(Methods in Experimental Physics 26) Daniel Malacara (Eds.) - Physical Optics and Light Measurements-Academic Press (1989) PDF

Methods of
Experimental Physics
VOLUME 26
PHYSICAL OPTICS AND LIGHT MEASUREMENTS

METHODS OF
EXPERIMENTAL PHYSICS
Robert Celotta and Judah Levine, Editors-in-Chief
Founding Editors
L. MARTON
C. MARTON
Volume 26
Physical Optics and

Light Measurements
Edited by
Daniel Malacara
Centro de lnvestigaciones en Optice
Leon, Gto
Mexico
ACADEMIC PRESS, INC.

Harcourt Brace Jovanovich, Publishers
Boston San Diego New York

Berkeley London Sydney
Tokyo Toronto
Copyright 0 1988 by Academic Press, Inc.
All rights reserved.
No part of this publication may be reproduced or
transmitted in any form or by any means, electronic
or mechanical, including photocopy, recording, or
any information qtorage and retrieval system, without
permission in writing from the publisher.
ACADEMIC PRESS, INC.

1250 Sixth Avenue, San Diego, CA 92101
United Kingdom Edifion publid~edbv

ACADEMIC PRESS INC. ( L O N D O N ) LTD.
24-28 Oval Road. London NWI 7DX
Library of- Congress Cbraloging-in-Publication Data

Physical optics and light measurements.
(Methods of experimental physics; v. 26)
Includes bibliographies and index.
1. Optics, Physical. 2. Optical measure-
ments. 1. Malacara, Daniel, Date-
11. Series.
QC395.2.P49 1988 535’.2 88-978
ISBN 0-12-47597 1-8
Printed in the United States of America

88 89 90 91 9 8 7 6 5 4 3 2 1
CONTENTS
. .
LIST OF CONTRIBUTORS . . . . . . . . . . . . . . . . . ix
PREFACE. . . . . . . . . . . . . . . . . . . . . . . . . . xi
...
LIST OF VOLUMESI N TREATISE . . . . . . . . . . . . . . . xi11
1 . Interference
by D A N I E LMALACARA
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . 1
1.2. Two-Beam Interferometers . . . . . . . . . . . . . . 7
1.3. Multiple-Beam Interferometers . . . . . . . . . . . . . 25
1.4. Multiple-Pass Interferometers . . . . . . . . . . . . . 37
1.5. Applications of Interferometry . . . . . . . . . . . . . 38
References . . . . . . . . . . . . . . . . . . . . . 45
2 . Diffraction and Scattering

by D A N I E LMALACARA
2.1. Diffraction ..................... 49

2.2. Fresnel Diffraction . . . . . . . . . . . . . . . . . . 53
2.3. Fraunhofer Diffraction and Fourier Transforms . . . . . 62
2.4. Diffraction Gratings . . . . . . . . . . . . . . . . . 70
2.5. Resolving Power of Optical Instruments . . . . . . . . . 79
2.6. The Abbe Theory of the Microscope . . . . . . . . . . 86
2.7. Scattering . . . . . . . . . . . . . . . . . . . . . . 92
References ..................... 102

V
vi CONTENTS
3. Optical Polarization
by FREDERIC R . STAUFFER
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . 107

3.2. Electromagnetic Description of Light . . . . . . . . . . 108
3.3. Wave Propagation in Isotropic Media . . . . . . . . . . 113
3.4. Wave Propagation for Metals . . . . . . . . . . . . . 120
3.5. Thin Films ..................... 125
3.6. Wave Propagation in Anisotropic Media . . . . . . . . 133
3.7. Slits, Gratings, and Metal Grid Polarizers . . . . . . . . 157
3.8. Light Source and Detector Polarizations . . . . . . . . . 158
3.9. Polarization Determination and Mathematical Description 159
References . . . . . . . . . . . . . . . . . . . . . 162
4 . Holography
by R. D. BAHUGUNA
A N D D . MALACARA
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . 167

4.2. Theory of Holography . . . . . . . . . . . . . . . . 168
4.3. Different Types of Holograms . . . . . . . . . . . . . 175
4.4. Some Applications of Holography . . . . . . . . . . . 191
4.5. Experimental Procedures in Holography . . . . . . . . 199
References .. .. .. .. .. .. . . . . . . . . . 206
5 . Photometry and Radiometry

by WILLIAM L. WOLFE
5.1. Introduction ..................... 213

5.2. Symbols, Units. and Nomenclature . . . . . . . . . . . 214
5.3. Formulas for Blackbody Radiation . . . . . . . . . . . 219
CONTENTS vii
5.4. Simple Radiative Transfer . . . . . . . . . . . . . . . 223

5.5. Radiometric Temperature Measurements . . . . . . . . 239
5.6. Radiometric Instruments . . . . . . . . . . . . . . . 246
5.7. Measurements . . . . . . . . . . . . . . . . . . . . 263
5.8. Photometry: Radiometry of Visible Light . . . . . . . . 284
References ..................... 287
6. Detectors
by T. 0. POEHLER
6.1. Introduction ........... . ......... 291

6.2. Figures of Merit . . . . . . . . . . . . . . . . . . . 292
6.3. Thermal Detectors . . . . . . . . . . . . . . . . . . 293
6.4. Photon Detectors . . . . . . . . . . . . . . . . . . . 303
6.5. Noise . . . . . . . . . . . . . . . . . . . . . . . . 328
. . . . . . . . . . . . . . .
6.6. Optical Window Material 331
References ..................... 332
This Page Intentionally Left Blank
CONTRIBUTORS
Numbers in parentheses indicate pages on which the authors’ contributions begin.
RAMENDRADEOBAHUGUNA(167), Centro de Znvestigaciones en Optica,

Apartado Postal 948, 37000 Leon, Gto. Mexico
DANIELMALACARA ( 1 , 49, 167), Centro de Investigaciones en Optica,
Apartado Postal 948, 37000 Leon, Gto. Mexico
THEODORE 0 . POEHLER(291), Applied Physics Laboratory, Johns Hopkins
Road, Laurel, Maryland 20707
FREDERICR. STAUFFER(107), Sacramento Peak Observatory, Sunspot,
New Mexico 88349
WILLIAM L. WOLFE(213), Optical Sciences Center, University of Arizona,
Tucson, Arizona 85721
ix
PREFACE
Two books covering the field of modern optics have been prepared in
this series “Methods of Experimental Physics”, separating the material into
two parts, one with the title “Geometrical and Instrumental Optics”, and
the other with the title “Physical Optics and Light Measurements”.
The purpose of these books is to help the scientist or engineer who is
not a specialist in optics to understand the main principles involved in
optical instrumentation and experimental optics.
Our main intent is to provide the reader with some of the interdisciplinary
understanding that is so essential in modern instrument design, develop-
ment, and manufacture. Coherent optical processing and holography are
also considered, since they play a very important role in contemporary
optical instrumentation. Radiometry, detectors, and charge coupled imaging
devices are also described in these volumes, because of their great practical
importance in modern optics. Basic and theoretical optics, like laser physics,
nonlinear optics and spectroscopy are not described, however, because they
are not normally considered relevant to optical instrumentation.
In this volume, “Physical Optics and Light Measurements”, Chapter One
describes the theory and applications of interference and interferometers.
Chapter Two studies diffraction, its basic theoretical fundamentals, and
some practical applications. Polarized light and its uses are considered in
Chapter Three. Holography and holographic methods are studied in detail
in Chapter Four. The photometric and radiometric principles are covered
in Chapter Five. Finally, Chapter Six considers detectors.
There might be some overlapping of topics covered in different chapters,
but this is desirable, since the points of view of different authors, treating
different subjects, may be quite instructive and useful for a better under-
standing of the material.
This book has been the result of the efforts of many people. Professor
H. W. Palmer started this project and spent many fruitful hours on it.
Unfortunately, he did not have the time to finish his editorial work due to
previous important commitments. I would like to express my great appreci-
ation of and thanks to Professor Palmer and all of the authors, without
whom this book could never have been finished. I also thank Dr. R. E.
Hopkins and many friends and colleagues for their help and encouragement.
Finally, I appreciate the great understanding of my family, mainly my wife
Isabel, for the many hours taken away from them during the preparation
of these books.
DANIEL MALACARA
Leon, Gto. Mexico.
METH0DS OF
EXPERIMENTAL PHYSICS
Editors-in-Chief
Robert Celotta and Judah Levine
Volume 1. Classical Methods

Edited by lmmanuel Estermann
Volume 2. Electronic Methods, Second Edition (in two parts)

Edited by E. Bleuler and R. 0. Haxby
Volume 3. Molecular Physics, Second Edition (in two parts)

Edited by Dudley Williams
Volume 4. Atomic and Electron Physics-Part A: Atomic Sources and

Detectors; Part B: Free Atoms
Edited by Vernon W. Hughes and Howard L. Schultz
Volume 5. Nuclear Physics (in two parts)

Edited by Luke C. L. Yuan and Chien-Shiung W u
Volume 6. Solid State Physics-Part A: Preparation, Structure,

Mechanical and Thermal Properties; Part B: Electrical, Magnetic, and
Optical Properties
Edited by K. Lark-Horovitz and Vivian A. Johnson
Volume 7. Atomic and Electron Physics-Atomic Interactions (in two

Parts)
Edited by Benjamin Bederson and Wade L. Fite
Volume 8. Problems and Solutions for Students

Edited by L. Marton and W. F. Hornyak
Volume 9. Plasma Physics (in two parts)

Edited by Hans R. Griem and Ralph H. Lovberg
Volume 10. Physical Principles of Far-Infrared Radiation

By L. C. Robinson
Volume 11. Solid State Physics

Edited by R. V. Coleman
xiv M E T H O D S OF EXPERIMENTAL PHYSICS
Volume 12. Astrophysics-Part A: Optical and Infrared Astronomy

Edited by N. Carleton
Part B: Radio Telescopes; Part C: Radio Observations
Edited by M. L. Meeks
Volume 13. Spectroscopy (in two parts)
Edited by Dudley Williams
Volume 14. Vacuum Physics and Technology
Edited by G. L. Weissler and R. W. Carlson
Volume 15. Quantum Electronics (in two parts)
Edited by C. L. Tang
Volume 16. Polymers-Part A: Molecular Structure and Dynamics; Part
B: Crystal Structure and Morphology; Part C: Physical Properties
Edited by R. A. Fava
Volume 17. Accelerators in Atomic Physics
Edited by P. Richard
Volume 18. Fluid Dynamics (in two parts)
Edited by R. J. Emrich
Volume 19. Ultrasonics
Edited by Peter D. Edmonds
Volume 20. Biophysics
Edited by Gerald Ehrenstein and Harold Lecar
Volume 21. Solid State: Nuclear Methods
Edited by J. N. Mundy, S. J. Rothman, M . J. Fluss, and L. C. Smedskjaer
Volume 22. Solid State Physics: Surfaces
Edited by Robert L. Park and Max G. Lagally
Volume 23. Neutron Scattering (in three parts)
Edited by K. Skold and D. L. Price
Volume 24. Geophysics-Part A: Laboratory Measurements; Part B: Field
Measurements
Edited by C. G. Sammis and T. L. Henyey
Volume 25. Geometrical and Instrumental Optics
Edited by Daniel Malacara
Volume 26. Physical Optics and Light Measurements
Edited by Daniel Malacara
1. INTERFERENCE
Daniel Malacara
Centro de lnvestigaciones en Optica A.C.
Apdo. Postal 948
37000 Leon, Gto. Mexico.
1 .l.Introduction
The luminous phenomenon called interference is a direct consequence

of the wave nature of light. Using the interference of light, we can make
interferometers, which are instruments that use this phenomenon to measure
very accurately many physical parameters. The general subject of interfer-
ence has been treated extensively in many classical textbooks on optics like
those by Born and Wolf,’ Cook,’ Franqoq3 Candler,4 Steel,’ and Tolansky.6
There are also special chapters on the subject of interference in many
advanced books like those by Baird,’ Baird and Hanes,’ and Dyson: and
others. This chapter describes very briefly the interference phenomenon and
some interferometers. Special emphasis is placed on the applications of
these useful instruments, which have played a very important role in the
development of physics due to their extremely high accuracy.
1.1 .l. Methods to Obtain Interference Fringes

To obtain interference fringes, the phases of the two interfering waves
must be synchronized, that is, they must be coherent. Before the advent of
lasers, this was possible only if both waves originated from the same light
source. In order to produce two waves from a single source, we must have
either a division of the wave front or division of its amplitude.
Division of Wave Front. This class of interference is produced when
the two interfering wave fronts are taken from different portions of the
original wave front. The typical examples are Young’s experiment, the
Fresnel biprism, and Lloyd’s mirror, but there are many others.
Young’s double-slit experiment is performed as shown in Fig. 1. The
lenses are not strictly necessary, but their addition makes the theory easier.
The light source S must be either a very narrow slit or a point, and lens L
collimates the light to obtain an approximately flat wave front. Normally,
1
M E T H O D S OF E X P E R I M E N T A L PHYSICS Copyright 0 1988 by Academic Pres,. Inc.
Val. 26 All rights of reproduction in any form reserved
ISBN 0-12-475971-8
2 INTERFERENCE
screen
1.0--
.8-
(b)
6-
.4 -
.2-
FIG. 1. Young's experiment. (a) Experimental arrangement and (b) Diffraction pattern.
the interference pattern would be observed at infinity, but lens L serves to

bring the pattern closer over the screen. It can be found in almost any good
textbook on optics' that the interference pattern is given by
7T-1 sin{ ?sin e]

I = I,,, cos'{ A sin e ] 7TD 9 (1.1)
sin 0
A
where I,,, is the irradiance at the center of the pattern, and 8 is the angular
deviation from the optical axis. This radiation pattern also appears with
INTRODUCTION 3
&l--
4 ----
screen
s*
(a)
SO" rce
FIG.2. ( a ) Lloyd's mirror and (b) Fresnel biprism.
dipole antennas and radio waves. The total width of the central maximum
is given by sin 0 = A l l .
The Fresnel biprism and Lloyd's mirror also produce interference fringes
by division of the wave front, as shown in Fig. 2. If two waves are to
interfere producing fringes with good contrast, the polarization states of
both waves must be the same. This condition is always satisfied in the
Fresnel biprism. However, in Lloyd's system, only one beam is reflected.
Therefore, the reflection coefficients and the phase shifts under reflection
must remain nearly constant over the range of incident angles used. This
is possible only near grazing incidence.
4 INTERFERENCE
A second reason for using grazing incidence in the Lloyd system is that
the spacing between fringes decreases rapidly as the separation between
the virtual sources S, and S2 is increased. The sources must be quite close
to each other to make fringes visible, and this is possible only near grazing
incidence. If the screen in the Lloyd system is placed near the edge of the
mirror, we can observe that a dark fringe appears at this edge. This happens
because there is a phase shift upon reflection with grazing incidence. Wolfe
and Eisen" have studied the coherence requirements in Lloyd's mirrors.
Division of Amplitude. This class of interference occurs when both
interfering beams are obtained by division of the amplitude of the original
wave front by means of a partially reflecting optical surface. Then, both
beams travel different paths, and interference occurs when they are recom-
bined. Typical examples are Newton rings and the Michelson interferometer
described in the next section.
1.1.2. Classification of Interference Fringes

A classification for interference fringes can be made according to the way
they are observed, namely, fringes of equal thickness and fringes of equal
inclination.
Fringes of Equal Thickness. Each fringe represents the locus of all
points in which two optical surfaces (or wave fronts) have a constant
separation. This is much better understood by means of the following
example.
Consider the optical arrangement in Fig. 3 where the flat or convex surface
of one lens is placed against the flat surface of another lens. Monochromatic
light enters the first lens and impinges upon the second surface from which
observing eye
observed fringes
FIG. 3. Arrangement to observe Newton fringes.

INTRODUCTION 5
some of the light is reflected. Some of the light that goes through the first
lens is reflected from the upper flat surface of the second lens. The two
reflected beams interfere constructively when the phase difference is an
integral multiple of 27~.We can see that the optical path difference is twice
the surface separation. If one of the surfaces is spherical, each fringe is a
ring that represents the locus of points with equal surface separation. These
circular fringes, also called Newton rings, are a particular case of fringes
of equal thickness.
Fringes of Equal Inclination. I n this case, each fringe is the locus of
points in the field of view with the same angle of the incidence 8 at the
interferometer. As an example, let us consider two parallel reflecting surfaces
as in Fig. 4. As in the previous example, the two interfering beams are
produced by division of amplitude. Fringes of equal thickness cannot be
formed because the optical path difference (OPD) is the same for the entire
field. One way to change the phase difference and hence to observe fringes
is to introduce a range of angles of incidence by using an extended light
source.
Circular fringes are observed with an angular radius 8 given by
mA
cos 8 =-)
2d
where rn is an integer smaller than or equal to 2 d / A .
FIG. 4. Arrangement to produce fringes of equal inclination.

6 INTERFERENCE
To directly observe these fringes, the lens and the screen may be replaced
by the unaided eye (focused at infinity). Then, the observation must be
made almost perpendicularly to the optical surfaces, and the eye must be
placed very close to them, because of the small diameter of the pupil of
the eye. If the observing distance d between the surfaces is so large that
the fringes cannot be resolved with the naked eye, a telescope must be used.
1 .I .3. Light Sources for Interferometers

If two light waves are to interfere, they must have their phase synchro-
nized. Then, the two waves are said to be coherent to each other. When the
two interfering waves are taken from the same light source, either by division
of the amplitude or by division of the wave front, this condition is satisfied
under certain conditions that we will describe.
Let us first assume that the light source is perfectly monochromatic, but
that the light source is extended. Each point of the light source may produce
a complete interference pattern. Good fringe contrast is obtained only if
the phase difference for both waves is the same for all points over the
extended light source.
If the above condition is not satisfied, the light source has to be very
small. In that case, it is said that the interferometer has to be illuminated
by a spatially coherent beam. If the condition is satisfied, the light source
can be extended, and the interferometer is said to be source-size
compensated.
Let us now assume that the illuminating beam is spatially coherent or
that the interferometer is source-size compensated. The light, however, may
not be perfectly monochromatic. Then, each wavelength component of the
light produces an interference pattern. Good fringe contrast is obtained
only if all patterns produced by the interference between both light beams
are the same for all wavelengths. In general, this condition can be achieved
only if this OPD is zero for all wavelengths, the interferometer is then said
to be white-light compensated. Otherwise, the light has to have a narrow
bandwidth whose maximum permissible value depends on the particular
interferometer. A light beam with a narrow bandwidth is said to have
light-temporal coherence. Very few interferometers are white-light com-
pensated, and, therefore, their light sources have, in general, to be quite
monochromatic.
The monochromaticity of a light beam can be measured either by its
bandwidth or by its coherence length, which is the maximum optical path
difference that can be introduced before the fringe contrast is completely
lost. The coherence length of light sources varies from a fraction of a
millimeter to as much as one meter for the KrR6standard lamp (Baird and
TWO-REAM INTERFEROMETERS 7
Howlett”), or even several hundreds of meters for a single-frequency gas

laser.
1.2. Two-Beam Interferometers
Another possible classification for interferometers takes into account the

number of light beams that form the interference pattern. Then, we may
have either two-beam o r many-beam interferometers. In this section, the
most common two-beam interferometers will be briefly described.”
1.2.1. The Newton Interferometer

We call a Newton interferometer any arrangement of two surfaces in
contact with, and illuminated by, a monochromatic source of light, as shown
in Fig. 5 . This interferometer has been described in detail by Murty.”
monochromatic source
of light
4
in contact
FIG.S . Newton interferometer.

