Академический Документы
Профессиональный Документы
Культура Документы
V O L U M E XIV
EDITORIAL ADVISORY BOARD
EDITED BY
E. WOLF
University of Rochester, N.Y., U S A .
Contributors
J. C. D A I N T Y , A. L A B E Y R I E
L. A. RISEBERG, M. J. WEBER
M. A. D U G U A Y , G. S C H M A H L , D. R U D O L P H
P. J. V E R N I E R , P. J. B. C L A R R I C O A T S
1976
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photocopy-
ing, recording or otherwise, without the prior permission of the Copyright owner.
PUBLISHERS:
C O N T E N T S O F V O L U M E I1 (1963)
I. RULING.TESTINGAND USE OF OPTICAL GRATINGSFORHIGH-RESOLUTION
SPECTROSCOPY. G . W . STROKE. . . . . . . . . . . . . . . . . . . . 1-72
I1. THEMETROLOGICAL APPLICATIONS OF DIFFRACTIONGRATINGS. J . M . BURCH 73-108
111. DIFFUSIONTHROUGH NON-UNIFORM MEDIA.R . G. GIOVANELLI . . . . . . 109-129
I v. CORRECTION OF OPTICAL IMAGES BY cOMpENSATION OF h E R R A T I O N S AND BY
SPATIALFREQUENCY FILTERING. J . TSUJIUCHI . . . . . . . . . . . . . 131-180
v. FLUCTUATIONS OF LIGHTBEAMS.L . MANDEL . . . . . . . . . . . . . . 181-248
VI . METHODS FOR DETERMINING OPTICAL PARAMETERS OF THINFILMS. F.A B E L ~249-288
C O N T E N T S O F V O L U M E I11 (1964)
I. THEELEMENTSOF RADIATIVE TRANSFER.F. KOTTLER. . . . . . . . . . 1-28
I1. P . JACQUINOT
AFODISATION. AND B. ROIZEN-DOSSIER. . . . . . . . . . 29-186
111. MATRIXTREATMENT OF PARTIAL
COHERENCE. H . GAMO. . . . . . . . . 187-332
C O N T E N T S O F V O L U M E IV (1965)
I. HIGHERORDER ABERRATION THEORY.J . FOCKE. . . . . . . . . . . . 1-36
11. APPLICATIONSOF SHEARING INTERFEROMETRY. 0. BRYNGDAHL . . . . . . 37-83
111. SURFACE DETERIORATION OF OPTICAL GLASSES. K . KINOSITA. . . . . . . 85-143
IV . OPTICALCONSTANTS OF THIN FILMS. P . ROUARD .
AND P BOUSQUET . . . . 145-197
V. THEMNAMOTO-WOLF DIFFRACTION WAVE.A . RUBINOWICZ. . . . . . . 199-240
VI . ABERRATIONTHEORY OF GRATINGS AND GRATING MOUNTINGS. w.T . WELFORD 241-280
VII . DIFFRACTIONAT A BLACK SCREEN. PART I : KIRCHHOFFS THEORY. F. KOTTLER281-314
C O N T E N T S O F V O L U M E V (1966)
I. OPTICALPUMPING. C. COHEN-TANNOUDJI AND A . KASTLER. . . . . . . 1-81
I1. NON-LINEAR OPTICS.P . S . PERSHAN. . . . . . . . . . . . . . . . . 83-144
I11. TWO-BEAM INTERFEROMETRY. W . H . STEEL. . . . . . . . . . . . . . . 145-197
IV . INSTRUMENTSFOR THE MEASURINGOF OPTICALTRANSFER FUNCTIONS. K.
MURATA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199-245
V. LIGHTREFLECTION FROM FILMS OF CONTINUOUSLY VARYING REFRACTIVE INDEX.
R . JACOBSSON . . . . . . . . . . . . . . . . . . . . . . . . . . . 247-286
VI . X-RAY CRYSTAL-STRUCTURE DETERMINATION AS A BRANCHOF PHYSICAL
OPTICS. H . LIPSONAND C . A . TAYLOR . . . . . . . . . . . . . . . . 287-350
VII . THE WAVEOF A MOVINGCLASSICAL ELECTRON. J . PICHT . . . . . . . . 351-370
C O N T E N T S O F V O L U M E VI (1967)
I. RECENT ADVANCES IN HOLOGRAPHY. E. N . LEITHAND J . UPATNIEKS. . . 1-52
I1. SCATTERING OF LIGHTBY ROUGHSURFACES. P. BECKMANN . . . . . . . . 5369
I11. MWLIREMENT OF THE SECOND ORDER DEGREE OF COHERENCE. M . FRANCON
AND S . MALLICK .......................... 71-104
Iv. DESIGNOF ZOOM LENSES.K . YAM^ . . . . . . . . . . . . . . . . . 105-170
V . SOMEAPPLICATIONS OF LASERS TO INTERFEROMETRY. D . R . HERRIOTT . . . . 171-209
VI . EXPERIMENTAL STUDIES OF kT"TarITY FLUCTUATIONS IN LASERS.J . A . ARM-
STRONG AND A . W . SMITH . . . . . . . . . . . . . . . . . . . . . . 21 1-257
VII . FOURIER SPECTROSCOPY. G . A . VANASSEAND H . SAKAI . . . . . . . . . 259-330
VIII . DIFFRACTIONAT A BLACKSCREEN.PART11: ELECTROMAGNETIC THEORY. F.
KOTTLER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331-377
C O N T E N T S O F V O L U M E VII (1969)
I. MULTIPLE-BEAM INTERFERENCE AND NATURAL MODESIN OPENRESONATORS.
G. KOPPELMAN. . . . . . . . . . . . . . . . . . . . . . . . . . 1-66
I1. METHODS OF SYNTHESISFOR DIELECTRIC MULTILAYER FILTERS.E . DELANO AND
R . J . PEGIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67-137
111. ECHOESAT OPTICALFREQUENCIES. I. D. ABELLA. . . . . . . . . . . . 139-168
IV . IMAGEFORMATION WITH PARTULLYCOHERENT LIGHT.B. J . THOMPSON. . 169-230
V. QUASI-CLASSICAL THEORYOF LASERRADIATION. A . L. MIKAELIAN AND M . L .
TER-MIKAELIAN .......................... 231-297
VI . THEPHOTOGRAPHIC IMAGE.S . Oom . . . . . . . . . . . . . . . . . 299-358
VII . IN~RACTION OF VERYINTENSE LIGHTWITH FREEELECTRONS. J . H . EBERLY 359415
C O N T E N T S O F V O L U M E VIII (1970)
I. O ~ n c s J. . W . GOODMAN
SYNTHETIC-APERTURE . . . . . . . . . . . . . 1-50
I1. THEOPTICAL PERFORMANCE OF THE HUMAN EYE.G . A . FRY. . . . . . . 51-131
111. LIGHTBEATING SPECTROSCOPY. H . Z . C u m m s AND H. L . SWINNEY. . . 133-200
IV. MULTILAYER ANTIREFLECTION COATINGS. A . MUSSET AND A . THELEN. . . 201-237
V. STATISTICAL OF LASERLIGHT.H . RISKEN. . . . . . . . . . .
PROPERTIES 239-294
VI . COHERENCE THEORY OF SOURCE-SIZE COMPENSATION IN INTERFERENCE
MICROSCOPY.T. YAMAMOTO ..................... 295-341
VII . VISIONIN COMMUNICATION.L . LEVI. . . . . . . . . . . . . . . . . . 343-372
VIII THEORYOF PHOTOELECTRON COUNTING. C. L. MEHTA . . . . . . . . . 373-440
C O N T E N T S O F V O L U M E I X (1971)
I. GAS LASERSAND THEIR APPLICATION TO PRECISE LENGTHMEASUREMENTS.
A . L. BLOOM . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-30
I1. PICOSECOND LASERPULSES. A . J . DEMARIA . . . . . . . . . . . . . 31-71
111. OPTICALPROPAGATION THROUGH rn TURBULENTATMOSPHERE. J. W.
STROBMN . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73-122
IV . SYNTHESIS OF OPTICAL BIREFRINGENT NETWORKS. E . 0. AMMANN . . . . . 123-177
V. MODELOCKING IN GASLASERS. L.ALLENAND D. G .C . JONES . . . . . . . 179-234
VI . CRYSTAL OPTICS WITH SPATIAL DISPERSION. V. M . AGRANOVICH AND V. L .
GINZBURG. . . . . . . . . . . . . . . . . . . . . . . . . . . . 235-280
VII . APPLICATIONS OF OPTICAL METHODS IN THE DIFFRACTION m 0 R Y OF ELASTIC
.
WAVES K . GNIADEK AND J . PETYKIEWICZ . . . . . . . . . . . . . . . 281-310
WII. EVALUATION.
DESIGNAND EXTRAPOLATION
METHODSFOR OPTICAL SIGNALS.
B. R . FRIEDEN
FUNCTIONS.
ON USEOF THE PROLATE
BASED . . . . . . . . 31 1407
and the excitation and decay modes of the rare earths and also covers such
topics as relaxation by radioactive decay, multiphonon processes and
ion-ion interactions. Examples of some applications are also given.
The fourth article, written by M. A. Duguay discusses a useful recent
application of picosecond laser pulses, namely the development of ultrafast
shutters based on the optical Kerr effect. This device utilizes optically
induced birefringence to obtain gating times of the order of a few pico-
seconds. After a discussion of gating in different substances and the factors
that limit the resolution, some applications of the ultra-fast shutters are
described. Among them are their use in ultra-high speed photography,
which has made it possible, for example, to photograph a light pulse in
flight and the development of a technique for the sampling of ultra-short
optical signals by means of which molecular fluorescence signals can be
displayed on the picosecond time scale.
A relatively new technique for making diffraction gratings is described
in the fifth article, contributed by G. Schmahl and D. Rudolph. The
grating profile is provided by the intensity distribution of holographically
produced interference fringes that are stored on a glass blank coated with
a thin film of photoresist. After a presentation of the basic principles of
such holographic diffraction gratings, their production is described. The
properties of such grating are then discussed and their performance is
compared with that of gratings of more conventional type.
The sixth article, by P. J. Vernier, is concerned with a basic question con-
cerning the photoelectric effect, namely the origin of photoelectrons. It has
long been known that generally only a very thin layer of an irradiated solid
gives rise to photoelectrons. A more accurate knowledge is, however,
required in connection with efforts to improve photocathodes for use in
photometry and for a precise interpretation of the results of electron
spectroscopy. In this article the theoretical foundations of this subject are
first discussed. A detailed review is then presented of researches on the
escape depth of photoelectrons, and experimental and theoretical results
are compared. Investigations on surface photoexcitations are also reviewed.
The last article in this volume, contributed by P. J. B. Clarricoats,
presents an account of theoretical researches on optical properties of fibre
waveguides. Since about the early 19503, when glass fibres appear to have
been first seriously considered as optical elements, much research has been
conducted in this field. Today optical fibres promise to play an important
role in the field of telecommunications. In this article the basic properties
of optical fibre waveguides of various types are discussed and their relative
PREFACE XI
merits are brought out. The modal as well as the ray methods of treatment
are employed in analysing their properties.
This volume attests once again to the vigor and the breadth of current
research in optics.
July 1976
This Page Intentionally Left Blank
CONTENTS
4 . THEIMAGERECONSTRUCTION PROBLEM . . . . . . . . . . . . . . . . . . . . 76
4.1 The visibility phase problem with direct interferometry . . . . . . . . . .. 76
4.2 The triple interferometer . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3 The seeing compensation approach . . . . . . . . . . . . . . . . . . . . 78
5. CONSTRUCTION OF A SYNTHETIC-APERTURE ARRAYOF OPTICALTELESCOPES . . . . 19
6. INTENSITY INTERFEROMETRY . . . . . . . . . . . . . . . . . . . . . . . . . 82
7. HETERODYNE INTERFEROMETRY . . . . . . . . . . . . . . . . . . . . . . . . 84
8. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
VI . PHOTOEMISSION
by P. J . VERNIER
(DijoN)
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
1.1 The 3-step model . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
2. THEORETICAL BASISOF THE PE . . . . . . . . . . . . . . . . . . . . . . . . 250
2.1 Calculation of the density of absorbed photons (DAP) from the bulk dielectric
constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
XVI CONTENTS
BY
J. C . DAINTY
Queen Elizabeth College,
Campden Hill Road, Lonaon, W8 7AH U.K.
CONTENTS
PAGE
0 1. INTRODUCTION . . . . . . . . . . . . . . . . . . . 3
0 2 . NORMAL SPECKLE PATTERNS . . . . . . . . . . . 5
5 3 . PARTIALLY COHERENT ILLUMINATION . . . . . . 18
0 4. SURFACE-DEPENDENT FEATURES OF SPECKLE
PATTERNS . . . . . . . . . . . . . . . . . . . . . . 33
9: 5 . CONCLUDING REMARKS . . . . . . . . . . . . . . 44
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . 44
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . 44
9 1. Introduction
Fig. 1. Speckle pattern produced in the Fraunhofer plane of an optically rough diffuser
illuminated by a He--Ne laser.
where the unimportant phase factor has been ignored. Substituting for
A ( < ,q ) and evaluating the Fourier transform we obtain,
It is clear from this expression that the complex amplitude in the observing
plane is given by the sum of a large number of random phase and amplitude
vectors. As a result of the central limit theorem (see for example CHANDRA-
SEKHAR [1943], MIDDLETON [1960]) the random process A(x,y) tends to
a complex Gaussian process. The real and imaginary parts of the field are
identically distributed with zero mean and variance s2/2 and at any single
point they are statistically independent. The joint probability density
1 9 0 21 NORMAL SPECKLE PATTERNS I
function for the real and imaginary parts of the field, denoted by A R and
A , respectively, is given by
Provided that the phase of the scattered field at the scatterer is uniform
in the interval -n to and that the number N of scatterers is very large,
the complex amplitude of the speckle pattern will be a complex Gaussian
process regardless of whether the scatterers have a uniform or random
modulus. The Convergence of the process to a Gaussian form will in general
depend on the statistics of the aj and this is discussed further in 84.2. In
many practical situations the number of scatterers is so large that problems
of convergence are not important.
If the real and imaginary parts of the field have a joint Gaussian distribu-
tion, then it follows using a probability transformation that the modulrs
has a Rayleigh distribution, the intensity has a negative exponential distribu-
tion and the phase is uniformly distributed in the interval - n to n :
(4)
=o A<O
=o I<O
I
- THE STATISTICS OF SPECKLE PATTERNS
I
statistical fluctuation
I
- I
h\
\ 5 6 7 8
INTENSITY
Ye vertical
change
scale
xl00
Fig. 3. A measured histogram based on some 23000 intensity measurements taken from a
speckle pattern. Because the sampling aperture was small in relation to the speckle size,
the histogram should have the same form as the negative exponential reference curve which is
shown (MCKECHNIE [1974b]).
Fraunhofer plane, provided that a large number of scatterers lie with the
area of the point spread function of the imaging system in object space.
This condition is likely to be satisfied in many practical situations, but will
not hold for high resolution optical systems. The area of the point spread
function of an aberration-free lens is -approximately L2/NA2where NA is
the numerical aperture, whereas the minimum possible phase decorrelation
area of any diffuser is approximately L2 ;the maximum number of scatterers
contributing to a particular image point is therefore approximately equal
to 1/NA2.This maximum number may be further reduced depending upon
the actual phase decorrelation area of the diffuser used in practice. We
cannot expect the statistics of the intensity to follow the negative exponential
form for optical systems with numerical apertures greater than approx-
imately 0.1. The resulting statistics for small numbers of scatterers and
other surface dependent features are discussed in 0 4.
that the dimension of the finest structure in the scattered field is inversely
related to the effective diameter of the scatterer (just as the dimension of
the Airy disc is related to the diameter of the diffracting aperture).
The most general second order statistic of the intensity is the second
order probability function, p(Z1,I,) and was first derived by VON LAUE
[1916]. If 1, and Z2 are the intensities at two points (xl ,yl) and ( x 2 , y 2 )in
the Fraunhofer plane, then
where 4ois the modified Bessel function of zero order and CI2 is the
modulus of the autocorrelation function of the complex amplitude and
is given by
-m
where S(<,q) is the intensity distribution at the scattering plane. In the
derivation of (7) and (8), we assume that a large number of randomly
phased scatterers lie within the scattering aperture. It can be seen that
p(Z, ,Z2) is completely defined in terms of the autocorrelation function C , ;
it is a property of Gaussian processes that all probability density functions
are completely specified by the autocorrelation function. Thus in practice
the only second order statistic we need to evaluate is the autocorrelation
function or its Fourier transform, the Wiener spectrum.
The autocorrelation function of the intensity in the Fraunhofer plane of
a rough diffuser has been derived by many authors (VON LAUE[1916],
GOODMAN [19631, GOLDFISCHER [19651, SUZUKIand HIOKI[19661, Ross
[1970] and YAMAGUCHI [1972, 19733); here we follow Goodmans
approach. The autocorrelation function of the complex amplitude C,(xl ,yl)
is defined as
CA(X1,Yl) = (A(x+x,,y+y,)A*(x,y)), (9)
where ( ) denotes the ensemble average. (For a statistically stationary
speckle pattern the ensemble average may be replaced by a space average.)
The complex amplitude in the Fraunhofer plane of a rough scatterer is
given by eq. (2), and combination of equations (2) and (9) leads to
where the various cross-products reduce to zero upon taking the average.
10 THE STATISTICS OF SPECKLE P A T T E R N S [I, 2
-00
-m
-m
1, 21 N O R M A L SPECKLE P A T T E R N S 11
since u = </AD and v = q/AD. Thus the Wiener spectrum of the intensity
fluctuation is simply equal to the autocorrelation function of the intensity
distribution across the scattering aperture suitably scaled and normalised ;
it should be noted that
ir
-02
W(U,v)dudu = C(0,O) = o2 = ( I ) .
Since both the autocorrelation function and Wiener spectrum depend only
on the intensity distribution across the scattering aperture, the above
analysis applies also to planes other than the Fraunhofer plane, provided
that a large number of scatterers contribute to the intensity at any point
in the plane.
For a uniformly illuminated circular scattering area of radius r, the
autocorrelation function and Wiener spectrum of the intensity fluctuation
of the speckle pattern are given by
and
where y l ( x ) is the first order Bessel function and w, is the cut-off spatial
frequency given by w, = 2rfAD.
Several thousand values of the speckle pattern intensity are required to
accurately verify the above results and the experimental measurements of
HOHN[1968] and DAINTY [1970] only approximately verified the theore-
tical results. However recent experimental results by MCKECHNIE [1974a1
shown in Fig. 4 indicate that the expressions given above are verified to a
high degree of accuracy. The results of Fig. 4 also show that the roughness
of the scatterer does not influence the Wiener spectrum (or autocorrelation
function), provided of course that the illumination is perfectly coherent,
that the phase fluctuation at the scatterer is uniform in the interval --n:
to 71 and that a large number of scatterers contribute to the intensity at
any point in the Fraunhofer plane.
The autocorrelation function and Wiener spectrum of speckle patterns
produced in the image plane of rough diffusers may be found either by
using an extension of the Fraunhofer case (ENLOE[1967]) or by straight-
forward application of linear filter and square law detection theory (BURCK-
--
12 THE STATISTICS OF SPECKLE P A T T E R N S CL 2
T I I I ~ I x I ' 7 . ~
ERROR DUE TO S T A T I S T I C A L FLUCTUATION
-
z .L.
2
0
SPATIAL FREQUENCY ( t ) U N I T S OF
Fig. 4. Measured Wiener spectra of the intensity of speckle patterns observed in the Fraun-
-
hofer plane of two optically rough diffusers. The solid line is the form predicted using eq. (17)
(MCKECHNIE [1974a]).
A'(X>Y) = ~ ~ ~ ~ ~ , . Y l ~ P ~ x - x l ~ Y - Y ~ ~ d x l d Y l .
-m
If the diffuser structure is very much finer than the diameter of the point
spread function (this condition is also necessary for Gaussian statistics for
the image amplitude), the Wiener spectrum of the object amplitude WA(u,u)
will be constant for the values of (u, u) for which IT(u, u)I2 is significantly
non-zero, and combination of equations (18) to (20) gives, for this special
case of a white noise object,
m
r r
-m
The statistical properties we have evaluated so far all relate to the complex
amplitude or intensity at one or more points in a speckle pattern. In prac-
14 T H E S T A T I S T I C S OF S P E C K L E P A T T E R N S CL D 2
tice, speckle patterns are more likely to be averaged over some non-zero
area by, for example, a scanning aperture and we shall call this averaged
or integrated intensity the measured intensity. The measured intensity
Z(x, y ) is related to the intensity of the speckle pattern Z(xl, yl) by a con-
volution formula,
Z(x7y) = ~ ~ ~ ( x l . ~ ~ ~ ~ ~ ~ - ~ l , ~ - ~ l(22)
) d X , d y
-a
where
diameter of scanning aperture - 2a
k=
Rayleigh resolution limit of lens 0.61Aflr.
The quantity k is a measure of the aperture size relative to the speckle
size, since we showed in Q 2.2 that the speckle size is of the same order
of magnitude as the Airy disc. In Fig. 5, ob/(Z)is plotted as a function of
the relative aperture diameter for both top-hat and Gaussian scanning
apertures.
L O 21 NORMAL SPECKLE PATTERNS 15
01 I 1
0 2 4 6 8 10 12
K
Fig. 5 The standard deviation of the intensity fluctuations relative to the mean intensity for
top-hat and Gaussian scanning apertures for a speckle pattern formed by an optical system
with a top-hat pupil function. For the top-hat aperture, -, k = (diameter of scanning
aperture)/(Rayleigh resolution limit of lens). For the Gaussian aperture, k = (diameter
of aperture at l/e points)/(Rayleigh resolution limit of lens).
c$ N W(0,O)
rJ
-W
(b(u,v)lz dudv.
where d b is the area of the scanning or averaging aperture, and d pis the
area of the pupil in spatial frequency units (i.e. d,= nvz/ln2f2).
Finding the first order probability density function for the measured
intensity, p(I), is not straightforward. The problem is analogous to that
of square-law detection and low-pass filtration in the one-dimensional
time domain, which was analysed by KACand SIEGERT [19471 and SLEPIAN
[1958]. The extension* to the case of the measured intensity in a speckle
* All three authors in fact made errors in extending the results of K A c and SIEGERT[1947] ;
these arose as a result of failing to include all aspects of the fact that the amplitude is a complex
Gaussian process.
16 T H E S T A T I S T I C S OF S P E C K L E P A T T E R N S [I7 2
pattern was made by CONDIE [19661, DAINTY [19711 and BARAKAT [19731.
Equa.tion (22) for the measured intensity may be rewritten in terms of the
complex amplitude in the speckle pattern A(x, ,yl) as,
I(x, Y ) = I.j
-02
14x1, Y1)I2B(x-x1, Y-Y1)dX, dY,.
Y) =
.r9r
-w
C A ( ~ - ~ I , Y . - Y ~ )Yl)$n(Xl,
B ( ~ , , y,)dx1dy1, (28)
p(T) = -
2n s -w
exp (-izZ)
(1-i.z~~)
dz,
(1) = Eon0
1, Q 21 NORMAL SPECKLE PATTERNS 17
and
0; = &;no.
0 1.0 2.0
Fig. 6 . The approximate probability density function of a normal speckle pattern scanned by
an aperture. This distribution has two parameters, the mean intensity (assumed to be unity)
and the standard deviation ubof the intensity fluctuation (relative to the mean intensity).
18 THE STATISTICS OF SPECKLE P A T T E R N S CL 3
Speckle patterns were first observed in situations where the light incident
on the scatterer was both spatially and temporally partially coherent, and
the properties of such patterns are somewhat less straightforward than
those for perfectly coherent illumination. For each case of spatial and
temporal partial coherence we again make the subdivision into speckle
patterns formed in the region of the Fraunhofer plane and those formed in
the image plane, although a perfectly general theory may of course be used
to apply to both cases (PARRY[1975b]).
In Fig. 7 we show an optical system that might be used to form a speckle
pattern in the Fraunhofer plane of a diffuser illuminated by spatially
partially coherent light. The diffuser is uniformly illuminated by light
which has a mutual coherence function r((,,tZ,q l ,q 2 ) given by the
so urce L1 L2
Fig. 7. Formation of a speckle pattern produced in the Fraunhofer plane ofa diffuser illuminat-
ed by spatially partially coherent light.
1, o 31 PARTIALLY COHERENT ILLUMINATION 19
I(x, Y ) = ij.
-m
m1-52, ?l-YIZ)H(Sl, ?)Z(tl,? ) H * ( 5 2 , v ] 2 ) 2 * ( 5 2 3 r z )
u = - 51-4;2 , v=- 91 - Y 2
Af Af
we obtain
x exp { - 2 4 u x + uy)} du dv
20 THE STATISTICS OF SPECKLE P A T T E R N S [I, 0 3
which further simplifies to
1(x7Y ) = S(X, Y) e 1 rJ
-m
H(Afu, Afu)Z(Afu,A f u ) exp { - 2ni(ux + uy))dudu ,
where no is the number of (coherent) speckles that lie with the source
function s(x, y).
The Wiener spectrum of the intensity fluctuation in a speckle pattern
formed by spatially partially coherent illumination of a diffuser is given by
wu, 0) = K(U,W W u , nfo)12 (34)
where r(5,q) is the mutual coherence function and Wc(u,u) is the Wiener
spectrum of a coherent speckle pattern. Equation (34) is the basis of a
method of measuring the mutual coherence function ;providing the diffuser
satisfies the conditions laid down in 0 2, the function Wc(u,u) is of a known
form (eq. (21) in the present notation) and by measuring W(u,v ) the squared
modulus of r(<, q) can be found. An example of an experimental result is
given in Fig. 8; the same basic method is also used in astronomy for the
determination of stellar diameters and binary star separations.
I, 5 31 PARTIALLY COHERENT ILLUMINATION 21
separation 2 mm
Fig. 8. An experimental measurement of the spatial coherence of a light source using speckle
patterns (FUJIIand ASAKURA [1973]).
C%Y) =~~~~(~l-xz,Y~-Yz)h(x-x~,Y-Yl~
-m
Fig. 9. Formation of a speckle pattern produced in the image plane of a diffuser illuminated
by spatially partially coherent light.
22 THE STATISTICS OF SPECKLE PATTERNS [I, 0 3
where h(x,y ) is the amplitude point spread function of the imaging system
and Z(x,y) is the complex amplitude transmittance of the diffuser. In
terms of the effective source J(u, v) = s(a/AD,b/AD),
m
j j T(x, y ) exp {
n n
I(x, Y ) = ?;s -m
J(u,v)lQ(x,Y; u, 41, du dv, (36)
Q ( x , Y ; ~0 ), =
S:F
-00
h(x-x,,y-yi)Z(xi,y,)ex~ f - 2 z i ( u x i + v ~ i ) ) d x i d ~ i .
Equation (36) states that the speckle pattern produced in the image plane
in spatially partially coherent illumination is just the (weighted) sum of
the intensities of speckle patterns produced by different angles of coherent
illumination. The exact first order probability density function is therefore
given by an expression similar to eq. (27), where the eigenvalues E, are
given by an expression similar to (28). The approximate probability density
function is given by equation (30), where the value of no is approximately
equal to the ratio of the diameter ofthe point spread function to the diameter
of the coherence patch at the diffuser. In an optical system which has an
illuminating condenser of numerical aperture NA, and an imaging objec-
tive numerical aperture of NA, then no N Y = NA,/NAo (for no B 1).
The Wiener spectrum of the intensity fluctuation in a speckle pattern
produced in a spatially partially coherent optical system has been derived
for optically very rough surfaces by DAINTY [19701and YAMAGUCHI [19731,
and for surfaces of an arbitary magnitude of roughness but with a Gaussian
phase profile by FUJIIand ASAKURA [1974c]; this latter case is discussed
in 8 4.3, and only the result for a very rough surface is found here.
Equation (35) for the image intensity may be rewritten in the form
I(X) = rJ
-m
V ( x , , ~~)Z(x-x,)Z"(x-x,)dx,dx,,
PARTIALLY COHERENT ILLUMINATION 23
where
x ((Z(x-x,)Z*(x-x,)Z(x+x, - x , ) Z * ( x + x , -x,))
- (Z(X- xJZ*(x- xz))(Z(x +xo - X3)Z*(X+xo- x,))}
x dx, dx, dx, dx,. (37)
-00
dx1 dxZdx3dx4,
x (Z*(X-X,)Z(X+X,-X~))}
r{
or in terms of the autocorrelation of the diffuser complex amplitude Cz(x),
-00
where Wz(u)is the Wiener spectrum of the object amplitude, and 9-(ul, u,)
is the transmission cross-coefficient (BORNand WOLF[1970]) and is given
by
24 THE STATISTICS OF SPECKLE PATTERNS [I? 03
+
J(u)H(u u,)H*(u - u2)dul du2.
and for a diffuser whose lateral structure is small compared with the point
, (40)
ul,ul)l2 du, .
IS(U+ (41)
It should be noted that the result of eq. (41) for a white noise diffuser is
not restricted to the case of Gaussian statistics for the diffuser complex
amplitude transmittance ; if the function Z(x, y ) has an autocorrelation
function that is a delta-function (i.e. white noise), then it can be shown
that eq. (41) must follow regardless of the statistics of Z(x,y).
In Fig. 10, the scale value W(0)of the Wiener spectrum is plotted as a
function of the coherence parameter Y
radius of effective source NA,
Y = --
radius of entrance pupil NA, .
Fig. 10. The zero spatial frequency value of the Wiener spectrum of the intensity fluctuation
in a speckle pattern formed by an aberration-free partially coherent optical system with a
top-hat effective source distribution, plotted as a function of the coherence parameter 9
(DAINTY [1970]).
1, o 31 PARTIALLY COHERENT ILLUMINATION 25
Fig. 1 1 . Speckle patterns produced in the Fraunhofer planes of diffusers with three surface
roughnesses illuminated by an argon laser (wavelengths present 514, 496, 488, 476 nm);
(a)uh u lpm,(b)u, z 3pm,(c)uh u PARRY [1974b]).
PARTIALLY COHERENT ILLUMINATION 21
[1974a, b, 1975a, b] and PEDERSEN [1975a, b]. The results given so far
in this article have applied to all surfaces whose r.m.s. roughness is greater
than one wavelength, regardless of the detailed statistical properties of
the surface. However we can no longer accept this naive picture of the
scattering surface and it is necessary to introduce surface dependent param-
eters using the Beckmann model (BECKMANN and SPIZZICHINO [19631).
We consider the scattering geometry shown in Fig. 13 (PEDERSEN [1975b]).
The surface lies in the (x,y)~ x p l a n eand has a profile h(x) whose r.m.s.
fluctuation is CT,, > A. The surface is illuminated by a unit amplitude plane
wave with wave vector k , , and the scattered wave k is observed in the
Fraunhofer plane, where (kl = Ik,J = 27c/A. The scattered amplitude is a
\ incident wave
Z
I
surface profile
h(L0
Fig. 13. Diffraction geometry for the formation of speckle patterns (PEDERSEN
C1975b-j).
28 T H E STATISTICS OF SPECKLE P A T T E R N S 11, 5 3
I(m) = lLm2(k)I(km)dk.
and provided that this mean is locally stationary over regions of speckle
correlation, the normalised angular autocorrelation function Cw(m ,m z ) ,
of the white light speckle intensity fluctuation is given by
is the r.m.s. path deviation of the light contributing at the point m in the
speckle pattern and e, is a unit vector in the z direction. Equation (46) for
the angular autocorrelation of the intensity can now be evaluated to give
30 THE STATISTICS OF SPECKLE PATTERNS [I. P3
Let the mean path vector be s = (sl +s,)/2 and a difference path vector be
d = s 2 - s l . In the limit d = 0 eq. (48) reduces to
C,(O; s) = a2/(I>2 = l/JiXmy, (49)
where s is in general given by eq. (47); for normally incident illumination
and viewing at the centre of the Fraunhofer plane s = 2a, and we obtain
a2/(I>2 = l/Jiqmq. (50)
This simple expression clearly shows how the speckle contrast decreases
as either the surface roughness a),or the illumination bandwidth Wincreases.
If we now restrict our problem still further by assuming that we are
using relatively narrow band sources such that W <QC k , , then eq. (48)
reduces to
In order to show how the speckles are elongated and in what way this
elongation depends on the surface and illumination properties, it is con-
venient to assume that the illumination is normal to the scattering surface
and consider paraxial diffraction angles. We then have (Fig. 14),
s = i s x , s,, SJ ={re, 0, 2 4 9
Fig. 14. Diffraction geometry for the calculation of polychromatic speckle correlation with
normal incidence (FEDERSEN [1975b]).
I, D 31 P A R T I A L L Y COHERENT I L L U M l N A T I O N 31
where the x-axis lies along s, and 8 is the polar diffraction angle. Similarly,
d = {rAOX,rAO,, 0},
where A8, and A8, are the components of the change in diffraction angles
in the radial x-direction and the azimuthal y-direction respectively. Equa-
tion (51) becomes
Cw(A@;8)= Cw(0;8)exp
It can be seen from eq. (52) that the degree of angular correlation is
locally stationary with an elliptical Gaussian correlation function. In the
azimuthal direction, the width of the correlation function (i.e. the speckle
width) is determined only by the normal diffraction value,
This is illustrated in Fig. 15. The elongation of the speckles in the radial
direction due to angular dispersion is given by
I
,"'y /
diffr
/
>
I 0,
I I
I
I
!
I f *I
Fig. 15. Illustration of the correlation regions in polychromatic speckle patterns. On axis the
correlation region is diffraction-limited and circular. Off axis the angular dispersion causes
the region to be radially elongated (PEDERSEN [197Sb]).
32 THE STATISTICS OF SPECKLE PATTERNS CL 3
where w, is a correlation bandwidth given by
w, = w / d T Z . (54)
It is clear from eqs. (53) and (54) that the speckle length in the radial direc-
tion depends on both the surface roughness a, and the illumination band-
width W.
The probability density function of intensity in a polychromatic speckle
pattern can be found by considering the pattern to be the sum of a number
of partially correlated monochromatic speckle patterns, the degree of
correlation depending upon position and surface roughness. Using an
analysis similar to that given in $2.3 for the first order statistics of the
measured intensity in a normal speckle pattern, it can be shown that the
probability density function of intensity in a polychromatic pattern is
approximately given by a gamma variate (eq. (30)) or by exact expressions
similar to eqs. (27) and (28) (PARRY[1975a]).
The above results indicate that the observation of white light speckle
patterns formed in the Fraunhofer plane may produce useful measures of
surface roughness and this has been suggested by SPRAGUE[1972] and
TRIBILLON [19741, both of whom present preliminary experimental results.
However, it must be borne in mind that the above analysis contained a
large number of assumptions. In particular it was assumed that single
scattering occurred (first Born approximation) and that the surface could
be characterised either by a Gaussian distribution of height with non-zero
correlation length, or by a white-noise distribution; it is not clear whether
these assumptions are valid in many practical scattering problems.
Speckle patterns formed in the image plane of a diffuser illuminated by
spatially coherent, polychromatic light have been studied by ELBAUM,
GREENBAUM and KING [1972], GEORGE and JAIN[1972, 1973, 19741, and
MCKFXHNIE[1975]. The main application here is the reduction of speckle
in the images of diffuse objects. The analysis of the statistics is very similar
to the above case for the Fraunhofer plane. The main results are (i) the
speckle pattern is statistically stationary for a uniform diffuse object
provided that the aberrations of the imaging system are not field-dependent,
(ii) the speckle contrast decreases as the bandwidth W of the illumination
increases and as the surface roughness a, increases, (iii) the speckle pattern
depends on the aberrations of the imaging system. GEORGE and JAIN[1974]
and MCKECHNIE [1975] have given detailed analyses of the statistics using
Goodmans model of the scattering suface.
S U R F A C E - D E P E N D E N T FEATURES 33
cmn = Ipmn12,
or
= 0 otherwise,
= 0 otherwise,
I I I I I I
1.0 2.0 30
I
Fig. 16. The probability density functions for the intensity of the sum of two partially
correlated normal speckle patterns with equal mean intensity and c,* = 0, 0.6 and 1.0
(GOODMAN [1975a, c]).
5
random medium
W(4 = rSH*(;,)rr(C)H(i,+;)Hli2+<~
--m
where u = (/A$ It is clear from this equation that the Wiener spectrum of
the speckle depends on the fourth order moment of the complex amplitude
of the scattered field. If this complex amplitude is a white noise process,
the fourth order moment can be expressed in terms of delta functions, and
the expression for the Wiener spectrum reduces to that given by either
1, o 41 SURFACE-DEPENDENT FEATURES 37
eq. (15) or (21) in 92. To consider the effect of the autocorrelation of the
complex amplitude on the Wiener spectrum of the speckle intensity, we
consider a particular example in which it is assumed that A(5) is a complex
Gaussian process (DAINTY [1973]). The fourth order moment can then
be written in terms of the autocorrelation function and eq. (55) becomes,
after some rearrangement,
-m
0 1.0 W 2.0
I 4\
0 I I
0 1*o W 2-0
Fig. 18. Wiener spectra of the intensity fluctuation for speckle patterns observed in the Fraun-
hofer plane of a defocussed lens for defocus values of 0, 1 in,2/n and 4/nwavelengths. TJpper
(a), R = 0.2; lower (b), R = 0.1 The parameter R is equal to the diameter of the correlation
area relative to that of the optical system (DAINTY [1974]).
N
+-
4N
exp
L I I I I I I
0 0.2 0.4 06
sin2e
Fig. 19. Angular dependence of the speckle contrast for a small number of scatterers
(PUSEYand JAKEMAN [1975]).
can also be found using eq. (57) and evaluating the appropriate moments
40 THE S T A T I S T I C S OF S P E C K L E P A T T E R N S [I, 0 4
T
3
E
Area- x103 ( rn-2)
Fig. 20. Dependence of speckle contrast on illuminated area of diffuser for a fixed size of
inhomogeneity (PUSEY and JAKEMAN[1975]).
where small 8 and 8 have been assumed for simplicity, A 8 = O-Q, and
where C,(A8) is the autocorrelation function of the normal speckle pattern
produced when N + co as given by eq. (14). Clearly eq. (59) gives the
correct result of C(8,6) -+ C,(A8) as N + 00. However for finite N two
characteristic scale lengths appear in the autocorrelation function. The first
is the normal speckle size governed by the total extent of the scattering
volume, whilst the second depends on the variance and scale length of the
phase immediately behind the scatterer. This result is essentially the same
as that given in eq. (56) but derived and used in a different context (DAINTY
[1973]). Some experimental measurements of C(6,e) in the scattering by
liquid crystals are shown in Fig. 21 and broadly confirm the above theoret-
ical results. The statistics of speckle patterns produced when only afinite
number of scatterers contribute to the pattern at any point clearly depend
S URF ACE - DE P E NDE NT FEATURES 41
m
'r 2 0 volts
01
t
10
1 I
20
I I
30
I
40
I I
50
I J
O2 degrees
Fig. 21, Angular autocorrelation functions in the Fraunhofer plane for speckle patterns
produced by many scatterers (100 v) and a few scatterers (20 v) (PUSEY and JAKEMAN [19751).
I
r =lo
IRRADIANCE I
Fig. 22. (a) Experimental probability distributions of speckle intensity in the images of
slightly rough diffusers (FUJIIand ASAKURA [1974b]). (b) Probability density functions for
a range of specular/diffuse beam ratios (DAINTY [1972]).
1, o 41 S U R F A C E - D E P E N D E N T FEATURES 43
The final term on the right hand side of eq. (60), (oh/r,&is closely related
to the r.m.s. gradient of the surface height distribution and this equation
indicates that a simple measure of speckle contrast in the image cannot
uniquely determine the surface roughness.
Recently OHTSUBO and ASAKURA [1975] have observed that the speckle
contrast is somewhat smaller in the image plane than in out-of-focus planes.
GOODMAN [1975b, c] has shown that this is because the scattered field
does not have circular Gaussian statistics (as assumed above) and that
Fig. 23. Speckle contrast as a fmction of roughness characteristic a:/rb for four conditions
of partial spatial coherence (a) - (d) (FUJI[and ASAKURA[1974c]).
