. de Vries, Practical X-ray Spectrometry (McMillan, London, 1967)
19. I. Adler, J. Axelrod, J.J. Branco, Adv. X-Ray Anal. 5, 2 (1962)
20. H.J. Rose, R.P. Christian, J.R. Lindsay, R.R. Larson, Geol. Surv. Proj. Pap. (US) 650, 8128
(1969)
21. M.C. Nichols, D.R. Boehme, R.W. Ryon, D. Wherry, B.J. Cross, G. Aden, Adv. X-Ray Anal.
30, 45 (1987)
22. R. Ryon, H.E. Martz, J.M. Hernandez, B.J. Cross, D. Wherry, Adv. X-Ray Anal. 31, 35
(1988)
23. C.D. Wherry, B.J. Cross, T.H. Briggs, Adv. X-Ray Anal. 31, 93 (1988)
24. D.B. Wittry, in Proceedings of the 50th Annual Meeting of the Electron Microscopy Society
of America (EMSA/MAS), Boston, (1992)
25. R.G. Tissot, D.R. Boehme, Paper presented on 42nd Denver Conference on Applications of
X-Ray Analysis, (Book of abstracts, 1993), p. 152
26. G.J. Havrilla, Paper presented on 42nd Denver Conference on Applications of X-Ray
Analysis, (Book of abstracts, 1993), p. 153
27. M.C. Nichols, R.W. Ryon, Adv. X-Ray Anal. 29, 423 (1986)
28. E. Spiller, A. Segmuller, Appl. Phys. Lett. 27, 101 (1974)
29. D. Mosher, S.J. Stephanakis, Appl. Phys. Lett. 29, 105 (1976)
30. A. Rindby, Nucl. Instr. Meth. A249, 536 (1986)
31. D.A. Carpenter, X-Ray Spectrom. 18, 253 (1989)
32. D.A. Carpenter, M.A. Taylor, C.E. Holcombe, Adv. X-Ray Anal. 32, 115 (1989)
33. D.A. Carpenter, M.A. Taylor, Adv. X-Ray Anal. 34, 217 (1981)
34. Y. Yamamoto, Y. Hosokawa, Jpn. J. Appl. Phys. 27, 2203 (1988)
35. A. Rindby, P. Engström, K.H. Janssens, J. Osan, Nucl. Instr. Meth. 124B, 591 (1997)
36. Y. Yinming, X. Ding, Nucl. Instr. Meth. 82B, 121 (1993)
37. M. Haschke, W. Scholz, U. Theis, in Proceedings of EDXRF Conference, Bologna, 1998,
p. 157
38. S. Bichlmeier, K.H. Janssens, J. Heckel, D. Gibson, P. Hoffmann, H.M. Ortner, X-Ray
Spectr. 30, 8 (2001)
39. M. Haschke, W. Scholz, U. Theis, J. Nicolosi, B. Scruggs, L. Herczeg, J. de Phys. IV 12, 83
(2002)
40. A.A. Bzhaumikhov, N. Langhoff, J. Schmalz, R. Wedell, V.I. Beloglazov, N.F. Lebedev, in
Proceedings of SPIE, vol. 3444 (1998), p. 430
41. B. Holynska, J. Ostachowicz, A. Ostrowski, D. Ptasinski, D. Wegrzynek, J. Trace
Microprobe Tech. 13, 163 (1995)
42. A. Longoni, C. Fiorini, P. Leutenegger, L. Sciuti, G. Fronterotta, L. Strüder, P. Lechner,
Nucl. Instr. Meth. 409, 407 (1998)
43. R.D. Evans, The Atomic Nucleus (McGraw-Hill, New York, 1955)
44. J.M. Jauch, F. Rohrlich, The Theory of Photons and Electrons, 2nd edn. (Springer, Berlin,
1976)
45. B.K. Agarwal, X-ray Spectroscopy: An Introduction, (Springer, Berlin, 1991). ISBN 0-387-
092684
46. E.P. Bertin, Priciples and Practice of X-ray Spectrometric Analysis (Plenum Press, New
York, 1975)
47. P. Hahn-Weinheimer, A. Hirner, K. Weber-Diefenbach, Röntgenfluoreszenz-analytische
Methoden, Grundlagen und praktische Anwendung in den Geo-. Material- und
Umweltwissenschaften (Springer, Berlin). ISBN-10: 3540670211
References 17
48. R. Jenkins, R.W. Gould, D. Gedcke, Quantitative X-ray Spectrometry (Marcel Dekker Inc.,
New York, 1981). ISBN 8: 0-8247-1266-8
49. G. Schmahl, D. Rudolph, in Proceedings of the International Symposium Göttingen
(Springer, 1984)
50. G. Schneider, High-Resolution X-ray Microscopy of Radiation Sensitive Material (Cullivier,
1999)
51. H. Legall, G. Blobel, H. Stiel, W. Sandner, C. Seim, P. Takman, D.H. Martz, M. Selin, U.
Vogt, H.M. Hertz, D. Esser, H. Sipma, J. Luttmann, M. Höfer, H.D. Hoffmann, S. Yulin, T.
Feigl, S. Rehbein, P. Guttmann, G. Schneider, U. Wiesemann, M. Wirtz, W. Diete, Opt.
Express 20–6, 18362 (2012)
52. P.A.C. Takman, H. Stollberg, G.A. Johansson, A. Holmberg, M. Lindblom, H.M. Hertz, J.
Microsc. 226–2, 175 (2007)
53. http://www.xradia.com
54. J.H. Scofield, Trans. Am. Nucl. Soc. 55, 200 (1987)
55. J.A. Bearden, Rev. Mod. Phys. 39, 78 (1967)
56. C.G. Barkla, C.A. Sadler, Phil. Mag. 16, 550 (1908)
57. H.J. Beattie, R.M. Brissey, Anal. Chem. 26, 980 (1954)
58. R.T. Beatty, Proc. R. Soc. Lond. A87, 511 (1912)
59. H.A. Kramers, Phil. Mag. 46, 836 (1923)
60. M. Siegbahn, The Spectroscopy of X-rays (Oxford University Press, London, 1925)
61. H.G.J. Moseley, Phil. Mag. 26, 1024 (1913)
62. H.G.J. Moseley, Phil. Mag. 27, 703 (1914)
63. P. Auger, J. Phys. et Le Radium 6, 205 (1925)
64. G. von Hevesy, Chemical Analysis by X-rays and Its Applications (McGraw-Hill, New York,
1932)
65. R. Glocker, H. Schreiber, Ann. Phys. 85, 1089 (1928)
66. J.J. Thomson, G. Thomson, The Conduction of Electricity Through Gases, 3rd edn.
(Cambridge University press, Cambridge, 1933)
67. D.V. Rao, T. Takeda, Y. Itai, T. Akatsuka, R. Cesareo, A. Brunetti, G.E. Gigante, J. Phys.
Chem. Data 31–3, 769 (2002)
68. A.H. Compton, Phys. Rev. 21–5, 483 (1923)
69. K. Molt, R. Schramm, X-Ray Spectr. 28, 59 (1999)
70. W. Friedrich, P. Knipping, M. von Laue, Interferenz-Erscheinungen bei Röntgenstrahlen, in
Bayerische Akademie der Wissenschaften, Sitzungsberichte (1912), p. 303
71. H. Bragg, W.L. Bragg, Proc. R. Soc. Lond. A 88, 428 (1913)
72. P. Debye, P. Scherrer, Nachr. Ges. Wiss. Göttingen, Mathematisch-Physikalische Klasse 1
(1916)
73. W. Kleber, Einführung in die Kristallographie (VEB Verlag Technik, Berlin, 1969)
74. W.H. Zachariasen, Theory of X-ray Diffraction in Crystals (Dover Publications, New York,
1967)
75. W. Clegg, Crystal Structure Determination (Oxford University Press, Oxford Chemistry
Primer, 1998)
76. W. Massa, Crystal Structure Determination (Springer, Berlin, 2004)
Chapter 2
Main Components of X-ray Spectrometers
In the case of electron excitation the electrons are emitted mostly by a heated
filament or by field emission and accelerated by high voltage to the sample or to a
target. If these electrons hit the target they will be decelerated and lose their
energy. This can happen in different ways.
One is the loss of energy by an inelastic scattering process which can happen in
only very few larger steps or also in a lot of small steps. The amount of the energy
loss then will be different for every scattering process. This energy can be emitted
as electromagnetic radiation. It is called bremsstrahlung because the generation is
a deceleration process of the electrons and is continuous because the steps of
energy loss vary.
Another possibility for the energy transfer of electrons is the absorption pro-
cess. In this case the incident electron removes an electron from an inner atom
shell and excites the atom by the generation of a vacancy. The excited atom goes
into the ground state by filling this vacancy with an outer electron. The energy
difference between both electron shells can be emitted as X-rays. Because this
energy differences are characteristic for the atom this radiation is called charac-
teristic radiation. In that case the energy of the incident electron needs to be higher
than the bonding energy of the electron in the atom shell.
Electrons will be stopped in the first few surface layers and generate there both
bremsstrahlung and characteristic radiation. That means the excited volume and
also the analyzed volume is very small and it is necessary to prepare the sample
that this volume is representative for the material. But most of the electrons are
scattered and absorbed i.e. the efficiency for the generation of X-ray is very low.
The emitted radiation will be emitted isotropic in all directions. Radiation that
is directed into the material can generate secondary fluorescence i.e. characteristic
radiation that is excited by the primary characteristic radiation or by the brems-
strahlung. Increasing electron energies enlarge the excited volume because the
electrons penetrate deeper into the sample and the bremsstrahlung has higher
energies. Therefore the secondary fluorescence also can come from deeper sample
regions.
These processes of electron excitation are illustrated in Fig. 2.1. It shows the
thin layer of electron scattering where both secondary and backscattered electrons
are generated that makes the image in electron microscopes and it shows the
‘ionization pear’ which is emitting X-rays both from primary and secondary
fluorescence.
The generation of X-Rays by an incident electron beam is used in X-ray tubes
as well as in electron microscopes. In X-ray tubes the target material typically is a
pure element that emits only the characteristic radiation of this element. In contrast
to that in an electron microscope the target is the sample itself which consists of
different elements. In that case the characteristic radiation of the sample can be
used for the determination of the sample composition.
The electrons loss their energy by a multiple step process by scattering and
absorption. A part of this energy is transferred to electromagnetic radiation—
which is due to the high energy of the electrons in the range of X-ray. The intensity
of the bremsstrahlung can be described in a first approach by the Kramer’s
equation [1] as:
E0 E
NBrems ðEÞ dE ¼ b Z dE ð2:1Þ
E
where:
2.1 Excitation Source 21
E0 E
NBrems ðEÞ ¼ b Z expðl q tÞ ð2:2Þ
E
where:
l mass attenuation coefficient of the window material
q density of the window material
t window thickness
These influences the emitted spectrum, i.e. low energies will be absorbed in the
window. For a high intensity of the emitted tube radiation also at low energies
these windows typically are thin and from materials with low mass attenuation
coefficients, i.e. from light elements like Be or Al.
Nevertheless, this absorption influences the spectrum outside of the tube. This is
displayed in Fig. 2.2, which shows that the low energetic part of the brems-
strahlung is completely absorbed in the window.
22 2 Main Components of X-ray Spectrometers
where
N Number of characteristic X-rays
wi Weight fraction of the element i
mi Atom mass of the element i
EC Energy of the absorption edge of the element i
Several corrections are used for that relation to determine the correct electron
excited intensities [5–7].
The efficiency of electron excitation is very low. The power of the electron
beam is the product of tube voltage (U) and current (i) as Ptube = i U.
On the other hand the power of the emitted radiation is according to Bertin [8]
Pradiation ¼ C Z I U2 ð2:4Þ
where Z is the atomic number of the target material and C is a factor that depends
on the tube geometry and is in the range of 10-6 kV-1. So far the tube efficiency
can be estimated as:
Ptube
g¼ 106 Z U ð2:5Þ
Pradiation
This is a very low efficiency in the range of 1 % only for common excitation
conditions. The largest amount of the power of an X-ray tube is transferred into
heat and has to be accordingly dissipated by cooling.
2.1 Excitation Source 23
Atoms can also be excited by X-rays themselve if they have sufficient energy. The
use of X-ray tubes is an excitation with photons. This is the most common used
excitation for X-ray fluorescence spectrometers. X-ray tubes are available in very
different modifications and can be adapted to the analytical task in regards to
energy of radiation, intensity, beam size and shape, geometry etc.
Other photon sources are synchrotrons and laser induced sources. Synchrotrons
deliver a high brilliant radiation that can be modified according the special ana-
lytical requirements. These were the first sources that were used for l-XRF
applications because they have a high intensity which can be used for X-ray optics
with small entrance apertures. But they are fixed on special sites and require a high
effort for maintenance and support. Therefore they are preferred for special
applications.
Laser induced sources can be interesting in case of time resolved analysis
because these sources are mostly pulsed i.e. they have a time structure that can be
used for the analysis of dynamic processes.
In X-ray tubes the radiation will be generated by accelerated electrons that hit the
target material of the tube. Therefore characteristic as well as bremsstrahlung will
be emitted. This is the same process as described for the excitation with electrons
in Sect. 2.1.1. The difference is that for an X-ray tube the sample is called target
and consists mostly only of one pure element. Common target elements are
Tungsten (W), Molybdenum (Mo), Rhodium (Rh), Copper (Cu), Chromium (Cr).
The parameters of an X-ray tube depend on the analytical task. The main target
for the selection of a tube is the efficient excitation of the sample or even only of
the element(s) of interest. Therefore the knowledge about the contributions of the
different spectrum ranges to the excitation of an element is important.
The absorption of X-rays is described by the Lambert-Beer-law (1.2). In this
formula the mass attenuation coefficient l of an element describes the probability
that a photon of energy E will be absorbed and can then generate fluorescence
radiation. The mass absorption coefficient strongly depends on energy as displayed
for Ti in Fig. 2.3.
The dependence of the linear mass absorption coefficient l from energy E is
l * E-3, i.e. l drops down with energy rapidly. That means the probability for
absorption will be reduced with increased energy.
If the X-ray energy is large enough to excite a new shell—for Ti for example
the K-shell at 4.97 keV—the absorption will be increased. But then the decreasing
is continued i.e. photons with energies slightly higher than the absorption edge will
24 2 Main Components of X-ray Spectrometers
Fig. 2.4 Contribution of different energy ranges of an Rh-tube (50 kV) to the excitation of pure
Ti
strongly absorbed and contribute the excitation of the element, here Ti, but for
higher energies the absorption is significantly reduced.
Therefore, the intensity of the excitation spectrum on the absorption edge needs
to be high for an efficient excitation of X-ray fluorescence. This is also demon-
strated in Fig. 2.4 which shows the contribution of different energy ranges of an
Rh-tube spectrum working with 50 kV to the excitation of pure Ti.
The figure shows both the spectrum of an Rh-tube influenced by the trans-
mission function of a poly-capillary optic and the contributions of several energy
ranges of the this spectrum to the excitation of Ti. This demonstrates that the most
important contribution results from the bremsstrahlung in the energy range shortly
above the absorption edge. In contrast the high intensive fluorescence radiation of
the tube has only a marginal contribution.
The fluorescence radiation of a tube has high excitation efficiency only for
elements with absorption edges with up to 5 keV lower energies than the
2.1 Excitation Source 25
The X-ray tube can be prepared in different geometries which differ for parameters
like spot size on the target, grounding of the tube and cooling possibilities. The
main geometries are:
• Side window tubes
In side window tubes the electrons are emitted typically from filaments like coils
or tips. Then they are accelerated along the tube axis to the target. The emitted
radiation exits the tube by a window which is on the cylindrical face of the tube.
In that case it is possible to focus the electrons on a defined target area—this
area can be line, a small spot or also a circle. In that way it is possible to adapt
the tube-spot to the best conditions of sample illumination. The focusing mostly
is passive only by the geometrical arrangement of the different construction
parts of the tube like filament, Wehnelt cylinder etc.
In side window tubes the high voltage is on the target and the filament is
grounded to the tube housing. This makes the generation of the filament heating
26 2 Main Components of X-ray Spectrometers
easy but the target needs to be electrically isolated from the tube housing. That
means it is necessary that the tube window have a distinct distance from the
target which limits the solid angle of tube radiation that can exit the tube.
Furthermore electrons are scattered on the target. They can reach the tube
window and heat it up due to the absorbed electron energy. In particular for
high power tubes this heating can be significant. To dissipate the heat and to
have a sufficient vacuum sealing the tube window has to be thick enough. But
this increases the absorption of tube radiation especially for low energies. A
common window material is Be which has a low mass absorption coefficient
and can be prepared as vacuum tight thin films.
The scheme of a side window tube is shown in Fig. 2.5.
target. But this is both target and tube window. This allows a focusing of the
electrons to the target and the generation of small spots. In that case the target is
the tube housing and must be grounded. Accordingly the filament must be on a
negative high voltage.
Because the electrons hit the window directly it will be heated up by the power
of the accelerated electrons. That means if the tube window is thick it can well
dissipate the heat but the generated fluorescence radiation will be absorbed in
the window itself or on the other side if the window is thin the excited radiation
can penetrate the window but the heat cannot be dissipated which limits the tube
power.
The main advantage of the transmission window is the large solid angle of
radiation that can be used.
The scheme of a transmission window tube is shown in Fig. 2.7.
28 2 Main Components of X-ray Spectrometers
The tube housing has to fulfill different functions—the isolation of the internal
tube components for high voltage, the sealing of the tube vacuum and also the
shielding of tube radiation. There are different housing materials in use:
• Glass tubes where all components are included in a glass bulb which itself is
typically enclosed by a metallic envelope. The volume between both of them is
filled with oil for electrical isolation and heat transport between the target and
tube housing. This technology is old but still very successful and often used. For
high power tubes a direct water cooling of the target is usual.
• Metal-ceramic tubes have a housing of metal and ceramic. In that case the tube
housing offers both—electrical isolation and also shielding of radiation but
requires special technologies for the connection of metal and ceramic
components.
Also the tube window belongs to the tube housing. Windows are mostly from
Be to limit the absorption of radiation in the tube window. But in case of using
only high energetic radiation the window can be from glass—in case of glass
tubes—or other light metals like Al.
Typically the tube target is a pure element. Common elements are W, Mo, Rh, Cu,
and Cr. But for special applications also other materials can be interesting.
Most of the targets are fixed and embedded in a Cu-rod because Cu has high
heat conductivity and can dissipate the heat from the target. For a fixed target
position maximum tube powers are in the range of 4–5 kW. For higher power the
target position has to be changed to avoid its melting. One possibility is a rotating
anode. In this case the target point is distributed on a ring of a fast rotating anode.
The tube power then can be enhanced into the range to 20 kW. Even higher tube
powers are possible with jetstream targets i.e. with targets that are prepared by fast
moving liquid streams for example from Ga, In or low melting Sn-alloys [9].
The selection of the target material is important. It depends on the application
of the tube. In particular for spectroscopic applications the following criteria
should be considered:
• Resistance against the generated heat in the target
As discussed in Sect. 2.1.1 more than 98 % of the tube power is transferred into
heat. For that reason the target will be heated up to very high temperatures.
Therefore the target material should have a high melting point or a high heat
conductivity to ensure the heat can dissipated quickly. The best option would be
that the product of melting point and heat conductivity should be as large as
possible. Table 2.1 summarizes for few typical target materials this product.
2.1 Excitation Source 29
This product is very important, in particular, for fine focus tubes because for
them the focus needs to be small and therefore a high concentration of absorbed
power heats up the tube target locally.
• Conclusion
This third criterion is the most important for the selection of the target material
for X-ray spectroscopy because it influences the measured spectrum. Never-
theless, Rh has also acceptable ratings for the other criteria i.e. good combi-
nation of heat conductivity and melting point and an atomic number that
generates acceptable bremsstrahlung intensity.
Therefore Rh will be very often used as target material in X-ray spectroscopy.
But also other target materials are in use in special cases.
One example is a W-target if the measurements are performed only in air. In that
case low energetic fluorescence radiation will be absorbed in air and therefore it
30 2 Main Components of X-ray Spectrometers
has not to be excited i.e. the tube window can be relatively thick (for example
from glass). In that case the W-L-radiation is absorbed in the tube window and
cannot influence the fluorescence spectrum.
Another reason is the requirement for the very efficient excitation of a special
element group, for example, for trace analysis. In that case target material can be
used that have characteristic radiation slightly above the absorption edges of the
elements of interest.
The excitation conditions influence the measured fluorescence intensity. The target
of excitation is to get high fluorescence intensity together with a low spectral
background. The high fluorescence intensity assures a low statistical error and the
low spectral background is important for a high peak-to-background ratio, in
particular, in case of trace analysis.
The excitation efficiency depends mainly on the intensity of tube radiation on
the absorption edge of the element that has to be analyzed. This was demonstrated
already in Figs. 2.3 and 2.4. Therefore in most cases the bremsstrahlung is used for
the excitation. The bremsstrahlung intensity can be influenced by the tube target as
discussed in Sect. 2.1.3.4, by the tube voltage and current and by the use of filters
in the excitation beam.
• Influence of tube voltage
The dependence of bremsstrahlung intensity of tube voltage is described by
Kramer’s equation (2.1). But the influence to excitation efficiency in particular
for l-XRF can be better demonstrated with the tube spectra for different tube
voltages as shown in Fig. 2.8 because the concentrating optic changes the
energetic distribution of the tube spectrum. These spectra are measured by
scattering tube radiation on a PMMA block. For the lowest tube voltage of
20 kV the Rh-K-radiation cannot yet excited and the bremsstrahlung intensity is
fairly low. For increased tube voltages the energy is sufficient to excite the Rh-
K-fluorescence and also the bremsstrahlung intensity is growing but only for the
energy range around the bremsstrahlung maximum and higher. For the energy
2.1 Excitation Source 31
range between approx. 3 keV and the bremsstrahlung maximum increased tube
voltage prolongs the range of increasing bremsstrahlung intensity with energy
but the intensity itself is not increased in this range. But this energy range is
important for a lot of elements i.e. for the K-radiation of elements from
Potassium to Arsenic and for the L-radiation of elements from Silver to Lead.
From this relation it can be concluded that for the excitation of a flexible and
wide element range the highest tube voltage is advantageous.
There is another increasing of bremsstrahlung intensity with growing tube
voltage for low energies (below approx. 3 keV) which is caused by a multiple
scattering of electrons in the tube target. This increased bremsstrahlung intensity
has to be considered for the measurement of light elements because the tube
radiation not only influences the fluorescence intensity but also the spectral
background due to their scattering on the sample. The scattering has a high
efficiency especially in case of light sample matrices i.e. the higher brems-
strahlung intensity at low energies increases the spectral background for that
energy range and decreases the peak-to-background ratio. Therefore for the
analysis of light elements in light matrices the tube voltage should be reduced—
this has a negligible influence to the fluorescence intensity of these elements but
decreases the spectral background and improves the peak-to-background ratio.
10 keV and also in comparison to the also displayed 50 kV tube spectrum (blue)
which has the same tube power as the red spectrum for 30 kV and 666 lA. That
means, with 30 kV and increased tube current higher fluorescence intensity can
be reached for energies between 3 keV and the bremsstrahlung maximum.
So far it can be concluded that in case of analysis of elements with excitation
voltages up to 15 keV it can be advantageous to reduce the tube voltage and
enhance the tube current. But it has to be mentioned that the maximum tube
power mostly is limited and depends on the tube voltage.
• Influence of filters
Filters in the excitation path influence the excitation spectrum in particular they
reduce the low energetic part of the excitation spectrum. So far they have to be
considered already as primary optics. Their influence to the tube spectrum is
demonstrated in Fig. 2.10 for filters with different absorption (different thickness
and/or different materials).
Filters can be used
2.1 Excitation Source 33
The main requirement for a l-XRF source is radiation with high brilliance. Small
spots on the sample can be realized only with collimators or optics and for them small
sized sources are essential. This is possible only with side window tubes. End
window tubes allow the use of a larger part of emitted radiation but not the required
small spot sizes in the range of 100 lm and even less. Therefore they have not the
required high brilliance. The same is valid for high power side window tubes. Their
spot sizes in the mm-range allow a high total flux but the brilliance is less than for low
power tubes. The relation is approx. 4,000 W for spot sizes of 1 9 3 mm2 or 30 W
for spot sizes of 50 9 50 lm2 which relates to 1.3 or 12 kW/mm2, respectively.
That means, the best brilliance is available for low power tubes with smallest
target spots i.e. fine focus tubes.
As target material mostly Rh is used due to the low line overlaps. For special
applications also W is an option. The higher atomic number of W generates a
higher bremsstrahlung intensity which has advantages for the excitation of ele-
ments with higher atomic number or in case of measurements in air where the W-
lines are absorbed.
The primary optics is positioned between the X-ray source and the sample. They
are used to change the beam shape and/or the energy distribution of the source
radiation. Similar optics also can be used also as secondary optics. Also in that
case beam shaping and influencing the energy distribution is the main task. In the
case of wavelength-dispersive spectrometers the dispersive element (crystal,
synthetic multi-layer) can be considered as a secondary optic because it changes
the energetic composition and often it also influences the beam shape.
Possible primary optics are:
34 2 Main Components of X-ray Spectrometers
• Filters for the absorption of special parts of the spectrum (even the tube window
can be considered as a primary optic)
• Collimators and apertures for the definition of the beam shape
• Pure elements that are used for generation of secondary fluorescence
• Monochromators for selecting a monochromatic beam from the tube spectrum
• X-ray optics both for beam shaping and energy changing and combinations of
the different mentioned types
Filters are thin films of different materials that absorb parts of the tube radia-
tion. This can be interesting for XRD for the suppression of the ß-radiation of the
tube, the already mentioned suppression of characteristic W-radiation in glass
window tubes or the reduction of the tube spectrum in certain energy ranges for an
improvement of the peak-to-background ratio.
These effects can be improved by monochromators which are a special version
of X-ray optics.
Collimators and apertures are mechanical limitations for the X-ray beam. This
ranges from different shaped holes in high absorbing materials to Soller-colli-
mators, see Fig. 2.11 which are prepared for example from a set of parallel sheets
or by poly-capillary structures to prepare a parallel beam. But typically they does
not influence the tube spectrum.
Another possible component between source and sample can be secondary
targets which consist generally of a pure element. This is excited by the primary
source and emits its mono-energetic fluorescence radiation which then is used for
the excitation of the sample. This is a method for the efficient excitation of selected
elements with absorption energies at slightly lower energies than the secondary
fluorescence and for the reduction of the scattered background. This excitation
improves the peak-to-background ratio for the analyte elements.
All X-ray optics use the already discussed interaction of electromagnetic radiation
with matter i.e. absorption, scattering, refraction or/and diffraction. In particular
the following optic types can be distinguished:
2.2 Primary Optics 35
With the knowledge of the Bragg-law (1.9) the interference of scattered X-rays on
periodical structures was found and could be explained. But this relation could be
also used for X-ray spectroscopy (Table 1.1). Crystal optics not only changes the
energetic distribution of the beam but also its direction and in the case of shaped
optics even the beam shape. So far crystals were the first type of X-ray optics.
They can be used in the position of primary optics as a monochromator. But the
most frequent use of crystals is in WDX-spectrometers as dispersive element i.e. in
the position of secondary optics.
The energy range that can be covered by a crystal with given d-spacing is defined
by the Bragg-law. The end for long wavelengths or of low energies is reached for
sin 0 = 1 i.e. for Bragg-angles close to 90. That means the longest wavelength is
defined by the d-spacing of the crystal. Typical d-spacing’s are in the range of a few
tenth of a nanometer and therefore the spectrometry of long wavelength with crystal
optics is limited for these wavelengths too. The spectroscopy of longer wavelength
i.e. of elements lighter than Si requires special crystal structures for the diffraction
for example organic crystalline structures or compound crystals.
The short wavelength or high energy side is limited by the angular resolution of
the spectrometer. The angular resolution of a crystal can be calculated from Bragg-
relation as
d0 n 1
¼ ð2:6Þ
dk 2k cos 0
• the angular resolution can be improved for higher scattering order n but this
reduce the reflectivity,
• the analysis of short wavelength with large d-spacing is possible with very low
Bragg-angles but this reduce the angular resolution
So far it can be concluded that spectrometry of different wavelength require the
use of various crystals with different d-spacing’s. This can be either realized by
altering the reflection planes in the same crystal or by the use of different crystals.
Changing the reflection plane requires a tilt of the crystal in regards to the beam
and causes non-homogeneous an intensity loss because these planes are not par-
allel to the crystal surface. Therefore typically different crystals are used for the
Bragg-optics.
But there are additional influences to the energy resolution of a crystal [11].
Some of these factors are independent of distances, for example the intrinsic
reflection broadening D0intrinsic (Darwin width) and a contribution if the crystal is
flat and not shaped to the Rowland circle. The influence of the intrinsic reflection
broadening to energy resolution is given by
Other contributions are influenced by the distances between source and crystal
F1 and between crystal and detector F2. These can be described by
Fig. 2.12 Reflection on a flat crystal of a parallel beam (left) and a divergent beam (right)
Table 2.3 gives a selection of often used crystal structures with their application
for the analysis of different elements.
Diffraction optics are typically used as monochromators—which is possible as
primary optic but mostly as a secondary optic as the dispersive element in the
spectrometer.
The brightness of crystal optics depends not only from the reflectivity but even
more on the amount of reflected radiation i.e. on the captured solid angle. The
incident beam can be parallel or divergent (Fig. 2.12) [12]. For a parallel beam, the
reflection angle is the same for all photons, and the detector see only mono-
energetic radiation. The parallel beam can be prepared by a Soller-collimator. But
that requires a large analyzed area which is available for conventional XRF but not
for l-XRF.
For l-XRF the sample fluorescence is divergent because it is emitted from a
small sample area only. The divergent beam can be paralyzed with optics but that
would reduce the intensity or even change the energetic distribution due to the
transmission function of the optic. The wavelength dispersive Parallel-Beam-
Spectrometers (PBS) in electron microscopes use poly-cap half lenses to prepare a
parallel beam for that purpose [13–16].
If a divergent beam is reflected on a flat crystal photons with different reflecting
points have a different scatter angle i.e. different energies are reflected in every
reflection point. Therefore the detector needs an entrance slit for the selection of a
special scattering angle. But this reduces the collected intensity as can be seen in
Fig. 2.12.
To combine a high brightness with good resolution it is necessary to use curved
crystals. But then a special geometry and shaped crystals are required. The
entrance slit to the spectrometer or sample, the crystal and the detector needs to be
on a circle—the Rowland circle.
Two often used geometries are shown in Fig. 2.13. If the crystal is bend to the
double diameter of the Rowland circle (Johann-geometry) all reflected beams have
the same scatter angle i.e. the same wavelength but they are not yet focused to the
38 2 Main Components of X-ray Spectrometers
Fig. 2.13 Johann-geometry (left) and Johannson-geometry (right) with curved crystals
same point on the Rowland circle. If the shape of the crystal surface also is adapted
to the Rowland-circle (Johannson-geometry) not only all photons have the same
scatter angle but will be focused to the same position on the Rowland-circle.
In both cases due to the curved crystal a larger angle of sample radiation can be
captured which increases the collected intensity in the detector. But bending and
machining of crystals can be a difficult and expensive procedure.
If the crystals can be prepared as thin layers it is possible to bend the crystal. In
this case the preparation of even doubly curved crystals (DCC) is possible. The
arrangement of these DCC allows the capture of very large solid angles of
divergent radiation and it’s directed Bragg-reflection i.e. a combination of Bragg-
reflection and optical reflection is possible. The bending of the crystal has to be
performed carefully because the crystal structure can be damaged due to this
procedure. This limits also their bending diameter.
DCC can be used as primary optic for the very efficient excitation with mono-
chromatic radiation of special elements with a low spectral background for that
element. The large captured angle that would be enabled by a bended crystal offers
high excitation intensity with mono-energetic radiation that not overlaps with the
fluorescence radiation of the sample.
But it can be interesting also in secondary optics as dispersive element. Then a
very large solid angle of fluorescence radiation can captured. Examples are given
by XOS [17]. But these arrangements are usable only for one wavelength i.e. for
one element. In the case of changing the wavelength of interest the scatter angle
has to be changed which requires also a change of the bending radius of the crystal.
This limits the application of these bended crystals for flexible spectrometric
applications.
Another possibility to enlarge the captured angle for a divergent beam with
reflecting the same wavelength is a change of the d-spacing in the crystal. This is
possible, for example, by changing the composition of compound crystals like in
SiGe-crystals [18].
2.2 Primary Optics 39
Crystal optics are used as monochromators in the position of the primary optic
but mainly as wavelength-dispersive spectrometers in the position of a secondary
optic.
For long wavelengths or low energies large d-spacing are required. With natural
structures this is possible only with molecule crystals. But these crystals are often
soft and are not stable against environmental conditions. Therefore other structures
were required that could be used for the spectroscopy of low energies. For that
purpose Synthetic Multi Layers (SML) was developed [19–21]. This are layers
systems which are prepared by repeated layers of different elements with different
density like W/Si, Mo/Si or Ni/C etc. These layer systems need to be prepared in
sequences of 20–100 layers with extreme flatness and highest reproducibility of
thickness. There are different manufacturers for SML available [22–25]. For these
layer systems the scattered radiation has a path difference which produces inter-
ferences like for the scattering on crystals.
The challenge for SML is their manufacturing because the distance between the
structures i.e. the thickness of the separate layers must be in the range of the
wavelength of the X-rays. This would be for P-K radiation approx. 0.6 nm, for Na-
K-radiation approx. 1.2 nm and for C-K-radiation approx. 4.5 nm—all these are
dimensions which are in the range of only few atomic diameters.
But on the other hand SML structures have a few big advantages. They are
prepared synthetically and therefore it is possible to tailor them for an intended
application. This is valid for their d-spacing by using different thicknesses of the
layers, for their scattering efficiency by selection of different layer materials and
also for their shape by using curved substrates to perform a beam shaping.
40 2 Main Components of X-ray Spectrometers
volume is increased. But on the other hand for a thicker crystal the mosaic spread
is enlarged which influences the energy resolution. This allows the adjustment of
these parameters according the analytical requirements.
The special type of manufacturing of this graphite generates the high orienta-
tion of the crystallites in the same direction i.e. produces the mosaic structure [27,
28]. By a special treatment it is possible to reduce the mosaic spread even down to
0.04. This material is called Highly Annealed Pyrolytic Graphite or HAPG [29].
The reflectivity of HAPG with different thicknesses and in dependence of
energy is displayed in Fig. 2.17. This indicates the dependence of reflectivity of
both parameters. But it demonstrates also that the integral reflectivity per rad of
HAPG is better by at least one order of magnitude than that of Si-crystals.
The high reflectivity and the d-spacing of HOPG allow the generation of
polarized radiation [30] which is generated by a 90 scattering. In that case, the
Bragg angle would be 45. For Rh-L-radiation the Bragg angle of HOPG is:
sin 0 = Rh - L/2 dHOPG = 0.460 nm/0.6708 nm = 0.6859 i.e.
0 ¼ 43:3 ð2:9Þ
42 2 Main Components of X-ray Spectrometers
Fig. 2.18 Cu-Ka-spectra measured with different HOPG’s, HAPG and LiF (after [29])
Table 2.4 Energy resolution for different optics (working distance \300 mm)
Crystal Energy resolution E/DE
HOPG 50 lm 460
HOPG 150 lm 240
HAPG 15 lm (first order) 650
HAPG 15 lm (second order) 1,300
LiF 200
This is very close to 45 and produces a high degree of polarized radiation. The
high integral reflectivity of HOPG and the possibility to focus a large incident
beam diameter by an appropriate curved HOPG mirror overcomes the problem of
low scattering intensity.
Pyrolytic Graphite can be prepared as a thick block but also as very thin
layers—in thickness ranges down to 10 lm. Therefore these layers can be applied
on curved substrates without damaging the mosaic structure even in case of double
curved substrates and with small bending radii down into the range of few mm.
This allows the combination of Bragg-scattering with optical reflection on curved
mirrors. The result is the capture of large solid angles of radiation combined with a
beam shaping i.e. focusing, parallelizing etc. [30–35]. Optics from pyrolytic
graphite are used both for primary and secondary optics.
If the mosaic crystal is used as dispersive element in a spectrometer its energy
resolution should be as good as possible. Here HAPG offers the better perfor-
mance. This is demonstrated in Fig. 2.18 which shows Cu-Ka-spectra measured
with HOPG and HAPG as well as with a LiF-crystal. The measurements with the
HOPG/HAPG were performed with a von Hamos-spectrometer with a total
measurement time of 1 s for the HOPG and of 5 s for the HAPG [29]. The
measurement with the LiF-crystal was performed with a scan in a WDX
2.2 Primary Optics 43
spectrometer. The intensities are normalized because they cannot be compared due
to the different measurement conditions.
These measured spectra show
• a considerably improvement of energy resolution with the mosaic crystals is
possible. The Ka-peaks for the LiF are not separated but for HOPG as well as
HAPG they are clearly separated.
• the spectra for the thickest HOPG (150 lm) and the LiF crystal are comparable
but the considerably better reflectivity of the HOPG improves its performance
• the energy resolution of the HOPG depends on its thickness, for thinner layers it
can be improved
• a further improvement of energy resolution combined with a slightly loss of
reflectivity is possible with HAPG crystals
The energy resolutions E/DE of these measurements are listed in Table 2.4.
They indicate that there is a significant improvement against the LiF-crystal and
also of the HAPG in comparison to HOPG.
Compound Refractive Lenses (CRL) are imaging optics that use the refraction of
X-rays on the interface between different materials [35–44]. The refraction index
for X-rays is n = 1 - d, where d is in between 10-5 and 10-7 i.e. n is very close
to unit and for all materials negative. Therefore focusing optics has a concave
shape. Because the refraction for X-rays is very small refraction cannot strongly
influence the direction of X-rays. Therefore refraction optics uses multiple
refractions to have a sufficient beam deviation. Figure 2.19 shows a single
refraction lens and an optic with multiple lenses.
The focal distance f of a single refractive lens can be approached with help of
the radius R of the concave curvature by
f R=2d ð2:10Þ
Due to the small value of d the focal distance would be very large even if the
curvature R of the refracting surface is very small. But that means the dimensions
of these optics will be very small—typical radii are in the range of less than
0.1 mm. Therefore only narrow beams of radiation can be captured and focused by
refraction optics.
Another possibility to reduce the focal length is the use of multiple lenses. For
N single lenses the focal length is expected to
44 2 Main Components of X-ray Spectrometers
f N R=2d N ð2:11Þ
If the radius R of the lens is 0.1 mm the focal length would be for a single lens
in the range of 50 m, for 100 repeated lenses still 0.5 m. The refractive index for
X-rays depends slightly on the density of the material. Therefore metals like
Beryllium, Aluminum, Silicon, Nickel or X-ray resistant polymers are preferred.
The high number of single lenses which are required for a sufficient focusing
effect causes a certain length of the optics—for 100 lenses it would be in the range
of 15 mm. Then absorption of radiation by penetrating the optic has to take into
account. The absorption is different for a beam path in the middle or on the edges
of the lens. In outer paths the absorption of low energies is stronger i.e. the
radiation is hardened which is a change of the energy distribution of the beam.
Also for limiting the absorption the refraction optics should be from a material
with a small mass absorption coefficient. Nevertheless refraction optics can be
used therefore, but only for higher energies. For an average thickness of approx.
5 mm the transmission for Be and Al is shown in Fig. 2.20. This shows that for
energies less than 10 keV for CML from Be and 30 keV for CML from Al no
2.2 Primary Optics 45
significant radiation will penetrate the optic and the average radiation energy will
be increased. But it is still a continuous spectrum in the focus.
For an acceptable refraction the bending of the refracting interface needs to be
very high i.e. their radius R needs to be small as shown in (2.10) and (2.10). But
then the entrance aperture of refractive optics is very small and for a good spot
intensity a high parallel flux on this small area is required. This can be reached
only with synchrotron sources.
The small entrance diameter can be slightly enlarged by folding the parabolic
shape of the lens as shown in Fig. 2.21. They are called kinoform-lenses.
Refractive lenses commonly focus the beam only in one direction. Two dimen-
sional focusing requires two successive series of lenses that are rotated by 90.
Fresnel zone plates are circular lattices with decreasing lattice distance for larger
diameters. They are prepared from materials that have different refraction or phase
shifts for the transmission beams in the different zones [45–55]. These beams
interfere after the transmission and generate intensity maxima and minima. For X-
rays the refraction for all materials is very small and similar. Therefore Fresnel-
structures in that case are prepared by different thicknesses of the same material or
by coating with high absorbing materials. For a good efficiency of the zone plate a
high difference in absorption or phase shift is essential. This is possible by a high
aspect ratio.
For the interferences of the waves of two neighboring zones their width has to
generate path differences between the zones which are multiples of k/2 (see
Fig. 2.22). Calculations show that for a first approach the area F of every zone
needs to be independent of their radius R.
The focal length f of a Fresnel-optic for a parallel incident beam can be
determined as:
f ¼ F=n k ð2:12Þ
That means the focal distance depends on the wavelength k of the incident
radiation. That means the focal distance depends on energy—for higher energies
46 2 Main Components of X-ray Spectrometers
the focal length is enlarged (Fig. 2.22). For Fresnel-optics mostly diffractions in
first order are used because the efficiency of the diffraction is inversely propor-
tional to the square of the diffraction order i.e. the intensity drops down consid-
erably with diffraction order.
Fresnel lenses are imaging optics and their spatial resolution is determined by
the size of the outer ring. Therefore very small structures are required—down into
the nm-range. The preparations of these small structure sizes with a high aspect
ratio requires special technologies and they are prepared mostly from Si by etching
or by Au-coating because these technologies are known from micro-electronics.
Due to the small entrance diameter Fresnel optics require a high flux which is
only available from synchrotron or special laser induced X-ray sources. For lab-
oratory instruments Fresnel-optics are used rarely.
2.2 Primary Optics 47
from the substrate material is focused due to the Fresnel-structure. These structures
are called Bragg-Fresnel optics.
For reflecting Fresnel structures a tailoring for special energy ranges and also
for special measurement geometries is possible. This is shown in the following
example for a simultaneous spectroscopy in the low energetic range [60]. In that
energy range there are a lot of lines which can overlap—K-lines of light elements,
L-lines of medium elements and M-lines of heavy elements. Therefore in that
range the energy resolution needs be very good. This is possible by wavelength-
dispersive spectrometers but they are measuring usually in a sequential mode
which requires a long measurement time. With reflecting Fresnel optics as shown
in Fig. 2.24 even simultaneous spectroscopy for a selected energy range is pos-
sible. The Fresnel structures are changing in different areas of the optic and with
the structures also the reflected energies are changed.
A spectrum measured in an appropriate arrangement is displayed in Fig. 2.25.
The measurement was performed with electron excitation in an electron micro-
scope on a stainless steel sample. The measurement time was 100 s.
The spectrum shows the spectra of C, N and Ti-L. The used optic had different
sections for different energy ranges therefore the Nitrogen line is doubled in the
spectrum because this energy could be measured in two overlapping ranges of the
optic. The energy resolution per pixel corresponds to 1.7 eV.
The acceptance angle for this optics is caused by the small angle of total
reflection relatively small which limits the brightness of the spectrometer. It also
requires a very accurate alignment of the arrangement but it can be an interesting
solution for a simultaneous spectroscopy with high energy resolution in particular
in the range of high line density.
Due to the small refraction index for X-rays and correspondingly only small
differences of the refraction index for different materials X-rays can be only total
2.2 Primary Optics 49
reflected. This effect is also used for X-ray optics. The critical angle for total
reflection Hcrit can be derived from the Snellius’s law as
with:
n1,2 Refraction indices and n1 [ n2
Because both n1; n2 1 their ratio is close to 1 (but \1) and the critical angle
of total reflection for X-rays is close to 90. That means total reflection is possible
only for a grazing incidence of X-rays (Fig. 2.26).
The critical angle of total reflection (in deg) depends on the density q (in
g cm-3) of the material but also on the energy E (in keV) of the reflected radiation.
It is given in deg by
pffiffiffi
0crit ¼ 2:04 102 1=E q ð2:14Þ
That means the critical angle will be reduced with higher energies but can be
increased with a higher density of the reflecting material.
The critical angle for higher energies for example Cu-Ka radiation on an Au-
coated mirror would be approx. 1 but for lower energies for example O2-Ka-
radiation approx. 17.
50 2 Main Components of X-ray Spectrometers
The reflected intensity of radiation depends on the scattering angle but also on
the absorption of reflected radiation. Absorption for total reflection typical is very
small but in case of rough surface and also for light materials this effect is possible
and influences the reflection efficiency. This is illustrated in Fig. 2.27 where b is
the absorption coefficient and d is the dispersion of the complex refraction index
(see 1.5). This shows that for larger absorption (b) the reflectivity is reduced as
also for reduced dispersion (d).
Total reflection optics are used mainly in two versions—as total a reflection
mirror or as capillary optics.
Due to the very small incident angle and the reflection of a broad energy band total
reflection mirrors are used as X-ray optics for beam shaping. The advantage of
total reflection mirrors is their reflectivity [61]. Therefore they are used for high
intensive radiation like in synchrotrons and even for FEL’s. Their high radiation
intensity could damage X-ray optics in case of considerably absorption.
Total reflection mirrors are also used for low energetic radiation for example for
energies\1 keV and, in particular, for the EUV-range. In that case, also larger inci-
dent angles are accepted which offers the possibility for higher curved mirrors [62].
Total reflecting mirrors are produced from substrates with extremely low
roughness. If they are used for beam shaping the bending radii must be very large
to ensure that the incident angle is smaller than the critical angle of total reflection.
To enlarge this angle they can be coated with heavy materials with a high density
like Au or even Re.
Total reflection can be also performed in capillaries. First ideas to use capillaries for
propagation of radiation were mentioned already in the 1930s by Jentzsch [63]. But
then it took a long time for the use of capillaries for that purpose [64–66]. Now-
adays capillary optics are used very often for the propagation of radiation but also
for beam shaping, often for the concentration of radiation to small areas [66–70].
Capillaries are mostly prepared from glass because this material can be easily
treated for capillaries with a high ratio between length and diameter. Metal cap-
illaries would be better in regards to the critical angle of total reflection because
their density is larger than for glass. But manufacturing of long and very thin
capillaries with a very flat internal surface from metals is difficult. The advantage
of a slightly larger acceptance angle for a metal capillary is overcompensated by
the intensity loss by total reflection on a rough surface (see Fig. 2.27).
• Propagation of radiation by cylindrical capillaries
2.2 Primary Optics 51
where 0crit is the critical angle of total reflection and dentrance is the entrance
distance of the capillary.
In that case an intensity gain is possible caused by the larger captured angle in the
capillary Xeff than that without capillary for the same spot size (X0). This is
demonstrated also in the upper part of Fig. 2.28. The intensity gain g is given by
G 1=E2 ð2:17Þ
The middle part of Fig. 2.28 shows the situation for different capillary diam-
eters. A smaller capillary diameter with the same entrance distance would reduce
52 2 Main Components of X-ray Spectrometers
the captured angle of radiation and therefore also the flux. Then the gain is
changed because the relation between Xeff and XCap is altered. But in that case also
higher energies can be captured at the beginning of the capillary.
The lower part of Fig. 2.28 shows that the same solid angle of radiation can be
captured also for smaller capillary diameters if the entrance distance is decreased.
That would produce a smaller spot size with the same amount of captured radi-
ation. But the intensity in the spot will be decreased due to the higher number of
total reflections in smaller capillaries. Total reflections are not completely free of
intensity loss. Every reflection is combined with a small reduction of intensity. For
the propagation over 100 mm and an assumed scattering angle of 0.3 which
would correlate for example with Au-L-radiation the number of reflections for a
capillary diameter of 300 lm would be approx. 2 but for a capillary diameter of
10 lm it would be approx. 500. For an assumed intensity loss of only 0.2 % for
every reflection that would give an intensity reduction of approx. 0.5 and 64 %,
respectively.