8 INTERFERENCE
This instrument forms equal-thickness fringes; and a bright fringe is

formed whenever the phase difference is equal to an integral multiple of
2 7 ~If
. we take into account the fact that there is a phase shift under reflection
for one of the two beams, we can say that a dark fringe occurs when
2nd = mA, (1.3)
where m is an integer, d is the thickness of the gap between the two reflecting
surfaces, and n is the refractive index of the gap medium. This expression
assumes that the fringes are observed normally to the reflecting surfaces,
and hence the two following rules should be followed to obtain good
accuracy:
(a) If the surfaces are flat, a collimating lens is required in order to
observe the fringes from the focus ofthis lens. Another acceptable alternative
is to use a large extended source of light and then to observe the fringes
from a minimum distance of about five times the diameter of the flat surface
under test.
(b) If the surfaces are spherical, the observation point should effectively
be placed at the center of the curvature, as shown in Fig. 6.
Figure 7 shows some typical defects that may be observed with this
interferometer.
1.2.2. The Fizeau Interferometer

This instrument, shown in Fig. 8, has also been described in detail by
Murty.” Equal-thickness fringes are formed by the two reflected beams.
The lens is used to collimate the light and to optically place the observing
eye at infinity. The optical path difference between the two beams is
OPD = 2nd, ( 1.4)
where d is the thickness of the gap, and n is its refractive index. If the field
is free of fringes, we infer that the product nd is constant over the gap. We
can conclude that d is a constant only if n is assumed to be constant. If
straight and parallel fringes with separation S are observed, the two faces
forming the gap are flat and with an angle 8 in radians between them, given
by
A
e=-.
2 nS
This makes this interferometer quite useful for measuring small wedges
in glass plates.
TWO- BEAM INTERFEROMETERS 9
observation
k point
I
surface (a 1
under t e s t
observat Ion
polnt
k
't2 urf ace

(b)
under test
F I G .6. Testing of spherical surfaces with Newton interferometer.
1.2.3. The Michelson Interferometer

Probably the most famous of all interferometers was invented by Michel-
son for his well-known experiment. Figure 9 shows this interferometer. An
extended light source sends the light to the beam splitter P which divides
the beam in two, with one going to mirror MI and the other going to mirror
Mz. Both mirrors send the beams back to the beams splitter and from there
to the observing eye.
When the mirrors M , and M2 are exactly perpendicular to their optical
axes, the fringes are of the equal-inclination type. In this case, the fringes
10 INTERFERENCE
FIG. 7. Typical interference patterns with Newton interferometer.
are circles with angular radii given by Eq. (1.2), where d is given by
d = -OPD,
2 ’
and OPD, is the optical path differences along the optical axis.
TWO-BEAM INTERFEROMETERS 11
light source t- c
FIG. 8. Fizeau interferometer.
The interfering beam coming from mirror MI traverses the beam splitter
three times, while the beam from mirror M 2 passes through the plane only
once. To equalize the thickness of glass that each beam traverses, a plate
P2 of the same thickness and orientation as P, is inserted into the beam to
M 2 . This extra plate is called a compensating plate, and the interferometer
containing one is said to be compensated.
When the interferometry is not compensated, only one beam goes through
the inclined beam splitter, and thus the optical path difference for a given
value of 8 is different in different planes. In that case, the Haidinger fringes
are not circular but elliptical in shape, as described by Guild.I3
light
source
4observed
pattern
F i c . 9. Michelson interferometer.
12 INTERFERENCE
Fic;. 10. Three kinds of Michelson fringes.
The fringes have good contrast only if the OPD, is smaller than the
coherence length. Therefore, the fringes are observed with white light only
if the variation of the OPD within the spectral region being observed
becomes smaller than about A/4. Since the index of refraction is a function
of the wavelength, the former condition cannot be satisfied unless the total
amount of the glass traversed by both beams is the same, and then the
OPD, is adjusted to be smaller than a few wavelengths. Hence, white-light
fringes can be observed only in a compensated interferometer.
If the mirrors MI and M 2 are not perpendicular to their optical axes,
equal-thickness fringes will be observed instead of equal inclination fringes.
However, they are exactly equal-thickness fringes only if the observing eye
is very far from the interferometer, or a lens is used as in Fig. 9 to optically
place the eye at an infinite distance from the interferometer.
If the eye is close to the interferometer and if the mirrors are not
perpendicular to their optical axes, intermediate fringe shapes are observed.
These fringes are curved, with the convexity directed toward the narrower
part of the wedge formed by the two mirror images. The three kinds of
Michelson fringes are shown in Fig. 10. The extended light source permits
the fringes to be observed only at a well-defined plane near the virtual
images of the light source. These fringes are said to be “localized.”
1.2.4. Modifications of t h e Michelson Interferometer

Several modifications have been made of the Michelson interferometer.
The most important ones are the Mach-Zehnder, Jamin, and Twyman-
Green interferometers.
r-i
I 1
I 1
sample 1 I
t
I 1
FIG. 11. Jamin interferometer.
The Jamin interferometer is shown in Fig. 11. The light from the light
source is split at the surfaces of a thick glass plate. Then, the two beams
are recombined at a second identical glass plate. For some applications,
the advantage of this arrangement over the Michelson interferometer is that
the light passes through the sample only once. The plates have a large
thickness so that enough room is left for the sample. This interferometer
was used for refractometry, but it has now been replaced by other kinds of
interferometers.
The Mach-Zehnder interferometer, shown in Fig. 12, can be considered
as a Jamin interferometer in which the four reflecting surfaces have been
replaced by four plates. This instrument is generally used to detect small
changes or space variations in the refractive index of the samples.
Both, the Jamin and the Mach-Zehnder interferometers, are compensated
if correctly adjusted, so that white-light sources can be used if desired.
However, it must be pointed out that this adjustment is particularly difficult
in the case of the Mach-Zehnder.
The Twyman-Green interferometer14*" is one of the most useful
modifications of the Michelson interferometer, used mainly to test the
quality of optical components. The system is illuminated with a collimated
14 INTERFERENCE
,attern 2
9 -
p2 pattern I
2
+
3
\
Q
F I G . 12. Mach-Zehnder interferometer
beam of light, as shown in Fig. 13. The fringes that are observed by placing
the eye at the focus of lens L are of equal-thickness.
The Twyman-Green interferometer is not generally compensated, and
large values of the OPD are frequently obtained, so that the light source
Y
F I G . 13. Twyman-Green interferometer.
--(C)
FIG. 14. Testing optical components with a Twyman-Green interferometer.
mu t be monochromatic. Many different kinds of optical components ca

be tested with this interferometer as described by Malacara.I6 The simplest
one to be tested is a plate of glass with plane parallel faces, which is inserted
into one of the beams as shown in Fig. 14(a). The optical path difference
introduced by the presence of the glass plate may be found considering
that in the space with thickness d occupied by the plate, the index of
refraction is changed from 1 (without the plate) to n (with the plate). A
factor of 2 appears because the light passes twice through the plate. Thus,
O P D = 2 ( n - 1)d. (1.7)
If the interferometer is adjusted so that no fringes are observed before
introducing the plate into the beam, and so that fringes are observed after
the plate is inserted, we establish that the quantity ( n - l ) d is a constant
throughout the plate. Assuming that the index n does not vary over the
plate, we may conclude that d is a constant. The variations in d and in n
can be obtained independently only if another measurement is made with
a Fizeau interferometer.
16 INTERFERENCE
If the plate faces are flat but not parallel and n is constant, straight and
parallel fringes appear.
Using the arrangements in Figs. 14(b) and 14(c), the quality of a lens or
photographic objective can be tested. It is assumed in this test that the
convex or concave mirrors are perfectly spherical. If the lenses are assumed
to be perfect, the quality of the spherical mirror is tested. Interference
patterns associated with various lens aberrations are shown in Fig. 15. Many
FIG. 15. Some Twyman-Green interferograrns.

other types of optical elements, such as prisms and diffraction gratings, can
also be tested with this interferometer.
1.2.5. Shearing Interferometers

These are a class of interferometers in which a perfect reference beam is
not needed. The reference wave front is the same as the one under test but
with a different orientation, position, or lateral extension. Basically, there
are four types of shearing interferometers as illustrated in Fig. 16. Lateral
shearing interferometers are probably the most popular of the four and
have been extensively described by Murty.” Radial, rotational, and reversal
shearing interferometers have also been described in detail by Malacara”
and Bryngdahl.”
The lateral shearing interferometer measures the wave-front deformation
by comparing the wave front with a laterally displaced reproduction of
itself, thus avoiding the need for a flat reference wave front.
Probably, the first interferometer of this kind was designed by Bates.*’
This is basically a Mach-Zehnder interferometer used with a convergent
wave front and with the plate P2 tilted to produce two laterally displaced
I ate ra I radial
rotational shear reversal shear
FIG.16. Four types of shearing interferometer operations.

18 INTERFERENCE
front
sheared
wavef ront s
microscope
objective
gas LASER
rl
lens
under test
FIG. 17. Two lateral shearing interferometers.
wave fronts as shown in Fig. 17(a). The lens under test is placed between
the light source and the interferometer.
Murty” has proposed a simpler version of a lateral shearing interferometer
in which the large temporal coherence of laser light permits large values
for the OPD [see Fig. 17(b)].
Lateral shearing interferometers provide rapid and easy identification of
lens aberrations as shown in Fig. 18. Given an interferogram, the wave front
that produced it can be computed by using several methods, such as those
described by Saunders,” M a l a ~ a r a , ’and
~ Rimmer and W ~ a n t . * ~
Many radial shear interferometers have been designed, but the first one
was invented by Brown.25Basically, this is a Jamin interferometer that works
in convergent light and has a small meniscus lens in one of the beams. To
compensate the interferometer, a small parallel plate is placed in the other
beam [see Fig. 19(a)]. Another interesting radial shear interferometer, shown
in Fig. 19(b), was designed by M ~ r t y . ~ ~
Radial shear interferometers are sensitive to all types of aberrations, as
the Twyman-Green interferometers, but to a lesser degree.
TWO-REAM I N T E R F E R O M E T E R S 19
FIG. 18. Lateral shearing interferometer patterns.
Rotational shear interferometers are not sensitive to rotationally sym-

metric aberrations like spherical aberration but are quite useful for the
detection of nonrotationally symmetric aberration, like coma or astigmatism.
These interferometers have been described by Murty and Hagerott2’ (see
Fig. 20).
Reversal shear interferometers are insensitive to rotationally symmetric
aberrations and also to astigmatism if the axis of reversion coincides with
the x or y axes. It is quite sensitive, however, to coma. Common examples
of this kind of interferometers are the Koester prism interferometers shown
in Fig. 21, and described by Gates28 and Saunders.”
1.2.6. Common-Path Interferometers

Where the reference and test beams follow separated paths in an inter-
ferometer, they can be differently affected by vibrations, turbulence, and
20 INTERFERENCE
RADIAL L I SHEARED
WAVEFRON rs
BROWN'S INTERFEROMETER
WAVEFRON r
UNDER T E S r
I . ."& .^CC."
EYlSPHERlCAL
AM SPLITTER
MURTY'S INTERFEROMETER
FIG. 19. Brown's and Murty's radial shear interferometers.
temperature variations. An interferometer in which both beams follow the

same path is called a common-path interferometer. They have been extensively
treated by M a l l i ~ k . ~ '
Common-path interferometers have zero optical path difference for all
wavelengths and can therefore be used with white light. They are also very
stable. Most of these instruments are either of the lateral shearing or radial
shearing types. A large radial shearing is obtained when one of the beams
is made to traverse a small area of the system under test, while the other
beam traverses the whole aperture.
The most popular instrument of this kind is the so called scatter-plate or
Burch interfer~meter,~' illustrated in Fig. 22. After passing the light through
the first scatter plate, there will be two beams: a diffracted one, which we
FIG. 20. Murty and Hagerott's rotational shear interferometers.
FIG. 21. Koester's prism interferometers. (a) Gates interferometer and ( b ) Saunders
interferometer.
22 INTERFERENCE
FIG. 22. Burch’s scatter plate interferometer.
call DF, and an undiffracted one (zero order), which we call DO. The
undiffracted beam DO covers only a small area near the center of the surface
under test, while the diffracted beam covers the whole surface. The first
scatter plate is imaged onto the second plate, which is a mirror image of
the first. The beam DO produces two beams upon passing through the second
plate, one again undiffracted (DOO) and one diffracted for the first time
(DOF). The beam D F also originates two beams, one undiffracted (DFO)
and one diffracted a second time (DFF). The beam DO0 is observed as a
small bright spot near the center of the mirror. The beam D F F is very faint
and unobserved. The beams DOF and DFO form a fringe pattern. The pattern
is sometimes interpreted as a radial shear interferogram with a very large
amount of radial shear; however, it is best to think of it as a Twyman-Green
pattern.
There i s a large variation in common-path interferometers, many of them
using birefringent materials, as described by FranGon’ and Mallick.”’
1.2.7. Other Two-Beam Interferometers

There are many other two-beam interferometer designs as, for example,
those that involve division of the wave front rather than division of the
amplitude. Two very important instruments of this kind are the Rayleigh
and the Michelson stellar interferometers. M i ~ h e l s o nwas
~ ~the first to apply
interference phenomena to the measurement of refractive indices. Later,
Lord Rayleigh designed an interferometer for this purpose that has been
fringes
FIcj. 23. Rayleigh interferometer.
well described by Candler.4 It is very useful for the measurement of refractive

indices of gases or liquids (see Fig. 23). The light from a narrow slit is
collimated by means of a lens, it then goes through two glass tubes containing
two gases whose indices are to be compared. The beams are then recombined
by another lens in order to form a set of interference fringes. The working
principle is the same as that of Young's double-slit interferometer. When
the gases or liquids in the two tubes have different refractive indices, the
fringes are laterally displaced. The Rayleigh interferometer (or refrac-
tometer) has more accuracy in the measurements of small refractive indices
variations than any other instrument. It can detect changes in the refractive
index of the order of lo-', or one unit in the sixth decimal place.
The Michelson stellar interferometer ( Michelson6") is also based on the
Young interferometer, used in order to measure the angular diameter of
stars. The measurement is indirectly made by measuring the degree of spatial
coherence of the incoming light. This interferometer is illustrated in Fig. 24.
f l a t mirrors
s c r e e n or
d photographic
plate
--La"
'IrnI"
flat mlrrors
F I G . 24. Michelson stellar interferometer.

24 INTERFERENCE
The light received from opposite sides a and b of the star disc has an
angular separation 8. The optical path difference from point a assumed to
be on the axis is represented by OPD,. Thus, the optical path difference
from point b, off-axis, is given by
OPD,, = OPD, - ed, (1.8)
where d is the separation between the two small mirrors receiving the light
from the star. The irradiances produced by the two sources at points a and
b can be obtained from Eq. (1.1) as
I, =21[1 +cos(koOPD,)],
and
Ih = 21[ 1 + COS( koOPDh)]. (1.9)
Since the two point sources are not coherent with respect to each other,
the total irradiance on the pattern is
1, = 1, +Ib
=41 + ~ I [ c okoOPD,)
s( + cos ko(OPD, - Od )]. (1.10)
The fringe visibility versus the slit separation is plotted in Fig. 25. The
angular diameter of the star may be calculated from the measurement of
the distance d, which produces the first minimum in the visibility. As an
example, the star Betelgeuse has an angular diameter of about 0.05 sec of
arc, which lets d equal about three meters for green light.
A modern version of the Michelson stellar interferometer is the irradiance
interferometer in which the light from each slit is not made to interfere.
Instead two irradiance detectors are used, and then the irradiance variations
in the two signals are electronically correlated. This instrument has a
t" t"
X./28 1.22 X/8
01 Two point source8 b) C l r c u l a r rxtrndrd SOUICI
FIG.25. Fringe visibility in Michelson stellar interferometer.

M ULTIPLE-BEAM I N T E R F E R O M E T E R S 25
quantum theory explanation, after an experiment devised by Brown,

Hanbury, and Twiss.3’
1.3. Multiple-Beam Interferometers
Some interferometers form their interference patterns with the superposi-

tion of more than two wave fronts. These interferometers have many advan-
tages, as we shall see in this section.
1.3.1. The Fabry-Perot Interferometer

Multiple-reflection interferometers have been described in detail by Roy-
c h ~ u d h u r and
i ~ ~ Polster er a/.” The most popular of this class of inter-
ferometers is the F a b r y - P e r ~ t .The
~ ~ irradiance pattern of a two-beam
interferometer is sinusoidal, and therefore the exact position of a fringe
maximum cannot visually be located with an accuracy greater than about
one tenth of the fringe spacing. The Fabry-Perot interferometer produces
extremely narrow equal-inclination fringes, and thus the fringe maxima can
be located with an accuracy better than one hundredth of a wavelength.
The principle involves the use of multiple reflections, so that interference
is produced by an almost infinite number of beams as shown in Fig. 26.
Sl s2
FIG. 26. Multiple reflection in a Fdbry-Perot interferometer.
26 INTERFERENCE
(b)
FIG. 27. Fabry-Perot configurations.
The two parallel surfaces may be in two plates as in Fig. 27(a), or in one
plate as in Fig. 27(b). Such plates are sometimes called etalons, from the
French word for "standard."
The transmitted irradiance 1 ( 6 ) as a function of the phase difference,
assuming the amplitude transmission and reflection coefficients in Fig. 26,
may be given by
(1.11)
where it has been assumed that rl and rz are equal, in other words, that
the phase shift upon reflection is either 0" or 180". Here, I is the incident
irradiance, the asterisk denotes the complex conjugate, and the phase
difference is given by
477
6 =- nd cos 0 2 , (1.12)
A
where d is the separation between the reflecting surfaces, n is the refractive

index of the material between the reflecting plates, and O2 is the angle of
incidence of the rays inside the gap.
If the reflectances R , and R2 and the transmittances T, and Tz are always
measured in air, not inside of the glass plates as sometimes defined, Eq.
MULTIPLE-BEAM I N T E R F E R O M E T E R S 27
( 1 . 1 1 ) reduces to
T , T2
I ( 6 ) = I” (1.13)
6’
( 1 - R,)2+4R, sin2-
2
as shown by Ba~meister,~’where the average reflectance R, is defined as
(R~R~)~’~.
This expression can be generalized to include the case in which there are
phase shifts E , and E~ upon reflection as measured in the medium of the
spacer (shown by Baumeister and Jenkins3’) by writing Eq. (1.12) as
47T
6 = - nd cos O2 -$(
A
+ E ~ ) . (1.14)
As pointed out before, the great advantage of multiple-reflection over

two-beam interferometers is the increased sharpness of the fringes. The
width of the fringes depends very much on the reflectances of the plates as
illustrated in Fig. 28. If we define the half-width A6 of the fringe as the
width at half the maximum fringe irradiance, it is given by
1-R,
A6=2- (1.15)
R;I2 ’
The “finesse” N R is a number used to specify the sharpness of the fringes.

It is defined as the ratio of the fringe spacing to the width in Eq. (1.15). Thus,
(1.16)
As shown in Fig. 28, the finesse can be extremely large even for relatively
small values of R,.
FIG. 28. Fringe profile in a Fabry-Perot interferometer.

28 INTERFERENCE
As pointed out by Baird and Hanes,' with a Fabry-Perot interferometer

we can observe fringes of equal inclination, fringes of equal thickness, and
fringes of equal chromatic order, which will be described in this section.
Fringes of equal inclination are formed when the light source is extended
or convergent, so that a wide range of angles of incidence are present and
the two plates are parallel to each other. The fringes are observed at an
infinite plane by means of a lens with a screen at its focus or a small
telescope focused at infinity. The fringes are circles located at positions
such that
OPD = 2nd cosz 8 = mA, (1.17)
where m is an integer. The smallest circular fringe corresponds to the largest

value of m, given by
2 nd
m,,,=-- A e, (1.18)
where e is the fractional value of ( 2 n d l A ) . Thus, the value of m for the

pth fringe is
(1.19)
hence the angular radius of this fringe is given by

A
cos e2= 1 --2 nd ( p - 1 + e ) . ( 1.20)
Two rings with different wavelengths may have the same angular
diameters, because their value of m may also be different. However, this
ambiguity is eliminated by measuring two sets of rings with different values
of d.
Fringes of equal thickness may also be formed with a Fabry-Perot
interferometer by forming a small wedge with the two reflecting surfaces.
In this case, an extended source is not needed, and the light is normally
collimated by a lens. The instrument may now be called a multiple-reflection
Fizeau interferometer.
Fringes of equal chromatic order appear when a white light source is
used to illuminate the interferometer adjusted with parallel faces, and then
the outgoing light is seen through a spectrograph. Narrow bright bands are
seen in the spectrum. This effect is called a channeled spectrum, and the
fringes are called fringes of equal chromatic order or Edser- Butler fringes.
Tomkins and Fred39 have used these fringes to calibrate a spectrograph
by using the fact that they are located at positions A N = 2 n d / ( N - 4 / 2 ~ ) ,
MULTIPLE-BEAM INTERFEROMETERS 29
where N is an integer, and C#I is the total phase change on the two internal
reflections.
T ~ l a n s k y ~ has
~ . ~also
* ~used
' these fringes to measure the microtopography
of crystal surfaces and films.
1.3.2. The Spherical Fabry-Perot Interferometer

The optical path diff erence in a Fabry-Perot interferometer changes with
the angle of incidence. This property is not desirable in photoelectric
spectrometry where a large solid angle of acceptance of the light is required
together with a small variation in the optical path difference. In this tech-
nique, the light spectrum is analyzed not by changing the angle of incidence
to form rings but by changing the plates' separation by a small amount,
usually of the order of a wavelength or less.
An interferometer in which the optical path difference does not change
with small variations in the angle of incidence was designed by C ~ n n e s . ~ *
It consists of two spherical confocal surfaces as shown in Fig. 29. The two
spherical surfaces have their focal planes coincident at some plane between
them. If the two mirrors have the same radius of curvature, the center of
curvature of each mirror is located on the opposite mirror.
An incident ray reaching mirror M , at point A travels to point B, where
it is partially reflected. The reflected ray then follows path BDCAB, so that
it comes back to the original trajectory AB at A, where it is again partially
reflected. The optical path difference between two consecutive rays is
OPD = 4t, (1.21)
where r is the separation between the vertices of the two mirrors. The lower
half of each mirror is totally reflecting, while the upper half is partially
reflecting.
We can see that within the limits of the paraxial optics (small angles of
incidence), the phase difference is independent of the angle of incidence.
/ \
FIG. 29. Connes's confocal interferometer.

30 INTERFERENCE
We should also notice that the emerging rays are all coincident along the
incident ray and not the parallel and laterally separated rays as in the
Fabry-Perot interferometer. To keep the rays within the limits of the paraxial
optics, two small circular diaphragms are frequently used in front of the
mirrors.
The only adjustment that this interferometer needs is to fix the correct
distance between the mirrors with a tolerance of a few microns. Since the
mirrors are spherical, there is no need to adjust their parallelism.
The main use of this instrument is to examine the spectrum of lasers, as
described by Herriott4' and Fork et al.,44by oscillating one of the mirrors
a small distance (usually about h / 4 by means of an electromechanical
device).
1.3.3. Other Multiple-Reflection Interferometers

In some other interferometers, the principle of multiple reflection is also
used in order to increase the accuracy of the instrument. These inter-
ferometers are not used much and are mainly of historical interest. An
example is the Lummer-Gehrke plate4 shown in Fig. 30(a): It may be
FIG. 30. Other multiple-beam interferometers. ( a ) Lummer-Gehrke plate and ( b ) Echelon.

M ULTIPLE-BEAM INTERFEROMETERS 31
considered as a Fabry-Perot etalon using a very large angle of incidence.