44 T H E STATISTICS O F S P E C K L E P A T T E R N S CI
8 5. Concluding Remarks
We began this review article by stressing that in a large number of practical
situations normal speckle patterns are produced ; the statistics of these
speckle patterns are well-established both theoretically and experimentally
and in particular they do not depend on the detailed scattering properties
of the scattering medium. When examining the effects of polychromatic
illumination on speckle statistics we found the value of the r.m.s. height
variation of a scattering surface strongly influenced the statistical properties
of the scattered intensity. Finally in 54 it was seen that the general relation-
ship between the scattered intensity and the scattering surface may be very
complicated, even for surfaces with a Gaussian distribution of surface
heights. However, it must be stressed that through the whole review we
have glossed over the details of the interaction of an electromagnetic wave
with a scattering medium and have used relatively naive models of scattering
surfaces. Hopefully this unsatisfactory state of affairs in the subject as a
whole will be rectified by thL: advances of future workers.
Acknowledgements
I wish to thank Professor W. T. Welford of Imperial College, London
for his encouragement, advice and criticism of my work on speckle patterns
over a number of years. I also wish to thank Dr. G. Parry for his advice on
the original manuscript.
References
ALLEN,L. and D. G. C. JONES, 1963, Phys. Lett. 7,321.
ARCHBOLD, E. and A. E. ENNOS,1972, Opt. Acta 19, 253.
ASAKURA, T., H. FUJIIand K. MURATA,1972, Opt, Acta 19, 273.
BARAKAT, R., 1973a, Opt. Acta 20,729.
11 REFERENCES 45
I1
BY
A. LABEYRIE
Observatoire de Paris,
92190 Meudon, France
CONTENTS
PAGE
Q 0. INTRODUCTION . . . . . . . . . . . . . . . . . . . 49
Q 1. ATMOSPHERIC OPTICS . . . . . . . . . . . . . . . . 51
Q 2 . DIRECT INTERFEROMETRY . . . . . . . . . . . . . 59
Q 3 . INTERFEROMETER DESIGNS AND RESULTS . . . . . 64
Q 4 . THE IMAGE RECONSTRUCTION PROBLEM . . . . . . 76
5 5. CONSTRUCTION OF A SYNTHETIC-APERTURE
ARRAY O F OPTICAL TELESCOPES. . . . . . . . . . 79
Q 6. INTENSITY INTERFEROMETRY . . . . . . . . . . . 82
Q 7 . HETERODYNE INTERFEROMETRY . . . . . . . . . . 84
Q 8. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . 84
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . 85
0 0. Introduction
0.1. INTRODUCTION
0.2. HISTORY
0 1. Atmospheric Optics
1 . 1 . THE ATMOSPHERIC HETEROGENEITY
The optical wave received at ground level from a point source located
above the atmosphere has both phase and amplitude perturbations. The
amplitude fluctuation pattern, also referred to as the shadow pattern, is
responsible for the well known twinkling of stars. Because absorption
cannot account for the effect, it is generally interpreted as resulting from
the action of high-altitude turbulent layers according to elementary stio-
scopic or Scblieren effects. Both geometric ray deflection and interference
may contribute in this effect.
A simple method for viewing the high-altitude turbulent layers responsible
for the shadow pattern was used extensively by BOYER[unpublished] for
wind monitoring purposes. It consists in observing the lunar limb with a
small, amateur-type telescope. By defocusing slightly the eyepiece outward,
it is possible to focus on the turbulent layers and to see the flowing stream
of turbules, their invisible phase pattern being translated into an amplitude
pattern by a Schlieren effect. The altitude, velocity and direction of the
turbulent flow may thus be determined. More elaborate methods involving
large mirrors and shadow pattern correlations have recently been
developped by Roddier and his collaborators (MARTIN,BORGNINO and
RODDIER [1975]).
Phase corrugations on the wave are generated by turbulence at all alti-
tudes, but with increased efficiency at decreasing altitudes. The low-altitude
phase cells are observed easily on the aperture of large telescopes when
conducting knife-edge tests. In addition to the steady cell flow outside of
the dome, a stationary turbulence component occurring inside the dome
is also generally observed. RMS values for the fast phase fluctuation have
been estimated to vary in the range from 1 to 20 radians. Available data
do not cover the long-wave components of phase corrugations.
aperture sizes are smaller than seeing cells. Thus the phase is nearly uniform
over their aperture. Consequently, these telescopes are nearly diffraction
limited, and star images have the appearance of the classical Airy disk.
Considering now the other extreme case of very large instruments, their
aperture may cover several thousand seeing cells, having random phases.
The Airy-disk pattern is completely destroyed under such conditions ; and
classical diffraction experiments, in accordance with elementary Fourier
transform analysis, show that the angular spread of the image is of the order
of Ajd, d being the characteristic size of seeing cells. 12-cm cells thus corre-
spond to one arc-second as the approximate width of the image projected
onto the sky plane; and this figure is rather typical of average observing
conditions with large instruments in the best observatory sites. Only very
exceptionally are seeing cells larger than 24 cm; and this explains why
larger instruments do not benefit from their theoretical resolution per-
formance.
Some astronomers, especially binary star observers, have long remarked
on the existence of a fast-moving fine structure inside the typical one-second
image (Fig. 1). Recently (LABEYRIE [1970]), this structure has been inter-
preted as a speckle phenomenon, same as the well known effect in diffused
laser beams, and this identification led to the interferometric method
called speckle interferometry (sect. 3.3). As is apparent in the recent
review of DAINTY [1976], the speckle phenomenon has been extensively
studied in the recent years. In the astronomical case, it results from the
fact that any point inside the star image indeed receives coherent contribu-
(a) (b)
Fig. 1. The corrugated optical wave and corresponding image; (a) the full aperture of a large
telescope produces a speckled spread function; (b) a Fizeau mask with two small apertures
produces Youngs fringes in the image.
11, 4 1 ATMOSPHERIC OPTICS 55
tions from many of the phase cells in the aperture plane. The resulting
amplitude at the image point considered is thus a sum of many vibrations
with random phases. Summing vibrations with random phases has been
a classical problem since Rayleigh: the squared modulus of the sum may
have any value between 0 and IZaI2,a being the amplitude of the individual
vibrations. In the image plane, the particular value found in the summing
process depends on the image point considered since individual vibration
phases vary when the point is displaced in the image plane by more than
l/a,a being the angular aperture. The scale size for intensity variations in
the image plane is thus l/u.As discussed by GOODMAN [1965], in the case
of laser speckles, and by KORFF,DRYDEN and MILLER[1972], in the
astronomical case, the addition of coherent, but randomly phased, vibra-
tions corresponds to a two-dimensional random walk in the complex
plane if component vibrations are represented by vectors in this plane.
The sum vector is distributed according to a Gaussian law, and the ampli-
Fig. 3. Synthetic aperture systems and corresponding spread functions : top row - apertures
consisting of 1, 2 and 6 telescopes, as well as giant monolithic aperture; middle row - corre-
sponding spread functions, in the absence of atmosphere and optical aberrations; bottom row -
short-exposure, monochromatic, spread-functions in the presence of the atmosphere. This
is a laboratory simulation result obtained by photographing a monochromatic point source
(a laser-illuminated pinhole) through a mild diffuser and aperture mask. The conditions
simulated correspond to 1.5 'meter telescopes and typical 1 sec. seeing.
n, 11 ATMOSPHERIC OPTICS 57
plicative term applied to the infinite incident wave, and the speckles in the
image plane are thus convolved, in complex amplitudes, with the diffraction-
limited spread function. Speckled images recorded at the Palomar 200-inch
telescope are reproduced in Fig. 2. The laboratory equivalent of synthetic-
aperture systems involving several large telescopes produces the speckled
images in Fig. 3. These images were obtained as photographs of a mono-
chromatic point source, using a multi-aperture diaphragm and a mild
diffuser in front of the camera lens. Compared to the speckles from the
monolithic aperture, these contain an additional finer interference struc-
ture which takes the form of fringes with two apertures, of a honeycomb
pattern with 3 apertures, etc..
The number of speckles contained within the image envelope increases
with aperture size D as (D/dj2.When dealing with low-altitude seeing,
no ensemble translation of speckles occurs in the image even when seeing
cells are carried as a rigid pattern by the wind flow. Instead, speckles
appear and disappear locally much like vapor bubbles at the surface of
boiling water. It is apparent from the video images recorded at Palomar
that the lifetime of speckles increases at increasing wavelengths, from near
ultra-violet to near infra-red.
In small telescopes, few seeing cells are present across the aperture at
any instant, which results in few speckles in the images. Consistently with
simple statistics, experience shows that such images are subject to rapid
wander and size fluctuations. In such cases, image selection or exposure
triggering techniques such as used by RATT [1957], may improve markedly
the image sharpness. Telescopes up to a meter in aperture diameter may
thus produce nearly diffraction-limited images durilig a few milliseconds
every hour or so. Using short-exposure electronographic photographs of
binary stars, R ~ S C HWLERICK
, and BOUSSUGI~ [1961] have shown that the
instantaneous speckle patterns are identical for closely spaced pairs but
different in the case of widely spaced pairs. The angular extent over which
speckle is invariant, called the isoplanatic patch, is on the order of 3 to
10 arc-seconds. Patch size .is mainly dependant upon the presence of
turbulent layers at high altitudes, since these layers are crossed in different
regions by the light beams coming from different stars into the telescope.
Temporal and spatial coherence have both been assumed in the above
analysis. The size d of speckles being proportional to wavelength, some
degree of monochromaticity is indeed required for purity of the pattern.
If atmospheric and telescopic aberrations are smaller than the wavelength
2, the requirement may be written I/d2 > D/d. In white light, at the 200-inch
telescope, the image of a star close to zenith shows contrasted speckles
58 HIGH-RESOLUTION TECHNIQUES [II, 1
at the center of the typical one-second envelope. The few central speckles
are surrounded by radially-oriented coloured streaks generated by the
chromatic spreading of speckles at higher interference orders. This is well
observable visually in conditions of moderate wind speed, using a strong
eyepiece. It takes a filter with 200 fingstrom or narrower bandpass to
observe pure speckles all the way to the edge of a one-second image. When
pointing at a star lower toward the horizon, dispersion tends to elongate
the speckles and it takes a much narrower filter to remove this effect unless
some form of prismatic compensation is used. The experimentally-observed
proportionality of speckle size to wavelength supports the diffraction-
interference interpretation of the speckle phenomenon, as opposed to the
ray-deflection interpretation which had sometimes been proposed before
the advent of lasers and speckle theory.
The optical components of large astronomical telescopes are rarely
made to the accuracy meeting the Rayleigh criterion. Residual coma,
spherical aberration and astigmatism amounting to 0.5 arc-second are
usually tolerated since the effect of the atmosphere is even worse. Such
transverse aberrations, as long as they remain inferior to seeing effects,
have no influence on the speckle patterns : somewhat paradoxically, bright
speckles retain their similarity with Airy peaks in the presence of aberrations
which would destroy the Airy peak if they acted alone.
Realistic simulations of astronomical speckle phenomena may be carried
out in the laboratory using a bright artificial star, a sheet of polyethylene
or other diffusing material representing the atmosphere, a lens aperture to
represent the telescope, and filters. In addition to laboratory simulations,
a number of authors have used computers to derive image spread functions
by Fourier-transforming random distributions of phase cells generated
across some circular aperture. In all cases, the speckle patterns obtained
resemble closely those obtained with large telescopes.
(MT F
It has been shown that the RMS profile, obtained by averaging the squared
modulus of successive Fourier transforms, is identical to that for the
diffraction-limited MTF except in the central region where a central peak
is added.
The long-exposure MTF, obtained as the simple time-average of the
short exposure MTF, or equivalently as the Fourier transform of the
long-exposure image, consists of the central peak only, the feet being
cancelled in the averaging process.
Rather than being Fourier transformed, the star image may be auto-
correlated. The central peak appearing in the short-exposure case may be
shown to be identical to what would be obtained under diffraction-limited
conditions. It follows that the speckle pattern has certain similarities with
a random array of diffraction-limited images, i.e. Airy peaks in the case
of a circular aperture. Under certain conditions of oceanic storms where
organized swell is replaced by random waves, sailors have learned to
fear the sudden appearance of monster waves, much higher than average
(ADLARD COLES[19671). Similarly, and because speckled electromagnetic
fields are governed by the same Rayleigh statistics as gravity-induced
oscillations at liquid surfaces, there is a rare occurrence of exceptionally
bright speckles in astronomical images. These are not dangerous, fortu-
nately, and may in fact be exploited for diffraction-limited imaging purposes.
Q 2. Direct Interferometry
2.1. BASIC PRINCIPLES
The summed atmospheric term in eq. ( 3 ) tends towards the well defined
limit mentioned in sect. 1.4. In the Fourier space, this limit consists of the
diffraction limited MTF with an additional central peak. This term being
known, frow experiment or theory, a division gives the modulus of the
object function, which corresponds to the visibility curve obtained by
Michelson and Pease.
This general procedure can be applied regardless of aperture geometry
and gives spatial information on the object with diffraction-limited MTF
characteristics. It can be used not only with a Fizeau interferometer or
the full aperture of a large telescope, but also, in principle, with a coherent
array of large telescopes (Fig. X). It is interesting to note that early observers
seem to have attributed magic virtues to the Fizeau screen, without realizing
that information is actually gained when this aperture plate is removed
from the top of the telescope. However, some justification of Fizeau-
aperturing practice lies in the fact that the Fizeau screen simplifies the
image structure so that it can be processed by the eyes and brain of the
visual observer. Instead the considerable information content in speckled
images from a large aperture exceeds the data processing power of the
human eye-brain system.
11% 21 D I R E C T INTERFEROMETRY 63
A B C
Fig. X (additional): Principle of synthetic-aperture telescope : the giant (for example 100-
meter) parabolic mirror in A may be apertured with a multi-aperture mask as shown in B
while retaining the same limiting resolution. B is optically equivalent to the array of mirrors
shown in C if component mirrors are accurately adjusted in tilt and axial position to reproduce
the B geometry with 4 4 accuracy. This difficult tolerance can be relaxed to a few microns
when using filtered light. C is also equivalent to the coudi. arrangements in Figs. 12 and 13.
The above theory does not take into account the noise component
appearing when few photon-events are recorded in each image. However,
this happens to be a serious practical limitation to the accuracy of measure-
ments in most practical cases, since the requirements for short exposures,
narrow spectral bands and slow focal ratios (highf-numbers) do result in
photon-starved images. The problem is especially relevant when observing
faint objects with the new generation of image sensors working in the
photon-counting mode (BOKSENBERG [19721). Because amplification is
virtually noise-free in these receivers, the discrete nature of photon-events
becomes the dominant source of noise. The low level image is seen in the
form of few bright scintillations occurring on a dark background, and
distributed in time and space according to a compounded Poisson law.
When observing fringes at decreasing illumination levels, it eventually
becomes impossible to decide whether fringes are present or not. If the
fringe pattern were fixed, prolongated integration would solve the problem.
The fringes are however moving in our case, and higher order statistical
averages must be performed to extract the fringe signal. The analysis
procedure described in section 2.2 may still be used, but the validity of
eq. (3) under such conditions may be questioned. Classical results pertaining
to compounded Poisson distributions show that the equations are still
valid at low level if care is taken to remove the only distorsion arising in
the summed autocorrelation function, a narrow central peak appearing
at low level due to the correlation of each photon-event with itself. Assuming
low levels, pure photon noise, and many images, the signal-to-noise ratio
in the summed autocorrelation function is found to be
N,N~N~ (4)
The original Fizeau interferometer, used for the first time by Stephan at
Marseilles in 1873, involved only the aperture screen as special equipment
on the telescope. A strong eyepiece was used to observe the fringes. Tele-
scope optics are usually sufficiently good to meet the one-micron or so
tolerance on path equality which allows fringe observations in white light.
11, 31 I N T E R F E R O M E T E R DESIGNS A N D R E S U L T S 65
Fig. 5. Diagram of the 50-foot Michelson interferometer built by Hale and Pease at MtWilson.
66 HIGH-RESOLUTION TECHNIQUES CIL P3
fringe acquisition in white light was achieved by observing the star image
through a direct-view prism. Reference fringes of adjustable contrast were
provided by an auxiliary Fizeau interferometer for visual contrast
measurements, and corrections had to be made for a systematic effect
which decreased the apparent fringe contrast at long baseline settings.
In spite of appreciable operating difficulties, Michelson and Pease suc-
ceeded in resolving and measuring nine stars. These angular measurements
confirmed the enormous linear dimensions of objects such as Betelgeuse
in comparison to our sun, and gave the foundations for a scale of stellar
temperatures. A second interferometer having a 50-feet beam was later
constructed by Hale and Pease at Mt Wilson (Fig. 5). Similar in its principle
to the 20-foot system, the larger interferometer failed to produce many
additional results for reasons which are not completely clear but seem
related to operational difficulties greater than were expected. During a
recent visit to Mt Wilson, I found the interferometer well preserved in its
building. In spite of some very excellent design features, it appears that
the 50-foot cantilever beam may have suffered from poor stability about
its symmetry axis. The instrument could be revived at moderate cost, and
modern electronics for guiding and fringe sensing could certainly make
its operation much easier than in Peases days. It is not clear, however,
whether the modernized interferometer could compete with recent instru-
ments involving two independently mounted telescopes.
Photomultiplier tubes are hardly more sensitive than the human eye
for short exposure work at medium illumination levels, but they have faster
response and better photometric accuracy. Before the newer generation of
television sensors appeared, these advantages led several groups to develop
photoelectric fringe sensors replacing the human eye in Fizeau-type systems.
ELLIOTT and GLASS[1970] have used a picket-fence mask to generate a
photoelectric signal from Youngs fringes, but most workers in the field
have preferred beam-splitter arrangements to obtain a flat interference
field (Fig. 6), following Michelson who had already considered using
beam-splitters for variants of the basic Fizeau arrangement. He apparently
found no advantage in doing so for visual work, but the single-pixel nature
of photomultiplier tubes obviously makes it easier to work on flat inter-
ference fwlds than on Youngs fringes. This advantage of beam-splitter
arrangements no longer holds with the new generation of multi-pixel sensors.
A wide variety of configurations are possible with beam-splitter arrange-
INTERFEROMETER D E S I G N S A N D RESULTS 61
-14-
-1-1-
(C)
Fig. 6 . Types of single-pixel photoelectric fringe sensors: a, Kosters prism (K) arrangement
used by Currie with two polarizers (P), the waves made to interfere have opposite orientations
in the figure plane; b, system used by Cagnet, in which waves are identically oriented and polar-
izers unnecessary; c, standing wave phototube used by the author, in which the thin S11
film probes a standing-wave pattern produced by the two beams; d, picket-fence mask used
by Elliott and Glass in a Youngs fringe pattern.
beam splitter arrangements where the two incident beams are subjected
to equal numbers of reflexions before meeting the beam-splitter. The
phenomenon has been responsible for appreciable frustration in some
attempts at operating stellar interferometers. Once the effect is understood
however, fringes may be retrieved easily by using an additional mirror
or polarizing element. In the centrosymmetric case, a two-dimensional
display of the objects visibility function may be obtained directly if the
waves are tilted so as to produce narrow fringes : the fringe contrast varies
locally in proportion to the local visibility value. The shadow pattern on
the wave however destroys somewhat this display. Devices belonging to
the first class have been used on telescopes by KULAGIN[1970], CAGNET
[1973] and C. RODDIER[1971]. The fringe signal in the Cagnet system is
obtained as the difference between the outputs of two photomultipliers
located on each side of the beam splitter (Fig. 6b). The device was operated
in the Fizeau mode at the Haute Provence 193 cm reflector. C. RODDIER
[19711 worked with a different system which in fact uses the full aperture of
the telescope ; interference thus tends to vanish when using large apertures
with a single-pixel sensor.
CURRIE,KNAPPand LIEWER[1974] have developped an axisymetry
interferometer (Fig. 6a) which they mounted on the 100-inch and 200-inch
telescopes of the Hale observatories. They used a differential detection
scheme similar to that of Cagnet, but working in the photon-counting
mode with an on-line digital processor. With a pair of 2-cm apertures and
10 Angstroms spectral bandwidth, they were able to measure visibility curves
on Betelgeuse and alpha Hercules. In this system, polarizers are used to
avoid the above mentioned polarization incoherence problem associated
with axisymmetry devices. The system appears to work at rather low
counting rates, and this leads to expect considerable sensitivity for future
large-aperture, multi-pixel, devices working in the photon-counting mode.
Centrosymmetry systems have been operated in the laboratory by different
authors. BEAVERS [1963] has experimented with several forms of photo-
electric fringe sensing on the reduced-scale version of the Michelson 20-feet
interferometer which he has built.
Yet another type of single-pixel sensor was used by the author con-
currently with a Cagnet-type device in initial attempts with the two-tele-
scope interferometer at Meudon. As shown in Fig. 6c, the device is based
upon a special photomultiplier tube (built to the authors specifications
by the collaborators of Prof. Lallemand at Observatoire de Paris). The
S11 photocathode may be illuminated from both sides with plane waves. If
these are coherent, they interfere in the form of a standing-wave pattern
11, 31 INTERFEROMETER DESIGNS A N D RESULTS 69
which is probed by the semi-transparent and very thin S11 film. For flat
interference, the photocathode plane has to be parallel to the nodal planes
of the standing waves much as is the case with the beam-splitter films used
by CAGNET [1973] and CURRIE,KNAPPand LIEWER[1974]. If the phase is
made to oscillate with 180 amplitude by an auxiliary mirror drive, the
phototube signal may be submitted to lock-in detection, thus providing
with a single phototube the equivalence of the two-photomultiplier systems
mentioned above. Used concurrently with the Cagnet-type sensor, the
standing-wave sensor was generally preferred. However, both sensors were
discarded after a photon-counting television sensor was built and its
superiority was recognized (sect. 3.3).
F300
Irn
&
Fig. 7. Speckle interferometer used at the prime focus of the 200-inch telescope M, field-
finder mirror, with 0.1 mm hole; L, magnifying lens; G, concave field grating (Jobin-Yvon
holographic type) in plane of magnified image;U, spherical mirror; S , spectral mask; TV,
television camera ; R, auxiliary spectrum-viewing lens ;C, geometry-calibration grid. The optics
provides magnification and tunable filtering (color and bandwidth are separately adjustable).
Atmospheric dispersion is corrected by translating TV axially and rotating the complete system
about the telescope axis.
Fig. 9. Typical integrated power spectra of 200-inch images. obtained optically, showing resolution of six sellar disks and two binaries.
Object-reference pairs are indicated by a bar. The alteration in the case of p Canis Majoris is believed to result from aberrations resulting
from flexure of the 200-inch mirror for certain orientations. A mirror mask suppresses the bright central peak.
11, 9 31 INTERFEROMETER D E S I G N S A N D R E S U L T S 73
Fig. 10. The two-telescope interferometer at Nice. Narrow coudC beams from both telescopes
are received in the central building, where they recombine to produce Youngs fringes in the
synthetic image. The telescopes have special alt-alt mounts built from heavy-gage materials
for dimensional stability. Tracks are currently being designed for a variable baseline.
INTERFEROMETER D E S I G N S A N D RESULTS
Fig. 11. One of the two tclescopes operated as a Michelson interferometer. The coudii
beam exits through the hearing visible in front. Also visible are the massive secondary spider,
yoke and concrete support.
,=I
Fig. 12. Optical layout of the Meudon/Nice two-telescope interferometer: Tn, TS- Northland
South telescope; M- primary mirror cf= 850 mm); m- Cassegrain secondary cf= 7.5 mm);
F- coude flat; L-field lens; rm-roof mirror in pupil plane; D-dichroic mirror; TV1-guiding
camera; bl-bilens serving to separate the North and South guiding fields; S and P- slit and
direct view prism used for fringe acquisition; TV2- photon counting camera (tunable filter
or disperser not represented); Tr- tracks on which table moves (programming mechanism not
represented).
of the art has long been incompatible with such attempts, but the idea has
recently been revived by different groups at Berkeley (Lawrence Radiation
Laboratory), Itek Corp. and Hughes Labs.. MULLERand BUFFINGTON
119741 at Berkeley, in collaboration with DYSON[1974] used computer
simulations to study the performance of arrays having a few dozen aperture
elements with individually controllable phases. A parameter suitable for
deriving an error signal was found to be the intensity at one preselected
image pixel: the first aperture is phased to maximize intensity, than the
second, etc., and a few cycles of adjustments performed within the seeing
lifetime suffice to increase dramatically the image sharpness, even though
no attempt is made to suppress the shadow pattern on the aperture.
A different system, resembling more the original Babcock device has
been demonstrated in the laboratory by HARDY,FEINLrEB and WYANT
[1974] at Itek. In this seeing compensator, a wavefront-shearing inter-
ferometer serves to measure the phase distribution on the wave. A simple
analog computer derives correction signals, and these are supplied to a
deformable piezoelectric mirror. The spectacular image improvement
already obtained with this other simple system in its present stage of
development suggests a bright future for the seeing compensation approach.
A rather modest limiting magnitude, of the order of 10 to 13, has been
predicted for devices of that kind. However, for bright enough objects, not
only imaging cameras but also spectrographic equipment should benefit
from reconcentrating the energy which is scattered by the atmosphere.
can grow, i.e., it is possible to start with a relatively modest system and
progressively expand it once satisfactory operation is demonstrated.
However, the greatest potential interest of the array approach lies
perhaps in the coherent synthesis applications. Interferometric baselines
on the order of 100 meters can indeed be envisaged with an array since
system cost depends mostly on the total collecting area and not on the
telescope spacing. C6mpOnent telescopes for an array do not need to
involve unduely novel techniques, except perhaps for the coudC beam
arrangements, which should require as few reflexions as possible. Telescopes
should preferably be movable for a flexible baseline geometry, unless their
individual size is excessive. Optical path equality can be achieved by means
of tunable delay lines, as discussed by MILLER[1971]. Fringe detection in
the image may be achieved by speckle interferometry or using any of the
numerous possible beam-splitter arrangements. Figure 3 shows, under
conditions of laboratory simulation, the possible appearance of images
produced by a two-telescope arrays and a 6-telescope array, component
apertures being on the order of 1.5 meter. More complicated interference
structures would be observed within the speckles in the case of more than
6 telescopes, but the image analysis procedures used with a single telescope
remain valid. Fig. 13 shows the general layout of an array proposed by the
author. The array will have variable baseline geometries. Altitude-altitude
or spherical mounts are envisaged for the telescopes. These are 1.5 meter
Cassegrainians with a small interchangeable secondary and single-flat
Fig. 13. Proposed synthetic-aperture array of telescopes. Narrow coudB beams propagate
from each telescope into the central station, where they recombine. The telescope mounts
represented consist of ferro-cement spheres tracked on fluid pads. They are expected
to provide better dimensional stability than conventional coudi. mounts. Sphere surfaces
are precision ground. Conventional 1.5 meter (60-inch) telescope optics are mounted inside
the spheres.
11, 51 C O N S T R U C T I O N OF S Y N T H E T I C - A P E R T U R E A R R A Y 81
Fig. 14. Cross-section of concrete spherical mount studied for the array of 1.5 telescopes.
Reinforced concrete has good vibration damping characteristics, short-term stability and
low cost. The outer sphere surface is ground smooth and supported on rollers, water bearings,
or piezoelectric pads.
82 H I G H -RE SO L UT I O N T E CH NIQU ES [IL 0 6
avoid vibrations at the 0.1 micron scale. This implies particularly thick
spider arms and a very rigid yoke or sphere. In the prototype system, it
was found necessary to sacrifice partially the possibility of operating at low
elevation angles in order to meet these requirements. This is no great loss
for coherent work, since atmospheric turbulence and dispersion increase
sharply at large zenith angles. An attractive alternate design involves
spherical mounts. For low-cost production and excellent vibration-damping
characteristics, these mounts could consist of a shell made of reinforced
concrete or ferro-cement and supported on three rollers or fluid pads, as
shown in Fig. 14. This design appears to have attractive advantages, and
the unusual tracking problems appear to be quite solvable with the help
of modern minicomputers. Among the advantages are : 1. structural
simplicity; 2. no dome is needed since the sphere includes self-enclosed
laboratory space; 3. the spherical mount may be tracked either in the
alt-alt, equatorial, or even alt-az modes.
A project along these lines is currently being worked out at Meudon.
A 3.5 meter ferro-cement sphere has been constructed and will shortly
be equipped with a 1.5 meter mirror to develop the spherical tracking
technology. A second telescope will then be built, and institutions from
different countries will later be invited to contribute additional telescopes
for progressive array growth. The systems luminosity will surpass that
of Mt Palomars 200-inch instrument if the array grows to include eleven
1.5 meter telescopes or six 2-meter telescopes. Because it is still difficult
to predict the maximum baseline dimensions that will be usable, some
flat expanse of terrain at least one square kilometer should be selected,
in a region having low nebulosity and jet-stream activity in addition to
other desirable astronomical characteristics.
0 6. Intensity Interferometry
In order to overcome the problems which stellar interferometry had to
face in the years 1950, Hanbury Brown and Twiss proposed the novel
method which they called intensity interferometry. The principle may
be presented in the following elementary fashion, readers being referred
to Hanbury Browns articles for more details.
Neglecting atmospheric effects, which have no influence on this method,
the illumination produced on the ground by a stellar source is not uniform
if mapped during a period shorter than the coherence time z = l/f of the
beam ( f being the frequency bandwidth). The non-uniformity results from
interference of light emitted by different parts of the source. Indeed, the
11, 61 INTENSITY INTERFEROMETRY 83
electric field behaves as if the source were spatially coherent during such
a short interval. The phase is however not uniform on the source, so that
the emitter may be compared to a piece of diffusing glass illuminated by
a laser beam. If the diffusing glass is further assumed to fluctuate randomly
with lifetime z, an accurate model of the beams spatio-temporal structure
is obtained. The theory of speckle phenomena mentioned in section 1.3
applies to this case and shows that a distant screen, the terrestrial ground
in the present case, is illuminated with a speckle pattern fluctuating with
the time-constant z. The size of the speckles d is related to the angular size
M of the source by the usual relation d = A/M. This is equivalent to saying
that the whole spatio-temporal structure of an incoherent beam is speckled:
temporal as well as spatial speckles are present. The beam structure may
be described as a flow of random cells or speckles (also known as field
modes or coherence cells) propagating in space with velocity c while
deforming themselves. Fast detectors located on a transverse surface (the
terrestrial ground) see simultaneous intensity fluctuations if they are
spaced by less then the transverse dimension of speckles. They see un-
correlated fluctuations in the opposite case. The cell size may thus be
determined by comparing the signals from two detectors having a variable
spacing.
This is the technique used by HANBURY BROWN,DAVIS and ALLEN[1974]
at Narrabri observatory. A pair of 6.5-meter light collectors mobile on a
300-meters diameter circular track are each equipped with a fast photo-
multiplier and narrow-band filter. The two photoelectric signals are multi-
plied, and the result is integrated for several hours or days until adequate
signal-to-noise ratio is obtained. Because of the limited electrical band-
width in the photomultipliers and amplifier circuits, the optical frequency
bandwidth effectively utilized is extremely narrow, on the order of
Angstroms. This implies a comfortable tolerance on the dimensional
stability of mechanical structures, but also a very inefficient use of the
incident energy. For this last reason, the method has been applicable only
to the very brightest blue stars up to magnitude 2.5. It nevertheless permitted
remarkably accurate measurements on 32 stars with the unrivaled resolution
of l o p 3 arc-second.
Following the recent results with two telescopes operated as a Michelson
interferometer (sect. 3.4), J. Davis and myself have discussed the potential
applicability of both methods for work at very long baselines, on the order
of 2 kilometers. Whereas baselines as long as one kilometer may be usable
with direct interferometry, it seems that propagation of an electrical signal
is easier over long distances than the undisturbed coherent propagation of
84 HIGH-RESOLUTION TECHNIQUES [II, 8
0 7. Heterodyne Interferometry
Heterodyning techniques have been successfully employed at radio
wavelengths for the very-long-baseline observations involving antennas
located several thousand kilometers apart. The method consists in beating
light from the star with that from a local oscillator, on a suitable sensor.
Simultaneous work at two stations using the same oscillator frequency,
provides interference information. Heterodyne interferometers for work
at 10.6 microns are currently developped by TOWNES and his collaborators
[1974] at Berkeley, as well as by GAYand JOURNET [1973] at Observatoire
de Paris. At the University of Utrecht, VANDE STADT[1973] and Nieuwen-
huyzen build a system intended for work at 3.4 microns. Like intensity inter-
ferometry, and for the same reason, the heterodyne approach restricts
considerably the spectral band used. Its potential usefulness is generally
considered as marginal in the ultra-violet, visible and near infra-red regions
where photoemissive sensors are available. In the 2 to 10 micron infra-red
range, heterodyning becomes more efficient, but even there it is not yet
clear how it will compete with direct interferometry. Work on direct inter-
ferometry at 10 microns is carried out at Berkeley by D. Cudaback and
J. Franck.
0 8. Conclusions
Present trends in the fields concerned with high-resolution observation
at optical wavelengths indicate the likelihood of major improvements in
the coming decade. The directions which appear to hold most promise are:
1. synthetic-aperture arrays of telescopes. It is unclear yet how far it will
be possible to push this technique which is still in its infancy, but orders
111 REFERENCES 85
References
Symposia
Restoration of atmospherically degraded images, 1966, Woods Hole Summer Study, Nat.
Acad. of Sciences, Nat. Res. Council.
Synthetic aperture Optics, 1967, Woods Hole Summer Study, Nat. Acad. of Sciences, Nat.
Res. Council.
Synthetic Aperture Optics, 1970, ed. M. W. Stockton, Optical Sciences Center Report 58,
Tucson.
I11
BY
L. A. RISEBERG
GTE Laboratories Incorporated,
Waltham, Massachusetts, 02154, U.S.A.
and
M. J. WEBER*
Lawrence Livermore Laboratory,
Livermore, California, 94550, U.S.A.
* Work performed under the auspices of the U.S. Atomic Energy Commission.
CONTENTS
PAGE
Q 1. INTRODUCTION . . . . . . . . . . . . . . . . . . . 91
Q 2 . HISTORICAL DEVELOPMENTS . . . . . . . . . . . . 93
Q 3. RARE-EARTH ENERGY LEVELS . . . . . . . . . . . 98
Q 4. EXCITATION AND DECAY IN RARE-EARTH SYSTEMS 102
Q 5. RADIATIVE DECAY. . . . . . . . . . . . . . . . . 106
Q 6. MULTIPHONON RELAXATION . . . . . . . . . . . . 116
Q 7 . COOPERATIVE RELAXATION . . . . . . . . . . . . . 133
Q 8. SELECTED APPLICATIONS . . . . . . . . . . . . . . 150
9 9. CONCLUDING REMARKS. . . . . . . . . . . . . . . 155
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . 156
0 1. Introduction
8 2. Historical Developments
2.1. SPECTRA
38-
36.
34-
32.
30.
28.
26.
24.
22-
20.
18.
16.
14.
12.
10-
8.
,
6.
4.
-. -,*
t
2 -5
--
I-%
-
-9
-I
.....
r-i -*
0- -2- -
-r
- - 2 - 1 - 2 - R
2 3 4 8 7 6 5 , 3 2
F5(z H, 1% I, 6H5/z IFo 'S F6 Hi5+ I, Iiq H, FT,~
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb
Fig. 2.1. Observed energy levels of the trivalent rare earths. The width of the levels indicates
the approximate magnitude of the Stark splitting in LaCl, . The semicircles denote those levels
from which fluorescence has been observed in LaC1, (after DIEKE [1968]).
114 21 HISTORICAL DEVELOPMENTS 95
2.2. RELAXATION
At this time, with the lower energy-level structure of the rare earths
established, activity became directed toward some of the more subtle
phenomena, among which were the relaxation processes accompanying
the luminescence.* A particular stimulus for these studies from the early
sixties onward was the development and increasing importance of lasers
involving rare-earth ions in crystals, glasses, and liquids.
Over the years, the field of relaxation phenomena has received review
treatment in various forms. The monumental treatise of ELYASHEVICH
[1953] contained a summary and detailed description of work up to that
time. Review papers in the early 1960s by DIEKE [1961,1963] were in the
nature of updated reports. More recently, reviews by Moos [1970], GRANT
[1971] and KUSHIDA [1973a, b, c] contained discussions of one or more
aspects of this field, and a broad coverage was given in the,book by DI-
BARTOLO [19681.
Although only recently studied in detail, relaxation effects were mani-
fested very early in the study of rare-earth luminescence. The question
of the source of the radiative transitions was, as noted above, elucidated
by Van Vleck. This model was confirmed experimentally by BROER,GORTER
and H ~ ~ G S C H A G[1946].
E N The symmetry properties of the crystal field
were invoked in early spectroscopic studies to account for the selection
rules and polarization observed in these radiative transitions. With the
availability of good energy level assignments and wavefunctions, cal-
culations of oscillator strengths for magnetiedipole transitions could be
made. Theoretical calculations of electric-dipole strengths also became
possible with the techniques and phenomenological approach introduced
by JUDD [1962] and OFELT[1962]. Experimental verification of spectral
intensity calculations is now well established.
As for nonradiative relaxation phenomena, very early qualitative
observations indicated their importance in affecting the luminescence
properties. It was first noted in the hydrated salts that only the ions in the
center of the rare-earth series exhibited fluorescence. A variety of param-
eters appeared to be related to their stability, but the importance of the
H 2 0 vibrations in luminescence quenching was soon recognized. The
stability of these central ions was explained by HELLWEGE [1942] in terms
x
- 600
II
- 700
- 800
- 900
I - I000
- 1500
- 2000
- 3000
- 4000
- 5000
- -10000
5
- _5 i
C e Pr Nd Pm Srn Eu Gd Tb Dy Ho Er Tm Y b Lu
Fig. 3.1. Approximate extent of the two lowest configurations of the trivalent rare earths.
White - 4fN;black - 4fN-'5d (after DIEKE [1968]).
Fig. 2.1 provides a good guide to the location of the J states of the tri-
valent rare earths since the centers of gravity of the J manifolds exhibit
only small variations with host. The order and separation of the levels
within a J manifold, on the other hand, vary considerably from host to
host. The overall extent of the crystalline Stark splittings is small on the
energy scale of Fig. 2.1.
100 RELAXATION PHENOMENA [IK 5 3
x
700
800
900
1000
I- 1
i500
2000
3000
4000
5000
Nd 0000
0 X
LZO k! -
Ce Pm Sm Eu Gd Tb HO Er Tm Yb
Fig. 3.2. Approximate extent of the two lowest configurations of the divalent rare earths.
White-4fN;black-4fN-' 5d (after DIEKE [ 19681).
Fig. 3.3. Schematic diagram of the splitting of rare-earth energy levels due to the electrostatic,
spin-orbit, and crystal-field interactions.
111, 31 RARE-EARTH ENERGY LEVELS 101
The complex energy level structures of the rare earths were successfully
unraveled with the advent of the powerful tensor operator techniques and
crystal-field theory. This work has been thoroughly discussed, for example,
by WYBOURNE [1965] and by DIEKE [1968]. The free ion states, obtained
by diagonalizing the combined electrostatic and spin-orbit energy matrices,
are linear combinations of Russell-Saunders states of the form
IfN[ySL]J> = c C(ySr)lfNySLJ>.
YSL
(3.1)
where the factors l$ are parameters describing the strength of the crystal-
field components, 0:) are tensor operator components which transform
as corresponding spherical harmonics, and the summation is over the i
electronsofthe ion. The e a r e related to the products A:(rk)of the commonly
used parameters A: and radial integrals ($). The number and types of
terms appearing in the expansion in eq. (3.2) are derivable using group
theory and the point symmetry at the rare earth site. For the f shell, k is
limited to values 5 6.
In the above approach the strength of the crystal field is given by a
small number of l$ parameters. Attempts to calculate these parameters
using lattice sums and including covalency and overlap effects have achieved
only very limited success. Therefore the values have been determined
experimentally. To do this the energy matrix including the crystal-field
interaction is diagonalized using an estimated set of starting parameters.