The entrance distance of a capillary optic is mostly limited by the tube
geometry. Because capillary optics requires point sources mostly side window
tubes are used in combination with capillary optics. But these tubes have a certain
diameter which limits the reduction of the entrance distance.
Finally it can be concluded that mono-capillaries can be used for the propa-
gation of radiation. This is combined with an intensity gain that depends on the
energy and the length of the capillary.
The spot size of a mono-capillary depends on the inner diameter of the capillary
dcap itself but also on the working distance dworking i.e. the distance from the
capillary exit to the excited area. Due to the total scattering there is a beam
divergence as demonstrated in Fig. 2.29.
This divergent beam enlarges the irradiated area. The final spot diameter is:
2.2 Primary Optics 53
For the small critical angles for X-ray which are in the range of 0.3 this
enlargement seems not important. But for working distances of 3 or 10 mm the
enlargement would be 30 or 100 lm, respectively.
This would be acceptable for capillary diameters in range of 300 lm but for
capillary diameters of only few lm this will be the dominant contribution to the
spot diameter.
This total reflected part also causes an energy dependence of the spot size. For
larger energies the critical angle is smaller and therefore also the spread by total
reflection is reduced. The spot size of capillary optics mostly is given for Mo-K-
radiation i.e. for 17.48 keV.
With elliptical capillaries, beam focusing is possible i.e. if the source is in one
focal point of the ellipse radiation will be focused into the other focal point.
This is possible with a magnification or demagnification of the image as
demonstrated in Fig. 2.32.
Elliptic optics are working only for a limited energy range which is defined by
the geometry of the ellipse. The reflection angle Hrefl on the ellipse for a beam
from the focal point is the same in every direction. It is given by tan Hrefl = e/
dfocal if dfocal is the distance between the focal points and e is the short diameter of
the ellipse. With these parameters the highest total reflected energy is given by:
pffiffiffi
Emax ¼ 2:04 104 dfocal =e q ð2:19Þ
All lower energies are reflected but higher energies are not. Higher energies can
only reflected if the e is reduced. For an Emax of 20 keV and a focal point
distance of 100 mm e would be in the low lm-range!
According to Fig. 2.32 it is possible to use different parts of the capillary. If the
part that is closer to the source is used, the captured angle is larger i.e. the flux
would be larger and the image of the spot then will be enlarged i.e. there is a
magnification. But it has to be taken into account that the entrance distance
depends on the source geometry—for a side window tube the distance is given
by the tube dimension i.e. it cannot be reduced without limits. If, on the other
hand, the part of the ellipse that is closer to the image is used, only a smaller
angle of radiation can be captured which reduces the flux but also the spot size
i.e. the image of the source is demagnified.
Nevertheless, the very small inner diameter of the elliptical capillary which
allows only capturing a small amount of radiation makes this optics not
attractive for bench top instruments. This situation is different for the brilliant
sources of synchrotrons.
2.2 Primary Optics 55
Single bounce optics transmits a wide energy range but this is limited to high
energies which are defined by the critical angle of total reflection. That means
their transmission function is like a low-pass filter i.e. all energies lower than
the energy given by the critical angle of total reflection will be reflected higher
energies are no reflected.
The generation of a linear spot on the sample is possible if two total reflecting
plates are arranged more or less parallel to each other. In that case the radiation can
be total reflected alternatively on both plates as in tapered mono-capillaries. This
allows the concentration of radiation in one direction (see Fig. 2.30). But as for the
tapered mono-capillaries also for this optics the intensity gain is not very high
because the reflecting angle will be increased and can be then larger as the critical
angle for total reflection and the captured angle has to be very small. Further, the
adjustment effort for this optic is higher than for a collimator system and the
working distance must be small which reduces the robustness of the instrument.
For that reasons this arrangement is not applicable for bench top instruments.
Capillary optics can be prepared not only from a single capillary but also from large
bundles of them. In that case every single capillary contributes to the propagation of
radiation [71–81]. The diameter of the single capillaries can be very small down to
the range of &1 lm. This is demonstrated in Fig. 2.33 which shows an SEM image
of the cross section of a poly-capillary system. The size of the single capillaries for
that optics is in the range of 1.5 lm. But the image shows also the complex
structure of the optics. The single capillaries are arranged as a honeycomb in
hexagonal bundles which are settled again as a honeycomb. The hexagonal bundles
in this case are covered with an additional glass capillary. This protects the outer
56 2 Main Components of X-ray Spectrometers
capillaries of a bundle against destroying and keeps their shape during the manu-
facturing process so that they can contribute to the propagation of radiation.
Poly-capillary optics can be used as full and half ‘‘lenses’’. Both types are
shown in Fig. 2.34. Full lenses capture the radiation from the source, propagate the
radiation combined with a change of their direction and concentrate the radiation
on a small spot. In a half lens the radiation exits the optic already when the beams
of every single capillary are parallel. In dependence of the direction of the half lens
it can be used to make a parallel beam from a divergent beam or to focus the
radiation of a parallel beam.
Poly-capillary optics mostly will be denoted as lenses but they have no imaging
functions only a beam shaping function. Therefore they should be better named as
‘concentrator’.
Poly-capillary optics are characterized by the following parameters [82, 83,
116]:
2.2 Primary Optics 57
• Spot size
The spot of a poly-capillary optic is typically smaller than that of mono-cap-
illaries because the contribution of the inner diameter of the single capillaries is
much smaller, see (2.18). Due to the small diameter dcap this contribution in
most cases even is negligible and the contribution of total reflection dominates
the spot diameter.
On the other hand it is necessary to consider a possible mis-alignment of the
single capillaries. That means if not all capillaries are looking to the same
point—like crossed-eyes another term has to be added to (2.18) that takes into
account this contribution according to
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
D2spotpolycap ¼ d2spotmonocap þ d2Alignpoly ð2:20Þ
That means, the spot size of poly-capillary optics depends also on their working
distance. But, of course, for a given poly-cap optic it is not possible to reduce
the working distance only by a change of the distance between the optic and
sample due to the focusing of the optic.
Smaller working distances are possible only by another optic that is ‘‘longer’’.
In that case the beam convergence is stronger i.e. the focal distance shorter and
the result is a smaller spot. This is shown in Fig. 2.35. Due to the stronger
curvature it is possible that the critical angle for total reflection in particular for
high energies is exceeded. Then this radiation penetrates the optics and the
intensity of high energetic radiation is reduced.
With short working distances the spot size can be reduced down to the range of
10 lm. But then the working distance must be \4 mm. This requires a very
careful sample positioning to avoid touching sample and optic and damage one
of them.
Fig. 2.38 Arrangement for the pin-hole method for spot size measurement
reflection. A typical spot size variation with energy of a poly-cap optic is shown
in Fig. 2.36.
For high energies the critical angle of total reflection is small and therefore the
spread also which reduce the spot size. Vice versa the angle is larger for small
energies and therefore also the spot size.
The spot size for poly-cap lenses is mostly defined for Mo-K-radiation i.e. for
17.48 keV. For Mo-L radiation at 2.92 keV it is approx. doubled.
2.2 Primary Optics 59
Table 2.5 Spot sizes of different beam shapes determined with pin-hole and knife edge method
Shape Pin-hole Knife edge Relation pin-hole/knife edge
Gauss 0.84 0.84 1.00
Gauss cut 1.2 1.1 1.09
Gauss pin 0.8 0.66 1.21
Parabolic 2.0 1.6 1.25
Cylinder 0.7 0.68 1.03
Cone 2.0 2.2 0.91
The measurement with the knife edge gives a distribution like a step function. This
distribution has to be differentiated to get the intensity distribution in the beam.
Spot sizes can be given as FWHM of these distribution, but then only 67 % of
the total spot intensity are included or as 3*FWHM which included the com-
plete beam intensity (99.5 %).
It has to be taken into account that with the wire and backward knife edge
method the beam intensity is measured via the fluorescence radiation of the
wire or knife material. In this case not only the distribution of the energy of the
fluorescence radiation is considered but the complete spectrum from the
absorption edge to higher energies. Therefore this measurement gives only a
rough approach for the spot size at the energy of the fluorescence line. If two X-
ray series can be excited in the material of the wire or knife it is possible to
measure the intensity distribution also for two energies, for example in case of
Mo the Mo-K line (17.48 keV) and the Mo-L-line (2.92 keV).
It has to be assumed that the theoretical description for the intensity distribu-
tions in a spot is only an approach. This is demonstrated in Fig. 2.40 which
shows the measured intensity distribution from Rh-K radiation of a poly-cap
62 2 Main Components of X-ray Spectrometers
lens measured with the pin-hole method. For the measurement a pin-hole with a
diameter of 5 lm was used, the step size was 6 lm.
It can be seen that the intensity distribution is inhomogeneous. This can happen
if the different capillary bundles have different transmission efficiencies and
therefore contribute differently to the beam intensity.
Therefore it has to be assumed that even for a more or less Gaussian distribution
the knife edge method can deliver another result for the spot size diameter than
the pin-hole method.
It is necessary to know that the spot size information determined with different
methods can differ. This can explain the different spot sizes declared by the
manufacturer and measured in the instrument. Nevertheless, these differences
are relatively small, for sure smaller than their energy dependence. Therefore
this is not of significant influence for the use of capillary-optics. But the correct
spot size can be expected from the pin-hole method only. For the comparison of
spot sizes of different optics the measurement method has to be mentioned.
• Transmission function
The spectrum propagated in a poly-capillary will be changed by the optic. This
influence is described by the transmission function which can determined by
2.2 Primary Optics 63
measurements with and without optics in the direct beam as well as with a
scattered spectrum. A typical transmission function measured via scattered
radiation is shown in Fig. 2.42. It shows a high transmission for medium
energies and reduced transmission for low and high energies [83].
At high energies the behavior is determined by the different amount of captured
radiation as described already for mono-capillaries in Sect. 2.2.5.2. For low
energies it is determined by the intensity loss due to multiple total reflections as
discussed also in Sect. 2.2.5.2. This loss by multiple reflections is more
important for poly-capillary optics than for mono-capillaries due to their small
inner capillary diameter. Typical inner diameters of 2 lm require for 100 mm
long capillary approx. 250 reflections for the radiation transport. Further the
highly curved inner surface of the capillary can cause surface distortions with an
influence to the reflectivity.
The transmission function can be approached by an asymmetric Gauss-function
which is also called Gumbel-function according to
Fig. 2.43 Scatter spectra of different lenses measured with same conditions
Fig. 2.44 Scatter spectra of the same lens but with different alignment
the outer part of the optic, the reflection angle will be larger and can exceed the
critical angle which means that radiation with higher energy will not be
reflected i.e. the outer part of a lens transports mainly radiation with low
energies and only the inner part of the optic that have no strong bending,
transports radiation with higher energies.
Differences of the transmission function are displayed in Fig. 2.43 for several
lenses. The scatter spectra here are measured with same conditions i.e. with an
Rh-tube, 50 kV and 600 lA, scattered on a pure PMMA sample and normal-
ized to the same spot size.
The scatter spectra have significant differences as a result of their different
transmission functions. The peaks in the spectra are generated by the Rayleigh
and Compton scattering of characteristic lines of the tube target (Rh) and the
material of the detector collimator (Zr). These differences of the transmission
function have to be taken into account for quantification.
2.2 Primary Optics 65
Also the alignment of the optic can influence the transmission function. By the
alignment the position of the optic to the source is changed. Therefore a dif-
ferent part of the source radiation is captured by the optic and change the
transmitted radiation. This is demonstrated in Fig. 2.44.
Here the same capillary optics were aligned to highest intensity and then this
alignment was changed to decrease the total intensity to approx. 90 and 80 %,
respectively. The spectra show a very similar behavior for the low energetic
part but significant changes for the higher end of the spectrum i.e. due to the
mis-alignment the outer parts of the optic now contribute less to the propagation
of high energetic radiation.
The gain G of a poly-cap optic depends on energy and is given by
It is defined as the ratio of the intensities with optic and collimator normalized
to the same spot size and the same distance to the sample. This is the nor-
malized transmission function.
• Halo-effect
In particular for the high energy end of the transmitted spectrum the spot size
can be enlarged. That is happens if high energetic radiation is not totally
66 2 Main Components of X-ray Spectrometers
reflected and penetrates the capillaries—of the complete optic or, if due to the
bended capillaries the critical angle for total reflection is exceeded only of parts
of the optic.
The result can be seen in the knife edge measurements of an Rh-tube with a
poly-cap optic in Fig. 2.45. For the low energetic Rh-L-radiation (2.69 keV) the
step function is sharp but for Rh-K-radiation (20.2 keV) the function is already
blurred. This effect will be even larger for higher energies.
If these distribution distributions are differentiated the peak shape can be
determined. This is demonstrated in Fig. 2.46. It can be seen that for the lower
energy of Rh-L-radiation the FWHM of the peak is significantly wider due to
the higher divergence of the emergent radiation. For the higher energy of Rh-K-
radiation the FWHM of the peak is smaller but it show tails due to the pene-
tration of radiation without total reflection i.e. without focusing.
The fluctuations on both sides of the Rh-L-peak are caused by the limited
statistics which can also be seen in the knife edge measurements.
The simplest choice to generate a small spot is a collimator. The spot intensity then
depends on the captured solid angle of radiation which is determined by the
geometric conditions.
• The captured angle is enlarged for short distances d between source and colli-
mator with 1/d2. But the shortest distance is given for X-ray tubes due to their
geometry. In particular for side window tubes it is typically C25 mm.
• If the collimator is close to the source the beam divergence is large i.e. the
diameter of the collimator is not the diameter of the spot on the sample. To
reduce the spot diameter on the sample the collimator should be as close as
possible to the sample.
• The spot of the tube needs to be small against the diameter of the collimator. In
other case the spot size on the sample is even enlarged.
2.2 Primary Optics 67
For these reasons collimator diameters are typically in the range of [200 lm.
But also smaller collimators are in use. An interesting feature of collimators is that
they can be prepared with special cross sections, for example, with a rectangular
shape. This allows the irradiation of thin but long sample areas.
Collimators have the advantage that they does not influence the tube spectrum.
This facilitates the quantification of the measurements.
The difference between collimators and capillary optics is mainly the different
captured angle. This is demonstrated in Fig. 2.47.
The left part shows the situation for a collimator. Only a small part of the source
radiation can penetrate the pin hole in the collimator which allows small spot
intensities only.
For the mono-capillary in the middle part of the picture there are two contri-
butions to the spot intensity: one contribution is limited by the inner diameter of
the capillary at the exit which is comparable to the collimator. A second contri-
bution is the radiation propagated by multiple total reflection. This increases the
intensity for a mono-cap by a factor of approx. 10 (see Table 2.6). The amount of
additional captured radiation depends on the distance source—capillary entrance.
The spot size for a mono-capillary can not reduced too much because then the
captured radiation is too small.
Because the request for smaller spots is evident, the next step to poly-capillary
optics was necessary (right part of the picture). In that case the input diameter of
the optic can be enlarged which also increase the captured angle of source radi-
ation (see Table 2.6). This is for a poly-cap lens approx. two orders of magnitude
larger than for a mono-cap. But this advantage is partly compensated by the
reduced transmission efficiency (transmission function) of a poly-cap optics.
For the first available poly-cap optics the transmission efficiency for medium
energies was in the range of a few percent only. By technology improvements and
better glasses for the optics the transmission efficiency could be improved into the
range of 10 %. This increases the spot intensity considerably.
The most important benefit of poly-cap optics is that they concentrate the
radiation to a smaller spot which can increase the gain into the range of [10,000.
This chapter introduced and discussed a series of different X-ray optics. Many of
them are also used for micro-analytical tasks. Their use depends on the analytical
68
requirements itself and also on the quality of available sources. The main features
of the different optics and their usability are summarized in Table 2.7. The
selection of the best optic depends on the intended application but also on all other
instrument parameters [84].
Especially for l-XRF, the spot size and the excitation efficiency are important.
The excitation efficiency depends on the elements that have to be analyzed. For a
multiple purpose instrument it is helpful to have optics with a broad band trans-
mission—this could be, for example, capillary optics. Here are used poly-caps for
the very intensive excitation of sample areas in the range of\50 lm and in special
arrangements even down to &10 lm as well as mono-caps for the analysis of very
small sample areas down in the range of few lm but with lower excitation intensity
or for larger spot sizes in the range of 100–300 lm as substitute for collimators ton
increase the excitation intensity.
For certain applications also other optics are used in laboratory instrumentation—
for example focusing SML. Because in that case the transmitted radiation is mono-
energetic these arrangements can be used only for excitation of special elements
with absorption edges close below to the mono-energetic radiation [85, 86].
For synchrotron sources also optics with smaller entrance diameter can be used due
to their excellent brilliance. The small entrance diameters for example of refraction
optics as well as transmission Fresnel optics need a high brilliance but they can
generate significant smaller spots on the sample down into the sub-lm range. Even the
combination of different optics is possible to reduce the spot size and improve the
spatial resolution. Then even spot sizes in the low nm-range are possible [87].
during the measurement in particular a spinning of the sample can be helpful for
averaging sample inhomogeneities.
If the sample position is not defined by the positioning mechanism like a
rotating sample tray a check of the measurement position can be necessary for
example with a camera system. This is a special requirement for l-XRF
instruments which analyze small sample areas and these have to be positioned
correctly in the beam.
• Measurement medium
The measurement medium can be air, vacuum or also a gas flow. This medium
is important either for a reduction of the absorption of low energetic fluores-
cence radiation—for example vacuum or for samples that cannot be evacuated a
He-flush—or for the allocation of special conditions for the sample—for
example cooling of organic samples in a cold gas stream.
The measurement in special media typically requires additional instrumental
effort.
• Excitation direction
Different excitation directions are possible in l-XRF instruments. In bench top
instruments the sample typically is positioned horizontally. Then the excitation
is possible from top or down.
The excitation from down has the advantage that the position is defined with a
high reproducibility by the sample support independent of the sample thickness.
But this is valid only for flat samples. But in the case of small samples special
sample supports are required and in the case of powder tablets small pieces of
the sample can fall into the source or spectrometer and contaminate the
instrument.
The excitation from the top allows the examination also of non-regular shaped
and small samples but in that case the positioning system has to handle different
sample heights. This excitation direction typically is used for l-XRF instru-
ments in connection with optical microscopes for the check of the correct
sample position. Recently this positioning is used also for WDX-spectrometers
to avoid instrument contaminations.
In case of very large or valuable samples that cannot be moved or even placed
into an instrument it is also possible to have a vertical sample position and
horizontal measurement direction.
2.3 Sample Positioning and Radiation Shielding 71
• Radiation shielding
The sample chamber has also another important function for X-ray spectrom-
eters—the radiation protection. X-rays are ionizing radiation which can damage
the human tissue and therefore special care is required for working with X-rays.
On the other hand, the sample chamber design has to avoid any contamination of
the measured spectrum by scattered radiation from the sample support or sample
chamber.
For the analysis of small sample areas it is important to excite the correct part of
the sample. Because the typical sample for l-XRF can be both non-homogeneous
and also non-regular shaped or even a very small particle this requires a special
positioning system. For an excitation from down this can offer several problems—
for example non-regular shaped samples cannot easily fixed in a special position,
small samples needs to be supported for example by a thin film which influence
both excitation and fluorescence radiation by absorption, overlapping the sample
spectrum with fluorescence radiation from the sample support etc.
These disadvantages can be avoided with a top-down geometry. In that case the
sample can be positioned on a stage which needs to be moved in X-Y-Z-direction.
Herewith we consider the movement normal to the excitation direction as X-Y and
in the excitation direction as Z.
For the top-down arrangement the sample lies on a stage it can be moved in
every direction. Non-regular shaped samples can be fixed in a special position on
the stage mechanically or in the easiest case with plasticine. Then it is possible to
examine every interesting part of the sample from the top. Small particles can be
put on a support without an influence of absorbing films or fluorescence radiation
of the support (see also Sect. 5.3).
The excitation from the top is possible with a normal incidence to the sample
surface as well with as a tilted beam. These two alternatives have to be considered
with the possible geometrical arrangement of the tube with X-ray optics, the
detector but also the optical microscope for the control of the measurement
position. All these three components need to be directed to the same sample
position. At least the microscope should have a perpendicular look to avoid
aberrations of the sample image. But there are also few analytical differences that
have to be taken into account.
72 2 Main Components of X-ray Spectrometers
• Tilted incidence
For a tilted incidence the radiation hits the sample for a given position of the
stage in a point that depends on the sample height. The microscope looks due to
its normal incidence every time to the same sample position (XM, YM), see
Fig. 2.48. Only for the correct Z-position the excitation beam hits the sample in
the position displayed by the microscope. This requires an exact Z-setting. The
positioning in Z-direction is possible either manually or with an autofocus
system. With an autofocus system the setting is more objective but with respect
of the various sample qualities the result can vary with sample surface quality—
for example reflectivity, contrast of different sample structures, tilted surfaces
etc. An advantage of a tilted incident angle is a reduced scatter intensity of tube
radiation which improve slightly the peak-to-background ratio.
Further, the tilted incidence causes that the spot on the sample is not a circle but
an ellipse. For an incidence angle a the long axis of this ellipse will be enlarged
to dellipse = dspot/sin a. This influences of course the spatial resolution but only
in one direction. Perpendicular to the tilted incidence the spot size is not
changed and therefore in this direction there is also no influence to spatial
resolution.
• Normal incidence
In case of a normal incidence the position of the excited area is not changed by
changing the Z-position of the sample. That avoids the requirements for an exact
positioning in Z-direction to measure in the correct sample position which is
difficult in particular for very small spot diameters. Nevertheless, this exact
Z-positioning is necessary at all due to the influence of the distance between
sample and detector to the collected intensity for a following quantification.
Further it has to be taken into account that both a normal sample view by the
optical microscope and a normal incidence of the X-ray simultaneously are not
possible. That means, the image must be frozen and then the optic or the sample
position has to be changed or the optical image looks slightly tilted and have a
parallax aberration.
But there is also another advantage of the normal incidence. In that case the
angle between the incident beam and the detector direction UInc is small i.e. the
2.3 Sample Positioning and Radiation Shielding 73
scatter angle 0scatter is relatively large (0scatter = 180 - UInc). Therefore dif-
fraction peaks of crystalline samples are moved into the low energetic spectrum
region. Diffraction peaks can be generated by Bragg-reflection of the poly-
chromatic tube radiation on crystalline sample areas (see also Sect. 2.5.6.4).
Larger scatter angles 0scatter but same d-spacing increase the Bragg-reflected
wavelength according to (1.9) which means the diffraction peaks moved toward
lower energies and have less influence to the fluorescence spectrum.
For both variants of beam incidence an exact setting of the sample height is
required because the distance d between sample and detector influences the col-
lected intensity I according to I * 1/d2. The setting can be performed manually
but also by autofocus systems.
The following autofocus systems are in use:
• Tilted incident of a laser beam
In this case it is possible to use different information. The position of the
reflected laser beam can be measured with a position sensitive detector. If the
reflected beam meets the detector the sample is in the correct height. This
triangulation method is automatable but requires sample surfaces without tilting
and with a comparable reflectivity to avoid detector saturation. For slightly tilted
samples the positioning will be not correct.
Therefore another possibility is more often used—a check of the overlay
between the laser spot and the optical image of the measurement position. But
also this version offers some problems because the laser spot can look very
different in dependence of sample color or reflectivity. The laser spot can also be
blurred which is a problem for the image processing. Therefore this method also
is not automatable but can be helpful for a manual Z-positioning.
value is changing in dependence of the Z-position. The focal plane is reached for
the highest contrast value. This method requires a good sample illumination and
for highly polished sampled it can be difficult to detect any contrast. The
accuracy also depends on the sample and the image magnification.
For testing this autofocus procedure, repeated settings with the autofocus system
were performed for different samples (polished white plastics with low struc-
turing and a machined bronze) and with different magnification of the used
image. The results are summarized in Table 2.8. It can be seen that the precision
for structured samples is in the range of 10 lm which is sufficient for spot sizes
of 25 lm but for low magnification it drops down to [100 lm and for weak
structures it is also too large for these small spots.
The requirement regarding accuracy for the sample position in Z-direction
depends on the tilting angle of the incident beam. The displacement of the spot
on the sample should be limited to approx. a quarter of the spot diameter. That
means for an incident beam of 45 and a spot size of 25 lm the sample posi-
tioning is Z-direction should be approx. 6 lm. For smaller spot sizes or for
smaller incident angles this requirement can even be stronger.
The influence of a different sample height to measured intensity is negligible
only for variations of sample height \50 lm. The measured intensity is
influenced considerably by the distance D between sample and detector due to
the dependence on D2. Typical distances are in the range of 20 mm. For this
distance the intensity change for a 50 lm variation would be approx. 0.5 %.
diameters down to 20 lm step sizes in the range of 2–5 lm are required. Mostly
in instruments the step sizes are better.
• Repeatability
The accuracy for setting the position has to be considered for two different
situations—the movement into a certain position for a single point measurement
and the movement for multi-point measurements and distribution analysis.
For the measurement of single small points the requirements are not high
because the measurement position is counterchecked with the image of the
optical microscope. A wrong positioning can be easily corrected if the step size
is small enough.
In the case of multi-point and distribution analysis the accuracy must be realized
even for larger areas and for longer measurement times. The accuracy given by
the minimum step size would be sufficient but a high reproducibility of the stage
movement is required.
• Speed
The speed of the stage depends on different parameters and should be adjustable.
For the fine positioning the stage should move slowly, on the other hand for the
movements over larger distances it should be faster. In all cases it has to be
considered that the sample should not be moved on the stage i.e. it needs to be
fixed on the stage or the stage movement needs to have a continuous acceler-
ation and deceleration.
These requirements depend also from the measurement mode for distribution
analysis. For a ‘stop and measure’ mode the speed needs not very high because
during the measurement the sample moves only short distances. For measure-
ments ‘on the fly’ the constancy and repeatability of movement also over longer
distances is very important to avoid shifts between the different scans.
It is necessary to see an image of the sample for the correct sample positioning.
The small spots of l-XRF with diameters less than 1 mm and even down into the
low lm-range allow a correct positioning only with help of optical images of the
sample with sufficient magnification. This magnification needs to allow the correct
positioning for the actual spot as well as the display of a sample area that is large
enough for an orientation. This applies even to large samples. Due to the large
range of sample sizes this can require images with different magnification.
The highest magnification is determined by the spot size. The image should
show the spot in a sufficient size at least with diameters in the mm-range in the
image.
76 2 Main Components of X-ray Spectrometers
That would mean for a spot size of 20 lm the image should show a part of the
sample which is in the dimension of approx. 1 mm or less, for a spot size of
0.5 mm the image can show a part of the sample which is in the dimension of
approx. 1–2 cm. For the first case a second optical microscope would be helpful
for a fast orientation, the second image size would be already sufficient for this
orientation.
But sample sizes can be also significantly larger. In that case the generation of
an image of the complete sample can be very helpful. This is possible by stitching
a series of adjacent images like a mosaic. That allows the preparation of sample
images which are large enough for orientation.
If images with different magnification are available it is possible to switch in
between them. In that case it is possible to have a sufficient overview for larger
sample areas as well as the possibility for a correct final positioning which can be
supported by the display of the spot in the image.
This is demonstrated in Fig. 2.49 for a piece of a concrete with different
magnification. The left part of the image shows a mosaic image of the sample with
a size of approx. 5 9 4 cm2, the image in the middle shows a sample part of
approx. 1 cm2 and the right part with highest magnification an image of a sample
part with a size of approx. 1 mm2 which is sufficient for the spot with diameters in
the range of 20 lm.
Fig. 2.50 Knife edge distribution with intensity fluctuations and their differentiation (left).
Determination of spatial resolution from intensity change (right)
Fig. 2.51 Linear distribution measured with spot size of 14 lm (black) and 25 lm (gray)
Fig. 2.52 Influence of spatial resolution from spot size (left 25 lm; right 14 lm)
Fig. 2.53 Mapping of a periodic structure with a step size of 4 lm (left) and 15 lm (right), the
lower images show the corresponding linescans
Fig. 2.54 Measurements with different step size (50, 25 and 10 lm—from left to right) but with
same pixel time (10 ms)
So far it can be concluded that a reduction of spot size will be the most efficient
possibility for an improvement of spatial resolution.
Fig. 2.56 Measurements with different pixel time (0.3, 1 and 5 ms—from left to right) but with
same step size (4 lm)
number of pixels should be large enough to have a sufficient ‘dpi’ for a good
image quality. Further the measurement time has to be considered—for the
images of the Cr-pattern in Fig. 2.54 the increase is significant, from 80 s over
335 to 885 s, respectively. This requires a compromise between image quality
and measurement time.
• Influence by intensity
Another influence to the spatial resolution is given by measured intensities per
pixel. Every measured X-ray intensity N has an intensity error that is determined
by Gauss-statistics i.e. the variance of a measured intensity is 1/HN. Therefore
the contrast changes in dependence of the collected intensity. That is demon-
strated in Fig. 2.55 which shows the statistically distributed intensities of a
linescan along a line with an intensity change of 20 %.
These distributions are calculated with medium intensities in every pixel of 10
and 104 counts, respectively. It can be seen that for the lower intensities no
structure can be detected because the step is completely overlapped by the
statistical fluctuations. But for the higher intensity the relative statistical fluc-
tuations are smaller and it is possible to detect the structure clearly.
2.3 Sample Positioning and Radiation Shielding 81
– With an increasing number of pixels also the total measurement time will be
increased. It has to be decided which time is available for a certain
measurement.
– The collected data needs also to be displayed i.e. if the number of measured
pixels is significantly larger than that of the screen for data display not all data
can be displayed in one step only by zooming in for certain details.
– The number of pixels is limited by the addressability of the instrument PC and
may be also by the size of the available memory.
From that considerations maximum pixels sizes in the range of 2,000–5,000 in one
direction should be the upper limit.
The fluorescence radiation will be absorbed on the way to the detector. This
absorption depends on the detector window (see also Sect. 2.5.4). An additional
influence to the fluorescence intensity is given by the distance between sample and
detector as well as the medium between them.
The most often used measurement media are air and vacuum. The absorption in
air is considerably only for light elements and can be reduced with a measurement
in vacuum. Figure 2.58 shows the intensity enhancement due to reduced absorp-
tion for measurements in vacuum in comparison to air for a sample—detector
distance of 20 mm.
From these results it can be seen that the detection of light elements requires the
measurement in vacuum for that sample-detector distance for the elements from
Na to P. The vacuum level needs not to be very high because due to the limited
excitation efficiency for light elements Na is the lightest detectable element at all.
A pre-vacuum of 20 mbar is sufficient. The absorption of Na-K radiation on a
distance of 20 mm in that pressure is still approx. 3 % but for a controlled pressure
the absorption is constant. This pre-vacuum can be realized with pre-vacuum
pumps and can reached in a short time.
2.3 Sample Positioning and Radiation Shielding 83
The secondary optic is positioned in between sample and detector. It can be used
only for beam shaping i.e. can be a collimator to avoid the detection of scattered
radiation from the instrument into the spectrometer and improve thereby the peak-
to-background ratio or it can generate a parallel beam for the dispersive crystal of a
following wavelength dispersive element.
But the secondary optic can be also the dispersive element of the spectrometer
itself—either a filter or the Bragg-reflecting element i.e. a crystal or a synthetic
multi-layer. In that case we discuss about wavelength dispersive spectrometers.
All optics which can be used in that position are the same like primary optics
and are already discussed in detail in Sect. 2.2. Therefore a recapitulation here is
not necessary.
But the spectrometer type will be defined by the secondary optic and the
detector. The following two spectrometer types has to be distinguished:
• Wavelength dispersive spectrometers
• Energy dispersive spectrometers
But a small entrance slit limits the captured angle of sample radiation and therefore
the brightness of the spectrometer. Therefore high excitation intensities are nec-
essary. WDX instruments works typically with X-ray tube in the power range of
1–4 kW.
N ¼ E=e ð2:23Þ
The number of primary charges is small and therefore the signal needs to be
amplified. After amplification a pulse height analysis determines the frequentness
of absorbed photons of certain energies. This represents the energy distribution of
the absorbed X-rays i.e. it is the measured spectrum.
Energy dispersive spectrometers have typically a less resolution than wave-
length dispersive instruments. This can be seen for WDX and EDX instruments in
Fig. 2.60.
For WDX instruments the energy resolution is considerably better especially for
low energies. But a series of different crystals (or multilayer systems) are required
for covering the complete wavelength or energy range, in particular, if the scatter
angle should be in a range of high resolution i.e. according to (2.6) for large scatter
86 2 Main Components of X-ray Spectrometers
angles. The energy resolution for every crystal decreases significantly with
increasing energy.
The degradation of resolution with energy of EDX spectrometers is not so
strong. In the range of 12 to 15 keV they show even better energy resolutions than
WDX instruments.
Regardless of the limited resolution of energy dispersive spectrometers there
are also some other differences between the two spectrometer types [88, 89].
• Spectrometer brightness
The brilliance of EDX-spectrometers is significantly better because the detector
can be arranged very close to the sample and can therefore capture a larger solid
angle of emitted radiation (see Fig. 2.61). This results in a brightness which is
by a factor of approximately 100 larger than for WDX instruments. That means,
EDX instruments needs accordingly less excitation intensity i.e. the tube power
can be also reduced by a factor of also approximately 100. Typical tube powers
for energy dispersive spectrometers are in the range of 1 to 50 W. This means a
significant reduction of cost and effort (tube cooling) against WDX instruments.
• Simultaneous measurement
Due to the collection of the complete sample radiation, photons of all energies are
captured at the same time i.e. all energies are measured at the same time and
therefore also all elements are detected simultaneously.
On the other hand count rates of energy dispersive detectors typically are lower
than for WDX instruments because the pulse height analysis requires a certain
time.
In recent years a new type of energy dispersive detectors has been introduced,
the Silicon-Drift-Detectors (SDD) which has not only an excellent energy reso-
lution but also very high count rate capability—up to several hundred thousand
counts per second. Together with the simultaneous measurement which avoids the
2.4 Secondary Optics: Spectrometer Type 87
change of scatter angles this can mostly compensate the count rate advantages of
WDX-spectrometers.
• Mechanical effort
The mechanical effort of EDX-spectrometers is very low. The only required
mechanics is the sample positioning. No high precise gear for the movement of
the detector and the analyzing crystal is necessary. That means there is a
reduction of costs and improvement of robustness of the spectrometer in com-
parison to WDX instruments.
A summary of the comparison between EDX and WDX instruments is given in
Table 2.9.
In particular for l-XRF the low brilliance excludes WDX spectrometers. For a
high intensity in the sample spot a high brilliance on the tube target is required.
But here the limit is reached with the existing fine focus tubes.
For detection of X-rays different types of detectors are used. At first for the
detection of X-Ray photographic plates or films was used. They allow the detection
of X-rays but not an exact intensity measurement—the intensity was measured via
the blackening of the photographic layer. Energy determination in that case was
88 2 Main Components of X-ray Spectrometers
possible only with a dispersive element like a crystal and the imaging of the
spectrum on the photographic film—i.e. in wavelength dispersive geometries.
Another detector used mainly in WDX-instruments is the scintillation counter.
In these detectors the X-rays is converted to light in a photosensitive layer which
then is amplified in a photomultiplier into an electrical voltage pulse. The energy
resolution of scintillations counters is very limited—typically in the range of 50 %
[90, 92]. With these detectors it was possible to measure the intensity by photon
counting. These detectors are used for wavelength dispersive instruments because
they perform the energy dispersion by Bragg-reflection and the detector needs only
to count the photons of the reflected energy. But then a high count rate capability is
required.
With the development of gas filled detectors an improvement of energy reso-
lution was possible. Their typical resolution is in the range of 12–25 % [90, 91].
These detectors are used either in wavelength dispersive instruments only for
counting but also in low level energy dispersive instruments. In WDX instruments
their energy resolution can be used for the suppression of Bragg-reflexes of higher
order and also for the improvement of the peak-background ratio.
In gas filled detectors the incident X-ray photons will be absorbed by the gas
atoms and generates gas ions and electrons. This filling gases are mostly precious
gases with ionization energies is in the range of few 10 eV. The electrons are
accelerated in the electrical field of the detector. In that way they can attain
sufficient energy to ionize further gas atoms. Close to the very thin counting wire
the electrical field will be increased and an electron avalanche is generated which
is collected in the counting wire. This internal gas amplification produces an
electrical current pulse that can be easily transformed into a voltage signal over a
large electrical resistance which is in series to the counting wire and can be further
processed by pulse height analysis. This detector principle is displayed in
Fig. 2.62.
Prop-counters can be prepared with large sensitive areas which allow the
detection of a large solid angle of radiation but the energy resolution is often not
sufficient for the analysis of unknown samples.
Energy dispersive detectors with high resolution are prepared from semicon-
ducting material. The incident X-rays excites an electron from the valence band to
the conduction band. These charges will be collected in an electrical field. The
excitation process requires only a small energy—in the low eV-range and therefore
2.5 X-ray Detectors 89
the number of primary charge carriers is higher than for gas filled detectors. This
improves the statistics but requires a careful collection of the charge carrier pairs
in the electrical field of the detector and of the signal processing to get a sufficient
signal.
A systematic overview for the different types of energy dispersive detectors is
displayed in Fig. 2.63.
The left part show the gas filled counters. They can be produced in different
shapes and sizes and with large sensitive areas. The gas filling is mainly by
precious gases because they are chemically inert against the influence of ionizing
radiation. In dependence of the gas and its density (pressure) the absorption of the
detector can be adapted to the energies that should be absorbed. Typical detection
gases are Ar, Kr and Xe. Additionally to the precious gas a quench gas is required
to stop the avalanche process. For that purpose different gasses are used. They can
change the chemical status during quenching. Therefore this quenching gas will be
consumed.
Depending of replacing the quench gas there are two different types of detec-
tors—sealed detectors where the amount of quenching gas is fixed and determines
the living time of the detector given in events that can be counted. Typically this is
in the range of 1012–1014 in dependence of the detector size. On the other hand in
flow counters the mixture of counting and quenching gas will be refreshed con-
tinuously. Flow counters are used for the detection of light elements that require a
very thin entrance window which cannot avoid the contamination of the detection
gas by air. This would decrease or even damage the detection properties of the
detector if the gas filling would not refreshed.
A special type of gas filled detectors is the Proportional Scintillation Counter
[93, 94]. This is a combination between a gas filled counter and a scintillation
counter. The incoming X-ray will be absorbed by the precious gas. When the gas
ions recombine to atoms by capturing an electron, the binding energy of the
electron will be emitted as electromagnetic radiation. These binding energies are
the range of 20–40 eV i.e. the emitted radiation is in the visible range. This light
90 2 Main Components of X-ray Spectrometers
can be captured and amplified with a scintillation process. In this case the internal
amplification is not a gas-amplification but an electrical process which improves
slightly the energy resolution. But the instrumental effort for these detectors is high
and therefore they are not in commercial use.
Nowadays mostly energy dispersive detectors are prepared from semiconduc-
tors [95–98]. They are summarized in the right part of Fig. 2.63. Here different
materials are available but the most often used is Silicon. This is because Si offers
a good combination of parameters that are important for good detecting proper-
ties—as a small energy e for the generation of a charge carrier pair, a relatively
long living time of the charge carriers, a high resistivity against radiation damages
and this material can also be easily managed both for manufacturing and for
handling as detector due to the high experiences from micro-electronics. High
effort was necessary for the development of low noise detector electronics for a
good energy resolution. It has to be adapted to the detector and needs not only for
the amplification of the signal but also for the suppression of noise and sum peaks
as well as perform a fast baseline restoration.
The energy resolution of semiconductor detectors is in the range of 120–200 eV
i.e. by a factor of 5–8 better than a gas filled detector but their count rate capability
is reduced—typically in the range of few 10,000 counts per second (cps). Also
their sensitive areas are only in the range of few 10 mm2. The mostly used
detectors that are used for l-XRF will be discussed later in Sect. 2.5.5.2.
These semiconductor based detectors have a limited sensitive area and collect
the radiation typically in one detector element which are read-out after every
single event. That means there is no adequate possibility for a position sensitive
collection of radiation as in photographic films. With Charge-Coupled-Devices
(CCD) [99–102] it is possible to fulfill also this function with electronic detectors.
In CCD’s the radiation is absorbed and transferred into electrical charges as in
other Silicon based detectors i.e. the number of charges are proportional to the
absorbed energy of one or several photons. The special structure of electrodes of a
CCD divided the complete detector in small areas which can be considered as
separate detector elements—as a pixel. In these pixels the generated electrical
charges are not immediately collected but stored. Due to the special arrangement
of electrodes all collected charges can be transferred from one pixel to the
neighboring by periodical changing of the electrode potential. In that way the
electrical charges of every pixel can be sequentially read-out. A comprehensive
signal processing generates a digital value that represents the absorbed energy of
every pixel. This read-out process requires a relatively long time—due to the high
number of pixels. During that time the detector needs to be closed that no other
radiation is absorbed. Due to the relatively long illumination time of the detector
the pixels can absorb more than one photon. In that case the generated electrical
charges represent the sum of the absorbed energy. But if during the collection time
only one photon is absorbed the electrical charge of the pixel represents its energy
i.e. in that case the detector would be both position sensitive and energy dispersive.
This requires a high read-out frequency of the detector and allows only a low
irradiation intensity to avoid the absorption of more than one photon in a pixel.
2.5 X-ray Detectors 91
Detectors with that functionality are available with pixel numbers up to 512 9 256
and pixel sizes down to 13 lm [100] and read-out frequencies of 400 s-1.
CCD’s are manufactured from Si-wafers with are relatively thin which reduce
the absorption of high energetic radiation (see 2.5.4). Therefore CCD’s have a high
detecting efficiency only for low energetic radiation. Because the absorption
probability of higher energies is small for higher energies an additional fluores-
cence materials from heavy elements in front of the detector is used which transfer
the high energetic radiation into radiation of lower energies which can then
detected by the CCD. But this is accompanied with a reduction of spatial reso-
lution because the fluorescence radiation is scattered and also of energy resolution
due to the transfer to low energetic radiation.
The potential of CCD’s to generate an image of the spatial distribution of
radiation can be used for different experiments—similar as a photographic plate
for the measurement of distributions of scattered X-ray in diffraction experiments
but also for the measurement of radiation which is reflected in different Bragg-
angles for a simultaneous wavelength-dispersive spectroscopy.
CCD’s are very helpful if a simultaneous collection of radiation fields are
advantageous for example for time resolved measurements or for the measurement
of very weak signals which require a long collection time.
The detector creates a number of charge carriers that are collected and generates an
electrical pulse. This pulse has to be amplified and filtered. Then a pulse height
analysis is performed i.e. the height of every pulse is converted with an ADC into a
digital value. This is interpreted as a channel of the spectrum and consequently this
channel will be incremented. After the detection of a high number of X-ray
photons and their pulse height analysis the contents of the channels represents the
probability of energy distribution of detected X-rays. This process is demonstrated
in Fig. 2.64.
92 2 Main Components of X-ray Spectrometers
The influence of filtering to the energy resolution is explained in Fig. 2.65. The
absorption of several photons of same energy will generate pulses with slightly
varying and statistically distributed heights because they are overlaid with different
noise contributions. Therefore the produced peak heights have a Gaussian distri-
bution. The energy resolution of the detector can be described by the Full Width at
Half Maximum (FWHM) of the distribution of the collected counts in this peak.
The quality of the filtering process depends on the shaping time. For long
shaping times the filtering is good i.e. the variations in pulse heights are small.
Then also the FWHM is small i.e. the detector have a good energy resolution. But
during the filtering the detector channel is closed (i.e. dead) which contributes to
the dead time and reduce the count rate capability of the detector.
On the other hand, for a short shaping time the filtering is limited and the pulse
height variation is larger. That means peaks in the spectrum are wider or the
energy resolution is reduced but due to the shorter shaping times, the dead time of
the detector is reduced and its count rate capability is improved.
with
E X-ray energy
F Fano factor (property of the detector material)
e Mean energy for creating an energy-hole pair
94 2 Main Components of X-ray Spectrometers
Additional to the statistical contribution there are also the electronic noise
contributions which enlarge the variance of the signal. The Electronic Noise
Contribution (ENC) can be influenced by the filtering process. This filtering
process is determined by the filter or shaping time s. There are three different
contributions described by:
with
Cdet Detector capacitance
T Temperature
Id Leakage current
s Shaping time
ai Coefficients
filtering which averages the noise contribution and results in a better energy
resolution. This noise also depends on the detector capacity C i.e. smaller detector
capacities reduce the noise contribution or improve the energy resolution.
capacity and their dark current differ strongly—the capacity of an SDD is approx.
20 times smaller and the dark current approx. 50 larger than for an Si(Li). It can be
seen that the sum has a minimum at a certain shaping time. Shorter as well as
longer shaping time results in higher noise contributions i.e. reduce the energy
resolution.
The energy resolution of the complete signal channel is the geometrical sum of
both the energy dependent part and the electronic noise contribution.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DEtotal ¼ ðDE2S þ ENC2 Þ ð2:26Þ
The energy dependence of this sum is displayed for different ENC in Fig. 2.67.
It shows that for high energies the total energy resolution is proportional to energy.
An increasing ENC shifts the FWHM for high energies only slightly towards
higher FWHM but for low energies the influence of the ENC is considerably.
The different peak shape with enlarged FWHM for increasing energy is dem-
onstrated in Fig. 2.68 for the same integral peak intensity and the same electronic
noise contribution. It can be seen that the peak intensity drops down for higher
energies caused by the energy dependence of the FWHM. Consequently, also the
peak-background ratio will be reduced for higher energies.
As described, the energy resolution depends on the shaping time s. For long
shaping times the noise suppression typically is better i.e. the energy resolution can
be improved. Further there is a dependence of count rate for Si(Li)-detectors (see
Fig. 2.69). For a given shaping time the FWHM is not changing for a certain range
of count rate but for higher count rates the FWHM is increasing significantly i.e.
the energy resolution is reduced.
During processing a pulse, the signal channel is closed or dead. During that
time no other pulses can be processed. The sum of these times is called dead time
tdead. The time where the measurement channel is available for the next pulse is
called live time tlive. The real or clock time is
2.5 X-ray Detectors 97
The dead time is growing with longer shaping times s and for high count rates.
Therefore the count rate capability of an X-ray spectrometer depends of the
shaping time of the signal channel. As demonstrated in (2.25) also the energy
resolution depends on that parameter. But the influence of s to both count rate
capability and energy resolution is different—high count rates are combined with
reduced energy resolution and vice versa. This is demonstrated in Fig. 2.70 and
requires a compromise for selecting the optimal shaping time in dependence of the
application. For that purpose shaping times of Si(Li)’s had to be switched between
different settings.
Due to the high count rate capability of SDD’s this now is not more necessary.
Even for high count rates the energy resolution is still very good for these detectors
and reduces the effort for different shaping times. The high count rate capability of
these detectors also change the requirement for best energy resolution to get a
good peak-to-background ratio as discussed in 2.5.5.2.
The first energy dispersive detectors were gas filled proportional counters. They
were used also for X-ray spectroscopy. But for higher energies i.e. for c-spec-
troscopy their detection efficiency was not sufficient because the low density of the
filling gas. Therefore semiconductor detectors are used for high energies. For high
energies their energy resolution is very good. Mostly detector materials such as Ge
were used for that purpose. With the reduction of electronic noise contributions in
particular, by the reduction of leakage current due to improvement of the detector
material, by improved manufacturing processes and also by the decrease of the
detector capacity these detectors could be also used step by step for lower energies.
One important step was the improvement of charge carrier living times in the
detector. That was possible by using high purity Si (HP-Si), but the preparation of
HP-Si is difficult. Therefore another possibility was the compensation of trapping
effects in the Si. This could be done by Li-atoms that are attached to Boron-doped
areas in the Si and reduces their activity for recombination. Another step was the
reduction of the electronic noise contribution by cooling the detector—down to
temperature of liquid nitrogen i.e. -196 C according to (2.25).