The angle of incidence of the light inside the plate is very near the critical
angle, so that a large reflectivity is obtained without any kind of coating
on the surfaces. Light enters the system through a small prism cemented
onto one end of the plate. This system was very much used4’ at the beginning
of the century for spectrographic analysis, but is has now been largely
replaced by the Fabry-Perot interferometer.
The e ~ h e l o n ~ ~of. ~Michelson
.~’ [Fig. 30(b)] is another multiple-beam
device, formed by a stack of flat parallel plates with constant thickness t.
The plates are arranged so that steps with approximate height of 1 cm and
width of 0.1 cm are formed. In the transmission echelon, the wave front is
divided in M portions with the relative optical path difference as a coarse
grating, blazed for a very high order. A complete description of transmission
and reflection echelons is found in the book by Candler.4
1.3.4. Interference Filters

A common application of the Fabry-Perot interferometer is to isolate
narrow spectral regions. Used in this way (Fig. 31), it is called an interference
filter. It should be used only with collimated light. The wavelength transmit-
ted by the filter is given by
2nd cos e
A= (1.22)
m
If a filter is to isolate a line with wavelength A, a very low order of

interference, m about 2, is chosen so that there is a large wavelength
difference between the different orders. The undesired orders are then
eliminated by means of any broad-band absorptive filter.
An interference filter can be made by evaporating a thin reflective film,
like silver or aluminum, on a piece of glass. Then a transparent film, like
magnesium fluoride, is evaporated on the metal. On this layer, another film
of metal is evaporated to complete the films. It is interesting to point out
that the glass substrate does not have to be extremely flat, since the important
thing is only the constant thickness of the evaporated spacer.
-glass
m
-c-i-elatl coating
dielectric spacer
-glass
Fic;. 31. Interference filter.

32 INTERFERENCE
ANGLE OF INCIDENCE
F l c . 32. Wavelength change with orientation in an interference filter.
An interference filter may be slightly tunned by changing the filter orienta-

tion with respect to the light beam. Keeping in mind that 8 is the angle
inside the spacer with refractive index n, and 4 is the angle of incidence
outside the filter, we can see that
( 1.23)
where A, and A. are the wavelengths for an angle of incidence 4 and for
normal incidence, respectively. Figure 32 shows a plot of (Ao- & ) / A o versus
4. A consequence of this angle sensitivity is that the narrower the passband,
the more accurate the angle of incidence and collimation have to be.
Half-widths of interference filters can have a wide range of values, between
about 80 A and about 1 A. In order to reduce the passband width (or
half-width), interference filters are often made with more than one cavity,
that is, with several spacers with reflecting films made of dielectric multi-
layers instead of metal.
1.3.5. Thin Films

A multilayer film consists of thin layers of different dielectric materials
deposited by evaporation, one on top of the other. Table I lists a few of
the materials used. A good review of this subject was written by Ba~meister,~’
and a more complete description is given in Chapter 13 in this book. Due
M ULTI PLE-BEAM I N T E R F E R O M E T E R S 33
TABLE1. Some Dielectric Materials for Thin Films
Refractive index Transparency

Formula Name A = 550 nrn range in pm
MgF, Magnesium fluoride 1.38 0.13-8

ZnS Zinc sulfide 2.3 0.39-14.5
Ti,O, Titanium oxide 2.3 0.39-12
SiOz Silicon dioxide 1.46 0.2-9
NalAIF, Cryolite 1.35 0.2-14
to the interference produced by multiple reflections at the interfaces between

layers, these filters are highly color selective in both transmission and
reflection. There are many uses for these films, for example:
(a) Antireflection coatings. A lens can be coated with one or more
dielectrics for the purpose of minimizing reflected light, thereby enhancing
the transmitted light. The flare around the image is reduced by eliminating
the unwanted light that would otherwise come from reflections at the lens
surfaces.
(b) Enhancement of the reflection and/or protection of metal mirrors.
(c) Beam splitters and semitransparent mirrors for interferometers and
other kinds of optical instruments. For example, partially transmitting
mirrors are used in lasers. These mirrors have an advantage over the metallic
ones in that they do not absorb energy.
(d) Band-pass filters. These filters reflect all the colors that are not
transmitted, as opposed to the gelating or glass filters that absorb all the
energy that is not transmitted. They are often called dichroic mirrors, but
the term dichrom strictly applies to materials that absorb two polarizations
differently.
Among the uses for these filters are cold mirrors for movie projectors
that do not reflect infrared but only visible light; color beam splitters for
cold television cameras; and others.
Single-layer Coatings. A single-layer coating deposited over a substrate
can be used to reduce the reflectance of a dielectric surface or to enhance
the reflectance of a metallic surface. Let us consider the case illustrated in
Fig. 33. The minimum reflectivity is obtained when the optical thickness of
the layer is h / 4 and its refractive index is
- nl/2 (1.24)
2- s 9
where n , is the substrate refractive index. This condition cannot be exactly

34 INTERFERENCE
surface
surface 2
I
I
V/II
I coating Y ' ' H
I
+ @ N 2 I f d
glass
substrate fl @ ,& N,
FIG. 33. Single-layer coating.
satisfied with practical dielectrics. It may, however, be approached very

closely, for example, by the magnesium fluoride.
The reflectance of the system varies with the optical thickness of the
coating as shown in Fig. 34. We can see that the maximum reflectance that
can be obtained for any thickness is the reflectance value with no coating.
In other words, the reflectance can be reduced, but never increased, when
this coating is deposited on a dielectric surface. This result can be shown
to be true whenever the coating has a refractive index smaller than that of
the substrate. The minima have zero reflectance only when the condition
in Eq. (1.24) is exactly satisfied.
In order that the wavelength interval between the selected minimum at
A,, and its adjacent maxima be as large as possible, the minimum optical
thickness that produces a minimum is used, that is, Ao/4. Then, if R is the
,reflectance
uncoated substrate
0 x/4 A/ 2 3x/4 OpllCOl thickness N d
FIG. 34. Change in the reflectance of a single antireflecting layer with the optical
thickness nd.
MULTIPLE-BEAM I N T E R F E R O M E T E R S 35
0 600 700nm
’0° wavelength
FIG. 35. Reflectance for a single antireflecting film.
reflectance of the uncoated substrate, the final reflectance becomes

R h).
= Ro cos2( .rrA 0
(1.25)
as shown graphically in Fig. 35.

If the reflectance is to be increased rather than decreased, the index of
refraction of the coating should be as high as possible with respect to that
of the substrate. In this case, there is a phase change under reflection at
one of the interfaces. Therefore, when increasing the optical thickness, the
first maximum of the reflectance occurs at Ao/4 (see Fig. 36).
Multilayer Films. Several films with different thickness and alternatively
of high and low index of refraction can be superimposed by evaporation.
By selecting the appropriate thickness and number of layers, it is possible
to obtain almost any desired bandwidth, reflectance, and transmittance of
the multilayer.
A particularly interesting multilayer is the “quarter-wave stack.” As shown
in Fig. 37, this multilayer consists of an even number of layers of alternatively
low and high refractive indices. All layers have the same optical thickness
Ao/4. Figure 38 shows the reflectance curve of a quarter-wave stack with
four periods of high and low index layers.
36 INTERFERENCE
.3c
.2c
-
0
-"
-
.I 0
( I I I
4 1 so0 600 700nrn
wavelength
FIG.36. Reflectance of a single reflecting layer.
FIG.37. A quarter-wave stack.
I reflectance
F I G .38. Reflectance curve for a quarter-wave stack.

M ULTI PLE-PASS I N T E R F E R O M E T E R S 37
-==--+I ,test lens
l M 3
FIG. 39. Double-pass Twyman-Green and Fizeau interferometers.
1.4. Multiple-Pass Interferometers
These are interferometers in which at least one of the wave fronts traverses
the normal trajectories more than once. The instruments, which have been
described in detail by H a r i h a ~ - a nand
~ ~ by Langenbe~k,~’
have some advan-
tages, as we shall see in this section.
A double-pass interferometer is designed with the purpose of separating
the symmetrical and the antisymmetrical parts of the wave aberration and
to display them in separate interferograms. Figure 39 shows the double-pass
Twyman-Green interferometer described by Hariharan and Sen.” The two
beams going out from the interferometer are reflected back to it by a small
mirror M 3 . Then, the patterns are observed by means of the beam splitter
S 2 , where there are four wave fronts. By moving the light source slightly
off-axis, two different patterns are observed. One pattern represents the
symmetrical component and the other one the antisymmetrical component
of the wavefront.
Double-pass Fizeau interferometers have also been designed by Sen and
Pu~~tambekar’”’~ and by Puntambekar and Sen” in order to reduce the
coherence requirements of the light source.
The multiple-pass principle can also be used in order to increase the
sensitivity of an interferometer. Two examples are the multiple-pass
Twyman-Green and Fizeau4’ interferometers shown in Fig. 40.
38 INTERFERENCE
aux i lio r y be a rn
divider
test
surf ace
coated mirror
surface
FIG.40. Multiple-pass (a) Twyman-Green and (b) Fizeau interferometer.
1.5. Applications of Interferometry
Interferometric techniques are very powerful tools in many different

branches of science for very accurate measurements of a great number of
parameters. A complete list of all applications of interferometry would be
almost infinite,47~s4*ss~sh
but a brief description of the main uses of these
techniques is given in this section.
1.5.1. Relativity Measurements

One of the crucial experiments in modern physics was the Michelson-
Morley e ~ p e r i m e n t , 'which
~ is now described in many modern textbooks
of p h y s i ~ s . ~The
' experiment was performed with the famous Michelson
interferometer in order to determine the frame of reference (the aether)
with respect to which the light moves. It was assumed that the light had a
constant speed with respect to the aether and that the earth moves in space,
APPLICATIONS OF INTERFEROMETRY 39
with the aether being stationary. Hence, if the interference pattern is

observed with one arm of the interferometer being parallel to the earth's
direction of movement, a small movement of the fringes is to be expected
when the interferometer moves with the earth at 90" with respect to the
former direction of movement. The negative result of the experiment led
to many hypotheses trying to explain it,SXbut the satisfactory explanation
came only with the special theory of relativity.
Modern versions of this experiment have recently been performed with
the same results.
1.5.2. S t a r Diameter Measurements

Michelson used his stellar interferometerh0in order to measure the angular
diameter of stars (see Section 1.2.7). Although this instrument has been
successful in measuring stellar diameters, it has many stability problems
that limit its accuracy.
1.5.3. Refractometry
Interferometry can also be used to measure very small changes in the
refractive index of a gas or to compare the refractive indices of two liquids.
As mentioned before, Michelson3' applied interference for the first time for
this purpose. The most commonly used instrument for this application is
the Rayleigh interferometer4 described in Section 1.2.7. A Jamin or a
Fabry-Perot interferometer can also be used to measure refractive
indices.
1.5.4. Schlieren Techniques

The well-known Foucault or knife-edge test and many variations
described under the general name of schlieren techniques can be considered
to be interferometric. A good description of this test can be found in many
books" and papers6'
As shown in Fig. 41, the general idea behind this test is to detect small
lateral displacements of rays focused to a point. This test can be applied
to measure wave-front or surface deviations in spherical mirrors or lenses.
Another application is to study the turbulence in a cell placed between the
wave-front generator and the knife-edge.
The Ronchi testh3 could be considered as a variation of the classical
schlieren techniques.
40 INTERFERENCE
FIG.41. Foucault test.
1.5.5. Wave-front Topography

Almost all interferometers can be applied to the study of the topography
of wave fronts or optical surfaces. An extensive bibliography on this subject
is found in a paper by Malacara et ~ 1 . "Brief
~ reviews are given by Briers"
and by Schulz and Schwider.66 An extensive study is presented in the book
edited by M a l a ~ a r a , ~where
' even modern techniques such as fringe-scan-
ning interferometers are described by Bruning."'
It would be impossible to describe the numerous testing techniques
in this chapter, and so the reader is referred to the specialized
publications.
1.5.6. The Transfer Function of Lenses

A way to specify the resolving power of a lens is by means of the optical
transfer function (OTF). This function gives the contrast modulation of the
phase shift for each spatial frequency in lines per unit length. There are a
large number of ways to measure the optical transfer function of a lens, as
explained by R o s e n h a ~ e r . "Some
~ of these methods are interferometric.
The obvious way is to measure the wave-front deviations by means of
any interferometer, as for example the Twyman-Green interferometer, and
then to compute the transfer function. A more direct way is to use a lateral
shearing interferometer and to make use of the fact that the transfer function
can be mathematically obtained from a convolution of the wave front, as
APPLICATIONS OF INTERFEROMETRY 41
shown by hop kin^.^' Some lateral shearing interferometers made for this
purpose are described by Fran~on.'
1.5.7. Length and Angle Measurements

Very precise length measurements can also be made by means of inter-
ferometry. For example, the separation between the two reflecting flats of
a Fabry-Perot interferometer can be accurately measured. From Eq. (1.17),
the order of interference m is given by
2nd cos 8
m= (1.26)
A
If this order of interference m is somehow determined and the wavelength

is known, the distance d can be determined. A shown by F r a n ~ o n ,there ~
are several methods to determine the order of interference.
M i ~ h e l s o n , ~at' the Bureau International des Poids et Mesures, was the
first one to make an experimental determination of the length of the meter
by means of interferometry. He used a Michelson interferometer with the
device in Fig. 42 instead of one with flat mirrors, and found that the meter
was equal to n, where n = 1,553,165.3 and h o = 6438.472 8, for the red
cadmium line.
A more recent determination of the length of the meter was made by
Baird and H ~ w l e t t , ~after * the adoption in 1960 of the new international
meter as 1,650,763.73 vacuum wavelengths of the 6057 8, line of Kr.x6The
precision with this method is about lo8.
Modern definitions of the meter, however, involve the use of lasers, with
the advantage of their great coherence length.5 The frequency of lasers is
calibrated against the cesium standard, thus defining the unit of time. The
I I
FIG. 42. Michelson etalon.

42 INTERFERENCE
second step was to measure the wavelength of the laser line by inter-
ferometric comparison with the Krypton line. Next, the vacuum speed of
light was computed from the product of the wavelength and the frequency,
and this result, equal to c = 299,792,498 meters/sec. was defined to be exact.
In conclusion, the definition of the meter fixes the speed of light, and any
source of known frequency may become a reference standard for length.
An obvious application of these techniques is the calibration of scales in
many instruments, for example in the fabrication of diff raction gratings.73
Interferometry using amplitude-modulated beams can be used to measure
longer distances. It is like having a much longer wavelength, but the accuracy
of the measurements can be much greater than that obtained by other
methods. Instruments using this principle in order to measure long distances
of the order of several meters have been made by Dukes and Gordon.74 A
good description of these instruments can be found in a paper by Bruning.68
Even the distance from the earth to the moon has been measured with this
principle by H a m m ~ n d . ~ '
as well as l a ~ - g e ~angles
~ * ~ ' can also be measured very accurately
by interferometric procedures.
1.5.8. Microscopy
The principle of multiple-beam interference in thin films can be used
with great advantage to study the microtopography of small crystal or glass
surfaces. This subject has been fully described by Tolansky4' and by Krug
et al." One of the many applications of this method is the observation of
the polishing defects of a metal or glass surface.
1.5.9. Doppler Interferometry

If a light source is moving with a velocity u with respect to the observer,
the observed wavelength and frequency shift by an amount given by
AA
-
Av - u
(1.27)
A v c
These small wavelength and frequency shifts can be measured very accur-
ately by means of interferometry as described by Foreman er a/.''' An
interesting application is the measurement of fluid-flow velocities by using
small light-scattering particles in colloidal suspension.".'*
Another application of the Doppler effect to interferometry is the measure-
ment of slow rotations, as for example the earth rotation, with very high
accuracy. This measurement is performed with the so-called Sagnac or ring
interferometer," shown in Fig. 43. Two beams travel in the interferometer,
APPLICATIONS O F I N T E R F E R O M E T R Y 43
MONOCROMATIC
COLIMATED
LIGHT SOURCE
OBSE E R OR
LIGH' ETECTOR
FIG. 43. Sagnac interferometer
one moves clockwise, and the other one counterclockwise. The result is that
there is a fringe shift proportional in magnitude to the angular rotation w ,
as follows:
4Aw
AN=- ( 1.28)
CA '
where A is the area of the interferometer ( A = 2R').

M i c h e l ~ o n ~ and
~ , ' ~Michelson and Gale'' were the first to use this inter-
ferometer to measure the earth's rotation. Macek and Davisx7 designed a
modern version of this instrument, using four laser tubes with a cavity in
a ring configuration. With lasers, a stronger signal is obtained, but the two
counter-running methods tend to lock together at low speeds, but with
special methods this can be avoided, as shown by Aronowitz.xx The cyclic
interferometer as well as the ring laser may be used as gyroscopes, as
described by Rowland and A g r a ~ a l . ' ~
44 1 NTERFERENCE
1.5.10. Spectroscopy
Probably one of the most popular and oldest applications of inter-
ferometers is the measurement of the spectral components of light beams,
as was already mentioned in Section 1.3. The resolution of interferometric
methods is so good that the hyperfine structure46and the wavelength values3'
can be measured with high accuracy.
Another very important and relatively new spectroscopic method that we
want to describe here is called Michelson-Fourier spectroscopy. This
method has been treated by P. Connes'" and Vanasse and Strong." It works
on the principle that the temporal coherence or wave-train shape determines
the frequency spectrum and vice versa. A Twyman-Green interferometer
can be used to measure the wave-train shape, and then the spectrum can
be computed mathematically. Suppose that we want to determine the spec-
trum of the light source illuminating the interferometer in Fig. 13. This is
done by moving one of the flat mirrors along its optical axis to change the
OPD. A light detector then replaces the eye in order to permit the average
irradiance of the interference pattern to be recorded. The irradiance
measured by the detector is a function of the optical path difference.
Defining (1/2)Zl(k)dk as the average irradiance due to either of the two
beams on a narrow band dk centered at k, the irradiance d l at the detector
will vary sinusoidally with the OPD as follows:
d l = 11(k ) [1+ COS( kOPD)] dk. ( 1.29)
Since different wavelength light beams are mutually incoherent, the irradi-
ance I due to all colors acting simultaneously is
I= lom
I,(k)[l+cos(kOPD)] dk. ( 1.30)
If we now define
lom
I,= I,(k) dk, (1.31)
and
I(OPD)=
I,' I , ( k ) c o s ( k . O P D ) dk, (1.32)
the irradiance at the detector can be written

I = I"+ I ( 0 P D ) . (1.33)
After I ( 0 P D ) has been experimentally measured, the irradiance spectrum
I ( k ) can be obtained from the cosine Fourier transform of I(OPD), which
APPLICATIONS O F I N T E R F E R O M E T R Y 45
gives
Z(0PD) C O S ( ~ OPD)
- d(0PD). ( 1.34)
In practice, the mirror displacements cannot have infinite magnitude, and

thus this integral is approximated by
I , ( k ) =- Z(0PD) cos(k- OPD) d(OPD), (1.35)
where the value of OPD,,, is twice the maximum mirror displacement. The
larger this OPD,,, , the better the spectral resolving power.
There are two clear advantages of interference spectroscopy over conven-
tional dispersing spectroscopy. The first advantage is called Jucquinot and
Dufour” or etendue advantage. The spectroscopic field of view of an
interferometer is circular and much greater than the narrow and long field
given by the slit of a spectrometer. For instance, with a Fabry-Perot
interferometer, the spectrum of the whole image of the sun can be simul-
taneously studied, while with a spectroheliograph, this has to be done by
dividing the image of the sun into slits.
The second advantage is called Fellgett9’ or multiplex advantage. In a
spectroscope, different wavelengths are studied either at different times if
a photoelectric detector is used or with different photographic regions if a
photographic plate is used. In interference spectroscopy, on the other hand,
all spectral lines are measured simultaneously with a single photoelectric
detector. This is equivalent to coding each optical frequency by an electrical
cosine signal with different frequency, and then decoding numerically by
Fourier analysis. In this manner, the signal-to-noise ratio is much greater
than in conventional spectroscopy.
We could describe many other useful applications of interferometry, but
it would be impossible to mention all of them in a single chapter.
Acknowledgment
The author wishes to express his gratitude to Professor W. H. Steel for his many valuable
suggestions.
References
1. M . Born and E. Wolf, Principles ofoptics, Pergamon Press, New York, 1959.
2. A. H . Cook, Interference of Electromagnetic Waves, Clarendon Press, Oxford, 1971.
3. M . Fraqon, Optical Interferometry, Academic Press, New York, 1966.
4. C. Candler, Modern Interferometers, Hilger and Watts, London, 1951.
5. W. H. Steel, Inferferomerry, 2nd ed., Cambridge University Press, London, 1983.
46 INTERFERENCE
6. S. Tolansky, An Introduction to Inrerferometry, Longmans Green, London, 1955.