102 RELAXATION PHENOMENA [IK 5 4
The resulting predicted energy levels are compared with the observed
levels and, by an iterative fitting procedure, the parameters are adjusted
to obtain a best overall fit to experiment. When the positions of many
levels have been measured, and the site symmetry is high so that only a
few terms appear in the expansion in eq. (3.2), root-mean-square deviations
'
of observed and calculated energies as small as 10 cm- have been obtained.
The crystal-field parameters have also been interpreted using a super-
position model pioneered by NEWMAN[1971]. The field is assumed to
arise from a sum of independent contributions from the other ions in the
crystal. All processes contributing to the crystal field are included in this
parameterization.
Once the crystal-field parameters for a given ion-host system have been
determined, a complete set of energy levels and eigenstates can be com-
puted. These states are labeled by a crystal quantum number ,u and are
of the form
These states can be used to calculate matrix elements for radiative and
nonradiative transitions between any rare-earth fN energy levels of interest.
Some of the levels from higher-lying electronic configurations (including
5d and 6s) have been studied for rare earths in crystals (see, for example,
LOH [1968]). In general, however, the locations of higher-lying con-
figurations are not well established because most of the levels are at energies
beyond the readily accessible optical region and above the fundamental
absorption edge of many materials. For the 5d states there is a strong
interaction between the outer 5d states and the static and dynamic crystal
fields. As a consequence, the locations of the 5d states vary by many
thousands of cm- in different materials and the optical transitions have
large linewidths.
The location of states having parity opposite to that of the 4fNcon-
figuration are of interest since, as discussed in section 5, electric-dipole
transitions between 4f states become allowed via admixing with these
states. Hence the locations of states such as 4fN-'5d and 4fN-'5g can
affect the rate of radiative decay from 4f states.
Absorption FI uorescence
V 2
V 1
Fig. 4.1. Schematic energy level diagram showing the radiative (straight line) and non-
radiative (wavy line) decay modes of a rare-earth ion following optical excitation.
a b b
where the summations are for transitions terminating on all final states b.
The radiative probability WR includes both purely electronic and phonon-
assisted transitions ; the nonradiative probability WNRincludes relaxation
111, 9 41 E X C I T A T I O N A N D D E C A Y 1N R A R E - E A R T H SYSTEMS 105
It is also equal to the ratio of the number of photons emitted from level a
to the number of ions excited by photons into level a.
Depending upon the specific rare-earth energy levels and host involved,
the relative probabilities for radiative or nonradiative decay between given
levels may range from comparable values to the two extremes where
W,"b-sz W a y or W$ > W2R. The radiative quantum efficiency of a level
may therefore approach zero or unity. Determination of the relative and
absolute magnitudes of WR and W N Rconstitutes the basic thrust in the
study of relaxation phenomena.
Determination of any tweof the quantities z, q, W R ,C W N Rin eqs.
1
(4.1X4.2) is sufficient to derive the other two quantities. Several
approaches, theoretical and experimental and combinations thereof, have
been applied to accomplish this. As will become evident later, meaningful
ab initio calculations of either radiative or nonradiative decay rates are
still beyond our present capabilities. Therefore all approaches involve
experiment and phenomenological treatment.
Experimentally, with the present existence of fast, intense pulsed light
sources and selective excitation techniques, measurements of excited-state
lifetimes present n6 serious problems. For a second quantity, measurements
can be made of the quantum efficiency q, but care is required to avoid the
introduction of systematic errors. As an alternative, measurements of the
line strengths in absorption have been combined with the Einstein ,443
relations to determine the corresponding emission probability W: . Knowl-
edge of the rare-earth concentration is required, however, to obtain the
absolute intensities of absorption spectra.
When transitions to several terminal J states are present, relative fluores-
cence intensities must be measured to determine the total 1 WR. This
becomes tedious if the wavelength and intensity ranges involved become
large. Different detectors may be required for different spectral ranges and
accurate calibration of the spectral sensitivity of the detection system is
required.
Success in calculating W$ using the Judd-Ofelt approach and phenom-
enological intensity parameters has prompted several studies wherein
1
calculations of W,"bare combined with measurements of z, to determine
106 RELAXATION P H E N O M E N A [m,0 5
the other quantities in eqs. (4.1) and (4.2). This approach can also be
tedious since it generally requires measurements of absorption spectra to
determine the best set of intensity parameters and computation of transition
matrix elements. In addition, whereas overall the r.m.s. errors between cal-
culated and measured line strength may be small ( 7 lo%), the strengths
of individual transitions may exhibit larger variations. As discussed in the
following section, the Judd-Ofelt approach works best for the lower-lying
states of the 4fNconfigurations.
9 5. Radiative Decay
5.1. THEORY
where
+(~,-~,)-1(6ul~~~d16~)(6,1P16b)1,(5.2)
where
P, = -e C~~(c;))~.
i
Any attempt at ab initio calculations of the matrix elements in eq. (5.2) are
plagued by numerous unknown quantities and difficulties. These include :
(1) the values for B,"(k odd) in V,, which, as discussed earlier in connection
with the even k terms in the static crystal field case, have not been calculated
satisfactorily ; (2) the energy denominators, since the energy levels E, of
the excited configurations are generally not known; (3) the intercon-
figurational radial integrals jR(4f)R(n'l')t dr, which require good radial
wavefunctions R for their evaluation, and finally ; (4) lengthy computations
involving consideration of many states.
A decade ago JUDD[1962] and OFELT[1962] independently showed
that calculations of ED probabilities could be made tractable by treating
the 4fNand the excited, opposite-parity configurations as degenerate with
a single average energy separation and by replacing Cil~s)(q5slC~') by the
tensor operator component U f i q of even order t. By so doing one can
invoke closure over the summation in eq. (5.2). The electric-dipole matrix
element for the pth component of polarization can then be expressed
simply as
($ulPpl$b) = y(t,4, p)(fNySLJJzIUt!kqlfNy'S'ZJ'J:), (5.3)
q , t even
t=2,4,6
The eigenstates are of the form in eq. (3.1) and the matrix elements of U @ )
can be derived (WYBOURNE [1965]) using tabulated doubly reduced matrix
elements of NIELSON and KOSTER[1964] and 3-j and 6-j symbols. The
expression for the electric-dipole line strength arising from one-phonon
vibronic transitions is identical in form to that given in eq. (5.4) (JUDD
[19621);hence this contribution is included in the intensity parameters 0,.
The oscillator strength f and the absorption cross section 0 for a transi-
tion of frequency v are related to S by
f ( a J ; bJ) = [87c2mv/3(2J+ l)he]S(aJ; bJ) (5.5)
and
s o(v)dv =
7ce2(n2+ 2)
9mcn
f
and thus can also be determined once the intensity parameters are known.
In eq. (5.6), n is the index of refraction of the host. Due to the ED for-
biddeness of f-f transitions, the oscillator strengths for transitions between
J states are small, of the order of
For relaxation one is interested in the spontaneous emission probability
given by
64x4v3x
A ( d ; bJ) = S(aJ; bJ). (5.7)
3(25 + i)hc3
x is the local field correction and for ED transitions is approximated by
n(n + 2)/9. The radiative lifetime and the fluorescence branching ratios
from a level a are defined by
and
n1, a 51 RADIATIVE D E C A Y 109
References
Rare-earth ion
Matrix elements Judd-Ofelt parameters
Pr3 +
Nd3+
Pm3+
Sm3+
Eu3+
Gd3+
Tb3+
Dy3+
H O+~
Er3+
Tm3+
Yb3+
A. WEBER[1968b].
B. KRUPKE[1971].
C. KRUPKE[1972].
D. AXE[1963].
E. WEBER[1967b].
F. KUBONIWA and HOSHINA [1972].
G . WEBER,MATSINGER, DONLAN and SURRATT[1972].
H. WEBER[1967a].
I. CHAMBERLAIN, EVERITTand ORTON[1968].
J. KRUPKE[1966].
K. DELSART and PELLETIER-ALLARD [1971].
L. WEBER,VARITIMOS and MATSINGER [1973].
M. WEBER[1968a1.
N. DETRIO[1971].
0. BECKER[1971].
P. KRUPKEand GRUBER[1965].
The radiative decay rates of rare-earth ions have generally been derived
either from the Einstein A-B relationship between the probabilities for
absorption and spontaneous emission or from the Judd-Ofelt approach.
In the first case the decay rate A,, for the transition from an excited state
a to the ground state 0 is determined from measurements of the associated
integrated absorption cross section 0 and the relation
(5.1 1)
where E,,, is the effective field at the emitting center, g b and go are the
degeneracies of the lower and upper electronic levels, and (rya) is the
electric-dipole matrix element between component states y and 6 of levels a
and b. Possible differences in the matrix elements for absorption and
emission should be considered when treating f ++ d transitions of the rare
earths.
The optical intensities of rare-earth spectra from liquids, glasses, and
crystals have been successfully rationalized using the Judd-Ofelt approach.
The intensity parameters for a given ion-host combination are derived from
a least-squares fit of calculated and observed intensities using as many
experimental values as available. Average deviations ranging from 5 to
20 % have been reported. Discrepancies for the intensities of transitions
TABLE
5.2
Measured and calculated oscillator strengths for the absorption spectrum of Er3+ in Y,O,
(UUPKE [1965])
and has obtained extremely good fits with rms deviations of only 5 % .
The resulting Judd-Ofelt parameters were used to calculate fluorescence
lifetimes and branching ratios and stimulated emission cross sections.
The literature of the Judd-Ofelt treatment applied to rare earths in
crystals is more dispersed. A summary of references to intensity studies in
crystals is given in Table 5.1. Unlike liquids and glasses, where the Stark
splittings are usually not resolved, transitions between individual Stark
levels can be observed in crystal spectra. Although most studies have con-
sidered only the total transition probabilities between Jmanifolds, in several
instances intensities of spectra between Stark levels have been analyzed
(see, for example, AXE [1963], KRUPKEand GRUBER[1965], BECKER
[19711, DELSART and PELLETIER-ALLARD [19713). A prerequisite for such
studies, as noted earlier, is the existence of satisfactory crystal-field eigen-
states.
Judd-Ofelt parameters are expected to vary throughout the lanthanide
series because of differences in the average energy separations of the
opposite parity configurations, the interconfigurational radial integrals,
and the relative contributions of vibronic intensities. In several studies
( K R U P [1966],
~ CARNALL, FIELDSand RAJNAK[1968], BUKIETYNSKA
and CHOPPIN [19703, PEACOCK [19733, WEBER, VARITIMOS and MATSINGER
[1973]) intensity parameters have been determined for a number of rare
earths in the same liquid or crystalline environment. Some systematic
variations in the R, values have been noted and used to predict parameters
for adjacent ions in the 4fNseries. Overall, however, the behavior based
114 RELAXATION PHENOMENA [Ill, 5 5
upon the results reported thus far are still somewhat unclear and additional
systematic studies are needed.
Radiative decay probabilities are calculated using Judd-Ofelt param-
eters and eqs. (5.7) and (5.8). In Table 5.3, calculated radiative lifetimes
of the 4F, state of Nd3+ in several different hosts are compared with
observed lifetimes. The latter values are those measured at low concentra-
-
tions and temperatures where a quantum efficiency of unity is expected.
In view of the 10% uncertainty in the Judd-Ofelt parameters, the agree-
ment is quite satisfactory. In general, agreement is good when, as in the
case of Nd3+, radiative transitions occur to several J manifolds, thus
providing a further averaging effect. Since the spontaneous emission
probability, eq. (5.7), varies as v3, radiative decay becomes more probable
for higher lying states. For 4f levels of rare earths, radiative decay rates
range from approximately 10 to lo6 sec-'.
TABLE
5.3
Comparison of the radiative lifetime calculated using reported Judd-Ofelt parameters and the
observed lifetime for the 4F, state of Nd3+ in different hosts
Lifetime (ps)
Host
calculated observed
and 4G,. In the last case, PEACOCK[1972b] found better results by ascrib-
ing separate fi2 values for each of the two hypersensitive transitions.
Because the energies of the final states differ significantly, a breakdown of
the average f d separation used in the Judd-Ofelt theory was suggested.
The occurrence of hypersensitivetransitions should therefore be noted when
employing the Judd-Ofelt method to calculate radiative decay probabilities.
For some ion-host combinations, charge transfer states occur at energies
which are lower than 5d states. This is observed, for example, for Eu3+
in the oxysulfides (STRUCKand FONGER C1971-j). These levels could be
responsible for the introduction of opposite parity states. In glasses,
REISFELD,BOEHM, LIEBLICH and BARNETT [1973] have found that increasing
covalency is correlated with lower charge transfer bands and larger Q2
parameters. Thus far no systematic study and treatment of possible effects
of charge transfer states on radiative decay of rare earths has been made.
The Judd-Ofelt theory is also applicable to the spectral intensities and
radiative decay of actinide series ions. Few intensity and decay rate measure-
ments have been reported for 5fNions. From absorption spectra, it is known
that the oscillator strengths of f-f transitions for these ions are generally
larger than for 4fNions, and therefore admixing of states from opposite
parity configurations may be greater.
In summary, the Judd-Ofelt method provides a useful description of
rare-earth intensities and can be used to predict radiative decay probabili-
ties. There are several limitations to its accuracy, however, due to assump-
tions inherent in the theory and to its application in practice. As examples
of the latter, when considering transitions between J manifolds, equal
ion populations in the Stark levels of the initial manifold are usually
assumed and J-state mixing is usually neglected. It should also be noted
that whereas the quality of Judd-Ofelt parameter fits are dependent
only upon relative absorption coefficients or fluorescence branching ratios,
spontaneous emission probabilities and absorption/emission cross sections
require accurate knowledge of the rare-earth concentration to be meaning-
ful.
The Judd-Ofelt treatment is expected to be most satisfactory for transi-
tions between low-lying levels of fNwherethe approximation of an average
energy denominator in eq. (5.2) is most valid. Studies of Pr3+ and Sm3+,
for example, have indicated that a single set of parameters is sometimes
not sufficient to account for all intensities and a second set of parameters
should be used for the higher-lying levels. This is indicative of a possible
breakdown of the closure approximation involved in arriving at eq. (5.3).
116 RELAXATION PHENOMENA
5 6. Multiphonon Relaxation
6.1. INTRODUCTION
where is the eigenfunction of the electronic state and the Ini) are
eigenfunctions of the phonon number operator and the ni are phonon
mode occupation numbers. Transitions involving no change in vibrational
quantum number n are purely electronic or zero-phonon lines. Transitions
In, 8 61 MULTIPHONON RELAXATION 117
6.2. THEORY
HCF = Vc,+ Ci
Qi . Vi V c F + . . . = V&+ 1 FQi++ C y ,
i i. j
jQiQj+ . . (6.2)
arises from two possible sources. The first is the orbit-lattice interaction.
For rare-earth ions, however, we assume that the interaction remains
harmonic and that the Vi , Vi. , . . . are essentially independent of tempera-
ture, at least up to temperatures of several hundred "C.The second and
dominant source arises from stimulated emission of phonons as the modes
become thermally populated. As we shall see later, a simple single-frequency
phonon model frequently yields good agreement with experimental observa-
tion.
Consider multiphonon relaxation across an energy gap AE to the next-
lowest level. The number of phonons p i of equal energy h o i , required to
conserve energy, and hence the order of the process, is determined by the
condition
pihwi = AE. (6.4)
The temperature dependence from eq. (6.3) is then
W ( T )= W,(n,+ 1)P[, (6.5)
where n, is the occupation number of the ith phonon mode and W , is the
spontaneous transition rate, that is, W(T)= W , at T = 0. Replacing n, by
its Bose-Einstein average
n, = [exp (ho,/kT)- 1]-', (6.6)
we have for the temperature-dependent multiphonon transition rate for a
single-frequency p-phonon process,
tensor operators with k odd cannot couple levels of the same configuration
because of parity. In addition, only those terms will contribute which
contain even-order operators satisfying the triangle rule
50
a
c
0
.-
0
c
TPK)
Fig. 6.2. Temperature dependence of the E(6Ft) to D(6Ft) multiphonon decay process in
LaC13:Dy3+(after RISEBERC and Moos [1967]).
The energy gap is taken to be that to the uppermost Stark level of the
terminal J multiplet. In this case it is not necessary to apply the detailed
model of eq. (6.8) because the emitting state consists of two Stark levels
separated by only a few cm-'. The theoretical fit clearly indicates that the
phonons correspond to the high-energy optical region, and that the process
is the lowest order consistent with energy conservation and the high fre-
quency cut-off of the phonon spectrum at 175 cm-'.
A more complex situation is provided by the F('F3) to E(5F4,5Sz)
relaxation of Ho3+ in LaF,. The theoretical fit to the experimental data
in Fig. 6.3 is given by eq. (6.7) with the following parameters:
P, = 6 A, =0 A, = 150~m-' A , = 5 o O ~ m - I
hoi = 3oOcm-' g1 = 0 92 = 1 93 = 5
W,= 2.2 x lo4sec-' W, = 0 w3= 0.
The effect of the Stark levels above the decaying level is accounted for by
a single non-decaying level at 150 cm-'. This reflects the reduced decay
rates for these levels because of the larger gaps. Similarly, the effect of the
126 RELAXATION PHENOMENA
Fig. 6.3. Temperature dependence of the F(F3) to E(F,, S 2 ) multiphonon decay process
in LaF, :Ho3+ (after RISEBERGand Moos [1968]).
-
for five different ions and fifteen different excited states can be fitted
approximately (within a factor of 2) by a simple exponential dependence
on the energy gap. (The two exceptions, the 'D1 and 'D2 states of Eu3+,
are subject to selection rules and the restrictions of eq. (6.1 l).) From knowl-
edge of the phonon spectrum of YAlO, and for the range of energy gaps
studied, a minimum of three and as many as seven or more phonons are
active in the relaxation. The close obedience to the exponential energy gap
law over a range of approximately six decades provides convincing testi-
mony to the phenomenological approach and the averaging out of the
detailed features of the electronic states and phonon modes in these high-
order processes.
The exponential dependence expressed by eq. (6.12) and illustrated in
Fig. 6.4, while applicable to processes involving many phonons, must
eventually break down for small energy gaps where relaxation by one- or
two-phonon processes is possible. In this regime the statistical averaging
that occurs for higher-order processes is no longer prominent; relaxation
rates are strongly dependent on the phonon density of states and on the
ion-phonon coupling. This is evident from the optical linewidths observed
for transitions to individual Stark levels. These widths, when determined by
lifetime broadening due to rapid one- or two-phonon processes, correspond
to relaxation rates which are much larger than would be obtained by
extrapolation of the energy gap dependence (WEBER[1973b1).
128 R E L A X A T l O N PHENOMENA
, I I
I
Neodymium
0 Europium
0 Holmium
I I I I I I I
1000 2000 3000 4000 5009
Energy gap to next-lower level (cm-1)
Fig. 6.4. Dependence of the rate of multiphonon emission on energy gap to the next lower level
for excited states of rare-earth ions in YAlO, at 77 K (after WEBER [1973b]).
For large energy gaps the multiphonon decay rates become smaller and
approach the regime where Z A i j > W. If in this limit W is obtained by
subtracting A from C1, then uncertainties in A can greatly affect the
value and accuracy of W. The multiphonon rates for YA103 were obtained
using A's found from Judd-Ofelt calculations. The uncertainty in their
values produces the large uncertainties in the resulting W values plotted
in Fig. 6.4. Use of relative quantum efficiency measurements is a more
desirable experimental approach in this case.
The selection rule in eq. (6.11) has generally not been found to be an
important factor for multiphonon relaxation when only one or two terms
of the ion-phonon interaction are forbidden. This, however, may be
because the effect is not expected to be significantly larger than the normally
observed variations from the exponential dependence. In cases where
111, 8 61 MULTIPHONON RELAXATION 129
I I I f I I I
Fig. 6.5. Spontaneous multiphonon emission rates from excited states of trivalent rare earths
as a function of energy gap to the next lower level (after WEBER[1973c]).
from differences in the phonon spectra (via the halogen masses) and thus
the order required to conserve energy.
Once sufficient data has been obtained for a given host to define the
exponential energy gap dependence and parameters in eq. (6.12), the
results can be used to predict the rate of spontaneous multiphonon decay
within a factor of -
from any 4f rare-earth level of interest. The results are generally good to
2-3. The only cautionary considerations are the
possible selection rule restrictions and the inapplicability of extrapolations
to small energy gaps.
Multiphonon relaxation in crystals containing high-frequency molecular
group vibrations in addition to the lattice phonon continuum requires
special consideration. An early treatment of CaWO, (RISEBERG [19683)
indicated, at least qualitatively, a convergence of the multiphonon rate
with energy gap. REEDand Moos [1973a, b], from a more detailed study
111, 8 61 MULTIPHONON R E L A X A T I O N 131
TABLE 6.1
Phenomenological parameters for multiphonon relaxation of rare-earth ions in crystals
of relaxation in YVO,, YAsO,, and YP04, have shown that the rates can
be critically dependent on the degree of resonance between the gap energy
and the sums of peaks in the vibrational spectrum.
Experimental results to date have been limited to crystalline hosts,
however, similar dependences of multiphonon emission rates on tempera-
ture, energy gap, and host may be expected for rare earths in liquids and
glasses. For liquids, in virtually all cases, there are high-frequency molec-
ular vibrations to consider. Relaxation in liquids has been treated by the
study of fluorescence lifetimes and efficiencies for a given transition in
different solutions (differing in some cases by a single substitution). The
vibrational spectrum is thus varied. Results have confirmed, in an ap-
proximate sense, the convergence of the multiphonon rates with the order
of the process (KROPP and WINDSOR[1965]). Although multiphonon rates
have not been extracted explicitly, it is probable that behavior similar to
that observed in molecular crystals will be observed.
Glasses also contain vibrational groups with frequencies higher than
most phonon frequencies in crystals. Although the degree of coupling of
these vibrations to rare-earth impurities is not well established, multi-
phonon decay bridging large energy gaps should be possible. This is
reflected, qualitatively, by the smaller number of fluorescing levels observed
for rare earths in glasses compared to crystals. Studies of radiative and
nonradiative relaxation of rare-earth doped glasses by REISFELD[19731
and co-workers have already shown some evidence of a systematic behavior
of multiphonon decay rates similar to that found in crystals.
132 RELAXATION PHENOMENA CIII, 06
Overall, the phenomenon of multiphonon relaxation is reasonably well
understood with a workable theory having predictive value and well
supported by extensive experimental work.
6.3.4. 5d + 4f relaxation
When excitation occurs via 5d states, 5d 3 4f relaxation is of interest.
As shown in Fig. 3.1, the 5d states of trivalent rare earths are located at
energies 5 50000 cm-'. Optical transitions to these states are therefore
frequently masked by absorption of the host material. The 5d states of
divalent rare earths (Fig. 3.2), in comparison, are located at lower energies.
For most divalent and trivalent lanthanide ions, levels of the 4fN-' 5d con-
figuration overlap those of the 4fNconfiguration. In such cases ions excited
into 5d levels rapidly decay nonradiatively to nearby 4f levels. This is
evident from the absence of 5d emission and the relative intensity and
quantum efficiency of 5d bands in the excitation spectra of 4f fluorescence.
Relaxation of 5d states has been investigated for trivalent rare earths in
Y3AI5Ol2(WEBER[1973a]). In this host the crystal-field splitting of the
5d states is large, resulting in low-lying 5d levels. For ions such as Nd3+
and Tb3+,the broad 5d bands overlap 4f levels; no 5d 4f emission was
--f
observed and rapid nonradiative relaxation was inferred. For Ce3+ and
Pr3+ , however, there are large energy separations from the lowest 5d level
to levels of 4f, approximately 16500 and 10000 cm- ',respectively. 5d + 4f
fluorescence was observed from these ions and the lifetime and intensity
were measured as a function of temperature. The results are shown in
Fig. 6.6. At elevated temperatures, the Ce3+ and Pr3+ 5d lifetimes exhibit
a rapid decrease. Since the d -+ f energy gap for Ce3 is larger, more phonons
+
Fig. 6.6. Temperature dependence of the 5d fluorescence lifetimes and intensities for Ce3+
and Pr3+ in Y,AISO,, (after WEBER [1973a]).
0 7. Cooperative Relaxation
7. I . ION-ION ENERGY TRANSFER
7. I.I. Introduction
Paramagnetic ions in solids can be treated as isolated ions only when
they are well separated. As the concentration is increased or if non-random
distribution occurs, the ion spacing may become sufficiently small that the
ions interact. The coupling of adjacent paramagnetic ions can arise via
134 R E L A X A T I O N PHENOMENA [In, 7
where c;:;
is a numerical factor dependent on the orientation of the
coordinate hxes and Or) is a multipole operator
136 RELAXATION PHENOMENA Cni, 0 7
(7.3)
The leading terms in the expansion in eq. (7.2) are the electric dipole-dipole
(EDD), dipole-quadrupole (EDQ), and quadrupole-quadrupole (EQQ)
interactions. These have radial dependences of R - 3 , R - 4 , and R-, respec-
tively. Higher-order terms are generally of negligible importance.
The matrix elements of HEsare subject to selection rules AS = 0 and
IALI, lAJl 5 k. As was discussed earlier for radiative decay, these rules
are substantially relaxed for transitions between states of the 4fN con-
figuration because of the large spin-orbit admixing of SL states and possible
J state mixing. Whereas EQ transitions are parity-allowed, ED transitions
require an admixing of opposite-parity states into 4fNand are correspond-
ingly weak for rare earths. Because of this forbiddenness, EQQ transitions
may be more probable than EDD transitions (AXEand WELLER[1964]).
However, due to the more abrupt radial dependence, R- versus R - 3 ,
EDD interactions should dominate at large separations. The intensity of
EDD transitions can be treated as in the Judd-Ofelt theory for radiative
transitions.
A second ion-ion coupling mechanism is the magnetic dipole-dipole
interaction
where p i = li + 2si and ( l i ,si), ( I j , s j ) are the orbital and spin operators for
the ith andjth electrons of ions A and B, respectively. The selection rules
AS, AL, AJ = 0, -t 1 for transitions between 4fNstates are again relaxed by
SLJ state admixing. The MDD interaction has the same long-range R T 3
radial dependence as the EDD interaction.
Finally we consider the exchange interaction given by
A B
A
(1 I
(b)
- B
(1)
Fig. 7.1. Schematic diagrams of cooperative relaxation processes: (a) transfer involving
excited state to ground state de-excitation of one ion and excitation of a neighboring ion
to the same excited state; (b) transfer involving dissimilar pairs of levels of like or unlike ions
(dashed lines for like ions).
138 RELAXATION PHENOMENA Cm7
The manner in which the interactions between two like or two unlike
ions lead to energy transfer and relaxation is illustrated schematically in
Fig. 7.1. The simplest process is shown in Fig. 7.la where an ion A in an
excited state (2) decays to its ground state (1) with corresponding excita-
tion of a neighboring ion B from its ground state (1) to (2). If A and B are
identical ions, this process involves resonant transfer of energy from ion
to ion. While this does not lead to net relaxation, it does give rise to spatial
energy migration. In concentrated rare-earth materials resonant transfer
can be fast. Discussion of energy migration is postponed to section 7.1.3;
below we consider only a single ion pair.
Fig. 7.lb depicts a process whereby an excited ion A decays from state
(2) to (2) while ion B, initially in its ground state, is excited to (1). Here
ions A and B may or may not be identical, as indicated by the dashed lines
for possible matching levels. In the absence of any additional interactions,
energy conservation imposes the constraint of resonance ; that is,
(E, - E,,) = (El,-El). In liquids and solids, however, due to the presence
of ion-lattice coupling, any energy mismatch may be taken up by the
emission or absorption of one or more phonons. The interactions in eqs.
(7.2), (7.4), and (7.5) can all be treated to include ion-lattice coupling
(ORBACH [19671).
The transition probability for energy transfer from ion A to B is given by
energy transfer in dilute materials and little attention was paid to the role
of exchange until recently.
DEXTER [19531also treated the concentration dependence of the lumines-
cence yield. For ion A, this is defined by
where WAB is given by eq. (7.7) and 7: is the radiative lifetime of state A(2).
As the ion concentration is changed, the ion-ion separation and the prob-
ability for relaxation both change correspondingly. For a given interaction
and functional dependence on R , the average yield yIA for a single-pair model
is obtained by taking an integral over all space of qA times the probability
that the nearest ion B is at a distance R . This yields
OD e-Yt
ijt)= l - y L -dt, (7.9)
1 + tnI3
with y = ynpB,the reduced density, and t = 4zR3/3yn.p B is the density of
B ions and yn is defined by 7:WiB = 3yn/47cR3for n = 6 (EDD), n = 8(EDQ),
and n = 10 (EQQ).
A great body of experimental work over the past decade was devoted
to applying the Dexter model to determine which multipolar term of HEs
was dominant for energy transfer and luminescence quenching in a partic-
ular dopant-crystal system. Van Uitert and co-workers, among others,
carried out an extensive series of studies in a wide variety of materials (see
GRANT C1971) for a bibliography of this work).
Although Dexter's treatment of luminescence yield describes the observed
concentration dependences in a reasonable way, the assignments of the
radial dependences and dominant interactions in a given case are often
ambiguous. In some instances, with the available data, there was not suffi-
cient differences in the theoretical dependences of yI(n) with n. There are
several features of the overall approach which made analysis difficult.
First, at concentrations large enough to show substantial ion-pair decay,
resonant transfer (as in Fig. 7.la) can be exceedingly fast among ions A,
particularly since the degree of resonance for ion-pair decay (as in Fig. 7.1b)
can be expected to be generally smaller. Therefore, those A ions surrounded
by a greater number of B ions than the average will dominate the decay,
and the short-range (but stronger) interactions will be enhanced (i.e. EDQ
and EQQ). At low concentrations, where the average separation is larger,
the longer-range interactions, such as EDD, will be dominant. Thus, in a
given material, different interactions dominate for different concentration
140 RELAXATION PHENOMENA [Ill, 9: I
ranges. In the case where A and B represent different species, for sufficiently
high concentrations of A (greater than a few %) resonant transfer within
the A system is sufficiently fast that the excitation may be considered
equally shared among all of the ions A. A rate equation approach for
A 4B transfer is then justified (GRANT [1971]).
An assumption of the Dexter approach was that the rate was dominated
by transfer to the nearest B ions. An extension to the entire environment
including the dynamics of the transfer was formulated by INOKUTIand
HIRAYAMA [1965] (hereafter referred to as IH). The difficulty in treating
the dynamics of these processes is that for an excited ion A decaying to a
random distribution of quenching ions B, the environment of the A ions
varies and the ion-pair relaxation rate depends on the particular environ-
ment of any excited ion A. Thus the decay of the system of excited ions A
is nonexponential. In the IH approach, an ion A is considered to be sur-
rounded by a set of quenching ions Bk at distances Rk. The energy transfer
rate from an A ion to the kth B ion is W A B r (Rk).
The time dependence of
the A excited-state population is then
(7.10)
[ 3 C t
0 A
where C is the acceptor (B) concentration and n = 6,8,10 for EDD, EDQ,
(7.11)
c, = m (7.12)
where R , is the separation at which the energy transfer rate for an isolated
A-B pair is equal to 2, .
The IH theory also treats the direct exchange interaction where @(t)has
the form
(7.13)
111, 0 71 COOPERATIVE RELAXATION 141
= 62 c m!(m+
rn (-z)"I
1)4 *
(7.14)
Thus energy transfer can be studied via the observation of the time depen-
dence of thedecay at a particular concentration. #(t) may also be used in
the analysis of other quantities. For example, the quantum yield ijA is
simply given by
(7.15)
where D is the diffusion constant and v(r - r,) is the probability for energy
transfer from an excited donor to the nth acceptor at r,. YOKOTAand
TANIMOTO [1967] obtained a general solution for the donor fluorescence
decay function including both diffusion within the donor system and
donor-acceptor energy transfer via EDD coupling. Their expression at
earlier times in the decay reduces to eq. (7.1 1) for n = 6 . The decay at
long times after excitation reduces to a simple exponential decay.
The asymptotic solution at long times is described by a characteristic
lifetime
_1 --- 1 +4nN,D, (7.18)
T 70
Tb - Nd
I= 8
0.8 -
0 0 exptl
Calc
.
CI
I
2
0
0 -c
0
I 0.6 t
I
y1
' c
3
F
>
0.4 a
20
0.2
0.01 0.1 1 10
I I
cO
I , I .
('0.95
I
E'O.Ol Yb0.04'23'
I I , I I
S=6
0
. I
-
Eu 5D0: T,, = 880 ps
s=10
LNT
Fig. 7.3. Luminescence decay of the 'Do level of Eu3+ in Y , O , : l % Eu3+, 4 % Yb3+ at
liquid nitrogen temperature. The theoretical curves show the IH calculations for EDD, EDQ
and EQQ interactions (after YAMADA, SHIONOYA and KUSHIDA[1972]).
>
t
v)
z
W
c
z
Y0 10'
z
Lu
P
W
z
s
4 \
'\
\
\
\
\
\
\
10-2
\
I 1 I
0 0.2 0.4 0.6 0.8 1.0
TIME (mi)
Fig. 7.4. Luminescence decay of the Yb3+ 'F, level in YF,: 1 % Yb3+, 10% H o 3 + .The solid
curve is experimental. Theoretical curves for exchange (dashed curve) and EDD (short and
long dashed curve) interactions using the IH theory are included (after WATTSand RICHTER
[19723).
146 RELAXATION PHENOMENA [In, 6 7
(7.21)
1o5 I I 9
Hd5S2) +Sm
1o4 Ed4S312)+Yb
a
2
II
a
7
-8
v)
Sm(4G512) +ELI
103
2
a
w Tm(G4) +Vb
U
v)
z
2I-
>
Q
E
2
102
w
10
I I 1 I \
Fig. 7.5. Energy gap dependence of multiphonon assisted energy transfer in Y,O, (after
YAMADA, SHIONOYA and KUSHXDA [1972]).
148 RELAXATION P H E N O M E N A
A
(1) 1:t
::: :
51 82
- I
(2)
3. (1)
1 (2
(1) (1)
B
A2
(b)
1)I- A lc)
B
hu
Fig. 7.6. Schematic diagrams of various cooperative relaxation phenomena : (a) multi-ion
relaxation, (b) cooperative excitation, (c) Raman luminescence.
111, 71 COOPERATIVE R E L A X A T I O N 149
0 8. Selected Applications
A significant impetus for the study of rare-earth relaxation during the
111, 81 SELECTED A P P L I C A T I O N S 151
Er3+, T m 2 + * 3 +Yb3+,
, U3+s4+)have been lased in crystals, glasses and
liquids. The single class of lasers enjoying the widest practical utilization
today is based on the 4F, + 'Iy emission of the Nd3+ ion at 1.06 pm.
In applications involving the luminescence of rare earths, the relaxation
processes discussed in this chapter are implicitly involved in a critical way.
After excitation, whether by optical, X-ray, cathode-ray, or other means,
it is the rates and relative importance of various relaxation processes that
determine what the emission spectrum will be, its intensity and efficiency,
and the lifetimes of the initial and final states. Below we discuss two rep-
resentative examples in which relaxation phenomena have been considered
explicitly in the design of luminescent devices.
A3 2 / :2
4s3/2
s2 - -2FY2 s2 -
.97um
Sl - Yb3+
2
F?/2
Yb3+ Er3+
Fig. 8.1. Schematic diagram of infrared upconversion in the Yb3+-Er3+ system. Sequential
relaxation of two Yb ions results in excitation ofvisible luminescence from the Er3+4S, level.
Since the total populations A and S are constant, one has the additional
condition n
CAi
i=1
= A; S,+S, = s. (8.3)
111, 0 81 SELECTED A P P L I C A T I O N S 153
10.1
Fig. 8.2. Power efficiency E (normalized to radiative quantum efficiency q ) for the Ho3+
green emission in YF, :Yb, Ho as a function of Yb excitation rate X (after WATTS[19703).
154 RELAXATION PHENOMENA [III, 5 8
8.2. LASERS
3 9. Concluding Remarks
In t h s chapter, we have attempted to survey the relaxation phenomena
affecting the luminescent properties of rare-earth ions in solids. Due to
the rapid growth of activity in this field in recent years, the literature is
voluminous and therefore no attempt was made to cover it completely.
Instead, examples were chosen for their appropriateness in illustrating the
discussion. The coverage has therefore been selective in detail, yet broad
in scope. The intention has been to acquaint the non-specialist with the
subject and to provide the applications-oriented individual with a familiar-
ity with the basic phenomena having implications on his project.
In the case of radiative relaxation, the Judd-Ofelt theory provides a useful
rationale for parametrizing the phenomena. Through a combination of
experimental measurements and theoretical calculations, the radiative
156 RELAXATION PHENOMENA [I11
References
ASAWA,C. K. and M. ROBINSON, 1966, Phys. Rev. 141, 251.
AUZEL,F., 1966, C. R. Acad. Sci. (Paris) 262, 1016.
AUZEL,F., 1973, Proc. IEEE 61, 758.
AXE,J. D., 1963, J. Chem. Phys. 39, 1154.
AXE,J. D. and P. F. WELLER,1964, J. Chem. Phys. 40,3066.
BARASCH, G . E. and G. H. DIEKE,1965, J. Chem. Phys. 43, 988.
BECKER, P. J., 1971, Phys. Status Solidi B 43, 988.
BECQUEREL, J., 1907, Radium 4, 328.
BETHE,H., 1930, Z. Physik 60,218.
BIRGENEAU, R. J., 1968, Appl. Phys. Letters 13, 193.
BLEANEY, B. and K. W. H. STEVENS, 1953, Rept. Prog. Phys. 16, 108.
BOTDEN, T. P. J., 1952, Philips Res. Repts. 7, 197.
BROER,L. J. F., C. J. GORTER and J. HOOGSCHAGEN, 1945, Physica 11, 231.
BOURCET, J. C. and F. K. FONG,1974, J. Chem. Phys. 60, 34.
BUKIETYNSKA, K. and G. R. CHOPPIN,1970, J. Chem. Phys. 52 (6), 2875.
CARNALL, W. T., P. R. FIELDS and K. RAINAK,1968, J. Chem. Phys. 49, 4412, 4424, 4443,
4447, 4450.
CARNALL, W. T., P. R. FIELDS and B. G. WYBOURNE, 1965, J. Chem. Phys. 42,3797.
CHAMBERLAIN, J. R., A. C. EVERITT and J. W. ORTON,1968, J. Phys. C 1, 157.
CHAMBERLAIN, J. R., D. H. PAXMAN and J. L. PAGE,1966, Proc. Phys. SOC.(London) 89,143.
DE GENNES,P. G., 1958, J. Phys. Chem. Solids 7, 345.
DELSART, C. and N. PELLETIER-ALLARD, 1971, J. Phys. (Pans) 32, 507.
DELSART, C. and N. PELLETIER-ALLARD, 1973, J. Phys. C 6, 1277.
DETRIO,J. A., 1971, Phys. Rev. B 4, 1422.
DEXTER,D. L., 1953, J. Chem. Phys. 21, 836.
DEXTER,D. L., 1962, Phys. Rev. 126, 1962.
DEXTER,D. L. and J. H. SCHULMAN, 1954, J. Chem. Phys. 22, 1063.
DIBARTOLO, B., 1968, Optical Interactions in Solids (Wiley, New York).
DIEKE, G. H., 1961, Spectroscopic Observations on Maser Materials, in: Advances in
Quantum Electronics, ed. J. R. Singer (Columbia University Press, New York) pp. 164186.
HI1 REFERENCES 157
DIEKE, G. H., 1963, Energy Levels of and Energy Transfer in Rare Earth Salts, in:
Paramagnetic Resonance, ed. W. Low (Academic Press, New York) pp. 237-252.
DIEKE,G. H., 1968, Spectra and Energy Levels of Rare Earth Ions in Crystals (John Wiley
and Sons, Inc., New York).
ELYASHEVICH, M. A., 1953, Spectra of the Rare Earths (State Publishing House of Technical-
Theoretical Literature, Moscow).
~RDCSS,P., 1966, J. Phys. Chem. Solids 27, 1705.
FEOFILOV, P. P. and V. V. OVSYANKIN, 1967, Appl. Opt. 6, 1828.
FEOFILOV, P. P. and A. K. TROTIMOV, 1969, Opt. and Spect. 27, 291.
FONG, F. K., S. L. NABERHUIS and M. M. MILLER,1972, J. Chem. Phys. 56,4020.
FCIRSTER,T., 1948, Ann. Physik 2, 55.