So for a long time LN2 cooled Si(Li)-detectors were the common detectors for
energy dispersive X-ray spectroscopy. They allowed energy resolution is the range
down to 128 eV for Mn-K radiation. But they had some significant limitations,
including:
• The sensitive area of Si(Li)-detectors needs to be small because an increasing
sensitive area enhanced the detector capacity and therefore also the electronic
noise contribution.
• Their shaping times needs to be relatively long because of a high sequential
noise contribution (2.25). But this limits the count rate capability due to high
dead times. For higher count rates shorter shaping times was required which
then decrease the energy resolution. For a good relation of energy resolution and
count rate capability often a series of 5–8 different shaping times were necessary
which complicates the handling of these detectors.
• The cooling with LN2 was for sure the strongest limitation of Si(Li)-detectors.
This could be a logistic problem but also the availability of the detectors was
limited because after a warm up of the detector the time for cooling down and
reaching temperature stability was in the range of 1 h.
Additionally to that inconveniences also the weight and size of the detector dewar
for the LN2-cooling cannot ignored and influences the size and design of complete
X-ray spectrometer.
Another type of energy dispersive detectors that used also for X-ray spec-
trometry was High Purity Germanium detectors (HP-Ge). They have a slightly
better energy resolution than Si-detectors due to the smaller energy e which is
102 2 Main Components of X-ray Spectrometers
required to generate a charge carrier pair (see Table 2.10) and the high absorption
allows also the detection of higher energies. But they had two main disadvantages:
• The HP-Ge is very sensitive against diffusion of the detector contact and win-
dow material into the bulk which destroys the HP-status of the Ge. Therefore
they need to be cooled continuously—also during transports and even if they are
not used for working. This makes the handling of HP-Ge detectors very
inconvenient.
• Ge has a relatively high atomic number and therefore a high absorption for X-
rays. But that means also that escape effects (see Sect. 2.5.6.2) have a high
probability also for radiation with higher energies. These escape peaks would
strongly influence the fluorescence spectrum.
Therefore the introduction of Silicon-Drift-Detectors (SDD) was an important
step in detector development [104–110]. SDD are prepared from Si-Wafers. The
absorption of X-ray photons and the use of their energy to generate charge carriers
is identical to Si(Li)-detectors. But the charge collection is different. It is executed
by an electrical drift field which is generated by a series of electrodes that all have
a different potential. The field strength is less than for Si(Li)-detectors which
allows the use also of not highly compensated material.
The electrons are drifting in this field to a collecting point in the middle of the
detector. Due to this arrangement the detector capacitance is reduced—approx.
20 times smaller than for a Si(Li)-detector. In this collecting point also the first
2.5 X-ray Detectors 103
stage of the preamplifier can be implemented which means that the signal is
already amplified directly on the detector, and noise contribution by cross influ-
ences are reduced.
The differences between Si(Li)-detectors and SDD’s can be discussed with help
of (2.25). This relation is displayed in Fig. 2.66 both for a typical Si(Li)-detector and
a typical SDD. It can be seen that the sequential noise and the flicker noise of a SDD
is significantly smaller than for a Si(Li) caused by the smaller detector capacity CDet.
Because this capacity contributes by square these influences are strong. On the other
hand the parallel noise contribution of a SDD is larger due to a larger leakage current
which is caused by the detector material itself but is also influenced by the FET in the
preamplifier. The result is a movement of the sum of all three contributions for the
SDD towards shorter shaping times s. That means their count rate capability is much
higher because the dead time of the detector is reduced. Additional due to the small
sequential noise it is also possible to work with a SDD at higher temperatures. This
makes the handling of SDD’s much easier (Fig. 2.75).
On the other hand, SDD’s have only thicknesses in the range of \1 mm. As
discussed in the last section, this reduces the high energy detection efficiency. The
first SDD’s was 300 lm thick—as Si-wafers for micro-electronics but now also
slightly thicker detectors are available with 450 lm and even 750 lm thickness.
The main parameters of Si(Li)’s and SDD’s are listed in Table 2.11.
PIN-diode has already better energy resolution but they are still limited in
energy resolution as well as for count rate. The sequential noise contribution for
PIN-diodes is relatively high. Therefore they need longer filter times, but then
for high count rates the dead time limits the performance.
Si(Li)-detectors have good energy resolution but their count rate capability and
handling due to the LN2-cooling limits their use.
2.5 X-ray Detectors 105
Fig. 2.79 P/B-ratio of a weight fraction of 10-5 of Mn for improved energy resolution and
increased count rate
pffiffiffiffiffiffiffiffiffiffiffi
LODelem ¼ 3 ðBGÞ=Ielem welem ð2:28Þ
with
BG Noise or spectral background
Ielem Intensity of the element of interest
welem Weight fraction of the element of interest
From that relation it can be concluded that a high peak-to-background ratio (P/B-
ratio) will improve the LOD i.e. the sensitivity. An improvement of the P/B-ratio is
possible for example for a better energy resolution. In that case the peak has a
reduced FWHM but higher peak-value as demonstrated in Fig. 2.78 for Mn-K-
spectra with different energy resolutions.
Therefore a good energy resolution was for a long time the main possibility for
increasing the sensitivity. But it can be already seen in Fig. 2.78 that the difference
between 125 and 145 eV is not evident.
2.5 X-ray Detectors 107
or by a higher count rate which is possible due to a more efficient excitation as well
as by a larger captured solid angle of radiation i.e. a larger sensitive area of the
detector. Both require a better count rate capability i.e. a shorter shaping time. But
this would reduce the energy resolution of the detector i.e. and then also the P/B-
ratio according to Fig. 2.78.
But for a considerably improvement of count rate this can have a larger
influence than a better energy resolution to the sensitivity as demonstrated in
Fig. 2.79.
These two images show the normalized Mn-K-spectra for an assumed weight
fraction of 5 % for different energy resolution in the left part and for different
count rates in the right part.
The left part assumes an average channel content of 1,000 counts and there are
spectra of the pure background (light blue) and two spectra for different energy
resolution, 125 eV (dark blue) and 145 eV (red). It can be seen that the
improvement of energy resolution do not considerably effect the peak to back-
ground ratio for that conditions.
In contrast, the right figure shows the spectra for the same energy resolution of
145 eV but with different intensity—one for comparison again with an average
intensity of 1,000 counts per channel (red) and one with 10,000 counts per channel
(dark blue) together with the corresponding background (light blue). Here not only
the statistical fluctuations are smaller but also the P/B-ratio is significantly
improved.
From that result it can be concluded that a higher count rate is much more
important for the improvement of the P/B-ratio that an improved energy resolu-
tion. That is valid, in particular, for higher energies i.e. [ 5 keV. Here the influ-
ence of energy resolution to the P/B-ratio is small. A good energy resolution is
required for very low energies because there the energy independent ENC (see
(2.26)) has a larger influence to the peak separation—in particular because in the
low energy range are the most lines, K-lines of atoms with low atomic number, L-
lines for atoms with medium atomic number and M-lines of atoms with high
atomic number. For detectors with low count rate capability the improvement of
energy resolution was the only possibility for a higher sensitivity but now for
detectors with high count rate capability an improvement of count rate is much
more efficient for an improvement of the P/B-ratio even if this is possible only
with a reduction of energy resolution.
108 2 Main Components of X-ray Spectrometers
An energy dispersive spectrum shows several structures which are not representing
the fluorescence radiation of the analyzed sample. This includes the diffuse scat-
tered (Rayleigh and Compton) tube radiation but also artifacts generated by the
detector or signal channel during the counting and signal processing in the detector
itself or due to interactions of the excitation radiation in the sample.
These artifacts have to be considered for the spectrum evaluation by corrections
or by their obviation.
Every detector has a response function that influences the intensity distribution of
the measured spectrum. The first influence is the broadening of every X-ray line
according to the different noise contributions. The fluorescence lines have a natural
width in the range of \ 1 eV but due to the different noise contributions they are
broadened. The result is a Gaussian distribution of every peak with an energy
dependent FWHM as described by (2.24). This broadening is valid not only for the
fluorescence lines but also for the scattered bremsstrahlung background of the
tube.
Furthermore there are influences of the spectrum due to incomplete charge
collection (ICC). This happens by two different processes [111–115].
2.5 X-ray Detectors 109
Fig. 2.81 Probability for generation of escape peaks (left) original and escape corrected spectra
of the Ti-K-lines (square root intensity scaling) (right)
hand, incident X-rays with low energies will be absorbed close to the detector
surface. That increases the escape probability. If the energy of the incident photon
is reduced to values that are smaller than the absorption edge of the detector
material the escape probability drops down to zero. The escape probability for a
thick Si-detector is shown in Fig. 2.81, left part [114].
This probability can differ for thin detectors. Then also escapes are possible on
the back-side. However, the escape peak probability for a detector type is fixed and
can be exactly described. Therefore they can also corrected by subtraction of the
escape peak. Often spectra are already corrected before displaying. This avoids the
erroneous interpretation of escape peaks as element lines. The right part of
Fig. 2.81 shows a spectrum with an intensive Ti-peak. For Ti the escape-proba-
bility is in the range of approx. 0.7 %. The escape peak is clearly seen in the
original spectrum (blue). The corrected spectrum (red) doesn’t show the escape
peak but it can be seen that the escape peak is overlapped by the Rh-L-line from
the tube.
X-rays are statistically distributed and therefore also the time difference between
the different photons. If this difference is shorter than the filter time a pulse can be
processed before the shaping of the previous pulse is finished. That means, the
filtering of the second pulse starts not on the baseline but at a slightly higher
potential. That would increases the pulse height and results in a count on the high
energy side of the peak. Because this pile up process is statistically distributed a
continuous contribution to the background would be generated.
The probability for this process can be reduced by a fast reduction of the
potential of the signal channel after the filtering process to the basic potential. This
is called baseline restoration. Despite of baseline restoration, overlaps of electrical
pulses are still possible.
2.5 X-ray Detectors 111
Therefore an additional very fast signal channel controls the time between
subsequent pulses. If the time between two pulses is larger than the detection time
of that fast channel the second pulse will be suppressed and not counted. This
process is called ‘pile up rejection’ and reduces significantly the high energetic
background of the spectrum.
Nevertheless if the time between two pulses is even shorter than the detection
time of the fast channel still pulses can be summed up which results in pile-up
peaks at the sum energy of the overlapping peak. The probability that pulses can
overlap depends on the probability of the single pulses in the spectrum i.e. for the
peak intensity and this is high for intensive spectrum lines. The intensity ISum of a
sum peak is given by:
with
IMother1,2 Intensities of mother peaks
P Probability for sum peak as function of shaping time and energy
The probability P for pile up peaks depends on the shaping time and is a complex
function of energy and intensity of both mother peaks. This dependence cause a
peak shift towards lower energies and a peak broadening for high energies of the
sum peak and high count rates. But this dependence is small and can be ignored for
a first approach. Sum peaks influence the spectrum and can be easily misinterpreted
as element lines. Therefore also their identification and correction is necessary.
The identification of sum peaks is possible by changing the excitation intensity.
In that case the mother peaks will change their intensity proportional to the
excitation intensity but the sum peaks according to (2.30) with the product of both
mother intensities. This can be seen in Fig. 2.82 which shows spectra of a Cu-alloy
measured with different count rates—25 kcps (red), 85 kcps (green) and 175 kcps
112 2 Main Components of X-ray Spectrometers
(blue). The spectra are normalized to the Cu intensity which normalize also the
intensities of other alloy elements (Ni, Zn and Pb).
But the sum peaks of the main element lines are Cu-Ka + Cu-Ka at
8.04 + 8.04 = 16.08 keV and Cu-Ka + Cu-Kb at 8.04 + 8.9 = 16.95 keV.
They show distinct intensity differences: with growing count rate of the mother
spectrum the sum peaks grow with the square of the intensity enhancement. In that
way sum peaks can be identified directly due to changing the excitation intensity.
In a normalized spectrum the fluorescence peaks have the same intensity but sum
peaks are changing their intensity. A reduction of excitation intensity can even
avoid pile up peaks.
Another possibility to handle sum peaks is their correction in the measured
spectrum itself. This is demonstrated in Fig. 2.83 that shows the spectrum of a Cu-
alloy.
In this case the sum peaks are calculated according to (2.30) and added to the
spectral background. The example shows not only a high Cu-intensity but also
reasonable intensities of the alloy-elements Ni, Fe and Mn which causes a series of
considerable sum peaks due to the combination of these elements. The image
shows the measured spectrum (red dots), the fitted pure element spectra of all
identified elements in different colors and their pile up peaks together with the
background spectrum (red bars).
The calculation of pile up peaks allows their identification or even their
immediate correction for high count rates and avoids wrong peak identification or
intensity calculation if the sum peaks overlap with element lines.
It is possible that either the excitation radiation or the fluorescence radiation in the
spectrometer are scattered and excites then parts of the spectrometer itself. If this
2.5 X-ray Detectors 113
radiation hits the detector fluorescence lines are detected in the spectrum which are
not from the sample. These effects are mostly weak but can be wrong interpreted.
The following effects are possible:
• Excitation of the Argon in air on the way from the sample to the detector. This
gives a small Ar-K-peak at 2.95 keV if the measurement is performed in air.
This can be used as a feature for the measurement in air.
• Scattering of the fluorescence radiation on the collimator or envelope of the
detector. These are typically small peaks hopefully of elements which are not
very often in the samples. Several spectra displayed here have a small Zr-peak
which is generated by a collimator in front of the detector.
• In case of thin detectors like SDD higher energetic radiation can penetrate the
detector and excite the back part of the detector (support, Peltier-elements). This
radiation then can be detected. This effect nowadays is mostly negligible.
• Scattering of radiation from the sample (scattered tube radiation, fluorescence
radiation) which excites parts of the spectrometer. These are second order
effects and therefore very weak but can give reason for a mis-interpretation as
trace elements.
will be scattered. This process is called ED-XRF (see Table 1.1) and happens for
X-ray spectrometry in general. But for l-XRF spectra diffraction peaks are
observed relatively often. If a large area of a poly-crystalline material is irradiated
as in conventional XRF every crystallite has another orientation and scatters
therefore a different but only small part of the tube spectrum. This enlarges the
spectral background for a wide energy range.
For l-XRF a similar amount of X-rays is concentrated on a small sample area.
If this area is crystalline the number crystallites with different orientations is
reduced and the Bragg-scattered radiation has a higher intensity at only few
energies. Therefore strong peaks can be often detected in l-XRF spectra which
cannot identified as element lines.
The correction of diffraction peaks is difficult—from scatter angle and mea-
sured energy it is possible to determine the d-spacing but because the crystal
orientation is unknown the scattering lattice plane i.e. their Miller indices cannot
be determined.
114 2 Main Components of X-ray Spectrometers
Fig. 2.84 Spectra of Pd with and without filter to avoid diffraction peaks
• Avoiding Bragg-scattering
Bragg-scattering can be avoided by using a primary filter. Then the continuous
part of the tube radiation is filtered. If there is no radiation in the spectrum it
cannot be scattered. In that case the diffraction peaks are clearly reduced or even
disappears but the fluorescence lines are still present is only with slightly
reduced intensity.
This is demonstrated for spectra of pure Pd shown in Fig. 2.84. The blue spectrum
in the background is measured without any filter and shows not only the Pd-K-line
but also the Pd-L-line and a series of large diffraction peaks. These peaks have a
larger FWHM then the Pd-peak and they are not at tabulated energies. These peaks
are reduced if the measurement is performed with filters between tube and sample
as can be seen for the green spectrum that was measured with a 100 lm Al-filter
and even more for the red spectrum that was measured with a 600 lm Al-filter.
But due to the filtering the tube radiation is reduced and therefore the fluorescence
too. Therefore longer measurement times can be necessary and low energetic
fluorescence lines are not efficient excited. For the Pd-K-line this effect is not large
because this line has a high energy for that the influence of the used filters is
negligible.
The Zr-peak in the spectra is from the detector collimator—this peak shows
already an intensity drop for the filtered spectra.
References
15. https://fscimage.fishersci.com/images/D10994*.pdf
16. http://www.bruker.com/fileadmin/user_upload/8-PDF-Docs/X-rayDiffraction_
ElementalAnalysis/Microanalysis_EBSD/Flyers/Fly_xsense_en_rev1_0_lowres.pdf
17. http://www.xos.com/products/X-ray-optics-excitation-systems-x-beam/optics/
doubly-curved-crystals/
18. A. Erko, I. Packe, W. Gudat, N. Abrosimov, A. Firsov, Nucl. Instr. Meth. A467, 358 (2001)
19. VDI/VDE Richtlinie 5575: Blatt 4 (2012), Röntgenoptische Systeme: Total-reflexions- und
Multischicht-Spiegel
20. M. Schuster, H. Göbel, L. Brügemann, D. Bahr, F. Burgäzy, C. Michaelsen, M. Störmer, P.
Ricardo, R. Dietsch, T. Holz, H. Mai, Proc. SPIE 3767, 183 (1999)
21. M. Montel; X-ray microscopy and microradiography, Academic Press. New York. (1957),
177
22. http://www.incoatec.de/
23. http://www.axo-dresden.de/
24. http://www.xenocs.com/
25. http://www.rigaku.com/products/optics
26. A.W. Moore, in Higly Oriented Pyrolytic Graphite. Chemistry and Physics of Carbon, vol
11 (1973)
27. I.G. Grigorieva, A.A. Antonov, X-ray Spectr. 32, 64 (2003)
28. A. Shevelko, A. Antonov, I. Grigorieva, Y. Kasyanov, O. Yakushev, L. Knight, Q. Wang,
Proc. of SPIE 4144, 148 (2000)
29. H. Legall, H. Stiel, V. Arkadiev, A.A. Bjeoumikhov, A. Antonov, I. Grigorieva, Proc. FEL
798 (2006)
30. J. Heckel, R.W. Ryon, in Polarized Beam X-ray Fluorescence Analysis. Handbook of X-ray
Spectrometry, 2nd edn. (Marcel Dekker Inc., New York, Basel, 2002), p. 603
31. H. Legall, H. Stiel, P.V. Nickles, A.A. Bjeoumikhov, N. Langhoff, M. Haschke, V.A.
Arkadiev, R. Wedell, Proc. SPIE 5918, 11 (2005)
32. H. Legall, H. Stiel, M. Schnürer, M. Pagels, B. Kanngiesser, M. Muller, B. Beckhoff, I.
Grigorieva, A. Antonov, V. Arkadiev, A. Bjeoumikhov, J. Appl. Cryst. 42, 572 (2009)
33. V.A. Arkadiev, A.A. Bjeoumikhov, M. Haschke, N. Langhoff, H. Legall, H. Stiel, R. Wedel,
Spectrochim. Acta B 62, 577 (2007)
34. M. Sanchez del Rio, M. Gambaccini, G. Pareschi, A. Taibi, A. Tuffanelli, A. Freund, Proc.
SPIE 3448, 246 (1998)
35. I. Uschmann, U. Nothelle, E. Förster, V. Arkadiev, N. Langhoff, A. Antonov, I. Grigorieva,
R. Steinkopf, A. Gebhardt, Appl. Opt. 44, 5069 (2005)
36. A. Snigirev, V. Kohn, I. Snigireva, B. Lengeler, Nature 384, 49 (1996)
37. B. Lengeler, C. Schroer, J. Tümmler, B. Benner, M. Richwin, A. Snigirev, I. Snigireva, M.
Drakopoulos, J. Synchrotron Rad. 6, 1153 (1999)
38. B. Lengeler, C.G. Schroer, M. Richwin, J. Tümmler, M. Drakopoulos, A. Snigirev, I.
Snigireva, Appl. Phys. Lett. 74, 3924 (1999)
39. C.G. Schroer, O. Kurapova, J. Patommel, P. Boye, J. Feldkamp, B. Lengeler, M.
Burghammer, C. Riekel, L. Vincze, A. van der Hart, M. Küchler, Appl. Phys. Lett. 87,
124103 (2005)
40. B. Lengeler, C.G. Schroer, M. Kuhlmann, B. Benner, T.F. Günzler, O. Kurapova, F.
Zontone, A. Snigirev, I. Snigireva, J. Phys. D Appl. Phys. 38, A218 (2005)
41. C.G. Schroer, B. Lengeler, Phys. Rev. Lett. 94, 054802 (2005)
42. V. Nazmov, E. Reznikova, M. Boerner, J. Mohr, V. Saile, A. Snigirev, I. Snigireva, M.
DiMichiel, M. Drakopoulos, R. Simon, M. Grigoriev, AIP Conf. Proc. 705, 752 (2004)
43. V. Nazmov, E. Reznikova, A. Somogyi, J. Mohr, V. Saile, Proc. SPIE 5539, 235 (2004)
44. VDI/VDE Richtlinie 5575: Blatt 7, (in preparation), Röntgenoptische Systeme: Refraktive
Röntgenlinsen
45. E. Spiller, Adv. X-ray Anal. 42, 297 (2000)
46. G. Hildenbrand, in Grundlagen der Röntgenoptik und Röntgenmikroskopie. Flügge und
Trendelenburg, Ergebnisse der exakten Naturwissenschaften. Band 30, (Springer, Berlin
1958), p. 1
References 117
Abstract This chapter starts with a short overview for the development of
instrumentation for position sensitive X-ray analysis. Then the different possibil-
ities for a position sensitive measurement are introduced and compared in regards
to their analytical results but also to their measurement effort. In a next part the
commercially available instruments are compared in regards to their main appli-
cations and their analytical performance. Finally, the different measurement modes
for l-XRF instruments i.e. single and multiple point measurements as well as one-,
two- or even three-dimensional distribution analysis are introduced and the pos-
sibilities for the evaluation of two-dimensional analysis are discussed based on
measurements of one sample.
The requirement for the investigation of the elemental composition even for small
sample areas already exists from the beginning of elemental analysis. There were
different motivations—the amount of the available sample was small, the ana-
lytical method needed to be non-destructive, the sample was inhomogeneous but
the elemental distribution of interest or the sample was a non-regular shape and
could not be prepared with a flat surface. At the beginning position depending
analyses were performed by preparation of a small piece from the sample and
carried out by wet chemical analysis. But this was not only an extensive and time
consuming process but also a destructive method. Therefore the need for instru-
mental methods was evident.
X-ray fluorescence was already known as a non-destructive method for the
determination of elemental composition—and in the case of excitation by electrons
it was even position sensitive. The deflection of electrons by electric or magnetic
fields is possible and enables the focusing of electrons on small spots. This allowed
the introduction of electron micro analyzers [1, 2]. In these instruments an electron
beam generates characteristic X-rays in the specimen which were analyzed with
material between tube and sample with a small hole. Leading companies at the
beginning of ‘‘Mini’’-XRF for coating testing were Twin City Plating, UPA, CMI
[14] and Helmut Fischer Instruments [15]. Similar instruments were also intro-
duced for the analysis of the content of precious metals in jewelry at first by
Spectro [16] but then also the other companies that prepared coating thickness
testers adapted their instruments for that application. Also in jewelry small and
intricate samples have to be analyzed. Because the qualitative composition of
jewelry often is known the separation of prop-counters was sufficient for this
analytical task.
But for more complex analytical tasks the energy resolution of the detector was
not sufficient for the separation of all element lines and therefore a flexible use for
the analysis of unknown samples was not possible. Further their spot size was also
too large for the analysis of filigree samples in particular for micro-electronics or
for distribution analysis.
Therefore further developments was directed to the use of detectors with higher
energy resolution and the reduction of the analyzed volume were requested. Better
energy resolution could be realized with Si-based detectors—in the first time
strongly cooled Si(Li)-detectors, than after their availability also Si-PIN-diode
detectors and in the last years also SDD’s because they do not require cooling with
liquid nitrogen.
The first application of these detectors was for the analysis of large sample
areas in ED-XRF instruments because the large excited area delivers high fluo-
rescence intensity for the small sensitive area of these detectors [15–18].
But for l-XRF the introduction of X-ray optics was important. Only with optics
it was possible to concentrate sufficient excitation intensity on small sample areas.
The development of X-ray optics was strongly influenced by the requirement of
small spots for synchrotron experiments. But due to the special beam qualities of a
synchrotron radiation—parallel, coherent and polarized with high intensity which
can be focused on spot sizes down in the low nm-range—this equipment is used
for very special applications.
Further the availability of synchrotrons for every day experiments is limited and
therefore the demand for laboratory instruments for spatial resolved experiments
which don’t need special highly sophisticated equipment and long preparation for
experiments was evident, especially for industrial application. Also for forensic
investigations the analysis of small areas of particles or inclusions is of high
interest. Therefore the existing technique for large area analysis was equipped with
collimators for beam size limiting [19, 20]. These instruments had a high reso-
lution Si(Li)-detector but their spot size was still in the range of approx. 1 mm.
The next step was the development of instruments with capillary optics—at first
with mono-capillaries. There were several prototypes developed in universities and
research laboratories [21–29].
The first commercial instrument with capillary optics was the XGT-2000 from
Horiba [30, 31]. This instrument was equipped with a mono-capillary with a
diameter of 10 lm and a low power tube. A Si(Li)-detector was used for the
122 3 Special Requirements for l-XRF
detection. In the instrument the measurement head was separated from the sample
chamber and could be evacuated to improve the light element sensitivity.
But the excitation intensity for mono-capillary instruments is very low.
Therefore with the availability of poly-cap optics also instruments with these
optics came to the market. EDAX with the Eagle [32, 33] as well as Veeco with
the VXR and MXR [34] introduced corresponding instruments.
With that first commercially available instruments the applications was
expanding and also other companies introduced new instruments—for example
Horiba with the XGT-5000 and XGT-7000 [35], EDAX with the Orbis [36] and
Bruker with the M4 Tornado [37, 38].
But at the same time the instrumentation for coating thickness testing had some
further development. The spot size was reduced, partly also by capillary optics (for
example from Shimadzu [39] and Seiko [40] ), by the possibility of measurements
in vacuum (Fischer [41] and Seiko [42] ), by implementations for distribution
analysis etc. At the same time also the quantification models were improved to
give more accurate results with less effort for calibration.
The different available instrument on the market can be distinguished in regards
to the components that determine the analytical performance—for example the
optics or the size of the analyzed area, the type of sample positioning and the
excitation direction, the use of the detection system or the light element sensitivity
given by the measurement medium.
But finally it has to be concluded that the instrument configuration depends on
the requested application and is also determined by the budget of the user.
In the case of a spatially resolved XRF the X-ray spectrum from selected sample
areas needs to be measured. In that case the analyzed sample area needs to be
small. This is possible by using collimators or X-ray optics for limiting the spot
diameter. This is demonstrated in Fig. 3.1 which shows the possibilities for the
excitation of sample areas with different size.
For large sample areas (left image) a limited collimating is sufficient and due to
the high intensity even WDX spectrometers but also EDX detectors can be used.
This type of instruments is typical for the examination of homogeneous samples
which don’t require a spatial resolution. Diameters of exited areas are in the range
of 5–40 mm.
In the case of a smaller collimator (middle image) the same configuration is
used but the excitation beam is limited by the collimator. This reduces the exci-
tation intensity and therefore it is necessary to use energy-dispersive spectrometers
because they have a higher brightness. But still there are limitations in spatial
3.2 Possibilities for Spatial Resolved XRF 123
Fig. 3.1 Possibilities for spatial resolved XRF (a) no spatial resolution (b) spatial resolution with
collimator (c) spatial resolution with capillary optic
resolution because the measured intensity drops down with the square of the spot
diameter which limits the reduction of the collimator size.
This type of instruments is very typical for the analysis of filigree and intricate
samples such as jewelry but much more often also for the analysis of layer sys-
tems. Layer systems are often prepared on final products for decorative reasons or
for special functional purposes. In that case the sample surface can be strongly
structured, for example in case of Printed Circuit Boards (PCB) or it can be
strongly curved as for screws which have a corrosion protection or for electrical
connectors which have a special coating to give a good electrical contact. In these
cases mostly spot size in the range of 0.2–2 mm are sufficient and adapted to the
analytical task.
For even smaller spots X-ray optics for example poly-cap optics (right part of
the image) are used to concentrate the tube radiation to a small sample area with
high fluorescence intensity. This type of instruments is used for the analysis of the
smallest sample areas and of elemental distributions. Typical applications are
forensic or archeological investigations, failure analysis, and elemental distribu-
tions in geological samples or examinations in micro-electronics with very small
structure sizes.
Fig. 3.2 Possibilities for spatial resolved XRF with illumination of a large sample area and
collection of fluorescence radiation (a) from a small sample area and (b) collection from the
complete sample area with an array detector
sizes but increase the probability for penetration of high energetic radiation
through the optics and can generate Halo-effects (see Sect. 2.2.5.3).
It is also possible to irradiate the complete sample and the sample is covered
with a collimator so that the sample is excited only on a small area. This has
recently been realized as an option for WDX spectrometers [44, 45]. For single
point analysis this configuration can be helpful but for distribution analysis the
effort is very high due to the following limitations:
• The excitation intensity of a WDX instrument is optimized for the excitation of
a large sample area. It cannot be concentrated to smaller areas that means the
fluorescence intensity from a selected area drops down with the ratio of the
analyzed area to the complete area.
• WDX spectrometers don’t allow simultaneous measurement of the complete
spectrum. For a distribution analysis therefore the separate acquisition for every
element is necessary which requires a high time effort.
• The optimum focusing conditions for a WDX spectrometer with curved crystal
geometry has different resolution for different points on the sample. That means
the sample needs to be moved into the optimal measurement position which
requires an additional sample movement mechanics and increase the effort for
the WDX-instrument.It is also possible to illuminate a large sample area and
collect the sample radiation with a position sensitive detector which has also an
energy dispersive function of every pixel of the detector (Fig. 3.2—right part).
These detectors are called X-ray color cameras and are available [46–49]. They
have energy resolutions comparable with Silicon-detectors for every pixel which
allows the separation of all element peaks.
For the spatial resolved analysis with this arrangement it is necessary to use
special X-ray optics in front of the detector which collects radiation only from a
certain sample area and transports that to a certain detector area.
The advantage of this configuration is a simultaneous measurement of larger
sample areas and the fact that no moving parts are required which improves the
reliability of the spectrometer. On the other hand some severe disadvantages exist
with this arrangement.
3.2 Possibilities for Spatial Resolved XRF 125
• The count rate capability of these detectors is still limited because every pixel
needs to be read out only after absorption of one absorbed photon. Due to the
statistical distribution of X-rays not every pixel then can have absorbed a photon
i.e. a lot of them will be empty but a few of them will have also more than one
photon i.e. a type of sum peaks will be generated.
• The intensity of excitation radiation is distributed over a large area. As already
discussed, that means high excitation intensity from high power tubes or syn-
chrotron is required.
• A change of the size of the analyzed area is only possible with a change of the
paralyzing optic in front of the detector. This demands additional mechanical
effort for change and adjustment of the optic and the requirement for different
optics is a cost factor too.
• The aspect ratio of the analyzed area is given by the aspect ratio of the detector-
array. If the sample area of interest has another aspect ratio an adjustment of the
analyzed area is not possible. That means large areas with no interesting parts or
several smaller areas which need to be stitched have to be measured.
• This detector is not well adapted for the measurement of small single sample
points. This analytical task happens very often in l-XRF for the identification
and quantification of small single particles, inclusions, and filigree parts or for
the examination of coating systems on small sample areas. Also the analysis of
single points arranged in patterns would be difficult.
• Another question that has to be considered is the homogeneity of the detection
efficiency of these area detectors. Due to differences of the structure of every
detector element or of the signal electronics the detecting efficiency can fluc-
tuate. That would require a correction of these deviations and their monitoring
over time. So far it can be concluded that the excitation of a large area and
collection of fluorescence radiation from a small area is a possibility for position
sensitive analysis but for a fast and flexible instrument this geometry is not
practical.
sample holders are required to fix these samples in the measurement position. For
small particles this can be a thin film which needs to be penetrated both by tube
and sample radiation and limits the sensitivity for light elements.
• Excitation from the top
This excitation direction avoids these positioning problems. It allows the measure-
ment of fluorescence radiation without any absorption by plastic layers. Samples can
be fixed in different positions with modeling clay and then the measurement is
possible in every sample position. Also the analysis of small particles is possible. But
for that excitation direction additional Z-adjustment of the sample is required.
The medium in the measurement chamber is an important parameter both for the
analytical performance for light elements and for the instrument costs.
Mostly these instruments are working in air, for example all instruments with
prop-counters for sure because their light element efficiency is limited by the
thickness of the detector window, but even instruments with semiconductor
detectors have not every time an evacuable sample chamber.
3.3 Instrument Types 129
The elemental range that can be detected for measurements in air depends on
the thickness of the detector window and on the distance sample—detector. For
prop-counter the lightest element that can be detected with sufficient sensitivity is
typically Ca or Ti. For detectors with a thin window the absorption of fluorescence
radiation on the path from the sample to the detector is dominant. For a sample-
detector distance in the range of 20–25 mm the lightest element that can be
analyzed is P or S. For the detection of lighter elements special efforts are nec-
essary (see Sect. 2.5.4) but even then the sensitivity is restricted by low excitation
efficiency and fluorescence yield for the light elements.
For reduction of this absorption there are two possibilities:
• Evacuating the sample chamber
The influence of absorption in air is demonstrated in Sect. 2.3.4 for a working
distance of approx. 20 mm. In this case the stability of the sample chamber must
be sufficient to withstand the air pressure. Because the measurement components
such as tube and detector are mounted on the sample chamber its stability must be
high enough for not changing the measurement position during evacuation. This
increases weight and costs of the instrument significantly. On the other hand a
stable sample chamber is automatically a good radiation shielding which is
required for X-ray instruments at all.
Further for generating a vacuum a pump is required together with the corre-
sponding accessories like power supply, valve, vent etc. But the vacuum does not
need to be of high level—pressures in the low mbar range are sufficient which
reduce the pumping time. If the pump is controlled it is even possible to set the
vacuum to a given level which can reduce the sample stress by evacuation—for
example for drying the sample.
A special possibility to reduce the absorption of low energetic radiation is the
permanent evacuation of the part of the sample chamber that contains the tube and
the detector. In that case no evacuation time is required and measurement can start
immediately after sample positioning. In this case tube and detector are separated
from the sample chamber by a window which is mostly a thin plastic film. This
window but also the air between sample and window absorbs still both excitation
and sample radiation. In both cases typically Na is the lightest element that can be
analyzed.
• He-flush in the radiation path
The other possibility to reduce the absorption is the use of He as a measurement
medium. The complete floating of the sample chamber with He is not possible due
to the high amount of required He. Therefore a He-flush is used mostly i.e. He is
pumped into the path between detector and sample and thus reducing the
absorption of fluorescence radiation only. The absorption of tube radiation is not so
important because also high energetic parts of the spectrum which are not absorbed
can excite the characteristic radiation.
The demand for this solution is that the He-flux must be relatively high to remove
all air from the radiation path but then the flux is not laminar which generates
fluctuations that also influence the stability of the measurement. Therefore a He-flush
130 3 Special Requirements for l-XRF
Typically energy dispersive detectors are used for l-XRF instruments. They can be
arranged close to the sample and have therefore a good brightness because they
can collect a high amount of fluorescence radiation. This good brightness and the
3.3 Instrument Types 131
In between there is a wide variety of bench top instruments for l-XRF in the
market. An overview for the situation in 2013 is given in Table 3.1 which sum-
marizes the discussed parameters and presents for the corresponding configura-
tions the actual instruments of main manufacturers. A distinct separation between
the different configurations is difficult because the instrument variety is very large
and there are several crossovers between the different types. The table summarizes
therefore only the most often used configurations.
It has to be mentioned that in particular for the ‘large scale’ instruments but
partly also for ‘high level’ instruments there are several systems available
developed and manufactured as prototypes in universities or research institutes.
These instruments are mostly designed for special applications and only for
internal use. That means, for example, the control and evaluation software is
dedicated for the application but not user friendly or flexible for multi-purpose
applications. Therefore and because mostly not every parameter of these instru-
ments is published they are not considered in the table.
Table 3.2 gives an overview of the preferred applications for the different
instrument types. It has to be mentioned that the field of l-XRF is very dynamic
and new applications are developed continuously. So far it has to be mentioned
that this table can only summarize the most important applications for the actual
time at 2013. Further it has to be mentioned that some applications can be solved
also with different instrument configurations but not all of them can offer the best
analytical performance. This is also labeled in the overview.
Table 3.1 Listing of different l-XRF laboratory instruments (status 2013)
132
Table 3.2 Main applications for the different types of l-XRF instruments
Instrument type Low level Low level High level Large scale WDX
prop-counter PIN, SDD
Jewelry production • •• •• •
Jewelry unknown oo • •• oo •
Coating automotive • •• •• oo oo
Coating—lE oo • • oo oo
Restricted elements Limited • • oo o
Failure analysis oo • • oo oo
Forensic oo • • oo oo
Research oo • • • oo
Particle oo oo • oo •
Element distribution oo oo • • oo
High throughout oo oo • oo oo
Archeometry oo oo • (small samples) • oo
Bore cores oo oo • (small ones) • oo
Legend: oo not usable, o limited use, • usable, •• over performed
For l-XRF instruments several measurement modes are available for the different
analytical tasks. This includes the measurement on single points and on different
points as well as distribution analysis both on lines or areas. But for these modes
not only the control of sample positioning is required but also specific evaluation
procedures for handling the comprehensive data material.
This measurement mode analyzes small single sample areas. This is required for
small particles, for inclusions in larger sample areas but also for special points on
larger samples. The measurement time for single point analysis depends on the
analytical task. For an overview analysis it can be short—in the range of few
seconds, for a more precise analysis or for trace analysis they should be longer—in
the range of 30–500 s in dependence of the weight fraction of the analyte and
precision requirements.
This mode is used for the identification of special elements in small areas but
also for the quantification both of homogeneous bulk material and of coating
systems. Single point measurements are very often used. Several applications are
described in Sect. 7.1.
134 3 Special Requirements for l-XRF
If not only the composition of a single point but of several ones is of interest the
efficiency of the measurement can be increased by a multi-point mode. In that case
several measurement points can be defined and saved as a stage program. Then the
instrument runs to all points according to this program sequentially and performs
the measurement automatically. All spectra are saved and can be evaluated with
the appropriate model—directly after the measurement or as post processing. The
single analysis points can be arranged as single points as well as points on a line or
in a grid. Typically in this mode every measured spectrum will be saved sepa-
rately. Measurement time per pixel is similar as for single points.
The efficiency can be even enhanced for repeated measurements of similar
samples by saving the stage program. In that case it is necessary to have the
possibility for a stage alignment because the samples cannot every time easily
positioned in exactly the same position on the stage. By using 2 or 3 reference
points which have a fixed relation to the measurement pattern or are even mea-
surement points it is possible by a comparison of the stage positions of these
reference points to adjust the measurement pattern to the actual stage position by
coordinate transformation.
Multi-point measurements are used for very different purposes—for example
for the analysis of coating thicknesses on contact layers of printed circuit boards,
for the determination of weight fraction profiles or for high throughput screening.
For an efficient data evaluation different methods can be used—display of
intensities or quantification results, ratios of intensities of selected elements or
even the use of chemometric methods for a compression of the huge amount of
data. More detailed examples are given in Sect. 7.4.3.
The special feature of l-XRF is the analysis of small sample areas which allows
the examination of inhomogeneity in the materials. But this can be also a disad-
vantage in the case of small inhomogeneities in the sample. A large analyzed area
would automatically average these fluctuations in composirion. For l-XRF, this
averaging is possible too if the sample is measured in different positions for a short
but identical time. The sum of all these single spectra represents the average
composition of the material and also the influence of the sample surface. For the
distribution of the single measurement points it should be possible to make this
periodically for example in a grind or randomly within the area that has to be
analyzed.
This measurement mode can be required for the analysis of powders, of geo-
logical samples or pressed tablets as demonstrated in Sect. 7.2.1.
3.4 Typical Measurement Modes for l-XRF 135
3.4.4 Linescan
3.4.5 Mapping
But highly focused X-ray tubes in connection with poly-capillary optics offer the
required high excitation intensity.
Two-dimensional analysis or mapping can give interesting insights to the
sample composition. This can be used for the investigation of geological samples
to understand geological processes, for the examination of electronic components
for the existence of restricted elements, for failure analysis in industrial products,
for analysis of art objects and a lot of more applications. Several examples for
mapping analysis are discussed in detail in Sect. 7.4.
D ¼ f ðx; y; E; IÞ ð3:1Þ
This relation is presented in Fig. 3.4 for a printed circuit board. For every pixel
a separate spectrum is collected and completely saved.
This method can generate large data files—for large mappings with more than a
million pixels into the range of Gigabyte’s but it allows a wide range of data
presentation, evaluation and post-processing.
• Presentation of multi-dimensional data sets
For the presentation of multidimensional data typically two dimensions are
available, in case of using different colors and varying brightness it is even
possible to visualize more dimensions. The following possibilities are mostly
used:
3.4 Typical Measurement Modes for l-XRF 137
I ¼ f ð EÞ ð3:2Þ
with: x, y fixed
This presentation allows for example the quantification for special sample areas (if
the spectrum has in case of sufficient intensity) and the comparison of the com-
position in different sample areas.
– Presentation of linear intensity distributions along arbitrary lines in the sample
(line-scan) i.e.
I ¼ f ðx; y; Ei Þ ð3:3Þ
with:
y mx + n, m, n as coefficients of the equation of the line
Ei energy
This presentation allows the display of intensity changes along a line. With help of
a color coding linescans are possible also can be displayed for several elements.
Because the intensity of single pixels can be very small it can be helpful to bin
several pixels perpendicular to the line. This increase the intensity per point in the
linescan and makes the distributions smoother. Element profiles can be also pre-
sented as weight fraction profiles after quantification.
– Presentation of the intensity distribution of single elements, of several elements
or of a selected energy range i.e.
B ¼ f ðx; y; E; IÞ ð3:4Þ
with:
B brightness of a selected color
x, y coordinates within the sample
E selected energy for the distribution
I intensity in a certain energy range of the spectrum
Fig. 3.5 Optical image of a meteorite (left) and distribution of its main elements with two
objects in the lower left corner (right)
is helpful to use a pixel binning which reduce their number and increase the
intensity in the pixel i.e. the calculation time will be reduced and the accuracy of
quant results improved.
Element distributions can give an impression about the composition for a larger
sample area but an objective assessment from these pictures is difficult. Further it
has to be considered that the amount of data in a mapping can be tremendous.
Therefore it is interesting and necessary to use additional procedures for a data
reduction, for the extraction of relevant data from mappings and for the objecti-
vation of distribution measurements. Different possibilities are available for that
purpose—for example chemometric procedures.
A few of these possibilities for data presentation and data processing are
demonstrated in the following with a data set measured on a meteorite (Campo del
Cielo).
Fig. 3.7 Spectra of two objects of different size in Fig. 3.5 (square root intensity scaling)
Fig. 3.8 Sum spectrum and maximum pixel spectrum of the meteorite of Fig. 3.5 (square root
intensity scaling)
spectrum and the MSP of the meteorite are displayed in Fig. 3.8. The maximum
pixel spectrum shows peaks of Cu and Zn which cannot be detected in the sum
spectrum. For Cr the intensity in the MSP is significantly higher than in the sum
spectrum but this element could also identified in the sum spectrum and therefore
its distribution was already calculated (Fig. 3.6). The distribution of Cr, Cu and Zn
is shown in Fig. 3.9.
It shows that Cr is concentrated in the core of the meteorite in certain grains that
have a high weight fraction of Cr. But the area of these grains is relatively small
compared to the complete sample area. Therefore in the MSP the Cr-peak has a
higher intensity than in the sum spectrum which averages the contribution of all
pixels. Cu is only concentrated in the right upper corner of the meteorite core. This
area is also small in comparison to the complete sample area and therefore the Cu
peak is not detected in the sum spectrum. Even more significant is the situation for
Zn. This element is concentrated in an area on top of the sample which includes
142 3 Special Requirements for l-XRF
(a) (b)
Fig. 3.10 Defined line in the mapping (a) and the calculated intensity distributions (b)
only approx. 90 pixels i.e. it corresponds to only 0.015 % of the complete area.
Therefore Cu and Zn are completely overlapped by the spectral background in the
sum spectrum but can be detected with help of the MSP.
• Linescan calculation
A further possibility of data evaluation is the calculation of linear distribution for
lines in every direction of the mapping. In that case all pixels on a defined line are
evaluated for their element intensities. This is demonstrated in Fig. 3.10. It shows
in the left part a line over the core of the meteorite and two corresponding
intensities distributions of selected elements in the right part.
Also in this case the determination of the intensity of only one pixel on the line
can have a very limited statistics i.e. intensity fluctuations are significant. This is
demonstrated in the upper right part. Here it is also possible to broaden this line. In
that case the intensities of all pixels perpendicular to the line axis are summed up.
This increases the intensity markedly and improves the statistics as can be seen in
the lower picture of the right part. The distributions in the upper and lower part
show distinct differences because the statistical intensity fluctuations of the
‘thicker’ line are reduced.
The presentation of analytical data as images is very popular because it can give a
lot of information about the analyzed material. But there are also requirements for
quantifications of this type of distributions. Often it is not sufficient to have only
the image but also quantitative values which characterized the distribution. This
means not necessarily only a quantification of every pixel spectrum. Very often the
extraction of content characterizing parameters can be much more helpful.
3.4 Typical Measurement Modes for l-XRF 143
The spectrum of a single pixel contains only very few counts in particular if the
measurement time per pixel is very short i.e. in the ms- or even sub-ms-range.
Then single channels often does not have even a single count. But also for that low
number of counts it is possible to identify very small intensity differences. This
will be demonstrated in Fig. 3.11 which shows again the meteorite, now only the
P-distribution together with 3 measurement areas—special objects-two single
points in a P-rich and P-depleted area (point 1 and 3, respectively) and one area
with approx. 800 pixels in the P-rich region. The corresponding spectra in the
energy range of the P-K-line are displayed in the right part of Fig. 3.11 (pixel
measurement time: 6 ms).
The both single pixel spectra (bars) show very low several channels does not
have even a single count. There is a small intensity for the P-rich point (blue) but
the differences between the two single point spectra (red and blue bars) are
marginal. Nevertheless, the areas of low and high P-content can be clearly dis-
tinguished in the distribution. That there is a remarkable P-content can be seen in
the spectrum of the rectangular with approx. 800 pixels (green bars) that have a
good statistics because the total measurement time is approx. 800 9 6 ms = 4.8 s
and shows a considerably P-peak.
Despite the low single pixel intensities the P-distribution of the structure of the
meteorite can be very well detected. This is surprisingly because the standard
deviation of a measured intensity I is HI for a Gaussian distribution. For small
intensities I the relative error 1/HI is large and can cover small intensity changes
(see also Fig. 2.54).
This behavior should be considered more in detail. The spectra of the area with
800 pixels shows a distinct peak because the measurement time is prolonged and
144 3 Special Requirements for l-XRF
Fig. 3.11 P-distribution with single measurement points in the P-rich and P-depleted area (left)
and the corresponding spectra of the single pixels: object 1: blue bars, object 2: red bars and the
area object 3: green dots (right)
the time dependence of signal and noise is different—the signal is growing with
time i.e. Ielem * t but the noise is growing with square root of time i.e. IBG * Ht.
That means the signal-noise ratio is improving with the square root of time
according to:
ffi pffi
IElem= t pffi ¼ t ð3:5Þ
IBG t
Because the binning of n pixels prolongs the measurement time by the factor of n
binning also improves the signal-noise ratio according to:
ffi
IElem= n tPixel pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffinffiffiffiffiffitffiffiPixel
ffiffiffiffiffiffi
ð3:6Þ
IBG n tPixel
The standard deviation of the Poisson distribution with a mean count rate m is
given by Hm. The probabilities for the occupancy of a channel with different
counts (0, 1, 3 and 6) are displayed in the left part of Fig. 3.12 in dependence of
the count rate m. These count rates are typical for measurement times in the low
ms- or sub-ms level per pixel.