7. K. M. Baird, “Interferometry: Some Modern Techniques,’’ in Advanced Optical Techniques
(A.E.S. Van Heel, ed.), p. 123, North-Holland Publ., New York, 1967.
8. K. M. Baird, and G. R. Hanes, “Interferometers,” in Applied Optics and Optical Engineering,
Vol. 4, p. 309 (R. Kingslake, ed.), Academic Press, New York, 1967.
9. J. Dyson, “Interferometers,” in Concepts ofClassical Optics (J. Strong, ed.), p. 377, W. H.
Freeman, San Francisco, 1958.
10. R. N. Wolfe and F. C. Eisen, “Irradiance Distribution in a Lloyd Mirror Interference
Pattern,” J. Opt. SOC.Am. 38, 706 (1948).
1 1 . W. H. Steel, “Two Beam Interferometry,” in Progress in Optics, Vol. 5, Chap. 3 ( E . Wolf,
ed.), North-Holland Publ., Amsterdam, 1966.
12. M. V. R. K. Murty, ”Newton, Fizeau and Haidinger Interferometers,” in Optical Shop
Testing ( D . Malacara, ed.), Chap. 1, Wiley, New York, 1978.
13. J. Guild, “Fringe Systems in Uncompensated Interferometers.” Proc. Phjjs. Soc. (London)
33, 40 (1920-1921).
14. F. Twyman, “Camera Lens,” British Patent 130224 (1919).
15. F. Twyman, “The Testing of Microscope Objectives and Microscopes by Interferometry,”
Trans. Faraday Soc. 16, 208 (1920).
16. D. Malacara, “Twyman-Green Interferometer,” in Optical Shop Testing ( D . Malacara,
ed.), Chap. 2, Wiley, New York, 1978.
17. M. V. R. K. Murty, “Lateral Shearing Interferometers,” in Optical Shop Testing (D.
Malacara, ed.), Chap. 4, Wiley, New York, 1978.
18. D. Malacara, “Radial, Rotational. and Reversal Shear Interferometers,” in Optical Shop
Testing (D. Malacard, ed.), Chap. 5, Wiley, New York, 1978.
19. 0. Bryngdahl, “Applications of Shearing Interferometry,” in Progress in Optics Vol. 4,
Chap. 2 (E. Wolf, ed.), North-Holland Publ., Amsterdam, 1965.
20. W. J . Bates, “A Wavefront Shearing Interferometer,” Proc. Phys. Soc. 59, 940 (1947).
21. M. V. R. K. Murty, “The Use of a Single Plane Parallel Plate as a Lateral Shearing
Interferometer with a Visible Gas Laser Source,” Appl. Opt. 3, 531 (1964).
22. J . B. Saunders, “Measurement of Wavefronts without a Reference Standard. I: The
Wavefront Shearing Interferometer,” J. Res. Nut. Bur. Stand. 65b, 239 (1961).
23. D. Malacara. Testing of Optical Surfaces, Ph. D. Thesis, Institute of Optics, University of
Rochester, New York, 1965.
24. M. P. Rimmer and J. Wyant, “Evaluation of Large Aberrations Using a Lateral Shear
Interferometer Having Variable Shear,” Appl. Opt. 14, 143 (1975).
25. D. S. Brown, Interferometry N.P.L. Symposium No. I / , p. 253, Her Majesty‘s Stationery
Office, London, 1959.
26. M. V. R. K. Murty, “A Compact Radial Shearing Interferometer Based on the Law of
Refraction,” Appl. Opt. 3, 853 (1964).
27. M. V. R. K. Murty and E. C. Hagerott, “Rotational Shearing Interferometer,” Appl. Opt.
5, 615 (1966).
28. J. W. Gates, “The Measurement of Comatic Aberrations by Interferometry,” Proc. fhvs.
Soc. B 68, 1065 (1955).
29. J . B. Saunders, “Inverting Interferometer,” J. Opt. Soc. Am. 45, 133 (1955).
30. S. Mallick, “Common-Path Interferometers,” in Optical Shop Testing ( D . Malacara, ed.),
Chap. 3, Wiley, New York, 1978.
31. J. M. Burch, “Scatter Fringes of Equal Thickness,” Nature 171, 889 (1953).
32. A. A. Michelson, “Interference Phenomena in a New Form of Refractometer,” Phil. Mag.
5, 236 (1882).
33. Hanbury R. Brown, and R. Q. Twiss, “A new Type of Interferometer for Use in Radio
Astronomy,” Phil. Mag. 45, 663 (1954).
REFERENCES 47
34. C. Roychoudhuri, “Multiple Beam Interferometers.” in OpficalShopTesfing ( D . Malacara,

35. H. D. Polster, R. M. Scott, R. L. Crane, P. H. Langenbeck, R. Pilston, and G. Steinberg,
“New Developments in Interferometry,” Appl. O p f .8, 521 (1969).
36. C. Fabry and A. Perot, “Thtorie et Applications d’une Nouvelle Mithode de Spectroscopie
Interferentielle,” Ann. Chim. Phys. 7, 115 (1899).
37. P. W. Baumeister, “Interference and Optical Interference Coatings,” in Applied Optics and
Opfical Engineering, Vol. 1, p. 285 (R. Kingslake. ed.), Academic Press, New York, 1965.
38. P. W. Baumeister and F. A. Jenkins, “Dispersion of the Phase Change for Dielectric
Multilayers. Application to the Interference Filter,” J. Opt. Sac. Am. 47, 57 (1957).
39. F. S. Tomkins and M. Fred, “Wavelength Measurements with a Concave Grating Spectro-
graph,” Appl. O p f . 2, 715 (1963).
40. S. Tolansky, Multiple-beam Interferomefry of Surfaces and Films, Clarendon Press, Oxford,
1948.
41. S. Tolansky, Surjace Microtopography, Wiley Interscience, New York, 1960.
42. P. Connes, “L‘italon d e Fabry-Perot Sphirique,” J. Phvs. Radium 19, 262 (1958).
43. D. R. Herriott, “Spherical-Mirror Oscillating Interferometer,” Appl. Opt. 2, 865 (1963).
44. R. L. Fork, D. R. Herriott, and H. Kogelnik, “A Scanning Spherical Mirror Interferometer
for Spectral Analysis of Laser Radiation,” Appl. Opt. 3, 1471 (1964).
45. K. W. Meissner, “Interference Spectroscopy. Part I,” J. Opt. Soc. Am. 31, 405 (1941).
46. K. W. Meissner, “Interference Spectroscopy. Part 11,” J. O p f . Sac. A m . 32, 185 (1942).
47. W. E. Williams, Applicafions of Inferferomefry,4th ed., Methuen, London, 1950.
48. P. Hariharan, “Multiple Pass Interferometers,” in Optical Shop Tesfing, ( D . Malacara,
ed.), p. 217, Wiley, New York, 1978.
49. P. Langenbeck, “Multipass Twyman-Green Interferometer,” Appl. O p f . 6 , 1425 (1967).
50. P. Hariharan and D. Sen, “The Separation of Symmetrical and Asymmetrical Wavefront
Aberrations in the Twyman-Green Interferometer,” Proc. Phys. Sac. (London) 77, 328
(1961 ).
51. D. Sen and P. N. Puntambekar, “An Inverting Fizeau Interferometer,” O p f . A c f a 12, 137
(1965).
52. D. Sen and P. N. Puntambekar, “Shearing Interferometers for Testing Corner Cubes and
Right Angle Prisms,” Appl. O p f . 5, 1009 (1966).
53. P. N. Puntambekar and D. Sen, “A Simple Inverting Interferometer,” Opt. Acta 18, 719
(1971 ).
54. W. E. Williams, “Applications of Interferometry,” in Concepfs of Classical Opfics (J.
Strong, ed.), p. 373, W. H. Freeman, San Francisco, 1958.
55. J. Dyson, Inferferomefry as a Measuring Tool, Machinery Publishing Co., Brighton, 1970.
56. P. Mollet, Opfics in Mefrology, Pergamon Press, Oxford, 1960.
57. A. A. Michelson and M. Morley, Silliman J. 34, 333 (1887).
58. R. B. Leighton, Principles of Modern Physics, Chap. 1, McGraw-Hill, New York, 1959.
59. T. S. Jaseja, A. Javan, J. Murray, and C. H. Townes, Phys. Rev. 133A, 1221 (1964).
60. A. A. Michelson, “Measurement of the Diameter of Orionis with the Interferometer,”
Asfrophys. J. 53, 249 (1921).
61. J. Ojeda-Castatieda, “Foucault, Wire and Phase Modulation Tests” in OpticalShop Tesfing,
ed. D. Malacara, John Wiley and Sons, New York, 1978.
62. V. Grigull and H. Rottenkolber, “Two-Beam Interferometer Using a Laser,” J. Opf. Soc.
Am. 57, 149 (1967).
63. A. Cornejo-Rodriguez, “Ronchi Test’’ in Optical Shop Tesfing (D. Malacara, ed.), Chap.
8, Wiley, New York, 1978.
64. D. Malacara, A. Cornejo, and M. V. R. K. Murty, “Bibliography of Various Optical Testing
Methods,” Appl. Opf. 14, 1065 (1975).
48 INTERFERENCE
65. J. D. Briers, “Interferometric Testing of Optical Systems and Components: A Review,”

Opt. Laser Technol. 4, 28 (1972).
66. G. Schulz and J. Schwider, “lnterferometric Testing of Smooth Surfaces,” in Progress in
Optics, Vol. 13, Chap. 4, (E. Wolf, ed.), North-Holland Publ., Amsterdam, 1976.
67. D. Malacara, ed. Optical Shop Testing, Wiley, New York, 1978.
68. J. H. Bruning, “Fringe Scanning Interferometers,” in Optical Shop Testing (D. Malacara
69. K. Rosenhauer, “Measurement of Aberrations and Optical Transfer Functions of Optical
Systems,” in Advanced Optical Techniques, Chap. 18, Wiley, New York, 1967.
70. H. H. Hopkins, “Interferometric Methods for the Study of Diffraction Images,” Optica
Acta 2, 23 (1955).
71. A. A. Michelson, “Ditermination Expirimentale de la Valeur du Metre en Longeurs
d’Ondes Lumineuses,” Trau. Mem. Bur. Int. Poids Mes. 11, 1 (1895).
72. K. M. Baird and L. E. Howlett, “The International Length Standard,’. Appl. Opt. 2, 455
(1963).
73. G. R. Harrison, N. Sturgis, S. P. Davis, and Y. Yamada, “Interferometrically Controlled
Ruling of Ten-Inch Diffraction Grating,” J. Opt. SOC.A m . 49, 205 (1959).
74. J. N. Dukes and G. B. Gordon, Hewlett-Packard Journal 21, 2 (1970).
75. A. L. Hammond, Science 170, 1289 (1970).
76. V. Met, “Determination of Small Wedge-Angles Using a Gas Laser,” Appl. Opt. 5, 1242
(1966).
77. D. Malacara and 0. Harris, “Interferometric Measurement of Angles,” Appl. Opt. 9, 1630
(1970).
78. D. T. Tentori and M. Celaya, “Continuous Angle Measurement with a Jamin Inter-
ferometer,” Appl. Opt. 25, 215 (1986).
79. W. Krug, J. Rienitz, and G. Schultz, Contributions to Interference Microscopy (English
trans.), Hilger and Watts, London, 1964.
80. J. W. Foreman, Jr., E. W. George, and R. D. Lewis, Appl. Phys. Lett. 7 , 77 (1965).
81. C. P. Wang and D. Snyder, “Laser Doppler Velocimetry: Experimental Study,” Appl. Opr.
13, 98 (1974).
82. Y. Yeh and H. Z. Cummins, Appl. Phys. Lett. 4, 176 (1964).
83. G. Sagnac, 1. Phys. Radium 4, 177 (1914).
84. A. A. Michelson, Phil. Mag. 8, 716 (1904).
85. A. A. Michelson, Astrophys. J. 61, 137 (1925).
86. A. A. Michelson and H. G. Gale, Asrrophys. J. 61, 140 (1925).
87. W. M. Macek and D. T. M. Davis, Jr., Appl. Phys. Lett. 2, 67 (1963).
88. F. Aronowitz, Laser Applications, Vol. 1, p. 133, (M. Ross, ed), Academic Press, New
York, 1971.
89. J. T. Rowland and Agrawal, Opt. Laser Technol. 13, 239 (1981).
90. P. Connes, “Recherches sur la Spectroscopie par Transformation de Fourier,” Rev.
d’Optrque 40, 116 (1961).
91. G. A. Vanasse and J. Strong, “Application of Fourier Transformations in Optics: Inter-
ferometric Spectroscopy,” in Concepts of Classical Optics (J. Strong, ed.), p. 419, W. H.
Freeman, San Francisco, 1958.
92. P. Jaquinot and C. Dufour, J. Rech. Cent. Natl. Rech. Sci. 2, 91 (1948).
93. P. Fellgett, “A Propos de la Thiorie du Spectromitre Interfirentiel Multiplex,” J. Phys.
Radium 19, 187 (1958).
2. DIFFRACTION A N D SCATTERING
Daniel Malacara
Centro de lnvestigaciones en Optica. A.C
Apdo. Postal 948
37000 Leon, Gto. Mexico
2.1. Diffraction
Diffraction phenomena have been some of the most interesting and most
thoroughly studied manifestations of light in the history of physics. Many
good textbooks present this subject in detail like those by Born and Wolf,’
Meyer,2 Marechal and F r a n ~ o n ,and
~ Ra~leigh.~ The
, ~ earliest
known mention of this phenomenon appears in the works of Leonard0 da
Vinci (1452-1519).However, Grimaldi7 is the first one to seriously treat the
subject in his book. It is interesting to know that Newton* was aware of
Grimaldi’s observations.
Huygens’ was the first one to develop a wave theory (see Fig. l), although
he also seems to have been aware of Grimaldi’s observations, since he did
not attempt to explain the existence of diffraction bands. An attempt to
explain diffraction qualitatively was made by Young, 1 0 * ’ 1 - 1 2 who assumed
that the wave passes undisturbed through the aperture, except at the edges,
where the edge wave is formed. In this case, he thought that the interference
between the direct wave and the edge waves formed the diffraction pattern.
It was not until 1818 that FresnelI3 used the wave theory together with the
principle of interference to explain diffraction quantitatively. He assumed
mutual interference between all secondary waves, adding on the screen the
amplitudes of each Huygens wavelet, and taking into account any phase
differences.
The Huygens principle, modified in this way, is called the Huygens-
Fresnel principle; it leads to quantitatively good predictions of the light
distribution on the screen. This theory, nevertheless, is not perfect. One of
the problems is that there is no explanation as to why the wavelets do not
travel backwards to the source, and another is that the resultant phase on
the screen, computed by this theory, is 7r/2 less than the experimental
value.
49
M E T H O D S O F E X P E R I M E N T A L PHYSICS Copyright 0 1988 by Academic Press. Inc.
Vol. 26 All rights of reproduction in any form reserved
ISBN 0- 12-475971-8
50 DIFFRACTION A N D SCATTERING
FIG. 1. Huygen's explanation of diffraction.
2.1.1. Kirchhoff's Diffraction Theory

In 1876 Kirchhoff 14.15,'6.17 established a complete mathematical theory of
diffraction. He then proved that the Huygens-Fresnel principle could be
regarded as an approximate form of his theory. Only a brief description
will be given here, since very complete and detailed accounts can be found
in several books.' Kirchhoff's theory starts with the so-called Helmholtz-
Kirchhoff theorem derived from the scalar differential wave equation. He
considers an arbitrary closed surface and an observation point P inside it.
The idea is to evaluate the field at the internal point P, from a knowledge
of the field at the closed surface (see Fig. 2). A point light source is placed
outside the closed surface, at a distance r much greater than A. Kirchhoff
then shows that the amplitude U ( P )at this point is given by:
This expression may now be applied to the solution of diffraction problems

by considering the surface S as formed by three parts: portion A, the
aperture through which the light passes, portion B, the diffracting flat screen,
DIFFRACTION 51
FIG.2. Helmholtz-KirchoH theorem
and portion C, a spherical shell which intersects the diffracting screen and
which has its center at the observation point P.
The amplitude over the diffracting screen must be zero. Therefore the
integral over portion B must also be zero. The integration over the sphere
C also goes to zero when its radius goes to infinity.
If the angle between the incident and the diffracted rays is called 8, it
may be shown that:
cos( n, s) + cos( n, r ) = 1 + cos 8. (2.2)
Hence, Eq. (2.1) becomes
(2.3)
rs
which is the well-known Kirchhoff diffraction integral, where u is the
diffracting aperture.
We notice here a few differences between the results of this theory and
those of Fresnel. The first important difference is the presence of the complex
number i in front of the integral. This means that the phase of the observation
point P, obtained with the Kirchhoff theory, has a difference 7r/2 with
respect to the phase obtained with the Huygens-Fresnel theory. This also
means that the phase of the illumination at point P, using a very small
diffraction pinhole (only one Huygens wavelet), and the phase of the
illumination at the same point with no diffraction, differ by 7r/2.
Second, we have the presence of the factor (1 + cos 0)/2. Usually this is
called the inclination factor, first derived by Stokes. It implies that the
amplitude of each secondary wave decreases when the inclination angle
increases. This inclination factor appeared naturally in Kirchhoff’s theory
and is only postulated in the Huygens-Fresnel theory.
The third difference is the presence of the product l / r s in the integral in
Eq. (2.3) which would appear in the Huygens-Fresnel theory only when
the decrease of the amplitude, as l / r s , is considered. Experimental
verifications of Kirchhoff’s scalar theory of diffraction were made by Silver”
in 1962.
Even Kirchhoff’s theory is incomplete, not only because of the approxima-
tions made, but mainly because the vectorial transverse nature of light waves
is ignored.
The scalar theory of diffraction produces, in general, very accurate results,
but a further refinement is achieved if the vectorial nature of light20.2’.22.23324.2s
and the optical properties of the screen or stops are considered by means
of the electromagnetic theory.
2.1.2. Other Theories

As pointed out before, Kirchhoff’s theory is not perfect. PoincarC26and
S ~ m m e r f e l dshowed
~~ that Kirchhoff’s hypotheses were inconsistent,
because the theory does not reproduce the assumed values on the screen
and in the plane of the aperture. As described by Bouwkamp28 (including
numerous references), many modifications to Kirchhoff’s theory were made.
One of the most important modifications of Kirchhoff’s hypotheses was
made by K ~ t t l e r , ~ who
~ . ~ ’showed that it is more appropriate to prescribe
some discontinuities at the edge of the aperture rather than continuous
boundary values. This is called a saltus problem, and it is especially interest-
ing when considering the diffraction at a black, completely absorbing screen.
Another interesting theory again concerns the existence of Young’s edge
waves. Young postulated these waves when noticing that the edge of the
aperture appeared luminous when observed from points within the
geometrical shadow. It is interesting to know that Fresnel considered these
edge waves but finally discarded them.
The first one to seriously consider the edge waves again was M a e ~ , but ~’
he did not go very far. Sommerfeld” used the edge waves to find an exact
solution to the problem of diffraction by an infinite half plane. The problem
was further studied by R u b i n o ~ i c z ,confirming
~ ~ . ~ ~ the existence of Young’s
edge waves for flat or spherical diffracted wave fronts.
Miyamoto3’ and Miyamoto and developed even further the
Rubinowicz theory showing that Kirchhoff‘s diffraction integral can be
FRESNEL DIFFRACTION 53
FIG.3. Some Fresnel diffraction patterns of recognizable objects.
separated into two waves, one coming from the aperture and another one
coming from its edge. In other words, they showed that a diffraction wave
appears as arising from the scattering at the edge of the aperture. Descrip-
tions of the Miyamoto-Wolf theory can be found in R ~ b i n o w i c z . ~ ~ . ~ ~
Even a corpuscular quantum approach to the diffraction problem in
gratings has been given, namely by Duane.40
2.2. Fresnel Diffraction
Depending on the positions of the light source and the observation plane,
we may consider two types of diffraction. Fresnel diffraction occurs when
either the light source or the observation plane or both are at finite distances
from the diffracting aperture or object. Fraunhofer diffraction occurs when
both the light source and the observation plane or both are at finite distances
from the diffracting aperture. In the following sections, we will consider
some selected configurations for Fresnel diffraction. Figure 3 shows some
interesting Fresnel diffraction patterns.
2.2.1. Single-Slit Diffraction

Let us consider the configuration shown in Fig. 4, with a point source at
0 and the observing point at P. A direct ray OFP and a diffracted ray O Q P
meet at P, with an optical path difference (OPD) equal to the small line
segment Q R .
c - c
a l b
\
\
\
\
\
FIG;. 4. Fresnel diffraction geometry.
Assuming that the distances a and b are very large compared to the
height S, the segment QR may be approximated by the sum of the sagittas
of the two arcs passing through Q and R, respectively.
S 2 S2 a + b
OpD=QR=-+-=-
2a 2b 2ab
s2, (2.4)
giving a phase difference 6:
S=kOPD=
n( a +b ) S2.
(2.5)
abh
To simplify the mathematics, we consider this as a two-dimensional
problem. Defining a nondimensional variable u as:
we can write the phase difference 6:
(2.7)
If we divide the wave front into narrow parallel strips of width ds, we
may use the Huygens-Fresnel principle and add the contribution of all the
strips to the amplitude at point P. This sum is made by considering the
FIG.5. Total amplitude at point P.
phase differences between the different strips and by adding them vectorially
as shown in Fig. 5.
Each vector, representing the contribution of each slit, has a magnitude
directly proportional to the width ds and hence to dv. Each vector forms
an angle 6 with the x axis. The resultant amplitude at the point P is directly
proportional to the resultant vector R.
We can mathematically generate a curve, as in Fig. 5, that can be used
to find the amplitude at point P for any width and position of the diffracting
slit. Using Eq. (2.7), we find:
dx = dv cos 6 = cos (q) dv,
dy = dv sin 6 = sin (G) dv, (2.9)
and by integrating, we obtain:
x= lo" ($)
cos dv, (2.10)
y= lo" ($)
sin dv. (2.11)
56 DIFFRACTION A N D SCAITERING
f q A2
FIG.6 . Cornu spiral.
Equations (2.10) and (2.1 1 ) are the Fresnel integrals, whose numerical values
are given in table form in many book^.^'.^^
The curve produced with these integrals is the Cornu spiral shown in
Fig. 6.
Strictly speaking, we should have used the inclination factor (1 + cos 8 ) / 2
here, since the amplitude of each secondary wavelet decreases when the
angle 8 increases. We assumed, however, that 8 remains very small
everywhere, and therefore the inclination factor is always very close to 1.
The Corm spiral allows us to find the amplitude of the illumination at
the point P on the axis. If the diffracting slit is centered on the optical axis
and has a width S, we first compute u by using Eq. (2.6). Then we choose
two symmetrical points A , to A, on the Cornu spiral, such that their
separation measured along the spiral be equal to u. The distance from A ,
and A * , measured along a straight line, represents the amplitude of the
illumination at P.
If the diffracting slit is decentered off the optical axis, we again use Eq.
(6.6) to compute the decentering and the slit width in terms of u. Then two
points A; and A2 are chosen such that their positions as measured along
the curve corresponding to the slit width and to the decentering. Again, the
separation between these two points, measured along a straight line, rep-
resents the amplitude at P.
The whole diffraction pattern for a diffracting slit, with a width corre-
sponding to a certain value of u, is found by moving two points, A; and
A;, along the curve in such a way that the distance between these two points
along the curve remains constant and equal to u. The distance between the
two points in a straight line will change continuously to give the amplitude
of the whole diffraction pattern. The irradiance which is the square of the
amplitude is shown in Fig. 7(a). In order to relate the resultant irradiance
to the value of the unobstructed irradiance, the squares of amplitudes
obtained from Fig. 6 must be divided by 2.
If we have only straight edges, we may consider that we have a decentered
slit, with one edge on the optical axis and the other edge at infinity. In this
case, we have one fixed point 2, and a moving point A; on the Curnu
spiral. We then obtain the diffraction pattern in Fig. 7(b). If there is no
INTENSITY
I I
I r
GEOMETRICAL LIMITS
(A)
INTENSITY
G GEOMETRICAL EDGE
(B)
FIG.7. Diffraction light distribution in (a) a wide slit, (b) a straight edge.
diffraction obstacle, the intensity everywhere on the screen is represented

by the distance along a straight line from 2, to 2,. In this case, we can see
that the phase at the point P, with no diffraction, is ~ r / 4(45") ahead of the
phase at the same point P, produced with the infinitely narrow diffracting
slit, and is 7r/4 ahead of the phase at P produced with an infinitely small
diffracting pinhole.
2.2.2. Circular Aperture Diffraction