F~RSTER,T., 1949, Z. Naturforsch. 49, 321.
FOWLER, W. B. and D. L. DEXTER,1962, Phys. Rev. 128, 2154.
FOWLER, W. B. and D. L. DEXTER,1965, J. Chem. Phys. 43, 1768.
FREED,K. F. and J. JORTNER, 1970, J. Chem. Phys. 52, 6272.
GANDRUD, W. B. and H. W. Moos, 1968, J. Chem. Phys. 49, 2170.
GASHUROV, G. and 0. J. SOVERS,1969, J. Chem. Phys. 50, 429.
GERMAN, K. R. and A. KIEL, 1973, Phys. Rev. 88, 1846.
GRANT,W. J. C., 1971, Phys. Rev. B5,648.
HELLWEGE, K. H., 1942, Ann. d. Phys. 40,529.
HELLWEGE,K. H., 1947, Naturwiss. 34,212.
HEWS, R. A. and J. F. SARVER, 1969, Phys. Rev. 182, 427.
HOSHINA,T. and S. KUBONIWA, 1971, J. Phys. SOC.Japan 31, 828.
INOKUTI, M. and F. HIRAYAMA, 1965, J. Chem. Phys. 43, 1978.
JENSSEN, H. P., 1971, Phonon Assisted Laser Transitions and Energy Transfer in Rare Earth
Crystals, Technical Report No. 16 (M.I.T. Crystal Physics Laboratory, Cambridge).
JENSSEN, H. P., D. CASTLEBERRY,.D. GABBE and A. LINZ,1973, IEEE J. Quant. Elec. QE9,665.
JOHNSON, L. F., J. E. GEUSICand L. G. VAN UITERT, 1966, Appl. Phys. Letters 8,200.
JOHNSON, L. F., H. J. GUGGENHEIM, T. C. RICHand F. W. OSTERMAYER, 1972, J. Appl. Phys.
43, 1125.
JOHNSON, L. F. and H. J. GUGGENHEIM, 1973, Appl. Phys. Letters 23, 96.
JI$RGENSEN, C. K. and B. R. JUDD,1964, Mol. Phys. 8, 281.
JUDD,B. R., 1962, Phys. Rev. 127, 750.
JUDD,B. R., 1966, J. Chem. Phys. 44, 839.
KIEL, A., 1964, Multi-Phonon Spontaneous Emission in Paramagnetic Crystals, in :
Quantum Electronics, eds. P. Grivet and N. Bloembergen (Columbia University Press,
New York) pp. 765-777.
KINGSLEY, J. D., 1970, J. Appl. Phys. 41, 175.
KISLIUK,P. and C. A. MOORE,1967, Phys. Rev. 160, 307.
KRAMERS, H. A,, 1930, Proc. Acad. Sci. Amsterdam 32, 1176.
KRASUTSKY, N. and H. W. Moos, 1973, Phys. Rev. BS, 1010.
KROPP,J. L. and M. W. WINDSOR, 1965, J. Chem. Phys. 42, 1599.
KRUPKE,W. F., 1966, Phys. Rev. 145, 325.
KRUPKE,W. F., 1971, IEEE J. Quant. Elec. QE-7, 153.
KRUPKE,W. F., 1972, IEEE J. Quant. Elec. QE-8, 725.
KRUPKE,W. F., 1974, IEEE J. Quant. Elec. QE-10, 450.
KRUPKE,W. F. and J. B. GRUBER,1965, Phys. Rev. 139, 2008.
KUBONIWA, S. and T. HOGHINA, 1972, J. Phys. SOC. Japan 32, 1059.
KUSHIDA, T., 1973a, J. Phys. SOC.Japan 34, 1318.
KUSHIDA, T., 1973b, J. Phys. SOC.Japan 34, 1327.
KUSHIDA, T., 1973c, J. Phys. SOC.Japan 34, 1334.
LAUER,H. V. and F. K. FONG,1974, J. Chem. Phys. 60,274.
LEASK,M. J. M., 1968, J. Appl. Phys. 39, 908.
158 RELAXATION PHENOMENA [111
BY
MICHEL A. DUGUAY*
Bell Laboratories,
Murray Hill,New Jersey 07974, U S A .
* Present address: Sandia Laboratories, Albuquerque, New Mexico 871 15, U.S.A.
CONTENTS
PAGE
0 1. INTRODUCTION . . . . . . . . . . . . . . . . . . . 163
5 ?,. THE ULTRAFAST OPTICAL KERR SHUTTER . . . . . 165
5 3. ULTRAHIGH SPEED PHOTOGRAPHY . . . . . . . . . 177
0 4. SAMPLING OPTICAL SIGNALS . . . . . . . . . . . . 183
0 5 . CONCLUDING REMARKS. . . . . . . . . . . . . . . 191
ACKNOWLEDGEMENTS. . . . . . . . . . . . . . . . . . 192
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . 192
8 1. Introduction
One of the new frontiers to be opened by the laser has been the area of
picosecond pulses. This began in 1966 when DEMARIA,HEYNAUand
STETSER [19663 reported the emission of ultrashort pulses by a neodymium :
glass laser Q-switched by a saturable absorber. Soon thereafter ARMSTRONG
[1967] measured these pulses and found them to have a duration on the
order of 6 picoseconds, a record at the time in terms of ultrashort optical
or electrical pulses.
The discovery of picosecond laser pulses stimulated the invention of new
techniques for measuring ultrashort optical pulses, notably the two-photon
fluorescence technique of GIORDMAINE, RENTZEPIS, SHAPIRO and WECHT
[1967]. Picosecond pulses were also applied to scores of experiments
ranging from measurements of molecular relaxation times to laser fusion
studies.
One particularly useful application of picosecond laser pulses, and one
that will be the object of this chapter, has been in the development of an
ultrafast shutter (or gate) based on the optical Kerr effect (also called AC
Kerr effect) by DUGUAY and HANSEN [1969a]. This effect, which had been
predicted by BUCKINGHAM [1956] well before the laser, and which was
first observed by MAYERand G I R E[1964],
~ refers to the very short-lived
birefringence induced in transparent mediums by high-power laser pulses
propagating therein. The ultrafast optical Kerr shutter uses this optically
induced birefringence to gate light on and off on the picosecond time scale,
much in the way in which the conventional Kerr cell uses an electrically
induced birefringence to gate light on and off on the nanosecond time scale.
To date, gating times as short as 5 ps have been achieved, when using ultra-
-
short pulses from neodymium glass lasers. By comparison, the fastest
electrically driven Kerr cells have an open time of 0.5 ns. Moreover, the
recent development of subpicosecond pulses in CW modelockcd dye lasers
due to SHANKand IPPEN [1974] has enabled these workers to achieve
gating times as short as 2 ps in a CS2Kerr shutter (IPPENand SHANK [1974]).
The ultrafast optical Kerr shutter has found a number of applications,
163
164 THE ULTRAFAST OPTICAL KERR SHUTTER [IV, P 1
The simplest, and what has been so far the most useful configuration
for the ultrafast Kerr gate (or shutter), is shown in Fig. 1.
A short cell containing carbon disulfide is placed between two crossed
polarizers. In the spectral range 5000-7500 A sheet polarizers (like Polaioid
-
HN22) can be used. Elsewhere in the spectrum good extinction and sizeable
angular acceptance ( 15") require the ust of calcite (or mercurous chloride)
crystal polarizers. The CS2 cell is made of low birafringence glass and is
held gently to avoid strain birefringence. The gating laser pulse is directed
into the CS2 cell at a small angle a from the axis AX, along which the
signal light to be gated is traveling. When working with Nd: glass pico-
second pulse lasers, the gating pulse will typically be 10 ps in duration
at a wavelength of 1.06 p. In the absence of the gating pulse the crossed
-
polarizers shut out the signal light; the gate transmission in the OFF
state is typically 0.001 %.
SIGNAL
F
GATING PULSE
Fig. 1. The ultrafast optical Kerr shutter in its simplest form. Polarizers PI and P, are crossed
and have their polarization axes at *45" to the plane of polarization of the gating laser pulse.
Filter F greatly attenuates the infrared gating pulse to prevent possible damage to P, .
In some cases F can be dispensed with (see Momou and MALLEY [1974]).
166 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R CIV? Q 2
The gating pulse (infrared pulse in Fig. 1) is plane polarized in the hori-
zontal plane, as is usually the case with practical lasers which have vertical
Brewster faced components. Polarizers P, and P, have their axes at +45" to
thehorizontal. Asthegating pulse propagatesthrough the CS, cell, it creates
a region of birefringence that travels along with it. That ultrashort portion
of the signal light that happens to enter the CS2 cell at the same time as
the gating pulse experiences the effect of this moving birefringent region,
and as a result undergoes a change in polarization that allows it to be
partially transmitted through polarizer P, . The birefringence induced in
CS, by a laser pulse whose plane polarized electric field is E(t) in the medium
is given by :
Here Gn,, and dn, refer to changes in the refractive index in directions
parallel and perpendicular, respectively, to the applied field E(t), n 2 B is
the AC Kerr effect coefficient,z is the relaxation time of the AC Kerr effect
in the gate medium, and the bar over E 2 signifies a local time average over
one optical period.
In CS, the bulk of the induced birefringence arises when molecules are
rotated so that their axis of easy polarizability moves closer to the plane
of polarization of the applied field E(t). The relaxation time associated
with this motion is about 1.8 ps in CS, at 25C (BROIDAand SHAPIRO
[1967], IPPENand SHANK[1975]). This means that it takes 1.8 ps after
the passage of a 6-function gating pulse in order for the molkcules to
randomize again, thereby returning the ultrafast gate to its OFF state.
This time defines the minimum open time of a CS, shutter. For pulses
much longer than z = 1.8 ps, eq. (2.1) can be approximated by:
(2.7)
As a specific example let us consider the experiment of DUGUAY and
HANSEN [1969a], where infrared ultrashort pulses from a Nd: glass laser
are used to gate ultrashort green pulses derived from them by second
harmonic generation. The various numbers in eqs. (2.5) and (2.6) are:
3T1T2= 0.14 (HN22 type sheet polarizers from Polaroid Corp.), L =
m, 1 = 0.53 x m, n = 1.60 and n2B = 2.2 x m2/V as measured
for CS2 by MAYERand GI= [I9641 and by PAILLETTE [1969].
Using these numbers in eq. (2.7) we get
T(t) = 0.14 sin (3.15 x l O - I 3 P ) (2.8a)
or, more conveniently, expressing P in MW/cm2, we have
Under these conditions the gate open time is somewhat shorter than the
gating pulse. For example, if P(t) is a Gaussian with a full width at half
maximum (FWHM) of 14ps, T(t)will also be Gaussian, but with an FWHM
equal to 14/$ = 10 ps.
168 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R [IV. 3 2
15 - CS2 SHUTTER
HN22 POLARIZERS
T (%I
Fig. 2. The transmission T of a CS, ultrafast Kerr shutter at a signal wavelength of 0.53~
is plotted versus the power density in the gating beam. For a shorter signal wavelength the
maximum would move towards smaller gating power densities. The CS, shutter has a length
L = 1 cm and the polarizers are of type HN22 (made by Polaroid). With calcite crystal po-
larizers, the peak transmission could reach 50 4for unpolarized signal light.
s
CY4)(t) = T(t)P2(t-T) dt.
s
G(4)(2)cc P2(t)P2(t- 7)dt.
t
1 0 - * 4
-50 0 50 T'(pS)
Fig. 3. The solid line represents the transmitted green light signal G'4'(~')as a function of the
delay T' between the powerful gating 1 . 0 6 ~pulse and the weak probing second harmonic
(0.53~)pulse. The width of G(4)(~') is 15 ps and is approximately equal to the duration (at half
Also shown in Fig. 3 is a plot ,of G(z) when nitrobenzene is used as the
optical gate medium. Nitrobenzene molecules are bigger than CS2 mole-
cules, and once aligned along a certain direction by the AC Kerr effect, it
takes them longer to relax to a random orientation state after the passage
of the laser pulse. This is evident in Fig. 3 where the gate transmission is
seen to decay exponentially with a time constant of 16f2 ps after the
gating pulse has left the gate. Since the gate transmission is proportional
to the square of the induced birefringence, the relaxation timc associated
with the latter is 32+4 ps in nitrobenzene at 25C. This relatively long
relaxation time supports the view that the mechanism responsible for the
AC Kerr effect in nitrobenzene is indeed orientational in nature, in agree-
170 T H E U L T R A F A S T OPTICAL KERR SHUTTER [ W > 2
PULSE TRAIN
INPUT X.0.60p
POLARIZER
POLARIZER P, CHOPPER
LOCK-IN
ROTATOR
AMPLIFIER
I
X-Y RECORDER
VARIABLE
DELAY
Fig. 4.Gating with subpicosecond pulses has been achieved by IPPEN and SHANK[1974] with
this set-up. A continuously pumped dye laser produces a day-long train of pulses at I = 0 . 0 6 ~ .
The major part (95 %) of each pulse gates on the K e n shutter, while the minor part (5 %) probes
the shutter transmission. The availability of day-long trains of pulses allows the use of highly
sensitive lock-in detection techniques.
power) pulses split off from the input beam (wavelength: 0.60 p). This
time the pulses are so short that the more exact expression (2.1) for 6nll-an,
must be used in eqs. (2.3) and (2.4). The result is that the transmission T(t)
is given by:
As one varies the delay z by which the probe pulse lags the gating pulse,
the transmitted signal G(z) varies like:
G($) cc jeffiffi
T ( t ) p ( t- z f )dt. (2.12)
-3-2-1 0 1 2 3 4 5
TIME DELAY (psec)
Fig. 5. Signal G(T)transmitted by the CS, optical Kerr shutter when driven and probed by
1.2 ps pulses from a mode-locked CW dye laser (IPPEN and SHANK [1975]). The ordinate is
linear and in arbitrary units. The shutter transmission at T = 0 is of the order of 0.1 /,.
-
The peak power density at the waist is lo8 W/cm and the beam waist
extends over about 1.0 mm. The product PL therefore falls somewhat
short of what is required for efficient gating (see Fig. 2), explaining the low
peak transmission (- 0.1 %) observed.
In order to get more signal and to bring out the effect of the relaxation
time z more clearly, IPPENand SHANK[1975] have also measured the
transmission T ( t )when the shutter is biassed half-way to the first maximum.
This is achieved by introducing a quarter-wave plate between the crossed
172 T H E U L T R A F A S T O P T I C A L KERR s H U T T E R [IV, I2
polarizers. For this case eq. (2.1la) becomes
T(t) = $Tl T2 sin2 [44+ (p(t)/2] (2.14)
where q ( t ) designates as before the laser induced phase lag. For small.&
eq. (2.14) approximates to
W )= fiT1 T,(1 +do), dt) d2. (2.15)
The expression for the transmitted signal G(z) becomes
G(z) oc
Lm+ (1 q(t))F(t
- 7) dt.
For pulse durations well under one picosecond, G(z) simplifies in this case
(2.16)
to
+
G(z) a const. exp (- z/z). (2.17)
The curve measured by IPPENand SHANK [1975], shown in Fig. 6, por-
trays a steep rise and an exponential decay indicating a time constant
z = 2.1 f0.3 ps, one of the shortest lifetimes to be measured directly by
optical techniques.
Forthcoming increases in the peak power of subpicosecond pulses
promise to render possible the study of light induced refractive index
changes in a great variety of liquid and solid systems and on a very fine
timescale. These studies should provide a most valuable complement to
light scattering techniques (FABELINSKII [19681) in the study of ultrafast
molecular dynamics.
W
OPTICAL KERR SHUTTER
z PULSE CORRELATION IN CS,
o_
UJ (HALF-WAY BIAS)
v)
-3-2-1 0 1 2 3 4 5 6
TIME DELAY (psec)
Fig. 6. Same as Fig. 5, except that this time the shutter is biassed half-way by inserting a
quarter-wave plate between the two crossed polarizers.
IV, 8 23 THE ULTRAFAST OPTICAL KERR SHUTTER 173
Fig. 7. This curve plots the transmitted signal for a glass ultrafast shutter driven by 1 . 0 6 ~
pulses from a mode-locked Nd:glass laser. The probe pulses (d = 0.53~)are harmonically
derived from the infrarcd ones. The width of the curve for G4)(~)at half-height is
approximately equal to the duration of the infrared pulses. Although harder to gate on, glass
Kerr shutters would ultimately allow gating on the femtosecond (10-l5s) time scale.
The curve for G(4)(z) obtained with LaSF-7 is consistent with a very
fast response. Any relaxation time, if present, is less than 5 ps. One dis-
advantage of glass optical gates is the high power density required to get
the same transmission as with a CS2 gate. With LaSF-7 the first trans-
mission maximum occurs at P = 15 GW/cm2 for a one-centimeter thick
gate.
These experiments have served to measure the nonlinear index of
174 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R CIV. 8 2
In the geometry shown in Fig. 1 two other factors limit the resolution
for pulses less than 5 ps in duration. One is the group velocity mismatch
between the gating and the gated beams. In CS2 a green light pulse will fall
behind an infrared (2 = 1.06 p) pulse by 2.0ps after 1cm of travel. In LaSF-7
glass that number would be 1.8 ps. A tenfold reduction in the gate thickness
down to one millimeter would reduce this differential delay to 0.18 ps in
LaSF-7. However, the power density required to reach the first transmission
maximum would then be increased to 150 GW/cm2, a relatively large
value, but one that is well within the state of the art.
Another factor lengthening the time response of the optical gate is the
spread in arrival times at the gates for different parts of the infrared gating
beam, due to the finite angle a between gated and gating beams. As can
be seen in Fig. 8, this spread amounts to (d tana)/c, where d is the gating
Fig. 8. Illustration of the time smearing factor (d tan a)/c introduced by the small angle tl
between gating and gated beams. In the figure the gating and gated pulses are assumed to be
nearly delta-functions and are represented by the solid and crossed areas. When the bottom
parts of the two beams enter the gate in perfect synchronism, the upper parts are (d tana)/c
apart. In the experiment of DUGUAY and HANSEN[1969a], this factor amounted to 2.0 ps.
N ,5 21 T H E U L T R A F A S T O P T I C A L KERR SHUTTER 115
--
SIGNAL LIGHT
-
c
Fig. 9. One way to avoid the time smearing factor of Fig. 8 is to employ a geometry where the
gating and gated beams are collinear. (a) In the geometry of MOUROUand MALLEY[1974]
the gating beam at 1 . 0 6 ~is sent through polarizer P, (type Polaroid HN22) where it suffers
negligible absorption, and therefore undergoes very little change in polarization (i.e., it
stays plane polarized at 45" from the axis of P,). The signal beam is coupled in by means of a
dichroic dielectric mirror. (b) In the set-up used by Yu and ALFANO [I9741 a dichroic mirror
is used to couple the gating beam into the shutter.
2.7. SELF-FOCUSING
(2.19)
Fig. 10. Set-up used by SHIMIZU and STOICHEFF [I9691 to study self-focused filaments in CS,,
and as a method of displaying the 1 . 0 6 ~pulse. When used as a Kerr gate, the transverse
geometry provides an open slit moving from A to B at the speed of light in CS, . In photo-
graphing the transmitted signal light, axis AB becomes a time axis and one can use this set-up
to display the signal (MALLEY and RENTZEPIS [1970]). Self-focusing of the gating beam can be
a problem in a transverse geometry Kerr gate.
from A to B in Fig. 10, the Kerr gate first opens at A and then later on at B.
This way the time profile of the pulse incident from the top appears as a
spatial intensity profile on the film.
The transverse geometry has also been used by FISCHER and ROSSMANITH
[19731 in studies of synchrotron radiation, and by RICHARDSON and SALA
[I9731 in their ultrafast framing photography of laser produced plasmas,
which we will return to later. TOPP,RJZNTZEPIS and JONES [1971] have
combined a transverse Kerr shutter with a spectrograph in order to time
resolve the spectrum of an optically pumped rhodamine 6G laser.
Self-focusing is a problem that is more severe in a Kerr gate of transverse
geometry. Because the minimum opening time at a given z is limited by
the transit time nd/c, it is necessary to keep d as small as possible. For 5 ps
resolution, for example, one must have d = 1.0 mm. In order to get the
same phase retardation 640 as one had in Fig. 1 over L = 1 cm, it is necessary
to increase the gating power density tenfold to 5 GW/cm2. As a result of
both the smaller diameter and the increased power density, the gating
beam can undergo self-focusing within the long CS2cell dimension (- 1 cm)
and break up in filaments as in the experiment of SHIMIZU and STorcHEw
[19691. This leads to severe inhomogeneities in the gate transmission.
Fig. 11. Experimental set-up used for the ultrahigh-speed photography of light pulses in flight.
Ultrashort infrared (1.06~)pulses from a Nd: glass laser open the ultrafast Kerr shutter for
about 10 ps. The green pulses are harmonically derived from the infrared pulses. As they pass
through a cell of milky water, light scattering makes them brightly visible from the side, so that
they can be stop-motion photographed in flight (see Fig. 12). Filter F is made of infrared
-
absorbing glass that is transparent to the visible; it attenuates the 1 . 0 6 ~pulse by a factor
of 1000.
[1971] and DUGUAY[1971]). The set-up that was used to accomplish this
is shown in Fig. 11. An ultrashort pulse of green light is directed into a cell
containing a colloidal suspension of milk particles in water. The milk
particles greatly increase the instantaneous light scattering that occurs in
pure water, thereby making the green pulse brightly visible from the side.
(A laser pulse propagating in vacuum would not be visible from the side.)
A 35 mm camera, whose mechanical shutter has been manually opened,
is placed behind a conventional ultrafast CS2 shutter. The latter is driven
by an infrared pulse (A = 1.06 p) produced by a mode-locked Nd: glass
laser. The green pulse is harmonically derived from the infrared pulse.
The lengths of the paths followed by the two pulses are adjusted in such
a way that when the shutter opens for about 10 ps it captures a picture of
IV, 31 ULTRAHIGH SPEED PHOTOGRAPHY 179
Fig. 12. An ultrashort pulse of green laser light is photographed in flight as it propagates
from right to left through a cell of milky water. The scale is in millimeters. The shutter open time
was about 10 ps. The red spot on the left side of the picture is the impression made on the high
speed Ektachrome film by the infrared laser pulse used to activate the ultrafast Kerr shutter
and incompletely attenuated by filter F in Fig. 11.
the green pulse in midflight through the milky water cell, as shown in
Fig. 12. Thus the green light bullet is stop-motion photographed in
flight. The red round spot to the left of the cell results from the direct
head-on impact onto the film made by the infrared pulse incompletely
attenuated by filter F in Fig. 11. Thus Fig. 12 also provides a pictorial
representation of second harmonic generation from infrared (in red) to
green.
In principle the ultrafast photography of light in flight constitutes one
of the most direct ways of displaying ultrashort laser pulses. Pulses obtained
from a mode-locked Nd: glass laser stand out bright and well isolated,
180 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R v, 03
the contrast between the brightness of the pulse and that of the background
being better than 200 to 1.
Even though the ultrashort pulse display shown in Fig. 12 is one of the
best obtained so far, it still is far from perfection because the shutter remains
open for about the same time (- 10 ps) as the green pulse duration. In 10 ps
light moves 2.2 mm in water, and as a consequence, the picture of the
green pulse is blurred out. One cannot recover the precise shape of the
green pulse. What would be needed in this case would be a 0.3 ps gating
pulse (perhaps from a mode-locked dye laser) and a glass Kerr shutter
because the green pulse is known to have a subpicosecond substructure
(SHAPIRO and DUGUAY [19693). Nevertheless, this technique has proved
useful in the study of weaker satellite pulses accompanying the powerful
ultrashort pulses generated by model-locked Nd : glass lasers (see DUGUAY
and MATTICK [197I]).
The technique of gated picture ranging has been used on the nanosecond
time scale to improve the visibility of targets, such as airplanes and ships,
obscured by fog or other obstacles. A powerful nanosecond light pulse is
sent out from the observation point towards the target. At a preselected
later time an electronic image converter tube is gated on for a few nano-
seconds and only the echo image scattered back by the target is recorded.
Earlier (or later) echoes from the fog corresponding to closer (or farther)
ranges are not recorded because the picture tube is gated off at those times.
With picosecond and now even subpicosecond laser pulses, the same
technique can be applied on the centimeter and even millimeter scales by
using the ultrafast Kerr shutter to gate the echoes. In a feasibility experiment
using the set-up shown in Fig. 13(a), a second-harmonic green pulse was
sent through a piece of thin paper tissue (facial tissue) towards a target
carrying the stylized drawing of a bell (shown unobscured by the tissue in
the upper left corner of Fig. 13(b)). When photographed under room light
illumination (see Fig. 13(b), upper right corner), the, target is completely
obscured by the tissue. When the green pulse is sent through the tissue and
the ultrafast shutter is turned on at the right time by the infrared pulse,
only the echo from the target is recorded by the camera. Two results are
shown in the lower part of Fig. 13(b).
In certain parts of the human body the skin is partially transmitting to
light, and veins, for example, can be seen. With the recent achievement of
0.5 ps laser pulses and 2 ps Kerr gating times, a spatial resolution of better
U L T R A H I G H SPEED P H O T O G R A P H Y 181
f TARGET
1 cm
Fig. 13. (a) Schematic description of set-up used for gated picture ranging through a piece
of paper (or "facial") tissue. An ultrashort green laser pulse first illuminates the tissue and then
33 ps (1 cmjc) later the target. The target echo, which carries the image information, lags the
tissue echo by 66 ps. (b) The top left picture shows the target under room lighting when the
tissue is removed. With the tissue back in place the target is completely invisible under room
lighting. The bottom two pictures show two examples of gated picture ranging through the
tissue. The target is visible, but the passage of the image carrying echo through the tissue has
degraded the quality of the picture (the ultrafast shutter itself does not degrade the picture, see
DUCUAY and MATTICK[1971]).
Fig. 14. Picosecond framing photography of a laser produced spark in air achieved by
combining an electronic streak camera with an ultrafast optical Kerr shutter. The latter is
opened for 10 ps every 6.7 ns. In the 6.7 ns interval during which the Kern shutter is opened,
the position of the image is electronically swept down in the streak camera tube, so that the
next 10 ps frame transmitted by the Kerr shutter is recorded well below the preceding image.
(a) Five frames in the initial stage of optical breakdown in air. (b) Five frames about 200 ns
after breakdown. RICHARDSON and SALA[19731.
IV, D 41 SAMPLING OPTICAL SIGNALS 183
-
elements needed in electronic sampling are an ultrafast gate (the shortest
gating times are 20 ps at present) and an amplifier to amplify the sample
cut out at a gven time from the signal.
An ultrafast optical Kerr shutter (or gate) and a photomultiplier con-
stitute the analogous key elements in applying the sampling technique to
optical signals. Since a Kerr shutter transmission of 50% can in principle
be achieved when using crystal (e.g., calcite) polarizers, and since photo-
multipliers have quantum efficiencies as high as 30%, this optical sampler
is close to the theoretical limit of sensitivity.
The optical sampling arrangement used by DUGUAY and HANSEN [1969b]
is shown in Fig. 15. The CS2 gate is driven, as before, by 1.06 p pulses
about 10 ps in duration generated by a Nd: glass laser. The 0.53 p second
harmonic pulses are sent into a cell containing a cyanine dye dissolved in
I I
Fig. 15. Picosecond fluorescence decay times were first measured by using an ultrafast shutter
together with a photomultiplier in order to do point by point sampling of optical signals
(DUGUAY and HANSEN[1969]). The fluorescent dye is excited by green pulses about 10 ps in
duration and emits a fluorescent signal at 1 = 0 . 7 5 ~ On
. a given laser shot, a 10 ps sample is
sliced from the incoherent fluorescence signal and is detected by the photomultiplier. As the
delay is vaned from shot to shot, the entire signal can be sampled as a function of time.
184 THE ULTRAFAST OPTICAL KERR SHUTTER [IV, 4
-cn 1
I- -DDI IN METHANOL
PROMPT RESPONSE(SHG)
-
4
I-
3
n
I-
0 I \
I
a I I I I I I I I
-40 -20 0 20 40 60 80 100
T I M E t ( pSeC 1
Fig. 16. The solid line shows the fluorescence signal emitted by the dye DDI sampled as a
function of time. Time is measured relative to the arrival time of the green pulse which excites
the fluorescence. The dotted line represents the prompt response of the measuring system.
It is obtained by removing the dye and by sampling the green pulse itself (it is the same as
G(4)(2) in the text). A deconvolution of the prompt curve from the fluorescence curve gives
a decay time of 14+3 ps for DDI dissolved in methanol or acetone.
X=l.OSp
Fig. 17. The echelon technique developed by TOPP,RENTZEPIS,JONFS [ 19713 makes use of an
echelon reflector to divide the signal beam into a number of segments spaced apart in time
by typically 3 ps. One ultrashort sample is cut out by the ultrafast shutter from each segment.
The samples are recorded on photographic film (as shown above) or by a linear array of
photodiodes.
The echelon step is chosen so that the various segments of the reflected
signal are progressively delayed at intervals of 4 ps, for example, as they
enter the shutter. When the shutter opens for a time dictated by the infrared
gating pulse, one ultrashort sample is cut out from each segment of the
signal beam. These samples are recorded photographically and by photo-
densitometry, a plot of segment height vs. number (Fig. 18) gives the time
profile of the signal. Recently this method has been improved by NETZEL,
RENTZEPIS and LEIGH[1973] by using a linear array of photodiodes instead
of film to record the signal segments.
TOPP,RENTZEPISand JONES [1971b] have also combined the echelon-
shutter technique with a spectrograph to time resolve the spectrum of the
stimulated emission from a rhodamine 6G laser. Interesting data on
the photobleaching of rhodopsin have been receniy obtained by using
these techniques (see NETZEL, RENTZEPIS and LEIGH[19733).
186 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R CIV, 4
Fig. 18. Experimental result obtained with the echelon technique. In this example the signal
beam is the second harmonic of the 1 . 0 6 ~gating beam. The various segments of the green
pulse probe the opening of the ultrafast shutter.
yi
Fig. 19. Set-up used to optically sample ultrashort laser pulses and display them on a real-time
oscilloscope. The pulse of green ( A = 0.53~) light to be displayed comes down from the top
left and enters a cell containing highly diluted milk (or a Ludox silica suspension type LS).
The ultrafast shutter is driven by the infrared pulse shown. The shutter opens only once and
cuts out a slice (or sample) from each scattered pulse. The samples are centered at 3.9 psec
intervals from the leading to the lagging edge of the incident green pulse. The fiber array
transforms the spatially distinct samples at points A, B, C, . . . . into temporally distinct
pulses on the oscilloscope screen. The fiber ends at A, B , C, . . . . are butted against a
glycerin-wetted glass window (not shown) for good optical coupling.
has the advantage over the echelon (see TOPP,RENTZEPIS and JONES [1971])
in this application, of replicating the signal pulse not only temporally but
also spatially. When the signal beam has spatial inhomogeneities, this
insures that all replica pulses entering the shutter are identical in shape.
The 10 samples cutout by the ultrafast gate are sent into 10 optical
fibers, cut to progressively longer lengths and giving delays of 5, 15,25, 35,
. . ., 95 ns. This set of optical fibers if referred to as an organ array, by
analogy with the progressively longer pipes of the musical organ. The out-
188 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R [IV, 0 4
VOGEL'S MULTIBE
~-n / n /n /n J
OPTICAL SIGNAL
put ends of the fibers are placed near the face of a single photomultiplier
of 3 ns response time, the output of which is displayed on a fast (0.8 ns
risetime) oscilloscope. This way the 10 samples, which were taken within
the same 5-10 ps interval, are spaced out at 10 ns intervals for detection
and display.
An example of a display obtained with Vogel's multibeamsplitter is
shown in Fig. 21. The second harmonic pulse at A = 0.53 p is being sampled
here. If the shutter had opened for, say, 2 ps in this case, the envelope of
the samples would represent the green pulse shape. For that laser shot,
however, the duration of the 1.06 p gating pulse was more like 16 ps,
leading to a gate open time of about 10 ps, a duration about equal to the
green pulse width. The envelope of the sample pulses in Fig. 21 represents
in fact the function G(4)(7)given in eq. (2.10).
The sensitivity of the OSO with Vogel's beamsplitter was such that
pulses with peak powers down to a few watts could be displayed. With
various improvements (see VOGEL,SAVAGE and DUGUAY [19741) sensitivity
down to the level of ten milliwatts in optical power seen possible in practice.
1 4 . 1 ~ 8SAMPLING STEP
Fig. 21. (a) Sampled display of an ultrashort laser pulse obtained with a Vogel multi-
beamsplitter, an ultrafast shutter and an organ array as shown in Fig. ,20. The envelopqof the
samples constituks the displayed shape. This shape is the convolution G4(5) described in the
text involving the green pulse shape ( - 12 ps wide) and the shutter transmission function T(t)
,.,
(also 12 ps at half maximum in this sample). The width of the display at half maximum is
about equal to the duration of the infrared pulse that drove the shutter in this example (17 ps).
The ordinate scale points up the excellent sensitivity available with this technique. (b) Picture
obtained when the ultrafast shutter is manually opened by removing P, . Unequal heights
reflect residual ineqiialities in the coupling and transmission of the various channels.
190 T H E U L T R A F A S T O P T I C A L KERR S H U T T E R v, P4
QATING PULSE
F
.LINEAR ARRAY OF
DE
Fig. 22. Set-up used by MOUROUand MALLEY[1974] to measure ultrafast relaxation times.
A second harmonic green pulse 4ps in duration propagates through the dye solution
N
(erythrosin in water) leaving an exponentially decaying tail of fluorescent light in its wake.
This fluorescence tail is imaged through the ultrafast shutter onto an array of 512 photodiodes.
When the shutter is opened for 4 ps by an infrared pulse, a record of the tail is captured by
the linear diode array.
2 4 -
z
a
K
a
K
s 3 -
-
K
a
J
a
E 2 - At = 4 psec
v)
0
x
K
530nm
0 I -
0
W
K
- 0
t(psec)- 0 100
Fig. 23. Result obtained by MOUROUand MALLEY119741 using the arrangement of Fig. 22.
The risetime of the fluorescence signal from erythrosin B in water (3 x 10- M) appears limited
by the widths of the excitation and gating pulses .,( 4 ps). When the dye is replaced by milky
water, a stop-motion image of the second harmonic (530 nm) pulse is projected onto the linear
photodiode array: the sharp curve obtained gives the prompt response of the system, as in
Fig. 16.
IV, 51 CONCLUDING R E M A R K S 191
tion derived from a Nd: glass laser are passed through a cell containing a
molar solution of erythrosin in water. The green beam has been col-
linated down to a diameter of 1 mm. The spontaneous fluorescence light
emitted along this narrow track is imaged through collinear Rerr gate onto
a linear array of 512 photodetectors (only 5 are shown in Fig. 19 for clarity).
Each photodetector collects light from a small volume element situated
along the fluorescence track. The light pulses emitted by each volume
element reach the ultrafast shutter after a delay proportional to the distance
along AZ. When the shutter is opened by the 1.06 p pulse (-6 ps in dura-
tion) it cuts out one sample from each pulse. Just as in the optical sampling
oscilloscope, the samples are spaced uniformly from the leading to the
lagging edge of the pulse shape.
The result obtained by Momou and MALLEY [1974] for erythrosin is
shown in Fig. 23. The number of samples taken is large enough to make the
recorded trace appear continuous. Mourou and Malley found a fluores-
cence signal risetime of 4f 1 ps, that is essentially equal to the time resolu-
tion of the apparatus. The implication is that the true fluorescence risetime
(i-e., the risetime under deltafunction pulse excitation) is less than 4 ps.
Another way of looking at the Mourou-Malley experiment is from the
point of view of ultrahigh speed photography (see DUGUAY and HANSEN
[1970]). The green pulse (or light bullet) leaves in its wave a tail of
fluorescence light that is stop-motion photographed in flight by the Kerr
shutter and detector array. The latter replaces the film and records one only
horizontal line of the picture, but that is all that matters here.
Momou and MALLEY [1974] have also used the ultrafast shutter and
their detector array in conjunction with a spectrograph to do time resolved
spectroscopy of the spontaneous emission of rhodamine 6G on the pico-
second time scale.
0 5. Concluding Remarks
The ultrafast Kerr shutter has established itself as a useful instrument
in fields of studies involving picosecond laser pulses. The shutter has been
driven by pulses derived from a variety of lasers, including COz laser pulses
(see DUGUAY and SAVAGE [1973], OWEN,COLEMAN and BURGESS[1973]).
The potential of this instrument in studies of ultrafast molecular dynamics
has been left largely unexploited so far, probably for a number of reasons,
one of which certainly being the difficulty and expense of producing stable
powerful laser pulses. If the day comes when laser pulses achieve the
reliability and flexibility of electronic pulses, use of the ultrafast Kerr
192 THE U L T R A F A S T O P T I C A L K E R R S H U T T E R CIV
shutter will become an easy task, a task that will not involve the difficulties
and dangers of high voltage pulses used in conventional Kerr cells.
The use of subpicosecond pulses in driving the shutter (IPPENand SHANK
[19751) has opened a most intriguing new frontier where the electronic
Kerr effect will certainly be called into play. Thus, one century after the
discovery of the DC Kerr effect (KERR[1875]), a closely related effect,
the AC Kerr effect, is playing an active role at the pinpoint of technology.
Acknowledgements
I would like to acknowledge the help of Mrs. Jeri Romaine and Mr. A.
Savage in preparing the manuscript for publication.
References
ALFANO,R. R. and S. L. SHAPIRO, 1972, Optics Commun. 6, 98.
ALFANO,R. R., S. L. SHAPIRO and M. POPE,1973, Optics Comm. 9, 388.
BROIDA,H. P. and S. L. SHAPIRO, 1967, Phys. Rev. 154, 129.
BUCKINGHAM, A. D., 1956, Proc. Phys. SOC.B69, 344.
CAMPILLO, A. J., S. L. SHAPIRO and B. R. SUYDAM, 1973, Appl. Phys. Lett. 23,628.
CAMPILLO, A. J., S. L. SHAPIROand B. R. SUYDAM, 1974, Appl. Phys. Lett. 24, 178.
CHIAO,R. Y ., E. GARMIRE and C. H. TOWNES, 1964, Phys. Rev. Lett. 13,479.
DEMARIA, A. J., D. A. STETSERand H. HEYNAU, 1966, Appl. Phys. Lett. 8, 174.
DUGUAY, M. A., 1971, American Scientist 59, 550.
DUGUAY, M. A. and J. W. HANSEN,1969a, Appl. Phys. Lett. 15, 192.
DUGUAY, M. A. and J. W. HANSEN,1969b, Optics Commun. I , 254.
DUGUAY, M. A. and J. W. HANSEN, 1970, NBS Special Publication No. 341, pp. 45-49 (Gov.
Printing Office, Washington, D.C.).
DUGUAY, M. A. and A. T. MATTICK,1971, Appl. Optics 10, 2162.
DUGUAY, M. A. and A. SAVAGE, 1973, Optics Commun. 9.212.
DUGUAY, M. A. and J. N. OLSEN,1975, Picosecond X-ray Pulses, IEEE J. Quant. Elect.
QE11, 170.
FABELINSKII, 1. L., 1968, Molecular Scattering of Light (Plenum Press, New York).
FISCHER, R. and R. ROSSMANITH, 1973, IEEE Trans. Nucl. Sci. NS-20, 549.
GIORDMAINE, J. A., P. M. RENTZEPIS, S. L. SHAPIRO and K. W. WECHT,1967, Appl. Phys. Lett.
11. 216.
JANSSEN, J. M. L., 1950, Philips Tech. Rev. 12, 52.
IPPEN,E. P. and C. V. SHANK,1975, Appl. Phys. Lett. 26. 92.
KELLEY, P. L., 1965, Phys. Rev. Lett. 15, 1005.
KERR,J.. 1875, Phil. Mag. 50, 337.
MALLEY. M. M. and P. M. R~NTZEPIS. 1970. Chem. Phys. Lett. 7, 57.
MAYER,G. and F. GIRES.1964. C. R. Hebd. Seanc. Acad. Sc. Paris 258, 2039.
MOUROU. G. and M. M. MALLEY,1974, Optics Commun. 11, 282.