For high count numbers the Poisson-statistics converges to the Gaussian sta-
tistics. What are the differences between both distributions for low count numbers?
3.4 Typical Measurement Modes for l-XRF 145
Fig. 3.12 Poisson probability for different counts per channel N in dependence of the mean
count rate m (left); difference of RMS values between Gauss- and Poisson-distribution (right)
If the expected count number is N and the standard deviation of the distribution is
as for the Poisson distribution HN the Gaussian probability is given by:
. ðN mÞ2
pffiffiffiffiffiffiffiffiffiffiffiffi
PðN Þ ¼ 1 2p N expð Þ ð3:8Þ
2N
The differences between the two distributions are very small and only relevant
for very small count rates as can be seen in the right part of Fig. 3.12. This shows
the root-mean-square differences (RMS) between both distributions if they are
summed over N for a fixed m. This shows only for less than a single count per
channel considerably differences between both distributions.
In a spectrum different Poisson distributed sources contributes to the counts in
every channel—for example the characteristic radiation and the scattered source
radiation. Therefore it is interesting to consider the behavior of both their sums and
differences. In the case of a single pixel that can be the
• Sum of counts in one channel from different sources, for example an element
line and the scattered background
For two different and independent count rates m1 for the element line and m2 for
the background the Poisson-probabilities P(N, m1) and P(N, m2) can be added
for the same occupancy probability N, i.e. P(N, m1) + P(N, m2) = P(N,
m1 + m2).
That means the probability for k counts in one channel is described by the
Poisson-distribution for the sum of the independent count rates m1 and m2.
• Difference between the counts of element lines and the background in one
channel
This difference of the counts from different sources in one channel is typically
used for the determination of netto counts of the element line. But in opposite to
the sum differences cannot be calculated as discussed in detail in [65]. The
reason is that for weak signals the number of counts could be negative which is
physically not possible.
146 3 Special Requirements for l-XRF
Due to the low number of counts per channel the addition of the counts of different
channels of one pixel and/or the binning of pixels is often used for an improvement
of the peak-to-background ratio or for the reduction of counting errors. The
understanding of the effects of these processes is important for the interpretation of
measured data. The following situations have to be considered:
• Determination of intensities in one pixel by addition of the content of several
channels of the spectrum
The counts of an element line are not concentrated in only a single but in several
channels due to the energy spread of the line by noise contributions of the
counting electronics. The sum of the count numbers of all spectrum channels of
the element line is correlated to the emitted counts of the element. This is
Poisson-distributed and given by the Sherman-equation (see Sect. 4.3.2.1).
Therefore the counts of all channels of an element line are Poisson distributed
(for higher emission rates the Gaussian distribution is valid too). The distribution
between the channels is given by the detector response function. That means for
the determination of element intensities the content of the energy range across
the element lines should be summed up. This increases the count number i.e.
reduces the statistical error and represents the element intensity more exactly.
Background intensities should be determined in the same energy range as the
peak intensity. Because this is not possible the background behavior can be
extrapolated from a line free energy range close to the element line.
• Sum of the counts from binned pixels
The counts in a single pixel can be very low and as already mentioned not
sufficient for a reasonable spectrum evaluation. Therefore a pixel binning is
often performed. Because the counts of these pixels are independent their sum is
also Poisson distributed. Their count rate m is the sum for the binned pixels i.e.
the binning increases the count rate according to:
X
mbinned ¼ mi ð3:9Þ
n
That means the count rate is growing with the number of binned pixels. This is
equivalent to an increase of the measurement time. In that way an improvement
of the count statistics as well as the peak-to-background ratio is possible. But
this pixel binning reduces the spatial resolution of the distribution.
• Difference of counts in two (neighboring) pixels for an element line
This situation is interesting for the detection of small intensity changes in dis-
tributions as for example the discussed P-distribution of the meteorite
(Fig. 3.11). Two statistical distributed signals can be definitely distinguished if
their difference is larger than the standard deviation of their fluctuation. For a
Poisson distribution the standard deviation of n counts is given by Hn. Then the
count numbers in one spectral channel or in a region of interest (for example the
energy range of an element line) should have at least the following difference for
a distinct distinguishing:
3.4 Typical Measurement Modes for l-XRF 147
pffiffiffiffiffiffiffiffiffiffiffi
npixel2 npixel1 ¼ k npixel1 ð3:10Þ
with:
npixel number of counts in a channel or region of interest in the pixel
k factor of confidence
This relation is displayed in Fig. 3.13. It shows that for high count numbers
which have a large absolute standard deviation the differences of measured
counts needs to be also large for a clearly separation of regions with different
intensity. This is valid even if their relative standard deviation is small. On the
other hand the count differences can be very small for low intensities. For count
numbers per pixel with n \ 1 even a single count is already significant i.e. these
small differences are valid and represent different concentrations in the pre-
sentation of distributions. That means distribution images have a high sensitivity
even for very low intensities (or concentrations) but only in case of displaying
the complete distribution and not for the comparison of single pixels.
This effect obtains a further amplification because typically a normalization of
the intensity signals will be performed. That means for the coding of intensities
the brightness of the pixel is normalized to the maximum intensity in this energy
range of all pixels i.e. it will be is displayed with high brightness even if there
are only very few counts.
Fig. 3.14 Distribution of S and Cr in the meteorite core with (a) and without (b) smoothing
• Image processing
Image processing is possible for example by smoothing or averaging. In that case
the number of pixels will not be changed. The effect of the convolution depends on
the used convolution kernel and is an improved S/N-ratio. But due to the procedure
also information and spatial resolution can be lost. This is demonstrated in
Fig. 3.14 for the core of the meteorite for the elements S and Cr.
• Pixel binning
Another possibility to improve the resolution of structures is the binning of pixels.
This binning is possible only by the summing up every time the same number of
pixels or by an adaptive binning. For the first case the spatial resolution is reduced
and the S/N-ratio is improved by the same ratio according to (2.26). The easiest
possibility is he binning of 2 9 2, 3 9 3 etc. pixels. In that case the S/N-ratio for
the larger pixels is increased but the spatial resolution is decreased and all new
pixels have different S/N-ratios.
S/N-ratios in the same range can be generated by an adaptive binning. In that
case a preselected S/N-ratio is used to perform the binning i.e. the size of the final
objects is changing in dependence of the intensity and structure changes. This can
be different for different elements for a mapping with several elements. As a result
the spatial resolution then depends on the S/N-ratio i.e. on the intensity
fluctuations.
For the adaptive binning different approaches are possible. For example the
Quadtree method [66] starts with the complete mapping image as one pixel. This
will be divided in four quadrants. The S/N-ratio of every quadrant is checked. If
this is larger than a preselected value the procedure has to be repeated if not the
pixel will not be divided again. This procedure has to be repeated for every branch-
pixel till the S/N-ratio is less than the preselected value or further splitting is not
possible because it is a single pixel. The method is called Quadtree method
because the pixels are squares and their splitting is like a tree in dependence of the
structure.
3.4 Typical Measurement Modes for l-XRF 149
But with this approach the binning is relatively coarse and therefore other
methods can be used with an adaptive binning for example the Voronoi tessellation
[67, 68] which calculates bins which covers the complete area with objects that are
compact i.e. as close as possible like a circle. For that tessellation the size of the
bins has to be adapted to the preselected S/N-ratio. If the S/N-ratio is reached the
binned area should not grow by a further binning step. This guarantees for all
pixels a similar S/N-ratio and for a given S/N-ratio the best spatial resolution.
But these binning methods have the disadvantage that the image structure of the
distribution is destroyed. They allow the evaluation of every binned pixel with a
similar sensitivity and analytical performance but the possibility to display the
distribution is lost. Further it has to be considered that the binning depends on the
distribution of elements i.e. the ‘binning structure’ can different for the different
elements.
Procedures for these binning procedures are typically not implemented in the
software package of instruments.
• Particle analysis
A special version of phase analysis is the particle analysis. In this case the ana-
lyzed sample contains or consists of small particles or has inclusions of same or
different composition. Phase analysis then identifies objects with a similar com-
position based on the elemental distribution. Particle analysis makes a classifica-
tion according their shape i.e. for parameters like size of their area, orientation,
different diameters, position etc. in dependence of the composition of the particles.
These results then can be summarized for the number of particles or their prob-
ability of presence according with respect to the different object parameters. This
is demonstrated for the core of the meteorite in Fig. 3.16. The left part shows the
Cr-distribution in the core of the meteorite together with the area which was used
for particle analysis. In the middle the identified and classified particles are dis-
played. The classification is performed according to the size of the particles; the
classes are displayed in different color. On the right side this result is presented in a
diagram which shows the number of particles in the different size classes.
Fig. 3.16 Cr-distribution of the meteorite from Fig. 3.5 inscribed the area for particle analysis
(left); identified and classified inclusions (coloring according size) (middle); number of particles
in every size class (right)
• Intensity distributions
The simplest possibility is the calculation of the intensity histogram for an element
i.e. the frequency of the presence of the different element intensities is calculated
for every pixel. The left part of Fig. 3.17 shows the distribution for Fe for the
intensity distribution of Fig. 3.5 and the right part shows the frequency of the
presence of intensity ratios between ni and Fe in the different pixels.
It can be seen that there are two different Fe-phases, that have different
intensities i.e. also different compositions. These two phases can be also detected
in the distribution (of intensity ratios in the right part of Fig. 3.17). This distri-
bution shows that the two Fe-clusters correspond to different Ni-weight fractions.
In comparison to the phase distribution in Fig. 3.16 here only two phases can be
detected which represents the core and the subsumed shells of the meteorite
because the Fe-content of the both shells are similar.
• Intensity ratios
Another possibility is the consideration of intensity ratios between different ele-
ments for every pixel. This allows the normalization of element distributions in
correlation to major components or can be used for the examination of changes of
weight fractions due to diffusion, leaching effects, enrichment procedures etc.
Examples for the determination of diffusion and leaching processes in the envi-
ronment are discussed in [69–72]. The determination of intensity ratios allows in
this case the determination of intensity profiles independent of further sample
components and sample structure.
Another example for the use of intensity ratios is the high throughput screening
for the development of pharmaceuticals (see 7.2.3.2). In this case intensity ratios
are used for the investigation of the reactivity between drugs and proteins.
• Multiple variate evaluation
Very often used methods for data compression are multiple variant evaluations.
These offer widespread possibilities for data analysis. There are various methods
for the data evaluation—for example for the comparison of spectra of different
materials and their classification without classical spectrum evaluation, but only
152 3 Special Requirements for l-XRF
Fig. 3.17 Histogram of Fe-distribution (left); probability for the frequency of intensity ratios
between Fe and Ni (right)
from their shape. These methods can also be used for the evaluation of elemental
distributions.
The easiest method for a data reduction is the Cluster analysis. In that case all
(element) intensities Ij of the spectra are written in a multidimensional space, every
element on an orthogonal axis and then the differences Dmn in that space between
every data point are determined according to:
pffi X
Dmn ¼ ½2 n6¼m ðIj;m Ij;n Þ2 ð3:11Þ
The two data points with the smallest difference are merged into one cluster.
Then this process is repeated iteratively for the remaining points. The criterion for
stopping this procedure can be a preselected number of clusters or also a minimum
distance between the remained data points that should not be undercut. An
example for cluster analysis with different stopping criteria is shown in Fig. 3.18.
Cluster analysis is possible from the original data i.e. channel contents but also
from evaluated data like fitted element intensities or weight fractions which can
reduce the influence of the spectral background. Due to the influence of noise
cluster analysis gives not in any case unambiguous and satisfying results.
Principle Component Analysis (PCA) is another often used method for
multivariate data evaluation. PCA is a mathematical procedure that converts sets
of correlated variables for example the intensities of single channels or elements or
also weight fractions in every pixel into a set of linearly uncorrelated variables—
the principle components. This is possible for example by a rotation of the
coordinate system. The number of principle components typically is smaller than
that of the original variables. Principle components are linear combinations of
original variables and defined according the size of their original variance.
3.4 Typical Measurement Modes for l-XRF 153
Fig. 3.18 Cluster analysis of same distribution but with different sensitivity
The variable with the largest variance is the principle component 1. Every next
component with a smaller variance is arranged orthogonal to the former one. For
normal distributed original data sets the principle components are independent of
each other. This procedure is demonstrated in Fig. 3.19.
With a PCA the number of variables will be reduced and data can be easier
interpreted. For example elemental distributions can be replaced by distributions
of compounds which are composed of different elements. That means, if the
principle components are composed of several elements with fixed ratios of mass
fractions i.e. they are compounds or they are characterized by special parts of the
spectrum (for example scattered radiation), the number of variables will be
reduced and allows the identification of different materials [73].
Also pattern recognition can be used for data reduction. In that case data are
classified for similar geometrical properties which can be used for a comparison
and identification of different objects.
These multivariate methods offer a wide range of possibilities for example for
structure identification, for identification of extreme compositions, for unambig-
uous phase identification etc. [74, 75].
The use of multivariate methods depends in any case of the distinct application
and has to be adapted to the current measurement conditions and results as well as
to the required analytical results.
154 3 Special Requirements for l-XRF
References
38. http://www.bruker.com/products/x-ray-diffraction-and-elemental-analysis/
x-ray-spectrometry/m4-tornado/overview.html
39. http://www.ssi.shimadzu.com/products/literature/XRAY/uEDX_Series.pdf
40. http://www.seiko-instruments.de/99-0-sft-9500.html
41. http://www.helmut-fischer.com/independent/31/XRAY_Overview.asp
42. http://www.seiko-instruments.de/139-0-sea-6000-vx.html
43. A.A. Bjeoumikhov, P. Chevalier et al., Proc. SPIE 3444, 598 (1998)
44. http://www.rigaku.com/products/xrf/primus
45. https://fscimage.thermoscientific.com/images/D01781*.pdf
46. N. Meidinger, H. Bräuninger, U. Briel, R. Hartmann, G. Hartner, P. Holl, J. Kemmer, E.
Kendziorra, N. Krause, G. Lutz, E. Pfeffermann, M. Popp, C. Reppin, R. Richter, H. Soltau,
D. Stötter, L. Strüder, J. Trümper, C. von Zanthier, SPIE 3765, 192 (1999)
47. L. Strüder, H. Bräuninger, M. Meier, P. Predehl, C. Reppin, M. Sterzik, J. Trümper, P.
Cattaneo, D. Hauff, G. Lutz, K.F. Schuster, A. Schwarz, E. Kendziorra, A. Staubert, E. Gatti,
A. Longoni, M. Sampietro, V. Radeka, P. Rehak, S. Rescia, P.F. Manfredi, W. Buttler, P.
Holl, J. Kemmer, U. Prechtel, T. Ziemann, NIM A288, 227 (1990)
48. O. Scharf, S. Ihle, I. Ordavo, V. Arkadiev, A. Bjeoumikhov, S. Bjeoumikhova, G. Buzanich,
R. Gubzhokov et al., Anal. Chem. 83–7, 2532 (2011)
49. I. Ordavoa, S. Ihlea, V. Arkadiev, O. Scharf, H. Soltau, A. Bjeoumikhov, S. Bjeoumikhova,
G. Buzanich, R. Gubzhokov, A. Günther, R. Hartmann, P. Holl, N. Kimmel, M. Kühbacher,
M. Lang, N. Langhoff, A. Liebel, M. Radtke, U. Reinholz, H. Riesemeier, G. Schallerg, F.
Schopperg, L. Strüder, C. Thamm, R. Wedell, Nucl Instr. Meth. 654–1, 250 (2011)
50. A. Erko, F. Schäfers, A. Firsov, W.B. Peatman, W. Eberhardt, R. Signorato, Spectrochim.
Acta B 59, 1543 (2004)
51. K. Janssens, K. Proost, G. Falkenberg, Spectrochim. Acta B59, 1637 (2004)
52. L. Vincze, B. Vekemans, F.E. Brenker, G. Falkenberg, K. Rickers, A. Somogyi, M. Kersten,
F. Adams, Anal. Chem. 76, 6786 (2004)
53. B. De Samber, G. Silversmit, K. De Schamphelaere, R. Evens, T. Schoonjans, B. Vekemans,
C. Janssen, B. Masschaele, L. Van Hoorebeke, I. Szaloki, F. Vanhaecke, K. Rickers, G.
Falkenberg, L. Vincze, JAAS 25, 544 (2010)
54. K. Proost, K.H. Janssens, L. Vincze, G. Falkenberg, N. Gao; P. Bly, Hasylab Jahresbericht
2003, available online
55. B. Kanngießer, W. Malzer, I. Reiche, Nucl. Instr. Meth. B211–212, 259 (2003)
56. G.J. Havrilla, T. Miller, Powder Diffr. 19–2, 119 (2004)
57. B. Kanngießer, W. Malzer, A. Fuentes Rodriguez, I. Reiche, Spectrochim Acta. B 60, 41
(2005)
58. T. Sun, Z. Liu, X. Ding, Nucl. Instr. Meth. B262, 153 (2007)
59. K. Tsuji, K. Nakano, X. Ding, Spectroch. Acta B62, 549 (2007)
60. T. Sun, M. Zhang, X. Ding, Z. Liu, X. Lina, H. Liua, J. Appl. Cryst. 40, 1169 (2007)
61. X. Lin, Z. Wang, T. Sun, Q. Pan, X. Ding, Nucl. Instr. Meth. B266, 2638 (2008)
62. K. Tsuji, K. Nakano, X-Ray Spectr. 36, 145 (2007)
63. R.B. Mott, J.J. Friel, J. Microsc. 193, 2 (1999)
64. P.G. Kotula, M.R. Keenan, J.R. Michael, Microsc. Microanal. 9, 1 (2003)
65. W. Thompson, Comput. Sci. Eng. 2, 84 (2001)
66. H. Samet, ACM computing Survey 16, 187 (1984)
67. M. Cappellari, Y. Copin, Mon. Not. R. Astron. Soc. 342, 345 (2003)
68. S. Diehl, T.S. Statler, Mon. Not. R. Astron. Soc. 368, 497 (2006)
69. M. Schumacher, I. Christl, R.D. Vogt, K. Barmettler, C. Jacobsen, R. Kretzschmar,
Biogeochemistry 80, 263 (2006)
70. A. Voegelin, F.A. Weber, R. Kretzschmar, Geochim. Cosmochim. Acta 71, 5804 (2007)
71. P. Langner, C. Mikutta, E. Suess, M.A. Marcus, R. Kretzschmar, Spatial distribution and
speciation of arsenic in peat studied with microfocused X-ray fluorescence spectrometry and
x-ray absorption spectroscopy environ. Sci. Technol., Article ASAP
156 3 Special Requirements for l-XRF
72. D. Vantelon, A. Lanzirotti, A.C. Scheinost, R. Kretzschmar, Environ. Sci. Technol. 39, 4808
(2005)
73. M.A. Rodriguez, P.G. Kotula, J.M. Griego, J.E. Heath, S.J. Bauer, D.E. Wesolowski, Powder
Diffr. 27–02, 108 (2012)
74. M.R. Keenan, P.G. Kotula, Surf. Int. Anal. 36, 203 (2004)
75. P.G. Kotula, M.R. Keenan, J.R. Michael, Microsc. Microanal. 9–1, 1 (2003)
76. J.S.B. Reed, Electron Microprobe Analysis (Cambridge University Press, Cambridge, 1975)
77. K.F.J. Heinrich, Electron Beam X-Ray Microanalysis (Van Nostrand Reinhold Co, New
York, 1981)
78. W. J. Thompson, Comput. Sci. Eng. 3, 78 (2001)
Chapter 4
Quantification
4.1 Introduction
• The excitation probability for these elements is very low (see also Sect. 6.1.2,
Fig. 6.4). That means the number of atoms with low atomic number that can be
excited by an incident photon which then can emit fluorescence radiation is
rather low.
• The fluorescence yield depends on atomic number. For atoms with low atomic
number the fluorescence yield is very small (see Sect. 1.2.2). That means only a
very small part of atoms that are excited will emit X-rays as result of filling inner
vacancies in the electron shell.
• The low energetic radiation will be strongly absorbed, in the sample itself which
reduce the information depth i.e. also the number of atoms that can contribute to
the signal, on the way from the sample to the spectrometer—but this can be
reduced by measurement in vacuum—and in the spectrometer itself for example
by absorption in the detector window.
A special benefit of X-ray fluorescence spectroscopy is the very high accuracy.
This is a result of low statistical errors. High counting rates and if necessary also
long measurement times with highly stable instruments allow the exact determi-
nation of element intensities.
But in the other hand X-ray fluorescence has a strong matrix interaction. This is
demonstrated in Fig. 4.1. The incident beam from the excitation source will be
absorbed on the way to the interaction with an atom. The excited atom then emits
fluorescence radiation. This fluorescence radiation will be emitted isotropically in
every direction. Radiation into the direction of the spectrometer (detector) can be
collected directly but it can also be absorbed on the way to the sample surface by
the matrix again. This absorption can also generate vacancies in atoms if their
absorption edges have lower energies than the fluorescence radiation. In that case
the intensity of the primary fluorescence is reduced and the intensity of the element
with the lower fluorescence energy will be increased. This interaction influences
the measured intensities of all element combinations in the sample. Therefore
calibration curves are not linear and do not depend only from the weight fraction
of the elements itself but also from the weight fractions of all other elements in the
sample.
4.1 Introduction 159
Figure 4.2 shows an example for the calibration curve of Fe in the binary
systems FeCr and FeNi. In the first case the Fe can be absorbed in the matrix and
excite Cr-radiation. Therefore the intensity of Fe is reduced in comparison to a
linear behavior. For the other binary alloy Fe-radiation can additionally excited by
the absorption of Ni-radiation. Therefore the intensity of Fe is increased in com-
parison to the linear behavior. In the case that all three elements (or still more) are
present the relation between weight fraction and intensity will be even more
complex and cannot be described by a single calibration curve but requires a set of
calibrations curves.
Generally, the excitation of elements with secondary fluorescence has to be
considered. But these are third order effects which can be mostly ignored in
particular if there are other influences for the quantification as for example sample
inhomogeneity’s, not correct fundamental parameters etc. that are larger and
cannot considered with sufficient accuracy.
XRF allows a very high precision because the statistical error can be reduced
due to very high count rates (see 4.5.2). But due to the matrix interaction relations
between weight fraction and element intensities are rather complex. For exact
quantifications therefore a calibration is required which relates the intensities of
reference samples with known weight fractions to their intensities. These relations
then can be used for the quantification of unknowns. Due to the matrix interaction
these calibrations are valid only for the calibrated elements and also for a limited
range of their weight fractions. This requires a high effort for the quantification—
the supply of references with exactly known weight fractions, their measurement
and finally the calculation of the calibration curves. Therefore a high effort is spent
to reduce the effort for the calibration i.e. enlarge the elemental and weight fraction
ranges of validity for the calibrations or even calculate the relation between weight
fraction and intensity from first principles [1–9].
160 4 Quantification
The level of spectra evaluation can be very different and depends on the given
analytical task. Most of elemental analysis will be performed with the target to
determine the weight fractions of the sample elements. But very often also easier
analytical results are sufficient. The following examples are a selection of them:
• Element identification
In that case only a qualitative analysis will be performed i.e. it has to be checked
which elements are in the sample or even if there is only a special element present.
That can be done with help of mathematical methods as well as only by a visual
monitoring.
• Positive material identification (PMI)
In that case the complete spectra or the intensities of all elements are compared
with a set of library data and are searched for similarities. This is a mathematical
procedure which does not requires any understanding of the physical conditions of
the measurement. But a library is required which contains spectra of all expected
materials and these spectra need to be measured with the same conditions as the
unknown sample.
This spectrum matching is a fast process and in case of limited material
qualities it can be a powerful tool because it can deliver directly the material or
trade name without a time consuming comparison with tables of weight fractions.
• Monitoring of weight fraction levels
A special analytical problem for process control can be the check of the weight
fraction range of an element. This can be monitored both by the intensity of an
element or by its weight fraction. This type of testing has to be performed if the
weight fraction of an element should be in a given range or below/above a certain
level. Then it delivers a ‘‘go’’—‘‘no go’’ message to the process or it allows a fast
reaction to process parameters.
But XRF allows the analysis of elemental composition with very high accuracy
because the statistical error can be very small. In general quantification in XRF has
two aspects:
• As already mentioned X-ray florescence has a very strong matrix interaction. In
general this is very similar for other techniques for elemental analysis as Atomic
Absorption Spectroscopy (AAS) or Inductive Coupled Plasma Spectroscopy
(ICP). But for these analytical methods the samples typically are strongly
4.2 Different Types of Quantification 161
diluted and therefore the matrix seems to be the same i.e. the matrix is the
solvent and the analyte can considered only as a small distortion. But XRF
mostly is a direct analysis i.e. without dilution and therefore the matrix is
strongly changing with every sample.
The strong matrix interaction of XRF causes that an approach of the measured
intensities to the corresponding weight fractions is not possible for a wide range
of weight fractions with linear calibration curves. Further for different quali-
tative compositions new calibrations are required because the matrix interaction
is changed. That means the measured element intensities depend not only on the
measurement conditions and the element weight fraction itself but are influ-
enced also by all other elements in the sample.
The basic model for the description of the relationship between intensity and
weight fraction is given by the Sherman-equation (3.7). This is the approach
only for interactions of first order. For a correct quantification, at least also terms
of secondary interaction have to be taken into account. In special cases even
terms of third order have to be considered.
Z
wi si ðEÞ S1
S pi xi
Ii ¼ G lðEÞ
I0 ðEÞ dE ð4:1Þ
sin wein þ sinlðiÞ
w
E aus
with:
Ii intensity of element i
G geometry factor
s linear absorption coefficient
(S - 1)/S jump ratio
p transition probability
x fluorescence yield
w weight fraction
l linear mass absorption coefficient
E energy
W incident and take off angle
I0(E) excitation spectrum
But the problem of this equation is that it has the ‘wrong direction’. It calculates
the intensity Ii of the element i as a function of the instrument and excitation
162 4 Quantification
with:
f functional relation
M description of all matrix interactions
with:
i index for the element of interest
j index for all other elements
This type of relation is called concentration correction models. But they have a
disadvantage—the weight fractions of the matrix elements are not known and
therefore a solution is only possible by iterative calculations.
To solve this problem it is possible to use a linear approach for the influence of
matrix elements i.e.
wj ¼ a Ij ð4:4Þ
with:
a sensitivity
This type of quantification model is called intensity correction models and they
can be solved directly without iterative procedures because the intensities can be
calculated directly from the measured spectrum.
All these models are described elsewhere in detail [15–29].
All these correction models need several parameters like sensitivities, correc-
tion factors, overlapping factors etc. for their correct application. The determi-
nation of these factors requires a calibration procedure. Calibrations are performed
by the measurement of samples with known weight fractions. These samples can
4.2 Different Types of Quantification 163
For l-XRF special conditions for the quantification have to be considered because
this method is used preferentially for inhomogeneous samples and analyze
therefore only a small sample volume. This causes special conditions for quanti-
fication and can be accompanied with different problems in particular in case of
using standard based quantification models.
• Variations of sample composition
The sample composition depends on the sample position for inhomogeneous
material. The first conclusion is that a standard based calibration would require a
wide range of different standards. But these standards are not available at every
time, and if they are available the effort for their measurement is very high.
164 4 Quantification
Fig. 4.5 Elemental distribution of a glass reference for Si, Ca and for Al in different depths
This measurement shows that even for glass references which are typically
assumed to be homogeneous this assumption can be wrong. For the examination of
large areas these small inhomogeneitie’s can be ignored because they are averaged
by the measurement procedure but for l-XRF which analyses small sample areas
only it can be crucial.
– Metal alloys
Also metal alloys typically are assumed to be homogeneous. This is correct for
larger areas. But is that still true for micro-areas? Here measurements were per-
formed on an Al-alloy and a Ni-alloy from Brammer [35]. The measurements were
performed in two steps. The first test was a repeatability test i.e. the measurements
were performed in the same sample position and the second test was performed
with same measurement conditions (60 s, 50 kV, 600 lA and 200 lA for the Al-
alloy and Ni alloy, respectively) in a grid of 3 9 3 points. The results are pre-
sented in Fig. 4.6 and in Tables 4.1 and 4.2.
The two diagrams show for several alloy-elements the standard deviation of the
measured intensities for the repeatability test as well as for the measurements in
different positions. The standard deviations for the measurements in the different
positions are significantly larger. This can only be explained by composition
changes of the sample.
166 4 Quantification
Fig. 4.6 Homogeneity test for an Al-alloy (left) and a Ni-alloy (right)
Table 4.1 Weight fractions and relative standard deviations for measurements on an Al-standard
Element Na Mg Al Cr Fe Cu Zn
Conc/wt-% 3.59 1.95 86.6 0.18 0.29 1.25 5.95
Rel STDev—repeat/% 4.60 3.12 0.17 1.77 9.82 0.74 0.79
Rel STDev—homo/% 6.52 5.27 0.19 5.10 28.26 5.04 4.33
Table 4.2 Weight fractions and relative standard deviations for measurements on a Ni-standard
Element Ti Cr Fe Co Ni Mo
Conc/wt-% 2.64 18.85 0.90 13.75 59.28 4.59
Rel STDev—repeat/% 0.37 0.11 0.47 0.08 0.03 0.26
Rel STDev—homo/% 1.13 0.24 2.07 0.21 0.06 0.67
The considerations of the last chapter show that the quantification for l-XRF offers
several problems because of the inhomogeneity both of the sample itself but also
of the standards. Therefore it has to be assumed that the accuracy of quantification
results in l-XRF with a standard based quantification will not correspond to the
high expectations regarding accuracy and not to the high effort which are required
for calibration with standard based models. Therefore a standard based model
should not be the preferred quantification model. Standardless models which don’t
require references and which are even independent of excitation conditions would
give satisfying accuracy and should be the preferred solution.
Nevertheless, for special applications also standard based quantification models
can be used and should be available for a bench top instrument. These models are
described elsewhere [2, 15–29].
The considered interactions are displayed in the following figure (Fig. 4.8):
The following interactions have to be taken into account:
• The tube emits their radiation with the spectrum Io(E). This spectrum can be
calculated with relations given by Ebel and Finkelshtein [36, 37]. This offers the
possibility to use the quantification model for different excitation conditions i.e.
with changing tube voltage or with filters in front of the tube.
• The tube radiation will be additionally influenced and changed by the trans-
mission through the X-ray optics. This can be described by the transmission
function T(E). This function depends on the optic and their adjustment to the
tube target, see Sect. 2.2.5.3. Therefore it is necessary to calibrate the influence
of the transmission function to get satisfying quantification results.
• The incident beam will be absorbed on the way through the sample to the
excitation of fluorescence radiation. This absorption depends on the matrix,
described by their mass absorption coefficient l(E) and the way through the
sample which depends on the incident angle Hincident.
• The incident radiation will be absorbed at the point of interaction. This is
described by the linear absorption coefficient s (E) and the jump ratio of the
absorption edge (S - 1)/S. These parameters determine the number of vacan-
cies in the electron shell. The number of excited atoms depends also on the
weight fraction w of the atoms of the element i.
• This vacancy has to be filled from outer electron which is described by the
transition probability pi.
• The energy attained by filling the vacancy can be emitted as X-ray photon or as
Auger-electron. The probability for the emission as X-ray is described by the
fluorescence yield xi where i describes the shell of primary excitation.
170 4 Quantification
• Finally the fluorescence radiation needs to go to the sample surface. In that way
the fluorescence radiation which has a given energy E will be absorbed. This is
described by the mass absorption coefficient li and the way through the sample
which depends on the take-off angle Htake off.
Equation (4.1) describes only interactions of first order, but for a sufficient
accuracy of quantification results at least second order interactions needs to be
considered. The second order terms Sij of the Sherman relation are given by:
Si 1 S 1
G Z wi si Si pi x: wj sj Sj pj xj Io ðEÞ
j
Sij ¼ l ðE Þ l ð iÞ
2
sin hincident þ sin htake off
s s
! !
sin hincident lðEÞs sin htake off ðlj Þs
½ ln 1 þ ffi þ ln 1 þ de
lðEÞs lj s sin hincident ðlj Þs ðlj Þs sin htake off
ð4:7Þ
Because the Sherman equation is a function between the intensity Ii of the element
i of its weight fraction wi for a calculation of the weight fraction an iterative
procedure has to be used which is demonstrated in Fig. 4.9.
The iterative procedure starts with a guess of weight fractions which are used
for the calculation of element intensities. For that purpose it is necessary to know
all elements in the sample, the measurement conditions, if necessary some stoi-
chiometric relations between the elements and all fundamental parameters [38].
For the calculation the optic transmission function has to be considered if an optic
is used. Further corrections are possible with special element dependent correction
factors from a ’type calibration’ (see next chapter). The next step is a convolution
of calculated intensities with the spectrometer function. The result is a calculated
spectrum which can be compared with the measured spectrum. This procedure has
4.3 Quantification for l-XRF 171
the advantage that also spectrum artifacts like escape-peaks or pile up-peaks are
already included in the comparison, that changing intensity relations between peak
of one series due to sample thickness or density of the sample are taken into
account and that the complete X-Ray series can be considered for the peak
intensity determination which reduces the statistical error against considering only
the main peak. The comparison between calculated and measured spectrum gives
hints for the improvement of weight fractions and the procedure can be iteratively
repeated up to the calculations show convergence. This is already reached within a
few iterations typically not more than 5 or 6.
In case of defined compounds the known ration between measured elements and
non-measured elements is considered for the matrix interaction and also the result
presentation.
• General accuracy
Testing this procedure of quantification is possible by the analysis of reference
samples. Of course it is necessary to consider the discussions in 4.3.1 about
172 4 Quantification
Fig. 4.10 Relation between given and analyzed weight fractions (left) and deviations from given
values (right) for a set of metal alloys
Table 4.3 Comparison of given (gray highlighted) and analyzed weight fractions for few
samples
Sample Si Ti Cr Mn Fe Co Ni Cu Zn Nb Mo W
CDA260 0.04 69.4 30.7
0.21 69.2 30.6
N-200 0.13 0.02 99.7 0.04
0.15 0.24 99.6 0.01
Inco 800H 0.35 0.56 20.9 0.78 44.6 31.9 0.27
0.45 0.46 19.4 0.74 44.5 32.0 0.42 0.35
20CB3 0.26 19.7 0.40 40.7 32.9 3.14 0.67 2.12
18.6 0.34 42.7 32.7 3.18 0.60 1.78
HA 25 0.30 19.9 1.57 2.3 50.5 10.1 15.3
17.4 1.55 2.5 50.5 11.1 0.45 16.3
310 0.54 25.3 1.72 52.6 19.3 0.10 0.40
0.95 23.9 1.62 54.4 19.4 0.17 0.35
430 0.39 16.6 0.44 82.0 0.29 0.11 0.05 0.06
15.4 0.68 82.5 0.25 0.10 0.05 0.07
Ti-GR7 99.8 0.08
99.9 0.11
optimizing the excitation conditions. For standard based quantification models the
measurement conditions needs to be fixed for the standard measurements as well
as for the unknowns.
The high flexibility of the standardless model can be demonstrated by results of
measurements on the same sample—a Cu-alloy—which was measured with dif-
ferent conditions. The quantification results for different measurement conditions
are displayed in Table 4.4.
The table shows in the first 4 columns the measurement conditions. These
conditions are strongly changing. The following parameters are changed:
– high voltage: from 30 to 50 kV
– tube current: from 200 to 600 lA, but this should have no influence to the
quant results due to normalization to 100 %, but high count rate can generate
pile up peaks that can influence the analytical result
– measurement medium: changing between air and vacuum
– different filters for the excitation path.
The next columns show the weight fractions—in the first line the given weight
fraction and the following lines the analyzed values.
The second half of the table is a summary of the results. Here are displayed for
all elements the mean values of all calculated weight fractions as well as their
absolute and relative fluctuations.
For few measurement conditions the deviations are relatively large. Here a
detailed consideration is necessary. This shows that the used element line needs to
be sufficient excited and detected to get reliable results. Deviations of the ana-
lytical results can be larger if this condition is not fulfilled. For example
174 4 Quantification
– Sn cannot efficient excited with 30 kV. In that case the Sn result has a large
deviation
– Si cannot efficient excited with a thick Al-filter. In that case the intensity of Si
is very small and the statistical error too large.
– For the highest excitation power (50 kV, 600 lA) the count rate is approx.
500 kcps which results in pile up peaks even of third order of Cu at an energy
of 24.1 keV. This pile up peak is overlapping with Sn-Ka and influence the
Sn-result for that measurement. In general pile up peaks can be corrected but
for pile ups of second order a correction mostly is not performed.
– Si radiation will be absorbed in air. For that measurement the result for Si is
wrong.
If the results of these 4 measurement conditions (highlighted in Table 2.3)
would be neglected the fluctuations are significantly reduced as can be seen in the
last part of the table.
In general it can be concluded that the quantification results don’t depend on the
excitation conditions if these allow an efficient excitation and if the collection of
fluorescence radiation is not disturbed. This is a noticeable advantage because the
conditions can be optimized for the actual sample.
• Type calibration
Also if the results of a standardless quantification procedure can be impressively
good there is a permanent request for improved accuracy. The described FP-model
allows an adjustment of the quantification results with only few standards. The
4.3 Quantification for l-XRF 175
procedure of this type calibration follows the flow chart in Fig. 4.9. The following
steps then are necessary:
– Measurement of the standards
– Quantification of the standard with the FP-model
– Comparison of the final result with the given weight fractions and determi-
nation of correction factors
– This correction factors can be calculated for single elements i.e. it is possible
to select the elements that should be corrected for further quantifications. This
allows to ignore elements with small weight fractions which have small
intensities and therefore large statistical error.
These correction factors then can be used for further quantifications. They can
be used together with the correction of intensities for the transmission function
(see Fig. 4.9) or also after the convergence of the iteration procedure for the final
weight fractions and before the normalization of the final result. That calibration is
called ‘‘Type calibration’’ because it can be performed for different sample
qualities separately.
This is demonstrated for a set of stainless steel samples in Table 4.5. The first
line of this table shows the average weight fractions of the set of reference sam-
ples. The next two lines show the average deviations and relative deviations for
every element as analyzed without correction. This shows that the weight fractions
of Cr was systematically to low and for Ni slightly too high. Similar deviations are
shown also in Table 4.3 for the stainless steels as well as for the Co- and Ni-alloys.
Therefore correction factors were calculated only for the main components Cr, Fe
and Ni.
The summarized results for the quantification with type calibration are shown in
the last two lines of Table 4.5. They show a reduction of the average deviation for
Cr and Ni by a factor of approx. 4 and for Fe an improvement by approx. 20 %.
For all other uncorrected elements there is no significant change of deviations.
Due to the small weight fractions of these elements or due to overlapping with
other peaks (for example overlapping of the Mn-Ka peak by the Cr-Kb-peak) their
intensities are small and have limited statistics. Therefore the correction factors of
these elements would be influenced by their large statistical error.
176 4 Quantification
4.3.3 Summary
Coatings on materials are very often used to protect the material for example
against corrosion, to realize a special functionality for example good electrical
contacts on printed circuit boards or to improve the appearance for a wide range of
consumables.
XRF offers possibilities for the analysis of coatings for thickness and compo-
sition. And this analysis can be done from top without any damaging of the
sample. The general principle is explained in Fig. 4.11.
The left part of Fig. 4.11 shows the excitation of a mono-layer on a substrate.
The excitation radiation penetrates the layer and generates fluorescence radiation
but also the substrate will be excited and emits fluorescence. Due to that excitation
4.4 Analysis of Coating Systems 177
Fig. 4.11 Analysis of a single layer with XRF (left). Layer and substrate signal in dependence of
layer thickness (right)
two different signals are available which can be used for the determination of the
coating thickness. The dependence of the fluorescence intensity of the thickness of
both the coating and the substrate are displayed in the right part of Fig. 4.11.
The layer emits the emission signal. In case of zero thickness of the layer there
is no signal but it will increase with growing thickness. For very thin layers the
intensity growing is linear but then the slope is reduced by self-absorption and
finally the signal goes into saturation because the fluorescence radiation is com-
pletely absorbed in the layer.
Also the substrate emits fluorescence radiation but this radiation will be
absorbed in the layer. Therefore the substrate signal has the highest intensity if
there is no layer i.e. if the layer thickness is zero. With growing layer thickness the
tube radiation will be absorbed on the way to substrate in the upper layer and also
their fluorescence radiation is absorbed in the layer. Therefore the substrate
intensity drops down with growing layer thickness.
In the displayed example the drop down of the substrate signal is faster than the
growing of the emission signal which means that the layer has a higher atomic
number than that of the substrate and absorbs the substrate radiation therefore
stronger.
From that consideration it can be concluded that there are different possibilities
to calculate layer thicknesses.
• Emission mode
In that case the emission signal of the layer is used for the calculation of coating
parameters. This mode is mostly used and the only possibility to examine also
multi-layer systems because for the determination of several parameters (thick-
nesses and composition of the different layers) the same number of independent
measurement parameters is necessary. For multi-layer systems at first the upper
layer is evaluated. With this knowledge the influence of this layer to excitation and
absorption of fluorescence radiation of the next layer can be considered for the
evaluation. Possible next layers have to be handled accordingly. Of course, the
uncertainty will be larger for every hidden layer due to error propagation.
178 4 Quantification
• Absorption mode
In that case the absorption signal of the substrate is used for the determination of
the layer thickness. This mode can be used only for single layers because only one
independent measurement parameter is available. But this mode is interesting in
the case of light layer materials which have a fluorescence radiation that cannot be
detected such as for the measurements of Al or Si in air or of Carbon. In that case
the absorption of the substrate radiation can be used. Another possibility is the
sensitivity enhancement for very thin layers against the emission mode due to the
higher intensities that have a smaller statistical error.
• Relative mode
The relative mode uses both signals from the layer and the substrate. In that case it
is possible to perform the measurement independent of the distance sample—
detector. This can be helpful in the case of special shaped samples i.e. for samples
like a cup or bowl that allow the measurement only with larger working distances.
It is necessary to know that the direct result of a coating analysis is the number
of atoms per area or the mass-per-area Q and not the thickness. But values of mass-
per-area cannot easily be handled—atoms per mm2 are numbers in the range of
1018 and the mass per area are value in lg/cm2. Therefore usually these values are
transferred to thickness with
t ¼ Q=q ð4:8Þ
with:
t thickness
Q Mass per area
q density
But for that conversion the density has to be known. Typically the density for
the pure element is used. This is an acceptable approach for relatively thick layers
but for very thin layers this can be wrong because the growing process of a layer
does not start always homogeneous. Even more difficult is the determination of the
density of alloys—here an average of the pure element densities is used. But this is
not in any case correct again in particular for very thin layers.
The analysis of coating systems has some special conditions that influence the
variety of coatings systems that can be analyzed.
• Knowledge about the elements in layer and substrate
The examination with XRF is non-destructive. But for a complete analysis it is
necessary to know which elements are in the layers and in the basic material i.e.
4.4 Analysis of Coating Systems 179
the complete coating structure. This is typically not a serious problem because
coating analysis is mostly performed as quality control for the manufacturing
process. In that case the elements in the substrate and layers are well known.
In general it is also possible to determine the layer structure with XRF but then
several measurements with different incidence or take off angles would be required
for the determination of the layer sequence from the different absorption of the
layer signals. But that requires special instrument configurations and will not be
considered here.
• The elements in the layers and substrate needs to be different
Because the spectrometer can only distinguish between the energies of the element
radiation but not between their sources it is necessary that all layers and also the
basic material are composed of different elements. Every layer needs to emit a
distinct signal which can be used for the determination of thickness. That is
realized if the layers and substrate contains different elements or if at least different
lines of the same element (K-L or M-lines) can be used for the evaluation.
This requirement for different elements in every layer is given not only by the
necessity for source identification but also by the mathematical demand for the
same number of independent measured parameters as layer parameters (thickness,
composition) has to be determined.
A simplification of layer analysis is given by the fact that the elements are not
only known but also the number of elements is mostly not very large. Very often
layers are mono-element layers—which is given by the often used galvanic coating
process which preferably produces mono-element layers. Coatings which are
prepared by condensation of evaporated material can also be prepared as alloys.
• Thickness range
The thickness range of layers that can be analyzed is limited. For thin layers it
depends on the signal-to-background ratio. That means it is influenced by the
spectrometer performance. For the upper end of the calibration curve it depends on
relationship of the fluorescence radiation energy and the self-absorption of this
radiation in the layer. These thickness ranges will be discussed more in detail in
4.4.4.
• Analyzed area
Another special condition of coating analysis which has not a direct influence to
the quantification is the size of the analyzed area. Examinations of coatings are
performed on final products which are mostly not flat but non-regular shaped.
Therefore only small flat areas are available for the measurement. That requires the
use of small spot for their examination i.e. it is necessary to guarantee sufficient
excitation intensity and a large capture angle to get satisfying fluorescence
intensity from a small sample area and a thin coating material.
180 4 Quantification
Also if there are only a few elements in the coating systems their interactions are
quite complex. The following interactions have to be taken into account:
• Absorption of the excitation radiation in the subsequent layers—this change the
excitation spectrum for every layer
• Intra-layer interaction between the elements within the layers in case of an
alloyed layer
• Inter-layer interaction between the elements of the different layers and between
the layers and the substrate—this interaction has to be considered in both
direction i.e. from upper layers to bottom layers and vice versa
• Interactions between the elements of the substrate
• Scattering of excitation radiation on the layers and on the substrate and their
interaction with the layer.
There are different approaches for the description for coating analysis [40–47].
An easy way to understand coating analysis requirements for a single self-sup-
porting layer is based on the Sherman-relation (4.1) which can be expanded to:
Z
wi si ðEÞ S1
S pi x i
Ii ¼ G lðEÞ
Io ðEÞ dE
sin hincident þ sin hltake
i
E off
ð4:9Þ
lð E Þ lf
f1 exp þ qt g
sin hincident sinhtake off
This relation did not yet consider the secondary intensity enhancement caused
by secondary fluorescence and scattered radiation between the layers in a multi-
layer system as well as from the base material for a supported layer system. This
requires an expansion of the relation and makes it more complex. But even then it
describes the interactions still correctly [40].
In general it is possible to use this relation similar as for bulk samples for a
forward calculation because the elements are known and very often only the
thickness has to be iteratively optimized. For alloyed layers both weight-fraction
and thickness needs to be changed.
But mostly this relation is not used directly. For coating analysis frequently the
intensity ratio Ri of the element i in the layer or in the substrate to the infinite thick
4.4 Analysis of Coating Systems 181
pure element measured with the same arrangement, is used. In that case (4.4) can
be reduced to:
Ii
Ri ¼ ¼ K f1 exp½ðle ðEÞ= sin Hins þ li = sin Htake off q tg ð4:10Þ
I1
because the integrals for layer and pure element are the same. The factor K is a
constant and does not depend on energy. This relation looks relatively easy but is
valid only for primary interactions—interactions of second order are not yet
considered. These effects by secondary fluorescence and excitation by scattered
radiation within the layer or between the different layers or the substrate of the
layer system which can be considered as a very thick layer can change the mea-
sured intensity.
The relation between intensity ratio and the thickness of the layer is already
shown in Fig. 4.11 as the emission signal. It shows that for very thin layers a linear
intensity gain which is decreased for larger thicknesses and then approaches
asymptotically the intensity value of infinite thickness. For half of the infinite
thickness the slope is already notably reduced i.e. intensity changes for increased
layer thickness are only marginal. This determines the limit for the measurement
of thick layers.
That means, for thick layers it is necessary to measure the intensity with very
high accuracy to reduce the statistical intensity fluctuations, i.e. the measurement
time can be determined from a preselected statistical error. This also determines
the upper limit for thickness determination. Nevertheless it can be assumed that the
covered thickness range ends for an intensity ratio of approx. 95 %.