Fresnel diffraction by a circular aperture can be easily studied for a point
on the optical axis. In doing so, we can obtain very interesting results, as
we shall see. Using the same geometry as in Fig. 2, we can find the amplitude
at the point P by dividing the wavefront into annular zones concentric with
the point F, where each zone has a radius S, and then by adding the
contributions of each zone to the final amplitude at P.
From Eq. (2.5), the phase difference between the light on a ring with
radius S and the light through the center is
6 = KS2, (2.12)
where
K=
v(a +b ) (2.13)
abh .
The contribution of each zone is directly proportional to its area, and
this area in turn is directly proportional to its width ds. Taking this into
account, we can find a graph analogous to the Cornu spiral, which can be
easily shown to be given by:
[
x + y-- ;I*=[&l'~ (2.14)
This is a circle with its center on the y axis, tangent to the x axis, and with
radius A/2K. This means that, if the radius S of the diffracting circle is
increased continuously, the amplitude on the axis at the point P oscillates,
as shown in Fig. 8(a). This is physically unacceptable, because these oscilla-
tions in the amplitude do not occur when the diffracting aperture is very
large. The problem disappears when taking into account the inclination
factor (1 +cos 8)/2, because the radii of the circles will become smaller
when the aperture grows, increasing the angle 8. The curve is now a spiral,
as in Fig. 9. When the aperture becomes very large, the amplitude reaches
the value A/2K at the point P on the axis. The color at the center of the
diffraction pattern produced by a circular aperture and illuminated with
white light has been studied by Bergsten and H ~ b e r t y . ~ ~
F R E S N E L DIFFRACTION 59
AMPLITUDE AT
POINT P
-
A
2K
AMPLITUDE
AT POINT P
FIG.8. Amplitude variation on axis with a diffracting circle, (a) without obliquity factor,
( b ) with obliquity factor.
2.2.3.Fresnel Zone Plate Diffraction

The amplitude on the axis of a diffracting circular aperture oscillates
because some annular zones contribute constructively to the interference,
while others contribute destructively.
If we mask all the zones that contribute destructively, the irradiance at
the point P is drastically enhanced. Then the diffracting hole becomes a
Fresnel zone plate, which is illustrated in Fig. 10.
FIG.9. Spiral for a circular aperture.
F I G .10. Fresnel zone plate

Since the irradiance is greatly enhanced at the point P, it must be greatly

reduced at all other points by the energy conservation principle. This means
that a Fresnel zone plate acts like a lens and can even form images.
The outer edge of a dark ring occurs for 6 = nm, while the inner edge of
the same ring occurs for 6 = (n - 1/2)7r. Thus, from Eqs. (2.12) and (2.13),
it may be shown that the radii S of the limits of a dark ring n are given by
(2.15)
and
ab
s . =[ -
n'n a+b
h ( n - 1/2) (2.16)
Defining So= SIin a n d f = 2Si/h (for a zone plate with a transparent center
zone), we may show that
(2.17)
which is the equivalent of the thin-lens formula for the positions of the
object and the image, at a and b, respectively.
In spite of the close similarity between a thin lens and zone plate, there
are some important differences. The first difference for a lens is that the
focus is not unique, but there are an infinite number of them with focal
lengths fm, given by
2s;
fm =- m =integer, (2.18)
mh '
where rn is any positive or negative odd integer.
It should also be noticed that for every convergent focus there is a
corresponding diverging focus. In other words, a zone plate acts simul-
taneously as a convergent and a divergent lens. A very useful application
of zone plates is for alignment, as described by Van
A pinhole camera uses a small pinhole, instead of a lens, to form the
image. A consequence of Eq. (2.17) is that in order to optimize the resolution
and light-collecting power, the pinhole must have a radius S, and the
photographic-plate-to-pinholedistance should be f:
A Gabor plate is quite similar to a zone plate, with the difference being
that its transmittance does not change in steps but in a quasi-sinusoidal
manner. It is normally produced by the interference between a flat and a
spherical wave front.
62 D I F F R A C T I O N A N D SCATTERING
/- I/ I
-I+-POINT
COLLIMATOR I \‘I
DIFFRACTING
APERTURE
FIG.1 I . Arrangement to observe Fraunhofer diffraction.
2.3. Fraunhofer Diffraction and Fourier Transforms
When the light source and the observing screen are both at infinite
distances from the diffracting aperture, we say that we are observing Fruun-
hofer difruction.
The light source may really be at infinity as in the case of a star, but
more frequently the light source is placed “optically at infinity” by means
of a collimating lens. In this case the light source is at the focus of a
convergent lens, as shown in Fig. 1 1 . The observing screen cannot be really
placed at infinity, but it is again done optically by placing it in the focal
plane of a convergent lens. The fact that the observing screen is at infinity
may be interpreted by saying that what we observe on the screen in Fraun-
hofer diffraction is the angular distribution of the light after the diffracting
aperture.
Figure 12 shows the slow transition from Fresnel to Fraunhofer diffrac-
tion, with a point source at infinity, by increasing the distance b of the
observing screen from the diffracting aperture. We say that we have reached
the far field or Fraunhofer diffraction region when b >> D‘/A, where D is
the diameter of the diffracting object.
The inclination factor is not important in the case of Fraunhofer diffrdc-
tion, and if we also take the case of normal incidence ( r = constant), the
F R A U N H O F E R DIFFRACTION A N D FOURIER T R A N S F O R M S 63
0
44
diffraction integral of Eq. (2.3) becomes
U(P)=i{ { AcdS,
S (2.19)
where A is a constant whose magnitude is directly proportional to the

amplitude on the diffracting aperture plane. If the aperture is on the x, y
plane, and the amplitude on this plane is a function of x and y , we consider
A also to be a function of x and y. The distance s is variable enough to
produce significant changes in the phase ks, but constant enough to include
l / s within the value of A. Thus,
U (P ) = J J ,~ ( xy ,) eiksdx dy. (2.20)
We may show that

s = x s i n O , + y s i n O,.; (2.21)
and by defining
k, =k sin 0,
and
k,, = k sin O v , (2.22)
we can write Eq. (2.20) in the form
U(k,, k,) = i { dx dy.

A(x, y ) e’(k\r+k”’) (2.23)
On the other hand, from Fourier theory, the function A(x, y ) can be given
by:
(2.24)
The functions A(x, y ) and U (k,, k,) are two-dimensional Fourier trans-
forms of each other.
Thus, except for unimportant constants, we can say that the angular
diffraction pattern is the Fourier transform of the amplitude function on
the diffracting aperture.
In a similar way, for diffraction apertures in one dimension, we may write
the one-dimensional Fourier transforms as
(2.25)
FRAUNHOFER DIFFRACTION A N D FOURIER TRANSFORMS 65
and
U ( k , ) e - l k \ ' dk,. (2.26)
Fraunhofer diffraction has been widely studied by many authors, among

whom we must mention S c h ~ e r d , ~whose
' excellent work includes many
fine handmade color diffraction pictures. This work was reviewed by Hoover
and Harris.46Harris4' has computed diffraction patterns for many different
apertures. The colors observed when the aperture is illuminated with white
light have been studied by Bergsten and P ~ p e l k a . ~ '
2.3.1. Single-Slit and Rectangular Aperture Diffraction

Let us take a slit on the x, y plane centered on the y axis. If the slit width
is 2 a , we can write Eq. ( 2 . 2 5 ) in the form:
U ( k , )= A ['
J -0
elk\' dx, (2.27)
assuming that the amplitude A ( x ) remains constant inside the slit. Integrat-
ing, and using Eqs. ( 2 . 2 1 ) and ( 2 . 2 2 ) , we obtain:
U (k ) = U,
sin( k a sin 6)
k a sin 6 ' I (2.28)
where U, is a constant; Fig. 13 shows this amplitude and its corresponding

irradiance versus k a sin 6.
The first dark zone in the irradiance pattern can be shown to be at an
angle 6 given by
A
sin 6 =-. (2.29)
2a
Therefore, the diffraction pattern becomes narrower when the slit width 2 a
increases.
For a rectangular aperture, with width 2 a and length 2b, we can use Eq.
( 2 . 2 3 ) to obtain
U(O,, 4.) = U"

sin( k a sin 0,)
k a sin Ox ][
sin( k b sin Or)
k b sin 6,
3. (2.30)
(B)
k o sin 6
FIG. 13. Fraunhofer diffraction pattern for single slit. ( a ) Amplitude distribution.
( h ) lrradiance distribution.
2.3.2.Circular Aperture Diffraction

The diffraction pattern of a circular aperture may be found by integrating
Eq. (2.23) inside the aperture, thus obtaining
u(e)=2u0 J,(ka[
sin 0 )
ka sin 0 1’
where a is the radius of the aperture and J , ( x )is a first-order Bessel function
(2.31)
- AIRY DISC
-- ka sin 8
(6)
F l c . 14. Fraunhofer diffraction pattern for a circular aperture. ( a ) Amplitude aperture.
( b ) Irradiance distribution.
of x. The function U ( 0 ) and its corresponding irradiance U 2 (0 ) are plotted

in Fig. 14.
The first dark ring, which outlines the central spot or Airy disc, studied
by Airy in 1835,49 can be seen to have a direction given by
h A
sin 0 = 1.22 -= 1.22 - (2.32)
2a D’
where D is the diameter of the aperture.
If the pattern is formed at the focal plane of a lens of clear aperture D

with focal length A the radius r of the Airy disc is:
f
r = 1.22h -. (2.33)
D
If a collimated beam of light illuminates a well-designed convergent lens,
the image is not a point but a Fraunhofer diffraction pattern of the free
aperture. This is the reason why the resolving power of any perfect optical
instrument (an instrument with no geometrical aberrations) is not infinity,
but is limited by the size of the diffraction image.
Sletten and Blacksmithso have shown that a paraboloidal microwave
antenna has also an image at the focus, limited in size by Fraunhofer
diffraction. It should be pointed out, however, that in a paraboloidal
antenna, the f / D ratio is between 0.2 and 0.5, and thus the image structure
differs from the Airy pattern because of the large angle of convergence, as
described by Minnett and Thomas.’’
Figure 15 shows the Fraunhofer diffraction patterns produced by four
different aperture shapes.
2.3.3.Babinet Principle
Two pupil functions, A l ( x , y ) and A 2 ( x , y ) are complementary to each
other if
A,(x,y)+A,(x,y)=constant. (2.34)
If the amplitude is constant over the whole x, y plane, then we say that
two diffracting apertures are complementary if the transparent parts in one
are the opaque parts in the other, and vice versa. For example, the com-
plementary aperture of a circular aperture is the whole x, y plane, excluding
only the opaque disc of the same size as the circular aperture.
Let us assume that we have two arbitrary complementary apertures, giving
diffraction pattern U , ( P ) and U , ( P ) . If U ( P ) is the amplitude of the
illumination produced without any diffraction apertures, the Babinet prin-
ciple’’ says that
(2.35)
which can be very easily shown to be true.

Applying this principle to the Fraunhofer diffraction patterns produced
by a circular aperture and an opaque disc, we can see that U ( P ) is a very
high and narrow peak at the origin, because all the light will travel in a
single direction if there is no diffracting aperture.
FIG. 15. Fraunhofer diffraction pattern for four different apertures.
Thus, except for the point at the origin, the Fraunhofer diffraction patterns
of complementary apertures are identical in shape but with amplitudes of
opposite sign.
2.3.4. Parseval‘s Theorem

As in any other physical process, energy is preserved, but it is probably
more evident in Fraunhofer diffraction, where all computations are made
by means of Fourier transforms as given by Eqs. (2.23) and (2.24). The
following result we quote here can be mathematically proved from the
definition of the Fourier transform.
(2.36)
70 D I F F R A C T I O N A N D SCATTERING
This equation is known as Parseval’s or the energy theorem, because it

may be used to show that energy is always conserved in Fraunhofer diffrac-
tion experiments.
2.4. Diffraction Gratings
A diffraction grating is just a diffraction screen formed by many parallel

and equidistant slits. The principle of the diffraction grating was discovered
by Rittenhouse” and later rediscovered by F r a ~ n h o f e r A. ~good
~ modern
review of diffraction gratings is found in the book by Hutleyss and in the
papers by Richardson” and Welford.” We could study the diffraction
gratings by applying Kirchhoff’s integral directly to obtain all their proper-
ties. However, by studying them using more elementary principles, we gain
better physical insight.
2.4.1. Angular Distribution of Light

As shown in Fig. 16, the maxima of irradiance occur at angles such that
the interference between the different slits is constructive. The interference
is assumed to take place at infinity, since we are considering Fraunhofer
diffraction.
In Fig. 16, we have represented only two consecutive slits. It is easy to
see that the maxima of the intensity are at an angle such that the optical
path difference between two consecutive slits is given by
OPD = CA - BD = mA, (2.37)
FIG. 16. Maxima of irradiance in a diffraction grating.

DIFFRACTION GRATl NGS 71
where m is any integer including zero, commonly called the order of

interference.
If d is the separation between the centers of the slits, and 8 , and Or are
the angles of incidence and diffraction, we can see that
d(sin -sin 8,) = mh. (2.38)
The angles 8 , and O2 are both positive, as shown in Fig. 16, because they
are each measured counterclockwise from the normal to the grating.
The complete angular distribution of the light may be obtained by adding
the amplitude contributions of each of the slits at a given angle 0,. Assuming
that the grating has N slits and that 6 is the phase difference between two
consecutive slits, the total amplitude at an angle 02 is:
E = A[1 + elS+ e2"+. . . + e ( N p l ) 'I,S (2.39)
which can be transformed to give
(2.40)
Irradiance is then given by
(2.41)
which can be shown to be equal to
(2.42)
i n order to visualize the form of this function, we have plotted the

numerator, the denominator, and their quotient for a grating with five slits
( N = 5) in Fig. 17.
The high peaks of the irradiance occur at values of 6, such that
_6 -- mr, m =integer, (2.43)

2
since at those points,
lim
S/2-mrr
[
sin'( y)
(I") (2.44)
and hence the peaks have a maximum irradiance A 2 N 2 .

DIFFRACTION A N D SCATTERING
I
I
I
I
I
I I I
I I I
A A
m=O m=l m=2 m =3

FIG.17. Irradiance in a grating with five slits.
The phase difference 6 is given by:

2~rd(sinO1 -sin 0,)
S=kOPD= (2.45)
A
Therefore, from Eq. (2.43) we may obtain Eq. (2.38) as expected.
There are ( N- 2) secondary peaks with much lower irradiance between
any two of the main peaks. If N is increased, the number of secondary
maxima increases, but their height decreases very rapidly.
Figure 18 shows the diffraction patterns produced by four gratings with
a different number of slits.
2.4.2. Chromatic Dispersion and Resolving Power

If light of several wavelengths (several colors) impinges upon a diffraction
grating at a given angle of incidence O,, there will be a different angle of
diffraction O2 for each color. This property is used to make spectrographs
and spectrometers using diffraction gratings instead of prisms. From Eq.
DIFFRACTION GRATINGS 73
FIG. 18. Diffraction patterns for four different arrays of slits.
(2.38), the chromatic dispersive power can be given by:

A02
--
m
(2.46)
AA - d cos 02'
The chromatic dispersive power increases linearly with the order of
interference. In a spectrograph using gratings, the angle O2 does not change
much within a given order; therefore, cos O2 may be considered to be a
constant. Thus A 0 J h A is approximately constant. In other words, the
chromatic dispersion is approximately linear with A, as opposed to that for
a dispersing prism.
The chromatic resolving power of a grating may be computed by means
of the Rayleigh criterion illustrated in Fig. 19. This criterion may be stated
by saying that two spectral lines with slightly different wavelengths are just
resolved when the maximum of one of the lines coincides with the first
minimum of the other.
We can see that the phase difference between the peak of the line and
the first minimum is given by:
(2.47)
but from Eq. (2.45), this difference in phase corresponds to a difference

SUM OF TWO LINES
FIG. 19. Two lines just resolved by a diffraction grating.
he2 in the angle 0 2 ,given by

2 .rrd
A~=--cos 8,AO2. (2.48)
A
From these two equations, the angular difference between the line maximum
and its first minimum is
A
(2.49)
= - Nd cos B2’
If two spectral lines with wavelength difference AA are separated by just

this angle, from Eqs. (2.49) and (2.46), we obtain:
_A -
- mN. (2.50)
AA
This is the well-known expression for the resolving power of a grating.
If we define the effective width We, of a grating as
we,,= w cos 82 = Nd cos 8 2 (2.51)
and use Eq. (2.49) we find a n alternative expression showing that the
resolving power of a grating is limited only by its effective width and the
wavelength, as shown by Michelson,5s
(2.52)
We may also write:
(2.53)
which shows that the chromatic resolving power of a grating is directly

proportional to its effective diameter times its dispersive power.
2.4.3. Energy Distribution Among Different Orders

Until now, we have been considering the grating formed by infinitely thin
slits. If the width of each slit is 2a, we cannot assume that the light diffracted
by each slit travels in all directions with equal irradiance but according to
the diffraction pattern for a single slit [Eq. (2.28)]. Thus, the irradiance in
a direction O will be the product of Eq. (2.42) times the square of Eq. (2.28):
sin'( y) sin ka(sin O1 -sin 0,)
sin (Ifi) [ka(sin O1 -sin O,)] '

(2.54)
We have plotted this irradiance in Fig. 20(a) for a grating with five slits
of widths 2a = d/3.
A line with some order of interference does not appear when its direction
coincides with the direction of a zero of the second factor in Eq. (2.54).
The zeros of the second factor may be shown to occur at angles such that
2a(sin O1 -sin Oz) = nh, n = a n y integer excluding zero. (2.55)
The line positions are given by Eq. (2.38). Thus, the missing order numbers
are:
d
m = n--, (2.56)
2a
where n is any integer, excluding zero, such that m is also an integer. The
influence of the slit width is illustrated in Fig. 21, which shows the diffraction
patterns for two slits with different widths but with the same separation.
2.4.4. Ghost Lines in Diffraction Gratings

Often spurious spectrum lines are seen in grating spectra, due to non-
constant separation in the slits. There are two main types of these so-called
ghost lines.
76 DIFFRACTION A N D SCATrERING
-6 -4 -2 I 2 4 6 m
0
(B)
FIG.20. lrradiance distribution in a grating with slits with finite width. ( a ) Finite-size grating.
( b ) Infinite-size grating.
Rowland ghosts,59 which have been studied by Michelson"" and Ander-

son,6' are due to periodic errors in the spacing of the slits because of
imperfections in the ruling engine. These spurious lines appear as two faint
lines symmetrically located with respect to the main or parent line.
Lyman ghosts have also been studied, among others, by Anderson6' and
by Runge.62They are due to nonperiodic errors in the line spacing. These
lines are usually observed at large distances from the parent lines.
0
5
rn
m
LI1
5
0
5
A ) T R A N S M I S S I O N TYPE B ) REFLECTION TYPE
FIG. 22. Basic types of diffraction gratings.
There are some other types of spurious lines and ghosts that may
appear in a diffraction grating spectrum, but all of them are due to ruling
imperfections.
2.4.5. Different Types of Diffraction Gratings