NETZEL, T.. P. M. RENTZEPIS and J. LEIGH, 1973, Science 182, 238.
OWEN,T. C., L. W. COLEMAN and T. J. BURGESS, 1973, Appl. Phys. Lett. 6,272.
OWYOUNG, A,, R. W. HELLWARTH and N. GEORGE, 1972, Phys. Rev. B5,628.
PAILLETTE, M., 1969, Annales de Physique 4, 671.
RICHARDSON, M. C. and K. SALA,1913. Appl. Phys. Lett. 23. 420.
IVf REFERENCES 193
BY
PAGE
9 1 . INTRODUCTION. . . . . . . . . . . . . . . . . . . 197
5 2. THEORETICAL CHARACTERISTICS OF SPECTRO-
SCOPIC DIFFRACTION GRATINGS. . . .. .. . . . 198
0 3. BASIC PRINCIPLES OF HOLOGRAPHIC DIFFRACTION
G R A T I N G S . . . . . . . . . . . . . . . . . . . . . . 200
5 4. PRODUCTION OF HOLOGRAPHIC GRATINGS . . . . 216
0 5. PROPERTIES OF HOLOGRAPHIC GRATINGS AND
COMPARISON WITH CLASSICAL GRATINGS. . . . . 223
5 6. FURTHER IMPROVEMENTS OF HOLOGRAPHIC
GRATINGS.. . . . . . . , . . . . . . . . . . . . . 242
ACKNOWLEDGEMENTS. . . . . . . . . . . . . . . . . . 242
REFERENCES. . . . . . . . . . . . . . . . . . . . . . . 242
0 1. Introduction
The first experiments with grating like structures were probably made
by the American astronomer David RITTENHOUSE [17861 in Philadelphia.
He used parallel hairs laid in a fine screw and observed diffraction effects
of light. The first ruled optical gratings were produced by Joseph VON
FRAUNHOFER [1821/22], who also discovered the fundamental properties
of optical diffraction gratings. Comprehensive reviews of the history,
theory and manufacture of gratings have been given by KAYSER [1900],
STROKE [1963, 19671 and HARRISON [19733.
Because of the severe mechanical problems of ruling gratings many
alternative methods of production have been considered. MICHELSON
[19271 suggested producing gratings by photographing stationary waves
using Lippmann plates. At the National Physical Laboratory in England
BURCH and PALMER [1961] made gratings by photographing fine inter-
ference fringes and measured the shift caused by processing of photographic
emulsions. LABEYRIE [1 9661 suggested using layers of bichromated gelatine,
Daguerre layers, photoconductive layers in connection with sputtering
techniques or thermoplastic layers as recording materials. None of these
methods gave gratings suitable for practical spectroscopic use. Stimulated
by the fact that in high resolution stellar spectroscopy with large telescopes
it is neccessary to use large diffraction gratings of good quality, the authors
of the present article proposed making diffraction gratings holographically
by using photoresist layers. The first results, obtained in 1967, already
demonstrated the good optical quality of such holographically made
diffraction gratings (RUDOLPH and SCHMAHL [1967a, b, c, 19681). In the
meantime several groups have started working in this field and holographic
diffraction gratings compete with classically ruled gratings from the
visible to the ultraviolet and soft X-ray regions.
The scope of this article is to describe the basic method of making gratings
holographically, to describe the results obtained and to compare these
results with those obtained from gratings produced by traditional means.
197
198 HOLOGRAPHIC DIFFRACTION GRATINGS [v. 8 2
0 2. Theoretical Characteristics of Spectroscopic Diffraction Gratings
The notation for the basic grating equations is introduced in this section
and a brief summary of the theory of spectral image formation by optical
gratings is given (STROKE [1963]). The relation between the wavelength
and the angle of diffraction is given by the grating equation
ma1
sin i+sin i' = -,
a
where i and 'i are the angles formed by the wave propagation vectors of
the incident and the diffracted waves with the normal to the grating surface.
The grating spacing is a, the wavelength of the diffracted radiation 1and
the spectral order m. The angle of diffraction 'i is positive if the incident
and diffracted wavefronts are on the same side of the normal and negative
if they are on opposite sides of the normal.
The angular dispersion, 6, is given by
di' m 1
g=-=____- --
dA cos ilea I cos i'
+
(sini sin it).
In autocollimation i = 'i
(2.2a)
With A = W .cos 'i and di' = Po in equation (2.2) and using equation (2.4)
3 21 SPECTROSCOPIC DIFFRACTION G R A T I N G S 199
Fig. I .
where the angular coordinates of the diffraction pattern in the focal plane
is given by fi = n x / ( A * f / A and
) q = u/(A/2),the normalised width of the
diffracted plane wave. The complex amplitude of the diffracted mono-
chromatic plane wave g(q) is given by
Here d(q) is the phase of the diffracted wavefront which can be derived
from wavefront interferograms. d(q) is constant for a plane wave. We
also assume that the amplitude is constant, which means that the efficiency
and polarisation properties are uniform over the whole area of the grating.
With %(q) = 1
I@)
1
=-
4n
%*(/9) = -
?I CB>
___
200 HOLOGRAPHIC DIFFRACTION G R A TIN GS cv. 4 3
where $I@) is normalised so that the total intensity is 1. Wave front aber-
rations A(?) are introduced by periodic, random and progressive errors
in the rulings and also by variations in the flatness of the blank or in the
layer in which the grooves are made. Aberrations also occur if $I(?) is not
constant. All aberrations give rise to deviations from the theoretical
diffracted intensity distribution.
High quality diffraction gratings should therefore have the following
properties :
a) A high ruling accuracy; i.e. periodic, random and progressive ruling
errors have to be as small as possible. Periodic errors result in ghosts in
the spectrum, random errors give broad wings to the diffraction pattern
i.e. scattered light over a broad wavelength region. Progressive ruling
errors (also known as error of run) over parts of the grating give rise to
deviations from the theoretical line profile near the centre of the line.
b) Good optical quality of the grating blank and the layer in which the
individual grooves are made.
c) A well defined profile for the individual grooves to obtain high
efficiency and small amount of light scattering.
d) The profile of the grooves should be uniform over the entire grating
to obtain constant intensity over the diffracted wavefronts when using
the grating in a spectrograph or a spectrometer.
and (3.1)
From Fig. 2 i = a+/?, 'i = or-8. 'i is negative if i and 'i are on opposite
sides of the normal in the xy-plane. Since we are only interested in the
spatial frequency and not in the position of the interference maxima and
202 HOLOGRAPHIC DIFFRACTION GRATINGS cv, 3
- A
a= - (3.4)
sin (cl+fi)+sin (a-fi) 2 sin clcos j? '
Using a symmetrical system (fi = 0) and I = 0.4 pm one obtains fringes
with the following spacings (Table 1). For a perfect plane grating the
grating spacing is constant over the whole area of the grating. Variations
in the spacing yield aberrations, A(?), in the diffracted wavefronts and
consequentlydeviations from the ideal intensity distribution in the spectrum.
TABLE1
Grating spacing as function of the intersection angle a for I = 0.4 pm
uc"1 a b l grooves/mm
1.1 10 100
11.5 1 1000
53.1 0.25 4000
Table 2 shows values of Ax for various values of i' and A(?) for A = 0.4 pm.
The larger i', the better Ax has to be. In addition, the value of Ax has to be
smaller as the wavelength at which the grating is used decreases. When
constructing the grating holographically according to Fig. 2, the wave-
fronts of the two beams S1and S2 will have aberrations due to the optical
elements used. We assume here that S, has an ideal wavefront and S2 has
an aberration A(?). Then a grating produced with these two wavefronts
- and on an ideal blank and photoresist layer - reconstructs the beam S2
- without considering amplitude factors - when the grating is illuminated
with the wavefront S1.We can call the reconstructed beam the + 1 order.
This example demonstrates that for the construction of a holographic
grating which will be used in the first (or low) order the optical elements
with which the beams S1 and S2 are produced do not neccessarily have
to be much better than the collimator and the camera of the spectrograph
in which the grating is mounted. To make gratings holographically which
are used in higher orders (echelle gratings) the accuracy of good optical
elements is not sufficient. The reason is, that the wavefront aberrations
for a grating (in autocollimation) are given by
204 HOLOGRAPHIC DIFFRACTION GRATINGS [v. 3
where x is the width of the grating up to the nth groove, measured from
the middle of the grating. Ax is the deviation of the nth groove from its
ideal position. For the change of the grating spacing as a function of n it
follows from eq. (3.7) that
-=I
d2x _
4 -_
1 nv-2
(3.8)
dn2 v=2 (v-2)!
The term d1 corresponds to a progressive linear ruling error and yields an
astigmatism of the grating, and higher terms of 6, influence the spectral
resolution.
The superposition of two divergent (or two convergent) beams yields an
interference field which can be described as a section of a family of two-sheet
hyperboloids of revolution (hyperbolic case). In Fig. 3 two divergent beams
originate from the two points F1 and F2 which are the focal points of th$
hyperboloids of revolution that are the geometrical loci of the interference
maxima. The hyperboloids are given by x2/a: -y2/ai - z2/a; = 1 with
2al = Ir, -r21 = nl, n = 0, 1, 2 ... and a: = a; = cz -$n2A2. In a plane
parallel to the xy-plane z = h the geometrical loci of the interference
maxima are hyperbolas
xz = $n2A2(1 + ( y 2 +h2)/(c2-tn2A2)). (3.9)
The spacings a of the interference maxima, which are given by the angle
V, o 31 BASIC PRINCIPLES 205
Fig. 3.
+
2a = F1PF2 ,are considered as a function of x in the vicinity of the plane
x = 0, i.e. the divergences of both beams are nearly equal. In all practical
cases c is in the order of several kilometers, that means c > nl.
With y = s and s2+h2 = b2 we obtain from eq. (3.9)
+ .. .)) (3.10)
and hence on taking the square root we obtain the following expansion
for x:
x = b2 *. n +
4 ( I + F') g(l+
$)-i* n3+. . ..
(3.11)
2
We note that eq. (3.11) contains n in odd powers only.
For the change of the grating spacing as a function of n it follows from
eq. (3.11)
da d2x
- - 6A3b2 $)-i
-=
dn dn2 -w *n+ .... (3.12)
A comparison of eq. (3.12) with eq. (3.8) shows that 6 , = 0, that means,
that in the hyperbolic case the grating is -to the first approximation - free
of aberrations.
The superpositionofonedivergent beam withoneconvergentbeam yields
an interference field which can be described as a section of a family of ellip
soids of revolution (elliptic case). In Fig, 4 one divergent beam originates
from the point F, and m e beam converges to the point F,. F, and F,
are the focal points of the ellipsoids of revolution that are the geometrical
loci of the interference maxima. The ellipsoids are given by x2/a: +y2/ai +
206 HOLOGRAPHIC DIFFRACTION GRATINGS
Fig. 4.
t
z 2 / u : = 1 w i t h 2 a 3 = ( r , + r 2 ) = 2 c + t * A a n d a ~ = a ~ = c - t - A + $ t 2 AI 2n .a
plane z = h parallel to the xy-plane the geometrical loci of the interference
maxima are circles
x2 = - y2 +( t k+y12)(1- h2(C ++A))- 2). (3.13)
The interference pattern is therefore a section of a zone plate pattern
(SCHMAHL and RUDOLPH [1969]). A grating made with one divergent and
one convergent beam is, therefore, a section of a zone plate at very large
values of t. We may, therefore, restrict the discussion to determine the
spacing a as a function of x for y g: x and z -jc x. In Fig. 2 for z = y = 0,
/3 = 0 it follows from eq. (3.13) that
x =(tAc)3 1 + ( i l r+ ...- ).
- - (3.14)
Because of the assumption that the grating is a section with large zone
numbers t can be replaced by t = to+n, with n = running number of the
grating; n = 0 represents the middle of the grating. With t = to +n and
expansion of the square root eq. (3.14) gives
In the expansions (3.14) and eq. (3.15) higher powers of A/toc are neglected.
The change in grating spacing as a function of n follows from eq. (3.15)
within an accuracy better than 10% for all practical cases:
da d2x
-
dn dn2
= --&A,)* * to%+ . . .. (3.16)
A comparison of eq. (3.16) with eq. (3.8) shows, that 6, # 0; that means
in the elliptic case that the grating has a linear progressive ruling error. In
Table 3 numerical values for Ax (cf. eq. (3.7)) are given for gratings with
V, Q 31 BASIC PRINCIPLES 201
TABLE3
Values of Ax calculated for gratings in the hyperbolic and elliptic cases.
100 1
500 0.3
1000 0.1
2000 J 0.05
PI
P
2a 3Yv
lengths and for gratings wiih a low line density for use as scales. To enhance
the ruling accuracy of such gratings one can use the following proce-
dure (SCHMAHLand RUDOLPH[1970]): Each wavefront affected with
aberrations can be assumed to consist of parts of wavefronts with different
v, 8 31 BASIC PRINCIPLES 209
using the beams R3 and R2. Illuminating the hologram H12with R1 and
the hologram H3, with R; , the beam R2 is reconstructed twice, except for
amplitude factors. The superposition of the two beams R2 yields the grating
H22. In a model experiment holograms H12 and H3, were made with the
beams R1, R2 and R 3 . The wavefronts of all three beams had aberrations
of several wavelengths. These aberrations were introduced by using bad
optical elements. IgFig. 7a and Fig. 7b Moirb patterns from the holograms
H,2 and H32 are shown. The Moire patterns were made by illuminating
the holograms with an interference fringe system made with two plane
wavefronts having aberrations 5 A/lO. The Moire patterns demonstrate
that these holograms- acccrding to the fact that they are made with distorted
wavefronts - have large ruling errors. From these two holograms two iden-
tical wavefronts R2 were reconstructed according to Fig. 6. Superposition
of these two wavefronts yielded the hologram H22. Fig. 7c shows the
Moire pattern of this hologram which demonstrates the high ruling accuracy
210 HOLOGRAPHIC DIFFRACTION GRATINGS
,
Fig. 7a. Moire pattern of the hologram H I .
from blank 7 both the exposed resist layers of blanks 5 and 7 are developed
and recoated. As in Fig. 8c the holograms on blanks 5 and 7 reconstruct
two beams with identical wavefronts which are superposed. The resulting
interference pattern with highly improved ruling accuracy is recorded in
a resist layer on the blank 12.
Fig. 8a.
a2
\ .
,L
BASIC PRINCIPLES 213
Fig. 8a-c. Optical arrangements for producing holographic gratings with improved accuracy by
the use of identical reconstructed wavefronts.
I
0 50 100 150 200 250 300 350
E [rnJ/ccm2]
Fig. 9. Characteristic curve for the positive working photoresist Shipley A 2 1350.
216 HOLOGRAPHIC DIFFRACTION GRATINGS CV-0 4
Fig. 10. Scanning electron micrograph of a grating with sinusoidal groove profiles.
W
d
M
.-N
s
U k
C
3
C
.- 'E
W
.-
U
k
4
4
L k
-
X
a
4
a
0
0
cl
0
s
M
iz
-
m
N
M
ii;
V, P 41 PRODUCTION O F H O L O G R A P H I C G R A T I N G S 219
'1
50 a' /
50
c
0
1
1
I
2 y
>
Imml
Fig. 13. Comparison of the uniformity o f a holographically and a classically made scale.
gratings with sinusoidal groove profiles have high efficiency values only
if the grating spacing is comparable to the wavelengths used, as will be
shown in section 5.3. For gratings with spacings large compared with
the wavelength, high efficienccy values can only be obtained by using saw-
tooth groove profiles. There are several methods of producing holographic
gratings with such profiles. The first method is to produce gratings with
a triangular profile by inclining the grating blank to the direction of the
7 \ Photoresist
h-
r
fringes
Blank
\ - v
development
Fig. 14. Method to make holographic gratings with sawtooth groove profiles. Refraction in
the resist layer and the blank has not been taken into account.
Fig. 15. Scanning electron micrograph of a grating with sawtooth groove profiles. made by
the method illustrated in Fig. 14. (Reproduced from HUTLEY[1974a] Fig. 2.)
The first two terms with an appropriate phase relation already yield a
good approximation to an ideal sawtooth profile. In practise it is, however,
very difficult to superpose two fringe systems with the required accuracy
when using different optical arrangements for the different fringe systems.
For a 100 mm grating with, e.g., 600 grooves/mm the accuracy of the adjust-
ment of the mirrors has to be better than 0.1 second of arc. To overcome
this difficulty one can use the following method. With a symmetric arrange-
222 HOLOGRAPHIC DIFFRACTION GRATINGS
I
2
P s i n d Z =2 sin Ul
Fig. 16. Arrangement for making holographic gratings with sawtooth groove profiles.
Fig. 17. Scanning electron micrograph of a grating with sawtooth groove profiles made with
the arrangement illustrated in Fig. 16, (I = 1.67pm.
V. P 51 C O M P A R I S O N WITH C L A S S I C A L G R A T I N G S 223
and 1 of the grating g form a fringe system at b, with the spacing a, whereas
+
the 2 and - 2 orders diffracted respectively by the parts 2 and 2 form
a fringe system with the spacing a/2. The proper phase relation between the
two fringe systems can be obtained, for example, by using a plane-parallel
plate p in a part of the parallel wavefront or by evaporating a step onto
the section 2 of the grating g. The method also works by replacing the
grating g of Fig. 16 by two identical gratings g, and g, made either on one
blank polished to optical tolerances or on two separate blanks. The latter
arrangement has the advantage, that the gratings g, and g, have to be only
about twice as large as the grating b. The asymmetric profiles shown in
Fig. 17 were made by successive exposure of a photoresist layer arranged
at b with the two fringe systems according to Fig. 16. For the shown profiles
the measured ratios of the intensity in the first positive to the first negative
order was 17 at the blaze wavelength. One advantage of this method is,
that it is possible to produce sawtooth profiles not only on flat but also
on curved blanks, e.g. it is possible to produce concave gratings with saw-
tooth profiles and uniform blaze over the whole area.
Finally it should be mentioned that it is possible to make asymmetric
groove profiles by copying holographic gratings using visible light, UV or
X-rays and by using ion etching processes.
Since the first holographic gratings for serious spectroscopic use were
made in 1967 in the Optical Laboratory of the Observatory of the University
of Gottingen such gratings have been competitive with classically ruled
gratings from the visible to the ultraviolet and soft X-ray regions. Today
it is possible to realise large gratings with a width of more than 600 mm,
to reach line densities of more than 10000 lines per millimeter, to make
gratings with a very low amount of straylight and completely free of ghosts.
It is possible to obtain good efficiency values for gratings with symmetrical
groove profiles and high line densities. In addition it is possible to attain
high efficiency values independent of the line density by making asymmetric
groove profiles holographically. Holographic gratings can be formed
independently of substrate curvature and, in principle with any desired
surface variation of grating frequency. Furthermore replicas of holographic
gratings can also be made.
224 H O L O G RAP H I C DIFFRACTION G R A T I N G S cv, 8 5
5.1. WAVE-FRONT INTERFEROGRAM, RESOLUTION AND INSTRUMENTAL
PROFILE OF PLANE GRATINGS
One method of obtaining information about the optical quality of gratings
is to examine wave-front interferograms. The spacing between two maxima
or two minima in an interferogram is called one fringe and corresponds
to a wa:le-front aberration d(q) of one wavelength of the light used to
make the interferogram. According to eq. (3.5) a wavefront aberration
d(q) of p fringes of a grating measured in autocollimation corresponds to
-
a ruling error Ax = p ,442 sin .)'i Normally wavefront interferograms are
made with large Michelson interferometers.
" T
Fig. 19. Wavefront interferogram of a grating with 1500 grooves/mm, made with the
arrangement illustrated in Fig. 18.
in the first order with 1 = 457.9 nm. The interferogram shows that the
wavefront aberrations are smaller than 1/10. From the wavefront inter-
ferograms one can deduce that the instrumental profiles are symmetrical
and that the spectral resolution reaches the theoretical values. Fig. 20a and
Fig. 20b show photoelectric recordings of hyperfine structure of the mercury
lines 435.8 nm and 546.1 nm in the second order in autocollimation, made
with a grating having ruled width 180 mm and 1465 grooves/mm, made
in the Gottingen Laboratory. The measurements were made with an
8-meter spectrograph in the Solar Tower of the Gottingen Observatory.
The theoretical resolving power of the grating is 527000 in the second
order. The measurements were made with 94% of the grating surface
- because of vignetting by the autocollimation lens of the spectrometer -
and with an entrance slit width of WE,= 20 pm, an exit slit width of
W,, = 10 pm and an uncooled gas discharge lamp, corresponding to
W , x 0.4 nm.
With W& = W&,,+ W&+ W&+ W t one obtains for the observed
width of a single isotope component at half maximum Webs = 1.2 & 0.1 nm
for the wavelength 435.8 nm.
This value is in full accordance with the measured widths of the single
226 HOLOGRAPHIC DIFFRACTION GRATINGS [v, 8 5
1 Hg4358 m = 2
180 x 130mm2.
Hg 5461 m = 2
180 x 130md
1465 llrnm 1465 llmm
h
Fig. 20a. Hyperfine structure of the mercury Fig. 20b. Hyperfine structure of the mercury
line 435.8 nm. line 546.1 nm.
components of Fig. 20a and means that the grating has full theoretical
resolving power.
The best way to measure the scattered light is to test a grating in the
same arrangement as used for spectral measurements (HUTLEY[19731).
Nevertheless, to give an idea of the amount of scattered light from a high
quality holographic grating we compared the scattered light from a grating
with that scattered from a mirror by measuring both under the same condi-
tions in a spectrometer.
Fig. 21 shows the results of our measurements and demonstrate that
the amount of light scattered from our gratings is comparable to that
scattered from good mirrors and is not caused by ruling errors. These
results are in good agreement with measurements of other authors (HUTLEY
[1974], HUNTER[1975j, PIEUCHARD and FLAMAND [1975]). The low level
of scattered light from holographic gratings is especially important in the
cases of Raman spectroscopy and high resolution stellar absorption
loo 'I
lo-'
10-2..
lo* "
1cP- *.
Fig. 21. Comparison of scattered light of a holographic grating and of a mirror, both coated
with aluminium.
5890 5896
--A [A1
Fig. 22. Solar spectrum near the sodium resonance lines.
228 HOLOGRAPHIC DIFFRACTION GRATINGS cv, 5 5
spectroscopy (SCHMAHL and RUDOLPH[1972]). Fig. 22 shows a photo-
electric record. of the solar spectrum near the sodium resonance lines, taken
in the middle of the solar disc. The NaD, line shows a residual intensity
of only 6 % without any rectification. This test was made with a spectral
range of about 200 nm entering the spectrometer. Up to now comparable
low residual intensities of these lines could be obtained with conventional
gratings only with a strongly reduced spectral range and/or double pass
systems.
Fig. 23 shows a comparison of a conventional and a holographic concave
grating in grazing incidence. Both had a radius of curvature of about 2
meters. The conventional grating had 1200grooves/mm and the holographic
grating had 1800 grooves/mm. Though one has to bear in mind that an
exact comparison would require equal grating spacings for both gratings,
the results indicate that the signal-to-noise ratio is much better in the case
of the holographic grating (HUNTER[1975]).
INTENSITY INTENSITY
loo{ loo{
90-
80- 181.987 Kr
WAVELENGTH X __c
5.3. EFFICIENCY
90 - /-\-
80 - / \
<
'.
70-
\
60-
50 -
LO- I \
30- / \
\
\
20- I \
\
Fig. 24a, b. Comparison of the efficiency values of a classical (a) and a holographic (b) grating
with sinusoidal groove profiles. Both gratings have 1800 grooves/mm. (Reproduced from
FLAMAND [1975] Fig. 3.)
230 HOLOGRAPHIC DIFFRACTION GRATINGS v, 5
10
'' 300 LOO 5;)O 6W 7M) 800 &m]
Fig. 24c. Efficiency of a holographic grating with sinusoidal groove profiles and 1800
grooves/mm. (Reproduced from HUTLEY[1974b] Fig. 6.)
Relative Relative
Fnicimcy efficiency
"/-I
/ -.
1260 llmn
E,
b],,'
70
- ,/-\, , //
r--.
, .. 1580 I/mm
50
40- I \
i: -
20
10.
fi *
Fig. 25. Efficiency values of holographic gratings with sinusoidal groove profiles and with
1260, 1580,2090 and 3600 grooves/mm.
v, 51 C O M P A R I S O N WITH C L A S S I C A L G R A T I N G S 231
Therefore one has to avoid regions which show strong anomalies, i.e.,
rapid variations of efficiency over a comparatively short range of wave-
length. Fig. 25a to Fig. 25c show the efficiency in the visible region of three
holographic gratings with 1260, 1580 and 2090 grooves/mm and sym-
metrical groove profiles, measured in an arrangement near autocollimation
in the first order (MIKELSKIS C19731). Fig. 25d shows the efficiency of a
grating with symmetrical'groove profiles and 3600 grooves/mm, measured
in the ultraviolet in the first order. The curves demonstrate that the peak
efficiency values occur in the range 0.8 5 ,?/a ,< 1, which means that the
peak efficiency shifts to shorter wavelengths with increasing line densities.
Measurements of the efficiency as function of the angle 6 between incident
and diffracted beam show that the peak efficiency shifts to longer wave-
lengths with increasing 6 relative to measurements in autocollimation
-(6 = 0). For example, the peak efficiency is shifted by about 80 nm for
6 = 34" for a grating with 1560 grooves/mm with respect to the position
for peak efficiency measured for 6 = 0. These results are in contradiction
to calculations of the peak efficiency with increasing S made under the
assumption that the blaze of classical gratings with sawtooth profiles
results from the fact that the facets of the grooves act as small mirrors which
reflect the light in the same direction as the grating is sending it by diffrac-
tion.
Fig. 25 shows that when J./a < 0.8 strong anomalies occur, in this case
for gratings used in the first order in autocollimation. Such anomalies are
characteristics of all gratings, irrespective of whether they are made
mechanically or holographically. The anomalies were first mentioned by
Wood. A general theoretical treatment shows that the Wood anomalies
are actually of two distinct types, a resonance type and a form first mentioned
by Rayleigh, which appears at wavelengths due to the emergence or reentry
of another spectral order at the grating surface (HESSEL and OLINER [1965]).
In special cases these two types are merged together. Anomalies are
connected with plasma waves in the electron gas in the metal coating of
the gratings. Such plasma oscillations are known as surface plasmons
(TENGand STERN[1967], HUTLEY [1973]). Anomalies are well suited to test
electromagnetic grating theories, which should reproduce the observed
anomalies with regard to strength and location in the spectrum.
When A/a < 0.8 one can obtain high efficiency values only with asym-
metric sawtooth groove profiles. As shown in section 4.2 such profiles can
be made holographically. Efficiency values of various gratings with saw-
tooth profiles made by HUTLEY [1974] according to the method of SHERIDON
are given in Fig. 26 and Fig. 27. Fig. 26 shows the efficiency of different
232 HOLOGRAPHIC DIFFRACTION GRATINGS cv, 8 5
D
Mirror
I I I I
Kx) 200 300 100 200 300
Wavelength trim)
Fig. 26. Efficiency values of holographic gratings with sawtooth groove profiles. A) 1200
grooves/mm, B) 1570 grooves/mm, C) 600 grooves/mm, D) 800 grooves/mm. (Reproduced
from HUTLEY[1974a] Fig. 5.)
As discussed in section 4.1 gratings for the soft X-ray region can be made
holographically in metal on glass form without any remaining organic
material. It has been found possible in this way to control groove profile
accurately and so produce square wave (laminar) grooves with heights of
C O M P A R I S O N WITH C L A S S I C A L G R A T I N G S 233
Spectral order
Fig. 27. Light distribution among the various diffracted orders of grating.A) of Fig. 26.
Relative efficiency (full line) and absolute efficiency. (Reproduced from HUTLEY[1974a1
Fig. 7.)
Fig. 28. Absolute efficiency of a 2 meter radius concave grating with 294grooves/mm at 4.5 mm
in the _+ 1 orders (right hand scale). The Lero order (left hand scale) shows modification due to
phase cancellation. (Reproduced from JOHNSON [1975].)
Fig. 30. Photograph of a conventional grating of ruled area 6.5 x 4 em2 and a holographic
grating of ruled area 8.6x26.8 cm'. (Reproduced from HUNTERf19751 Fig. 10.)
grating of ruled area 6.5 x 4 cm2 are shown. Both have 2400 grooves/mm
and a radius of curvature of 85 cm.In Fig. 31 and Fig. 32 a comparison
with the gratings shown in Fig. 30 is made (HUNTER[1975]). This com-
parison demonstrates that the efficiency values are comparable, but
also that the efficiency of the holographic grating is much more uniform
in spite of the larger aperture. Fig. 33 shows another comparison of a
holographic and a classical grating in the XUV (HUNTER[1975]) and
demonstrates that especially in classically ruled tri-partite gratings large
variations of the efficiency over the grating surface occur whereas holo-
graphic gratings of the samef-ratios are quite uniform.
In addition to these classical concave gratings special types of holo-
COMPARISON W I T H CLASSICAL G R A T I N G S 237
1500 2000
WAVELENGTH (1)
MEASURED EFFICIENCY OF A HOLOGRAPHIC GRATING AT 15' ANGLE
OF INCIDENCE. POS. FIRST ORDER 0, ZERO ORDER NEG. FIRST
Fig. 31. Comparisons of the gratings of Fig. 30. a) Efficiency of the holographic grating.
b) Efficiency of the classical grating. Zero ( x ), positive (A), and negative ( 0 )first orders at
15" angle of incidence. The dotted line represents the negative first order groove efficiency.
(Reproduced from HUNTER [1975] Fig. 13 and Fig. 12.)
238 HOLOGRAPHIC DIFFRACTION GRATINGS CV, 8 5
40-
-
+I 0 -I
7- -I- BEAM
WIDTH
0 J
60 -
50 -
-
s40 -
+I
t
0
z
E30-
-
LL
LL
u20- +BEAM
WIDTH
'0 O - A
Fig. 32. Comparison of the gratings of Fig. 30. a) Holographic grating. b) Conventional
grating. Efficiency maps of the zero and the positive and negative first orders at 144 nm.
The angle of incidence is 15". The two large peaks on either side of the zero order of the
conventional grating are caused by specular reflections from the unruled edges. (Reproduced
from HUNTER[1975] Fig. 14 and Fig. 11.)
graphic concave gratings with reduced aberrations have been made. The
first proposals and results in this field of special concave gratings were
given by CORDELLE, FLAMAND, PIEUCHARD and LABEYRIE [19691. Further
investigations and special designs of gratings with imaging properties for
the visible, UV and soft X-ray regions have been made by several groups
(e.g., POUEY[1975], SPEER,TURNER, JOHNSON, RUDOLPH and SCHMAHL
[1974], NODA,NAMIOKA and SEYA[1947b], NIEMANN, RUDOLPH and
SCHMAHL [19741).
COMPARISON W I T H CLASSICAL G R A T I N G S 239
HOLOGRAPHIC GRATING
I - POSITIVE ORDERS 1
NEGATIVE ORDERS
0
0 5 10 crn
GRATING SURFACE
Fig. 33a. Efficiency maps of a holographic grating at 121.6 nm in the zero order and positive
and negative first and second orders. 1200 grooves/mm, gold coating, and 1 m radius of
curvature. (Reproduced from HUNTER[I9751 Fig. 8.)
We will now discuss some special types of gratings and their imaging
properties.
1. Gratings on a spherical blank recorded with two spherical wavefronts :
By choice of suitable construction parameters it is possible to construct
concave gratings for Rowland circle geometry, with highly reduced aberra-
tions. Especially it is possible to completely avoid astigmatism and some
types of coma and spherical aberration for one particular wavelength. As
a rule of thumb one can say that the astigmatism can be reduced over a large
240 HOLOGRAPHIC DIFFRACTION GRATINGS cv. 5 5
Fig. 34. Principle of holographic concave gratings with non-equidistant spacings and three
stigmatic points.
TABLE
4
Wavelengths and positions of stigmatic images, according to Fig. 34, for the first spectral
order, I , = recording wavelength
Image location
Source location
0 A A
0 1=0 3, = I , I = mi,
A I = A, I = 21, a = (m+i)n,
A I =ml, I = (m + 1)I, I = 2mI,
using waves originating from A' and A, which yields a similar kind of
matrix as shown in Table 4 (FLAMAND [1975]). The line H in Fig. 34 is
the horizontal or sagittal focus. The curved line V is the vertical or tangen-
tial focus, i.e. - at least for smallf-ratios of the grating - V is the focus of
the spectral lines. But it must be stressed that for large apertures especially
severe aberrations occur between the three stigmatic points. These aberra-
tions must be calculated for every particular arrangement.
2. The holographic method allows the construction of gratings with
imaging properties on extremely curved substrates. One example has
been discussed in section 5.4.
3. Up to now examples where the recording waves were either plane or
spherical have only been discussed. Another possiblity is - as already
mentioned - the use of aspherical waves for the recording process, to
correct aberrations of gratings with imaging properties. A special case of
diffraction gratings are zone plates, i.e., circular gratings with radially
increasing line density. Zone plates with large zone numbers can be realised
holographically by superposition of two spherical waves or one spherical
wave and one plane wave. Such high power zone plates can be used for
imaging and/or spectrometric purposes in the soft X-ray region (X-ray
microscopy, X-ray astronomy, X-ray spectroscopy) (RUDOLPHand
SCHMAHL [1967], SCHMAHL and RUDOLPH[1969]). If such zone plates
are constructed with visible light and are used in the X-ray region large
spherical aberration occurs. This aberration has been corrected by use
of aspherical recording wavefronts ( R U ~ L P[H 19743, NIEMANN, RUDOLPH
and SCHMAHL [1974]). This method can, in principle, be applied to other
gratings with imaging properties.
242 HOLOGRAPHIC DIFFRACTION GRATINGS [v
Holographic plane and concave gratings with high groove densities and
sinusoidal groove profiles have only been widely used up to now. These
gratings have efficiency values comparable to those of classically ruled
gratings but have less scattered light and are completely free of ghosts.
Such gratings have, therefore, an improved signal to noise ratio. As dis-
cussed in section 4.2 it is also possible to make gratings with sawtooth
groove profiles holographically. These processes are more complicated
than the process of making gratings with sinusoidal groove profiles.
Holographic gratings with sawtooth profiles will, therefore, only be used
widely in the future when replica gratings in large series are available.
Furthermore, it can be expected that considerable progress will take place
in the field of gratings with imaging properties. Application of holographic
X-ray gratings is just in the beginning and it can be expected that it will be
possible to enhance the Ctendue of spectrometers in the grazing incidence
region. As shown in section 4.1 the holographic methods allow the produc-
tion of scales of high accuracy, especially by use of identical wavefronts,
as discussed in section 3.1.3. Although so far not used in practice, this
could be important in the future development of metrology.
Acknowledgements
References
BEUTLER,H. G., 1945, J. Opt. SOC.Amer. 35, 31 I .
BURCH,J. M. and D. A. PALMER, 1961. Optica Acta 8. 73.
CLARK,K. G., 1973, Electronic Components June-September, 553.
CORDELLE, J., J. FLAMAND, G. PIEUCHARD and A. LABEYRIE, 1969, Aberration-Corrected
Concave Gratings Made Holographically, in : Optical Instruments and Techniques, ed.
J. Home Dickson (Oriel Press, 1970).
DOWLLY, M. W., 1971, Coherent Radiation, Technical Bull. Nr. 106.
FLAMAND. J.. 1975. Rev. Physique-Chimie. in press.
vl REFERENCES 243
VI
PHOTOEMISSION
BY
P. J. VERNIER
Laboratoire de PhoioelectriciiP,
Facult6 des Sciences (M.I.P.C.),
Universit6 de Dijon,
Dijon, France
CONTENTS
PAGE
9 1 . INTRODUCTION . . . . . . . . . . . . . . . . . . . 247
5 2 . THEORETICAL BASIS OF THE PHOTOEMISSION (PE) 250
6 3. EXPERIMENTAL DETERMINATION OF THE ESCAPE
DEPTH OF THE PHOTOELECTRONS . . . . . . . . . 279
9 4. SURFACE PHOTOEXCITATION . . . . . . . . . . . . 305
9 5 . CONCLUSION . . . . . . . . . . . . . . . . . . . . . 319
ACKNOWLEDGEMENT . . . . . . . . . . . . . . . . . . 321
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . 321
0 1. Introduction
During the first years after its discovery in 1887 by Hertz, photoemission
(PE) was investigated by experiments of Hallwachs, Elster and Geitel and
Lenard. The results of these experiments suggested to Einstein the concept
of photon and his well-known equation that has been verified later on by
Millikan. Early studies that led to high-yield photocathodes still in use
(Ago-Cs, SbCs,) are well exposed in classical monographs (HUGHESand
DUBRIDGE [1942], SUHRMANN and SIMON [1958], ZWORYKIN and RAMBERG
[19493).
The investigations have been pursued for nearly a century in order to
produce higher-yield photocathodes in a wider spectral range and with a
better reliability. SOMMER [1970], who played an important role in the
development of trialkaline photocathodes, gave a very complete review
of the experimental knowledge and of possible technical applications.
Quite recently a new role has been found for PE in electron spectros-
copy namely the investigation of electron energy levels and the transitions
between them. After the pioneering work of Spicer, nearly all elements
and many alloys and compounds have been measured. The spectral range
investigated has widened during recent years and now extends to X-rays.
Applications to chemical analysis gradually became reliable and are now
commercially exploited (ESCA). Several review papers on electron spectros-
copy have been published recently (GCJRLICH and SUMI[ 19703, N. V. SMITH
[1971]). (See also SHIRLEY [1971].)
In this paper we shall only be concerned with the origin of the photo-
electrons. We shall deal with different theories, techniques and materials.
For a systematic review of important points such as technical applications,
electron spectroscopy and its relations with band structure, work function
(for this last point see RIVIERE [1969]), etc. we will refer elsewhere.
It was established a long time ago that the part of a solid responsible for
PE is generally a very thin superficial layer. A precise knowledge of the
exact origin of the photoelectrons is fundamental both to improving the
247
248 PHOTOEMISSION CwS 1
Early theories dealt with metals as free electron gases and with PE as a
surface phenomenon (FOWLER [1931], MITCHELL [1934,1935,1936]). The
emission of an electron was the result of the simultaneous interaction of
an electron of the solid with the surface and with the electromagnetic field.
Later theoretical (FAN[1945]) and experimental studies led workers to
describe PE as a volume process and to divide it into 3 parts:
1) Absorption of one photon and transfer of its energy to one excited
electron.
2) Transport of the excited electron towards the surface.
3) Transport of the excited electron across the surface.
The contribution of a small volume 6V to the PE is then the product of
three factors :
the total number of absorbed photons,
the proportion of absorbed photons that lead to the excitation of one
electron,
the escape probability of an electron excited in 6V.
The first two factors describe step 1, the last factor describes steps 2 and 3.
Insofar as we can make 6 V tend to zero, the photoyield Y can be written as
n
where D(r) is the density of absorbed photons per incident photon (DAP)
at r and p(r) is the electron escape probability for an electron excited at r .
To investigate the photoelectron energy distribution (PED) it is useful to
define the probability for an absorbed photon of energy hv to excite an
+
electron to an energy between E and E dE, which we can write in a normal-
% 11 INTRODUCTlON 249
ized form as
(1.2)
PO
where
Po =
s/?(E,hv)dE.
We shall also define the probability for an electron excited at energy E'
at r to emerge with an energy between E and E+ dE, as p(E', E, r)dE. The
number of electrons emitted with an energy between E and E+ dE is then
and
s
n(E)dE = dE D(r)/?(E',hv)p(E, E', r) dE' d3r, (1.3)
s
Y = n(E)dE.