From that consideration it is possible to discuss the thickness ranges that can be
analyzed with XRF.
• Limit of detection for layers
The limit of detection for layers can be estimated in the same way as for traces in
bulk material i.e. the 3 r criteria is usable in as
pffiffiffiffiffiffiffi
3 BG
tLOD ¼ tsample ð4:11Þ
Isample
with:
tLOD limit of detection for a layer system
tsample layer thickness of a given sample
Isample intensity measured on the sample
BG background intensity
182 4 Quantification
So far the important parameter for the analysis of thin layers is also the peak-to-
background ratio of the spectrometer. Better energy resolution of the spectrometer
or higher count rate capability (Sect. 2.5.5.2) improves the limit of detection. That
means spectrometers with semi-conductor detectors in particular with SDD can
have a higher sensitivity also for thin layers, but then a high fluorescence intensity
is required.
A further influence is given by the selected line of the element. As can be seen
in Fig. 4.12 the slope for the low energetic Au-M-radiation is higher than that for
the higher energetic Au-L-radiation. This effect also can be used to improve the
sensitivity for thin layer analysis.
The limits of detection for different layer systems are displayed in Fig. 4.13
which shows the complete thickness ranges for few typical layer elements (ele-
ments which are close in atomic number have similar thickness ranges) in
dependence of the different spectrometer type i.e. for spectrometers with a prop-
counter (PC) and spectrometers with a semi-conductor detector (here an SDD)
which is working in air or in vacuum.
The distinction between measurement in air and vacuum is necessary because
light elements like Al or Si can only be measured in vacuum because their fluo-
rescence radiation will be even absorbed in air. For elements around Au (i.e. Pt,
Pb) the use of M-radiation can be performed in air but also in vacuum. The
measurement in vacuum gives an intensity increase by a factor of approx. 3 which
of course also improves the limit of detection. A similar situation is valid also for
elements between Rh and Sn if their L-radiation is used.
In dependence of the element the thickness range covers up to 5 orders of
magnitude and goes down for heavy elements into the sub-nm-range. That means
mono-atomic layers or even thinner layers can be analyzed. Some examples will
be given later in 0.
4.4 Analysis of Coating Systems 183
Fig. 4.13 Thickness range for different elements in dependence of measurement conditions
(detector resolution, medium)
Due to error propagation the quantification error increases from layer to layer.
This effect will be forced because the intensity for hidden layers is reduced due to
the absorption effects i.e. an increased statistical error.
The influence of a 0.5 lm thick Au layer to the excitation spectrum is shown in
Fig. 4.15. This image shows the original excitation spectrum of the tube (blue)
which would excite the Au-layer and the spectrum after penetration of the Au-
layer (red) which would excite the hidden layer. It can be seen that the excitation
intensity drops down—more for the low energetic part of the spectrum which is
not required for the excitation but also for the higher energetic part which is used
for excitation. In the red spectrum the absorption edges of Au are clearly visible.
This changed excitation spectrum influences also the calibration curve of the
hidden layer. This is demonstrated for a Ni-layer which is covered by an Au-layer
in Fig. 4.16.
The fluorescence intensity of the Ni-layer is reduced by the 0.5 lm thick
covering layer because the excitation intensity at the absorption edge of Ni is
decreased by approx. 20 % and also because the Ni-fluorescence radiation is
4.4 Analysis of Coating Systems 185
The result of such correction is shown in Fig. 4.19 which shows the original
calibration curve, the results of two reference samples and the corrected curve
which contains these two references.
The achievable accuracy depends also for coating thickness testing on the
statistical error i.e. an improved accuracy could be expected for higher intensity of
the layer signal. This intensity is influenced by the measurement time, by the layer
thickness and in case of hidden layers by the thickness of covering layers.
Coating thickness testing is mostly used as quality control in production i.e. the
quant result needs to be available within short times for a fast feedback to the
manufacturing process. Measurement times in the range of only few seconds are
common. Therefore the accuracy typically is not very high.
But on the other hand also the homogeneity of the layer system has to be taken
into account. The layer homogeneity is influenced by the roughness of the sub-
strate which affects in particular very thin layers. For very thin layers also the
influence of the layer density has to be considered because this can be different
from that of the pure element.
But for also thicker layers non-homogeneous thicknesses are possible. They can
be caused by the coating process itself. For example the galvanic process prefers
the deposition of layer material in areas with a high electrical field on the sample.
These are often surface areas with high curvature. In the case of coating by
evaporation the areas which are perpendicular to the evaporation source have a
higher deposition than areas that are tilted. But that means, the analytical result can
be changing in dependence of the actual measurement point. Therefore the
expectation for high accuracy in coating analysis should be not extraordinary high.
Typically the uncertainties are in the range of few %. This also limits the
requirements for the statistical error and therefore also for the required measure-
ment time.
4.4.7 Summary
For every analysis the target is an accurate result. The effort to reach that intention
depends on the requirements. It can be very high and sometimes it can even spent
on the wrong problems. Therefore it is helpful to explore the different types of
errors and their contributions to the total analytical error.
The first question is what is accuracy? There is a wide spread literature and
even several standards available that describe analytical errors and their consid-
eration [48–53].
Accuracy can be considered as a combination of precision, reproducibility and
trueness. All these parameters should to be considered separately.
• Precision is the short term repeatability of the measurements. It considers the
analytical results of successive analysis. Only with a good repeatability can
correct results be expected from a single analysis.
• Reproducibility is the repeatability of an analytical result for different condi-
tions like different time, different operator, different instrument, different lab-
oratory etc. With reproducibility both the long term stability of the analytical
method and its robustness against different external conditions is subsumed.
• Stability is the long term repeatability of measurements on the same instrument.
It considers changes over a longer period of time (drifts). Often these changes
can be corrected by monitor measurements and drift corrections.
• Trueness is the difference between the analyzed and the correct value i.e. for
XRF this would be element weight fraction or layer thickness. This is finally the
dominant requirement but if precision and reproducibility are not given a high
trueness is not possible.
4.5 Errors in l-XRF 189
The combination of these parameters can be considered with the very often
used target model which is shown in Fig. 4.20. It shows targets with different
results.
• The distribution in the upper left image illustrates a good repeatability and the
results are also correct; they are randomly distributed but close to the middle of
the target.
• The distribution in the upper right image illustrates a bad repeatability but the
average have the correct value.
• The distribution in the lower left image illustrates a good repeatability but the
average have the correct value i.e. their trueness is bad; a systematic error causes
this deviation. In this case an adjustment of the method is required, for example,
a calibration.
• Finally the lower right image shows a randomly distribution they are not correct
and do not have sufficient repeatability. They cannot deliver a reliable analytical
result.
The problem in that picture is that the true value typically is not known. That
means only precision and not trueness can be assessed for an unknown sample.
Therefore the analysis of reference samples with known composition or thickness
is necessary for the examination of systematic errors.
Often the mean value from several repeated measurements is used as true value.
But that implies that no systematic error influences the analytical result.
From this picture a classification of errors can be derived—these are random
distributions and systematic deviations i.e. we have to consider also random and
systematic errors.
190 4 Quantification
The main randomly error is the statistical counting error, but also errors due to
sample positioning, fluctuations of measurement parameters etc. can be randomly.
Systematic errors are caused by systematic or repeated deviations from the correct
measurement condition like wrong sample preparation or positioning, use of not
correct data evaluation procedures or reference samples etc. Thereby the diversity
of errors can be very large.
The error depends on the functional relation between the measured value v and
the different influence parameters xi given as
v ¼ f ðx1 ; x2 ; . . .; xn Þ ð4:12Þ
If all influence parameters are independent of each other the total error which is
defined as total standard deviation Dv is given by
sffi
ffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffi
dv 2 2 dv 2 2 dv 2 2
Dv ¼ Dv1 þ Dv2 þ þ Dvn ð4:13Þ
dx1 dx2 dxn
with:
Dv total error
Dvi different error contributions
with this law it is possible to estimate the contribution of every single input to the
total error. This is demonstrated in Table 4.7 which shows that the total error is
dominated by the largest contribution i.e. it makes no sense to reduce anyone
contribution but the largest one should be reduced to get a small error.
Randomly error distributions typically have deviations from the true value in both
directions i.e. to higher and lower results. Therefore they allow an appropriate
treatment with statistical procedures. The results of random errors are typically
4.5 Errors in l-XRF 191
Gaussian distributed as shown in Fig. 4.21 for a good and bad precision in black
and gray, respectively.
There are different randomly distributed contributions to the error that has to be
considered.
• Statistical errors
For XRF the statistical error is the main contribution because this error can only be
reduced but not completely avoided. The statistical error so far is the smallest error
for XRF that can be reached. XRF is described by Poisson statistics which is the
same as Gaussian statistics if the intensities are high.
The standard deviation of the statistical error for a given number of counts N
can be described as
pffiffiffiffi
DN ¼ N ð4:16Þ
That means for a count number of 100 the error is 10 %, for a count number of
10,000 it is 1 % and for a count number of 1,000,000 it is 0.1 % i.e. the
improvement of error is related to the square-root of intensity.
This offers also the possibility for the reduction of the statistical error—higher
intensity. Higher intensities are available either by higher excitation efficiency or
excitation intensity or by longer measurement time. Because the excitation con-
ditions mostly are given the prolongation of measurement time is the preferred
method to reduce the statistical error. But then the instrument needs a high stability.
This knowledge of the statistical error offers the possibility for simple esti-
mations of other error contributions. For that purpose n repeated measurements
192 4 Quantification
with given measurement conditions should be performed. For the measured counts
N of every measurement the standard deviation from the mean value N can be
determined as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðN NÞ2
DNmeas ¼ ð4:18Þ
n
This standard deviation of the measurements can be compared with the standard
pffiffiffiffi
deviation of the statistical error i.e. with DNstatistic ¼ N . The relation between
both standard deviations i.e. DNmeas /DNstatistic should be less than 2. Then the
working of the analytical method can be assumed as satisfying.
This is also a way to investigate different randomly contributions to the ana-
lytical error. If the described procedure is performed with specific changes of the
measurement conditions and the ratio between standard deviation for the mea-
surement and the statistical errors are compared it is possible to estimate the
influence of different errors and to make conclusions for improvements of the
measurement strategy.
• Randomly distributed errors during the measurement
Only a few of the possible randomly distributed errors during the measurement
procedure with a l-XRF instrument can be mentioned. In dependence of the real
analytical task and the appropriate sample it is necessary to consider the different
error contributions for every analysis again.
The measurement can be influenced by random variations of the excitation
conditions as fluctuations of tube parameters during the measurement itself or in
case of their new setting.
Other random errors are changes of the analyzed volume by surface contami-
nations, by a tilted surface, and by fluctuations of composition for inhomogeneous
samples by changing the measurement position.
Also different distances between sample and detector can influence the captured
fluorescence radiation.
• Randomly distributed errors during data evaluation
Also after the measurement errors can occur during data evaluation. For example
the deconvolution of strongly overlapping peaks has an error that depends on the
actual measured spectrum i.e. is changing with the measurement. This contribution
can be reduced for higher intensities i.e. by longer measurement times which
improve the peak fitting procedures.
Systematic error contributions are deviations typically only in one direction. These
errors then are overlaid with a random distribution as demonstrated in Fig. 4.22.
4.5 Errors in l-XRF 193
That means the random distribution for a systematic error is not around the true
value but around another value. Here it was treated as a random error because the
correct true value is not known.
Systematic deviations can have manifold sources.
• During the sample preparation process for example by not sufficient homoge-
nization of the sample, by contaminations in particular of the sample surface, by
influences of surface topologies i.e. by tilted sample surfaces which influences in
particular the absorption behavior for light elements etc.
• During the measurement process by incorrect adjustment of the measurement
position of the sample, by wrong settings of excitation conditions, by the
influence of the measurement medium or contaminations of the instrument etc.
• By changes of the sample composition during the measurement for example by
the absorption of the excitation radiation, by sedimentation or precipitation
processes etc.
• During the data evaluation for example by incorrect qualitative analysis which
influence peak fitting results, by using a quantification model that is not ade-
quate for the analytical problem, by references with not correct defined com-
positions etc.
with:
Utotal total uncertainty
v given or analyzed values (weight fraction, thickness)
Dgiven-anal target performance comparison for both reference and sample
Dvreference standard deviation for the analysis of the reference
Dvcheck standard deviation for the check analysis
Dvsample standard deviation for the analysis of the sample
For this uncertainties can be used also for the unknown sample it should be
similar to the reference as well as the measurement and evaluation procedure
should be the same.
The top-down method for determination of uncertainties is relatively fast and
simple because
– all influences are considered automatically even hidden systematic errors and
– the effort for the determination is relatively low
But on the other hand there are also few disadvantages as
– the reference and the sample and also the measurement and evaluation pro-
cedure needs to be highly comparable
– the different contributions for uncertainty cannot examined separately which
reduce the possibility for their specific reduction and an optimization of the
analytical method
• Bottom-up method—consideration of all separate contributions and creation of
an uncertainty budget
With that method it is necessary to define and examine all different uncertainty
contributions and calculate or estimate their value. The total uncertainty then can
be calculated by adding all contributions according to (4.10). This method requires
a high effort but allows an overview for all influences to the analytical result. But it
offers also the possibility to concentrate to the largest contributions of uncertainty
and their influence or even correction to the analytical result.
196 4 Quantification
The effort for estimations can be controlled because after consideration of the
most important factors the procedure can be stopped. The disadvantages of the
method are
– The effort for the determination of single contributions can be high and
requires a series of measurements as well as complex mathematical
procedures
– An incomplete consideration of systematic deviations is not possible which
can lead to an underestimation of the uncertainty
The procedure for the determination of uncertainties in the bottom-up method
would be require
– a description of the analytical method and recording of all influences
– a selection of unimportant influences
– the definition of distribution models—experimental data according to
Gaussian distribution, benchmark figures by rectangular distribution
– the determination of standard deviations and parameter ranges, respectively
– a calculation of uncertainties with the help of sensitivity factors
– the calculation of combined uncertainty according to (4.9) and (4.10)
– the definition of the extension factor and the determination of the extended
uncertainty
References
1. B.K. Agarwal, X-ray Spectroscopy: An Introduction (Springer, Berlin, 1991). ISBN 0-387-
092684
2. M. Mantler, Quantitative analysis, in Handbook of Practical X-ray Fluorescence Analysis,
ed. by B. Beckhoff (Springer, Berlin, 2006)
3. E.P. Bertin, Priciples and practice of X-ray spectrometric analysis (Plenum Press, New York,
1975)
4. P. Hahn-Weinheimer, A. Hirner, K. Weber-Diefenbach, Röntgenfluoreszenz-analytische
Methoden: Grundlagen und praktische Anwendung in den Geo-. Material- und
Umweltwissenschaften. (Springer, Berlin, 1995)
5. R. Jenkins, R.W. Gould, D. Gedcke, Quantitative X-ray Spectrometry (Marcel Dekker Inc,
New York, 1981)
6. R. Jenkins, J.L. de Vries, Practical X-ray Spectrometry. (Springer, Heidelberg, New York
1969)
7. G. Lachance, F. Claisse, Quantitative X-ray Fluorescence Analysis (Wiley, Hoboken, 1994)
8. R. Tertian, F. Claisse, Principles of Quantitative X-ray Fluorescence Analysis (Heyden &
Son Ltd, London, 1982)
9. R. van Grieken, A. Markowicz, Handbook of X-ray Spectrometry (Practical Spectroscopy)
(Marcel Dekker Inc, New York, 2002)
10. J.W. Criss, L.S. Birks, Anal. Chem. 40, 1080 (1968)
11. J. Sherman, ASTM Spec. Tech. Publ. 1157, 27 (1954)
12. J. Sherman, Spectrochim. Acta. 7, 283 (1955)
13. T. Shiraiwa, N. Fujino, Jap. J. Appl. Phys. 5, 886 (1966)
14. J.V. Gilfrich, X-Ray Spectr. 20, 203 (2001)
15. H.J. Lucas-Tooth, B.J. Price, Metallurgica 64, 149 (1961)
16. H.J. Lucas-Tooth, C. Pyne, Adv. X-Ray Anal. 7, 523 (1964)
17. G.R. Lachance, R.J. Traill, Can. Spectr. 11, 43 (1966)
18. R.J. Traill, G.R. Lachance, Can. Spectr. 11, 63 (1966)
19. W.K. de Jongh, X-Ray Spectr. 2, 151 (1973)
20. N. Broll, X-Ray Spectr. 15, 271 (1986)
21. N. Broll, P. Caussin, M. Peter, X-Ray Spectr. 21, 43 (1992)
22. S.D. Rasberry, K.F.J. Heinrich, Anal. Chem. 46, 81 (1974)
23. G.R. Lachance, Can. Spectr. 15, 64 (1970)
24. R.M. Rousseau, X-Ray Spectr. 13, 115–121 (1984)
25. R.M. Rousseau, F. Claisse, X-Ray Spectr. 3, 31 (1974)
26. R. Tertian, X-Ray Spectr. 2, 95 (1973)
27. R. Tertian, Adv. X-Ray Anal. 19, 85 (1976)
28. JIS G 1256 (1997) Japanese Standards Association, Tokyo
29. N. Kawahara, T. Shoji, T. Yamada, Y. Kataoka, B. Beckhoff, G. Ulm, M. Mantler, Adv. X-
Ray Anal. 45, 511 (2002)
30. M. Rossbach, E. Zeiller, Anal. Bioanaly. Chem. 377, 334 (2003)
31. J. Pauwels, A. Lamberty, H. Schimmel, Accred. Qual. Assur. 3, 51 (1998)
32. Ph Quevauviller, Trends Anal. Chem. 20–8, 446 (2001)
33. L. Kempenaers, N.H. Bings, T.E. Jeffries, B. Vekemans, K. Janssens, J. Anal. Atom. Spectr.
16–9, 1006 (2001)
34. R. Dybczyński, B. Danko, H. Polkowska-Montrenko, J. Radioanal. Nucl. Chem. 245–1, 97
(2000)
35. Brammer Standard, www.brammerstandard.com
36. H. Ebel, X-Ray Spectr. 28, 255 (1999)
37. A.L. Finkelshtein, T.O. Pavlova, X-Ray Spectr. 28, 27 (1999)
38. W.T. Elam, B.D. Ravel, J.R. Sieber, Radiat. Phys. Chem. 63, 121 (2002)
39. W.T. Elam, R.B. Shen, B. Scruggs, J. Nicolosi, Adv. X-Ray Anal. 47, 104 (2004)
References 199
5.1 Introduction
For XRF sample preparation typically is relatively easy because the sample can be
measured as solid, powder, single particle or liquid. Nevertheless it is required to
perform the sample preparation very carefully because there are different influences
from the sample conditions to the measured intensities and therefore for the analytical
result. Only if they can be excluded it is possible to get a correct results. The effort for
the reduction of those effects should be adapted to the expectation for accuracy of the
analysis. Especially for standard based models in quantification it is very important
that standards as well as the unknown sample are prepared in the same way. Only
then it is possible to compare their intensities and use them for quantification.
This situation can be completely different for l-XRF because here very often
the unaltered original sample has to be analyzed—examples are forensic appli-
cations or art objects. In that case it is necessary to take possible surface effects
into account for the data interpretation or surface effects are part of the analytical
question. Nevertheless, also for l-XRF it is valid that the sample preparation has
an important influence to the measured signal and can strongly influence the result
of the analysis.
Another important factor to get a correct signal from the sample for l-XRF is
the sample presentation in particular for small or thin samples. Here it is necessary
to ensure that the correct sample position will be analyzed or that the sample
support cannot contribute to the measured intensities.
In XRF the excitation radiation penetrates into the sample and will be scattered
and absorbed. The absorbed radiation can excite atoms of the sample and these
atoms emit fluorescence radiation. This fluorescence radiation will be absorbed on
the way to the sample surface. Because the energy of the excitation radiation needs
to be higher than the absorption edge of the fluorescence atom their penetration
will be in any case stronger than that of the fluorescence radiation. So far the
thickness of the surface layer that contributes to the signal will be determined by
the absorption of fluorescence radiation in the matrix. This absorption depend both
on energy of the radiation and on the matrix of the sample.
This effect can be described by the ‘information depth’ which can be interpreted
as the layer that contributes to a certain content, for example 95 %, to the mea-
sured signal. That means from the Lambert–Beer-law
and with the assumption that 95 % of the radiation comes from a layer with the
thickness tinfo and a normal take off (5.1) can be changed to
This relation is displayed in the following diagram for four typical matrices i.e.
• for Au (which is typical for precious metal alloys but also for Pb- and W-alloys),
• for Fe (which represents all steels but also alloys of Cr, Co, Ni, Cu or Zn),
• for SiO2 (which can be used also for glasses and minerals) and
• for water (which is an approach for aqueous solutions).
From the diagram the information depth can be taken for different fluorescence
energies (displayed in the abscissa) or for the corresponding elements (displayed in
the upper axis). The information depth can be estimated for the different elements
(or energies) and for the different matrices by extrapolating from the curve for the
corresponding matrix to the ordinate. Of course not every real matrices are
described by the given diagram but the real relations are very close to that. Further
the information depth is not a fixed value at all. But it can be seen that the dynamic
5.2 Information Depth 203
range of the layer thickness that contributes to the X-ray signal is very large—up
to 6 orders of magnitude (Fig. 5.1).
The information depth is an important parameter that allows a guess of the
analyzed sample volume. This volume needs to be representative for the complete
material that has to be characterized or in the case of surface contaminations their
influence to the signal has to be considered.
On the other hand, it is also possible to estimate the influence of surface
roughness to the measured intensity with the help of the information depth. If the
information depth is large against the surface roughness their influence should be
negligible or vice versa in case of similar size of both parameters an influence has
to be expected.
For solid state samples the following preparation techniques can be used if this is
required for the analysis:
• Machining of the solid sample for example by machining or polishing
This allows the preparation of a flat surface for example from metal samples which
are required for an exact analysis—may be of larger sample areas. But in this case
204 5 Sample Preparation
Fig. 5.2 Intensity changes for a glass sample in dependence of roughness of polishing paper
Fig. 5.3 Machined surface of an Al-alloy (left) and the influence to the path of fluorescence
radiation
hand, for the randomly distributed points a significant difference can be observed.
But in comparison to the measurements shown in Table 4.1 which demonstrates
the influence of non-homogeneous distributions of the alloy elements (this sample
was highly polished) here is an additional influences due to a rough surface.
Table 5.1 shows that the standard deviation not only for the alloy elements is
larger than the statistical standard deviation but also for the basic element Al the
fluctuations are higher.
That can be seen in particular for the fluctuations of weight fractions which are
given in the last rows of the corresponding data sets for the (repeated) single point
and randomly distributed values. The influence of surface roughness to quantifi-
cation results in these figures is obvious. The fluctuations for the heavier elements
are smaller in general and in the same range for both measurement series. This is a
result of their smaller concentrations. The influence of surface roughness is neg-
ligible due to the larger information depth of these heavier elements. But for the
light and main elements (Al, Si) the fluctuations for the randomly distributed
measurements are significant—a result of the rough sample surface.
• Direct measurement
Very often l-XRF is used to measure layer systems, corrosion layers, contami-
nations, non-homogeneities or also valuable samples. In these cases a sample
preparation would destroy the structure or the sample itself that should be ana-
lyzed. This has to be avoided. A direct measurement is the only possibility but then
the influences of these surface influences to the analytical result have to be
accepted and take into account.
If the sample is non-regular shaped it is helpful to perform the measurement in
positions which are relatively flat and if possible perpendicular to the incident
beam. For the small spot of a l-XRF instrument it is mostly possible to find a
sample position that fulfills these conditions.
206 5 Sample Preparation
This is a very often analyzed sample type. Their preparation is determined mainly
by the analytical question. It is possible to measure the powder particles as a lose
bunch. In that case it is possible
• to analyze the single particles of the powder which means the excitation beam
needs to be positioned on the particle and the fluorescence radiation have to be
collected or
• to sum up the signals of a larger area as described in 3.4.3 to get an average
information from a larger sample area. But in this case not only the non-regular
shaped sample surface of the single particles but also the non-homogeneity of
sample density influence the analytical result.
If the target of the analysis is a more accurate analysis the powder has to be
prepared. In this case the first step should be a grinding of the powder—if possible
to a grain size which is small or similar against the spot size of the l-XRF. Then
the powder could be mixed with a binder and pressed to a tablet. This offers
relatively homogeneous and stable samples with a flat surface. But for very small
spots in l-XRF the inhomogeneities are still significant.
A next level for powder preparation, in particular for conventional XRF, is the
preparation of fusion bead samples. In that case the—mostly oxide—powder is
mixed with a reducing agent. This mixture is heated up for it’s deoxidation and then
melted at even higher temperatures. After a homogenization the melt can be casted
to a tablet. These fusion beads have the advantage that they are not only homo-
geneous with a flat surface but avoid also all mineralogical effects to the signal.
If the powder consist of metallic particles—for example chips from machining—
the sample preparation again depends on the analytical target. The analysis of single
particles can be performed directly, a rough averaging is possible with an area analysis
or it is possible to melt the sample and prepare a homogeneous casted sample.
A special situation is the analysis of single particles. Here it is necessary to
avoid an influence of the sample environment or of the sample support to the
accumulated sample signal. This is important, in particular, in the case of small
particles where the excitation beam penetrates the sample and interacts with the
sample support due to scattering or excitation. If the positioning of the particle on
a thin plastic film is very helpful (Fig. 5.4). If the excitation radiation penetrates
the sample it will be only scattered on the plastic film. But the scattering intensity
is very low due to the small thickness of the plastic film. As support sample cups
for the analysis of liquids can be used if they have the plastic film on top. In that
case it has to be taken into account that the excitation beam which is scattered on
the sample cup cannot be seen from the detector.
Sometimes it can be helpful to fix the sample on the plastic film to avoid the loss
of the particle. There are different possibilities—but if the particle has to be removed
from the sample cup hair spray can be useful because it can be easily solved by water.
It is necessary to test several spays for the lowest content of detectable elements.
5.3 Preparation and Presentation of Different Sample Qualities 207
Small particles can be collected for example with an impactor on filters which are
prepared from different materials like cellulose, glass, plastics etc. Sample prep-
aration in that case is not usual because the collected particles can be very small.
Therefore filters are measured mostly direct. In that case it is necessary to identify
the influence of the filter material itself by the measurement of the pure filter and
subtract this spectrum from that of the contaminated filter.
The measurement of the filter itself can be performed for single particles if they
are large enough or with an area analysis (3.4.3) to average a larger number of very
small pollution.
The preparation of liquid samples is easy because these samples are mostly
homogeneous or can be prepared easily as a homogeneous sample by shaking or
stirring. Liquids are filled into special sample cups. For a bottom-up measurement
the sample cup needs to have a bottom that allows the penetration of X-rays. Thin
plastic films are used in that case. For a top-down measurement this is not required
and the liquid can be measured without any absorbing material in the radiation
path. But in that case the adjustment of the correct sample height can be difficult.
For the analysis of traces in liquids it has to be taken into account that in
solutions the matrices are very light which results in a high scattering intensity.
This reduces the peak-to-background ratio and also the sensitivity. Therefore the
preparation of single droplets of the solution on a sample support that do not
scatter can enhance the sensitivity tremendous. This is demonstrated in Fig. 5.5.
The liquid droplet scatters the radiation due to the light matrix of the solution but if
208 5 Sample Preparation
Fig. 5.5 Excitation of a droplet and its residue with reduced scattering (left), spectra of the
droplet (gray) and the residue (black) (right)
the droplet is evaporated and only its residue is excited the scattering is reduced
significantly. This enhances the sensitivity for trace analysis in liquids [2–4].
Another advantage of this special sample preparation is the possibility to pre-
cipitate repeatedly several droplets on the same area and increase in that way the
weight fraction of the analyte but also the sensitivity. Further it is possible to add
internal standards to the sample which can be used for the quantification.
Quantification of these samples is easy because inside the very thin residue the
matrix interaction can be ignored. That means, weight fractions are proportional to
the measured intensity i.e.
The sensitivity factors Selem for the different elements have to be determined
only once. With an internal standard this relation is changing to
Extensive examinations for this type of sample preparation are performed from
Fitschen [3]. The different droplet parameters were examined which shows that the
droplet size influences the sensitivity. Finally it was possible to detect traces in the
range of only few nanograms.
References
Abstract There are several other analytical methods available that allow the
determination of elemental or structural composition—also as micro-analytical
methods for small sample volumes. The detailed discussion of l-XRF requires the
consideration also of these methods including the discussion of their benefits and
drawbacks. Additionally the combination of l-XRF with other micro-analytical
methods can be of interest. Known combinations are that with SEM-EDS and with
Raman-spectroscopy. Both combinations give complementary analytical result in
regards to elemental range or to inorganic and organic material composition,
respectively.
6.1.1 Overview
The fact that technical materials are highly complex structured generates the request
for analytical methods with spatial resolution that allow the examination of these
structures. Therefore several methods are available which can deliver various
information’s about composition and structure of the analyzed material. They have to
be distinguished in regards to the interaction of the used radiation with the material
itself but also in regards to their analytical performance and applicability [1, 2].
The most often used method for the determination of elemental composition in
small areas is the combination of SEM with EDS-systems (SEM-EDS). This has to
be discussed in detail in regards to its analytical performance and user friendliness
in comparison to l-XRF.
Another important comparison is necessary with Synchrotron excited XRF (S-
XRF) because this method is very similar to l-XRF but can cover a larger range of
applications in particular more sophisticated ones due to the higher flexibility for
the excitation (high brilliance, sweeping of energy, polarized radiation).
Fig. 6.1 Analytical methods and their spatial resolution and weight fraction range
In relation to l-XRF in particular the spatial resolution and the covered range of
weight fraction are of interest. For a few position sensitive analytical methods that
can be used for the determination of elemental composition or for the character-
ization of coating systems these relationships are displayed in Fig. 6.1. Some
further important parameters as depth resolution, elemental range, imaging pos-
sibility or instrumental effort that characterizes the analytical performance are
summarized in Table 6.1.
All of them have a different analytical performance and can be used with
different analytical targets. The instrumental effort for all of these methods is also
very different. Few of them require a very high instrumental effort for example for
the allocation of the source as for example for Proton Induced X-Ray Emission
(PIXE) [3] or Synchrotron excited XRF (S-XRF) [4], a high instrumental effort
and a complex sample handling is necessary if the method requires ultrahigh
vacuum conditions as for example Auger Electron Spectroscopy (AES) [5],
Transmission Electron Microscopy (TEM), X-Ray Photoelectron Spectroscopy
(XPS) [6, 7] or Secondary Ion Mass Spectroscopy (SIMS) [8–10], or they have a
very complex sample interaction as for example Laser Ablation—Inductive
Coupled Plasma—Mass Spectroscopy (LA-ICP-MS) [11].
But there are also few methods with a limited spatial resolution as conventional
XRF or Optical Emission spectroscopy (OES) that are very often used for ele-
mental analysis and have also considered for that comparison.
All of these methods give similar but in most cases also complementary
information. All of them are required for a comprehensive characterization of
materials. The frequency of the use of these methods depends not only on their
analytical performance but also on their availability—in particular their invest-
ment costs and their user friendliness but mainly on the actual analytical question.
Nevertheless, all methods contribute to the elucidation of the material properties.
Table 6.1 Overview for different micro-analytical methods
Meth. Incident Exit Resolution Light. Detect. limit Imaging Instrument Comment Literature
Elem. effort
Radiat. Energy Radiat. Energy Spatial Depth
XRF X, 30–60 keV X, 0.5-30 keV 20 mm 10 nm Na No Medium
l-XRF X, 30–50 keV X, 0.5–30 keV 20–500 lm 10 nm Na 20–50 mg/ Yes Medium [4]
kg
l-XRF X, 30–50 keV X, 0.5–30 keV [10 nm 1 nm N 5 mg/kg Yes Very high [4]
Synchr
SEM e-, 2-30 keV Secondary–e-, 5 nm 5 nm Be Yes Medium HV. no elem. [26]
50 eV \0.5 lm \0.5 lm Selectivity
Backscatter-e-,
E0
SEM-EDS e-, 2–30 keV X, 0.2-10 keV \0.5 lm \0.5 lm Be 1,000 mg/kg Yes Medium HV [12–15]
EPMA-WDS e-, 5–50 keV X, 0.2–20 keV \0.5 lm \0.5 lm Be 50 mg/kg Yes High HV
STEM e-, [100 keV Second-e-, 3 nm 3 nm B Yes High UHV, no elem.
50 eV selectivity
Auger (AES) e-, 5–20 keV Auger-e-, 0.2 lm 3 nm Li 1,000 mg/kg Yes High UHV [5]
6.1 Comparison with Other Micro-Analytical Methods
\200 eV
XPS/ESCA X, 1–3 keV Photo–e-, 0.5 nm 3 nm He 1,000 mg/kg No High UHV [6, 7]
[10 eV
SIMS Ions, 4–15 keV Secondary ions 5 lm 0.1 nm H 1 lg/kg Limited Destructive [8–10]
PIXE He++. [2 MeV X, 1–30 keV 2 lm 1 mg/kg Limited Very high Sample stress [3]
RBS He++, [2 MeV He++, \E0 2 mm 10 nm Li 1–1,000 mg/ No Medium [27]
kg
LA-ICP-MS Photons (laser) Ions–Light 100 nm H \1 lg/kg Limited Medium Destructive
(vis)
l-Raman Light, visible Light, visible \1 lm Yes Medium Molecules, phases [18, 19]
l-FTIR IR IR 20 lm Limited Medium
LAMMA UV Ions 2 lm H 1 mg/kg Limited Medium Destructive
l-XRD X, 6-20 keV X, same 50 lm 20 lm – – Yes Medium Structure info
213
EBSD e-, 5–30 keV e-, 5–30 keV 100 nm 50 nm – Yes High Structure info [28, 29]
214 6 Relations to Other Analytical Methods
First examinations of small sample areas with exitation of X-rays i.e. for a position
sensitive elemental analysis were performed with synchrotron sources caused by
their very special properties. They have not only a very high brilliance but also a
negligible divergence of the beam and produce polarized radiation in a wide
energy range with high stability. These properties, in particular the high brilliance
open the possibility to use even simple collimators for the generation of small
spots. But the high brilliance and the low divergence are also very helpful in case
of using X-ray optics. This optics typical has a small entrance diameter and
acceptance angle of incoming radiation (see Table 2.7). But with them the spot
size can be considerably smaller than in laboratory instruments—down into the
range of 100 nm or even 10 nm by using Zone-plates (Sect. 2.2.3.2), mono-cap-
illaries or Compound Refractive Lenses (Sect. 2.2.3.1). Spots with these dimen-
sions require automatically high sophisticated equipment for sample handling i.e.
for it’s positioning and viewing. But also the total analyzed areas needs to be
smaller—in the range of a few tenth of lm i.e. the sample has to be pre-examined
for a correct determination of the areas of interest. The use of laboratory instru-
ments with lower spatial resolution can be very helpful for this task.
Another difficulty is the availability of synchrotron sources. In the meantime
there are several synchrotron sources available but still these are not so many and
their beam lines are highly booked. The use of a beam line requires long-term
booking and a high effort for the preparation of the measurements because the time
for the experiment is limited. These conditions restrict the use of synchrotron
sources for very special applications only i.e. for the daily use laboratory instru-
ments are preferred. They are available at any time, they can be used also for
repeated and long term measurements, but they have often not the high analytical
performance like synchrotron sources.
On the other hand synchrotron sources are not only used for l-XRF applications
but also for the examination of large sample areas. Also for these applications the
high brilliance can be very helpful for a high analytical performance. Here in
particular the polarization of synchrotron radiation can be advantageous due to the
possibility to improve the peak-to-background ratios caused by the reduction of
scattering in the polarization plane.
6.1.3 SEM-EDS
One of the most often used analytical methods for spatial resolution is the adaption
of an X-ray spectrometer to an electron microscope [12–15]. Since energy dis-
persive spectrometers are available with their high brightness caused by their large
captured angle of fluorescence radiation this combination is very popular because
it combines the possibility of highly resolved images with elemental analysis.
6.1 Comparison with Other Micro-Analytical Methods 215
In this method the sample will be scanned by the electron-beam of the micro-
scope and the different interactions are used to produce the image (by the secondary
or backscattered electrons) and to perform the elemental analysis by excitation of
fluorescence radiation. Therefore the sample volume that contributes to the image
or to the elemental information is identical and can be directly compared. SEM’s
are often used instruments because the inspection of sample surfaces is a usual
request and because they allow this function with a high performance. Electron
microscopes are completed to approx. 80 % with an EDS-system.
The comparison between l-XRF and SEM-EDS shows that they have different
spatial resolutions (Table 6.1) and also differences in the elemental range that can
be analyzed. But the main differences exists for the sample handling and for the
general analytical performance.
• Sample handling
The sample handling in l-XRF is considerable easier than for SEM-EDS sys-
tems. There are the following differences:
– The sample for l-XRF needs not to be electrical conductive. In a SEM the
excitation will be performed with electrons which transport electrical charges
to the sample. These charges have to be removed and therefore the sample
needs to electrical conductive. If this is not the case charging effects influence
the image quality as well as the quantification.
If a sample is not electrical conductive it is necessary to change this situation—for
example by a coating with Carbon or Gold. This is an additional effort for the
preparation of special sample types like glasses, minerals or plastics. With l-XRF
all these samples can be measured directly. This reduces the time for the sample
preparation significantly.
– Samples in SEM typically require a high surface quality due to the small
penetration depth of electrons into the material—the sample needs to be
carefully polished. This is not required for l-XRF. In general the direct
analysis of every sample surface is possible.
– The measurements medium for l-XRF can also be air or pre-vacuum. This
enlarges the sample types that can be analyzed and reduce the preparation
time for a measurement tremendous.
Due to the possibility to measure in air it is possible to analyze material which is
sensitive against vacuum or can be even damaged by vacuum, for example organic
samples but also simple liquids or wet samples like pastes, slurry etc. This enlarges
the applicability of l-XRF for a wide range of sample qualities.
The required vacuum for l-XRF is only to avoid the absorption of the fluo-
rescence radiation in air—therefore pressures in the range of 10–50 mbar are
sufficient. In SEM the absorption of electrons has to be avoided. Therefore the
vacuum needs to be in the range down to 0.01 mbar. This means that the effort for
vacuum sealing is reduced due to the three orders of magnitude difference in the
vacuum level. But an important difference is the pumping time in particular for the
large sample chambers that are common for l-XRF instruments.
216 6 Relations to Other Analytical Methods
– Finally there are also differences between the excitation with electrons and
with X-rays in regards to the sample stress. The absorption of electrons in the
sample is accompanied with a higher impact of energy into the material. This
heat up the material and stress it—even its damaging is possible.
On the other hand, the absorption of X-rays generates not a high energy
impact to at the sample. Therefore their heating is negligible and the sample
stress is reduced. That means for l-XRF higher excitation intensities can be
used.
So far it can be concluded that l-XRF offers an easier sample handling which
allows a faster access to the analytical result and enlarges also the applicability to a
wider range of sample qualities.
• Analytical performance
A comparison of the analytical performance shows that there are mainly dif-
ferences for sensitivity. There are two aspects—the sensitivity for the analysis of
traces and for the analysis of coating systems.
– Sensitivity for trace analysis
The sensitivity for trace analysis depends on the peak-to-background ratio of the
measurement. This is determined by the peak as well as by the background
intensity.
The background intensity for electron excitation is higher because the back-
ground is generated by the bremsstrahlung of the electron beam. For X-ray
excitation the spectral background is generated mainly by the scattering of the
bremsstrahlung of the tube on the sample. This is a secondary effect i.e. signifi-
cantly smaller than the direct generation of bremsstrahlung as demonstrated in
Fig. 6.2 which shows spectra measured with electron as well as X-ray excitation in
the same position of a stainless steel.
Both spectra are normalized for the Fe-intensity but the electron excited
spectrum (red) has a significantly higher background. The inscribed part of the
figure shows the high energy part where traces of Nb and Mo in the sample gives
still a detectable signal in the X-ray excited spectrum (blue) whilst in the electron
excited spectrum these peaks are completely covered by the background.
The other influencing factor for peak-to-background ratio is the peak-intensity.
This is determined by the amount of the element but also by its excitation effi-
ciency i.e. the excitation conditions and the cross-sections for the excitation. This
is demonstrated in Fig. 6.3 with spectra of the multi-element glass standard NIST
612 which are measured again with both excitation modes. This sample contains a
series of elements in the weight fraction range of approx. 500 mg/kg. The spectra
in this figure are normalized to the main component Ca. Here the red spectrum
with X-ray excitation shows for high energies a higher spectral background—the
scattering of tube radiation on the light matrix of the glass is strong. But never-
theless, this spectrum shows strong element peaks whereas the spectrum with
electron excitation (blue) doesn’t. For that situation the different excitation
probabilities for the excitation with electrons and X-rays are responsible.
6.1 Comparison with Other Micro-Analytical Methods 217
Fig. 6.2 Spectra of a stainless steel excited with electrons (red) and with X-rays (blue)—(square
root intensity scaling)
Fig. 6.3 Spectra of the glass standard NIST 612—electron excitation (blue) and X-ray excitation
(red)
beam. In case of X-ray excitation the energy of the excitation radiation needs
to be higher than the absorption edge of the analyzed element and the
information depth is determined by the absorption of fluorescence radiation in
the sample (see Sect. 5.2).
This situation is demonstrated in the spectra shown in Fig. 6.6 which were mea-
sured again with electron and X-ray excitation on a layer system consisting of a
Cu-track, a Ni-diffusion barrier and an Au-contact layer.
These spectra show that the electron excitation gives a very good signal from
Au (both for Au-M at 2.1 keV and for Au-L at 9.7 keV), the Ni signal is still
detectable but the Cu signal cannot be detected because it is not excited. For the X-
ray excitation the elements of all 3 layers are detected.
This means with X-ray excitation it is possible to have a ‘‘deeper’’ view into the
sample which
– allows a more representative analysis for bulk material. The information
depth for l-XRF has approx. the same dimensions like the spot diameter in
case of using poly-cap optics and
– offers the possibility to examine also coating systems with thicker layers or
even multiple layer systems.
From this comparison it can be concluded that both methods i.e. SEM-EDS and
l-XRF have a high potential for the analysis of non-homogeneous material and
both of them gives partly also complementary information about the examined
material.
220 6 Relations to Other Analytical Methods
Due to these general problems there are not so many examples for a successful
integration of different analytical method. Sometimes they are used in as unique
prototypes designed in universities or research institutes.
Nevertheless, for l-XRF there are combinations with other methods known and
can give some complementary information about the analyzed material. This
complementary information can be
• an extension of the element range of l-XRF to lighter elements— by a
combination with SEM-EDS [16, 17]
• an extension to structural information by the combination with l-XRD
• the extension to organic information with methods of molecule spectros-
copy i.e. methods with optical spectroscopy that can work with spatial
resolution. One of them could be Raman-spectroscopy [18, 19]
They will be discussed in detail in the following.
Fig. 6.8 Spectra of the glass standard excited with electrons (blue) and X-rays (red) (logarithmic
intensity scaling)
This is demonstrated for a PVC-sample in Table 6.2. For this sample are given
only the weight fraction of Ca and Zn. The measurement with l-XRF gives no
information about the light elements i.e. C and O but these elements have a high
weight fraction in the sample. Therefore the normalization to 100 % overestimates
the content of all other elements. The EDS analysis gives a result for C, O and Ca
but the result for other elements are wrong because they are close to the limit of
detection of EDS. If the EDS-results for the light elements i.e. C and O are used for
the FP-calculation in XRF the final results for Cl and Ca are very close to the given
weight fractions.
Other interesting sample types are oxides which contains a lot of oxygen that
cannot directly measured with l-XRF. The advantage of the combined quantifi-
cation will be demonstrated for a glass standard.
The spectra were measured with EDS (20 kV, 2 nA) and l-XRF (40 kV, 350
lA) in the same sample position and are displayed in Fig. 6.8. They are nor-
malized for approximately the same total intensity and have a logarithmic intensity
scale to display better their background. The electron excited spectrum (blue) has a
higher background in the low energy range but shows there a significant peak of
Oxygen. On the other hand, the X-ray excited spectrum (red) has a lower
224 6 Relations to Other Analytical Methods
Table 6.3 Quantification result for the combined analysis of a glass standard
Elem Given EDS l-XRF EDS + l-
wt% XRF
wt% Stat. wt% Oxide w Oxygen Stat.
error wt% wt% error
O 56.7 55.9 nd 49.16 55.9
Mg 0.088 0.56 7.5 0.56
Al 20.71 21.9 1.4 39.90 48.3 25.57 1 21.47
Si 20.19 19.2 1.5 50.70 47.4 22.16 0.9 20.10
P 0.152 0.50 7.2 0.32 0.25 0.20 60.1 0.095
K 0.42 0.42 4.5 1.40 0.55 0.46 3.3 0.402
Ca 0.095 0.08 17.2 0.29 0.13 0.09 8.0 0.081
Ti 1.11 1.17 3.6 3.79 1.96 1.18 0.6 1.03
Cr nd 0.06 0.024 0.02 3.6 0.014
Fe 0.32 0.33 7.9 1.31 0.54 0.38 0.6 0.329
Cu nd 0.03 0.11 0.09 3.0 0.007
Ga nd 0.03 0.10 0.07 3.0 0.006
Sr 0.153 nd 0.81 0.261 0.22 0.5 0.191
Zr nd 0.28 0.104 0.08 0.8 0.067
Ba 0.062 nd 1.14 0.394 0.34 2.5 0.309
nd not detectable
background—in particular for the lower energies but shows the peaks of the heavy
elements (for example Fe, Ta, Pb, Zr) with much higher intensity.
The different quantification results are displayed in Table 6.3. They show in the
second column of the first part the given weight fractions for all elements.
In the second part the EDS results are presented together with their statistical
uncertainty of for every element. These results are acceptable for the main com-
ponents but in particular for the heavy elements with small weight fractions there
are no results available because the concentrations are less than the LOD’s.
The third part displays the quant results for l-XRF calculated with different
models. The first column shows the result without consideration of Oxygen. Due to
the high weight fraction of Oxygen in a glass these results cannot be correct.
Therefore in the second column (Oxide) Oxygen is taken into account by the
stoichiometric relations of O to every identified element. But this are weight
fractions for oxides and cannot be compared with the given pure element weight
fractions. Therefore the third column (w Oxygen) shows the results after the
transfer to element weight fractions. This improves already the agreement between
given and analyzed results. But still there are significant differences in particular
for Oxygen because it is only indirect analyzed—differences of the oxidation level
of several elements could be a reason for the deviations. The last column of the l-
XRF results shows the statistical error of the measured intensities.
The last part of the table presents the results of the combined analysis. In that
case the results with the lower statistical uncertainty from EDS (here for O and
Mg) are set as given results for the XRF quantification. For all other heavier
elements the results are calculated new by a FP-model for XRF.
6.2 Combination of l-XRF with Other Methods 225
The comparison with the given results shows now a pleasant agreement with
the given results.
The demonstrated examples shows that the combination of l-XRF and SEM-
EDS can be very helpful for a complete material characterization. This combi-
nation can instrumentally easy realized by an X-ray tube with an X-ray optic as an
option on electron microscopes.
The combination of XRF and XRD offers the possibility of using the information
about chemical composition from XRF for the phase analysis with XRD. With that
information phase analysis is faster and more accurate because the variety of
phases with similar structure parameters can be significantly reduced by the
knowledge of chemical composition.
Several instrumental realizations for this combination are available—mostly
with conventional instruments i.e. with analyzed area in the range of \100 mm.
This combination can be integrated in one instrument [20, 21] or realized by two
separate instruments which are combined in a workstation [22]. They are used for
example in cement industries for the characterization of the raw materials as well
as the final products.