There are reflecting as well as transmission diffraction gratings, as shown
in Fig. 22.
Both of these gratings usually have the groove faces made at an angle
(called the blaze angle) from the plane of the grating. The purpose of the
blaze angle is to shift the maximum of the light distribution among the
different orders, from the zero order to some other desired order, as first
described by W00d63.h4and by Babco~k.'~
To avoid absorption and other undesirable effects due to the light trans-
mission through glass, the grating type used most is the reyedon grating.
The blaze angle a in this case is given by
6 ,- 6 2
a=- (2.57)
2 '
and the total deflection angle is
+
y = o1 62. (2.58)
Then, from Eq. (2.38), we find:
mh - 2 cos -
_ Y sin a.
(2.59)
d 2
A case of special interest arises, y = 0; ( a = el = &), reducing Eq. (2.59) to
mh = 2 d sin a = 2 h , (2.60)
RESOLVING POWER O F OPTICAL I N S T R U M E N T S 79
GROOVE
\
.
B L A Z E D GRATING I/
ECHELLE
GRATING
FIG.23. Blazed reflection grating and echelle grating.
where h is the step height as shown in Fig. 23. The blaze angle is often
specified as the wavelength that fulfills this condition in the first order.
WoodG6pointed out that very efficient gratings for the infrared could be
obtained by using coarse-blazed gratings with this condition ( y = 0). Since
the light is reflected on the grooves' faces in a mirror like manner, he called
them echellettes, Harrison6736x considered that since resolving power depends
only on the total grating width, the echellette could be made with spacings
as large as about 100 grooves/mm. The typical blaze angle for these gratings,
named echelles by Harrison, is 63"26'. Echelles are very useful in order to
achieve high resolving powers in the infrared.
Gratings can also be concave and can be used in many configurations
that will not be described here. The reader is referred to the book by Sawyer69
for a more complete description.
The art of grating manufacture is a very complicated and interesting one.
The reader can consult the excellent papers by Harrison,67768S t r ~ n g , ~
and
"
Stroke7' for details.
In closing the subject of diffraction gratings, it should be mentioned that
there are also two- and three-dimensional gratings. Figure 24 shows the
diffraction patterns for two two-dimensional arrays of apertures.
2.5. Resolving Power of Optical Instruments
Because of diffraction effects, the resolving powers of all optical instru-

ments, even if they have no geometrical aberrations, are limited by the sizes
of their apertures.
FIG.24. Diffraction of a two-dimensional array of holes. (a) A regular array and (b) its
diffraction pattern. (c) An irregular array, and (d) its diffraction pattern.
If a lens is free of aberrations of any kind, its resolving power is determined

by the size of the Fraunhofer diffraction pattern produced by the lens source
in the Airy disc. We can see that the angular resolution of a lens in a
telescope is limited only by the diameter D of the lens.
If we define the image distance ratio ( L ' /T )as (L'/ D ) , which becomes
equal to the focal ratiof/T when L ' = J we may write the linear radius of
the Airy disc on the image plane as:
r = 1.22A (b). (2.61)
Let us consider two very close stars, with their Airy discs overlapping in
the image plane. The light of the two stars is mutually incoherent, so that
there is no fixed phase relation between the two images. The resultant
combined image is then obtained by adding the intensities of the two images.
RESOLVING POWER O F OPTICAL INSTRUMENTS 81
SUM OF TWO IMAGES
-
RADIUS OF AIRY DISC
FIG. 25. Two point images just resolved.
Rayleigh established that under these conditions of incoherent illumina-

tion, the two images are just resolved when they are separated by a distance
equal to the radius of the Airy disc, as illustrated in Figs. 25 and 26.
Until now, we have considered images produced by lenses with no
geometrical aberrations that produce perfectly spherical wave fronts. When
the wave fronts have deformations smaller than A / 4 (Rayleigh criterion),
the image is not much different from that produced by a perfect lens, as
shown by Marechal.’* If the wave-front deformations are large compared
with the wavelength, the image is that predicted by geometrical optics. The
intermediate case, however, is extremely complicated, as described by
Marechal and F r a n ~ o n Figure
.~ 27 shows the appearance of an image of a
point source in the presence of aberrations.
The angular resolution of a telescope increases, that is, it resolves a
smaller angle when the diameter is increased. However, a telescope with a
resolution greater than that permitted by the atmospheric “seeing” or
turbulence is unnecessarily precise, because in this case the resolution limiter
will be the seeing and not the telescope. Depending on the geographic
place, the seeing is of the order of a few tenths of a second of arc, which
may be matched by a telescope with a diameter of about 40cm. To take
full advantage of the resolution of a telescope with diameters much larger
than 40 cm, we have to use it outside of the terrestrial atmosphere. However,
very large telescopes are made because the light-gathering power increases
with the square of the diameter.
F i c . 2 6 . Two point images with equal irradiance. ( a ) Unresolved. ( b ) Just resolved.

(c) Completely resolved.
2.5.1. Optical Transfer Function

The different points in the image formed by a photographic lens are also
mutually incoherent. Therefore, we say that the object is incoherently
illuminated. The quality of the image in a photographic lens is seldom
completely determined by the maximum resolving power of the lens. For
example, two lenses might be able to resolve the same fine detail, but with
very different contrast. As in electronic amplifiers, it is not enough to know
ill
.-.-
C
g
0)
D
L
0
N
P
d
!i
the high-frequency cutoff of an audio amplifier, but the complete frequency

response curve must also be known.
The optical equivalent for lenses of the frequency response curve for
electronic amplifiers may be found by defining the spatial frequency content
of an image. To illustrate this concept, we may take any picture and take
a measurement of the irradiance along a straight line.
The graph of the irradiance, with respect to x, may be considered, by
means of Fourier analysis, as the superposition of many sine waves as
shown in Fig. 28, with different wavelengths L. Then, we define the spatial
frequency w as:
w=-
2T
(2.62)
L’
The frequency response curve of a lens described first by O’Nei1173 in
1954 is called the optical transferfunction (OTF) and is defined as the ratio
of the harmonic components in the image to those in the object. It has two
m
-wK
3
0
Lb
FIG. 28. Spatial frequency content of an image.

parts. The plot of the amplitude ratio versus spatial frequency is often called
the modulation transferfunction (MTF). The second part, the plot of the
phase ratio, is less frequently used.
The best way to measure the MTF is to form the image of an object that
contains all spatial frequencies in equal amounts. The plot of the amplitudes
of the spatial frequency components in the image thus formed would then
represent the frequency response of the lens.
An object that contains all spatial frequencies in equal amounts is a point
source, as may be proved without much difficulty. The image of this point
source is called point spreadfunction. The point spread function is the same
as the Fraunhofer diffraction pattern of the lens aperture only if the lens
is perfectly free of geometrical aberrations.
We shall assume a point source at infinity. The point spread function in
the focal plane of a lens with focal length F, with coordinate x F and y F
can be obtained from Eq. (2.23) if we first obtain the amplitude as follows:
A(XF, yF) =
II-: T ( ~y ), e ~ ( k / F ) ( r ~ f i + i ,)i ldx dy,
where T ( x ,y ) represents the pupil function, including any deformations of

(2.63)
the wave front due to aberrations of the lens. The point spread function
S ( x F ,y F ) is then given by:
and therefore, from Eq. (2.63):
. [I -‘m
The variables x and y in these integrals do not remain after the integration
1
T * ( x ,y ) e - l ( k I ~ ) ( X X f i + Y v F ) d x d y . (2.65)
and, therefore, may be replaced by the symbols 2 and J in the second

integral, so that we may move the entire expression under the same integral
signs.
(2.66)
Once the point spread function is computed from this expression or
becomes known by some other means, the optical transfer function
F ( w , , w , , ) may be computed by using the Fourier transform of S ( x F ,y F )
as follows:
F(w,,w,) =
I:I s ( x F y, F )e - “ W \ X F + ’ > ” ,
There is a simpler way to compute the OTF when T ( x , y ) instead of

) dXF dYF. (2.67)
S ( X , , y F )is known, as will be described here. If the function F ( w , , w , ) is

known, the point spread function can be computed by taking the Fourier
transform of F ( w , , w ) ) , as follows:
Then, except for an unimportant constant, Eq. (2.66) becomes identical

if we define:
k
w, =-(x-Z) (2.69)
F
and
k
w , =- ( Y
F
-?I, (2.70)
obtaining thus:
If a lens is perfectly free of aberrations, we can set T ( x , 1 ) = T*(x,y ) = 1

inside the clear parts of the aperture. Then F ( q , w.,.) is the common area
of two sheared images of the aperture, as in Fig. 29.
Summarizing, the optical transfer function is the two-dimensional Fourier
transform of the point spread function and also the autocorrelation of the
pupil function.
In Fig. 30, we have the optical transfer function of a perfect circular lens
for various defocusing magnitudes. The cutoff frequency in focus is given by:
kD
w, =-. (2.72)
F
2.6. The Abbe Theory of the Microscope
When we defined the resolution of a lens, we assumed that the two

neighboring points to be resolved were perfectly incoherent to each other.
T H E A B B E T H E O R Y OF T H E MICROSCOPE 87
COMMON AREA: F(wx,wy)
FIG.29. Computation of the optical transfer function of an aberration-free lens.
\ \ INFOCUS-,
\ \
0 I I 1 I I I I dr
-.2 -
0 0.2 0.4 0.6 0.0
(i&P
FIG.30. Some optical transfer functions.
FIG. 31. Abbe theory of the microscope.
The appearance of the image is quite different if the two points are coherent.
This is, for example, the case if the object is illuminated by a single point
source or a laser.
Ernest Abbe74 developed a theory, usually called the Abbe theory of the
microscope, which allows the computation of the resolving power of a lens
when the object is illuminated with coherent light. His main assumption
was that the spatial frequency structure of the object to be imaged acts like
a diffraction grating when illuminated by a coherent light source, as shown
in Fig. 31.
If the object acts as a diffraction grating, the light will be diffracted into
many beams with different orders of diffraction. In this case the image may
be regarded as the interference pattern arising from the superposition of
all diffracted beams.
If the lens is so small that only the zero order of diffraction passes through
the lens, no interference takes place at the image plane, and therefore no
image is formed.
The object structure is first detected at the image plane when the lens is
large enough to let the two first orders of diffraction pass through the lens.
Under these conditions, the image is not an exact reproduction of the object,
because the orders of diffraction with magnitude greater than 1 are not
reaching the image plane. The image, in this case, has a sinusoidal profile
with a spatial frequency equal to that of the object (see Fig. 32).
When the lens is larger, so that the order of diffraction + 2 and -2 passes
through the lens, the image will more closely resemble the object, as shown
in Fig. 32.
T H E A B B E T H E O R Y O F T H E MICROSCOPE 89
OBJECT FIRST ORDER IMAGE THIRD ORDER IMAGE

(A) (B) (C)
FIG.32. Explanation of Abbe theory.
The resolving power limit may be easily found by using the diffraction
grating equation, (2.38) for the first order rn and normal incidence:
d=- A (2.73)
sin 8’
where A, is the wavelength in the object medium, d is the minimum object

diameter, and 8 is the angular radius of the lens, as seen from the object.
Alternatively, we can write this diameter d as
(2.74)
where A. is the wavelength in vacuum, N is the refractive index in the

object medium, and NA is the numerical aperture of the lens.
We can see that the resolution is greater when the object is illuminated
with incoherent light. The resolution with coherent illumination is almost
half of that obtained with incoherent illumination. Another difference
between illumination by coherent and incoherent light is that the optical
transfer function in the coherent case is constant for all spatial frequencies,
dropping suddenly to zero at the cutoff frequency, while in the incoherent
case, it decreases continuously as the frequency increases. It must be pointed
out, however, that the differences between the two types of illumination are
so complex that a superficial examination must be avoided. Comparison
should be limited only to certain particular aspects.
The resolution in the case of partial coherence (illumination with a mixture
of coherent and incoherent light) is very interesting and has been studied
by many authors, like
Ar2
A6 2
A) NORMAL MICROSCOPE B ) P H A S E C O N T R A S T MICROSCOPE
FIG.33. Amplitudes in a phase-contrast microscope.
2.6.1. The Phase-Contrast Microscope

A logical application of the Abbe theory of the microscope is the phase-
contrast microscope developed by Zer~~ike,’~.’’ (he received the 1953 Nobel
Prize in Physics for this development).
This microscope permits the observation of transparent objects without
staining. It is only required that the object’s index of refraction be different
from that of the medium in which it is immersed. If the object has a refractive
index different from that of the surroundings, a wave front passing through
it becomes deformed because of the different optical paths. Thus the light
becomes diffracted as in any other kind of object. No image is visible,
however, because the interference at the image plane produces a constant
amplitude everywhere.
To understand this, we might consider (see in Fig. 33) that the amplitude
A, at a given point on the image is the sum of the nondiffracted amplitude
A,, (zero order) and the diffracted amplitude Ad (nonzero order). The
diffracted amplitude changes from point to point on the image, because it
contains the information about the object structure. The nondiffracted
amplitude is constant in phase and amplitude everywhere on the image
plane. The resultant amplitude A, has the same magnitude for all image
points, even though Ad and the phase difference A 6 are variables.
The object is transparent, but by shifting the phase of the nondiffracted
wave by an angle 7r/2, as we see in Fig. 33, the resultant amplitude A, will
be a function of the phase shift As. The object then becomes visible,
especially at the edges, where the light is diffracted.
As shown in Fig. 34, the phase shift in the zero order of diffraction is
obtained by means of a “phase plate” placed at the back focus of the
microscope objective. The phase plate is a thin glass plate, on which a ring
THE ABBE THEORY O F THE MICROSCOPE 91
CON DENSER OBJECTIVE
ANULAR OBJECT PHASE IMAGE

DIAPHRAGM PLATE PLANE
FIG. 34. Optical arrangement in a phase-contrast microscope.
of some dielectric material has been evaporated to produce the phase shift.
The illuminating system has a diaphragm with an annulus aperture and
a condenser. This illuminating system is designed so that all the nondiffrac-
ted light passes through the ring on the phase plate. The diffracted light
tends to pass out of the ring on the phase plate. Figure 35 shows the objected
observed in a normal and in a phase-contrast microscope.
2.6.2. Spatial Filtering of Images

This is a technique based on the Abbe theory of the microscope. We have
seen before that any irradiance distribution may be considered as the
(A) ( B)
FIG. 35. Pictures taken using a microscope. (a) Normal. (b) Phase contrast.
FRONT FOCAL PLANE BACK FOCAL PLANE

OF FOURIER L E N S OF FOURIER L E N S
FIG.36. Spatial filtering of images.
superposition of the light diffracted by the many spatial frequencies con-

tained in the object. Alternatively, it can be said that each order of diffraction
contributes to the formation of the image with a different spatial frequency.
From the point of view of Fraunhofer diffraction, it is important to
remember that the angular distribution of the light diffracted by an aperture
is the two-dimensional Fourier transform of the amplitude distribution over
the plane of the diffracting screen. Because of this property, if a convergent
lens is illuminated with a collimated beam of light, the light distribution
on the back focal plane is the two-dimensional Fourier transform of the
amplitude distribution over the lens plane, or pupil function, except for a
phase factor. The phase factor can be eliminated in order to obtain the
exact Fourier transform if the diffracting object is not in contact with the
lens, but in front of it on the front focal plane, as shown in Fig. 31.
The spatial filtering of images is used to analyze the spatial frequency
content of any image and even to remove undesired frequencies by means
of appropriate stops at the back focal plane of the Fourier lens. This is
usually done with the optical arrangement in Fig. 36. The object to be
filtered is at the front focal plane of the lens and the two-dimensional
Fourier transform is at the back focal plane, where the filtering stops are
placed. Examples of spatial filtering are shown in Figs. 37 and 38.
There have been a great number of applications for the use of this
technique in the analysis of many graphic results. For example, in seismic
studies where a Fourier transform of the picture is desired. It is often used
to remove the noise from a picture if the spatial frequency of the noise is
known.
2.7. Scattering
Most of the light that our eyes receive comes not directly from the light
source but through scattering. Looking at almost any object, what we see
SCATTERING 93
FIG.37. Spatial filtering of a two-dimensional grid.
is the light scattered from its surface. If we look at the blue sky or at clouds,
we also see scattered light. These examples give us an idea of the great
importance of scattering phenomena. Furthermore, the subject is very com-
plicated, and many good books have been written covering it, for example,
those by Van de Hulst” and Kerker.79
Scattering phenomena may be classified according to the size or shape
of the scattering object or to the wavelength. In general, scattering is an
anisotropic phenomenon, depending on the shape and orientation of the
FIG.38. Spatial filtering of a picture with TV scanning lines
object and on the orientation of the plane of polarization. The shapes of

scattering objects which are most easily analyzed mathematically are
spheres, ellipsoids, and cylinders.
Miex" developed a very complete theory for scattering by spheres of
arbitrary size, using electromagnetic theory.
2.7.1. Scattering by Large Objects

Depending on the size of the object, several models could be used to
explain scattering. If the object is very small compared with the wavelength,
the electric field of the wave forms electric dipoles with each particle. This
theory will be considered in Section 2.7.2.
If the size of the object is of the order of the wavelength or larger, the
main mechanism producing scattering is diffraction. As shown in Fig. 24,
if a scattering cloud is formed of many spherical particles of the same
diameter, their size can be found by examination of the diffraction pattern.
Using this idea, Young designed an instrument called eriometer to measure
the size of blood corpuscles.
SCAnERING 95
Young's eriometer does not work for particles of varied sizes, because
the Fraunhofer diffraction pattern is formed from the irradiance sum of all
the individual patterns. Parrent and Thompson" and Silverman et aLX2
pointed out this problem and proposed to observe the pattern in the far
field (s >> n 2 / A ) of the individual particles but close enough to be in the
near field of the whole sample. Under those conditions, the diffraction
patterns of each particle will be in different places instead of all being
superimposed. It is interesting that the diffraction patterns are not Airy
patterns, because they may be interpreted as Fresnel patterns, or alterna-
tively, as Fraunhofer patterns with a coherent background.
If the object is very large compared with the wavelength, the mechanisms
producing scattering are diffraction on the edge of the object and refractions
and reflections inside it. These particles may be water drops, such as in fog,
rain, mist, and haze.
The rainbow is one example of scattering by water drops. This
phenomenon has been considered by poets as well as scientists. The scientific
description is often assumed to be an exercise in geometrical optics, but it
is much more complicated than that, as shown by Khare and Nussenzveigs3
and Van de H ~ l s t . ' ~
The best approximate theory to explain the rainbow was formulated by
Airy. An exact but complicated theory is obtained applying the general Mie
theory.
The glory is another beautiful atmospheric phenomenon due to scattering
by water drops, as described in the book by Mir~naert.'~ It is observed as
a rainbow surrounding the sun or moon, covered by a thin veil of clouds.
More often, persons standing on a high point, observe their shadows
projected onto low clouds, showing a colored ring around each of their
heads. The glory can be seen by each observer, with the ring only around
his or her own head.
2.7.2. Scattering by Small Particles

When an electromagnetic wave passes through matter or a cloud of
particles much smaller than the wavelength, its electric field acts on each
molecule or particle, displacing the positive charges in the opposite direction
to the negative charges. The resulting deformed molecule or particle may
thus be considered as an electrical dipole, formed by two charges of opposite
sign joined by an elastic force.
An electrical dipole is the system formed by two charges to a -9 with
masses M and m, respectively, and joined by an elastic force ky, where k
is a constant, and y is the separation of each charge from the equilibrium
position. We shall assume that M >> m.Thus, if the electric charges are free
of any friction or driving force, their movement equation would be

m -d=2Y- k (2.75)
dt2 ’
with the solution
y = A e-‘wo‘ (2.76)
which, substituted into Eq. (2.75), gives the characteristic vibration
frequency
w =27ruo= & (2.77)
where vo is called the resonance frequency.

Let us assume that an electromagnetic wave with the electric field
acts as a driving force for the dipoles. With this force, the movement equation
would become
(2.79)
The choice of sign for the driving force is made by taking into account that
the electron feels the force in the direction opposite to E, because of its
negative charge -9. However, this equation is incomplete, because we know
that any accelerated charge radiates, and therefore the dipole charges will
emit an electromagnetic wave. Due to this radiation, the electrons experience
some kind of reaction force that opposes their movement. This reaction
force has been shown by Lorentz to be
F , = - . -2q2 d3y
(2.80)
3~x10’ dt3’
Thus, the complete movement equation is
- = - k y + - 2q2 --qEo
m -d- 2Y d3y
e-””. (2.81)
dt2 3cx lo7 dt3
The effect of the magnetic field of the driving wave is negligible compared
with that of the electric field and thus it is ignored. By proposing again a
solution of the type
Y = A e-‘”‘ (2.82)
SCATTERING 97
and by substituting the value for the amplitude, A is found to be equal to

-9Eo
A= (2.83)
m[wi-w2- iyw3/oi]'
where y is given by
(2.84)
Therefore, we see that the dipole oscillates with a frequency equal to that
of the driving wave. In order to study the amplitude and phase of the dipole
oscillation, we shall consider three different cases. We assume that y w 3 / w i
is very small compared with ( w 2 - w ' ) .
(1) If w << w o , as is the case for most materials at frequencies in the visible
range, A is negative, and hence y is opposite in sign to E.
Thus, the phase of the dipole vibration is the same as that of the driving
wave, as shown in Fig. 39. We say that the electrons vibrate in phase with
the driving wave if they always move following the force they feel, that is,
opposite to the driving electric field.
(2) If w >> w o , as in the case of free electrons or most materials at X-ray
frequencies, A is positive. Thus, the dipole vibration is 180" out-of-phase
with respect to that of the driving wave, as shown in Fig. 39.
DRIVING ELECTRIC FIELD
W<<Wo
w
0
a
FIG.39. Phase of the vibration of dipoles driven by an electromagnetic wave.