To obtain a more general validity for the expressions (1.1) and (1.3) we may
(1.4)
try to introduce into the DAP D(r) a surface term of the form A(z) where z
is the distance from r to the surface. A depends on the polarization and
on the angle of incidence of the light. Of course, no transport process
through the solid needs to be considered and the escape probability as-
sociated with the surface absorption ps is merely the ratio of the number
of emitted electrons to the total number of excited electrons. p, might not
be the limit of p(r) as z tend to zero and P(E, hv) can take different forms
8, and P, for surface and volume effects. Two objections can be made
against the 3-step model as it is described here.
or for semi-infinite solids i.e. slabs of infinite thickness. The DAP is then
a function of the distance z from the surface. We shall consider here the
DAP in slabs illuminated by a monochromatic parallel incident beam of
frequency v = 0427~that induces in the slab an electric field of complex
amplitude E(r) at the point r . We shall disregard the magnetic absorption
that is not important in PE. The response of the solid to the electric field is
then described by the complex amplitude either of the density of electric
current j(r) or of the electric displacement R(r). These quantities are
related by
j(r) = ioR(r), (2.1)
if we include in the density of current the displacement current. Within the
linear approximation, the most general relation between E(r) and R(r) has
the form:
Re j(r)&*(r) - Im R(r)l*(r)
D(r) = -
2Nho 2Nh
where N is the number of incident photon per unit time. As defined by (2.3)
the DAP may have a microscopic significance and can be used to define
the absorption in a volume 6V arbitrarily small. But in this paper we shall
not make use of this possibility.
In practice screening effect make E(r, r) tend to zero when lr-rl tends
to infinity. To define the DAP in a volume 6V large compared with the
size of the crystalline cell we may use a local dielectric constant that is
generally complex,
where i, is the angle of incidence of the incident light beam, I the wave-
length in vacuo, and &, the complex amplitude of the electric field in the
incident beam.
The smallest volume where (2.8) can be applied is one crystalline cell,
but this is only a very rough approximation if it is not embedded in a set
of identical cells. This restriction is still more important if we wish to go
beyond the local approximation by introducing the wave-vector-dependent
dielectric constant E ( q , o).If the gradient of the componentspf q r ) is large
enough, so that 4 r ) is significantly different from &(r)for values of r - r,
which do not cancel E(r, r), the expressions (2.5) and (2.8) cannot be
substituted for (2.2) and (2.3). This occurs when the screening length in
the solid is not much greater than the wavelength of the light. E ( q , o)can
then be defined as the ratio of the Fourier transforms of R(r) and &(r)in
an infinite solid. This adds new restrictions to the use of E ( q , o)in the
calculation of the dielectric (mean) response and the DAP in a finite volume
SV, if 6V is near the surface.
Even in an isotropic or cubic material, the isotropy of the relations
between b(r) and R(r) is destroyed by the definition of the light wave vector
q. E ( q , o)then becomes a tensor. In the simple case when q is real and the
material is isotropic, we have a cylindrical symmetry and we can define
scalar longitudinal and transverse dielectric constants EL(q, o)and +(q, o)
that are the components of the dielectric tensor in a particular reference sys-
tem (STERN [19631). We may note that the cylindrical symmetry is generally
broken when q is complex (heterogeneous wave).
(IW12
)
__ (2.11)
where (IS(r)12) is the mean value of the squared modulus of the complex
amplitude B ( r ) in the volume 6V for which gZolis defined. When the vector
potential of the electromagnetic field has the complex amplitude A(r), the
transition probability O k k , from the state $k to the state $ k , is
2n
Okk, = lJ&kk?16(Ek, - E k - ho). (2.12)
The matrix element is
=-($kl
m
( A ( r ) .P 3 - P * A(r))I$k)3 (2.13)
The Fresnel equations permit the calculation of the electric field excited
by an incident plane wave in a solid. At the interface of two semi-infinite
homogeneous media, characterized by indexes of refraction no and n, the
reflection coefficient is
no cos i , -n, cos i,
r, = (2.16)
no cos i, + n , cos i,
for s polarization, and
n1 cos i, -no cos i,
r = (2.17)
n , cos i , +no cos i,
(2.19)
l+r,
1 + r, r: exp ( - 2iq,z,)
+
x {exp (- iq, z ) r: exp (iq,[z - 2z,]))
where rs and rl are the coefficient of reflection for the electric field, given
by (2.16) when the incident light beam encounters respectively the first and
the second face of the film and
2714 cos i,
4, = (2.21)
A
is the z component of the wave vector q in the photoemitter. Because of
absorption q, is always complex. q, that appears in (2.18), is the imaginary
part of 1qZ/2n.For p polarization, because the angle of refraction i, in the
photoemitter is complex, we must add separately the energy densities
Dp,(z) and D,,(z) associated with the tangential (x) and normal ( z ) com-
ponents of the electric field:
+
(1 ~,)(cosi,/cos io)
+
1 rprb exp ( - 2iq, z,)
/j
x sin i, x { exp ( - iq, z ) - rb exp (iq,(z - 22,))) ; (2.22)
rp and rl, have for p polarization the same significance as rs and ri for s
polarization.
The calculations of the photoyield by COQUET and VERNIER[1966] and
by many other authors since that time have been based on equations (2.20)
and (2.22).
in the case of spatial dispersion and the CTCF associated with the dis-
persion equation (2.23) are insufficient to determine the amplitude of
several transmitted waves. HOPFIELD and THOMAS [19631 and many later
authors introduced an additional boundary condition by examining the
microscopic process that is responsible for the spatial dispersion e.g.
Hopfield and Thomas cancel the contribution of the exciton to the electric
polarization at the surface. MELNYK and HARRISON [1970] introduced an
additional boundary condition, namely the continuity of the normal com-
ponent of the electric field, to account for the excitation of longitudinal
plasma oscillations in metals in addition to the standard transverse waves.
For p polarization at high angles of incidence, they predicted oscillatory
variations of the transmittance and absorptance of thin films of K with
photon energy above the plasmon resonance. ANDEREGG, FEUERBACHER
and FITTON [I9711 observed such variations in the spectral distribution of
the photoyield.
The justification of the additional boundary conditions raised a good
deal of controversy. AGARWAL, PATTANAYAK and WOLF[1971a, b, 19741
substituted for the additional boundary condition a coupling between the
transmitted waves at the surface.
VERNIER [1973] noted that the use of boundary conditions implies that
the transition layer, where the bulk index of refraction of the homogeneous
medium does not represent the solid properties, has a negligible thickness.
We may ask whether the standard CTCF can be applied across this layer,
when the local approximation breaks down. The rigorous solution is to
solve the coupled equations that relate the motion of the charges and the
electric field in the vicinity of the surface. Such a solution has been given
for a free electron gas by SAUTER [1967], FORSTMANN [1967], and FUCHS
and KLIEWER [1969]. FUCHS and KLIEWER [1969] calculated the imped-
ances 2, and 2, that appears in the standard expressions of the reflectance
cos io-z,
for p polarization; R , = (2.24)
cos i, z,+
1-2, cos i,
for s polarization; R, = (2.25)
1+z, cos i,
These expressions would be equivalent to Fresnel equations (2.16) and
(2.17) if Z , and Z, could be deduced from a single dielectric constant E for
every angle of incidence by
(2.26)
260 PHOTOEMISSION [VI, Ei 2
Z, = ( - -sinio , (2.27)
This requirement is, however, not fulfilled. Moreover Fuchs and Kliewer
found a surface absorption term. We note here that such surface absorp-
tion implies that the normal component of the Poynting vector is not con-
tinuous at the surface. Therefore the CTCF is no longer valid. The process
of surface absorption considered by Kliewer and Fuchs does not lead to
one-electron excitation and therefore to PE. Later on we shall note other
processes of surface absorption that lead to PE and that are also inconsistent
with the CTCF and, therefore, with the Fresnel equations.
(2.29)
is replaced by
A A =
he
~ <Xk(r)uk,(r)eik'.r(
~ ~ -
- b(r.) (V,Xk(r))luk(r)eik").
mo
(2.31)
We assumed here that the gradient of the electron wave function is much
steeper than the gradient of the electric field. This assumption might not
be valid if the screening length is very short ; to calculate the matrix element
the expression (2.13) should then be used instead of (2.14).
SCHAICH and ASCHCROET [19711 noted that the surface contribution
262 PHOTOEMISSION [VI, 2
bending effects (ALLEN and GOBELI [1962]). But surface states have been
theoretically predicted at other energy levels in the band scheme both in
semi-conductors and metals (BORTOLANI, CALANDRA and KELLY[19731,
APPELBAUM and HAMANNC19731, GURMAN and PENDRY C19731, HEINE
[19721). We are concerned here with PE from surface states. It is generally
difficult to discriminate between PE from intrinsic surface states and other
levels that occupy the same position in the band scheme. No polarization
test is available and the best test is the dependence of the observed
phenomena on an exposure to very small amounts of gases.
+
t I
5 'I) E
Fig. 1. Mean free path of an electron of energy E above the Fermi level with respect to
electron-electron interaction in gold, calculated by KROLIKOWSKIand SPICER[19691 (con-
tinuous curve), and by SZE,MOLLand SUGANO[1964] (dotted curve). Experimental values of
the escape depth for photons of energy hv = E, obtained by: CROWELL, HOWARM,LABATE
and SPITZER[1962], x SZE, MOLLand SUGANO[1964], 0 KATRICH and SARBEI[1961],
+ VERNIER,COQUETand BIGUEURE [1966], V PONG.SUMIDA and MOORE[1970], A MEAD
[1962] (non photoelectric method), * KANTER[1970] (non photoelectric method).
266 PHOTOEMISSION CVI, 2
if both final levels E,+AE and E-AE are vacant, i.e., lie above the Fermi
energy EF (random-k approximation). Therefore
1 ( E - Eo@
6,
ZEF-E
dAEp(E,)p(E, + AE)p(E- AE). (2.37)
z(E) CC dEo J E F - E .
Results of the calculation of I(E) by KROLIKOWSKI and SPICER
[ 19691 using
(2.37) are plotted together with experimental data (see Q 3 or Fig. 1 for
gold). KANE[1967] developed the random approximation in the case of
A
x-
G-
-5a-
r
&
,a-
#
$6-
4-
2-
I I I I I
VI, 21 THEORETICAL BASIS OF T H E PE 261
(2.38)
(2.39)
The z components of the wave vector inside (ki) and outside (k;) the solid
are related to the work function W, by
We may try to extend the expressions (2.41) and (2.42) to the case of
Bloch electrons; 8' then represents the angle of incidence of the group
velocity and W, the energy of the electron above the bottom of the conduc-
tion band :
W, = E - E , (2.43)
W, = E , - E , . (2.44)
(2.46)
At least in metals the limit angle 6 , defined by eqs. (2.41), (2.43) and (2.44)
is small for small photon energies and we may write,
(2.47)
(2.48)
mean free path for an electron excited from the Fermi level. p o depends
only on the photon energy and can be written as
(2.50)
with
"
f(6,, a, E ) = - 1-cos 8,-
2
___
al(E)
In ( +al(E)
1 cos 6 , )]. (2.53)
Here 6 , depends on the energy Eand is given by (2.41). To apply the expres-
sion (2.53) KROLIKOWSKI and SPICER use the expression (2.37).
The ballistic approximation breaks down when secondary electrons keep
enough energy to get out of the solid. KANE[1967] performed a numerical
calculation of the energy distribution of secondary electrons in Si, after
scattering by pair creation. SMITHand SPICER [19691 attributed a structure
observed in the energy distribution of photoelectrons excited in alkali
metals by 10.2 eV photons to electrons that have lost energy by plasmon
excitation.
212 PHOTOEMISSION [w,4 2
9 d2f ~
f = POD(Z).
+ zh
- (2.54)
dz2
Here fi0D(z) is the density of electrons excited by unit time and 9 is the
diffusion coefficient.
The density of photoelectric current is given by the current at z = 0,
df
I = -9e-. (2.55)
dz
The surface properties of the cathode define boundary conditions that we
can describe by an escape probability p o . The surface treatment necessary
to obtain the NEA is the deposition of an activating film in which inelastic
scattering occurs and reduces p o .
In a semi-infinite solid we can deduce from (2.54) and (2.55) that
y = [
.o
m
Po exp (-Z/L)PoD(Z)dZ, (2.56)
Y =
s + +
[ A exp (- z/Ld) B exp ( z/Ld)]PoD(z) dz. (2.58)
VL 5 21 T H E O R E T I C A L B A S I S OF T H E P E 213
Here A and Bare constants which can be deduced from boundary conditions
at the interfaces of the film. The quantity in brackets can be identified with
the escape probability p ( r ) in the expression (1.1).
We can consider that the terms A exp (-z/Ld) and Bexp ( + z/Ld) result
from successive back and forth diffusions in the film with an escape prob-
ability p o at the surface and an absorption probability p A at the substrate
interface and we can write
Erfc (y) = -
>J, exp (- x) dx, (2.61)
(2.62)
L = 2;. (2.63)
214 PHOTOEMISSION
\ Erfc (X\
Fig. 3. Comparison of the functions Erfc (x) -and exp (-2x) -----
(2.64)
with the expressions of KANE[1962]. Insofar as the mean free path I@)
is a slowly varying function of the electrons energy E, the pair production
scattering determines the overa!l number of emitted electrons but does not
affect the form of the spectral yield nor the PED near the threshold. For
a non negligible energy loss Ephin phonon scattering, he found that if
the final states at threshold are not at an extremum in the conduction band
then
x (hv-hv,)3; (2.65)
The expressions are the same for direct and indirect transitions, only the
value of the threshold hv, is different. The expression (2.65) gave quite
good results for semi-conductors.
In the most general case a simple explicit form of the escape probability
cannot be given. STUART,WOOTENand SPICER[I9641 and STUARTand
WOOTEN[19671 performed numerical calculations for various values of
Eph,1, and lp. The results roughly agree with simple models.
of P(E, v). The transition probabilities given by the expressions (2.1 l),
(2.12), (2.13) have been derived within the one-electron approximation.
The k conservation selection rule in the volume effect is a direct consequence
of (2.13). Such a treatment may not be sufficient and we need to introduce
also the many-body effects in the calculation of the transition probability
6 j ) k k . . The indirect transitions with phonon creation may be considered
as a transition between two Bloch states of different wave vectors that is
allowed by a subsequent scattering of the electron in the final state by the
lattice. In a more general way, we may associate with each scattering
process of the excited electron a perturbation of the transition probability
O k k . .The perturbation may be an increase or a decrease only when the
transition probability is not zero without scattering e.g., for direct transi-
tions. In most cases the perturbation merely adds a new term to the absorp-
tion. Indirect transitions with phonon creation have a quite low probability
in comparison to direct transitions and have been observed by optical
means only in the spectral range where no direct transition can occur.
Because the mean free path of the electrons for the electron interaction is
often much shorter than for the phonon interaction we may expect in the
first case a much stronger perturbation that, quite strangely, is seldom
taken into account.
A semi-empirical way of taking into account the dependence of photon
absorption on the scattering processes is to introduce damping in the
Bloch waves that are used to calculated matrix element (2.13). We may
attribute to the initial state $k a quasi-infinite lifetime and a finite lifetime
z to the final state. The theory of Weisskopf and Wigner, quoted by DAVY-
DOFF [1965], when adapted to our notations gives, instead of (2.12),
(2.67)
The expression (2.67) can be very useful for comparing the results of optical
or PE measurements with the information deduced from escape depth
measurements. It must be applied with great caution when several scattering
events follow one another; z must be taken as some sort of an empirical
parameter that represents the time necessary for a complete thermalization
of the absorbed photon energy.
In the dynamical theory of electron diffraction (DEDERICHS [19721)
damping of the electron wave is often introduced by substituting for the
classical Bloch wave (2.9) the expression
(2.69)
(2.71)
several authors as artificial. For that reason, several formalisms have been
derived to treat PE as a one-step process.
MAKINSON [19491developed a theory of surface PE as a one-step process.
In order to include in the expression of the photo-yield a factor representing
the transmission of the electrons, Makinson introduced the coupling of
the electrons of the solid with the sets of electron waves obtained by
associating with each incoming wave the reflected and transmitted ones.
In more recent work the set of waves associated with each incoming wave
has been completed by the diffracted waves and its coupling with the
electrons of the solid has appeared as a basic element of the one-step
theories of PE.
ADAWI[1964] and MAHAN[1970] have introduced the methods of
scattering theory for the surface effect and the volume effect respectively.
They treated the PE as inelastic scattering of the electrons of the solid by
photons. Mahan used the asymptopic form of the Green functions to cal-
culate the photoelectric current dI emitted within a solid angle dQ. He
obtained the result
(2.72)
Mahan developed his calculation mainly for q5i representing free elec-
trons but the expression (2.71) is still valid when q5i represents Bloch elec-
trons. Let us note here that we may include in the double scattering term
of (2.72) double scattering by the electromagnetic field, and thus include
two-photon photoemission.
The same T-matrix elements appear in the Mahan theory of PE and in
the theory of Leed and Auger emission. In the latter cases the calculations
are much more advanced than in PE (see for instance TONG,RHODIN and
TAIT[1973]) and could be probably used in PE. In recent years, other one-
step formalisms have been developed by SUTTON[19701, SCHAICH and
ASCHCROFT [1970], THORNBER [1971], HERMEKING [1972,1973], TZOAR
and GERSTEN[1973], GERSTEN and TZOAR[1973], CAROLI,LEDERER-
ROZENBLATT, ROULETand SAINT-JAMES [19733, MEESSEN [19731. Based on
more rigorous principles than the 3-step model, they have been able to
predict new structures in the PE of X-rays (NOZIERES, DEDOMINICIS [19691).
But up to now no one-step theory of PE has been able to include in a proper
manner the effects of electron-electron scattering. In the papers of MAHAN
[19701and SCHAICH and ASCHCROFT [19703electron scattering is introduced
just like in the 3-step model. In any case the functions that appear in the
3-step model may be used at least as phenomenological parameters to
interpret the experimental data. We may hope that advances in the one-step
theory will afford a better link between the functions that can be deduced
from experimental data and the microscopic processes.
The integration of (3.1) when D(z) is given by (2.20) or (2.22) was performed
by COQUETand VERNIER[1966], VERNIER, GOUDONNET, CHABRIER and
CQRNAZ [1971], PEPPER [1970].
with its energy E is related to the density of states (ODS). We also have
seen (section 2.7.1) that the observed photoelectron energy distribution
(PED) depend on both 1(E) and ODS. Krolikowski and'spicer obtained
the energy dependence of the mean free path within a constant factor, by
fitting it with all PED data. They obtained the constant factor from one
absolute yield for hv = 8.6 eV by assuming the escape conditions described
in section 2.7.1.
GESELL and ARAKACVA [I9711 deduced the attenuation length of un-
scattered photoelectrons in A1 and Mg by two methods similar to that of
Smith et al. and that of Krolikowski et al. respectively. They also used the
variation of the yield with the angle of incidence of light (see section 3.5).
They found a steep decrease of the attenuation length from several hundreds
of Bngstrom for electrons of 5.6 eV above the Fermi level to a few Angstrom
for electrons of 10.2 eV. This decrease is much sharper than the one cal-
culated by the theory of RITCHIE and ASHLEY [1965].
Even if the scattered electrons are eliminated by energy analysis, the
models used to estimate the transmittance of the surface for the excited
electrons are very rough. Moreover, the correct calibration of a source of
light especially in the ultra-violet range is quite delicate, Therefore this
method of estimation of the escape depth is exposed to large systematic
errors and more redundant data give much safer results.
To obtain both p o and L we need two independent equations of type
(3.1), i.e.,we must illuminate the sample in two ways such that the associated
DAP are not proportional. Taking the ratio of the associated photoyield
eliminates po and the absolute calibration of the light source is no longer
necessary. This has most often been accomplished with several thin films
of different thicknesses or with one thin film illuminated through the
substrate (back illumination) and directly from vacuum front illumination.
With particular materials and in definite spectral ranges the special yield
distribution or the yield variation with the angle of incidence may give
sufficient data.
3.2. THE ESTIMATION OF THE ESCAPE DEPTH FROM THE VARIATION OF THE
RATIO OF FRONT Y+ TO BACK Y - YIELD VERSUS THE THICKNESS zo
OF THIN FILMS
An elementary analysis of the DAP gives for large values of zo and for
front illumination
///' /
' ""I "' / -' '00 I I I l
-
01 m
VI, 31 T H E ESCAPE D E P T H OF THE PHOT OE L E CT RONS 283
'h
I I
m 4
x
. zo(N 5
Fig. 5. Variation of the photoyield of a thin film of gold with its thickness, as calculated by
COQUET and VERNIER [1966] for 1 = 2804 A, i, = 65" and direct illumination polarization.
The circles represent the experimental data after VERNIER, COQUET and BIGUEURE [1966].
The photoyield of a film of infinite thickness has been taken as unity.
VI, P 31 T H E E S C A P E D E P T H OF T H E PHOTOELECTRONS 285
The fit between the theoretical and experimental curves that represent Y
versus film thickness is not always good. It cannot be good if the structure
or the work function of the film depends on its thickness (SHU'LMAN,
STRAKOVSKAYA and NEMCHENOK [1969], GAUDART [19733).
PONGet al. [1966,1967,1970,1972] extended the method into the far
ultra-violet. They measured the photoyield of films of increasing thick-
nesses with illumination through the LiF substrate. For insulators CuBr
(PONG[1966]), KBr (PONG[1967]) he first deposited a transparent film
of gold. Pong found L = 180 A in KBr, L = 30 10 A in PbTe and 51 A
286 PHOTOEMISSION [VL 5 3
for CuBr for 8 eV < hv c 11 eV. We may note that scattering by electron-
hole pair creation may be more important in CuBr than KBr where the
band gap is quite important. PONG,SUMIDAand Moo- [1970] found
L = 40+ 10 A for gold and L = 230 A for A1 in the spectral range 5.5-10.2
eV. The value for A1 is in good agreement with the theory of STUART and
WOOTEN [19673 if the mean free path for electron interaction is Z, = 500 A
and the mean free path for phonon interaction ,Z = 130 A. For gold the
value of L seems quite large when compared with the results of VERNIER,
COQUET and BIGUEURE [1966] and of PIERCEand SIEGMANN [1974].
3.4. ESTIMATION OF THE ESCAPE DEPTH FROM THE BACK AND FRONT
PHOTOYIELDS OF ONE THIN FILM
were consistent with one value of the thickness for each film. The index of
refraction did not depend on the film thickness zo when zo > 80 A. It can
therefore be assumed that the films were homogeneous and reproducible.
They deduced one value of the escape depth L for each angle of incidence
and for each polarization by comparing the observed value of Y + / Y -
with the result of calculation from the substitution of (2.20) or (2.22) into
(3.1). They have found L = 4Of. 10 A for every angle of incidence and
polarization in the photon energy range 4-5 eV. GOUDONNET,
TRUITARD and VERNIER[1973] applied the same method to silver in the
vicinity of the plasma resonance. For clean films the threshold is at hv, N 4
eV and all data in the range 4-4.5 eV are consistent with an escape depth
L = 40& 10 A (Fig. 7). When a silver film is actived by a submonolayer of
Cs, the escape depth is not changed for hv > 4 eV. The photoelectric
I I , I I , I
>
3; 3.6 3.7 i.0 39 4 42 43
!+I >,4 45
RKXCN E rml a / ]
Fig. 7. Estimated values of the escape depth of photoelectrons from the photoyield of a
280 b; thick film of Ag as a function of photon energy, for normal incidence 0 ; for s-polariza-
tion and an angle of incidence of 30" x and 60" 0 ;for p-polarizationand an angle of incidence
+
of 30" and 60" A.
threshold falls below hvb = 3 eV, but for hv < 4 eV, L seems to depend
on the polarization of light. We may try to attribute this anomaly to an
anisotropy of the photoelectric excitation that involves a real dependence
of L on polarization, but we shall discuss in section 4.2 interpretations
based on a surface effect.
288 PHOTOEMISSION CVI. 3
'2
Fig. 8. Comparison of experimental values of the angular yield ratio Y - ( i 2 ) / Y - ( 0 ) with
values calculated from the expression (3.1). No possible choice of film thickness zo and escape
depth L can account for the yield observed at the resonance peak (after CALLCOTT and
ARAKAWA [1975]).
290 PHOTOEMISSION [VI, 3
depth of the electron in Se was L = 40+ 10 A for hv = 7.8 eV; the mean
energy of the emitted electrons was then 4.7 eV above the Fermi level.
PONGand SMITH [1973] replaced the selenium by copper phtalocyanin
and found then L = 11 A for electrons 1.5 eV above the Fermi level (for
hv = 7.8 ev).
When a film and its substrate both contribute to the photoexcitation,
the contributions of the film and the substrate sometimes can be separated
by an energy analysis of the photoelectrons. EASTMAN [1970], deposited
thin films of yttrium onto a substrate of gold and measured the variation
of the integrated intensity Z of the d peak of gold as a function of the Y
film thickness. The experimental data could be fitted with the law
with L' = lOA in the photon energy range 5-8 eV. The same method
can be applied to thin films of any substance, if its PED has no structure
in the d band of gold. By substituting Gd or Ni for Y , Eastman obtained
the same value L' = lOA for hv = 7 eV. We must note that such a method
gives the elastic escape depth L'; the escape probability without energy
loss is proportional to exp (- z/L'). L' may be well approximated by the
mean free path 1. But the escape depth L for all emitted electrons, as defined
by (1.1), may be much larger, especially for high energy electrons excited
by soft X-rays or by electron bombardment. Applications of the Eastman's
method to this energy range will be discussed in section 3.10.
CAMPAGNA, PIERCE,SATTLER and SIEGMANN [1973] measured the spin
polarization P of the photoelectrons emitted by a ferromagnetic material
placed in homogeneous magnetic field normal to its surface,
(3.7)
where nt and nl are the respective numbers of spin up and spin down
photoelectrons. Of course if copper is substituted for the magnetic material
then P = 0. Many important results about magnetism have been obtained
from such measurements of P.We are concerned here with the possibility
of separating the electrons originating from a ferromagnetic and a non-
magnetic material. PIERCE and SIEGMANN [1974] deposited thin films of
copper of increasing thickness zo on a substrate of Ni and measured the
variation of P with z,,. For a uniformly magnetized film and a constant
photoelectric current, P was found to be proportional to the number of
electrons excited in Ni. Pierce and Siegmann could thus obtain the trans-
292 PHOTOEMISSION
1 1 I I 1
5 10 15 20 25
Cu THICKNESS (A)-
Fig. 9. Spin polarization P of the electrons emitted by an Ni sample coated with a film of Cu
of thickness z., The rectangular fields represent the statistical uncertainties for both P and zo.
Fields with the same cross hatching are for films successively evaporated on the same Ni
substrate. The solid curve is a least square fit by an exponential curve (after FIERCE and
SIEGMANN C1974-J).
parency exp( -z,/L) and the escape depth L for the Cu film (Fig. 9). Pierce
and Siegmann deduced in a similar manner the escape depth in Ni from
the variation of P for the electrons emitted by a substrate of Cu covered
with a thin film.of Ni. Pierce and Siegmann found L = 1 1 A in Cu for
electrons 5.2 eV above the Fermi level.
-
transmitted by the film with a collector and he could eliminate, with a
0 target Film
F""" c o l k ~
Fig. 10. Principle of the apparatus of KANTER [1970]. The beam is collimated by a 0.02
Tesla magnetic induction. The beam diameter is about 1 mm. The beam could be moved, with
the help of deflection plates, across the film surface so that the detection of pin holes and
other film nonuniformities was greatly facilitated. The normal component ofthe beam energy
spread, as measured at the collector without a target inserted, was 0.5 eV between the 10%
and 90 % points of the collector current versus retardation voltage curve. Typical bombarding
currents were 2 x 10-'A. Current leaving the film was in the 10-"-IO-'4A region and were
measured with a vibrating reed electrometer. The noise current was about 10-14A.
retarding potential technique, the electrons that had suffered large energy
losses. Kanter plotted the logarithm of the electron current versus the film
thickness, and found a straight line of slope l/Lo. Here Lo is the elastic
escape depth L', when the scattered electrons are eliminated, and the total
escape depth L, when all transmitted electrons are collected. In this energy
294 PHOTOEMISSION [VI, 3
range, where L and L .' are nearly equal, we may expect the escape depth
measured for electrons injected with a kinetic energy E, to be approximately
the same as for photoelectrons excited by photons of energy hv = E, WF, +
where WF is the work function of the film. In Fig. 1 the results of Kanter
are plotted for gold at that energy. Kanter found no significant difference
between the values of the escape depth for Al, Au and Ag. He found a
decrease from L = 4OA for an electron energy of E = 5 4 e V above the
Fermi level down to 15-20A for E = 10 eV. The experiments of Kanter
determine the electron interaction processes more precisely than PE,
because the angular and energy spreads of the electrons are smaller. The
main difficulty is to avoid holes and other defects in the films.
The injection of hot electrons into a thin film may be obtained in sand-
wiches of metal-insulator-metal thin films (Fig. 11). Such sandwiches emit
electrons into vacuum when an appropriate voltage V is applied between
the metal films. Electrons of the base metal film are transported across the
insulator either by thermoinjection into the conduction band or by the
tunnel effect. They are then transported across the outer metal film and
are emitted into vacuum. We may assume the transport probabilities across
the insulator and the outer metal film to be proportional to exp( - zI/L,)
-+ii substmte
Fig. 12. Band scheme of GaAs, showing the excitation and thermalization of electrons in
r and X minima of the conduction band, -=
for I .4 ihv 1.7 eV (a) and for hv > 1.75 eV (b and
c) (after JAMES and MOLL[19691).
For 1.4 eV < hv < 1.7 eV all excited electrons are near the r level and,
except for a small path before thermalization, they undergo the same
transport process, characterized by the same escape probability
P = Pr ~ X (P - z / L r ) . (3.8)
The escape probability at the surface p r and the escape depth Lr are in-
dependent of the photon energy. The photoyield Y for 1.4 eV < hv < 1.7 eV
depends on hv only because the absorption constant a depends on hv. We
may write
The plot of 1/ Y versus 1/ a (Fig. 13), for that spectral range is a straight line
that intersects the x axis at the abscissa - L,. Its slope is 1/pr . Except for
very strong doping a results from direct interband transitions and does
not depend on the sample. a can, therefore, be taken out of the literature.
Many determinations of the spectral yield of GaAs and other 3-5 NEA
VI,8 31 THE ESCAPE DEPTH O F THE PHOTOELECTRONS 291
Fig. 13. 1/Y versus l/aplot for Zn doped GaAs crystal (density of Zn atoms p = 2.8 x 10''
) , different Cs, 0 coverages (after GARBE
~ r n - ~with [1969a]).
Fig. 14. Diffusion length L, in GaAs as a function of the Zn doping concentration p and
crystal growth process (after GARBE
[1969a]).
298 PHOTOEMISSION CVh 3
Y, = PXF,
(3.1 1)
+
1 l/aL,
Because of surface states and because surface doping -is different from
volume doping, the distance 6E from the valence band to the Fermi level
depends on the distance z to the surface (Figs. 15 and 16). This effect makes
it possible to obtain NEA. When the electron affinity is positive VANLAAR
and SCHEER [1962] calculated the spectral yield distribution on the assump-
hS; (4
WKh band 3
--____
Ferrni level
Lbkcce band
-Cz
0
Fig. 15. Scheme of the band profile of a p-type Si crystal with positive electron affinity.
300 PHOTOEMISSION
Fig. 16. Band scheme just after cleavage of the degenerate n-Si cristal (density of donors 10
cm-) for which WAGNER and SPICER [I9721 measured the PED represented in Fig. 17.
tion that each layer of thickness dz contributes to the emitted current the
amount
hvd(z)= hv,(z)+ 0.3 eV. They assumed for each layer a contribution
s
dY = (c,(hv-hv,(z))~+c,(hv-hv,(z)) exp (-z/L)dz
the result with the experimental result of Scheer and Van Laar. They found
L = 25 i-5 A in the same energy range.
WAGNER and SPICER[19721 investigated just cleaved degenerate n-type
samples of Si (10" As atoms per cm3, 0.001 SZacm) for hv in the range
6-12 eV. After an exposure to oxygen at very low pressure (lo-'' torr),
they observed several changes in the PED N(E) (Fig. 17), viz.
-6 -5 -4 -3 -2 -1 0
INIT~ALELECTRON ENERGY lev)
Fig. 17. PED for n-Si just after cleavage (1) in ultrahigh vacuum and after exposure to residual
gases, (2) for hv = 10.2 eV (after WAGNERand SPICER [1972]).
SCATTEED VALENCE
ELECTFDNS
I
that the intensity of the elastic peak Z, from the valence band is given by
(3.14)
Fig. 19. Variation with gold thickness of the intensity of the gold x and chromium 0 peaks
for photoelectrons emitted by a chromium substrate. covered by a gold thin film illuminated
by 1252 eV photons. The continuous curves are calculated for L.' = 22A (after BAER.HEDEN.
HEDMAN, KLASSONand NORDLING [1970]).
X-rays also induce the emission of Auger electrons with an energy in-
dependent of hv and characteristic of the chemical elements because of the
deep holes that they create in the solids. KLASSON, HEDMAN, BERNDTSSON,
NILSSON, NORDLING and MELNIK[1972] used both Auger electrons and
directly photoexcited electron peaks, to determine the escape depth of
electrons in gold and alumina. They found in gold L' = 19 8, for E = 0.9
k e V a n d L f = 3 7 8 , f o r E = 1.4keV,inaluminaL'= 1 3 8 , f o r E = 1.4keV
and L' = 22 8, for E = 3.9 keV. STEINHARDT, HUDISand F'ERLMAN [19721
investigated L' in a thin film of carbon dtposited onto a gold substrate.
They found L' = 10 8, for E = 920 eV and L' = 13 8, for E = 1169 eV.
The same method applies when Auger emission is induced by electron
bombardment. We may assume a uniform excitation in the ?art of the film
and substrate that are responsible of the Auger emission, because the
incident electrons have a penetration depth greater than L' both because
L' increases with electron energy and because secondary electrons can also
"1. Q 41 S U R F A C E PHOTOEXCITATION 305
Energy(eV)-+ 48 60 110 120 262 350 355 920 935 1169 1400 1736 3900
Material
1
Be 4.7 (8.6) 10.0 12.7
'412 0 3
13 22
C 7.5 10 13
cu (6.1)
Ag 9.4
Au
Mo 5.2 6.7
W (7.2) 6
0 4. Surface Photoexcitation
A surface contribution to photon absorption may arise, within a
306 PHOTOEMISSION CVL P4
one-electron theory, from two distinct effects, surface PE from bulk states
(SPBS) and PE from surface states. We shall analyse experimental data
on the assumption that surface and volume effects can be separated in the
intensity of the emitted current. When necessary, we shall use the one-
electron approximation.
The surface photoexcitation is more likely detected by photoemission
than by any other of its physical consequences (e.g., reflectance or trans-
mittance studies), because the small value of the escape depth reduces the
relative importance of bulk phenomena with respect to surface ones. In
spite of that, and even for quite small values of the escape depth, most
analyses of PE data have been based on the assumption of a pure volume
photoexcitation and calculated from the bulk value of the dielectric con-
stant. The consistency of the calculated escape depths with all experimental
data is an evidence of the quite small value of the surface effect with respect
to the volume one. This result must be especially noted because in early
theoretical work on PE from metals, the free-electron model led theorists
to consider only the surface effect.
that the escape depth of photoelectrons is very small, at most a few mono-
layers. This fact would make the surface effect easier to detect if it exists.
The steep increase of photoyield at zero thickness, observed by MAYER
[1961], is consistent both with volume and surface photoexcitation and
we shall need further experimental tests to decide whether the mechanism
of the absorption, that can lead to photoemission only in the surface layers,
is different from bulk absorption.
I I I I I , ,
Fig. 20. Variation of the density of electrostatic energy per incident light power, near the
surface associated with 3 components of the electric field in a 280 b; thick silver film illuminated
at an angle of incidence i, = 60. ------ .r component parallel t o the surface (p wave),
-z component normal to the surface (p wave), - - - - - - - - - - y component parallel to the
surface (s wave).
plasma resonance, where the condition (c) might be satisfied. But the
sensitivity of the anomaly to the composition of the surface does not
allow us to rule out a PE from surface localized impurity levels.
ENDRIZand SPICER[1971b] observed in aluminium values of the ratio
Y,/ Y, larger than expected in a volume effects below surface plasmon
energy for hu < 10.5 eV. ENDRIZand SPICER[1971a,b], FLODSTR~M and
ENDRIZ[1973] and ENDRIZ[1973] discussed in detail the expected contri-
bution of the surface effect (SPBS) with the model developed by MITCHELL
310 PHOTOEMISSION [VI, 9 4
TABLE4.1
Value of the true quantum efficiency that gives maximum polarization dependence for alkali
opaque filmsdeposited a 77 K before and after reheating, for both polarization Q , and Q,,
at wavelengths 1, in lngstroms (after MONINand BOUTRY[1974])
cs Rb K Na
Fig. 21. Spectral distribution of the "true quantum yield" of an opaque film of Rb after
deposition at 77 K + and after reheating at 195" K * for s and p polarization (after MONIN
[19733).
Qp and Q, for Cs are not very different and this suggests that the relative
contribution of the surface effect is very small although the contribution
of the volume effect is reduced by the very small' value of the escape depth
that is confirmed here by the small value of both Qp and Q, (several orders
less smaller for other alkali metals). But the distinction between the volume
effect and emission from surface states becomes rather semantic when the
escape depth is reduced to a few or even to one atomic layer (section 3.1).
In such thin a surface layer the usual approximations of volume effect may
be quite rough, especially the description of the photon absorption by
the bulk dielectric constant and the assumption of an electric field dis-
continuous at the surface. It seems quite difficult then to deal with the
surface layer by using the model of an homogeneous solid limited by an
abrupt surface.
Every observed anomaly of the ratio Yp/Y,seems up to now to depend
on structural parameters that are not completely controled. If these param-
312 PHOTOEMISSION [VI, 5 4
Fig. 22. Spectral distribution d the true quantum yield Q of an opaque film of Cs after
deposition at 77 K + and after reheating at 195 K * for s and p polarization (after MONIN
[1973]).
In metals the good results obtained with the free electron model in many
fields, e.g.,electrical conductivity, led to the idea that no volume photo-
effect could occur and photoemission has been explained in the early
theories as a pure surface effect. When the possibility of a volume effect
was recognized (FAN[1945]), the k conservation selection rule led one to
expect a larger threshold hv, for the volume effect than hv, for the surface
effect, because in the latter case only the tangential component of the
electron wave vector k is to be conservedwhen the photon energy is absorbed
by the electron. WEISSLER [19561 interpreted the spectral yield distribution
VI, D 41 SURFACE PHOTOEXCITATION 313
of several materials (W, Mo, Pt) with two thresholds hv, and hv,. The
experiments described in the last paragraph did not confirm this inter-
pretation but the surface effect is still expected to afford a more noticeable
contribution near the threshold (ENDRIZ [19731, CHABRIER, GOUDONNET,
TRUITARD and VERNIER [I19731, CALLCOTT and ARAKAWA [19751).
The steep increase of the photoyield, that Weissler had interpreted as
a threshold for volume effect, is now attributed to a reflectance decrease
that permits a better penetration of the light into the solid when the photon
energy becomes larger than the plasmon (GdkLICH [1959]). An alternative
explanation of the PE at low photon energy, when no k conservative transi-
tion exists, lies in indirect transitions with phonon creation or annihilation
or in non-direct transitions where the momentum conservation is insured
by other many-body effects (see section 2.8).
GARTLAND, BERGEand SLAGSVOLD [19731 compared the spectral distri-
bution of the photoyields Yo for normal incidence and Y700 for an incidence
of 70". As expected they have found Yo to be proportional to ( h v ~ h ~ , ) ~
near the threshold, but this was not so for Y700.They attributed the
difference between Y700and the expected (hv - h ~ , law ) ~ to a contribution
of the surface effect. A calculation of the penetration of light for an angle
of incidence of 70" would be necessary to confirm this interpretation.
In a semi-conductor high values of the photoyields require transitions
from the valence band. A threshold must then be observed when the photon
energy equals the difference hvi between the vacuum level and the top of
the valence band. Another threshold hv, has been considered by ALLEN
and GOBELI [19621, because the corresponding transitions are ,generally
not direct, and they assumed much higher transition probability when
direct transitions are possible.
When photoelectrons come from surface states, a tail can be observed
in the spectral yield distribution below the threshold hvi . To give a correct
interpretation of this tail we must take into account the band bending
induced by the space charge due to surface states and, of course, a possible
shift of the spectral yield distribution due to a change of the work function.