X-Ray diffraction can be performed also position sensitive. For that purpose
conventional instruments will be equipped also with X-ray optics—capillary optics
in combination with a filter or monochromator or focusing monochromators for
example from synthetic multi-layer optics. These instruments can be additional
equipped with an energy dispersive X-ray detector for the measurement of fluo-
rescence radiation. It is necessary to mention that the excitation with the mono-
energetic radiation which is required for diffraction is not the best choice for the
excitation of fluorescence radiation. But for a rough elemental analysis it can be
already sufficient. Another possibility would be the excitation without filter which
enlarges the energy range of the excitation spectrum.
atoms and allow the determination of these parameters. Also for Raman the
incident radiation—preferred Laser radiation in different energy ranges—can be
focused at a small sample area which allows the examination of the identical
sample area like with l-XRF.
A prototype for this combination was prepared for the testing the possibilities
for a simultaneous determination of inorganic and organic sample components
[23]. The l-XRF part was prepared by a l-focus tube with Rh-target and poly-cap
optic and with an SDD-detector with 30 mm3 sensitive area. The sample could be
positioned on a motor driven X-Y-Z-stage. The sample was observed by an optical
microscope with one magnification. The microscope was also used for focusing of
the incident laser beam—two lasers (with 633 and 785 nm) were available for the
optimal analysis of different sample compositions. The laser light as well as the
Raman-spectrometer was connected to the microscope via optical fibers. A model
of this prototype is displayed in Fig. 6.9.
With that prototype several analysis were performed for the examination of
pigments and also of geological samples [24, 25]. But in general it has to be
mentioned that the problems discussed in Sect. 6.2.1 are valid also for that for the
combination. The methods are very different both for the measurement strategy
and for the philosophy of data interpretation. This requires a deep understanding of
both methods for an optimal interpretation of measured data and to get correct
analytical results.
References
1. O. Brümmer, Mikroanalyse mit Elektronen- und Ionensonden (VEB Deutscher Verlag für
Grundstoffindustrie, Leipzig, 1977)
References 227
The small spot of l-XRF offers the possibility to analyze small sample areas—this
can be single particles or inclusions, a small area in a non-homogenous sample but
also a small area on a non-regular shaped sample. The presented applications will
demonstrate the different analytical requirements and the results for example for
the high accurate analysis of the precious metal content in jewelry, will discuss the
special conditions for coating analysis on homogenous but non-regular shaped
samples, will show few results for the analysis of small particles and inclusions as
example for non-homogenous samples as well as the analysis of restricted ele-
ments in consumer goods as example for both non-homogeneous and non-regular
shaped samples.
The monetary value of precious metals is very high and continuously growing.
They are used in various products—jewelry, dental alloys, electronic components
etc. Mostly they are used as alloys to reduce the material costs and adapt their
physical or electrical properties as hardness, plasticity, color or resistance to their
utilization. The high costs of precious metals request the exact determination of the
content of the valuable components in that alloys.
For the analysis of precious metal alloys few special requirements has to be taken
into account.
• High analytical accuracy
This requirement is caused by two reasons—the monetary value of the material
is very high and therefore an analytical error can have a considerably impact to
the material value. The gold price was growing by a factor of more than 4 in the
last 10 years into the range of[50 USD/g. That means an analytical uncertainty
of 1 % correlates to a material value of approx. 5 USD for a jewelry piece of
10 g. But what about heavier pieces like bars or bullions. From that point of
view the availability of a high accurate analysis is required.
The second reason is the availability of a high accurate analytical method
especially for the determination of the Gold content—the fire assay method. The
high value of Gold requested already in the ancient time the availability for an
accurate determination of its weight fraction. The end of this long history is a
very exact method—the fire assay or cupellation. The first descriptions if this
method can be found already in the bible which means that it has a history of
more than 3,000 years.
For fire assay a small amount of the sample (approx. 0.25 g) is covered with a
Pb-foil and heated in an air stream to 1,000–1,200 C. In that case the alloy is
melted and the non-precious metals are oxidized and go into the ceramic cup
which supports the piece of metal. After cooling down a small ball of precious
metals is available which have to be hammered to a thin film which then is
cooked in a nitric acid to dissolve the Silver from the alloy. An accurate
weighing of the material in all steps of the process, i.e. of the original sample, of
the ball after oxidation and of the final metal film allows the determination of
the content of non-precious components, of the Silver and of Gold. If necessary,
a further dissolution process in aqua regia removes the Platinum from the Gold
and allows also the determination of the weight fraction of Platinum. Due to the
high accuracy and sensitivity of balances which allow the determination of the
weight down to the lg-range the precision and trueness of fire assay could be
improved down to 0.02 wt% which represent an very high accuracy. But fire
7.1 Point Analysis 231
assay is destructive because the sample has to be damaged to get the material for
the analysis and the analysis requires also a long time—in the range of few
hours—which not allows a short feedback to the manufacturing process.
Further fire assay gives only summary information about all alloying elements
but the concentration of these elements determines color, hardness etc. So far
other analytical methods which are faster, give information about all elements in
the sample and have comparable analytical accuracy are requested.
Fig. 7.1 Piece of a filigree jewelry (left) and magnified image of a small part (right)
demonstrating the request for small spot analysis
If the elements are correctly identified the energy resolution of the detector have
only a small influence to the analytical performance. The fitting or deconvo-
lution procedures are working very well for the peak area determination also in
case of overlapping peaks.
The use of l-XRF instruments for the analysis of jewelry alloys was discussed by
Brill and also others [1–6]. It was found that for jewelry this method offers a good
analytical performance. But the required accuracy is only possible with a standard
based quantification model. The range of weight fractions in jewelry is rather large.
The Au-content can vary from 8 to 24 karat i.e. 33.3–99.9 wt%, respectively.
Covering this wide concentration range requires the splitting of the calibration
ranges in several smaller ones. This is for Au-alloys in jewelry unproblematic
234 7 Applications
Fig. 7.2 Spectra of the same jewelry measured with and without a surface cleaning
because the karatage have few preferred ranges—8, 10, 14, 18, 22 and 24 karat,
which corresponds to weight fractions of 33.3, 41.7, 58.5, 75.0, 91.6 or 100 %,
respectively. That means mainly for these weight fraction ranges the calibrations
are required.
The varying weight fractions of Au in jewelry influence its fluorescence
intensity and therefore the analytical accuracy too due to the changes of the
statistical uncertainty. This is demonstrated with the results in the third column of
Table 7.2 which shows the average relative standard deviation for Gold for the
corresponding weight fraction range. The measurement conditions were: 40 kV,
1 mA and 60 s. The statistical uncertainty of the Au-intensity drops down with
increasing Au-weight fraction.
The other columns show average deviations of a large set of different jewelry
alloys (80 samples). The measurements were performed with instruments with
different detectors—with prop-counter (PC) and silicon-drift-detectors SDD. The
calibration for both instruments was prepared in the same way. The qualitative
composition was assumed to be known i.e. overlaps of unknown peaks had no
influence for the measurements with prop-counters. The measurement condition
for both detectors was also 40 kV, 1 mA and 60 s as mentioned above.
It can be seen that the average deviation for the different weight fraction ranges
improves as expected for higher weight fractions of Gold. The results for the
measurements with the SDD are slightly better even although the measured
intensity with a prop-counter is slightly higher and therefore its statistical uncer-
tainty is smaller. The smaller uncertainty for the SDD is a result of the reduced
peak overlap which simplifies peak area calculation. This compensates or even
slightly over-compensates the smaller statistical uncertainty of the prop-counter.
The last line of the table shows the average (absolute) differences between
analyzed results and given results for the both detectors.
The same results are displayed in Fig. 7.3. Here the absolute deviations from
the given weight fraction are displayed in dependence of them for every mea-
surement with both instrument setups.
7.1 Point Analysis 235
Table 7.2 Average deviations for measurements of Au-alloys in dependence of the Au-weight
fraction
Au range wt% Karatage range Statistical Std-dev Average deviation Average deviation
for Au (%) PC (wt%) SDD (wt%)
33–45 8–12 0.45 0.24 0.24
24–60 12–16 0.37 0.17 0.14
60–85 14–20 0.33 0.15 0.14
85–100 20–24 0.30 0.09 0.07
Average 0.134 0.122
0.5
-0.5
-1
-1.5
30 40 50 60 70 80 90 100
Given concentration / Wt-%
It can be seen that more samples with high weight fractions were available. The
deviations are in the range of\0.2 % for high gold weight fractions—as discussed
due to the smaller statistical uncertainty and slightly larger for low weight frac-
tions. The large number of samples with similar weight fractions (58.8, 75.0,
91.6 wt%) represents the often used karatage in jewelry alloys. The difference
between the two detector types is negligible.
The analysis of pure Gold is an usual analytical task for the determination of
chemical purity of the material. The material value is not so much different but the
effort for purification makes a considerably difference in costs. The challenge is to
distinguish between weight fractions of 99.5 and 99.9 wt% of Au. The determination
of these small differences is difficult because the analytical accuracy for the statis-
tical uncertainty for Gold is already in the range of 0.1 % and additional uncertainty
contributions in the same range by other influences has to be expected which then
goes into ranges of the differences of the weight fractions that have to be determined.
Therefore it is easier to determine the Au-weight fraction by difference. That
means, all traces are identified and quantified and their results are subtracted from
100 %. This result can be more accurate than for the direct analysis. For an
236 7 Applications
Fig. 7.4 Spectra of two Au-alloys with 0.1 wt% (blue) and 0.5 wt% (red) Ag, measurement time
30 s (square root intensity scaling)
instrument with SDD the limits of detection are very good—in the range of
\100 mg/kg. For weight fractions close to the LOD the uncertainty can be in the
range of half of the weight fraction i.e. 50 mg/kg. That means the uncertainty for
the Au-weight fraction by subtracting the trace concentration would be only
0.005 % i.e. significantly less than for the direct measurement. Even if there are
several traces the accuracy is improved for the quantification by difference. This is
also demonstrated in the following example.
Figure 7.4 shows two spectra with different Ag-weight fractions—0.1 and 0.5 wt%.
The Ag-peak can be clearly identified and the peak intensities show significant dif-
ferences. The measured intensities are for Au 36,414 and 36,200 counts, respectively,
with a statistical sigma-value of approx. 200 counts. That means the a-ranges are
overlapping. The intensities with their sigma-values for Ag are 79 + 9 counts and
264 + 16 counts, respectively. That means here a clear differentiation between the
alloys is possible and the statistical uncertainty in sum is significantly smaller.
So far the analysis by difference can be much more sensitive and accurate than
the direct analysis in case of high weight fractions and this is valid not only for
jewelry alloys.
Coatings are common in more or less all technical materials. They are used for
example for corrosion protection as Zn- or Cr-layer, they are used for functional
reasons for example to warranty a good electrical contact by Au-layers or to allow
a high efficiency of a catalyst for example by Pd- or Pt-layers and they are also
used for decorative reasons for example as Cr- or Ni-layer on automotive parts,
7.1 Point Analysis 237
consumer goods or instruments. All these coatings have to be controlled for their
correct thickness and in the case of alloyed coatings even for their composition.
The plating industry use XRF for quality control already for many years [7–14].
This testing needs to be non-destructive because it has to be performed on final
products which should be not damaged by the quality control. Further these
measurements very often can be performed only on small areas because final
goods are not homogeneous or mostly non-regular shaped without large flat areas.
These conditions can be fulfilled by l-XRF. The measurement is non-
destructive and the size of the analyzed can be adapted to the analytical task by
using collimators with spot sizes in the range from 0.2 to 2.5 mm. For smaller spot
sizes capillary optics are required for sufficient intensity and then the spot sizes can
go down into the range of few 10 lm.
The variation of different coatings systems is endless and continuously new
coating systems are developed. So far, it is also necessary to have a flexible
analytical method that can be easily adapted to the different coating systems. The
conditions that have to be take into account for the XRF-analysis of coating
systems are already discussed in Sect. 4.4.2.
For coating analysis very often simple instruments can be used which are
working with prop-counters because coating analysis mostly is a quality control
and the number of elements of the coating are limited and known i.e. peak overlaps
can be easily handled. This is advantageous because these detectors have a large
sensitive area. Due to the thin layer the fluorescence signal can be small which
requires the accumulation of the fluorescence radiation from a large solid angle.
The limited energy resolution of prop-counters which can cause strong peak
interferences is no problem because the elemental structure of the coating system
has to be known at all for a quantification and often coatings are manufactured
only of very few elements. Nevertheless, the development to very thin layers in
micro-electronics requires a better peak-to-background ratio of the measurement
which can be realized only with high resolution detectors as SDD’s.
In the following few typical examples for coating analysis are demonstrated.
For corrosion protection very often Fe is coated with Zn [7, 8]. The thickness of Zn
has to be monitored because too thick layers are too expensive and have the ten-
dency to spall and too thin layers assure no sufficient protection. The fluorescence
energies of Fe and Zn are well separated even for the measurements with prop-
counters. This can be seen in Fig. 7.5a. The image shows the measured spectrum as
red dots. The fitted pure elements spectra are split into the K-peaks for Fe (blue) and
Zn (red). The sum spectrum (gray) fits the measured spectra very well. A splitting of
the K-lines in Ka- and Kß-peaks can be helpful because the theoretical intensity ratio
between Ka- and Kb-peaks can be influenced due to secondary excitations and
different self-absorption in the layer which would deteriorate the fitting result.
The calibration curve for the thickness prepared with four reference samples is
displayed in Fig. 7.5b. The diagram shows the measured intensities both of the
238 7 Applications
Fig. 7.5 Spectra of a Zn_Fe coating measured with prop-counter (left). Measured intensities and
calibration curve for Zn-thickness (right)
layer (squares) and the substrate (diamonds—left hand scale). Both intensities are
changing as expected non-linear—the layer intensity is increasing and the sub-
strate intensity decreasing with thickness. Further the relation between given and
analyzed layer thickness is displayed (open circles—right scale). This example
demonstrates that for single layers a high accuracy is possible.
The measurement conditions influence the analytical results also for coating
analysis. Here the influences of the used analytic line, of the energy resolution of
the detector, of the spot size and the working distance are demonstrated. These
influences are valid not only for the examined layer systems but in general.
The examinations are performed on different Au-layer systems. This is an often
used system for electric contacts because it warranties good contact properties for
a long time due to their negligible corrosion. But the high Gold price limits the
layer thickness—they should be as thin as possible but on the other hand thick
enough for avoiding damages by multiple plugging [11].
• Influence of the analytic line
For the analysis of Au different analytic lines are available—the Au-L- as well
as the Au-M-line. The Au-M-line has a higher sensitivity for thin layers but the
maximum thickness for an analysis at this energy is restricted to less than 1 lm
(see Sect. 4.4.4). Further the low energy of Au-M prefers measurements in
vacuum. But this requires a higher instrumental effort and a longer analysis time
7.1 Point Analysis 239
with the standard deviation of the statistical uncertainty. For both lines the
repeatability is very good but due to the higher intensity of the Au-M-line their
statistical uncertainty is smaller and accordingly the intensities are less scattered.
These results are summarized in Table 7.3. It shows the mean intensities, their
standard deviation as absolute and relative value, the statistical standard devi-
ation and the relation between measured and statistical standard deviation.
These values show that for this layer the intensity of Au-M is by factor of
approx. 3.5 higher than the Au-L radiation which is caused by the very small
layer thickness for that the because for Au-M radiation it is already close to
saturation, for Au-L this is still a thin layer. Therefore both the measured and the
statistical standard deviations are better for the M-line.
The relation between measured and statistical standard deviation for both Au-
intensities are approximately the same i.e. additional uncertainty contributions
are negligible. For the thicker but hidden layers of Ni and Cu the intensities are
higher and the statistical uncertainty smaller but due to the influence of the
coating with Au the measured standard deviation is larger than the statistical
value—for both elements in the same range.
Fig. 7.8 Spectra of an Au–Ni-layer measured in air with prop-counter (black) and SDD (gray)
and correspondingly a larger captured angle of the prop-counter (see also Sect.
2.5.5.2). The higher intensity results in a smaller statistical uncertainty for the
prop-counter but the ratio between standard deviation and statistical uncertainty
is smaller for the SDD caused by the smaller uncertainty contribution from the
spectra evaluation due to the better peak area calculation by the less peak
overlaps. The thickness values for both measurements are comparable.
Further the improved peak-to-background ratio caused by the better energy
resolution of the SDD gives a higher sensitivity i.e. the limit of detections are
better by approx. a factor of approx. 20 as shown in the last row of Table 7.4.
quantification result. By using the relative mode (see Sect. 4.4.1) for quantifi-
cation both signals i.e. that from the layer as well as that from the substrate are
used. This allows the compensation of the influence of different working dis-
tances. This is demonstrated in Fig. 7.9 which shows the dependence of
Au-layer thickness from the working distance. Calculated both with the com-
mon emission mode and the relative mode
These results show that changes of the working distance by +5 mm would
influence the thickness result for the normal emission mode by more than +35 %
(black squares). But for the relative mode (gray diamonds) the thickness values
are significant less changing. Even for this large working distance range the
values for the Au-layer thickness are within approx. +10 %.
Table 7.5 Results of coating analysis on Au–Ni_Cu coatings with emission and relative mode
Colli/mm Degree of overlap Emission mode Relative mode
Au/lm Ni/lm Au/lm Ni/lm
0.1 9 0.3 100 % 0.248 2.17 0.256 2.47
0.2 9 0.2 approx. 95 % 0.218 1.92 0.261 2.52
0.5 9 0.5 approx. 80 % 0.181 1.65 0.265 2.51
[ 0.7 approx. 70 % 0.156 1.55 0.273 2.49
Fog ultra-thin coatings the repeatability can be reduced due a larger statistical
uncertainty from the small collected intensities. But still a high precision is
requested and even possible. This will be demonstrated for an Au–Pd–Ni_Cu layer
system. These systems are used for lead frames [15]. Previously lead frames were
partly coated with Ag for a good contact of the bonding wires. In that case, only
the end of the fingers of the lead frame was coated. This requires an expensive
masking of the fingers. A cost reduction was possible by a coating the complete
lead frames with ultra-thin Au-layer mostly on top of an ultra-thin Pd-layer. The
thicknesses of these layers are in the range of 10–30 nm. Such a lead frame is
displayed in Fig. 7.10 together with the spot that has a diameter of 30 lm.
Spectra measured in this point in air (red) and in vacuum (blue) are shown in
Fig. 7.11. It can be seen that for this thin layers the lower energetic lines of Au (M-
line) and Pd (L-line) have a higher intensity than the higher energetic Au-L- and
Pd-K-lines. This effect is supported if measurements can be performed in vacuum
due to the reduced absorption of the low energetic radiation in air. For very small
spots further the measurement of Pd-K-radiation is not very efficient because the
excitation in this case is limited by the low transmission of high energetic radiation
in poly-cap optics.
The repeatability was tested by 50 measurements with a measurement time of
30 s. These results are presented in Fig. 7.12 and Table 7.6. They show that the
reproducibility for the thin layers of Au and Pd are rather good. The standard
deviation for Au is in the range of 2 % and only slightly larger than the statistically
expected value. For Pd the fluctuations are slightly larger due to its lower intensity.
The standard deviation for the measurement of Pd is also only slightly larger than
the statistically uncertainty. The intensities of Ni and Cu are significantly higher
than for Au and Pd and therefore their fluctuations are less. But here the relation
between measured and statistical fluctuations is worse than for the upper layers
because their intensities are influenced by the absorption in the upper layers.
This example shows, even very thin layers can be analyzed, in particular, if low
energetic radiation can be used for the evaluation. This radiation has a higher
sensitivity for thin layers as discussed in Sect. 4.4.4. The reproducibility depends
on the collected intensity that can be increased by an improvement of the exci-
tation conditions or by a longer measurement time.
244 7 Applications
Also the analysis of very complex layer systems is possible with XRF. As an
example the examination of compound layers of solar cells will be discussed [16].
The typical structure of a CIGS-cell is displayed in Fig. 7.13. As substrate mostly
float-glass is used which is coated with a Mo-layer as the back-contact. The next
layer is the absorber which is prepared from compound semi-conductors like
CuInGaS or CuInGaSe (CIGS). The next layers are the upper CdS-contact and a
passivation with ZnO. For that structure both the thicknesses of all layers and
especially the composition of the absorber are interesting. This composition is
important because it determines the efficiency of the solar cell. Therefore the
control of this composition is required preferably with a short feedback to the
manufacturing process. This is possible if these measurements are performed in
the manufacturing line close to the coating process.
A typical spectrum of a complete CIGS is displayed in Fig. 7.14. It shows all
expected lines from the elements of the solar cell well separated.
For a fast feedback to the manufacturing the measurements often are performed
only for the absorber layer. Then the further coating processes can be avoided in
case of a wrong absorber composition. Additionally then all the analytical task is
simplified—the absorber has to be analyzed only. The quantification for alloyed
layers can be performed with the described models (Sect. 4.4). An example is
given in [16]. This paper compares the XRF results for composition and thickness
with reference methods as SEM-EDS for the absorber composition and with
profilometer results for its thickness. These results are summarized in Table 7.7.
They show a good agreement for the composition of the absorber measured in
different manufacturing levels—only the absorber without any further coating,
absorber with CdS-layer and finally with both CdS- and ZnO-layer—as well as
with the SEM-EDS measurement for the absorber only.
The thickness determination was performed with XRF also for the different
manufacturing levels. Also these results are consistent to each other and agree with
the result of a profilometer [17, 18].
This comparison shows an excellent agreement between the different analytical
methods. Compositions as well as thicknesses are measured with reference methods
give the same results like XRF. But the reference methods are destructive—the
sample has to be cut for a positioning in the SEM and the absorber can be deposited
246 7 Applications
only on a part of the sample to use the profilometer—the other part has to be
covered. Further these measurements require a long time both for sample prepa-
ration and for sample positioning while XRF is non-destructive as well as very fast.
7.1 Point Analysis 247
Fig. 7.15 Variations of thickness of the CIGS-absorber and the Cu-content in a 6 9 6 matrix
from an area of 200 9 200 mm2
Because only small sample areas are measured—for this example the spot size
diameter was approx. 1 mm—it was possible to investigate even the homogeneity
of the coatings by measurements in different positions of the large and flat solar
cell. The result of a 6 9 6 multi-point measurement for the absorber thickness and
the Cu-content is presented in Fig. 7.15. It can be seen that there are severe
inhomogeneities. The thickness of the absorbe is barrel-shaped and drops down to
the border. Also the composition of the absorber shows variations—the Cu-content
has a minimum in the upper right corner and is growing into the lower left corner.
Information’s about these inhomogeneous distribution allow the control of the
parameters for the manufacturing process.
Coating measurements typically are performed in air because often the results are
used for quality control of the manufacturing process and therefore the analysis
time should be as short as possible. But if the coatings are prepared from light
elements for the analysis vacuum is required and for organic coatings even this
would give no results. But measurements in vacuum would prolong the analysis
time by the puming time. The instrument costs would be also higher. For special
applications that has to be done—for example in wafer manufacturing. But it is
also possible to use the absorption signal for the thickness calculation. However,
this mode works only for single-layers and the signal from the substrate must have
a sufficient dynamic range i.e. the absorption of the layer must be strong enough to
decrease the peak intensity but should not absorb the complete substrate radiation.
As an example measurements on Al_Fe coatings will be demonstrated. Fig-
ure 7.16 shows the spectra of Al-layers of different thickness on a Fe-substrate
measured in vacuum and in air.
These spectra were measured with same measurement conditions (40 kV,
200 lA, 20 s with SDD) on two samples with an Al-layer on Fe but with different
thicknesses (12.5 and 100 lm). Both samples were measured in vacuum. The
green spectrum is from the sample with the 100 lm Al-layer and the blue spectrum
from that with the 12.5 lm layer. This sample was measured also in air (red
248 7 Applications
Fig. 7.16 Spectra of different Al-layers on Fe measured in vacuum and in air (intensities in a
square root scale)
spectrum). This spectrum has an Ar-peak due to the excitation of the Ar in the air
(approx. 1 wt%) between the sample and the detector. In general, this Ar-peak can
be used as a proof for the measurement in air.
The measured Al-intensity depends on the layer thickness as well as on the
absorption of this low energetic radiation on the way from the sample to the
detector. The absorption for Al in air is very strong—the intensity difference
between the red and the green spectra for the same layer thickness but different
measurement medium is approximately a factor 30 which corresponds to the
results in Fig. 2.57. The two Al-peaks measured in vacuum show not significant
intensity differences because these thicknesses are already larger than the satura-
tion thickness for Al which is approx. 4 lm (see Table 4.6).
That means that the dynamic range for the Al-emission signal is relatively small
but have a high slope which offers a high sensitivity—but the measurements needs
to be performed in vacuum.
The Fe-intensity varies strongly between the sample with the thick Al-layer
(green spectrum) and the sample with the thin Al-layer (red and blue spectrum). This
difference is caused by the absorption of Fe-radiation in the Al-layer—the thicker
layer has a stronger absorption. Additionally, the two spectra of the sample with the
12.5 lm Al-layer show no significant intensity difference for Fe, independent of the
measurement medium. The absorption of Fe-K-radiation in air is very low for the
usually short distances between sample and detector, only very few per cent (see
also Table 4.6). Thus use of this absorption mode can reduce the instrumental costs
tremendous because an evacuable sample chamber is not required.
If the absorption signal of Fe has used for the determination of the Al-thickness
also the dynamic range of the signal is larger. The Fe-signal would be absorbed for
7.1 Point Analysis 249
Fig. 7.18 Images of particles with different size but same composition
Then it is possible that the particle support contributes to the signal or the quan-
tification result is influenced by the analyzed volume [19–24].
Particles of same composition but of different size were analyzed for a test of
this situation. The particles are machining chips and they are shown in different
sizes in Fig. 7.18.
The particles are positioned on a free thin plastic film as described in Sect. 5.3.2
to avoid an influence from the sample support. The spectra of these particles are
displayed in Fig. 7.19. They were measured with 50 kV, 200 lA and 30 s with a
spot size of 25 lm but normalized to the same Fe-intensity.
The spectra (green: large particle, red: medium particle, blue: small particle)
show in general the same elements—namely Ca, Mn, Fe, Cu and Zn but the
spectral background is very different. For a better display of these differences a
logarithmic intensity scale is used. It can be seen the smaller the particle the higher
the spectral background. The peaks at 12.8 and 13.5 keV are the sum peaks of Fe.
Their intensity drops down if the particle size is reduced because less material
produce less fluorescence intensity and therefore also a lower pile up.
The quantification results with a standardless model are summarized in
Table 7.8. They show a very good agreement between the different sample sizes.
Even the smallest particle which has a size of approx. 100–150 lm only gives a
comparable sample composition.
For metallic particles the infinite thickness for the different elements is in the
range of few ten lm i.e. mostly smaller than the particle size. Therefore the
collected element intensities are only slightly influenced by the particle size. A
larger influence to the measured spectra is given by contaminations (here for
example by Ca) and by the surface roughness of the sample.
This situation can be changed for the analysis of small particles from light
material, for example from like glass. Then the information depth for several
elements in the light matrix can be larger than the particle size. That means the
measured intensities do not only depend on the weight fraction of an element but
also on the size of the analyzed particle. In that case a conventional determination
7.1 Point Analysis 251
Fig. 7.19 Spectra of particles with different size but same composition (logarithmic intensity
scale)
Fig. 7.20 Information depth in dependence of energy for the two glass standards
Fig. 7.21 Fluorescence intensity of Si-K (left) and Fe-K-radiation (right) of glass particles in
dependence of particle size
Fig. 7.22 PCA plots that show a clear separation between the different particle materials (left)
and which allows the determination of particle size for given materials (right)
high organic content by evaluating the complete spectrum [31–33]. In these cases
the probability for a correct material identification was very high.
Also for the characterization of particles this procedure is an appropriate
approach. It was applied to the measurement data from glass particles of different
composition and size [26]. The results presented in Fig. 7.22.
The left score shows that the different particle materials are concentrated in well
separated clusters which would allow the distinct determination of the different
materials. For that investigation not only the two mentioned glass materials (B2
and D2) but also 4 further particle materials were used. The plot on the right side
shows a PCA-score only for B2 and D2 which allow even a separation for different
particle sizes. The points in this diagram are arranged on a line and show a
dependence of particle size—smaller particles have a reduced value for the used
principle components. That offers the possibility even for an estimation of the
particle size with PCA.
This example shows that even for analytical tasks which show not a clear
relation between measured intensity and weight fraction a successful data evalu-
ation is possible. Chemometric procedures can be a powerful tool for this type of
data evaluation. But they need to be adapted to the special analytical question and
it is necessary to train the system with the measurement of a lot of data. This can
be a high effort but it has to be forced only once at the beginning of the analysis.
Fig. 7.23 Image of a particle in a plastic film (spot diameter 30 lm) (left) Spectra of the
inclusion (gray) and the plastic film (black) (right)
The control of consumer goods for hazardous components obtains more and more
significance. This is necessary to protect people against poisoning. Consumer
goods mostly are composed of different components and they are also non-regular
shaped. l-XRF is for these analytical problems therefore well adapted.
There are several regulations for restricted substances which define the maxi-
mum contents of these components in different materials. These limits can vary for
different countries and for different purposes. The most often used ones are:
• RoHS (Restriction of Hazardous Substances in consumer goods) [33] which
restricts the use of six hazardous materials in the manufacture of various types
of electronic and electrical equipment for example household appliances, con-
sumer equipment, lighting equipment, electronic and electrical tools, toys, lei-
sure and sports equipment, medical devices, semiconductor devices. The
maximum weight fractions are displayed in Table 7.10.
• A similar rule is given with WEEE (Waste of Electrical and Electronic
Equipment) [34] which supervise the same substances.
The determination of substances like Cr+6, PBB and PBDE are not possible with
XRF but it is possible to determine the complete content of Cr or and of Br. If
these contents are smaller than the given limits also the restricted substances
must be within the regulation.
This condition has to be fulfilled for all homogeneous components of the
mentioned articles. Because these components can also be very small they have
to be checked also on small sample areas. Therefore l-XRF is an appropriate
analytical tool for these problems.
• The toy guide-line is a new rule from the European Union to protect children
against hazardous substances. There are limitations for several organic flavors
but also for several elements for different conditions. These are listed in
Table 7.11. The limits are separately defined for different material qualities.
Partly they are relatively low which can offer problems for the analysis with
XRF, for few of them the analysis will be difficult or even not possible (B—
because this element is too light, Co, Ni, Se, Sb in liquids and lacquers, Cd, Hg
and Pb in general due to their low limits). On the other hand a lot of elements
can be monitored and a dwell control very often is sufficient in that cases.
The sensitivity of XRF depends on the element of interest but also on the mea-
surement conditions. In particular for l-XRF also the type of beam shaping—col-
limator or poly-cap optic—influences the excitation efficiency because the excitation
spectrum depends on the transmission function of the optic. In the case of using X-ray
optics the spot can be smaller and the excitation intensity for medium energies is
higher but the limited transmission of high energies limits the sensitivity for elements
that has to be measured at high fluorescence energies as for example Cd and Sb.
Anyway, the analysis of these elements requires a good energy resolution which
is possible only with Si-detectors i.e. PIN-diodes and preferred SDD’s. As an
7.1 Point Analysis 257
Table 7.11 Weight fractions Element For dry and ductile In liquids In scraped
of restricted elements in toy toy material or lacquer toy materials
materials according to toy layers
guide-line in mg/kg
B 1,200 300 15,000
Al 5,625 1,406 70,000
Cr+3/Cr+6 37.5/0.02 9.4/0.005 460/0.2
Mn 1,200 300 15,000
Co 10.5 2.6 130
Ni 75 18.8 930
Cu 622.5 156 7,700
Zn 3,750 938 46,000
As 3.8 0.9 47
Se 37.5 9.4 460
Sr 4,500 1,125 56,000
Cd 1.3 0.3 17
Sn 15,000 3,750 180,000
Sb 45 11.3 560
Ba 4,500 1,125 56,000
Hg 7.5 1.9 94
Pb 13.5 3.4 160
example a reference sample for RoHS in a polyethylene matrix and weight frac-
tions of 1,000 mg/kg for Cr, 1,100 mg/kg for Br and Hg, 1,200 mg/kg for Pb and
300 mg/kg for Cd were measured both with a collimator (1.5 mm, 50 kV, 800 lA,
60 s measurement time) and with a poly-cap optic (30 lm, 50 kV, 600 lA, 30 s).
The spectra are displayed in Fig. 7.24 with a logarithmic intensity scale to show
the complete range of absolute intensities for the both measurement conditions.
The intensity with poly-cap optic is significantly higher but the spectral back-
ground is also very high caused by the scattering of tube radiation on the light
sample matrix. Here the tube has a thin Be-window which transmits also low
energies. On the other hand, the spectrum from the collimator in general has
smaller intensities even for the larger spot but also the background intensity is
reduced because the tube spectrum is absorbed by the glass window of the tube.
258 7 Applications
The limits of detection determined from these spectra are for Cr, Br, Hg and Pb
much better than the RoHS requirements. For Cd it is close to this limit both due to
its lower concentration but also because the excitation efficiency of poly-cap optics
is reduced for high energies caused by its transmission function.
The sensitivity for the analysis of Cd can be improved by
• a larger collimator which utilize a larger content of tube radiation for the
excitation and generates a higher fluorescence intensity
• a higher tube voltage which increase the excitation efficiency in particular for
high energetic elements
• a longer measurement time which allows the accumulation of a larger amount of
fluorescence radiation
• the use of filters which reduce the spectral background but also the analyte
intensity due to a lower excitation efficiency. The final peak-to-background ratio
for Cd than depends on the filter material and thickness.
With a collimator excitation the lowest limits of detection for Cd are less than
the given RoHs-threshold value. The dependence of LOD from measurement time
and spot size are shown in Table 7.12. It shows the expected improvements for
larger spots and longer measurement time i.e. for a higher collected intensity.
The LOD for the poly-cap is worse due to the non-efficient excitation. But for a
long measurement time they can fulfill the RoHs-conditions too. Another possi-
bility in that case is the use of the Cd-L-radiation which will be efficiently excited
with a lens but in this case there can be overlaps with elements like K or Ag and if
the measurement is not performed in vacuum also with the Ar-line of air.
In general it can be concluded that l-XRF has a high potential for the exam-
ination of consumer goods for restricted and hazardous substances. The possibility
for a non-destructive analysis of samples which are non-homogeneous and non-
regular shaped with a sufficient sensitivity predestinates this method for that
analytical task.
Fig. 7.24 Spectra of a PE sample with restricted elements (logarithmic intensity scaling)
measured with a collimator (red) and a poly-cap optic (blue)
Table 7.12 LOD for Cd with Spot size Measurement time Measurement time
different measurement 100 s (mg/kg) 300 s ( mg/kg)
conditions (collimator)
0.7 mm 55 19
1.7 mm 25 10
2.5 mm 21 9
30 lm (poly-cap) 145 85
Fig. 7.25 Spectra of a powder sample measured in two single positions for 60 s (dots) and in 60
randomly distributed positions for 1 s
Table 7.14 Weight fractions of main components for area analysis in wt%
Element Na2O MgO Al2O3 SiO2 K2O CaO TiO2 MnO Fe2O3 RMS-value
Given 3.4 0.76 14.2 72.3 3.95 2.20 0.26 0.06 2.03
JG1-2 8.77 0.43 16.0 61.4 6.01 4.51 0.31 0.07 2.33 0.9785
JG1-1 11.10 0.03 17.4 60.1 4.48 5.02 0.18 0.04 1.51 0.9597
JG1 sum 0.60 0.71 13.4 74.8 4.35 2.96 0.39 0.08 2.56 0.9978
small amount of material is used for the tests which saves material costs. On the
other hand this requires analytical methods for the material characterization which
are fast and needs only a small amount of the sample material. l-XRF can do this
for the determination of chemical composition. But also other analytical methods
like IR-Spectroscopy for the determination of organic components or l-XRD for
the elucidation of the structure are used.
High throughput screening will be performed in two ways—with continuously
and typically small changes of the sample composition or by a selection of samples
264 7 Applications
from a randomly distributed sample set. For both methods an example will be
presented.
Another possibility for high throughput analysis is the selection of a special sample
from a randomly distributed sample variety. This can be used for example for the
development of pharmaceuticals [36–39].
7.2 Multiple Point Analysis 265
The variety of different agents is endless. Their effect very often is known but
their efficiency depends mainly on the reaction with proteins, for example, in the
human body. Proteins and also agents are composed mostly of hydrocarbons which
cannot be detected with XRF. But typically they have also few elements that can
be measured with X-ray fluorescence like halogens i.e. Cl, Br or I or others like P
or S for the agents and P, S, Zn, Se or others for the proteins. There are samples
available which are prepared from small polystyrene balls with a diameter of only
80 lm which are loaded with only one but always with a different agent. These
balls can be rubbed into small holes in a plastic film with a grid of holes. If this
grid will be brought into contact with only one protein the different drugs react
with the protein but every with a different efficiency. A high efficiency should be
recognized by a high content of the special elements of the protein because then a
lot of these proteins are accumulated on the agent.
Because the polystyrene balls have a small diameter the grid can be very
narrow. For the presented example the holes have a diameter of 0.1 mm and their
distance is 0.5 mm. That allows to place 10,000 holes on an area of 50 9 50 mm
with 10 9 10 single grids with 10 9 10 holes each. This is a very high number of
samples that can be analyzed only automatically. The measurement time for a
single point should be in the range of 1 s or even less. Even than the total mea-
surement time is in the range of several hours. The short measurement time
requires a high excitation intensity, for poly-cap optics no problem.
The elemental distribution for halogens of such a sample holder is displayed in
Fig. 7.31 [40]. This image shows that not all holes in the sample holder are filled.
Most of the points have a similar brightness but very few of them are significantly
lighter. For the corresponding agents a high efficiency for the used protein can be
expected.
The best possibility for data evaluation here is the display of intensity ratios for
different elements combinations—for example of the protein-element versus an
element of the agent. Such an intensity ratio plot is displayed in Fig. 7.32.
This example shows three different types of reactions:
• one with more or less no agent signal independent of the substrate signal. These
are the points close to X-axis, they could be caused by the empty holes.
• another set shows a more or less proportional behavior between the substrate
signal and the agent signal, they can be caused by a normal or no reaction
between agent and protein.
• a third group consisting of only very few points are close to the ordinate but with
a high agent signal which let expect a high reactivity between agent and drug.
The positions of these points have to be identified that the balls with the agents can
be separated from the sample holder for a further determination of the agent
composition by sequencing.
7.2 Multiple Point Analysis 267
This example shows that high throughput screening can not only be used for the
determination of the compositions of a high number of samples but also for the
selection of samples with a special composition. l-XRF allows a high efficiency
and low costs for that analysis even for extremely large sample number.
268 7 Applications
The slag of a steel cooking furnace is in a direct contact with the melt. Therefore
the alloy elements of the steel diffuse into the slag. The amount of these lost
elements can be influenced by the heat treatment of the melt. For an optimal
treatment it is necessary to determine this diffusion in dependence of the heat
treatment. For the investigation of the influence of the heat treatment to the Mn-
diffusion slags with different Mn-content were filled into Pt-capillaries. One slag
was free of Mn and the other one Mn-enriched. The Pt-capillaries with a diameter
of approx. 2 mm are welded together at the open ends and the welding was
protected with an additional cover. Then these tubes were heat treated at different
temperatures and for different times. For the analysis the capillaries then were
sliced along their axis and polished. The Mn-distribution could be measured on
this surface with a linescan along the capillary axis. Such a sample is displayed in
Fig. 7.33 (left part) [41].
A few of these Mn-distributions for different heat treatments are presented in
the right part of Fig. 7.33. For the given examples the heat treatments were per-
formed at the same temperature (1,400 C) but the timing schedule was chang-
ing—the durations were 1 and 2 h and in the cooling down time (1 and 3 h). It can
be seen that the diffusion is growing with the time of the heat treatment and also
with the cooling time of the sample.
From these results it is possible to determine the best heat treatment for the steel
cooking procedure but also the required amount of Mn in the melt to compensate
its loss due to the diffusion into the slag. Another result could be the determination
of diffusion coefficients in dependence of temperature.
Gems are minerals of high purity and special color. These colors are mostly caused
by traces of transition metals in the crystal lattice. If during their growing process
7.3 One Dimensional Distribution Analysis: LineScan 269
Fig. 7.33 Image of a sliced Pt-tube filled with a slag (left), Mn-profile in the slag for different
heat treatments (right)
the environmental conditions have been changed it is possible that also the colors
are changed. Interesting is the question which elements are responsible for which
color of the gem. Tourmalines often show these changing colors. Therefore the
examination of such a gem could be interesting.
As an example a zebra-tourmaline (Liddicoatite var) from Antsirable, Mada-
gascar [42] was examined. It is shown in the upper part of Fig. 7.34. The tour-
maline has a length of approx. 20 mm and was analyzed along a line on its axis
which is shown by the yellow line. The step size was 36 lm and the measurement
time per pixel 10 s. The lower part of the figure displays the distribution of the
intensities of different elements in arbitrary units.
The main components of the tourmaline—Al, Si and Ca show only very few
fluctuations. But the Ca-content shows an opposite behavior than the traces Ti and
Fe. These traces correlate with the darker stripes in the gem—the higher the
intensity of Fe and Ti the darker the color of the gem. Another trace is Mn which is
changing over the complete length of the tourmaline from low concentration on
the left side to high concentration on the right side. This is correlated to a change
of the basic color of the tourmaline from pink to red and orange i.e. the higher the
Mn-concentration the lighter the crystal.
The complete examination takes a time of only 10 min but gives a very good
overview for the composition of the gem and the influence of different traces to its color.
The rolls of roll bearings show often after a longer working time a coloring. It is
interesting to understand the reason for these colors and their differences. For that
purpose one roll was measured along a line of the surface as shown in Fig. 7.35.
The step size was 25 lm and the measurement time per pixel 10 s.
270 7 Applications
• In the middle of the mapping is the mentioned break of the sediments. This
break is in an wider area of high contents of Si, Ca, S and P. This is the result of
a volcanic eruption close to the maar which produces a lot of air pollution as
volcano ash which results in the thick Si- and Ca-rich layer and also a high S-
content.
A mapping of the complete area takes a lot of time because not only one line
has to be measured but a complete area with several lines. A linescan would be
faster but would be influenced by the not linear structure of the sediments.
Therefore a higher number of lines or a mapping gives better information about the
structuring.
The distribution of Si and Fe of a section of 10 mm of the presented bore core is
displayed in Fig. 7.38. The fluctuations of the Fe intensity are very high. The Si-
intensity is significantly lower caused by its lower excitation efficiency but shows
also fluctuations. The break of the bore core on the ‘‘volcano eruption layer’’ is
marked by a drop of both element intensities (approx. in the position 18–20 mm).
The Si content is increased short before the break—where the volcano dust is
already precipitated and behind the break.
The strong fluctuations can be better investigated with a stretched distance
scale. The high number of points of this linescan would allow this stretching
without any reduced resolution.
sea and of sediments transported by the rivers into the sea. In the case of not uniform
sediment layers the line for the measurement has to be defined within different non-
disturbed sections of the sample and then the different parts has to be stitched.
In the investigated bore core the Ti-distribution was of special interest. This
distribution is shown in Fig. 7.39 [from 44]. TiO2 is not a typical sediment in the
sea. The main source are river sediments transported into the Sea. That means,
the amount of Ti depends on the amount of water transported by the rivers. If the
water transport in the river is high also the Ti content in the sea is increased and
therefore the Ti-sediment layers are thicker. Vice versa they are reduced in time
of low river water. But the water level of the rivers depend on the amount of rain
which shows a seasonal changing but additionally also long term influences by
climatological changes.
The available bore core was measured with a resolution of 50 lm which relates to
a time resolution of approx. 2 month because the biological production in the
Carioca Basin is relatively high and the average thickness of an one-year sedi-
ment layer is approx. 0.3 mm. The Ti-content shows strong yearly variations
which represents the rainy and dry season in that area. But in few (darker marked)
time periods the Ti-content don’t reach high values. That means in these periods
which covers approx. 2 years beginning at 760 AC, approx. 9 years beginning
810 AC, approx. 3 years beginning at 860 AC and approx. 6 years beginning
910 AC there was a reduction of rain that have also reduced the efficiency of
agricultural production. The population on the Yucatan peninsula during that
time was already very high and required a high amount of food. If the agricultural
production was reduced this has an impact to the availability of food and a hunger
crisis could be the result. From these results it was concluded that this reduced
availability of food could be the reason for the collapse of the Maya civilization.
For the analysis of bore cores the instruments needs to fulfill special requirements.
For example:
• The scanning length should be adapted to the length of the bore core. For lakes this
cores typically are short—in the range of few 10 cm—but for bore cores from the
sea they are longer because their exploitation is more expensive. The bore cores are
274 7 Applications
Fig. 7.39 Distribution of Ti in a bore core from the Caribbean sea [44]
wet and they should be stored with comparable conditions as in the sea i.e. at 4 C
and in a wet atmosphere to avoid changes of their chemical and biological com-
position. The length of a bore core depends on the borer itself but much more on that
of the cooling boxes. These boxes have usually a length of 1.5 m and therefore the
instrument should be able to measure this length to avoid additional cuttings of the
bore core which would destroy the information of the core in this position.
• The requirements for spatial resolution of the measurement can be different—for
an overview or for the analysis of thick geological layers the spatial resolution
needs not very high, structure sizes in the mm-range are typical. On the other
hand, for the analysis of small structures like sediments with structures sizes
even in the lm-range high spatial resolutions are necessary. The spot sizes
should be in the range of the structure size itself i.e. in the mm-range for
overview measurements and in the range of ten’s of lm for the examination of
sediment structures.
• Due to the non-regular shapes of the geological or sediment structures a soft-
ware support for the determination of an undisturbed sequence would be helpful
either for the definition of the measurement line or by the data evaluation of
mapping measurements.
• For a complete analysis of bore cores not only the elemental composition is
interesting but also other parameters like color, magnetic properties, density and
others. There are instruments available that combine the measurement of ele-
mental composition with the determination of these parameters [45–49].
For all these reasons special instruments for the bore core analysis were
developed which have different spot sizes, which are combined with other ana-
lytical methods, which can be used for measurements directly on sea by assembly
the instrument into sea-capable containers, which have different software packages
for the control of measurements and data evaluation [50–52].
7.4 Two Dimensional Distribution Analysis: Mapping 275
If not only a single line is measured but many of them in equivalent distances the
elemental distribution of a larger area can be detected. This two-dimensional
elemental distributions can give a comprehensive information about the elemental
composition of the material. This method can be used for a wide range of appli-
cations. Only few of these possibilities can be discussed in the next sections for
some typical sample qualities like geological samples, archeological objects, and
biological applications, for material analysis, and also for forensic applications.
This should give a rough overview about the analytical potential and the diversity
of applications for element mappings with l-XRF.
Geological samples can be very different in composition and structure. But these
parameters are very important because geological material gives a lot of infor-
mation about the geological status, about natural resources and also about possible
technologies for their exploitation and treatment [53–59]. Therefore both original
and treated geological materials are often analyzed also with spatial resolved
methods like SEM-EDS. The analysis of a typical geological sample—a meteor-
ite—is already extensively described in Sect. 3.4.5 where the performance of
position tagged spectroscopy is explained based on the measurements on such a
sample.
Pitchblende is a Uranium ore which contains Uranium oxide (UO2). Due to the
high content of U it is one of the most intensive natural radioactive sources. The
end product of the radioactive decay of U is Pb which results in an enrichment of
this element in the rock. Finally, the Pb-content depends on the age of the material.
A large stone of pitchblende was analyzed on an area of approx. 5 9 3 cm and
with a pixel distance of 50 lm, the dwell time per pixel was 10 ms. An optical
image of the sample is shown in Fig. 7.40 as a mosaic image. The measured area is
the inclined green rectangle.
The sum spectrum of all pixels is displayed in Fig. 7.41. It shows a complex
composition with a lot of elements also with several elemental overlaps.