I DIPOLE
I
1
FIG.40. Radiation pattern for a dipole.
( 3 ) If w = w o , that is, near the resonance frequency, the value of A

becomes purely imaginary. Thus, the dipole vibration has a very large
amplitude and is 90" out-of-phase with respect to the driving wave.
When the dipole vibrates, it emits an electromagnetic wave with an
irradiance ( S ) that depends on the direction being considered. The symbol
( S ) represents the time average of the Poynting vector, in a direction as
follows:
q2w4 x 10-~
(S)= 8.rrcr2 - A: sin2 6, (2.85)
where 6 is the angle between the given direction and the dipole orientation.
The angular distribution of the dipole radiation is plotted in Fig. 40. The
complete radiation pattern would be obtained by revolving the pattern about
the dipole axis, thus obtaining a toroidal distribution.
The irradiance ( S ) is inversely proportional to the square of the distance
r from the dipole to the point under consideration. The symbol A represents
the maximum amplitude of oscillation of the dipole, as given in Eq. ( 2 . 3 8 ) ;
thus, if A,?,= A . A*,
q4E:x w 4 sin2 6
(S)= 8.rrm2cr2 . (2.86)
SCAITERING 99
In the vacuum, using the values of the refractive index n = 1 a n d po=

47r x lo-', the time average of the Poynting vector may be written as:
(2.87)
Therefore, Eq. (2.86) may be transformed into
The emitted wave has a phase difference kx with the electron vibration
only at distances from the dipole that are greater than the wavelength. At
very small distances from the dipole, the emitted wave has a phase shift of
180" with respect to the dipole vibration. The reason for this phase difference
at small distances is that at low frequencies, the charges in the dipole move
in a direction such that the electric field produced by the dipole tends to
cancel that of the driving wave.
2.7.3. Rayleigh Scattering

Let us think of a light wave entering a gas where the individual molecules
are randomly distributed with a n average distance between them that is
larger than the wavelength of the light.
If the size of the particles is much smaller than the wavelength of the
light, we have Rayleigh scattering.' In this case, we may think that each
molecule o r particle absorbs the luminous energy a n d then reemits it in all
directions. Due to the random distribution of the molecules, the light emitted
by each particle is assumed to be incoherent with the light emitted by the
others.
The energy scattered in a direction 0 by each particle is found by means
of Eq. (2.89). The total energy W scattered per unit time by the particle
may be found by integrating ( S ) over a sphere with radius r, in the center
of which is the particle, as shown in Fig. 41.
The total energy W emitted in all directions, per unit time, is given by:
w= I,: (S)(27rr sin e ) r do, (2.89)
and, using Eq. (2.88), we obtain by integrating

8 T q 4 w 4 10-'~(s~)
~
W=
3m2[ ( w t - w ' ) ' + I)$( 2, (2.90)
4
r sin t? -1
I
I
I
SCATTERED
LIGHT
-
INCIDENT
LIGHT
FIG.41. Total scattering by a molecule.
which may be written in the form:
where
(2.92)
The quantity u represents the units of area. Therefore, we could think

of the molecule as having an apparent area u that scatters only the light
that falls on it. For this reason, the quantity u is called the scattering cross
section.
Here we can actually distinguish the following three kinds of scattering:
(1) If w = w " , as in the case of most materials illuminated with visible
light, (T can be approximated by
8Tq4w4 10-1~
U = (2.93)
3m'w:
This Rayleigh scattering is directly proportional to the fourth power of the
frequency.
SCATTERING 101
( 1 ) If w >> w o , as in the case of most materials illuminated with X-rays

or free electrons illuminated with visible light, we can approximate u by
(2.94)
This is called Thompson scattering, and the magnitude of u is independent

of the frequency of the light in this case.
FIG.42. Two pictures taken with (a) normal black-and-white film, (b) infrared film and filter.
(3) If o = w o , a condition that occurs only if the frequency of the light

is near the resonance frequency of the material, we approximate by writing
W’-W;=~O~ (w-wO), thus:
2.rrq4w2x 1 0 - 1 ~
(2.95)
We see that if u = w o , the magnitude of the scattering becomes very large.

The resonance frequencies for most materials are in the ultraviolet region.
This kind of scattering is called resonance scattering.
The fact that the sky is blue is due to the o4 dependence of Rayleigh
scattering of visible light by the air molecules in the atmosphere. In other
words, the blue light is scattered much more than the red light. For the
same reason, the sun or the moon look red near the horizon, because the
blue light emitted by the sun or moon is scattered away and only the red
light reaches the observing eye.
Any distant object, for instance a mountain, is seen much more clearly
through a red filter than through a blue filter. This is the reason for using
infrared photography instead of the normal visible photography for distant
landscapes (see Fig. 42).
In closing this chapter, we must mention that the measurement of scatter-
ing are very useful tools that permit us to obtain interesting information
about the structure of the scattering particles. This tool is especially useful
in the fields of chemistry, meteorology, and astronomy, as is shown by Van
de Hulst.’’
Acknowledgment
The author wishes to thank Professor W. H. Steel for his very valuable comments.
References
1. M. Born and E. Wolf, Principles ofOptics, 5th ed., Pergamon Press, New York, 1975.
2. C. F. Meyer, The Diflraction of Light, X-Rays and Material Particles, 2nd ed., Edwards
Bros., Ann Arbor, Michigan, 1949.
3. R. W. Wood, Physical Optics, Macmillan, New York, 1936.
4. A. Marechal and M. Francon, Diflraction, Structure des Images, editions of the Revue
d’Oprique The‘orique et Instrumentale, Paris, 1960.
5. L. Rayleigh, Scientific Papers, Cambridge University Press, London, 1899.
6. L. Rayleigh, Phil. Mag. 41, 107 (1871); 41, 274 (1871); 41, 447 (1871).
7. F. M. Grimaldi, Physico-Mathesis de Lumine, Bologna, 1655.
8. 1. Newton, Opticks, London, 1704 (Reprint: Dover, New York, 1952).
9. C. Huygens, Traite‘ de la LumiPre, Leyden, 1960.
10. T. Young, Phil. Trans. Roy. Sor. 92, 12 and 387 (1802).
1 1. T. Young, A Course of Lectures on Natural Philosophy and Mechanical Arts, London, 1807.
REFERENCES 103
12. T. Young, Miscellaneous Works of the Late Thomas Young, Vol. 1 (George Peacock, ed.),
John Murray, London, 1855.
13. A. Fresnel, Oeuures ComplPtes, Vol. 1, pp. 89 and 129; vol. 2, pp. 261 and 479 (H. de
Senarmont, E. Verdet, and L. Fresnel, eds.), Imprimerie Imptriale, Paris, 1866.
14. G. Kirchhoff, Sitz. Ber. Kgl. Preuss. Akad. Wiss., 22 June 1882.
15. G. Kirchhoff, Sizt. Akad. Wiss. Berlin, 641 (1882). Reprinted in: Wied. Ann. Phys. 18, 663
(1883).
16. G. Kirchoff, Cesammelte Abhandlungen Nachtrag, Barth, Leipzig, 1891.
17. G. Kirchoff, Lecfures on Mathematical Optics, Part. 2 (Boltzmann, ed., Teubner, Leipzig,
1891). (In German.)
18. G. G. Stokes, Trans. Camb. Phil. Sac. 9, 1 (1849).
19. S. Silver, “Microwave Aperture Antenna and Diffraction Theory,” J. Opt. Sac. Am. 52,
131 (1962).
20. A. Sommerfeld, Optics, p. 325, Academic Press, New York, 1954.
21. J. D. Jackson, Classical Electrodvnamics, 2nd ed., p. 391, Wiley, New York, 1962.
22. F. Kottler, Ann. d. Physik 71, 457 (1923).
23. F. Kottler, Ann. d. Physik 72, 457 (1923).
24. B. B. Baker and E. T. Copson, The Mathematical Theory of Huygens Principle, 2nd ed.,
Clarendon Press, Oxford, 1950.
25. J. C. Maxwell, A Treatize on Electricity and Magnetism, Oxford, 1873.
26. H. Poincart, Mathematical Theory o f l i g h t , Part 2, Georges Carre, Paris, 1892. (In French.)
27. A. Sommerfeld, Nachr. Akad. Wiss, Gotringen, Math.-Phys., K 1 1, 338 (1894).
28. C. J . Bouwkamp, Rep. Progr. Phys. (London) 17, 35 (1954).
29. F. Kottler, “Diffraction at a Black Screen,” in Progress in Optics, Vol. 4, Chap. 7 and Vol.
6, Chap. 8 (E. Wolf, ed.), North-Holland Publ., Amsterdam, 1971.
30. F. Kottler, Ann. Phys. 70, 405 (1923a).
31. E. Maey, Wied. Ann. Phys. 49, 96 (1893).
32. A. Sommerfeld, Math. Ann. 47, 317 (1896).
33. A. Rubinowicz, Ann. Phys. 453, 257 (1917).
34. A. Rubinowicz, Ann. Phys. 473, 339 (1924).
35. K. Miyarnoto, A . General Theory of the Boundary Diflraction Wave and Related Investiga-
tions, Ph.D. Thesis, Inst. of Optics, The Univ. of Rochester, 1961.
36. K. Miyamoto and E. Wolf, “Generalization of the Maggy-Rubinowicz Theory of the
Boundary Diffraction Wave, Part 1,” J. Opt. Sac. Am. 52, 615 (1962a).
37. K. Miyamoto and E. Wolf, “Generalization of the Maggy-Rubinowicz Theory of the
Boundary Diffraction Wave, Part 11,” J. Opt. Soc. Am. 52, 626 (1962b).
38. A. Rubinowicz, Progress in Opfics, Vol. 4 Chap. 5 (E. Wolf, ed.), North-Holland Publ.
Amsterdam, 1965.
39. A. Rubinowicz, “Simple Derivation of the Miyamoto-Wolf Formula for the Vector Poten-
tial Associated with a Solution of the Helmholtz Equation,” J. Opt. Sac. Am. 52, 717 (1962).
40. W. Duane, frog. Nut. Acad. Sci. 9, 158 (1923).
41. E. Jahnke and F. Emde, Tables ofFunctions with Formulae and Curves, Leipzig and Berlin,
Teubner. (Reprint: Dover, New York, 1945, 4th ed., p. 35.)
42. T. Pearcey, Tables of Fresnel Integrals to Six Decimal Places, Cambridge University Press,
Cambridge, 1956.
43. R. Bergsten and S. Huberty, “White-light Fresnel Diffraction by a Circular Aperture,”
J. Opt. Sac. Am. 67, 643 (1977).
44. A. C. S. Van Heel, Advanced Optical Techniques, Chap. 13, Wiley, New York, 1967.
45. F. M. Schwerd, Die Beugungserscheinungen aus den Fundamentalgesetzen der Undulations-
theorie Analyfisch entwickelt und in Bildern dargesrellt, Schwan und Goetzsche Hofbuch-
handlung, Mannheim, 1835.
104 DIFFRACTION AND SCATTERING
46. R. B. Hoover and F. S. Harris, Jr., “Die Beugungserscheinungen: a tribute to F. M.

Schwerd’s Monumental Work on Fraunhofer Diffraction,” Appl. Opf. 8, 2161 (1969).
47. F. S. Harris, Jr., “Light Diffraction Patterns,” Appl. Opt. 3, 909 (1964).
48. R. Bergsten and S. H. Popelka, “White Light Fraunhofer Diffraction,” J. Opt. SOC.A m .
69, 584 (1979).
49. C. B. Airy, Trans. Cambridge Phil. SOC.5, 283 (1835).
50. C. J. Sletten and P. Blacksmith, Jr., “The Paraboloid Mirror,” Appl. Opf. 4, 1239 (1965).
51. H. C. Minett and B. MacA. Thomas, “Fields in the Image Space of Symmetrical Focusing
Reflectors,” Proc. IEE 115, 56 (1968).
52. A. Babinet, Compt. Rend. Acad. Sci. Paris 4, 638 (1837).
53. D. Rittenhouse, Trans. Amer. Phil. SOC.2, 201 (1786).
54. J. Fraunhofer, Ann. d. Physik 74, 337 (1823). Reprinted in his Collected Works 51, 117
(Munich, 1888).
5 5 . M. C. Hutley, Diflracfion Gratings, Academic Press, New York, 1982.
56. D. Richardson, “Diffraction Gratings,” in Applied Opfics and Opfical Engineering, Vol. 5 ,
Chap. 2 (R. Kingslake, ed.), Academic Press, New York, 1969.
57. W. T. Welford, “Aberration Theory of Gratings and Grating Mountings,” in Progress in
Optics, Vol. 4, Chap. 6 (E. Wolf, ed.), Academic Press, Amsterdam, 1965.
59. H. A. Rowland, Phil. M a g . V 35, 397 (1893).
61. J. A. Anderson, “Periodic Errors in Ruling Machines,” J. Opt. SOC. A m . 6, 434 (1922).
62. C. Runge, “False Spectra From Diffraction Gratings. Part 11: Theory of Lyman Ghosts,”
J. Opt. SOC. Am. 6, 429 (1922).
63. R. W. Wood, N a t u r e 140, 723 (1937).
64. R. W. Wood, “Improved Diffraction Gratings and Replicas,” J. Opt. SOC.Am. 34,509 (1944).
65. H. Babcock, “Bright Diffraction Gratings,” J. Opt. SOC.A m . 34, 1 (1944).
66. R. W. Wood, Phil. M a g . 20, 770 (1910).
67. G. R. Harrison, “The Production of Diffraction Gratings. I: Development of the Ruling
Art,” J. Opt. SOC.Am. 39, 413 (1949).
68. G. R. Harrison, “The Production of Diffraction Gratings. 11: Design of Echelle Gratings
and Spectrographs,” J. Opt. SOC.Amer. 39, 522 (1949).
69. R. A. Sawyer, Experimental Spectroscopy, Dover, New York, 1944.
70. J. D., Strong, “The Johns Hopkins University and Diffraction Gratings,” J. Opt. SOC.A m .
50, 1148 (1960).
71. G. W. Stroke, Progress in Optics (E. Wolf, ed.), Vol. 2 North-Holland Publ., Amsterdam,
1963.
72. A. Marechal, Rev. Opf. 26, 257 (1947).
73. E. L. O’Neill, The Modulation Function in Optics. Tech. Note 110, B.U.P.R.L., 1954.
74. E. Abbe, Gesammelte Abhandlungen, Fischer, Jena, 1904.
75. W. H. Steel, “Effects of Small Aberrations on the Image of Partially Coherent Objects,”
J. Opt. SOC.Am. 47, 405 (1957).
76. F. Zernike, “Beugungstheorie des Schneidenverfahrens und Seiner Verbesserten Form,
der Phasenkontrast Methode,” Pbysica 1, 44 (1934).
77. F. Zernike, “Phase Contrast, a New Method for the Microscopic Observation of Trans-
parent Optics,” Physics 9, 686 (1942). Reprinted in Achieuements in Optics by A. Bowers,
Elsevier, New York, 1950.
78. H. C. Van de Hulst, Light Scattering by Small Parficles, Wiley, New York, 1957.
79. M. Kerker, The Scattering of Light and Other Electromagnetic Radiafion, Academic Press,
New York, 1969.
80. G. Mie, Ann. Physik 25, 377 (1908).
REFERENCES 105
81. G. B. Parrent, Jr. and B. J . Thompson, “On the Fraunhofer (Far Field) Diffraction Patterns
of Opaque and Transparent Objects with Coherent Background,” Oprica Acra 11, 183
(1964).
82. B. A. Silverman, B. J. Thompson, and J. H . Ward, J. Appl. Meteor. 3, 792 (1964).
83. V. Khare and H . M. Nussenzveig, Phys. Rev. Lert. 33, 976 (1974).
84. M. Minnaert, The Nature of Light and Color in rhe Open Air, Dover, New York, 1954.
3. OPTICAL POLARIZATION
Frederic R. Stauffer
Sacramento Peak Observatory
Sunspot, New Mexico 88349
3.1. Introduction
Polarization effects are a result of the wave interaction of light with the
physical world and have long been demonstrated with a double refractive
index crystal known as calcite. Light waves in two orthogonal planes are
laterally separated after passing through calcite, resulting in two polarized
images. Christian Huygens in the late seventeenth century experimented
with and successfully described calcite’s double images by using wave-front
constructions. But, at this time, light was thought to be a longitudinal wave
like sound, and polarized light remained undiscovered. In 1808, E. L. Malus
discovered that reflected light viewed with a properly oriented calcite crystal
passed a single image and extinguished the other image. Malus concluded
that the reflected light was polarized.
In 1815 Brewster showed the significance of the incidence angle on the
polarization of the reflected beam. Maximum polarization (lOOo/~ for dielec-
trics) occurs at an incident angle, now termed Brewster’s angle, depending
on the material’s determined optic parameters. In 1816, Fresnel and Arago
demonstrated with Brewster angle polarizers that two orthogonal polarized
beams did not interfere, which led to the important conclusion that light
is a transverse wave propagation. Fresnel proceeded to simply state the
polarization effects produced by light’s reflection and refraction at boun-
daries between materials further supporting light’s wave nature. Later in
the century, Maxwell’s and Faraday’s efforts produced Maxwell’s equations
describing the fundamental interrelationship between the basic electromag-
netic vectors. The necessary theories to understand polarization were
complete.’.’ The theories are based on observations. To understand
polarization, it is best to experiment with its different aspects. Computer
optical design programs have also made the polarization equations easier
to use.
107
METHODS OF EXPERIMENTAL PHYSICS Copyright 63 1988 hy Academic Press. Inc.
Vol. 26 All rights of reproduclion in any form reserved
I S B N 0-12-475071-8
108 OPTICAL POLARIZATION
3.2. Electromagnetic Description of Light
3.2.1. Field Equations and Wave Equations

Maxwell’s equations are vector and scalar products written in terms of
the basic fields, the electric field E and the magnetic induction B, and in
terms of the derived fields (due to the fields’ effects on matter), the electric
displacement D and the magnetic field H, taking into account the average
contribution of the charge distribution p and the electric current density J.
To determine the interaction of light with matter, Maxwell’s equations need
to be supplemented with equations describing the effect of an electromag-
netic field on matter known as the material equations. Generally, these
equations can be written for complicated three-dimensional anisotropic
and/or relativistic situations. But, assuming static and isotropic conditions
makes the problem simpler and realistic, and the material equations reduce
to constant relations between the electromagnetic vectors. The quantities
needed to describe the effect on matter are the dielectric constant or
permittivity E, permeability p, and conductivity u.
VxE+B=O
VXH-D=J Maxwell’s equations (3.1)
V-D=p
V.B=O
D = EE
B=pH material equations
J=oE
where E , p and u may vary with wavelength.

Using the material equations and vector identities to separate the cross
product equations and assuming the medium is homogeneous results in a
second-order differential wave equation describing the space-time distribu-
tion for an electromagnetic field.
V2E= PEE+ puE. (3.2)
Plane and spherical wave solutions exist for this differential equation. Plane
waves in Cartesian coordinates are useful in describing electromagnetic
effects and are used as the example. For dielectrics, the conductivity is zero,
ELECTROMAGNETIC DESCRIPTION O F LIGHT 109
and the wave equation simplifies to
1
=- for a vacuum
C
6= -V1 for a dielectric
A =wavelength
s = propagation direction unit vector

k = ks
S = constant phase term
V * E=O
V * B=O
B=JCLESXE
E - volts/meter
D - coulombs/(meter)2
B - tesla
H - amps/ meter
The plus or minus exponential of the wave solution means that waves
can propagate in two directions. It is important to define and use a sign
convention for consistent results. The sign convention in the literature varies,
but a standard for polarization work has been defined. The convention used
is light traveling in the negative direction or a positive time d e p e n d e n ~ e . ~
The propagation direction for the electromagnetic field is s, and r is the
vector distance from an arbitrary origin to a plane in space of constant
phase. The phase term 6 offsets the wave front in space for a fixed time.
Applying Maxwell’s equations to the solutions for E and B shows their
divergence equal to zero and thus oscillating in a plane perpendicular to
the propagation direction.
The constant electric field plane intersecting at r is defined by k * r, and

the planes with equal electric fields are spaced apart by the wavelength A.
A snapshot of a plane wave or a fixed position in time shows the electric
field constant across planes perpendicular to the propagation direction; and
the electric field would be a sinusoidal modulation along the propagation
direction. Another view is to fix a point in space and to allow the time to
vary. Then, the electric field would be constant across the plane perpen-
dicular to the propagation direction, and it would be time-dependent and
modulated by w t as the plane waves go by.
A coordinate system about the propagation direction and a plane projec-
tion can be used to describe the resultant electric field. A linear resultant
oscillation is termed linear polarization. Linear polarizations with varying
azimuth can be constructed by combinations of linear orthogonal com-
ponents shifted by zero or multiples of half-wavelengths corresponding to
a phase shift of zero or integral multiples of r. For example, equal amounts
of horizontal, E H , and vertical, Ev, linear polarizations with zero phase
difference give a resultant linear polarization with a 45" azimuth, A. The
linear polarizations add together as a planar electromagnetic field distribu-
tion along the propagation direction propagating at light's velocity. The
resultant plane projection of the light beam results in a linear vector.
+ E~
E = EH e - J ( k r-wr) e-J(k r-wr-8)
E =(E H E H + e") e "k'r-"'
(3.4)
S=O
A = arc tan( Ev/ E H )
where E~ and E~ are the dialectric constants in the horizontal and vertical
directions. Now, if the phase difference is changed to an odd multiple of
7r/2, equal intensity orthogonal linear polarizations add together, resulting
in a helical electromagnetic field distribution along the propagation direction
propagating at light's velocity. The resultant plane projection of the light
beam results in a circular spinning vector produced by the helical distribu-
tion moving through space.
E = ( E H E+~E , , E ~e J * )e-J(k r - w r )
S=~ / 2 , EH = Ev for circular polarization,
E, E E~V eJTI2)
=E ~ ( f e-J(k r - w r ) (3.5)
A right-handed helix, E-, produces a clockwise rotating plane projection
looking towards its source, and vice versa for a left-handed helix, E,. These
spinning projections define right- (R)and left- (L) handed circular polariz-
ELECTROMAGNETIC DESCRIPTION OF LIGHT 111
ations, re~pectively.~ Physically, Ev is advanced or retarded with respect to

EHby an odd multiple of A/4,for example by an anisotropic medium such
as calcite.
Right and left circular polarizations describe another orthogonal coordin-
ate system perpendicular to the propagation and give another perspective
to view polarization.
Normalized circular polarization vectors are,
&, = (2)l/z(EHfjEV) e-~lh.r-uf)
E(r, 1 ) = (E+&+eJd +E-E-) epJ'h'r-wf) (3.6)
The oppositely spinning E+ and E- vectors add around their paths, resulting
in an ellipse whose major axis, a, is the sum, and whose minor axis, b, is
the difference of the intensities. The major axis is tilted at an angle A, which
is half of the phase difference between the vectors. Figure 1 shows various
combinations of L and R and phase.
a=E++E
b=E+-E
ellipticity = b / a
handedness: E- > E, right handed
E- = E, indeterminate
E- < E, left handed
azimuth = A = 6/2
L/E, E, E_ Result.
= 1.0 00- 00./00.1
6 = 0.0' 6 = 90.0' 6 = 180.0'

FIG. 1. Plot of L and R polarizations with different phases and amplitudes and the resultant
vector. The polarized light sources are in the page, and the resultant vector is produced by
adding the sources.
3.2.2.Poincare Spherical Description

It takes three independent values to determine the polarization state of
the monochromatic wave. For example, the amount of left and right circular
polarizations and their phase relationship describes the ellipticity, azimuth,
and the handedness (the resultant vector’s rotation direction). A three-
dimensional spherical surface for a given radius or beam intensity can be
used to map the three values needed to describe polarization. Orthogonal
polarizations, such as L and R or H and V, are at opposite ends of the
sphere’s diameters. The poles of a sphere are taken to represent pure L or
R polarization with increasing amounts of the other component added until
equal amounts of L and R polarization meet at the equator, producing
Linear polarization with a varying azimuth, as in Fig. 2. This results in
ellipticities ranging from - 1 to 0 to +1, or from pure R at one pole, to
linear polarization at the equator, to pure L at the other pole. The sphere’s
latitude, LAT, is related to the ellipticity and the handedness. The longitude,
LONG, is equal to twice the azimuth of the elliptical polarization or the
phase difference between the L and R components.
tan (y)i,
a =
LONG = 2A.
-45.0 +45.0’
R
FIG. 2. Polarization ellipticity and handedness at positions as mapped on the Poincare
sphere. The handedness is L for the upper and R for the lower hemisphere. The sense of
rotation for handedness is as in Fig. 1, with the polarized light source in the page.
WAVE PROPAGATION IN ISOTROPIC MEDIA 113
An elliptical polarization is a point on the sphere, P(LAT, LONG).