In a theoretical analysis KANE[19621 has made a survey of 11 different
possible surface and volume emission processes and associated with eaw
one a contribution to the spectral yield:
q = Ci(hV--hVi)P*. (4.1)
The threshold hvi and the exponent pi depend on the process. VANLAAR
and SCHEER [1965] fitted the measured photoyield of a cleaned slightly
314 PHOTOEMISSION [VI. D 4
The PED has been the most direct way to investigate the energy levels
of electrons in a solid. If a piece of structure obseryed in the PED appears
or disappears when a very small amount of gas is adsorbed we think at
first of a surface localized level.
We reported in section 3.9 and Fig. 17 the PED observed by Wagner
and Spicer for highly n-doped samples of Si. The attribution of the peaks
A and B to surface states is strongly suggested by arguments other,than
their disappearance after an exposure to oxygen. Their distance to the
Fermi level and their intensity are independent on the photon energy, in
contrast to the peaks C and D that have been attributed to direct interband
transitions. Wagner and Spicer estimated a density of surface levels of
about n, = 8 x lo4 cm-.
VI, 41 SURFACE PHOTOEXCITATION 315
EASTMAN and GROBMAN [1972] measured the PED for cleaved surfaces
of lightly n-doped crystals of Si ( l l l ) , Ge (111) and GaAs (110). The
resistivity of the Si and Ge crystals were respectively 5 0 * cm and 4 Q * cm.
They used synchrotron radiation in the spectral range 7-25 eV. Like
Wagner and Spicer, Eastman and Grobman found a structure near the
Fermi level that disappeared after an exposure of the sample to the residual
gases for a few hours (lo-'' torr). Except for the peak B, that has not been
found by Eastman and Grobman, the estimated density of surface states
in Si is roughly the same in both experiments. RANKEand JACOBI[I9731
have observed oxygen sensitive structure in the PED from polar faces of
GaAs crystals ((111) Ga and (1 11) As), that they attributed to surface states.
In Si samples that had been exposed to residual gases, Eastman and
Grobman observed the appearence of a broad band of filled states centered
at 7 eV below tbe vacuum level, that they attributed to extrinsic levels of
(Si-0). Eastman and Grobman gave a rough estimation of the escape
depth of bulk excited electrons by comparing the number of electrons,
that had been excited in bulk and in surface, and assuming the same matrix
element for both transitions. They obtained L = 40 A for hv = 8.5 eV,
L = 17A for hv = 10eV and L = 6 A for hv = 12eV. The agreement
with other estimates (WAGNER and SPICER[1972], GOBELIand ALLEN
[ 19621, KANE [1967]) is satisfactory, although the difference between
the values of L for hv = 8.5 eV and hv = 12 eV seems quite large.
The cleavage of a crystal under an ultra-high vacuum is not a sufficient
condition for obtaining a perfect or perfectly reproducible surface. ERDU-
BACK and FISCHER [I9721 observed the surface of cleaved crystals of Si by
LEED and by Auger electron spectroscopy, as they determined the PED
for photons of low energy before and after annealing at different tempera-
tures. In agreement with the observations of ALLENand GOBELI[1962],
CALLCOTT [19671, FISCHER [19681, and EASTMAN and GROBMAN [19721,
before annealing, they observed no structure that could be attributed to
PE from surface states, as expected if the levels associated with the peak
B of Wagner and Spicer were present. Erduback and Fischer observed
such structures only after annealing. The LEED pattern of the Si surface
which was of type 2 x 1 just after cleavage became 1 x 1 after annealing at
550 "K and 7 x 7 after annealing at 800 OK.
SEBENNE, GUICHAR, BOLMONTand BALKANSKI [1973] deduced from the
spectral yield distribution of n-Si and its modification by an exposure to
oxygen a density of surface levels in agreement with results described above.
But LAUDB[19731 suggested that the structures previously attributed
to surface states should be attributed to direct transitions from the r;,
316 PHOTOEMISSION [VL 04
level of bulk Si. The disappearance of the structure with contamination
should be explained by surface inelastic scattering.
In metals PE from intrinsic and extrinsic surface states has been invoked
to explain structures in the PED that were sensitive to residual gases.
FORSTMANN and HEINE[1970] suggested intrinsic surface states for
anomalous structures previously observed for Ni and Cu e.g., by CALLCOTT
and MAC RAE[1969] who had measured the PED for (1 11) faces of Ni
monocrystals that had been controled by LEED. EASTMAN [I9711 believes
that these structures should rather be attributed to PE from extrinsic
surface states. EASTMANand CASHION {1971] measured the PED for thin
films of Ni exposed to oxygen and CO and found characteristic electronic
levels at 5.5 eV below the Fermi level for oxygen and at 7.5 and 10.7 for CO.
BAKERand EASTMAN [19731 also observed characteristic impurity levels
when (1 11) and (100) faces of a W monocrystal were exposed to 0 and CO.
They could control the structure of the absorbed films by LEED and Auger
spectroscopy, and identify in the PED from CO absorbed layer a peak at
-8.9 eV due to CO molecule.
In metals most of the effects, that could reveal a gap between the energy
bands associated with each direction of the wave vector k, are masked by
overlapping, when all directions of k add their contributions. Nonetheless
VERNIER, COQUET and BOURSEY [19681predicted a possible increase of the
photoelectric threshold for crystallographic faces of a crystal such that
the vacuum level lies in the band gap associated with k normal to this face.
FEUERBACHER and FITTON[1973] suppose in that case that a new type of
surface effect occurs with direct transition of a bulk Bloch state to a free
electron state in vacuum. A surface coupling between these states is possible
because in a semi-infinite solid each Bloch state in the solid includes an
exponentially damped part outside the solid and each free electron wave
outside the solid includes a damped part inside the solid.
A determination of the PED for one direction of emission can separate
the overlapping bands and is especially useful to detect the surface states
associated with one direction of k. The energy distribution for the electrons
emitted normal to the surface within an angle of 12-15" was measured by
WACLAWSKI and PLUMMER [19721 with a retardation potential technique
and FEUERBACHER and FITTON [1972,1973] with a 127"cylindrical analyser.
FEUERBACHER and FITTON[1972] observed (100) faces of W monocrystal
and WACLAWSKI and PLUMMER [19721 observed polycristalline samples,
but purified the sample by a heat treatment, that induced a dominant (100)
orientation. In both publications 7.7 eV and 10.2 eV photons induced a
PED with a peak that has been attributed to electrons excited from a sur-
S U R F A C E PHOTOEXCITATION 317
12
11
x)
12
11
x)
I
->
3L
12h
rlz
4
1
a
Fig. 23. Energy distribution of photoelectronsemitted normally to (IOO), (1 10) and ( 1 1 1 ) faces
of a monocrystal of W for several photon energies (after FEUERBACHER and FITTON [1973]).
318 PHOTOEMISSION CVL 5 4
face level at 0.4 eV below the Fermi energy. Feuerbacher and Fitton made
the - 0.4 eV - peak disappear by an exposure of W to H, . Waclawski and
Plummer observed the same effect with H,, 0,, N, and CO and noticed
at the same time appearance of peaks characteristic of the absorbed gas.
FEUERBACHER and FITTON [1973] gave the energy distribution of the
photoelectrons emitted normal to (loo), (1 10) and (1 11) faces of a mono-
crystal. They confirmed the results obtained earlier with (100) faces and
found very different distributions for the other faces (Fig. 23). The surface
peak 1 that had been found at -0.4 eV for (100) face, does not exist for
other orientations. This confirms its surface origin. The structure no. 2
could be explained at first sight by non-direct transitions from bulk levels,
but Feuerbacher and Fitton attributed it to the photoexcitation of bulk
states allowed by surface, because it is observed only for (100) faces. The
peaks no. 4 and 6, observed for face (1 lo), can be associated with a strong
density of levels at E4 = -0.3 eV and E, = - 1.4 eV but no Bloch state
+
exists at the energy E4+ hv and Es hv on the line T N of the band diagram
(Fig. 24). Feuerbacher and Fitton explain these structures by the surface
coupling between the Bloch state inside the solid and the free electron
state outside the solid that we have previously considered in this paragraph.
Other structures 3,8,9,10,11 can be readily interpreted as direct transitions.
(See also TURTLE and CALLCOTT [1975].)
a!-
~ ~~
r
(A) (do) ($0) (000) (1%
Fig. 24. Band structure of tungsten after Christensen. (Quoted by FEUERBACHER
and FITTON
[1973]).
VI, 51 CONCLUSION 319
0 5. Conclusion
The surface excitation, that was alone considered in the early works
based on the free electron model, appears now as a minor contribution that
can be proved only by very refined technique.
In a good approximation the PE is localized by electron scattering within
a more or less thin superficial layer. The attenuation length of a light beam
is generally at least in the range of one hundred Angstrom and is larger
than the escape depth of the electrons. The escape depth of the electrons
varies in a very large range according to the photon energy. It depends to
a quite small extent on the material when the electron-electron interaction
is not forbidden by a band gap. As was first recognized by SPICER[1960]
in alkali-antimonides, much higher values of the escape depth have been
obtained for small photon energy in semi-conductors when the electron
affinity is smaller than the band gap. The drop of the elastic escape depth
in IK by one order of magnitude when the electron energy becomes large
enough to allow scattering by pair production (BAERand LAPEYRE [1973])
is another instance of the dominant efficiency of the electron-electron
interaction. Still higher values of the total escape depth are obtained when
the electron affinity is negative. These abnormal values, which violate the
general rule, have, of course, a special importance in the fabrication of light
detectors.
We have represented in Fig. 25 the general trend of the variation of the
escape depth with energy. The number of published data is so high that
we have not been able to indicate individual measurements. For high energy
photons quite a large difference appears between the inelastic and elastic
escape depths. In principle, the values represented in Fig. 25 are associated
with elastic escape depth. But we must be aware that the distinction between
elastic and inelastic escape depth is based on a finite energy resolution of the
experimental apparatus. Obviously the inelastic escape depth results from
the combination of an increasing number of mean free paths and in far
ultra-violet and soft X-ray range the inelastic escape length may be several
orders of magnitude larger than the elastic one.
The escape depth is often reduced to a few monolayers (less than one
for Cs). Then the dependence of the light absorption process on the elec-
tron scattering is certainly very important for purely optical as well as for
photoelectric phenomena. The least perturbation, that we can expect,
concerns direct transitions. When electron scattering becomes more
important, increasing differences between initial an final wave vectors of
the electron are allowed. It should be much more satisfactory to treat such
320 PHOTOEMISSION [Vl, 05
1 I I
1 10 ZT) m >
ELECTRCN ENERGY ABOVE FERMI LEVEL lev)
Fig. 25. General behavior of the variation of the escape depth with electron energy.
cases within a one-step theory and we may hope that one-step theories will
be soon able to take into quantitative account the electron-electron scatter-
ing.
Of course, the surface is one of the most important elements in photo-
emission. The work function is determined by surface phenomena and the
interaction of an incident electron with the surface can include elastic or
inelastic scattering that cannot be fully described by the work function
alone. The PE has much to receive from and much to contribute to progress
in surface physics. Here too, it would be desirable to gather into a one-step
theory the interaction of the electrons with the surface, with light and with
other elements of the solid. Especially when the escape depth is very small,
it is very crude to describe the absorption of light by the surface layer with
the same dielectricconstant as the absorption by the bulk. The interpretation
of reflectance data is subject to the same objection as the interpretation of
photoelectric data, especially in the case of cesium.
VII REFERENCES 321
Acknowledgement
The author wishes to thank T. A. Callcott for many stimulating sugges-
tions as well as critical reading of the manuscript.
References
ADAWI,I., 1964, Phys. Rev. M A , 788.
AGARWAL, G. S., D. N. PATTANAYAK and E. WOLF,1971a, Optics Commun. 4, 255; 1971b,
Phys. Rev. Lett. 27, 1022.
AGARWAL, G. S., D. N. PATTANAYAK and E. WOLF,1974, Phys. Rev. B10,1447.
ALLEN,F. G. and G. W. GOBELI,1962, Phys. Rev. 127, 150.
ANDEREGG, M., B. FEUERBACHER and B. FITTON, 1971, Phys. Rev. Lett. 27, 1565.
ANTYPAS, G. A., L. W. JAMESand J. J. UEBBING,1970, J. Appl. Phys. 41, 2888.
APPELBAUM, J. A. and D. R. HAMANN, 1973, Phys. Rev. Lett. 31, l,M.
ARAKAWA,E. T., A. J. BRAUNDMEIER, M. W. WILLIAMS, R. N. HAMMand R. D. BIRKHOFF,
1974, Communication at the 4th Intern. Conf. on Vacuum-Ultra-Violet Radiation Physics,
Hamburg.
ASHLEY,K. L., D. L. CARRand R. ROMANO-MORAN, 1973, Appl. Phys. Lett. 22,23.
BAER,A. D. and G. J. LAPEYRE, 1973, Phys. Rev. Lett. 31, 304.
BAER,Y., P. F. HEDEN, J. HEDMAN, M. KLASSON and C. NORDLING, 1970, Solid StateCommun.
8, 1479.
BAKER,J. M. and D. E. EASTMAN, 1973, J. Vac. Sci. & Technol. 10, 223.
BALLANTYNE, J. M., 1972, Phys. Rev. B6, 1436.
BARTELINK, D. J., J. L. MOLLand N. I. MEYER,1963, Phys. Rev. 130, 972.
BELL,R. L., L. W. JAMES, G. A. ANTYPAS, J. EDGECUMBE and R. L. MOON,1971, Appl. Phys.
Lett. 19, 513.
BELL,R. L. and W. E. SPICER,1970, Roc. IEEE 58, 1788.
BELL,R. L. and J. J. UEBBING,1968, Appl. Phys. Lett. 12, 76.
BERGLUND, C. N. and W. E. SPICER,1964, Phys. Rev. 136 A, 1030.
~LECHSCHMIDT,D., M. SKIBOWSKI and W. STEINMANN, 1970, Phys. Stat. Sol. 42,61.
BORTOLANI, V., C. CALANDRA and M. J. KELLY,1973, J. Phys. C6, 349.
BURTON, J. A,, 1947, Phys. Rev. 72, 531.
CALLCOTT, T, A., 1967, Phys. Rev. 161, 746.
CALLCOTT, T. A. and E. T. ARAKAWA, 1974, J. Opt. SOC.Amer. 64, 839.
CALLCOTT, T. A. and E. T. ARAKAWA, 1975, Phys. Rev., B11, 2750.
CALLCOTT,T. A. and A. U. MCRAE,1969, Phys. Rev. 178, 966.
CAMPAGNA, M.. D T. PIERCE, K. SATTLER and H. C. SIEGMANN, 1973, J. Phys. 34, C 6-95.
CAROL!,C., D. LEDERER-ROZENBLAIT, B. ROULETand D. SAINTJAMES, 1973, Phys. Rev. B8,
4552.
CHABRIER, G., J. CORNAZ, J. P. G o m o m and P. J. VERNIER,1970, Optic. Commun. 1,391.
CHABRIER, G.,J. P. GOUDONNET, J. F. TRUITARD and P. J. VERNIER, 1973, Phys. Stat. Sol. (b)
60.K 23.
322 PHOTOEMISSION IVI
VII
BY
P. J. B. CLARRICOATS
Department of Electrical ana Electronic Engineering, Queen Mary College, University oflondon,
England
CONTENTS
PAGE
0 1 . INTRODUCTION'. . . . . . . . . . . . . . . . . . . 329
6 2. FIBRES WITH CORE A N D CLADDING POSSESSING
UNIFORM REFRACTIVE INDEX . . . . . . . . . . . 331
9 3 . FIBRES WITH NON-UNIFORM REFRACTIVE INDEX 381
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . 400
REFERENCES . ....................... 400
9 1. Introduction
During the period 1969-75, the use of optical fibres for telecommuni-
cation purposes has grown from a position of speculative research into
one of commercial reality. This growth has been largely due to the efforts
of those engaged both in fibre manufacture and in the development of
light sources. Against this background many workers have been engaged
in research into the theory of optical waveguides and it is with their con-
tribution that the present review is concerned.
Work on optical fibres goes back at least to the early 1950's and a
resume of the early history is to be found in the book by KAPANY [1967].
Credit for the proposal to use a cladded optical fibre for telecommuni-
cations purposes is attributed to L o and HOCKHAM [1966], whose classic
paper appeared in July 1966. Thereafter much of the early work aimed
at the development of long-haul systems was undertaken in the United
Kingdom at the British Post Office and by Standard Telecommunication
Laboratories. By 1969, workers in the U.S.A., especially at the Bell Tele-
phone Laboratories, in Germany and in Japan, were enlarging their
efforts. At first, fibre attenuation was prohibitively large and solid-state
laser sources inefficient and of very short lifetime. A major breakthrough
occurred in 1970 with the announcement in the U.S.A. by KAPRON, KECK
and MAURER [I9701 of the Corning company, of a quartz fibre with losses
below 20 dB km-'. Thereafter, other laboratories also achieved excep
tionally low loss fibres based on quartz. In 1972, another low loss fibre
was announced using an organic liquid (tetrachloroethylene) as a core
material. This was proposed independently by OGILVIE,ESDAILEand
KIDD [1972] in Australia and by KAISER,TYNES, CHERINand PEARSON
[1972] in U.S.A. Their announcement was soon followed by one by ~ A Y N E
and GAMBLING [I9721 in England, who proposed the use of a related
organic liquid with even lower loss. In Japan, effort was concentrated
mainly on fibres with graded refractive index since these have the consider-
able advantage of neutralising modal pulse dispersion, which is the principal
disadvantage associated with multimode fibres. Their researches were
3 29
330 OPTICAL FIBRE WAVEGUIDES [VIL 0 1
92. Fibres with Core and Cladding Possessing Uniform Refractive Index
2.1. INTRODUCTION
* References to the numerous contributions of these workers will be given in later sections.
332 O PT I CAL F I B R E W A V E G U I D E S
Fig. 2. Optical fibre configurations:(a) fibre with finite cladding boundary, (b) infinite cladding
approximation, (c) cladding mode approximation.
TABLE 1
Typical fibre parameters
field confined to the core region and in practical fibres designed to support
many modes, the condition is usually accomplished by making the core
radius much larger than in the monomode case.
A first step in evaluating the transmission properties of a fibre involves
the calculation of the propagation coefficients of the modes. An exact
analysis for the configuration of Fig. 2(a) has been made by CLAWCOATS
and CHAN[1973], however, their study serves to validate the widely used
infinite cladding assumption which reduces the configuration to that of
Fig. 2(b). A further assumption applies to modes below their core-mode
cut-off when the fields are virtually unperturbed by the presence of the
core, the fibre can then be represented by Fig. 2(c).
Studies of the propagation behaviour of a cylindrical dielectric embedded
in an infinite medium of different permittivity as in Fig. 2(b) or 2(c) began
vn, 0 21 FIBRES WITH CORE A N D C L A D D I N G 333
The general form of the characteristic equation for modes with azimuthal
dependence* n in the structure of Fig. 2(b) is given below, the derivation
of which is presented in numerous texts and papers (see references),
where
* Note: Throughout this text n is the azimuthal mode number which should not be confused
with refractive indices n l , 2 . . ..
334 OPTICAL FIBRE W A V E G U I D E S
TABLE2
U,and U, for first 28 modes (nl x n2)
0 2.405
2.405 3.832
3.832 5.136, 5.520
5.136 6.380
5.520 7.016
6.380 7.588
7.016 8.477, 8.654
7.588 8.772
8.447 9.761
8.654 10.173
8.772 9.936
9.761 11.065
6-
5-
-
OI
3
0
C
&
._ L-
al
v normalized frequency
Fig. 3. Eigenvalue U as a function of normalised frequency V. The solid lines are determined
numerically from equation (5); the dashed lines are the approximate solution given by equa-
tion (T3.2) of Table 3.
that the modes HE, + l m and EH, - l m are degenerate. When V + co, W + co
thus U satisfies Jnr ( V , ) = 0 in that limit. For propagating HE,, and
EH,, modes, U lies between consecutive roots of the Bessel function
J,* 1 ( U ) . Figs. 3 and 4 show U and W as functions of V while special cases
of U and W are contained in Table 3. The behaviour of modes below
cut-off will be deferred to the section on leaky modes.
Table 3 contains approximate analytic expressions for U derived
originally by SNYDER[19691 and GLOGE[1971a]. Equations (T3.1a),
(T3.lb) and (T3.2), are very accurate while (T3.2) is useful for many cal-
culations. When A = 0.05 equation (T3.1 b) giving U for the HE, l mode
is accurate to within 0.5 % in the range 1.5 < V < 5 .
. From the definitions of U and V given in equations (1) and (2), GLOGE
[1971a] writes
336 OPTICAL F I B R E W A V E G U I D E S
b
t -+
Ib
h 3
5 e
VII, 21 FIBRES WITH CORE A N D C L A D D I N G 331
"s IB
I
d
338 O P T I C A L F I B R E WAVEGUIDES
10-
8-
6-
W
L-
2-
OO'
Fig. 4. The eigenvalue Was a function of normalised frequency V for the lowest order modes
of the fibre shown in the configuration of Fig. 2(b): 01 = HE,,; 1 1 = HE,,, TM,,;
21 = HE,,, EH,,; 02 = HE,,; 31 = HE,,, E H , , ; etc.
Thus with equations (T3.1) or (T3.2) and (7) 8, the normalised phase-change
coefficient, can be easily evaluated as a function of V. Fig. 5 shows results
due to GLOGE [1971a] who uses a linearly polarised modal notation and
is discussed in section 2.7. The relationship to the conventional notation
is explained in Fig. 4.
V
Fig. 5 . Normalised propagationcoefficientb defined by equation (8) as a function of normalised
frequency V for optical fibre configuration of Fig. 2(b).
where 8, is the complement of the critical angle for total internal reflec-
tion at the interface between core and cladding regions. We now see that
as fl increases from n2 to n, the angle 8 which the ray vector n,k subtends
with the generatrix of the cylinder, changes from 8, to zero.
-.. - .
U plane
Fig. 6 . Rays in an optical fibre waveguide: (a) Dotted region corresponds to trapped modcs such as ray I . Ray 2 describes a leaky ray; ray 3 des-
cribes a refracting ray. (b) Section through plane containing ray 1. (c) Detailed view of trapped leaky and refracting ray regions, Locus A B C
corresponds to increasing meridional angle 6, while 4 remains constant. 2.t
V", 21 FIBRES WITH CORE A N D CLADDING 341
1.40 -
a
I30 -
1.20-
I10 -
0 1 2 3 4
V
I0 I I I I
0 1 2 3 4
V
Fig. 7. Normalised propagation coefficient Bas a function of normalised frequency V. Fibre
parameters n, = 1.53; n2 = 1.50; r 2 / r 1= 5.
The first term in equation (14) arises from the material dispersion and is
independent of the waveguide mode, the second is due to waveguide dis-
persion and is mode dependent. By using the relation for dUjdV given in
Table 3
d(Vb)
_ _-
U2
- 1 - -[l-2un]
dV V2
= 1+ c[l-
V2
i] when V > 1
d(Vb)/dV = 0.
While for HE,, and EH,- 2m modes
d( Vb)
dV
VIL 21 F I B R E S W I T H CORE A N D C L A D D I N G 343
F
J
0 2 6 8 10 12
V
Fig. 8. d(Vb) /dV as a function of normalised frequency Y for optical fibre configuration of
Fig. 2(b).
GLOGE [1971a] has plotted d( Vb)/dV as a function of V for the lowest order
modes ofa multimode fibre as shown in Fig. 8. For fibres with large Vvalues,
which support modes with large values of n, the spread in delay AT is
obtained from equations (19) and (16) as
In the limit of large n, the spread in arrival time of the fastest and slowest
modes, as predicted by equation (20), is precisely that obtained using a
ray approach. Extreme skew rays cross the axis of the fibre at an angle
8, = (2A)* and for these, the time difference relative to the axial ray is
given by the limiting form of equation (20). GAMBLING, PAYNEand SUNAK
[1971], GAMBLING, DAKIN, PAYNE and SUNAK[1972] and ROSMAN [1972]
have used the simple ray model to investigate the response of fibres and
have reported good agreement with experiment. In these cases mode
coupling must presumably have been small for, as discussed later, when
present it leads to a dependence on length of the form L* rather than L.
344 OPTICAL FIBRE W A V E G U I D E S Cvn, 0 2
I I I I I I I
0.3 0.4 0.5 0.6 0.7 08 0.9 1.0
ho* Pm
Fig. 10. arjaf for HElI mode as a function of V and L for optical fibre comprising Schott
K, and K, glasses as core and cladding materials. Dotted curves show dispersion in the
constituent materials. Fibre parameters r J r , = 5.
For a single mode waveguide, the first term dominates and published
values of dN/dk may be used to estimate pulse dispersion. To demonstrate *
this point, Fig. 10 shows computed values of a~/ affo rthe HE,, mode in
a glass fibre waveguide without approximations and including the effects
346 O P T I C A L FIBRE W A V E G U I D E S CVIL 0 2
Fig. 1 1 . Bandwidth limit due to material dispersion: variation with wavelength for constant
line width AI = 40 nm. Curves correspond to respectively silica and various glasses.
VII, 0 21 FIBRES WITH CORE AND CLADDING 347
Gbitk. ,GaAs-laser
He!&-laser '\\\
- \
-2
Lz
a
10-
-a
E
1-
X
!i
1 I
0.1
with
m = 1, 2, 3, ...
For constant Urn,the integer values of n and m satisfying equation (23) lie
along parallel lines as in Fig. 13. Neglecting the term 3 in equation (23),
all values of n and m lie within the triangle. Because there are two
348 O P T I C A L FIBRE W A V E G U I D E S
L
n
Fig. 13. n, m,mode number plane. Dashed curve indicates the location of points n + 2m = con-
stant.
polarisations and two mode types (HE and EH) for each value of n and m,
the total number of modes is four times the area of the triangle. As the
maximum values of n and m are 2 V/nand V/nrespectively, the total number
of (trapped) modes MT is given by
V, (cut-off V value1
Fig. 14. Computed total number of modes MT as a function of cut-off V value. Curve B
includes modes of the form HE,, and EH,, with both polarisations present. Curve A
includes modes with only one polarisation present.
due to PASK,SNYDERand MITCHELL [1975] gives the same result. The use
of a compound mode number by GLOGE [1972a]
M = (n+2rn) (26)
together with the knowledge that for all V, U z U, (the cut-off value of U )
allows us to write
u, = (2M)Q (27)
then from equation (8)
V d2(Vb) 2u2 2M
-
- _ - N
I v - - -
(29)
dV2 V2 MT
350 OPTICAL FIBRE W A V E G U I D E S CVII. 0 2
The first term of equation (30) is positive whereas the second term is negative.
Thus, as proposed by GLOGE[1971b], the possibility exists for some
compensation of material dispersion by waveguide effects in the case of
higher order modes for which M w M T .
The possibility of compensating material dispersion by appropriate
choice of core and cladding glasses has been proposed by DYOTT and STERN
[1970], SMITHand SNITZER[1973], also JURGENSEN [1975]. The latter
author shows that no suitable glasses exist which will provide compensation
in a monomode fibre at wavelengths below about 1.3 p. Also, so far,
attenuation considerations have dominated the choice of materials for
optical fibres and the limitations on bit-rate imposed by material dis-
persion have taken second place.
2.7. FIELDS
Table 4 presents the field components of the HE,, and EH,-, mode
under the assumption that n , x n 2 . Fig. 15 shows computed transverse
fields of the HE11 , and HE2 modes and the linearly polarised nature of
these fields is evident. It may also be demonstrated as follows. If the Cartesian
components of the transverse electric field are formed from the cylindrical
components we have
TABLE 4
Field components when n, = n2
Component Function
r < r1 r > r,
* Upper sign denotes HE,, mode, lower sign denotes EH,-,, mode.
initials because the field is almost linearly polarised. The notation LP,,
identifies the above mode pair. An extensive discussion of this topic has
been presented by MARCUSE[1972b, 1 974a).
If initially we assume that a artesian field, as expressed in equation (33),
represents a solution for the configuration of Fig. 2(b) and provided
n, z n2, then, following ARNAUD119741, continuity of Ey and aE@r at
r = r1 leads directly to the characteristic equation
where the functions are defined with eq. (1). This solution corresponds to
either and HE,,. l)m mode or an EH,,- l)m mode. The appropriate character-
istic equations for HE,,,,, and EH,,, modes are then
Fig. 15. Normalised field intensity as a function of radial position within cladded optical
fibre corresponding to Fig. 2(a). Parameters as Fig. 7.
while
Pcore
~ = 1-
U2
-[l-Ky]
PT VZ
and
--
Pclsd u2[l-K,].
--
PT v2
For large V,
1
'K, = 1- -
V
then
and
--
Pclad
- -u2
P, v3
354 O P T I CAL FIBRE WAVEGUIDES
V
Fig. 16. (power in cladding)/(total power) and (power in core)/(total power), as a function
of normalised frequency V.
Fig. 16, due to GLOGE [1971a], shows the above ratios as a function of V.
CLARRICOATS and CHAN[I9731 have produced similar results for the
configuration of Fig. 2(a) and have found that the role of the outer cladding
boundary is not significant other than near the core mode cut-off. The above
figures show that for all modes above their core mode cut-off, their power
flow is concentrated into the core. If equal power flows in every mode, as
could be the case if excitation is by means of an LED source, the total power
flow in the cladding for all modes is given by
If the loss region has only a finite thickness dr and its inner boundary still
lies at r = r 2 ,
\ lC01 \
Fig. i7. Attenuation coefficient as a function of normalised frequency for optical fibre con-
figuration of Fig. 2(a). Parameters as for Fig. 7.
356 O P T I C A L FIBRE W A V E G U I D E S [vn, 52
tion (48)over the linear range. The very substantial differential attenuation
which exists between the HE,, mode and next higher order mode triplet,
provided V does not approach 2.4, makes it possible to provide powerful
suppression of all higher order modes in monomode fibres. This requires
a careful choice of fibre parameters and the application of a lossy paint
or a lossy plastic coating to the cladding boundary. KUHN[1975] has
discussed this optimisation problem in detail and has included in his study
Fig. 18. Leaky mode attenuation coefficient as a function of ratio r J r , with Vvalue as param-
eter, for configuration of Fig. 2(a) with n 1 = 1-05; n2 = 1.025; n3 = 1.5.
VI1, 21 FIBRES WITH CORE A N D CLADDING 351
the case when the outer lossy region (which is assumed to be semi-infinite)
has a real part of refractive index greater than n, . Under these circum-
stances so-called leaky mode conditions prevail and the attenuation rises
rapidly with increasing n, or decreasing V. CLARRICOATSand CHANG [I9721
have analysed the structure of Fig. 2(a) when n3 > n, > n, and Fig. 18
shows computed attenuation coefficients as a function of V, due to CHAN
[1973]. Although the parameters are relevant to a microwave application
the main features apply in optical waveguides.
101 I I I I I 1 I I I
I
9t
1 2 3 1 5 6 7 8 9
V
Fig. 19. Complex eigenvalue U as a function of normalised frequency V for optical fibre con-
figuration of Fig. 2(b).
358 O P T I C A L FIBRE W A V E G U I D E S [VII. 2
1 2 3 4 5 6 7 8 9
V
a where
AV
=- n 2 3
n-I
VII, 0 21 FIBRES WITH CORE A N D C L A D D I N G 359
ZAV
ui = -
In 1AVI2
n=2
For n > 1 the modes are only weakly attenuated compared to the below
cut-off modes of a dielectric slab waveguide. A physical explanation is
to be found in the ray diagram of Fig. 21. There we find that weakly leaky
modes correspond to rays which satisfy 8 > 8, with 6 < 8,. These rays
would appear to be trapped according to geometric optics since the angle
made with the tangent plane 6 , is less than the critical angle for total internal
reflection. However, as demonstrated by SNYDER[19731, MARCATILI
[1973] and MARCUSE [1973b] for cylindrical fibres, it is the angle 0 which
the ray makes with the generatrix of the cylinder which determines whether
a ray is trapped or the corresponding mode is bound.
When both 8 and 6 are greater than 8,, the corresponding ray is refracting
and the leaky mode attenuation is large, see Fig. 20. Such modes are called
leaky refracting modes by SNYDER[1974] and the associated rays are leaky
refracting rays. An anatomy of modes on dielectric structures, due to
SNYDER [1974], is contained in Table 5 .
The field of a weakly leaky mode decays exponentially to a radius rtp
whereafter propagating conditions are found. This behaviour is precisely
analogous to that found in cylindrical antennas where a stored energy
Fig. 21. Ray diagram indicating that some leaky rays in the configuration of Fig. 2(b) are
trapped in the configuration of Fig. 2(a). See also Fig. 6.
360 O P T I C A L FIBRE W A V E G U I D E S [vir. 0 2
TABLE5
Anatomy of modes on dielectric structures
(1) Energy is unattenuated (1) Energy with the structure is attenuated due to radiation, i.e.
as it propagates, ix., the the fields are made up by waves that undergo partial
fields are made up by reflection.
waves that undergo total
internal reflection.
(2) Fields external to the (2) Fields oscillate (radiate) at finite distances from structure.
structure are evanescent.
(3) Characteristics found (3) Characteristics found from certain complex roots of the
from the real roots of eigenvalue equation. On structures that support bound
eigenvalue equation. modes, leaky modes are the bound modes below their
cutoff frequency.
(4) modes approximate the radiation field within and
modes gives bound external to the structure far from the source.
energy of structure.
Examples
1
Exist only on non-finite Exist an all dielectric structures
structures with an index of
refraction greater then their Refracting leaky modes Tunnelling leaky modes
surrounds. Examples:
Cylinder or slab in free space
Attributes
( I ) Radiation appears to (1) Radiation appears to
originate at boundary of originate some distance
the structure, i.e.,the fields from the boundary of the
everywhere external to structure. The fields
the structure have an external to, but immediately
oscillatory behavior. adjacent structure are
evanescent and oscillate at
greater distances from the
structure.
(2) These modes are strongly (2) These modes are weakly
leaky. As I, -P 0, the leaky. As I , -+ 0, the
Examples
leakage vanishes.
where
Qnm= U 2 - V 2 . (55)
Fig. 22 shows the evanescent and propagating regions asso ia d with
weakly leaky rays. As the energy appears to tunnel across the region rtP- Y, ,
Snyder designates such modes and rays as tunnelling modes and tunnelling
rays respectively. Expressions for the attenuation coefficients of leaky
rays are presented by SAMMUT and SNYDER [1975] while Fig. 23 shows
ale, as a function of OlO,.
Although leaky modes can play a significant role in the transport of
power in the configuration of Fig. 2(b), especially if V is very large, it should
be recognised that Fig. 2(b) is only an approximation to that of Fig. 2(a)
and there leaky modes will not prove significant unless n3 > n,. This is
evident in Fig. 21 which shows that a leaky ray of the configuration of
Fig. 2(b) is trapped in that of Fig. 2(a). Because of the large difference
between A = (n,- nz)/nl and A , = $[n; - l]f/n,, a very large number of
modes which would be leaky in Fig. 2(b) are in fact trapped by the outer
cladding boundary and these are the so-called cladding modes of the struc-
ture. However, if n3 is lossy, many of these modes will be highly attenuated.
362 OPTICAL FIBRE W A V E G U I D E S [w 2
Fig. 22(a). Geometrical interpretation of the fields inside and outside of optical fibre con-
figuration of Fig. 2(b) under leaky ray conditions. Ci, denotes interior caustic; C,,, is the
caustic beyond which radiation occurs. The unshaded regions correspond to rays i.e. oscilla-
tory fields.
-2
Fig. 22(b). Field of a leaky mode for the configuration of Fig. 2(b). For r less than rtpr the
field resembles a bound mode. For r greater than rip, energy is radiated. 0 is the direction of
radiation in the far field.
FIBRES W I T H CORE A N D C L A D D I N G 363
Q/Qc
Fig. 23. Diagram showing the U plane for leaky (LR) and refracting (RR) rays.
n 3 s n2
(b)
Fig, 24. Configurations for curved slab and cylindrical optical waveguides. (a) Curved cylindri-
cal fibre; (b) Curved slab waveguide.
Finally, the attenuation coefficient a (= 2pa' where a' is the coefficient used
in SNYDER[19741) is given by
thus
a = +k'(ef - e2)exp -(+k'po(8: - e2)t).
According to SNYDER [19741, for a circular fibre supporting many modes,
only meridional rays which strike the boundary at a radius po suffer signif-
icant bending loss and their attenuation coefficient will be the same as
for a slab waveguide, except for the factor q(8) which can be found from
Table 3.
Near cut-off we can neglect K in the expression for q(8), then on sub-
stituting for (U/V)' from equation (12)
Q/Qc
Fig. 25. Curvature loss as a function of normalised ray angle for graded refractive index and
stepped refractive index planar waveguides.
366 OPTICAL FIBRE W A V E G U I D E S [VII, 2
shows the curvature loss as a function of eje, for slabs of thickness t with
uniform and graded profiles where - and 0 are related through
e2
- = (ez - 2t/p0).
From the above description, we see that it is the energy in rays whick.
are close to the critical angle in the straight fibre, which is lost first when the
fibre is gradually bent. In modal terms, it is those modes which are closest
to cut-off which first couple to the radiation modes and which give rise to
attenuation. Only for a severe bend would low order modes in a multimode
fibre, couple directly to the radiation field.
Mode coupling due to perturbations in optical waveguides including
bends, has been extensively considered by MARCUSE [1969, 1972, 1973a,
1974a) following an initial treatment of the problem by SNYDER [1969a, b].
GLOGE [1973] and MARCUSE [1972a] have shown how mode coupling in-
fluences pulse propagation while SNYDER[1969~1,also CLARRICOATS and
CHAN[19731 have considered how isolated inhomogeneities couple power
between modes and from a mode to the radiation field.
When the core refractive index is perturbed by h ( r , 8',z) the coupling
coefficient between the pth and qth modes is given by MARCUSE [1974a] as
where ep and e, are the linearly polarised vector fields associated with the
modes of the unperturbed fibre as given by the equations of Table 4.
where
P, iP,.
Notice that the power oscillates between the modes with a period which
VIJ, 21 F I B R E S W I T H CORE A N D C L A D D I N G 361
,guided modes
radiation modes I
I
---- I
I
I
I
I I I
f
I l
I ,
l
c
PN 4 P
4 = 27Wp-8,8,)
x 4r,p
where 8 x ep x 8, is the ray angle within the fibre as defined in the previous
section. Equation (66) shows that coupling between modes due to a bend
is inversely dependent on the sixth power of the bend radius.
The effect of random bends in a fibre has been considered by GLOGE
[1972] using a ray treatment and by MARCUSE[1969, 1973a1,who uses
modal methods to show that the power coupled between the pth and 4th
modes Ppqis
PjJq = I ~,,I2F(P,- 8,). (67
In equation (67), the z dependence of K,, (see equation (63)) has beell
removed leaving Kp4 with F(B, -p,) representing the power spectrum of
the z dependent curvature function. Only the spectral component with
wavelength A, given by equation (65) influences coupling between the pth
and qth modes. For modes far from cut-off,
40-~
8
0 0 0
- 300
C=7x10-5radZm VI
(u
0
N
6-
E!
r
-0
0
NL
a - 200
n
E
C
3
0 0 0 0
- 100
2 I I I I I
0 2.5 5.0 7.5 10.0 12.5 15.0
length m
Fig. 27. The output beam angle and total number of modes as a function of length in a multi-
mode fibre which propagates 700 modes. The straight line represents a best fit to a theory
which assumes mode coupling arises due to random bends. C is the curvature spectrum func-
tion.
total mode number, varies with length along a fibre where coupling of
neighbouring modes is due mainly to random bends. The straight line
represents a best fit to theory in which the coefficient, labelled C by GLWE
[1972aJ, is the curvature spectrum function. C is given by
where 6 is the amplitude of sine waves sin (2nz/A) used to model the random
curvature and v] the number of undulations per unit length.