Already the sample image shows the inhomogeneous composition which can be
validated by the single element distributions of the mapping. The distributions of
the main components are shown in Fig. 7.42.
From these distributions the following conclusions are possible:
276 7 Applications
• The video image shows that the sample in the lower part is smoothed by
polishing but in the upper part it has the original rough surface. This is not
reflected in the elemental distributions. Its spatial resolution seems to be the
same for the complete mapping area. The slightly divergent spot on the not
polished parts has no visible influence to the elemental distribution.
• The elements are strongly structured. The grains in the upper part have medium
size, the content of Si and Ca is relatively high. Further Fe and Cu are mainly
concentrated in this region.
• The elements U and Pb are concentrated in the lower part of the rock. Because
the sample contains both Pb and As (see spectrum in Fig. 7.41) and the a-lines
7.4 Two Dimensional Distribution Analysis: Mapping 277
Elemental overlap Si
Ca Fe
Cu As
U Pb
of these elements are overlapping the distributions are calculated for their b-
lines (L and K respectively).
• The As is distributed similar like Si or Fe in the upper part and shows a distinct
different distribution than Pb.
• The Pb has a significantly higher weight fraction in the areas with U than in the
other ones. In particular it shows even the same fine structure than the U-
distribution which means that this Pb indeed is a product of the U-decay. A
278 7 Applications
standardless quantification of a U-rich area gives approx. 11.8 wt% UO2 and
3.8 wt% PbO. From that ratio the age of this mineral could be estimated with the
Uranium-Lead dating to approx. 250 Mio years.
7.4.1.2 Manganese-Nodules
Mn-nodules are deposits formed of more or less concentric layers of iron and
manganese hydroxides around a core which can be found at the bottom of the sea.
Other elements like Co, Ni, Cu or Zn are also present in weight fractions in the
range of \1 %. The cores can be different in size and composition [60]. The Mn-
nodule analyzed here was cut and polished. The size of the analyzed area was
approx. 28 9 28 mm with a step size of 36 lm which gives about 800 9 800
pixels. The dwell time was 30 ms. Figure 7.43 displays the sum spectrum of the
complete mapping. It can be seen that the main components are Mn and Fe but
also small weight fractions of Si, Ca, Ti, Ni and Sr can be detected.
The results of the distribution analysis are displayed in Fig. 7.44. The distri-
butions of the different elements show
• a well resolved distribution for all elements, in particular the shell structure as
result of the growing process in dependence of the environmental conditions
• a core shaped like an ellipse with approx. 7 9 3 mm with a distinct different
composition i.e. a high Si-content with small Fe-rich inclusions. The sur-
rounding shells consist mainly of Mn but also of Fe, Ni and some Ca and other
traces.
The composition of the core and few additional areas of the shells are given in
Tables 7.17 and 7.18. The areas used for the quantification are highlighted with
red circles in the overlay of Fig. 7.44.
These results show that the core are mainly oxides without Mn but already with
a relatively high Fe-content. The composition among the shells is quite similar but
it seems that there are gradients with growing content of CaO, Ti and Fe from the
inner to the outer part and of Mn contrarily. Halogens like Cl and Br which are
solved in the sea are in small weight fractions only in the shell but not in the core.
The shell structure of the nodule can be also seen in the linescan that is cal-
culated from the mapping along a line perpendicular to the long axis of the elliptic
core. The distributions of few main elements are shown in Fig. 7.45 with an
intensity normalization to their highest value.
The linescan shows that the structuring of Mn and Fe as the main components is
not the same. That can be a result of changing environmental conditions during the
very slow-approx. only 5 mm in 106 years—nodule growing process. Other ele-
ments which are present only in traces have also their own structure according to
the changing growing conditions. The assumption of an opposite trend of the
concentrations of Fe and Mn on the radius can be confirmed by that linescan.
7.4 Two Dimensional Distribution Analysis: Mapping 279
Fig. 7.43 Sum spectrum of the mapping on a Mn-nodule (square root intensity scaling)
Ca Mn
Fe Ni
280 7 Applications
Table 7.18 Quant results for different areas of the Mn-nodule/wt%—minors and traces
Elem Ti V Cu Zn As Br Sr Y Mo Ba
Shell 3 3.14 0.19 0.06 0.10 0.02 0.03 0.38 0.08 0.09 0.53
Shell 2 0.98 0.19 0.13 0.23 0.13 0.04 0.34 0.13 0.20 0.53
Shell 1 0.89 0.19 0.13 0.24 0.16 0.03 0.34 0.09 0.21 0.55
Core 1.00 0.00 0.03 0.02 0.01 0.00 0.09 0.03 0.00 0.00
In geology very light elements can be of interest for example to determine if metals
are present as pure elements, as oxides, carbonates or in other forms. Therefore, the
analysis of elements with peak energies \1 keV for example Oxygen or Carbon
would be helpful. But their analysis with energy dispersive detectors is very difficult
for different reasons—the excitation of fluorescence radiation with these low
energies has a very low efficiency (see Fig. 6.4), additionally the fluorescence yield
of low energetic lines is very small (see Fig. 1.3) i.e. the intensity of light elements
peaks is very small. Further these energies are strongly absorbed on the way from
the sample to the detector i.e. they have a very thin information depth only (see
Fig. 5.1) which means that only a very thin surface layer of the sample contributes
7.4 Two Dimensional Distribution Analysis: Mapping 281
to the signal. Further this low energetic radiation is absorbed in the measurement
medium (air or vacuum—see Fig. 2.57) and in the detector window (see Fig. 2.71).
The excitation efficiency for low energies can be influenced by the thickness of
the tube window, but for l-XRF it is mainly influenced by the low energy
transmission of the capillary optic. For the detection of energies of \1 keV an
ultra-thin detector window is required (see also Sect. 2.5.4). These windows are
common for EDS-detectors in SEM’s but for l-XRF instruments they are not usual
because they are very thin and therefore also extremely sensitive for a mechanical
damaging. The measurements demonstrated here were performed with such a
detector. Another influence that will be discussed in particular with the following
measurements is the quality of the vacuum.
For that purpose mappings are performed on a mineral with some light element
oxides (mainly MgO). The measurement are performed with the following con-
ditions: 35 kV, 600 lA, 666 9 666 pixel with a step size of 12 lm and a pixel
time of 20 ms which results in a total time of approx. 2.5 h. The mappings were
collected at different vacuum levels (800, 300, 75, 10 and 1 mbar, respectively)
which could be were controlled by the pump.
Figure 7.46 presents the low energy part of the sum spectra of the mapping for
the different vacuum levels with the same intensity scaling as well as the distri-
butions of Magnesium and Oxygen at the corresponding pressures. Further the
complete sum spectrum and the overlay of all main elements measured at the
lowest pressure of 1 mbar are shown in the last row of this figure.
From these measurements the following conclusions are possible:
• It is possible to measure the distribution even of ultra-light elements i.e. ele-
ments with fluorescence energies of \1 keV. But in this case the detector needs
a mechanically sensitive ultra-thin window and the vacuum level must be
adapted to these measurements.
• The vacuum level influences strongly the detectability of the low energetic
radiation as demonstrated with these measurements. Vacuum is required for
these elements as can be seen for measurement at 800 mbar which is close to
atmospheric pressure—no structures can be recognized. The presence of air is
documented by the Ar-peak in the corresponding sum spectrum.
• With the reduction of the air pressure the peak heights of the low energetic
elements are growing and then also their structures can be recognized.
For Mg already at 300 mbar the structure can be detected, the improvement of
image quality for lower pressures is marginal despite an increased height of the
Mg-peak for lower pressures. But even that is finished for pressures \100 mbar.
Then the absorption of Mg-radiation in air can be ignored.
For Oxygen the required vacuum level is significantly less—first structures can be
detected only at approx. 10 mbar (see also Fig. 2.72). For the here available final
pressure of 1 mbar the structure is relatively clear and the sum spectrum shows an
acceptable Oxygen-peak. But it has to be take into account that this peak corre-
sponds to a weight fraction of approx. 50 %. This low intensity is a result of the
282 7 Applications
800 mbar
300 mbar
75 mbar
10 mbar
1 mbar
Fig. 7.46 Sum spectra and distributions of Mg and O in dependence of the vacuum level
7.4 Two Dimensional Distribution Analysis: Mapping 283
low excitation efficiency and means that the presence and distribution of oxygen
can be detected but a quantification for O cannot be recommended.
• For the lowest pressure even the Fe-L-radiation can be separated. Fe has a high
content in the outer grains of the sample and only therefore the low energetic L-
line can be seen despite of their low excitation efficiency. For further L-lines the
weight fractions are too small.
From these measurements it can be concluded that with a special instrument
configuration even the detection of ultra-light elements with l-XRF is possible. But
this possibility is usable only for a distribution analysis. The use for point analysis
with quantification seems to be difficult because the information depth for these
light elements is very small i.e. in the thickness range of possible surface con-
taminations in particular with oxide layers. Further and the intensity is too small for
a reasonable quantification due to the high statistical error for these main elements.
Therefore the quantification of oxides by considerations of known compound
structures as discussed in Sect. 4.3.2.2 should give more trustable results.
The analysis of art objects and archeological objects is an important analytical task
because the knowledge of elemental composition provides valuable information on
their history and authenticity, for the understanding of degradation processes and
for required or already in the past performed restorations or for the selection of the
best restoration technology. A wide variety of applications in this field were
performed with l-XRF because the analysis is non-destructive, because the sample
can be analyzed on small and individual points, because it is even possible to make
the analysis on site on the sample (see also Sect. 5.3.5). The analytical and
instrumental requirements for that purpose can be very different because the
objects can be different in size, in shape, in composition, and are mostly unique
valuable that they neither can be destroyed for the analysis nor sometimes even
moved for the analysis. Examinations have to be performed typically on pigments
for example in paintings but also on scripts, on metal objects like coins or
sculptures or on oxides like gemstones, pottery, enamels, of glass and a lot more
[61–82]. For that purpose different XRF instruments are used; laboratory built
prototypes [83–90] but also commercially available instruments [91–95].
In the following sections only few of these applications can be presented.
Old photographs are often yellowed. This is a result of changes of color and
contrast both of the paper and of the pigments in the photograph. The change of
color of the paper is caused by an ageing process of the bleach due to the influence
284 7 Applications
of the weakly acidic cellulose. The loss of color of the pigments is a result of a
chemical reaction of the pure silver-particles with environmental aerosols. In that
case the black color of the silver is reduced and changed to brown. The yellowing
reduces the quality of the photographs significantly. It can even completely destroy
the image that its content can not more recognized. But the silver is still on the
paper not as pure Ag but as nitrite or sulfide—both not with the black color. With a
l-XRF distribution analysis it is be possible to detect the Ag-distribution inde-
pendent of the chemical bonding. This reconstructs the image content.
Because the photo paper is relatively thin the excitation radiation can penetrate
the paper and also excite the sample support. Therefore it is helpful to position the
sample on a free strained thin plastic film. Then the radiation can penetrate the
photo paper but will be only slightly scattered on the thin plastic film and the
enhancement of the spectral background is negligible.
A typical example for a yellowish old photograph is shown in Fig. 7.47. This is
a mosaic image prepared by the l-XRF instrument. Therefore the images shows
small squares due to the shadowing of the single pictures. The photograph has only
a weak structuring but the image content can still recognized. The inscribed dark
square on the edges in the image is the analyzed sample area.
To restore the image it is necessary to measure the Ag-distribution on the photo.
This should be done with a good spatial resolution—down into the range of the
grain size of the photographic layer to get a comparable resolution and also with
high sensitivity because the Ag-content on the thin photographic layer is not high.
The high spatial resolution requires the use of poly-cap optics. But then the
excitation of Ag-K-radiation is not very good due to the low transmission of high
energetic radiation in poly-caps. Therefore the Ag-L-radiation should be used for
the imaging. In that case the measurement at low pressure would be preferred to
reduce the absorption of Ag-L-radiation in air. On the other hand vacuum can
damage the paper by drying. That has to be avoided. Therefore the vacuum level
should be [100 mbar even better in air. The use of a even better in air.
The use of a W-tube for this application is advantageous because its brems-
strahlung background in the position of the Ag-L-absorption edge is higher which
means the Ag-L-radiation will be better excited. This can be seen in Fig. 7.48
which shows two spectra measured on a position with a relatively high Ag-content
(dark color in the photograph). The blue spectrum is excited with an Rh-tube and
the red spectrum with a W-tube. The measurement conditions were the same with
50 kV, 600 lA. It can be seen that the excitation with the W-tube the Ag-L-
intensity is increased by a factor of approx. 3 and that the W-line from the tube do
not influence the other elements in the sample. On the other hand, the Rh-L-line
has an overlap with the Ag-L-line which would give a further reduction of the
signal-to-noise ratio for the mapping.
The result of the analysis of the photo is shown in Fig. 7.49. The measurement
conditions were the same as for the single point i.e. 50 kV, 600 lA. The mea-
surements were performed on an area of 75 9 75 mm with a step size of 50 lm
and a pixel time of 15 ms.
7.4 Two Dimensional Distribution Analysis: Mapping 285
Fig. 7.47 Old yellowed photograph (original color is even more yellowish)
This result shows that with help of l-XRF it would be possible to identify and
restore old photographs. There is a relatively high time effort for the measurement
but the measurement can run automatically and don’t need any interaction during
its run.
The determination of the elemental distribution of paintings also can give inter-
esting information about its history and the ‘‘technology’’ of the painter. In that
way it is for example possible to proof the authenticity of a picture or define, if
necessary, restoration strategies [95–102].
Large paintings often can be measured only directly on the wall in a vertical
position because their transport would require a high effort. But also their high
value prohibits often transports because this would require a high security effort
and accordingly also high costs. Similar problems arise if the art objects are very
large and heavy or even immobile, like murals. In these cases it is necessary to
investigate the art objects on site, for instance hanging paintings or sculptures on
their pedestals. This requires the positioning of the instrument in front of the
object. Then the measurement head with tube and detector should be movable to
perform the measurements of single positions as well as of mappings with its
controlled movement in front of the sample (see Sect. 3.3.5). For this purpose
different instruments—mostly prototypes—were developed. Only very few com-
mercial instruments are available for this special application [91, 95].
For the distribution analysis of large paintings high spatial resolution is not
required because the structures are in the sub-mm range which means the step size
could be in the same range. This is also helpful to limit the pixel number and the
7.4 Two Dimensional Distribution Analysis: Mapping 287
measurement time because the size of the samples can be very large, in ranges of
up to 1 m2. With small step sizes this would increase the number of pixels tre-
mendous and prolong the total measurement times and requires a large data
memory. Therefore the step size for these measurements can be in the range
[100 lm. But then the coverage of the sample is not completely for the small spot
of a poly-cap optic. On the other hand, the high excitation intensity of an X-ray
optic would be necessary. In that case a spot size variation would help. This is be
possible by using a poly-cap optic with a convergent exit beam by changing the
working distance.
These requirements are realized in a commercial instrument—the M6 JetStream
from Bruker [94]. This instrument has a travel range of 800 9 600 mm with a
motor driven head i.e. the sample is fixed in their position. The excitation is
performed with a Rh-tube and a poly-cap optic which can be used in different
working distances which allows the adaptation of the spot size to the sample
structure. This is shown in Fig. 7.50. The working distance and with that also the
spot size can be defined by the focal plane of the optical microscope which can be
changed in 5 steps in a range of approx. 5 mm. The change of the spot size
depends on the final adjustment of the optic, it can vary for example between 30
and 300 lm or between 100 and 800 lm. For the detection of fluorescence radi-
ation a 30 mm2 SDD with high count rate capability is used.
For large sample areas it is necessary to have relatively short measurement
times but a sufficient sensitivity. The measured counts for elements in the range
from Z = 20–51, i.e. Ca to Sb, respectively, is displayed in Fig. 7.51 (from [95]).
That means the limits of detection are better than 1,000 mg/kg for measurement
times of 10 ms per pixel. This is sufficient for most examinations. Longer mea-
surement times of course would improve the sensitivity.
With that instrument several paintings were analyzed. Few of them with few
very interesting results, for example, that original paintings are completely or
partly repainted, that restorations were already performed or that color degrada-
tions occurred caused by chemical changes due to environmental influences.
Old paintings are often very dark and have low brilliance. An interesting
question is if this dark impression is only due to contaminations by dust or if they
are caused also by changes in pigment chemistry. If so, the original appearance of
the painting was not only lighter but colors could be even different.
Such investigations were performed by Dik et al. [89, 95, 98]. Here it will be
demonstrated with measurements on the painting ‘‘Homer’’ by Rembrandt (mea-
surement performed by Dik et al.). The size of the picture is 107 9 82 cm. It was
measured with a step size of 500 lm (total approx. 3.5 Mio pixel) and a dwell
time of 20 ms which results in a total measurement time of approx. 20 h.
Figure 7.52 shows in the first row the original picture and the distribution of Co,
in the second row the distributions of both detectable Pb-radiations i.e. of Pb-L and
Pb-M. Due to the different fluorescence energies of these radiations they have a
different information depth—the Pb-L radiation comes from deeper layers than the
Pb-M radiation. All elemental distributions are displayed in white i.e. the lighter
the white the higher the measured intensity and concentration of the elements.
288 7 Applications
Co pigments are typically blue. Depending on further components this blue can
have different shades. Pb oxide is a white pigment.
The examination of the element distributions provides the following results:
• The Co weight fraction is high for example at the cap on Homer’s head and also
on his waist belt. But in the painting these parts appear not as expected for Co-
pigments in blue but in a yellowish brown. It is quite likely that the paint in
these areas originally had a more bluish tone but is changed now by environ-
mental influences.
• The Pb-L distribution shows the presence of PbO as white pigment through all
paint layers, the low energy Pb-M-distribution shows its presence only at the
surface. This is of interest because oil paints have the tendency to become more
transparent over time. The Pb-M lines visualize the modeling of light and dark
in the paintings that have altered over time. This is caused by a change of the
PbO-pigment on the surface due to environmental influences. The degradation
results in a loss of color and increased transparency, causing an overall dulling
of the painting. Pigments in deeper layers are less affected by the degradation
and therefore resemble the original lead oxide pigment distribution of the
painting more closely. That allows the conclusion that the reduced brilliance of
the colors of old paintings must not be only a result of contamination with dust
but also of chemical changes caused by the environmental influences.
7.4 Two Dimensional Distribution Analysis: Mapping 289
Frescos are very often contaminated by environmental influences and then they
loss their stability. The surface of the walls then is crumbly and the frescoes are
damaged step by step. For the restoration as a first step it is necessary to stabilize
the surface of the wall [103].
For that purpose the wall typically is plastered by lime (CaCO3). Due to the
influence of sulfate from the environment the lime is changed to gypsum (CaSO4)
290 7 Applications
But this lime is still not stable against environmental influences and therefore a
second step is necessary for a sustainable conservation. In that step, a pulp with a
Barium-hydroxide Ba(OH)2 solution stabilize the material by the conversion of
CaCO3 to BaCO3 according to:
The effect of this process can be checked and monitored by the measurement of
the penetration of Ba into the wall. That requires the determination of the Ba-
distribution perpendicular to the wall surface which is possible with measurements
on a broken piece of the wall after the treatment. Because the wall plaster is
inhomogeneous i.e. a mixture of small stones and lime a line scan perpendicular to
the surface can give a wrong picture. But with a mapping a more detailed
examination of the distribution should be possible. This is demonstrated in
Fig. 7.53 that shows on the left an optical image of the analyzed sample and on the
right the distributions of S and Ba, respectively, both in a thermal false color
presentation. The sample surface is on the right.
The distributions show clearly that there is still a high S content in the bulk of
the plaster. But on the wall surface on the right the S-intensity is clearly decreased
and the Ba-distribution shows the opposite behavior. The Ba-concentration on the
surface is increased but only in a relatively thin layer of approx. 1.5 mm.
This can be seen also on the linescans in Fig. 7.54 which are calculated from
the distributions over a line perpendicular to the surface. For a more representative
distribution which is not too much influenced by embedded stones these linescans
were averaged for the complete mapping.
These results show that this technology for stabilization of walls works very
well and it can be expected that the BaCO3 surface shows better resistivity against
environmental influences.
Also for life sciences and biology the distribution of special elements are
important and can be very interesting—for example for an environmental moni-
toring with help of organisms, for the understanding of the influence of essential
elements to the growing process, for checking the authenticity of pharmaceuticals,
7.4 Two Dimensional Distribution Analysis: Mapping 291
Fig. 7.53 Image of a piece of the wall together with the measurement direction
and a lot more [104–130]. Here again only few examples can be discussed that
demonstrate these possibilities. Other examples for this type of application are also
presented in Sects. 7.2.3.2 and 7.5.2.4.
Pharmaceuticals have some active ingredients but their main components are
fillers. The fillers are not forced to have a therapeutically function but they can
influence the people tolerance for the drug. But these fillers can also be used for an
identification of manipulated products with help of their composition and/or dis-
tribution. The presented examination was performed with two pills of acetylsali-
cylic acid (Aspirin)—one original from Bayer and one generic product of a no
name manufacturer. They was cut, polished and then measured on the cross section
in an area of approx. 12 9 3.5 mm2 (step size of 25 lm, i.e. approx.
480 9 140 pixel, dwell time 20 ms). The sum spectra of both mappings are
292 7 Applications
Fig. 7.55 Sum spectra of the mappings of both pills (blue aspirin; red no name)
displayed in Fig. 7.55. Both spectra are very similar but the no name product (red
spectrum) has more inorganic components: the content of Ca is significantly higher
but there are also traces of Si, P, K and Fe. This causes a heavier matrix which is
the reason for the slightly reduced scatter background.
Astonishing are the traces of Pb in the generic product. The weight fraction can
be estimated in the low mg/kg-range but a poisonous ingredient in a pill is sur-
prisingly and should be avoided.
Figure 7.56 shows the distributions of the mentioned elements both of the
original (left column) and the generic product (right column). The original prod-
uct, in general, has a more homogeneous structure. The no name product has some
distinct inclusions with a high P-content. Further the distribution of S and Fe traces
in the original product is more homogeneous. The no name product shows a lower
Fe-intensity in areas with higher S-intensity.
This example shows that l-XRF allows the identification and examination of
distribution of inorganic components in pills. Because these components can differ
for original and generic products this offers also the possibility for the identifi-
cation of fake products.
There are several cancer therapies. One of them is the Neutron Capture Cancer
Therapy (NCT) [131, 132]. For that method a large area of the body is irradiated
with thermal neutrons. Their cross sections with the light elements of the human
tissue are very low. But if atoms with higher cross section for thermal neutrons are
positioned close to the cancerous cells neutrons can be captured by their nuclei.
7.4 Two Dimensional Distribution Analysis: Mapping 293
Sulfur
Iron
Multi -
element
overlay
Fig. 7.56 Element distributions of Aspirin (left Bayer, right no name), (high weight fractions
correlates to light colors)
Then a conversion of the nucleus is possible and low energetic radiation can be
emitted which is absorbed adjacent to the emitting atom and can damage the
cancerous cells.
For this therapy it is important to find pharmaceuticals which can be accu-
mulated preferentially close to the cancerous cells and this has to be proofed. The
analysis of tissues by optical microscopy is possible. This shows the cancerous
cells but not the adsorbed pharmaceuticals. For that purpose the elemental dis-
tribution performed for example with l-XRF has to be compared with the optical
image.
In case of an irradiation with thermal neutrons it is necessary to consider the
tissue elements i.e. H, C, O, N as well as elements that can be used for neutron
capture. The neutron cross sections of the isotopes with the highest probability are
given in Table 7.19. It can be seen that the cross section for the tissue elements are
several orders of magnitude less than for Boron and Gadolinium which are used in
radioactive pharmaceuticals. The cross section for Gd is even 5 orders of mag-
nitudes larger than that for B.
The used nuclear reaction for Boron is: 10B(n, a)7Li. The emitted a-radiation
will be absorbed in immediately environment environment to the emitting atom
and damaged on his way the cancerous cells.
The used reaction for Gadolinium is 157Gd(n, c)158Gd with an energy of
7.8 MeV. This radiation has to high energy and penetrates the light tissue without
damaging. But the decay of Gd is very complex and one of the products are Gd-
Auger-electrons with an energy of 41 keV which have a high potential for dam-
aging the cancerous cells.
294 7 Applications
P S
Fe Gd
Fig. 7.57 Element distribution in an aorta of a rat (high weight fractions correlates to brightness
of the colors)
7.4.4 Electronics
Fig. 7.58 PCB of a digital watch as mosaic image (left), total intensity distribution (mid) and
overlay of several elements (right)
Fig. 7.59 Distributions of restricted elements (Pb, Br, and Cr) in the PCB
Fig. 7.60 Distributions of Rh-L and compton scattered radiation of the PCB
7.4 Two Dimensional Distribution Analysis: Mapping 297
Rh is not in the sample i.e. these images are generated only by the scattered
tube radiation. Rh-L has a low energy (2.7 keV) and do not penetrate into the
sample. Therefore the left part of the image shows the scattering only from the
sample surface. The intensity depends on the scattering element and shows
therefore an elemental contrast. The right image shows the scattered Compton-line
from Rh-K-radiation. This radiation has a higher energy (20.2 keV) and can
penetrate the sample. That means the light matrix elements of a PCB (C, O, Si)
cannot absorb the radiation but the electronic components on the PCB absorb the
incident as well as the backscattered radiation which produce an image of the
components of both sides of the PCB. Therefore it is possible to see the horizontal
Cu-tracks on the backside in this image. This would offer the possibility for failure
analysis also on hidden parts of components.
Fig. 7.61 Image and Au-distributions of Au-structure on SiO2-glass. a Optical image. b Au-L;
4 lm. c Au-L; 10 lm. d Au-M; 4 lm
Table 7.20 FWHM of Measurement conditions (ms) FWHM (lm) Structure size (lm)
distributions with different
4 lm step, Au-L, 230 30.7 17
measurement conditions
4 lm step, Au-M, 230 36.8 16
4 lm step, Au-L, 1,600 23.7 10
10 lm step, Au-L, 90 32.2 18
analytic treatment with spatial resolved methods. In the following only few
examples can be demonstrated that should demonstrate the analytical potential for
these type of problems.
Figure 7.63 shows a blade of a damascene knife. The structure of the different
types of steel can be nicely recognized in the image. With a distribution analysis it
is possible to visualize the different alloys. For the measurement of this blade a
pixel distance of 100 lm was sufficient because the structures are relatively large.
The analyzed area had a size of approx. 130 9 30 mm which results in a pixel
number of 1,320 9 310. The dwell time was 15 ms and the total measurement
time 2.2 h.
The pictures in Fig. 7.64 show the overlay of the distributions of the main
components as well as their single distributions. These distributions reflect the
optical structure of the knife.
There are strong fluctuations in composition over the blade. Cr and Fe have a
relatively homogeneous distribution in all parts of the knife, the edge has a high
content of Mo and V, the region between edge and blade have increased Ni and
Cu-content and Ni is distributed in the structure of the blade.
For that structure a phase analysis can be very interesting. The result of a cluster
analysis (see also Sect. 3.4.5.6) is displayed in Fig. 7.65. Most of the pixels can be
classified into 4 phases—two in the blade (structured red and blue), one in the edge
(green) and one in the junction between blade and knife edge (light blue).
The spectra of all pixels of the same phase can be summed and used for
quantification. The results for the four phases together with their area fraction are
summarized in Table 7.21.
It can be seen that the composition of the edge and the blade are quite different,
in particular the Mo- and V-content are much higher in the edge as already seen in
the elemental distributions. This makes the edge hard. The two phases of the blade
itself differ mainly for the Ni-content. The interface between the two parts with the
high content of Ni and Cu also could be nicely seen in the distribution.
These relations between the different elements can be confirmed by a chemo-
metric evaluation as demonstrated in the binary distribution diagrams in Fig. 7.66
that plots the measured intensity ratios of two selected elements. In the case of a
300 7 Applications
Overlay Fe
Cr Ni
Cu Mo
higher number of pixels with a given intensity ratio the color changes to green and
blue—the intensities are normalized to their highest value.
From the binary diagrams the following conclusions are possible:
• Fe/Cr: the intensities of both elements are in a linear relation but the Fe intensity
is higher by a factor of approx. 5. The largest number of pixels (blue colored
points) exists for approx. 70 % of the maximum Fe-intensity which relates to
20 % of Cr-intensity. This higher numbers are related to the blade of the knife.
• Fe/Ni: this relation shows two different relations: the first—for all Fe intensities
the Ni-intensities are relatively low. A high number of pixels (blue points) exists
again for a Fe-intensity of approx. 70 % which relates to the blade. The second
relation are increasing Ni intensities related to decreasing Fe-intensities. This
behavior describes the interface between blade and edge of the knife.
7.4 Two Dimensional Distribution Analysis: Mapping 301
Fe Ni
Cr Fe
Cu Mo
Ni Fe
• Ni/Cu: this relation is similar to that of Fe and Cr. The Ni-intensities are by a
factor of approx. 2 larger. A larger number of points are close to zero because
not all pixels of the distribution contain a Cu-content.
• Fe/Mo (normalized to highest intensities): there are two regions with a higher
pixel number: most pixels have a very low Mo-intensity with a Fe-intensity of
302 7 Applications
again approx. 70 %—this represents the blade of the knife. Another pixel
concentration can be observed for a slightly lower Fe-intensity and a Mo-
intensity of approx. 1 %. This correlates to the edge of the knife.
For environmental analysis the examination of the air pollution is important. Due to the
low weight fractions and small particles sizes the accumulation of pollutions particles
over longer time periods is necessary. This can be done with collectors which deposit
the particles on a filter for example from paper. If the filter is moving during the period
of collection it is possible to inspect even temporal changes of the deposition. For the
examination of these filters l-XRF is an often used tool—for the analysis of single
particles that are prepared from the filter as well as for the distribution analysis of the
complete loaded filter when the particles are not removed from the filter. Then the
analysis has to be performed on the filter i.e. the filter itself contributes also to the
fluorescence radiation. Therefore, the filter needs to be homogeneous and clean—at
least regarding the elements of interest in the air pollution.
The here analyzed filters had a size of approx. 16 9 2 cm which was deposited
along the filter axis. They were collected in a park inside of a city which is
contaminated by the traffic pollutions but also by industrial pollutions within the
city. The filter represents a time period of one day.
Measurements were performed along the axis as linescan with a step size of
100 lm and a pixel time of 500 ms. The existing elements on the filter could be
determined from the sum spectrum of all pixels. The linear distributions of a few
elements of interest (S, Cl, Ca and Fe) along the filter axis are displayed in
Fig. 7.67 with normalized intensities.
For these elements the distribution shows strong fluctuations which are widely
identical for all elements i.e. it seems that they are emitted from the same sources.
The structuring can be related to the emission of pollutions. During the day time
(6–22 h) the pollutions is high with some peaks of high concentration. During
night the pollution is reduced and shows only one strong peak at approx. 2 am.
These fluctuations now can be allocated to the actual emission situation and it is
possible to influence the different emitters.
Solar cells are manufactured from different materials. Poly-crystalline Si is still the
mostly used one. The efficiency of these cells depends on the structure of the Si
and on implemented impurities in particular on the grain borders of the single
grains. The crystallites in the Si-wafer have different orientations and therefore
they generate diffraction peaks at different positions. For a demonstration of that
situation a Si-wafer was analyzed on an area of approx. 14 9 14 cm with a step
size of 180 lm and a dwell time of 45 ms. The sum spectrum is shown in
7.4 Two Dimensional Distribution Analysis: Mapping 303
Fig. 7.68. It shows an intensive Si-peak but also various small peaks in the spectral
background which are partly wider and do not correspond with the energy of
element lines. These are diffraction peaks of the different crystallites (see also
Sect. 2.5.6.4). They are Bragg-scattered on different lattice planes in dependence
of the orientation of the crystallites.
For the different diffraction peaks it is possible to calculate the intensity dis-
tribution like a mapping. These distributions are displayed in Fig. 7.69 (right part)
together with the mosaic image of the wafer (left part).
The distribution of the different diffraction peaks shows a high structuring and
this is comparable to that of the optical image of the wafer.
Spectra of single grains are displayed in Fig. 7.70. All of these spectra show
again the strong Si-peak and every spectrum a different set of diffraction peaks.
These have partly considerably intensities. According to Sect. 2.5.6.4 they can be
identified as diffraction peaks because
• most of them have indeed a symmetric Gaussian shape but their widths are
larger than it would be expected for fluorescence peaks at the same energy and
• the peak positions are not in agreement with that of element lines or the com-
plete line series is missing.
But every diffraction peak corresponds to a special d-spacing of the lattice but
for another orientation of the crystallite. The intensities of the diffraction peaks are
relatively small because often very high indexed lattice planes contribute to the
diffraction peaks which therefore have a low scattering intensity only.
Very often the lattice structure of an analyzed sample is known or the few
possible variants can be determined with help of the chemical composition that can
be determined by the fluorescence analysis. A comparison of the lattice distances
determined from the diffraction peaks with all possible lattice distances and their
intensity relations should allow a determination of the grain orientation to the
304 7 Applications
Fig. 7.69 Mosaic image of the Si-wafer (left) and intensity distributions of the diffraction peaks
(right)
l-XRF has a large potential for failure analysis because it allows the examination
of small sample areas both on faulty as well as on good products. Here an example
will be presented for the analysis of a conveyor belt prepared as compound
material based on rubber which is used for the transport of wet pastes like pulps
7.4 Two Dimensional Distribution Analysis: Mapping 305
Fig. 7.70 Object spectra of different grains of the poly-Si with a series of diffraction peaks
(square root intensity scaling)
[135]. The belts have a width of approx. 3 cm and a length in the range of few ten
meters. Typically the life expectancy of a belt is in the range of several month but
few of them are broken already after approx. 1 month. A section of such a
damaged belt is displayed as mosaic image in Fig. 7.71. It shows large cracks on
the surface which are going even deep into the material.
The examination starts with an elemental mapping of the belt. Figure 7.72
displays the sum spectra collected from a faulty belt both from the top rubber layer
(blue spectrum), from the inner side of a detached piece of the rubber (red spec-
trum) and that of a good belt (yellow spectrum). The spectra of the inner side of
the defect belt and the good belt are widely identical but the spectra from the front
side of the defect belt (blue) shows additional peaks for Si, Cl, Ti and Pb. It seems
that these are contaminations due to the transported material on the belt i.e. that
give not yet hints for the reason of the defects.
Rubber typically is cross-linked by Sulfur. Therefore the distribution of S
distribution can be interesting to get information about the strength of the material.
The S-distributions of all three parts are shown in Fig. 7.73. Whereas the S-
distribution of the good belt is homogeneous show the distributions for the faulty
belt on both sides areas with different S-intensity i.e. changing S-weight fractions.
The difference of the S-intensity is demonstrated in Fig. 7.74 which shows two
object spectra of areas with higher as well as lower S-intensity.
From these result it can be concluded that the cross-linking of the rubber is not
homogeneous for the faulty belts—areas with high S-content are very strong cross-
linked and areas with lower S-content are less cross-linked. The too strong cross-
linked areas have less plasticity and tend to break in case of bending during the
movement of the belt and areas of low S-content are not sufficient cross linked and
can be therefore cracked due to the high forces on the belt. The non-homogeneous
306 7 Applications
Fig. 7.72 Sum spectra of different mappings of defect and good belts (square root intensity
scaling)
distribution, in the damaged belt this is partly substituted by Ti. The distribution of
these two elements is caused by a polymer net which is embedded in the belts for
their stabilization. In the durable belts this net contains Ca-stearates but the faulty
belt contain Ti-dioxide. It is known that Ca-stearate has a positive influence to the
durability of the polymers and therefore also to the belt but Ti-oxide not. This
result is also confirmed by linescans perpendicular to the belt surface calculated
from these distributions. They show from left to right the distribution of Ti and Ca
in the belt from top to down as already discussed i.e. for the faulty belt a higher Ti-
content but for the good belt a higher Ca-content.
Finally, a further difference which can cause the damaging process are thin steel
lacings along the belt axis which are embedded for a further increase of its sta-
bility. They are visible in mappings of the cross section parallel to the belt axis.
These Fe-distributions are displayed in the lower section of Fig. 7.75. They show a
different structure of the lacings—in the faulty belt the single wires are slightly
thicker and only twisted but in the good one they are thinner, twisted and have an
additional covering wire in distances of approx. 5 mm.
308 7 Applications
Good Faulty
S-distributions
Ca-distribution
Distribution of Ca (blue bars),Ti (green dots) along a line normal to the belt surface
The thicker wires in the faulty belt should have a stronger resistance against
bending which stress the embedding rubber and can force its breaking. And if the
steel is not completely embedded in the rubber it can corrode. The corrosion along
the wire will damage the connection between the rubber and the lacings which
reduce the stability of the belts.
In this way a complex failure structure could be investigated and revealed by
the distribution of different elements and hints for the improvement of belt quality
were possible.
Forensic offers a wide range of analytical questions which can be solved with XRF
and are discussed elsewhere [136–138]. This can be the identification of small
particles for example from accidents or from crime scenes [139–144], the
7.4 Two Dimensional Distribution Analysis: Mapping 309
Gunshot residue is one of the most often occurring analytical problems in foren-
sics. Every time when a gun is fired small particles of the explosive but also from
the bullet, the cartridge and the firearm itself are emitted. These small particles are
distributed over a relatively large area both in the shooting direction but also in the
backward direction. If they are detected this can be considered as a proof that a
gun was fired. Explosives consist of special elements like Ba, Sb or Pb. If these
elements can be identified in a single particle there is a high probability that this
particle was generated by an explosive process. On the other hand, the detection of
the elements in separate particles is not yet considered as this proof.
Gunshot Residues (GSR) particles are very small in the range \1 lm and
therefore the spatial resolution of l-XRF is not sufficient for their identification.
Therefore most of these investigations are performed in electron microscopes in
combination with EDS because this offers the required spatial resolution. But this
method needs a high effort—the sample has to be collected on special Carbon-tabs
which then have to be introduced into the high vacuum system of the SEM and
only after that procedure the analysis can starts. But this again is relatively time
consuming itself because the particles are distributed over a larger area and can be
only identified by scanning of the complete area of Carbon-tab area which is
approx. 1 cm2.
But also if the large spot size of l-XRF cannot be used for the identification of
single GSR-particles it can be used for two investigations in GSR:
• Pre-investigations if typical GSR elements are present
In that case a larger area of the carbon tab can be scanned. Also if the single
particles cannot be identified it is possible to detect the elements of the explosive
material itself. If these elements cannot be detected with XRF the probability to
find GSR-particles is negligible and it is not necessary to spend the high effort
for the SEM-analysis. On the other hand this probability is very high if the
typical GSR-elements are detected with l-XRF. Then a SEM-EDS analysis has
a good chance to be successful.
310 7 Applications
The analysis of money in particular of coins is a typical analytical task for ar-
cheometry—to check the changes of their composition during time, to get infor-
mation about the development of technology in precious metal metallurgy, to
check the correct composition of coins etc. But also in forensics coin analysis is an
interesting task in particular or the identification of fakes. Here a special case for
the identification of a faked coin will be presented [154]. The collector’s value of
coins strongly depends on the age but mainly on the frequency of occurrence of a
coin. The number of manufactured coins is changing from year to year and differs
also for the different manufacturing mints. Because the monetary value of a coin
can differ tremendous it happens that the date or the manufacturing mark on the
coin are changed.
7.4 Two Dimensional Distribution Analysis: Mapping 311
Fig. 7.77 Coin of 2 Reichsmark from 1927 (left) and spectra measured on the ‘‘2’’ (blue) and on
the ‘‘7’’ (red) (right)
7.4.6.3 Pigments
Pigments are used for all coloring but also for writing. Examples for pigments
analysis are already given in Sect. 7.4.1.3. But also for forensics the identification
and detection of pigments can be very interesting. This is valid for example for
handwriting or lettering to compare the pigments of different pens or visualize
erased writings but it is also interesting for the identification of fakes of banknotes
or other certificates.
• Pigments on documents
A lot of data and information are written on paper with pigments of different
kind. In former time pigments often was prepared with help of minerals or other
natural materials. These compounds typically have some inorganic components
which easily can be detected by X-ray fluorescence (see for example
Sect. 7.4.2). Nowadays also organic compounds are used as pigments. Their
examination with XRF is much more difficult because they contain not in any
case detectable elements.
The identification of pigments, the distinction between different pigments or also
the detection of deleted or erased letters is a typical forensic task. This will be
demonstrated on an example—a document with an organic pigment which is
also manipulated. The analyzed part of the document, some of the elemental
distributions as well as the sum spectrum of the distribution measurement are
shown in Fig. 7.79. The measurements were performed on an area of approx.
46 9 17 mm2 with a pixel distance of 60 lm i.e. 770 9 280 pixel within
slightly more than 30 min.
From the displayed distributions the following conclusions are possible:
• On the optical image some printed and also hand written letters as well as a
stamp can be seen.
• The sum spectrum of the complete mapping shows a relatively complex com-
position with several elements. Few of these elements are typical for paper—as
for example Si, K and Ca as filler and Fe and Ti (only traces) as pigments. But
the distribution of most of these elements is homogeneous i.e. no structure
which is identical to the letters can be detected.
• The Ti-distribution is non-homogeneous. It shows randomly scattered hot spots.
These hotspots have different intensity but same size. It can be supposed that
these spots are generated by TiO2-particles which are embedded within the
paper. Then the low energetic Ti-radiation will be absorbed by the cellulose
fibers of the paper.
• Only the distribution of S shows very clearly the hand written letters and also
with less intensity the stamp. All hand written letters can be well recognized
7.4 Two Dimensional Distribution Analysis: Mapping 313
only the ‘‘6’’ from 1968 of the DoB (date of birthday) seems to be blurred—in
the S-distribution more than in the optical image. For a better investigation this
part of the S-distribution was enlarged and displayed separately in the last row
in false color presentation. Here not only a ‘‘6’’ can be seen but also some other
structures. These can be interpreted as ‘‘4’’ i.e. it seems that the ‘‘4’’ was partly
removed and overwritten by a ‘‘6’’. The fluorescence of the small left traces of
the original letter is sufficient for the identification of the original writing.
This assumption is supported by the distribution of Si which is displayed in false
colors too to improve the contrast between the small intensity differences. This
distribution is also not homogeneous but shows a higher intensity in the position of
the ‘‘6’’. It seems that the surface of the paper in this position is different than in other
positions i.e. it could be changed in this position by erasing the original writing.
• The printed letters cannot be detected in any element distribution. These letters
seems to be completely from organic pigments.
From these investigations it can be concluded that the examination of pigments
based on their X-ray fluorescence is possible. It is possible to distinguish in
between different pigments due to their different composition but it is also possible
to detect pigments which are not visible because they are covered by others or
because they are erased and still some traces are available that can be excited by
X-rays. This can be used as a powerful tool for forensic investigations.
314 7 Applications
Distribution of Mg Distribution of Al
Distribution of Fe Distribution of Ti
Distribution of S Distribution of Zn
• Banknotes
Banknotes are high-tech products with a lot of different safety features which
should protect the money against falsification. A few of these features can be
easily recognized, for example the quality of the paper, the coloring or some
special features like metal wires or holographic images. But there are much
more that can require even special equipment for checking. One of these fea-
tures are the different pigments which are used for printing. These pigments and
their distribution can be examined also with l-XRF. With that method both the
distribution of pigments as well as the composition of the paper can be analyzed.
7.4 Two Dimensional Distribution Analysis: Mapping 315
As an example here a 50 Hong Kong Dollar bank note was analyzed with a step
size of 120 lm and a dwell time of 30 ms which results for the a complete
measurement in a total measurement time of approx. 6.5 h i.e. an overnight run.
Few of the single element distributions are displayed in Fig. 7.80. Here it can be
seen that various pigments are in use for the different structures on the bank-
note. Few of the pigments structures are identical with the optical image (upper
left image) but others differ. This can be observed for example for the digit
‘‘50’’ in the upper right corner of the Ca-distribution which is an inverted image
in the Ca-distribution. Few elements are present only in very special structures
others in more or less all of them—often independent of their colors.
As already mentioned in the last section nowadays also often organic pigments
are used. These pigments cannot be identified in any case with XRF because
often they show no detectable XRF-signal. But for inorganic pigments XRF is a
powerful analytical method.
Products and raw material have not only inhomogeneities in the plane parallel to
the surface but also perpendicular to that. And these layers needs not to be plain
and flat. Therefore exists also a request for a 3 dimensional distribution analysis
for the investigation of different materials. This is another task than coating
analysis which assumes layers with known element composition, homogeneous
distribution within the plating layers and a flat surface. With the confocal
arrangement described in Sect. 3.2.3 the non-destructive investigation of
3-dimensional element distributions is possible. But this is limited by the infor-
mation depth of the sample. That means, confocal measurements can be performed
preferentially for heavy elements in samples with light matrices. This will be
demonstrated in the Sect. 7.2 of this chapter.
But also for heavy matrices the 3-dimensional elemental distribution can be of
high interest. This is even possible with a common l-XRF instrument but only
with a destructive sample preparation method. For that purpose a 2-dimensional
distribution can be measured conventionally. Then a thin layer of the sample has to
be removed, for example, by polishing and the new surface plane will be analyzed
again. If this process is repeated several times a data set which represents the 3-
dimensional element distribution can be generated. But this analysis is not only
destructive but causes also a high manual effort because the sample has to be
polished for every new layer—every removed sample layer should have the same
thickness—and the sample needs to be positioned in the instrument exactly in the
same position that the different 2-dimensional distributions fit each other. But this
is a possibility for 3-dimensional investigations of samples with heavy matrices
with help of l-XRF due to their small information depth. An example for the
analysis of a geological sample with that method will be demonstrated in the
following section.
316 7 Applications
The analyzed sample was a strongly structured natural rock that is shown in
Fig. 7.81. It was composed of quartz (white area in the back part), granite (gray
area in the front part) and it had few inclusions (dark areas) from Wurtzite which is
a Zink-sulfide (b-ZnS).
For these measurements the sample was embedded in epoxy which had on the
back side pins to define exactly the measurement position on the sample tray. The
sample was polished in steps of 200 lm (checked by a gauge) an measured in
25 sections in 324 9 253 pixels with a step size of 80 lm i.e. the investigated
volume was 26 9 20 mm2 with a thickness of 5 mm. The information depth is less
than the step size which was chosen with 200 lm to get a relatively large analyzed
volume. But this is possible because the structures of the rock are larger than this
step size. The dwell time was 30 ms that this gives a total measurement time per
layer of approx. 1 or 25 h for the complete data set.
One of the 2-dimensional distributions is shown in Fig. 7.82 as an overlay for
several elements. It shows in the upper part regions with a high Si-content—this is
the quartz. The high Ca- and Fe-content in the lower part represents the Granite and
the light circles are the Wurtzite inclusions here represented by the high Zn-content.
For the reconstruction of 3-dimensional distribution every measured 2D-dis-
tribution were evaluated for the single element distributions. These then are
merged together with a special 3D-presentation software [155, 156]. This gives as
result the 3D-distribution for all elements which can be displayed in various
modes. Figure 7.83 shows the 3D-distribution for the main elements of this
mineral in the same sample position. With appropriate software packages it is
possible to display cross sections in every direction, complete views from different
directions for single elements or as overlay of different elements.
This 3D-analysis allows a complete view into the material and the investigation
of its different phases which can be helpful for the understanding of the material
behavior. It means, for a lot of analytical questions this can be a very interesting
tool. But it has to take into account that the effort for the preparation of the
different sample sections is very high and the analysis time also can be extended to
several hours. But the main drawback is that the sample will be destroyed due to
this type of analysis.
• Analyzed volume
The other possibility for a 3-dimensional analysis is the application of a con-
focal geometry as described in Sect. 3.2.3. Here, two optics are used, one to
7.5 Three Dimensional Distribution Analysis 317
focus the tube radiation onto the sample and a second one to restrict the field of
the vision of the detector. The foci of the two optics are aligned to overlap each
other and form a probing volume from which fluorescence radiation is detected.