The above results refer to a multimode fibre. In a single mode fibre,
radiation loss can occur due to bends, core diameter fluctuations, or on
account of inhomogeneities, while coupling to cladding modes can also
occur. Fig. 28, due to CLARRICOATS and CHAN[1973], shows for a mono-
mode fibre with finite cladding boundary, how the normalised power
which is.both coupled into the cladding modes from the HE, mode, and
is transformed into radiation, varies with the location of an isolated in-
homogeneity. The curve labelled HE,, , shows the power coupled back into
the HE,, mode by the inhomogeneitv. Following SNYDER[1969a, b], the
valid assumption is made that equal amounts scatter in both forward and
backward directions. Evidently, coupling to radiation predominates over
coupling to the cladding modes. The latter varies with r/rl in a manner
"11, 21 FIBRES W I T H C O R E AND CLADDING 369
rlr
Fig. 28. Normalised coupling coefficient between .HE,, mode and other modes of optical
fibre, in the configuration of Fig. 2(a), as a function of the position of an inhomogeneity of
volume V and permittivity difference A&. Also shown is the coupling into radiated power.
where the first term on the right-hand side arises from attenuation effects
at the cladding and core cladding interface, which increase as 0. The
coefficient A is measured in m-l rad-; 0 independent loss is omitted
but this can be incorporated in the final solution. The second term is
associated with the differential group delay according to ray optics. The
third term arises from mode coupling between closely adjacent modes
and takes the form of a diffusion process in the ray picture. D = (4/n2)C,
see equation (69), is the coupling coefficient which is assumed in the follow-
ing to be independent of 6. Gloge also discusses the influence of a 8 depen-
dent coupling coefficient. Equation (70) may be solved using the Laplace
transformation
(71)
then equation (70) becomes
0, = (4D/A)* (77)
yrn = (4DA)*. (78)
VI1, 6 21 FIBRES W I T H CORE A N D C L A D D I N G 371
The denominator expresses the loss in the short distance z, while the terms
in the numerator indicate that the distribution has been preserved and the
delay is precisely that predicted by ray optics in the absence of coupling.
By integratingp(8, z, s) with respect to 8 and taking the Laplace transform,
the impulse response Q(z, t ) of the fibre is obtained as
Fig. 29. Imuulse resuonse according to eauation (81). Normalised for eaual ueak values and
plotted-as a function of norkalised time for different fibre iengths.
1.2-
1.0 -
0.8 -
c
cp
L
0
t 0.6-
10
0.4 -
u
0.2 04 0.6 0.8 1-0 1. 2 1.4 1.6
Y -2
Fig. 30. Delay and time spread of the fibre output on-axis and at an angle 0 = 0,.
FIBRES WITH CORE A N D C L A D D I N G 373
Y w2
Fig. 31. Relative width of the impulse response as a function of the normalised fibre length.
The two straight lines show the asymptotic behaviour for very short and very long fibres.
and reveals the important and desirable result that in the presence of
coupling in long fibres, the pulse dispersion increases only as L*.
In the above treatment, it has been assumed that the steady state distribu-
tion at z = co is such that 0, < 8, ,i.e. that the mode angle corresponding
to the 1/e point in the Gaussian power distribution is much less than the
complement of the critical angle. If this is not the case, the steady state
distribution is generally determined by a sharply rising loss term at
8, =8, rather than by the quadratic term A8 as in equation (70).
Although there will be an increase in overall attenuation, the pulse width
can then be smaller by a factor of several times compared to that shown
in Fig. 31.
374 OPTICAL FIBRE W A V E G U I D E S [VII, 52
2.15. REDUCTION OF PULSE DISPERSION BY INTENTIONAL INHOMOGE-
NEITIES
MARCUSE [1974b] has made a very detailed study of the compromise
between pulse width reduction and the excess loss caused by mode coupling.
He has also explored the possibility of providing a selective coupling
mechanism which would prevent all but a minimal amount of power from
escaping due to coupling between high order modes and the radiation
field. This is possible, in principle, if refractive index fluctuations could
be created with a spectrum function possessing a sharp cut-off at high
values of (pp-fi4), i.e. at short beat wavelength A . Since power in modes
with high azimuthal number v, is confined to regions of the core such
that r > rmax= vmaxrl/V, the influence of the fluctuations on these modes
can be further reduced if they are confined to the region Y < rmax. By
assuming a flat spectrum between A = 00 and A = 0.13 cm; MARCUSE
[1974b1 predicts a ten fold reduction in pulse width for a 1 km fibre length
(compared to the case without coupling). For this calculation V = 40,
A= [((An)2)]* = and rmax/rl= 0.8. Although such improve-
ments are spectacular, they have not yet been achieved experimentally be-
cause of the difficulties associated with imposing controlled refractive index
fluctuations during manufacture as discussed by MARCUSEand -BY
[1975]. Differences in mode coupling in fibres with similar parameters are
evident and these are probably associatedwith differences in the random fluc
tuations which are always present in fibres. They occur for example, when
fibres are pulled using the double crucible method of manufacture especially
01 1 10 100 200
modulus of drum surface in kg I mm
Fig. 32. Distortion loss as a function of drum surface modulus. Fibre diameter = 120 pm.
Core diameter = 80 pm. Relative index difference = 0.5%. rms drum surface roughness = lpm.
Correlation distance = 1 mm. Mean lateral force per unit length is the parameter.
VIL 5 21 FIBRES W I T H CORE A N D C L A D D I N G 375
2.16. EXCITATION
2.16.1. Incoherent
The excitation of an optical fibre by an incoherent source was first
considered by POTTER[1961] and more recently by PASKand SNYDER
[1974]. The latter authors have also considered the influence of leaky ray
attenuation on the power carried to the output of the fibre.
If the source illuminates the whole end face of the fibre and the angular
distribution of rays Z(8,O) is given by
r(e, 0)
= z, COS e (83)
corresponding to a Lambertian source, the power in the trapped rays
-=
satisfying 8 8, is
PTR= nr: I. sin2 8,. (84)
The total power radiated is n2r:Zo thus the ratio
PTR/PT
= sin 0, x 8: = 24. (85)
However, rays which, according to Fig. 6, satisfy 8 > 8,, S < 8, also trans-
port power, only with increasing attenuation as 6 -+8,. For leaky rays,
Because the leaky rays are attenuated, only a percentage of the total power
guided by the fibre at z = 0, remains after distance z. Fig. 33 due to PASK
and SNYDER [1974] shows how the percentage varies with z/rl and I/. For
a typical multimode fibre with V = 50, and 8, = 0.1, about 10% of the
376 OPTICAL FIBRE WAVEGUIDES L W , 9; 2
p! geometric optics
v- 200
zlr,
Fig. 33. Fraction of incident power within fibre as a function of fibre length normalised to
fibre radius. Parameter is the V value for the fibre.
2.16.2. Coherent
We have just seen that an incoherent source with illumination angle
exceeding the acceptance angle of the fibre, excites all the bound modes
with near uniform intensity. By contrast, a coherent source may be arranged
to excite one mode efficiently although, in the case of a multimode fibre, this
generally calls for the use of a lens. Insight into the general problem of
coherent excitation is provided by considering the case of on-axis excitation
by a parallel beam of monochromatic energy. This situation approximates
very well a focussed gas laser and, with a compound-lens, that of a solid-
state laser. The problem was first studied in the present form by SNYDER
[1966] following earlier work on the coupling of fibres by JONES[1965].
vn, 21 FIBRES W I T H C O R E A N D C L A D D I N G 311
TABLE
6
Summary of results of analyses for disk and strip geometries with and without a lens
(YANG and KINGSLEY [19751)*
31 pW 0.031 W
D
Rectangular strip 2.16 B1 TDn, tan ( 0 , / 2 ) JV -no cot 0,
2
width = D
D
length = no - cot O1
2
Fiber axis in center of strip 0.4 mW 6.2 W
3s B1 Tn, FD tan O1 F3
Focused rectangular strip - J V . 8no - tan2 0,
2 (1-(2F tan O,/D)) D
width = 2F tan 0,
4n0 F tan 0,
length = 0.26 mW 0.28 w
D(1-(2F tan OJD))
Lens, fiber coaxial with center
of strip
* The numerical values of the coupled powers and the electrical power input required assume
a radiance of 30 W(ster. cm) for a power input, J K of 2 x lo4 W/cm2, and a fiber having a
numerical aperture of 0.14, or an acceptance half angle of 8. T = transmission coefficient,
B , = radiant power per unit sold angle.
The inset to Fig. 34(a) shows the configuration and for the present, we
consider the angle of incidence 8, to be zero and ri = rl. The field just
inside the fibre, at z = 0, is assumed to have the same distribution as in the
incident wave. We can then expand in terms of the normal modes and on
applying orthogonality, obtain an expression for the amplitude coefficients
of the modes:
E = cage4 where E = E,$, r < rl (87)
4
= 0, r > r,
H = Ca,h,. (88)
4
378 O P T I C A L F I B R E WAVEGUIDES
Fig. 34. Excitation of an optical fibre in the configuration of Fig. 2(a). (a) Normal incidence
excitation. Fibre parameters as Fig. 7. ~ rz/rl = 5, --- r z / r l = 15, ----- core
mode approximation.
Let
J.'px h:. ds = sq
and
then
aq = sols,
FIBRES W I T H CORE A N D C L A D D I N G 379
bEfl
10 *
/ (b)
(b) Excitation efficiency as a function of angular incidence for cylindrical beam of uniform
intensity incident at an oblique angle to the axis. V = 10; ri/ri = 1 . Fibre parameters as in
Fig. I .
380 O P T I C A L FIBRE W A V E G U I D E S [vn, 0 2
and the ratio, power in qth mode/incident power
- 4w2
-- for HE,, modes
u2v2'
-= 0, for EH1, modes.
For HE,, modes well above cut-off
w w v,
then
For off-axis incidence, HE,, and EH,, modes are excited since the incident
plane wave gives rise to a phase-variation about the azimuthal coordinate
0'. The above method can be applied, although the integrals are more
complicated to evaluate. SNYDER [1966] gives a result when r2 = co and
n, < = n, which leads to equation (99), while CHAN[1972] has under-
taken a more general study including the case when r2 is finite corresponding
to the inset to Fig. 34(a):
VII, Q 31 F I B R E S WITH N O N - U N I F O R M R E F R A C T I V E I N D E X 38 1
8 is the ray direction associated with the qth mode in the fibre while 8, is
the direction of the refracted incident ray within the fibre.
When 8 =Or, equation (99) simplifies to give
We see from equation (100) and Fig. 34(b) thatthe peak excitation occurs
when the refracted ray direction and modal ray directions coincide and
that the intensity decreases with increasing n. Fig. 34 also quite accurately
describes the radiation pattern of a fibre excited in the appropriate mode.
The above results refer to a uniform intensity input beam. The Gaussian
beam at normal incidence was first considered by STERNand DYOTT [1971]
while the oblique incidence case has been studied by CHAN[1972], MAR-
CUSE [1970] also IMAI and HARA [1975]. Although excitation efficiencies
are in general slightly higher than for the uniform beam case, as the input
and modal fields more closely correspond, the behaviour is otherwise very
similar.
3.1. INTRODUCTION
In the preceding section we have seen that single mode fibres possess
a large bandwidth which is limited mainly by material dispersion. They
suffer the disadvantage that they are difficult to join without appreciable
coupling loss because the power carrying region is only a few microns in
diameter. They are also completely unsuitable for excitation by light
emitting diodes (LEDS). These disadvantages are largely overcome in
multimode fibres but with a substantial bandwidth penalty arising mainly
from modal dispersion. While mode coupling can limit this dispersion,
controlled mode coupling, which provides optimum coupling without
additional loss, has yet to be achieved experimentally. Efficient single mode
propagation may be achieved through a potentially multimode fibre by
using a lens to improve coupling from a much smaller laser aperture, but
such a configuration may prove hard to produce in a fully engineered
system.
Several alternative possibilities to the above fibre types have been
investigated with a view to overcoming most of the above disadvantages.
Among those with circularly symmetric refractive index profiles, two
main types have emerged, both 9s a result of initial development in Japan.
382 O P T I C A L FIBRE W A V E G U I D E S CVIL 0 3
These are the so-called W-fibre, which has the profile displayed in Fig. l(c)
and which can exhibit monomode properties. The other has a continuously
graded core refractive-index as shown in Fig. l(e); this fibre exhibits multi-
mode properties but with inherent delay equalisation properties. In ray
terms, the fibre has self-focussing properties and the registered trade
name SELFOC has been applied to such fibres. Fibres manufactured using
different techniques have refractive-index profiles which differ from the
near parabolic profile but these can have quite satisfactory bandwidth
properties, an example is shown in Fig. l(Q.
The double-clad fibre with a low index inner cladding (named the W-fibre
because of its refractive index distribution) has been extensively investigated
by KAWAKAMI and NISHIDA [1974, 19751. In common with the single-clad
fibre with finite cladding radius, see Fig. 2(a), the dominant mode trans-
forms from a core mode to a cladding mode at a non-zero Vvalue. However,
a) i )
Fig. 35. Characteristics of the eigenvalues and fields in a doubly clad (W) fibre. (ai) Was a
function of U.Parameter 6 = r2 - rl . (aii) W as a function of U,showing shift of higher mode
,
cut-off. Note that for the HE, mode Wis substantially greater at higher mode cut-off value of
V for W fibre as compared with singly clad fibre. The field is therefore more tightly confined
to the inner core. (b) & (c) Transverse field distributions within singly clad (b) and doubly
clad (c) fibres.
VII, 31 FIBRES W I T H N O N - U N 1 F O R M REFRACT IVE I N D E X 383
Thus the field in the doubly-clad fibre is contained more closely to the
the core than in the singly-clad case which reduces the attenuation if the
cladding has a higher loss than the core; it also minimises bending loss.
Fig. 35(a) shows that the value of V = U at cut-off for the TE,, mode
(which is degenerate with the TMol and HEzl modes) occurs at a value
of 3.5 for the W-fibre as compared to 2.4 for a singly-clad fibre with the
same core and outer-cladding refractive indices. This implies that a larger
core radius can be chosen for a W-fibre which eases the problem of coupling
from a laser.
radius r -
Fig. 36. Refractive index profiles according to equation (101).
where the constant I in equations (103) and (1 05) 'corresponding to the ray
period, has been rewritten as a/(2d)*. Then, if a/(24)* B 1, the wave
equation becomes
= 0. (108)
By separating the variables
386 OPTICAL FIBRE WAVEGUIDES
and
where
K 2 = o2&,p0ng-fl2
E Po
p 2 = 2A02 L n ; ,
a2
Whend=Oandif*+Oasr+oo,
K 2 = 02eopOng-f12 = k2- B2
k
= 2 - (2A)*(2m
a
+n + 1).
LL is a Laguerre polynomial and the functions t,4n,, form an orthogonal
set with n, m = 0, 1, 2, . . .. First order perturbation theory then leads to
2k 6
K2 =
a
+
-(2A) *(2m n+ 1)-
a
+ + +
(2A)[6m2 +6m(n+ 1) (n l)(n 2)], (1 16)
V
ooooo results due to Kirchoff
results obtained by a five steps stair
case approximation of the permittivity profile
I I 1 I I I I 1
I 1 2 3 L 5 6 7 0
V
0 0 0 0 0 results due t o Kirchoff
- results obtained by a five steps stair case
approximation of the permittivity profile
Fig. 38. Normalised group velocity as a function of normalised frequency V for graded refrac-
tive index fibre with parameters as for Fig. 37.
VII, 5 31 FIBRES W I T H N O N - U N I F O R M REFRACTIVE I N D E X 389
Fig. 39. Propagation regions for a graded refractive index fibre of radius a . (a) R , corresponds
to the inner ray caustic R , the turning point for the ray. (b) Conditions corresponding to a
leaky ray within a graded refractive index fibre. Propagating conditions occur between R ,
and R , and beyond R , . (c) Conditions corresponding to refracting rays.
outside of the region R , < r < R,, the phase of the field within the region
rnn = JR2u(r)dr =
R1
r2
must sum to an (approximately) integer number of half. periods so that,
Rt
F ( r )- B2 - r2
4[[ [k'-/?:-
(k2-8E18
M, = z o o gpndr (122)
where a is the radius at which n(r) = n,. On integrating (122) with respect
to n we obtain
M, =
L [k2(r)-P,"]rdr = k2
l [n'(r)-n;]rdr.
For small index differences the integral represents the volume under the
(123)
where the upper limit in the integration, R2(0), is the radius at which
kn(r) = B. For fibres with refractive index profiles of the form of equations
(101) and (102) shown in Fig. 36,
392 OPTICAL FIBRE WAVEGUIDES [VlI. P 3
where
ki = k2ni.
2)l+
7, =-
C
a-2-Y A (~r+"+ 301-2-2y
+
2(a 2)
+ O(A3)] (130)
where
In the limit that the derivatives of no and A vanish, equation (132) reduces
to the form given originally by GLOGEand MARCATILI [1973al; we continue
with that assumption for the moment.
To obtain the impulse response if all modes are excited equally, we
need to know the density of modes arriving during an element of time dt,
where t is the incremental time associated with the Mth mode after removal
VII, 5 31 F I B R E S W I T H NON-UNIFORM R E F R A C T I V E I N D E X 393
while
a-2
O<t<- A, except for a x 2 ( 136)
a+2
+ + 4 4 A2(2a+2)*]1
4c, c2 d(a 1)
(141)
(2a+ 1) (5a + 2)(3a+ 2)
394 O P T I C A L FIBRE WAVEGUIDES
A*
-7
Fig. 40. The impulse response for graded refractive index fibres with refractive index profiles
corresponding to Fig. 36.
where
a-2-y 3~ - 2- 2y
c1 = 9 c2 =
a+2 2(a+2) '
~
-$ :material limit I
1;
2
a
m
E
-
- '
11
'source wavelength
900nm
L -
\I wave guide n.a.0.16
0.01
I
-- I
-- 1
1 , 1 . 1 # 1 , 1 . 1 . 1 ,
Fig. 41. r.m.s. pulse width as a function of refractive index parameter u. _ _ result obtained
assuming that the refractive index and refractive difference are independent of wavelength;
_--__ a result obtained from equation (141) and for a fibre with dispersion characteristics
of a TiO, doped silica core with a silica cladding at a wavelength of 0.9 p.
source wavelength and find for the above fibre, a monotomic decrease
between 2.5 and 2.2 in the range 0.5-1.1 p. Also shown in Fig. 41 are
the limits on pulse broadening imposed by the finite linewidth of the
source for a typical LED and GaAs laser. Evidently there is a strong
incentive to manufacture a fibre with near optimum a if a narrow line-
width source is to be used.
STEINBERG [1974] has shown that for a source of finite bandwidth B
and with neglect of material dispersion,
= A~B, 2. (144) a =
In the above analysis, it has been assumed that all modes are excited
equally. TIMMERMANN c1974b-J and JACOMME[1975] have considered the
more general problem where a source with an arbitrary distribution of
power as a function of solid angle excites the modes of the graded-index
fibre.
396 O P T I C A L FIBRE W A V E G U I D E S CVK 03
3.4. FIBRES WITH RING-SHAPED REFRACTIVE INDEX PROFILES
radius r
Fig. 42. Refractive index profile for a ring shaped distribution.
VII, P 31 FIBRES W I T H N O N - U N I F O R M REFRACTIVE I N D E X 397
+=k,R, (147)
and
k, = kn,.
The axial velocity of the ray is given by
C
u=-cos~=-
n1 n,
1--
'[ I: .
On substituting for 4 from equation (147), the time for the ray to progress
through a length L of the fibre is then
[1973a1provided
which agrees with the derivation of GLoGE and MARCATILI
a.
nz zs- The maximum value of 4 is (2A)* so
v,: =~A(/C,R~)~. (1 50)
The incremental delay t = T - LnJc, is then given by
mode number is the same as for a slab with parabolic refractive index pro-
file namely Sp = (2A/a)*,then
dp k,a (24-4')
mmax= - - ~
(153)
Sp - 2 (2A)t
= (;Pea [ 1- ($)I.
In passing we note that, on summing equation (154) between v = 0 and
v = v,, , we obtain the expression for the total number of modes given by
GLOGEand MARCATILI [1973a1, namely,
&zkgR , A . (155)
If allowance is made for degeneracy of the modes and the two states of
polarisation the total mode number is four times the above value. Finally,
we obtain the impulse response from equation (1 54) as
mdv - kgR,ac [ ( __
dt 2L0 n
L i l y - __
ct
Ln1
( >'3 (156)
The function has been plotted by GLOGE and MARCATILI [1973a] in Fig. 43,
who have also obtained results for the case when Ro/ais not large. It is easily
I I
0.2 04
relative delay TC/A n1 L
Fig. 43. Impulse response for a graded refractive index fibre with profile corresponding to
Fig. 42,
VII, 0 31 FIBRES WITH N O N - U N I F O R M R E F R A C T I V E I N D E X 399
shown that half the power in the output pulse arrives in a time
t = 0.124~1,
L/c (157)
the remainder being drawn out into a tail of length An, L/c, caused by
the higher order azimuthally dependent modes.
3.5. CONCLUSIONS
Acknowledgements
The author of this review is indebted to many for their helpful suggestions
over the years on the subject of optical waveguides. He thanks especially
Dick Dyott who gave considerable help to the author in the preparation
of this review. He also thanks Leo Felsen, Charles Kao, John Midwinter,
Murray Ramsay, Charlie Sandbank and Allan Snyder for their helpful
comments.
References
ABRAM,R. A., R. W. ALLANand R. C. GOODFELLOW, 1975, The coupling of light emitting
diodes to optical fibers using sphere lenses, to be published.
ARNAUD, J. A,, 1974, Bell Syst. Tech. J. 53, 675.
ARNAUD, J. A,. 1975, Bell Syst. Tech. J., to be published; Electron. Letters, 11, 8.
BIERNSON, G. and D. J. KINGSLEY, 1965, IEEE Trans. on Microwave Theory and Techniques
MTT-13, 345.
CHAN,K. B., 1972, Electromagnetic wave propagation in optical waveguides, University of
London Ph.D. Thesis.
CHAN,K. B. and P. J. B. CLARRICOATS, 1973, Queen Mary College Technical Report EE/73/10.
CHECCACCI, P. F., R. FALCIAI and A. M. SCHEGGI,1975, Resume of Colloquium on the
Optics of Guided Waves, Paris.
CLARRICOATS, P. J. B., 1960. Proc. IEE, Part C, Monograph No. 409E and 410E.
CLARRICOATS, P. J. B., 1970, Lecture Coursenotes, University of Colorado.
CLARRICOATS, P. J. B. and K. B. CHAN,1970, Electron. Letters 6. 695.
CLAKRICOATS, P. J. B. and K. B. CHAN,1973, Proc. IEE 120, 1371.
CLARRICOATS, P. J. B. and D. C. CHANG,1972. Proc. G-AP Intern. Symp., p. 296.
COLLIN,R. E., 1960, Fleld Theory of Guided Waves (McGraw Hill).
DYOTT,R. B. and J. R. STERN,1970, IEE Conf. Publication No. 71, pp. 176-181.
DYOTT,R. B. and J. R. STERN,1971, Electron. Letters 7,624.
DYOTT,R. B., C. R. DAYand M. C. BRAIN,1973, Electron. Letters 9, 288.
DYOTT.R. B.. 1974. The disperslon limited bandwidth of fibre made from optical glasses
(unpublished memorandum).
ELSASSER. W. M.,1949. J. Appl. Phys. 20, 1193.
ENOCH,J . M., 1961. J. Opt. SOC.Am. 51, 499.
FLETCHER, A,, T. MURPHY and A. YOUNG,1954. Proc. Roy. SOC.223, 216.
GAMBLING, W. A., D. N. PAYNEand H. SUNAK,1971, Electron. Letters 7, 549.
GAMBLING, W. A., J. P. DAKIN,D. N. PAYNE and H. SUNAK,1972, Electron. Letters 8, 260.
GAMBLING, W. A., D. N. PAYNEand H. MATSUMURA, 1974, Electron. Letters 10, 148.
GARDNER, W. B., 1975. Bell Syst. Tech. J. 54, 457.
GLOGE,D., 1971a, Appl. Optics 10, 2252.
GLOGE,D., 1971b, Appl. Optics 10, 2443.
GLOGE,D., 1972a, Bell Syst. Tech. J. 51, 1767.
GLOGE,D., 1972b, Appl. Optics 11, 2506.
GLOGE,D., 1973, Bell Syst. Tech. J. 52, 801.
GLOGE,D. and E. A. 3. MARCATILI, 1973a, Bell Syst. Tech. J. 52. 1161.
GLOGE.D. and E. A. J. MARCATILI, 1973b, Bell Syst. Tech. J. 52, 1563.
GI.OGE,D., 1975, Bell Syst. Tech. J. 54, 243.
HONDROS,D. and P. DEBYE,1910, Ann. Phys. 32, 465.
HUBBARD, W. M., 1972. Appl. Optics 11. 2495.
VII] REFERENCES 40 I
SANDBANK, C. P., 1975, Proc. Topical Meeting on Optical Fiber Transmission, Williamsburg.
SCHELKUNOFF, S. A,, 1943, Electromagnetic Waves (Van Nostrand).
SMITH,L. and E. SNITZER,1973, Appl. Optics 12, 1592.
SNITZER, E. and H. OSTERBERG, 1961, J. Opt. SOC.Am. 51, 499.
SNYDER, A. W., 1966, J. Opt. SOC.Am. 56, 601.
SNYDER, A. W., 1969a, IEEE Trans. Microwave Theory and Techniques MlT-17, 1130.
SNYDER, A. W., 1969b, Ph.D. Thesis, University of London.
SNYDER, A. W., 1969c, IEEE Trans. Microwave Theory and Techniques, MlT-17, 1138.
SNYDER, A. W., 1973, Electron. Letters 9, 437.
SNYDER. A. W. and D. J. MITCHELL, 1974, Electron. Letters 10, 11.
SNYDER, A. W., 1974, Appl. Phys. 4, 273.
SNYDER, A. W. and J. D. LOVE;1975, IEEE Trans. Microwave Theory and Techniques 23,134.
STANDLEY, R. D. and W. S. HOLDEN,1974, Bell Syst. Tech. J. 53, 779.
STEINBERG. R. A., 1974. Electron. Letters 10, 375.
STERN,J. R. and R. B. DYOTT,1971, Electron. Letters 7, 52.
STRATTON, J . A.. 1941, Electromagnetic Theory (McGraw Hill).
STREIFER, W. and K. B. PAXTON,1971, Appl. Optics, 10, 769 and 116.
TIMMERMANN, C. C., 1974a, A.E.U. 28, 144.
TIMMERMANN. C. C., 1974b. A.E.U. 28, 186.
TYNES, A. R., 1974, J . Opt. SOC.Am. 64,
UNGER.H. G.. 1964, A.E.U. 19, 189.
VIGANTS,A. and S. P. SCHLESINGER, 1962, IRE Trans. G-MTT, p. 375.
YANG,K. H. and J. D. KINGSLEY, 1975, Appl. Optics 14, 288.
YIP, G. L. and Y. H. AHMEW,1974, Proc. URSI Symp. on Electromagnetic Wave Theory,
p. 272.
YIP, G. L. and J. C. Y. HUANG,1975, Resume Colloquium on Optical Waveguides, Paris.
AUTHOR INDEX
W Y
A -, speckle, 29
axisymmetry, 67
Abbe sine condition, 235
aberration free lens, 8, 35 B
aberrations, 57
- in gratings, 202 ballistic approximation, 270
- introduced by blank, 21 1 band bending, 299
absorption coefficient, 282 bandwith, spectral-. 55, 64
actinides, 92 baselines, 73
active phasing, 78 beamsplitter, 66 et seq.
age theory, 273 beat wavelength, 367
Airy disk, 3, 54 et seq. Beckman-model, 6 et seq.
alkali halides, 290 bends in fibres, 363 et seq.
angular acceptance, 164 binary stars, 20, 53, 57, 70
- dispersion, 198 birefringence, 174
anisotropic media, 258 bit rate, 347
annealing, 315 blaze waveleugth, 2 0
anomalies, 23 1 Bloch states, 254 et seq.
anomalous structure, 316 Bohr radius, 138
aperture shape, 56 Boltzmann statistics, 146
apertures, large-, 60 Born-approximation, 32
aplanatic system, 235 Born-Oppenheimer wavefunctions, 1 16
Apollonius' circle, 240 borosilicate fibres, 330
apparent stellar diameters, 50 Bose-Einstein average, 120
array of fibres, 187 branching ratios, 1 13
- - photodiodes, 185, 188 bright speckles, 59, 78
array of telescopes, 49, 62, 79 et seq. Brillouin zones, 254 et seq.
aspherical waves, 241 Brownian motion, 52
astronomy, 20, 36, 49 et seq.
asymmetrical groove profiles, 219, 231 C
atmosphere, 19
- optics, 51 carbon disulfide, 165 et seq.
- phase shift, 61 Cassegrain telescopes, 80
- scattering, 4 CdS, reflectivity of-, 258
attenuation Ce3+, 150
- coefficient, 365 et seq. cell
- in lossy layers, 354 et seq. -, seeing, 54
- length, 263 -, phase, 55
Auger electrons, 267, 279, 304, 315 central limit theorem, 6, 35
autocorrelation, 4, 9 et seq., 59 centrosymmetric objects, 3, 77
characteristic curve of photoresist, 21 5
412
SUBJECT I NDE X 41 3
E F
echelle gratings, 203 fibre
echelon technique, 164, 185 et seq. -, double clad, 382
ED, 106 er seq. - bandwidth, 347 et seq.
EDD, EDQ, 136 - optics, 84, 186, 329 et seq.
efficiency, 228, 231 - with uniform index, 331 et seq.
eigenvalues, 16 - with nonuniform index, 381 et seq.
eigenvalues of modes, 333 et seq. -, square, 386
Einstein relations, 105 -, W, 382
electric displacement, 251 field components in fibres, 350 et seq.
electric field amplitude, 25 1 filter, linear, 12
electron Fizeau aperture, 60
- bombardment, 291, 304 - interferometer, 64, 66
- diffraction, 276 flight, light photographed in, 177
- hole pair creation, 253, 264 fluctuation
- hole recombination, 268 -, density, 52
- spectroscopy, 247 -, intensity, 10
-. one-approximation, 254 -, partial pressure, 52
-, (quasi) free-gas, 253 -, temporal, 4
-, scattered, 302 fluorescence
-, secondary, 302 -, chromium, 149
electronographic photographs, 57 - decay time, 164, 183, 191
electrostatic interaction, 135 - lamps, 151
elongation of speckles, 31 - quencing, 134
energy density required, 216 Forster-Dexter theory, 134
- levels of rare-earths, 98 Fourier transform, 6 et seq., 58 et seq.
entropy, 7 - space, 62
EPR, 134 - synthesis method, 221
EQ, 109 Fraunhofer plane, 5 et seq.
EQQ, 139 Fredholm equation, 16
Er, 113 frequency, normalised, 333 et seq.
ESCA, 247, 303 Fresnel equations, 256
escape fringes, spacing o f , 202
-. elastic-depth, 302 et seq. frustrated total reflection, 260, 289
- depth, 248 et seq., 284, 288
- depth trends, 319 G
- probability, 248 et seq., 263
etalon, oven stabalized, 214 GaAs, 295 et seq.
etendue, 242 GaAs laser. 346
- efficiency, 379 gain, resolution, 60
- of fibres, 375 gamma distribution, 16
Eu3, 143, 150 gas laser, 376
evanescent region, 361 gated picture ranging, 164, 180 et seq.
- waves, 260 gating
exchange interaction, 136 -, collinear, 174
excitation in rare-earths, 102 - in glass, 173
excited state lifetime, 104 -, transverse, 176 er seq.
experiments on Gaussian, 55
- - cooperative relaxation, 142 et seq. - distribution, 5
- - escape depth, 219 et seq. - scanning aperture, 14 et seq.
- - multiphonon relaxation, 124 et seq. - spread function, 60
- - radiative decay, 111 et seq. ghosts, 226, 242
SUBJECT I N D E X 415
glasses, 131 index variation, parabolic, 384
Goodman model, 6 infrared, near, 57
graded index fibres, 329, 351, 383 - upconversion, 97
grass, 226 inhomogeneities, intentional, 374
grating intensity interferometer, 50, 82 ef seq.
-, concave, 235 interference
-, echelle, 203 -,diffraction, 58
- blank, 200 - fringe system, 201
-equation, 198 - grating, 200
- interferometer, 224 interferometer
- with imaging, 235 -, direct, 59 ef seq.
-, holographic, 197 et seq. -, heterodyne, 84
-, metallic, 260 -, intensity, 82 et seq.
-, special types Of, 239 -, triple, 78
-, tripartite, 236 internal modes of molecular groups, 123
grazing incidence, 219 intramultiplet transitions, 110
Green functions, 274 instrumental profile, 224 et seq.
groove profiles of gratings, 219, 229 ion
grooves, nonuniform, 235 - ion energy transfer, 133 et seq.
group delay, 342 et seq., 392 - lattice interactions, 91, 116
- theory, 101 - pair energy transfer, 96
-velocity, 174, 264, 342, 388 - phonon coupling, 119 et seq.
isoplanatic patch, 57, 60
- spread function, 12
H
Hamiltonian J
-, crystal field, 117
-, ion pair coupling, 138 Judd-Ofelt theory, 106
helical rays in fibres, 385
Hermitian coherence matrix, 34
heterogeneous wave, 252 K
heterodyning, 50, 84
Hg-line, hyperfine structure, 225 Kerr shutter, 162 er seq.
history of rare-earths, 93 Ketteler-Helmholtz formula. 253
H O ~ +125,
, 154 knife-edge test. 53
holographic diffraction gratings, 197 ef seq. Kramers-Kronig relations, 252
- - -, theory of, 198 et seq.
- - _ , principles of, 200 et seq.
honeycomb pattern, 57,62 L
host dependence, 129 ei seq.
humidity, 52 La, 124, 129, 149
hyperfne structure of Hg line, 225 laminar profiles, 149, 219
hypersensitivity, 114 lanthanides, 92
Laplace transformation, 370
Laportes rule, 106
I laser, 95, 154 et seq, 346
-, display of ultrashort-pulses, 179
IH approach, 140 -, gas, 376
image intensifier, 69 - excitation in rare-earths, 150
imaging properties of gratings, 235, 242 - speckle, 55, 56
impedance, 259 lattice vibrations, 116 et seq.
impulse response in fibres, 371, 387 leaky modes, 357 ef seq.
416 SUBJECT INDEX
ABEL~S, F., Methods for Determining Optical Parameters of Thin Film 11, 249
ABELLA, I. D., Echoes at Optical Frequencies VII, 139
AGARWAL, G. S., Master Equation Methods in Quantum Optics XI, 1
AGRANOVICH, V. M., V. L. GINZBURG, Crystal Optics with Spatial Dispersion IX, 235
ALLEN,L., D. G. C. JONES, Mode Locking in Gas Lasers IX, 179
AMMANN, E. O., Synthesis of Optical Birefringent Networks IX, 123
ARMSTRONG, J. A., A. W. SMITH,Experimental Studies of Intensity Fluctuations in
Lasers VI, 21 1
ARNAUD, J. A,, Hamiltonian Theory of Beam Mode Propagation XI, 245
BALTES, H. P., On the Validity of Kirchhoff s Law of Heat Radiation for a Body in a
Nonequilibrium Environment XIII, 1
BARAKAT, R., The Intensity Distribution and Total Illumination of Aberration-
Free Diffraction Images I, 67
BASHKIN, S., Beam-Foil Spectroscopy XII, 287
BECKMANN, P., Scattering of Light by Rough Surfaces VI, 53
BLOOM, A. L., Gas Lasers and their Application to Precise Length Measurements IX, 1
BOUSQUET, P., see P. Rouard IV, 145
BRYNGDAHL, O., Applications of Shearing Interferometry IV, 37
BRYNGDAHL, O., Evanescent Waves in Optical Imaging XI, 167
BURCH,J. M., The Metrological Applications of Diffraction Gratings 11, 73
CHRISTENSEN, J. L., see W. M. Rosenblum XIII, 69
CI ARRICOATS. P.J. B.. Optical Fibre Wiiveyuides A Review X I V . 327
COHEN-TANNOUDJI, C., A. KASTLER, Optical Pumping v, 1
CREW, A. V., Production of Electron Probes Using a Field Emission Source XI, 221
CUMMINS. H. Z., H. L. SWINNEY, Light Beating Spectroscopy VTII, 133
DAINTY. J. C.. The Statistics of Speckle Patterns XIV. 1
DECKER Jr., J. A,, see M. Harwit XII, 101
DELANO, E., R. J. PEGIS,Methods of Synthesis for Dielectric Multilayer Filters VII, 67
DEMARIA,A. J., Picosecond Laser Pulses IX, 31
DEXTER, D. L., see D. Y.Smith X, 165
DREXHAGE, K. H.. Interaction of Light with Monomolecular Dye Layers XII. 163
DUciuAY. M. A.. The U l t r a f d Optical Kerr Shutter XIV. 161
EHERLY.J. H.. Interaction of Very Intense Light with Free Electrons VTI. 359
FIORENTINI, A., Dynamic Characteristics of Visual Processes 1, 253
FOCKE, J., Higher Order Aberration Theory IV, 1
FRANCON,M, S. MALLICK,Measurement of the Second Order Degree of Co-
herence VI, 71
FRIEDEN, B. R., Evolution, Design and Extrapolation Methods for Optical Signals,
Based on Use of the Prolate Functions 1X. 311
FRY,G. A., The Optical Performance of the Human Eye v111, 51
GABOR,D., Light and Information I. 109
420
CUMULATIVE INDEX 42 I
R~II)OI.PH D..
. see G. Schmahl XIV. 195
SAKAI, H., see G. A. Vanasse VI. 259
SCHMAHL. G.. D.RUWLPH,Holographic Diffraction Gratings XIV. 195
SCHULZ,G., J . SCHWIDER, Interferometric Testing of Smooth Surfaces XIII, 93
SCHWIDER, J., see G. Schulz XIII, 93
SCULLY, M. O., K. G. WHITNEY, Tools of Theoretical Quantum Optics X, 89
SITTIG,E. K., Elastooptic Light Modulation and Deflection X, 229
SLUSHER, R. E., Self Induced Transparency XII, 53
SMITH,A. W., see J. A. Armstrong VI, 211
SMITH,D. Y., D. L. DEXTER, Optical Absorption Strength of Defects in Insulators X, 165
SMITH,R. W., The Use of Image Tubes as Shutters x , 45
SODHA,M. S., A. K. GHATAK, V. K. TRIPATHI,Self Focusing of Laser Beams in Plas-
mas and Semiconductors XIII, 169
STEEL,W. H., Two-Beam Interferometry V, 145
STROHBEHN, J. W., Optical Propagation Through the Turbulent Atmosphere IX, 73
STROKE, G. W., Ruling, Testing and Use of Optical Gratings for High-Resolution
Spectroscopy 11, 1
SVELTO,0,, Self-Focusing, Self-Trapping, and Self-Phase Modulation of Laser
Beams XII, 1
SWINNEY, H. H., see H. Z. CLJMMINS VIII, 133
TAYLOR, C. A., see H. Lipson V, 287
TER-MIKAELLAN, M. L., see A. L. Mikaelian VII. 231
THELEN, A,, see A. Musset VIII, 201
THOMPSON, B. J., Image Formation with Partially Coherent Light VII, 169
TRIPATHI, V. K., see M. S. Sodha XIII, 169
TSUJIUCHI, J., Correction of Optical Images by Compensation of Aberrations and
by Spatial Frequency Filtering 11, 131
UPATNIEKS, J., see E. N. Leith VI, 1
VANASSE, G. A., H. SAKAI,Fourier Spectroscopy VI, 259
VAN HEEL.A. C. S.. Modern Alignment Devices I. 289
VERNIER, P., Photoemission XIV, 245
WEBER,M. J., see L. A. Riseberg XIV, 89
WELFORD, W. T., Aberration Theory of Gratings and Grating Mountings IV, 241
WELFORD, W. T., Aplanatism and Isoplanatism XIII, 267
WHITNEY, K. G., see M. 0. Scully x , 89
WOLTER,H., On Basic Analogies and Principal Differences between Optical and
Electronic Information I, 155
WYNNE,C. G., Field Correctors for Astronomical Telescopes X, 137
YOSHINAGA, H., Recent Developments in Far Infrared XI, 77
YAMMI,K., Design of Zoom Lenses VI, 105
YAMAMOTO, T., Coherence Theory of Source-Size Compensation in Interference
Microscopy VIII, 295
YOSHINAGA, H., Recent Developments in Far Infrared XI, 17