This probing volume can be moved through the sample by displacing the sample
[167–169] in three directions.
The shape of the probing volume is a rotational ellipsoid with an eccentricity
normal to the sample surface. This eccentricity depends on the angle between
both the excitation and fluorescence radiation. For small angles the eccentricity is
large i.e. probing volume has a larger dilation normal to the sample surface which
reduces the spatial resolution perpendicular to the sample surface. But this angle
also influences the analytical performance—if the angle is close to 90 the
scattering intensity is low and the peak-background-ratio will be improved but on
the other hand then the path of excitation and fluorescence radiation in the sample
is prolonged which means a further reduction of the real information depth due to
the tilted incidence. For smaller incident and take off angles the information depth
can be enlarged but this reduces also the angle between the two beams and enlarge
the probing volume i.e. the ellipsoid will have a larger eccentricity. Therefore a
compromise between information depth and depth resolution has to be used—
typically the better depth resolution is the preferred aim because samples with
light matrices and larger information depth can only be analyzed at all.
318 7 Applications
Distribution of Si Distribution of Zn
Distribution of Ca Distribution of Fe
The size of the probing volume is in the range of the spot size of the capillaries
i.e. 15–30 lm in the sample surface plane and approx. 30–50 lm normal to that.
But that means for samples with heavy matrix the probing volume is in the
range of the information depth and the spatial resolution normal to the surface is
limited. For sample with light matrices the situation is different—the informa-
tion depth is larger and 3D-measurements for a larger sample volume are
possible.
Further there is also a dependence of the size of the analyzed volume from
energy. This dependence has two reasons—one is the energy dependence of the
spot size of the capillaries i.e. the spot size is increasing with decreasing energy
(see also Sect. 2.2.5.3). The other and more important effect is the
self-absorption of fluorescence radiation in the sample i.e. the relation between
probing volume and analyzed volume. This will be demonstrated with the
following discussions of depth profiles of homogeneous samples with light
matrices, for example glasses [164].
This kind of depth profiles are calculated and displayed in Fig. 7.84 for different
elements i.e. for different fluorescence energies for a 200 lm thick glass disk
(SiO2-matrix) with contributions of CaO, Fe2O3, PbO2 and SrO2 each with
50 mg/kg. This is a light matrix and therefore the information depth for the
fluorescence lines of the composition elements is large against the probing
volume. In Fig. 7.84 the depth profiles are normalized to each other.
7.5 Three Dimensional Distribution Analysis 319
The Ca-profile (for Ca-K radiation with 3.7 keV) looks symmetric but has a
narrow width—notable less than the sample thickness. The Ca-radiation has a
strong self-absorption in the glass i.e. not the complete glass thickness can con-
tribute to the signal. The Fe-distribution (for Fe-K-radiation with 6.4 keV) looks
similar but with a larger width due to the higher fluorescence energy. For the
higher energies as the Pb-L-radiation (10.5 keV) or even Sr-K-radiation
(14.1 keV) the fluorescence can penetrate completely the glass sample. Therefore
the intensity drops down with larger sample depth; this decreasing is determined
by the self-absorption in the sample and can be approached by an exponential
behavior. On the right hand side these distributions drops down at 200 lm i.e. at
the thickness of the glass sample. But the Sr-distribution has a higher intensity
than the Pb-intensity. Due to the higher energy of Sr-radiation its absorption is
smaller than for Pb-radiation. That means, the measured intensity depends close
to the surface from the immersion of the probing volume into the sample i.e. from
the real size of the probing volume but if the probing volume is completely in the
sample it depends from the self-absorption of the fluorescence radiation.
As a result the position of the layer is the point of inflection of the intensity
increase only in case of negligible self-absorption—then the half analyzed
volume is the half of the probing volume. But if there is a considerably
absorption the position of the layer is between the point of inflection and the
maximum of the intensity peak [164].
values. For special prepared test samples the model delivers an acceptable
accuracy [163]. For real samples with non-homogeneous layer structures and
non-homogeneous distributions in the sample plane the modeling of the struc-
ture would be difficult but gives also acceptable results for the layer structure.
But even the qualitative determination of distributions in different sample depth
or the measurement of depth profile in special sample points can be a big
advantage and give already a lot of information about the sample composition.
For 3-dimensional analysis the number of voxels can grow tremendous. Further
the measured intensity drops down due to the absorption from deeper layers.
Therefore the measurements are performed often not with high spatial resolu-
tion. Measurements of only few hidden planes or of planes perpendicular to the
sample surface or even only of a depth profile normal to the sample surface in a
few points are common analytical tasks.
In case of l-XRF the information depth is also limited by the absorption but
here of the energy of the fluorescence radiation. This can be smaller but also
larger than the radiation energy for tomography. But the limitations for the
analyzed sample are only for one dimension. For the other two dimensions
there are no limitations i.e. the analyzed volume can be larger.
• Tomography shows the distribution of different elements due to their different
absorption. The absorption depends both on the mass absorption coefficient of
the sample for the fluorescence radiation of the element and also on the
thickness of the component. Therefore even the determination of the quali-
tative composition of the sample is not possible with absorption tomography.
This is no problem for the confocal l-XRF. This offers much more infor-
mation for the material characterization.
It has to be mentioned that in case of detecting the fluorescence radiation in
different directions in tomographic geometry also the elemental distribution
can be determined. But then the absorption of fluorescence radiation determines
again the analyzed sample volume i.e. it has to be smaller than the information
depth, Further this method requires a relatively high instrumental effort.
• Finally l-XRF offers as already discussed the possibility for quantification
which is not possible with tomographic methods.
In between there are already samples of different qualities investigated with
confocal equipment. Art objects [165–171] as well as samples from life sciences
[172, 173] and the analysis of environmental objects [174, 175] or also quality
control of industrial products [176, 177]. A few examples for these applications
will be discussed in the following.
The analysis of geological samples with confocal geometry is limited due to the
high absorption of fluorescence radiation in the sample which results in small
information depths. Here as an example a not polished Mn-nodule is analyzed i.e.
the sample surface is rough like a broken structure. The measurement was per-
formed on an area of 18 9 14 mm with a step size of 40 lm. The step size in Z-
direction was 100 lm. For a dwell time of 15 ms the measurement time for each
plane is approx. 1 h and the total measurement time for 20 layers accordingly
approx. 20 h.
The distribution of several elements in one of these layers is shown in Fig. 7.85.
The signal can be collected only for a thin surface layer. Deeper areas cannot
contribute because their radiation is absorbed in the covering part of the sample i.e.
the size of the probing volume is in the range of or only slightly larger than the
information depth. That means, only a narrow surface range which has the correct
working distance can contribute to the signal and all other areas which are more far
or more close to the optics do not. Therefore the combination of the different
distributions represents the topology of this not polished sample as ‘‘isolines’’.
322 7 Applications
This is demonstrated for one element, here for Fe, for all measured layers in
Fig. 7.86. This represents the surface shape of the sample. The distribution shows
the irregular shaped sample surface even with a crack and a hole which looks like
open areas in the complete data set. In these areas the movement to deeper layers
was stopped. For deeper layers even these areas would contribute to the signal.
For samples with light matrices the information depth should be larger. For that
type of samples a real 3-dimensional distribution can be interesting and should be
also measureable [165, 166]. Interesting objects are micro-electronic components
because they often are prepared from light elements and are fine structured in all
dimensions.
One example is the analysis of integrated circuits (IC) on printed circuit boards.
IC’s are Si-chips that realize different functions. For that purpose the chip needs
several electrical connections to the lead frame by bonding wires. The complete
chip is typically protected by a plastic cover. Such an IC is displayed in Fig. 7.87
7.5 Three Dimensional Distribution Analysis 323
in the environment of a PCB. The chip is covered by the plastic protection and the
only visible parts are the outer electrical connections.
This IC was examined with a 2-dimensional mapping with a step size of 50 lm
and a dwell time of 10 ms. The distributions of few single elements and their
overlay are shown in Fig. 7.88.
Around the IC are the contacts of the lead frame corresponding to the optical
image. These contacts are coated with a SnPb-solder for a good contact to the Cu
tracks on the PCB. The Cu contacts of the lead frame are coated in the inner part
close to the chip with Ag. But also the Ag in the back-contact of the chip can be
detected. This is possible because the Ag-distribution was measured with the high
energetic Ag-K radiation (22.1 keV) which can penetrate not only the plastic cover
but also the Si-chip. Finally the Au-distribution shows the bonding wires. These
images already give a lot of information about the internal structure of the IC.
But a 3-dimensional elemental distribution can show even more details. This is
demonstrated for another IC which was measured on an area of 13 9 7 mm2 with a
step size of 25 lm and a dwell time of 25 ms. The total time per plane then is approx.
1 h. For the 35 sections with a distance of 40 lm the total measurement time was
approx. 35 h. The distributions for few elements are prepared with a special 3D-
viewing-software and shown for two different directions of the view in Fig. 7.89.
These distributions show much more details of the internal structure of the IC.
In that case the lead frame is also a Cu-alloy and the outer part coated with a Sn-
solder. The bonding wires are again made from Au. With appropriate software
packages it is possible to display the distribution as 3D-representation. These
images allow a non-destructive inspection of the design and internal structure of
the IC.
Daphnia are planktonic crustaceans and members of the order Cladocera. Because
of their small size of 0.2–5 mm and their jumpy swimming style, Daphnia are
commonly known as ‘‘water fleas’’. Their aquatic living environment can range
from rivers, streams, lakes and ponds to acidic swamps. Their short lifespan, their
reproductive capabilities and especially their sensitivity to contaminants make
them to indicator species in the field of environmental toxicology. Daphnia are for
instance used to test the effects of toxins in ecosystems. Because Daphnia are
almost transparent, their internal organs are easy to examine in live specimens.
The performed examination demonstrates the capabilities of l-XRF as a
method for the analysis of these biological species regarding elemental compo-
sition and the possibility to use it also for environmental control [172, 173].
Daphnia’s are interesting also for examinations in confocal geometry due to their
light matrix and small size. Figure 7.90 shows the photo of a Daphnia.
The photo shows the thin, partly transparent body and the head with a large eye.
The body is protected by a thin and transparent carapace. The most prominent
features are the eyes and the two antennae.
324 7 Applications
Overlay Cu Pb
Sn Ag Au
For the measurement the Daphnia was prepared on a plastic film. The mea-
surements for a 2-dimensional mapping were performed on an area of approx.
3 9 3.5 mm2 with a step size of 20 lm. Figure 7.91 displays the distribution of
few interesting element.
The distributions of Ca and S represent the crustace of the animal. The Ti
distribution shows the alimentary tract. The Daphnia was living in water with Ti-
oxide particles. These particles were assimilated together with the water in the
alimentary tract.
7.5 Three Dimensional Distribution Analysis 325
The distribution can give an interesting view inside the animals but the correct
distribution cannot be investigated. Therefore also a 3-dimensional distribution
could give more information.
For that purpose two confocal measurements were performed on an area of
2.5 9 2.5 9 1 mm3 with a step size in a plane parallel to the surface of 8 lm and
in Z-direction of 25 lm as well as with 50 lm. The dwell times were 25 and
100 ms and the total measurement times 40 and 60 h, respectively.
The results are shown in Fig. 7.92 as overlay for 3 elements. The distributions
show that the longer measurement time (right image) offers a lot more information.
In that case the loss of sections due to the longer Z-step size is not so important.
The resolution in Z-direction is still sufficient but the higher collected intensity
gives a better signal-noise-ratio and a reduced digital error.
326 7 Applications
Fig. 7.92 Element distributions of Ca, S and Fe for two different dwell time (color scaling in the
right picture)
For these conditions Fig. 7.93 shows few views through the Daphnia in dif-
ferent directions. This allows the investigation of the internal structure of the
animal in particular also the influence of environmental contaminations.
In particular the possibility to differentiate between the distributions of the
different elements is very important to understand the structure and distribution of
the water flea and then in dependence of the contaminating substances in the water
also their influence to its living conditions.
Another often occurring analytical problem is the analysis of pigments (see also
Sects. 7.4.2.1, 7.4.2.2, 7.4.2.3). Pigments are used in paintings on paper or canvas
as well as on documents on paper or also parchments. This are all light matrices
i.e. a sample quality that can be analyzed with confocal geometry.
7.5 Three Dimensional Distribution Analysis 327
The main problems for the analysis of pigments layers are that these layers are
very thin which results in a weak signal, that their thickness is even varying i.e. the
signal intensity is also varying and that the pigment signal is overlaid by the signal
of the paper itself.
As example the pigment analysis of an old handwritten document will be
presented. A fraction of the paper is shown in Fig. 7.94 together with the analyzed
detail (green inscribed rectangular) with a size of approx. 2 9 2 cm2. Measure-
ments are performed with a step size of 50 lm which is adapted to these relative
large structures, the dwell time was 100 ms to get a sufficient signal from the thin
pigment layer.
The paper was positioned on a thin plastic film to avoid high spectral back-
ground from the sample support. Spectra from different areas of the mapping are
summed up and displayed in Fig. 7.95.
The green spectrum is from the paper itself. It shows peaks of Cl as bleaching
element, K and Ca as filler of the cellulose and traces of Mn, Fe and Cu. The red
spectrum is from the large capital which has a high intensity for Pb and Hg i.e. it
can be assumed that this pigment contains vermilion (HgS) and lead white (PbO).
Finally the blue spectrum is from the other small letters which have high inten-
sities for Ca and Fe which suggests that this pigment is an iron gall ink.
These conclusions from the different spectra are confirmed by the elemental
distributions which are shown in Fig. 7.96.
The left image shows the distribution of Cl. This is in the paper and will be
absorbed in the pigments if they are on top of the paper. Therefore the letters are
displayed with less brightness which represent a lower intensity.
The two images in the middle show the distribution of Ca and Fe, respectively.
The Ca-distribution shows the small letters with higher intensity—the pigment is
on top of the paper and it contains Ca. The capital is displayed with decreased
intensity. That means, the Ca-fluorescence of the paper will be absorbed in the
pigments of the capital. The Fe-distribution shows not only the letters that can be
328 7 Applications
seen on the paper but additionally also others letters. These letters are on the back
side of the paper. The Fe-radiation has sufficient energy to penetrate the thin paper.
For the lower energetic Ca-radiation this is not possible. This demonstrates again
the ability to detect hidden or weak pigment structures which are here on the
backside but that can be also erased.
Finally the right part of Fig. 7.96 shows the distribution of both Pb and Hg.
These elements are used only in the capital.
This examination demonstrates again that the identification of pigments and
their distributions can be performed with l-XRF—not only for visible but also for
hidden pigments. But the identification can be performed preferably by finger-
printing. A quantification of pigments on paper is a difficult task due to the pen-
etration of the pigment into the paper and the overlapping of the pigment signal
with that of the paper [70]. But it was demonstrated that the identification of the
ink is possible by a fingerprint model independent of the used paper [65].
Because this sample has a light matrix it is possible to examine it also with in a
confocal geometry. These measurements were performed only for a single letter on
an area of 4.8 9 5.4 mm2 with a step size of 20 lm (i.e. 240 9 270 pixel). The
step size in direction of the sample was 25 lm for 17 layers i.e. a total thickness of
approx. 0.4 mm; the measurement time per pixel was 15 ms which results in a
total measurement time of approx. 45 min per layer.
The results of these measurements are displayed in the following figures.
Figure 7.97 shows the optical image of the analyzed sample (left side) and the
distributions of Fe and Ca of different depth levels.
These distributions show that the paper seems to be slightly uneven, the first
distribution shows only the top of the letter on the front side, the next distribution
shows the lower part of this letter. The next distribution shows the upper part of the
letter of the backside and the last distribution finally shows the lower part of the
7.5 Three Dimensional Distribution Analysis 329
Fig. 7.95 Sum spectrum of the several objects (paper green; capital red; small letters blue)
Distribution of Cl Ca Fe Pb and Hg
backside letter. The two distributions are weaker than that from the front side due
to the absorption of the radiation in the paper. Further these distributions show also
that only the Fe-radiation can penetrate the paper but not the Ca-radiation due to
their lower energy. But in the third distribution the relatively high Ca-intensity
represents the filler in the paper.
If these different distributions are arranged in a data cube for the selected
energies of Ca and Fe the complete 3-dimensional distribution of the pigment will
be visible. They are shown in Fig. 7.98 in the left two images for Ca and Fe
together and in the 3 right images for Fe only each with different viewing
directions.
The Ca–Fe-distributions show that the Fe is concentrated in the letters only and
the Ca is mainly in the paper. It can be also seen that the letters on both sides of the
paper are well separated. The distribution of only Fe confirms this situation. The
most right image shows that the pigment layers are really separated and it can be
also seen that the paper is slightly bended—the upper part is curved to the left.
This was already expected from the single layer measurements of Fig. 7.97.
330 7 Applications
Fig. 7.97 Optical image of the letter (left) and distributions of Ca and Fe (blue and red,
respectively) in layers with an distance of 0.125 lm (right)
Art objects have very often light matrices—this can be paintings with an organic
matrix where the used inorganic pigments are of interest but this can be also
enamel with pigment structure or others. Therefore for these type of samples
several examinations were performed [151, 153, 156, 157, 160 and 161].
• Layer structure of a painting
As an example the investigation on a painting will be presented which shows
that with confocal geometry the pigment of different layers can be identified. For
a demonstration typical paint layers were prepared on pure cotton paper. The
paper was coated with the white pigment lead white (2PbCO3 Pb(OH)2)) and
then with colored pigments like chrysocolla (green, CuSiO3 nH2O). The pig-
ments were mixed with organic binding material which contains Ca and Zn. To
compare the confocal measurements with the real structures a cross section of
the painted paper was prepared and displayed as optical image from a micro-
scope in Fig. 7.99. It shows the paper on the bottom coated with a thick lead
white and on top a thinner chrysocolla layer.
The depth profile measurements was performed with the following conditions:
tube parameters 35 kV, 600 lA, step size in Z-direction 10 lm, measurement time
per depth 100 s live time. The profiles of few elements are shown in Fig. 7.100.
The surface of the paint layer is approx. at 750 lm, the paper is on the right hand
side. The different element intensities are normalized.
The profiles show different depth distributions for the different elements. On top
there is a narrow Si-profile. This seems to be part of the organic binding material in
the crysocolla layer and is not a separate layer. The fluorescence of Si starts
already earlier than the Cu-fluorescence because of the larger probing volume for
Si. This lower energy causes also a high self-absorption of Si-radiation. Therefore
the information depth for Si is very small even in this relatively light matrix i.e.
only upper layers can contribute to the Si-signal.
The Cu-signal is still growing at deeper layers than Si and reaches also the
maximum for a deeper position in the sample. But this seems to be not a result of
different layers but of the less absorption of the Cu-radiation in the matrix (see also
Fig. 7.84).
7.5 Three Dimensional Distribution Analysis 331
Fig. 7.98 3D-distribution of Fe and Ca (red and blue, respectively) and of Fe only (gray)
The Pb-distribution is clearly different from Cu. The signal starts and reaches
also the maximum value at deeper layers. But the fluorescence energies of Cu and
Pb are not so much different. That means there is a layer containing Pb(PbO)
below the crysocolla—as the sample was also prepared. If we assume that the
layers are flat and not mixed the beginning of the crysocolla layer is approx. at
750 lm and the PbO-layer starts at approx. 825 lm i.e. its thickness is in the range
of 75 lm. The thickness of the Pb–O-layer is not known—it can be even infinite
thick.
Preliminary XRF analysis on the object has shown that the black color is an
enamel color which consists of ground lead glass with Fe and/or Cu oxide as
pigment. Due to further optical examinations it has to be assumed that on this
sample the color was not fired onto the glass. That would imply that this typical
enamel color here was used as a stained glass color for cold painting. But then
different painting techniques would be mixed. This should be verified with the
confocal examinations.
For a differentiation between cold painting and stained glass two reference
samples were prepared with reverse paintings on a glass slide with the black
332 7 Applications
enamel pigment—one of them was dried in air (VP1) the other was fired for
30 min at 700 C (VP2). The examination was performed with a bench top
instrument in confocal geometry in air with a probing volume in the range of
30–50 lm depending on the fluorescence energy. Measurements of the depth
profile are carried out on both sides i.e. from the painted side and from the glass
side of the samples.
The first measurement was performed in a position without the black enamel
pigment on both sides of the glass. The depth profiles for several elements are
shown as not normalized distributions in Fig. 7.102. Ca, Fe, As and Sr are in the
glass (broken lines) but Cu, Zn and Pb are contained in the black enamel (solid
line). Additionally also Mn and Co are in the black enamel which is not expected
in historical material but in the here used actual pigment.
The sample surface here is on the left hand side of the diagram. The profiles
from both sides of the air dried sample (VP1) and that from the backside (not
painted side) of the fired sample (VP2) looks similar. But the profiles of the
painted side of the fired glass shows significant changed which can be caused only
by the heat treatment. During firing the lead glass in the black enamel melts and
fuses together with the glass. Additionally Cu and Zn have been diffused into the
7.5 Three Dimensional Distribution Analysis 333
Fig. 7.101 ‘Lüneburger Meditationstafel’ of the 14th century; National Museum Schwerin, Inv.
no. G 2627
surrounding glass. For Mn and Co this cannot be observed. It seems that they are
not mobile. The profiles of these elements are not shown to avoid confusing by too
many lines.
The diffusion of Cu and Zn could not be finished to the not painted side of the
glass because it was too thick. Therefore these profiles from the glass side look like
the profiles of the not heat treated samples.
Further measurements were performed on positions of the black enamel of the
test samples. The corresponding profiles are displayed in Fig. 7.103. Here the
intensities are normalized. Again on the left side the air dried sample is shown and
on the right side the heat treated sample. From these profiles it can be recognized
that for the air dried sample the pigments of the black enamel (Mn, Co, Zn, Pb)
have similar profiles and are on top of the glass. The Ca-profile which charac-
terizes the position of the glass starts at deeper sample positions only and is
considerably noisy caused by their absorption in the paint layer. But the Ca-profile
rises up significantly later than for the paint layer. That means, the glass substrate
is covered by the black enamel pigment. For the heat treated sample these ele-
ments are diffused into the glass. The Ca-profile starts as the first one due to the
low föuorescence energy of Ca (see also Sect. 7.5.2.1), the profiles of the elements
of the pigment are distributed according their energy i.e. they are moved slightly
334 7 Applications
Fig. 7.102 Depth profiles of different elements in test samples for ‘‘reverse painting’’
towards deeper layers for higher fluorescence energies. That means the fired
sample is homogeneous but the air dried sample has a layered structure.
With this knowledge the art object—the Lüneburger Meditationstafel—could
be now investigated. This was done with a step size of 10 lm normal to the sample
surface and live times of 50–100 s, the excitation condition was 50 kV, 600 lA.
Measurements are performed on different sample positions only on the painted
side of the object. The upper row of Fig. 7.104 shows the profiles measured of one
of the glass spots (left) and of one of the black enamel spots (right) and the second
row shows the corresponding normalized profiles.
7.5 Three Dimensional Distribution Analysis 335
Fig. 7.104 Depth profiles measured in two spots of the ‘Lüneburger Meditationstafel’
References
69. G. Vittiglio, K.H. Janssens, B. Vekemans, F. Adams, A. Oost, Spectrochim. Acta. B54,
1697 (1999)
70. K.H. Janssens, B. Vekemans, L. Vincze, F. Adams, A. Rindby, Spectrochim. Acta. B51,
1661 (1996)
71. C.T. Yap, S.M. Tang, Archeometry 26, 61 (1985)
72. B. Kanngieser, W. Malzer, I. Reiche, Nucl. Instr. Meth. B211, 259 (2003)
73. S. Larsson, PhD Thesis, Department of Ohysics, Chalmers University (1991)
74. O. Hahn, D. Oltrogge, H. Bevers, Archaeometry 46, 273 (2004)
75. C. Neelmeijer, I. Brissaud, T. Calligaro, G. Demortier, A. Hautojärvi, M. Mäder, L.
Martinot, M. Schreiner, T. Tuurnala, G. Weber, X-Ray Spectr. 29, 101 (2000)
76. A. Denker, J. Opitz-Coutureau, Nucl. Instr. Meth. B213, 677 (2004)
77. K. Janssens, J. Dik, M. Cotte, J. Susini, Acc. Chem. Res. 43, 814 (2010)
78. I. Reiche, K. Müller, M. Eveno, E. Itié, M. Menu, JAAS 27, 1715 (2012)
79. O. Hahn, Restaurator 31(1), 41 (2010)
80. O. Hahn, M. Wilke, T. Wolff, Restaurator 29(4), 235 (2009)
81. S. Castro, A. von Reden, U. Baumer, P. Dietemann, H. Stege, I. Fiedler, C. Tilenschi, J.
Klaas, Zeitschrift für Kunsttechnologie und Konservierung 1, 177 (2010)
82. S. Bretz, U. Baumer, H. Stege, J. von Miller, D. von Kerssenbrock-Krosigk, Stud. Conserv.
53, 209 (2008)
83. A. Gianoncelli, G. Kourousias, Appl. Phys. A89, 857 (2007)
84. R. Cesareo, A. Brunetti, S. Ridolfi, X-Ray Spectr. 37, 309 (2008)
85. R. Cesareo, G.E. Gigante, I.S. Iwanczyk, A. Dabrowski, Nucl. Instr. Meth. A380, 440
(1996)
86. R. Lu, A. Le, Y. Gu, J. Zhu, Nucl. Instr. Meth. B104, 595 (1995)
87. S. Bichlmeier, K.H.E. Janssens, J. Heckel, P. Hoffmann, H.M. Ortner, X-Ray Spectr. 31, 87
(2002)
88. G. Vittiglio, S. Bichlmeier, P. Klinger, J. Heckel, W. Fuzhong, L. Vincze, K. Janssens, P.
Engström, A. Rindby, K. Dietrich, D. Jembrih-Simbürger, M. Schreiner, D. Denis, A.
Lakdar, A. Lamotte, Nucl. Instr. Meth. B213, 693 (2004)
89. M. Alfeld, K. Janssens, J. Dik, W. de Nolf, G. van der Snickt, JAAS 26, 899 (2011)
90. K. Tsuji, K. Nakano, JAAS 26, 305 (2011)
91. H. Bronk, S. Röhrs, A. Bjeoumikhov, N. Langhoff, J. Schmalz, R. Wedell, H.E. Gorny, A.
Herold, U. Waldschläger, Fresenius’ J. Anal. Chem. 371, 307 (2001)
92. http://www.horiba.com/scientific/products/X-ray-fluorescence-analysis/micro-xrf-analyzer/
details/xgt-7000-X-ray-analytical-microscope-488/
93. http://www.edax.com/Products/Micro-XRF/Orbis/Orbis-Micro-EDXRF-Analyzer.aspx
94. http://www.bruker.com/en/products/X-ray-diffraction-and-elemental-analysis/
X-ray-spectrometry/m6-jetstream/technical-details.html
95. M. Alfeld, J. Vaz Pedroso, M. van Eikema-Hommes, G. Van der Snickt, G. Tauber, J. Blaas,
M. Haschke, K. Erler, J. Dik, K. Janssens, JAAS 28, 760 (2013)
96. L. de Viguerie, V.A. Sole, P. Walter, Anal. Bioanal. Chem. 395, 2015 (2009)
97. J. Dik, K. Janssens, G. Van der Snickt, L. van der Loeff, K. Rickers, M. Cotte, Anal. Chem.
80, 6436 (2008)
98. J. Dik, K. Janssens, M. Alfeld, K. Rickers, E. van de Wetering, in Nouveau regard sur
Rembrandt, ed. by E. van de Wetering (Local World BV, Amsterdam, 2010)
99. M. Alfeld, K. Janssens, K. Appel, B. Thijsse, J. Dik, Zeitschrift fur Kunsttechnologie und
Konservierung 25, 157 (2011)
100. L.S. van der Loeff, M. Alfeld, T. Meedendorp, J. Dik, E. Hendriks, G. Van der Snickt, K.
Janssens, M. Chavannes in Van Gogh Studies 4: New Findings, ed. by L. van Tilborgh, D.
van Halsema, J. House, G. Weisberg (WBOOKS, Zwolle, 2012), p. 33
101. D.L. Howard, M.D. de Jonge, D. Lau, D. Hay, M. Varcoe-Cocks, C.G. Ryan, R. Kirkham,
G. Moorhead, D. Paterson, D. Thurrowgood, Anal. Chem. 84, 3278 (2012)
102. F.P. Hocquet, H.C. del Castillo, A.C. Xicotencatl, C. Bourgeois, C. Oger, A. Marchal, M.
Clar, S. Rakkaa, E. Micha, D. Strivay, Anal. Bioanal. Chem. 399, 3109 (2011)
References 339
139. E.O. Espinoza, M.J. Mann, L. Leonardo, A. Copeland, J. Forens. Sci. 39, 839 (1994)
140. Y.W. Liu, A.R. Harding, D.E. Leyden, Anal. Chim. Acta. 180, 349 (1986)
141. S. Becker, L. Gunaratnam, T. Hicks, W. Stoecklein, G. Warman, Prob. Forens. Sci., XLVI,
80 (2001)
142. J.A. Buscalia, Anal. Chim. Acta. 228, 17 (1994)
143. J.M. Curran, C.M. Triggs, J.R. Almirall, J.S. Buckleton, K.A. Walsh, Sci. Justice 37, 241
(1997)
144. P. Engstrom, S. Laesson, A. Rindby, B. Stocklassa, Nucl. Instr. Meth. B36, 222 (1998)
145. J.I. Thronton, Forensic paint examinations, in Forensic Science Handbook, ed. by R.
Saferstein (Prentice Hall, New Jersey, 1982), p. 529
146. A. Zeichner, N. Levin, E. Landau, J. Forens. Sci. 37, 542 (1992)
147. J. Zieba-Palus, Sci. Justice 39, 123 (1999)
148. B. Charpentier, C. Desrochers, J. Forens. Sci. 45, 447 (2000)
149. D. DeGaetano, J.A. Siegel, K.L. Klomparens, J. Forens. Sci. 37, 281 (1992)
150. R.H. Keeley, P.J. Nolan, Rep. To Interpol Meet, Paris, (1986)
151. P.V. Mosher, M.J. McVicar, E.D. Randall, E.H. Sild, Can. Soc. Forens. Sci. J. 31, 157
(1998)
152. F.S. Romolo, P. Margot, J. Forens. Sci. 119, 195 (2001)
153. A. Zeichner, N. Levin, M. Dvorachek, J. Forens. Sci. 37, 1567 (1992)
154. B. Kämpfe, M. Haschke, MGM-Magazin (1998)
155. D. Stalling, M. Westerhoff, H-Ch. Hege, in The Visualization Handbook ed. by Ch.D.
Hansen, Ch.R. Johnson (Elsevier, 2005), p. 749
156. http://www.vsg3d.com/amira/overview
157. G.J. Havrilla, T. Miller, Powder Diffr. 19(2), 119 (2004)
158. B. Kanngießer, W. Malzer, A. Fuentes Rodriguez, I. Reiche, Spectrochim. Acta B60, 41
(2005)
159. B. Kanngießer, W. Malzer, I. Reiche, Nucl. Instr. Meth. B211–B212, 259 (2003)
160. W. Malzer, PhD Thesis, University Bremen (2004)
161. W. Malzer, B. Kanngießer, Spectrochim. Acta. B60, 1334 (2005)
162. I. Mantouvalou, W. Malzer, B. Kanngießer, Spectr. Acta. B77, 9 (2012)
163. T. Wolff, W. Malzer, I. Mantouvalou, O. Hahn, B. Kanngießer, Spectrochim. Acta. B66,
170 (2011)
164. I. Mantouvalou, W. Malzer, B. Kanngießer, Spectrochim. Acta. B 77, 9 (2012)
165. I. Mantouvalou, K. Lange, T. Wolff, D. Grötzsch, L. Lühl, M. Haschke, O. Hahn, B.
Kanngießer, Anal. Chem. 83, 6308 (2011)
166. B. Kanngießer, W. Malzer, I. Mantouvalou, D. Sokaras, A. Karydas, Appl. Phys. A106, 325
(2012)
167. I. Mantouvalou, K. Lange, T. Wolff, D. Grötzsch, L. Lühl, M. Haschke, O. Hahn, B.
Kanngießer, J. Anal. At. Spectrom. 25, 554 (2010)
168. B. Kanngießer, A.G. Karydas, R. Schütz, D. Sokaras, I. Reiche, S. Röhrs, L. Pichon, J.
Salomon, Nucl. Instr. Meth. B264, 383 (2007)
169. I. Mantouvalou, W. Malzer, I. Schaumann, L. Lühl, R. Dargel, C. Vogt, B. Kanngiesser,
Anal. Chem. 80, 819 (2008)
170. Z. Smit, K.H. Janssens, K. Proost, I. Langus, Nucl. Instr. Meth. B219, 35 (2004)
171. X. Wei, Y. Lei, T. Sun, X. Lin, Q. Xu, D. Chen, Y. Zou, Z. Jiang, Y. Huang, X. Yu, X.
Ding, H. Xu, X-Ray Spectr. 37, 595 (2008)
172. B. De Samber, R. Evens, K. De Schamphelaere, G. Silversmit, B. Masschaele, T.
Schoonjans, B. Vekemans, C.R. Janssen, L. Van Hoorebeke, I. Szaloki, F. Vanhaecke, G.
Falkenberg, L. Vincze, JAAS 23, 829 (2008)
173. B. De Samber, G. Silversmit, R. Evens, K. De Schamphelaere, C. Janssen, B. Masschaele,
L. Van Hoorebeke, L. Balcaen, F. Vanhaecke, G. Falkenberg, L. Vincze, Anal. Bioanal.
Chem. 390, 267 (2008)
174. M. Bielewski, M. Eriksson, J. Himbert, R. Simon, M. Betti, T.F. Hamilton, Paper Submitted
to the 2007 Annual Report of the Synchrotron Facility ANKA, FZ Karlsruhe, Germany
References 341
175. M.A. Denecke, K.H. Janssens, K. Proost, J. Rothe, U. Noseck, Environ. Sci. Technol. 39,
2049 (2005)
176. C.M. Sparks, E.P. Hastings, G.J. Havrilla, M. Beckstead, Powder Diffr. 21, 145 (2006)
177. B.M. Patterson, J. Campbell, G.J. Havrilla, X-Ray Spectr. 39, 184 (2010)
Chapter 8
Prospectives for l-XRF
8.1 Instrumentation
For special applications it can be also advantageous to select the target mate-
rials. For flexible use Rh-targets are the best choice (see also Sect. 2.1.3.4). But
for example in case of for the analysis of heavy elements a W-target can be helpful
because the bremsstrahlung background of the tube spectrum has a higher intensity
for high energies. Tubes with W-targets are common for most coating thickness
testing instruments and also for jewelry analysis.
The incident angle of excitation radiation is also important. For smaller spots
the normal incidence to the sample surface will be advantageous because in case of
a tilted incidence positioning errors due to not correctly adjustment of the distance
between optic and sample are possible which can be reduced or even avoided with
a perpendicular incidence.
In regards to the concentration optics improvements are possible and desirable.
The request for improvements of spatial resolution is valid every time. That
requires mainly smaller spot sizes. The spot diameter is physically determined by
the divergence due to total reflection inside the capillary and therefore by the
critical angle of total reflection and the working distance (see Sect. 2.3.3). A
reduction of the spot size down into the range of 10 lm is possible but only with
working distances in the range of \4 mm. This would limit the flexibility of the
instrument mainly and deteriorate the sample handling.
Optics influences also the excitation intensity due to their transmission function.
Improvements can be expected for the transmission efficiency. The transmission
efficiency of poly-capillary optics was growing in the last 10 years from a few
percent to actual approx. 10 %. Further improvements are possible and can be
expected.
Another interesting improvement would be wider energy ranges of high
transmission efficiency. Actual transmission functions (see Fig. 2.41) have high
efficiency for a limited energy range between approx. 3 and 12 keV. This range of
high efficiency can be shifted towards higher or lower energies. A wider range of
high transmission would be advantageous for a broad band excitation i.e. for
excitation of a large range of elements.
The detection of the fluorescence radiation nowadays is performed with SDD’s.
These detectors combine high count rate capability with sufficient energy reso-
lution. For energy resolution every time improvements would be helpful but the
SDD’s are already very close to the physical limits and these detectors are also
adequate for most applications. But for special purposes it can be interesting to use
more than one detector. This can be interesting for three reasons:
• It would increase the captured angle of radiation and therefore also the detected
intensity which can improve the sensitivity in particular for the detection of
trace elements.
• It can increase the speed for distribution measurements because a doubling of
intensity can be reduce the dwell time per pixel. If the detectors have different
directions to the incident beam the diffraction angle for the detectors are dif-
ferent and that could offer the possibility for the identification of Bragg-reflexes
by the comparison of the spectra measured with different detectors.
8.1 Instrumentation 345
The use of position sensitive detectors for scanning systems could be interesting
if higher read-out frequences are available and the analyzed area can be easy
adapted to different sizes and aspect ratios of the sample. It can be also interesting
for an improvement of spatial resolution with a longer distance between samle and
detector. But in the moment the position sensitive detection of large excited areas
has not real advantages against scanning systems as discussed in Sect. 3.2.2.
Sample positioning influences the efficiency of the complete analysis but
depends strongly on the actual analytical task. For single point measurements a
fast and correct positioning is important especially in case of small spots. Due to
the high count rates measurement times can be short. In that case the time for
sample positioning has a higher influence to the total analysis time.
For that reason fast and exact movement is requested but also a good sample
view i.e. a good sample illumination and variable magnification of the optical
microscope.
These requirements depend on the spatial resolution of the instrument. For spot
sizes generated by a collimator which are [0.3 mm often manual positioning is
sufficient. Motorized positioning for spot sizes in that range are helpful for auto-
mated multi-point measurements of measurement patterns. For spot sizes in the
lm-range motorized stages are required because only with them an exact posi-
tioning in these small dimensions is possible. In that case an automatic sample
positioning reduces the positioning time and avoids errors. An automatic posi-
tioning in Z-direction i.e. autofocus is already mostly available. But also an
automatized positioning in X-Y for example with pattern recognition procedures in
case of repeated measurements could be increase the efficiency tremendous.
Important for the correct sample positioning is also the optical image of the
sample. This image should have a magnification that allows the identification of
sample details in the dimension of the spot size. For small spots this is possible
only for a small part of the complete sample and therefore different magnifications
are required which allow both the presentation of small details but also an over-
view for larger sample areas. The switch between these magnifications can be
either discontinuous or also continuous.
For a high image quality a good sample illumination is helpful. For that purpose
homogeneous shadow-less illumination from different directions, in special cases
also transmitted light from the bottom can be helpful.
Another requirement is a flexible sample positioning. The typical movements
are in X-Y-Z-direction. For large wafers a movement in polar coordinates would
be allow the analysis of larger wafer diameters in the same sample chamber.
Another advantage of a sample rotation around a Z-axis in any sample position
would be the possibility to change the orientation of the crystallites in a poly-
crystalline sample. Then also the scatter angle and the position of diffraction peaks
would be changed which could allow the identification of diffraction peaks. But for
that purpose a sample rotation is required where the rotation center can be posi-
tioned in every measurement point i.e. the rotation axis has to be corrected with an
X-Y-movement.
346 8 Prospectives for l-XRF
according the intended application and according the accepted price range for a
special application.
The most interesting advances can be expected from improvements of focusing
optics. Here are, as discussed, smaller spots and wider transmission curves of
highest interest.
For instrument control and data evaluation there are several options for further
improvements which are targeted mainly to a faster analysis with reduced
uncertainties by the human interface—for example by mis-positioning of the
sample—and finally a higher analytical accuracy. This is valid for single point
measurements as well as for distribution analysis. Further there is a request for
high sophisticated evaluation procedures and their flexible use for the interpreta-
tion of distribution data i.e. for the extraction and objectivation of elemental
distributions.
In regards to instrument control mainly algorithms for sample handling have
potential for improvements. The sample positioning needs to be fast and accurate.
Actual this will be done in most cases manually. Only for repeated measurements
on similar samples automated procedures are already available. Automated sample
positioning procedures—for example with algorithms of pattern recognition—
would reduce positioning times as well as positioning errors.
Important is the compilation and archiving of all measurement conditions
together with the collected spectra to follow up the examination of the sample.
The general targets for data evaluation differ for single point, multiple point and
distribution measurements.
For single point measurements the main task is the quantification both of bulk
material and of layer systems. Accuracy and precision of quantification every time
can be improved. For non-homogeneous bulk material an important request would
be the consideration of influences by the environment of the measured volume (see
Sect. 4.3.1). This influence the measured intensities and therefore also the deter-
mination of mass fractions.
For the standardless quantification a better knowledge of fundamental param-
eters would be helpful to improve the accuracy both for bulk samples and layer
systems. For standard based analysis the implementation of the data of reference
samples and their easy handling is important for the use of them.
The exploitation of additional information about the analyzed material should
be possible by a more detailed evaluation of diffraction peaks in combination with
extended sample position possibilities. If a single crystallite could be analyzed in
different orientations i.e. if the sample is rotated or even tilted the diffraction peaks
changing the energy position. The knowledge of the sample chemistry from the
fluorescence analysis would allow the determination its compound and then its
possible d-spacings. These can be compared with the d-spacings determined from
348 8 Prospectives for l-XRF
Reference
1. B.K. Agarwal, X-ray Spectroscopy: An Introduction (Springer, Berlin, 1991). ISBN: 0-387-
092684
2. L.V. Azaroff, X-Ray Spectroscopy (McGraw Hill, New York, 1974)
3. B. Beckhoff, B. Kanngießer, N. Langhoff, R. Wedell, H. Wolff, Handbook of Practical X-ray
Fluorescence Analysis (Springer, Berlin, 2006). ISBN -10: 3540286039
4. E.P. Bertin, Priciples and Practice of X-ray Spectrometric Analysis (Plenum Press, New
York, 1975)
5. M.A. Blokhin, Methods of X-Ray Spectroscopic Research (Pergamon Press, New York, 1965)
6. T.G. Dzubay, X-ray Fluorescence Analysis of Environmental Samples (Ann Arbor Science,
USA, 1977)
7. H. Erhardt, Röntgenfluoreszenzanalyse: Anwendung in Betriebslaboratorien (Springer,
Berlin, 1989). ISBN-10: 3540186417
8. R. van Grieken, A. Markowicz, Handbook of X-ray Spectrometry(Practical Spectroscopy)
(Marcel Dekker Inc., New York, 2002). ISBN-10: 0824706005
9. P. Hahn-Weinheimer, A. Hirner, K. Weber-Diefenbach, Röntgenfluoreszenz-analytische
Methoden: Grundlagen und praktische Anwendung in den Geo-. Material- und
Umweltwissenschaften (Springer, Berlin, 2000). ISBN-10: 3540670211
10. K.H.A. Janssens, F.C.V. Adams, A. Rindby, Microscopic X-ray Fluorescence Analysis
(Wiley, New Jersey, 2000). ISBN-10: 0471974269
11. R. Jenkins, R.W. Gould, D. Gedcke, Quantitative X-ray Spectrometry (Marcel Dekker Inc.,
New York, 1981). ISBN8: 0-8247-1266-8
12. R. Jenkins, J.L. de Vries, Practical X-ray Spectrometry (McMillan, London 1967). ISBN:
0333006135
13. G. Lachance, F. Claisse, Quantitative X-ray Fluorescence Analysis (Wiley, New Jersey,
1994). ISBN: 0-471-95167-6
14. A. Rindby, K.H. Janssens, in Microbeam XRF, ed. by R.E. Van Grieken, A.A. Markowicz.
Handbook of X-ray spectrometry, 2nd edn revised and expanded. (Marcel Dekker Inc, New
York, 2002), pp 631–718
15. R. Tertian, F. Claisse, Principles of Quantitative X-ray Fluorescence Analysis (Heyden &
Son Ltd., London, 1982). ISBN: 0-855-01709-0
16. K. Tsuji, J. Injuk, R. van Grieken, X-ray Spectrometry: Recent Technological Advances
(Wiley, New Jeresy, 2005). ISBN-10: 047148640X
17. R. van Grieken, A. Markowicz, Handbook of X-Ray Spectrometry (Practical Spectroscopy)
(Marcel Dekker Inc, New York, 2002). ISBN-10: 0824706005
18. G. Zschornack, Handbook of X-ray Data (Springer, Heidelberg, 2007)
19. E.R. Cohen, Quantities, Units and Symbols in Physical Chemistry—the IUPAC Green Book,
3rd edn. (RSC Publishing, Cambridge, 2007)
A Angle
Absorption, 4 capture, 51
mass attenuation coefficient, 4 critical, of total reflection, 50
Absorption mode, 247 incident, 71
Analysis Application
art objects, 283 electronics, 295
banknotes, 315 forensic, 309
bore core, 271 Area analysis, 259
by difference, 235 Auger-electron, 6
cancerous cell, 293 Auger electron spectroscopy, 212
coating thickness, 236 Autofocus, 73
damascene structure, 299 accuracy, 73
drugs, 266
failure analysis, 305
faked coins, 311 B
filters, 302 Binary phase diagram, 301
frescos, 290 Bottom-up method, 196
gems, 269 Bragg
geological sample, 275 equation, 14
glass particles, 250 reflex, 115
gunshot residue, 309 Bragg-angle, 84
hidden pigments, 312 Bremsstrahlung, 20
inclusion, 254 Brightness
jewelry alloys, 233 of spectrometer, 86
metal particles, 249
minerals, 275
multi point, 258 C
of compounds, 283 Calibration curve, 159, 163
of pure Gold, 235 Calibration procedure, 163
paintings, 286 Capillary
pharmaceuticals, 291 cylindrical, 51
photographs, 284 intensity gain, 69
pigments, 327 single bounce, 54
precious metals, 230 spot diameter, 53
restricted elements, 256, 295 tapered, 53
roll bearing, 270 Characteristic radiation
smallest structures, 297 intensity, 22
Analyzed area Chemometric evaluation
size, 231 intensity ratios, 266
D
3D-analysis E
analyzed volume, 317 ED-XRD, 14
Darwin width, 36 Elastic scattering
Data set intensity distribution, 9
multi-dimensional, 136 Electron probe micro analyzer, 120
Dead time, 97 Element
Detection analysis, 7
efficiency, 98 distribution, 138
of ultra-light elements, 280 identification, 160
Detection efficiency Energy resolution, 93
high energies, 98 dependence on count rate, 107
low energies, 99 dependence on energy resolution, 107
very low energies, 99 Environment of the measured area, 167
Detector, 15 Error contribution
artifacts, 108 randomly, 191
CCD, 90 statistical error, 191
count rate capability, 102, 103, 106 systematic, 193
energy dispersive, 89 Error propagation, 190
energy resolution, 102 Errors
HP-Ge, 102 randomly distributed, 192
material, 94 Escape peak, 110
Index 353
N
Noise contribution, 93 Q
electronic, 94 Quantification, 157
energy dependent, 93 l-XRF, 163
flicker noise, 95 3D-analysis, 319
influence of count rate, 97 by concentration correction model, 162
parallel, 95 by fundamental parameter model, 168
sequential, 95 by intensity correction model, 162
element range, 157
errors, 188
O matrix interaction, 157
Optic Quantification in l-XRF
crystal, 37 special conditions, 163
diffraction, 35 Quantification methods, 160
HAPG, 41
HOPG, 40
mono-capillary, 51 R
primary, 15, 33 Radiation shielding, 71
refraction, 43 Rayleigh-compton-ratio, 13
secondary, 15 Refraction, 8
index, 8
Relative mode, 242
P Reproducibility, 188
Particle analysis, 150 Resolution
Pattern recognition, 154 energy, 85
Phase analysis, 150, 299 image quality, 79
Pigments on banknotes, 315 spatial, 77
Pigments on documents, 312 RoHS, 256
Index 355