Cartesian coordinates can describe the sphere’s latitude and longitude, too.
Taking the axes z through R, x through H, and y through +45”, linear
polarization gives three equations to mark a polarization, P ( x , y, z).
x = cos( LAT) sin( LAT)
y = cos( LAT) sin(LONG) (3.7)
z = sin( LONG)
PoincarC first suggested the description, and the sphere is referred to as the
Poincare‘ sphere.’
3.3. Wave Propagation in Isotropic Media
3.3.1. Fresnel Equations

When light is incident on a dielectric medium with different constants
E’(o,r) and p ’ ( w , r), Maxwell’s equations and their continuity at the boun-
dary need to be considered for waves parallel to the plane defined by the
boundary’s normal and the incident beam’s propagation direction (also
known as P, T , or TM-transverse magnetic waves), and for waves perpen-
dicular to this plane (S, a,or TE waves). Recall that the tangential com-
ponents of the electric vector are continuous across the boundary, the normal
component of the electric displacement changes across the boundary due
to surface charges, the normal component of the magnetic induction is
FIG.3. Sign convention for the Fresnel equation interpretation and derivation. r, is perpen-
dicular and r,, is parallel to the page. 0 is a vector out of, and x is a vector into, the page.
continuous, and the tangential component of the magnetic vector changes

across the boundary due to a surface current density. From the boundary
conditions and measuring angles about the boundary’s normal, it is found
that for a beam incident at an angle equal to di,( 1 ) a reflected beam leaves
at an angle equal in magnitude to the incident angle (4, = -bi),(2) a
refracted beam proceeds at an angle 4, obeying Snell’s law, and (3) the
perpendicular and parallel components are described by Fresnel’s
equations6
( p ~ / p ~ ~ ~ J ” 4i
’ s=i n( ~ sin 4,
L ’ E ‘ / ~ ~ E ~ ) ’ ’ =
~ n sin +i = n’sin 4t, Snell’s law
n = refractive index = c/ o
n cos 4 i - Py n’cos cpl

EsreAected - P
- - rq= 9
P
bsincident
n cos 4i+ y n’ cos
P
Estransmitted 2n cos 4i
= t, = , Fresnel equations (3.8)
P
sincident
n cos c$~ + y n‘ cos C#I~
P
Epretlected - P
6n’ cos 4I - n cos 4t
- rp = ,
pincident
6
P
n’cos 4 , + n cos 4t
Eptransmitted - 2n cos 4,
- tp=
E p incident
6
P
n‘cos + , + n cos
lrs12+lt,12= 1, 1,
Irp12+lfp12=
R, = rsr?, R , = rprp*,
where R, and R , are the reflectance and are related to the energy flow by
Poynting’s equation. In most cases, p / p ’ - 1 , simplifying the equations.
The Fresnel equations are derived according to Fig. 3 and predict the
phase and amplitude for S and P polarizations. The incidence plane is the
page, and it is also the rp plane. The vector perpendicular to the page is r,,
and it is an x or 0 into or out of the page. The equations’ phase shifts are
interpreted in Fig. 3. For example, at normal incidence, rs and rp are
indistinguishable, and the minus sign difference is because the rp vector is
antiparallel to r, in the figure.
WAVE PROPAGATION I N ISOTROPIC M E D I A 115
The vectors r, and rp have some interesting and useful mathematical

relations that simplify the equations. Effective indices are an example. The
equation form for S and P polarizations for the lim 4i+ 0 reduces to the
normal incidence coefficients even though there is no physical difference
between S and P.
n-n’ n’- n
r, = - rp = -
n + n” n + n” (3.9)
For non-normal incidence, the reflection coefficient equations can be rewrit-

ten with effective indices p and p ’ to reproduce the forms of the normal
incidence equations.’
p s= n cos c$~, p:= n‘cos 41,
n‘
CL;
=-
cos cpl ’
CL,-cL:
r, = - rp = CL; - CLP (3.10)
C L S + d ’ CLP+CLb’
The reflection coefficients can also be written in terms of Snell’s law and
incidence angle only.
n’
v=-
n
cos 4, - ( v’ - sin’ 4i)”’
r, =
cos 4, + ( Y’ -sin2 4 ~ ” ”
(3.11)
v’ cos bi- ( v 2 - sin’ 4 ~ ” ’
rp =
+
v’ cos 4i ( v’ - sin’ 4i)’/2*
The positive sign of the square root term was chosen because the replaced
cosine term is positive real valued. When the square root goes to zero and
then becomes imaginary, the negative root is chosen because it predicts an
exponential decrease in the refracted beam’s amplitude, which is the real
situation. Eliminating v from equation (3.11) for rp gives an equation for
rp in terms of r, and the incident angle only.’
(3.12)
There are two cases for v, IvI > 1 and Ivl< 1 . For IvI > 1, for example a
beam incident from air to glass, the value of the square root term is always
positive, real, and less than cos so that r, is negative. The negative sign
is interpreted as a r phase shift for 8 , . The rp goes from a positive value,
-
- I
-
4
- I
.om
- - 10 n ___________________
-
4
-
L -
-
..-.-.-._._, y.- , , ,
6, = 0, for 4i = 0, through zero for Brewster’s angle, 4B,to a negative value

8, = .rr, for 4i > c#J~.Figure 4 is a plot of r, and rp versus the incidence angle
for Ivl< 1.
Brewster angle conditions are,
v 2 cos 4 B = ( v 2 -sinZ 4B)’”,
(3.13)
rs = cos 2 4 B .
At Brewster’s angle, the reflected polarization parallel to the incident plane
is zero and produces a highly polarized reflected light beam perpendicular
to the incident plane. For incident angles less than Brewster’s angle, the S
and P reflections have a n phase difference, and the reflected polarization
is orthogonal to the incident polarization. A light beam can be optically
isolated with a circular polarizer and a reflection, assuming there are no
other polarization effects.
For IvI < 1, a beam traveling from glass to air, for example, r, is always
positive and reaches unity at a critical angle, &, which gives total reflection
and an imaginary phase related component due to the square root term.
sin’ 4, = v2. (3.14)
An S-beam incident from glass to air does not have a phase shift upon
reflection until it is incident at c#+ > &. For +i 2 &, Ir,l= 1. For r p , its value
is less than zero up to Brewster’s angle, where rp equals zero and then is
positive real valued until the critical angle where the imaginary phase related
term influences r p , and lrpl = 1. Figure 5 is a plot of r, and rp versus the
incidence angle for I vI > 1. We can write r p and r, as complex expressions.
r P = p P ej S p , r s = p s e jSs . (3.15)
WAVE PROPAGATION IN ISOTROPIC M E D I A 117
10 n
4
4
-
-
L
4
0.0
0.u v@ ‘pe 9o.a
Incidence angle, 9
FIG. 5. Absolute magnitude for the reflection coefficients r,,p and phase shift S,,,, from a
more to a less dense medium.
In ellipsometry, the following notation is also used:
(3.16)
For total internal reflection, expressions for the phase changes a,, 6, are,
(3.17)
Figure 6 is a A ( 8 p - 6,) plot, and the extremum in the curve is found by

maximizing A with respect to the incident angle.
(3.18)
Also, [A( 6,- &)/A&] slowly varies about the extremum, and this fact can
be used to introduce a phase shift between P and S which is fairly insensitive
to the incident angle. A Fresnel rhomb is a total internal reflection device
to produce a phase shift between P and S. The incident beam totally
internally reflects for both r, and rpoff of the two sides inclined so that the
incidence angle gives a ?r/4 phase difference between P and S on each
reflection. This gives a total retardation between P and S of ~ / 2 A. linearly
polarized beam with equal P and S components incident on the Fresnel
0
0.d 9O.U
Incidence angle, (pl
AV>I
- - - - - - -. &<I
FIG. 6. A for u < 1 and u>1 as a function of incidence angle
rhomb will become a L or R circularly polarized beam depending on the

incident beam's orientation.
The A phase shift between the reflected S and P components translates
to a A rotation of the PoincarC sphere about the axis through the S and P
linear polarizations. For example, linear polarization incident with equal
components in the S and P planes will reflect with a different ellipticity.
3.3.2. Complex Plane Representation

Another way to view Fresnel's coefficients is to map relations in the
complex plane. For example, the material constants determine r,, which,
with the incidence angle, solely determine the behavior of r p . If r, is
described as the upper half-plane area (positive time dependence in the
wave equation) bounded by the unit circle about the origin, then the values
of r, map into r, an area bounded by a unit circle in the complex plane.
Constant phase lines radiate from the origin, and constant amplitude lines
are circles about the origin bounded by the unit circle.'" Figure 7 is a plot
of r, and several plots of rp for different incidence angles.
A noteworthy feature of Fig. 7 and the equation relating r p to r, is that
bi= 45". At this angle, rp = r f , 6, = 26,, and a known polarization is pro-
duced, which can calibrate an instrument in terms of phase and amplitude
changes through the entire system for both P and S polarization. The P
polarization reflected at 45" is unaffected by small changes in the media
due to temperature, stress, and electromagnetic fields. This result works for
a film-free substrate, since P is written only in terms of S and the incidence
angle.
Another complex transformation is from a normal incidence reflection
coefficient, z, to a coefficient, w,at other incidence angles."-'2 The coefficient
w is an analytic function of z and is dependent on the incident angle. The
WAVE PROPAGATION I N ISOTROPIC M E D I A 119
l5
.
3
rP (s=45.0' rP , q=75.0'
FIG. 7. Complex plane representation for r, in terms of constant phase (radial lines) and
amplitude (semi-circles), and r p as a function of r, and incidence angle.
normal incidence equation form is,

1-v
r =-r =-- - z,
l+v
(3.19)
1-z
v=-
l+z'
The transformation from z to w, and the inverse transformation from w to
z for S polarized light are,
Z=
( a +2p) * [(a' - 4 ) + 4 ( a +2)p]'12 9
2(1 - P )
(1 + z ) - ( z 2 - a z + 1)'12 (3.20)
W = '
( 1 + z)(z' - az + 1
a=4tan2cpi+2,
p = [ 'I+w
-"]'.
This transformation is particularly useful. If an S reflectance measurement

is made at normal incidence R,( &) and at a known incidence angle R,( cpi),
then the complex normal reflection coefficient, z, can be determined. In Fig.
8(d), a circle with Izl= R , ( I $ ~ ) " *is plotted on the complex plane with the
image for IwI = R S ( ~ J ' l Where
2. the two curves meet determines v and z.
...'..,
[O
.'......'.
(4
FIG. 8. (a) Constant amplitude circles representing either IzI or IwI, (b) IzI mapped into w
for 4, = 30". (c) IwI mapped into z for +a= 30". and (d) IzI normal incidence measurement
overplotted with a I wI measurement for 4, = 30" mapped into z.
3.4. Wave Propagation for Metals
3.4.1. Polarization Effects

Metals include the conductivity coefficient which affects the Fresnel
equations. The damping term due to conductivity in the wave equation is
observed as Joule heating. The wave equation solution is written as a
monochromatic wave as for dielectrics with E satisfying the wave equation
with conduction included.
V2E+ k2E = 0, k 2 = 0 2 p (E ,):- (3.21)
whe k is the complex wave number. A solution for a conducting medium

wave equation is found by substituting the complex dielectric constant i,
which expresses the attenuation due to the conductivity term, into the
nonconductive medium solution and by making the appropriate changes
in phase velocity, wave number, and refractive index. This also carries over
into the Fresnel equations and Snell's law.
WAVE PROPAGATION FOR METALS 121
(3.22)
where K is the absorption index, n and K are real, and n' is the complex
refractive index. The absorption coefficient k = nK is most often used. It is
important to remember the distinction between k and K .
Snell's law provides a complex-value refracted sine term when light is
incident from a dielectric to a metal, i.e., n , sin 4i= nl2 sin 4,. This means
that for a refracted wave in a conductor, the complex wave number produces
an imaginary term which represents constant beam amplitude surfaces
parallel to the boundary and a real term depending on the incidence angle
which relates constant phase surfaces. Since the constant amplitude and
constant phase surfaces do not necessarily agree, the electromagnetic wave
is said to be inhomogeneous (see Ref. 2, p. 611).
In Fig. 9, Eq. (3.11) is used to plot the reflection coefficients for aluminum
and silver as a function of the incidence angle. The effect of the attenuation
term smoothes the extremum termed the second Brewster angle. This also
means that the rp term is not extinguished as for a dielectric and increases
*T - - - - - - \
\
4
Lo \
0 . 6 " ' ' " " ' >

acr 9O.U
Incidence angle, 9
FIG.9. Absolute reflection amplitude coefficients and the phase shift for S and P polariz-
ations as a function of incidence angle for aluminum (top) and silver (bottom).
0
0.0‘ 90.0-
Incidence angle,
AAI
_------- AA9
FIG. 10. A as a function of incidence angle for aluminum and silver.
with K . Silver has a large absorption coefficient and consequently has only
a small variation in reflectance as a function of the incidence angle. However,
aluminum or rhodium have a lower absorption coefficient and produce
relatively large polarization effects compared to silver.
The difference between S and P phase shifts is plotted in Fig. 10. An
incidence angle with a 7r/2 phase shift in the plot reflects a linearly polarized
beam as an elliptically polarized beam whose axes are along rp and r,. If
rp = r,, circular polarization is produced. This is called the principal angle,
and it is a way to produce a known elliptical polarization.
3.4.2. Reflection and Transmission Polarizers

Reflection or transmission polarizers can be made by using the fact that
-
R , is close to zero ( Tp 1) at Brewster’s angle.’ For transmission polarizers,
a pile of plates is needed to eliminate the S component. The angular
acceptance has to be limited, because R,/ R , increases away from Brewster’s
angle.13 Their advantages, especially for reflection polarizers, are that they
cover wavelength ranges not covered by crystal polarizers, that they are
generally achromatic, have wide apertures, have uniform polarization across
the aperture, have high extinction ratios, have relatively high transmission,
and are easily fabricated. Their disadvantages are limited angular acceptance
(depending on tolerances and polarizer throughput), physical size of the
polarizer, possible beam deviation, and multiple reflections.
Reflection polarizers are used primarily in the vacuum ultraviolet (VUV),
ultraviolet (UV), and infrared (IR) spectral regions. 14-18 A good polarizer
needs to have a high R , / R , value which is dictated by the surface’s optic
constants. Metal coatings are most often used and are sometimes overcoated
with a protective layer whose thin-film interference effects need to be
considered and are discussed later. The metal coatings are to increase the
WAVE PROPAGATION FOR METALS 123
r, reflection. According to Fresnel's equations, r, increases with the refraction

index of the reflecting medium, and the Brewster angle or minimum rp angle
shifts towards increasing incidence angles, as seen in Fig. 9. With increasing
absorption, rp increases and R,/ R , decreases. This suggests a material with
a large n and small K in order to get the largest values of R , / R , . It is also
important to consider how the material properties change with wavelength
for VUV and UV applications. Magnesium fluoride, for example, becomes
absorptive around 1140 A, and the choice of material for the spectral region
is important. The literature gives a fairly complete description and the
( R , / R p ) m a xR, , transmittance, and angle of incidence for ( R , / R p ) m a xfor
different materials from 300b; to 2000A which can possibly be used as
mirror coatings: gold, iridium, osmium, rhodium, and rubidium." Gold is
the best coating for a polarizer to cover from 300 b; to 2000 A, because its
( R s / R p ) m astays
x fairly uniform and is generally the largest. If a selected
wavelength range is used, the other materials, such as rhodium from 900 A
to 1300 A will give better performance (higher R,/ R,) than gold. For the
IR, the choice of metal coatings is not as difficult as for the VUV, and
overcoated aluminum mirrors have been used as polarizers.
The angular acceptance and R J R , values can be increased by multiple
reflections from more than one mirror (at the same time decreasing through-
put). Single Brewster-angle polarizer plates deviate the beam which might
be unwanted, and three- or four-mirror arrangements can be used to bring
~ * ' ~increases R,/ R , versus
the beam back in the original d i r e c t i ~ n . ' ~ , ' " . 'This
the incidence angle by multiple reflections at Brewster's angle, because the
values ( R J R , ) " and (R,)" ( m is the number of reflections at Brewster's
angle) increase and decrease accordingly. The acceptance angle and polariz-
ance can be increased, while sacrificing throughput by using multiple
reflections at Brewster's angle.
A three-mirror polarizer is drawn in Fig. 11. The middle mirror is set
with p = Brewster's angle; LY is the entrance incidence angle and is greater
than p, resulting in a higher R , and R , than for the middle mirror, with
R , being significantly diminished before the middle mirror reflection. Two
disadvantages of the three-mirror arrangement are (1) the image is rotated,
and (2) the polarization varies across the acceptance angle in a complex
I I
I
FIG. 1 1 . S polarizer made of three mirrors.

FIG. 12. S polarizer made of four mirrors.
manner, because the incidence angle is not the same for the three mirrors.
The four-mirror arrangement in Fig. 12 has all the mirrors at Brewster's
angle, thus a higher polarizance, low output of R,, the image is not rotated,
and polarization varies across the acceptance angle as for a single mirror.
Transmission Brewster-angle polarizers or pile-of-plane parallel plates
transmit the P component according to the amount of absorption and
therefore, theoretically, 100% for no absorption. Realistically, though, there
exist scattering, surface films, and crystal imperfections. Also, the beam is
deviated perpendicular to its propagation direction, and multiple reflections
make T, difficult to extinguish. These two problems can be corrected by
proper arrangement of the elements and by making them wedge-shaped to
reduce multiple reflection effects. These polarizers are useful in the IR with
materials such as selenium, silver chloride, as shown for Saran Wrap," and
in the UV with calcium fluoride, lithium fluoride, and magnesium fluoride.
There are several advantages of transmission polarizers: an undeviated
beam, high extinction, and high energy use. They are shorter than reflection
type polarizers and are an easy way to produce polarized light.2'
Reflection circular polarizers have been demonstrated, particularly for
the UV. As was shown above for metals, there is a principal angle where
the phase shift between r, and r p is 7712, so that an incident linearly polarized
beam adjusted to give r, = rp produces a circularly polarized beam. In a
paper by McIlrath et al., the authors show results for an aluminum (Al)
mirror exposed to air (overcoated by an aluminum oxide (AI,O,) layer of
about 25 A ) which produces /rsl = Irpl over all incidence angles and thus
can be used as a phase shifter between r, and rp. At an incidence angle
equal to 52" at Lyman-a, the phase difference is ~ / 2 Their . experimental
arrangement was an A1203 linear polarizer at Brewster's angle which fed
the A1 phase shifter mirror, and then analyzed by an LiF Brewster-angle
analyzer in combination with a detector. They produced a circular polarizer
which transmitted 12% with E - / E + = 6 to 10 at A =1216A with a 5"
acceptance angle at c#+ = 52". Circular polarizers are considered later in
discussing thin films and crystal polarizers.22
THIN FILMS 125
3.5. Thin Films

3.5.1. Single-layer Films
3.5.1.l.Polarization Effects. In general, most optics have thin films on
their surfaces, and these introduce, via Fresnel's equations and phase
interference, an effect on the P and S components. Figure 13 is an example
of a light beam in a medium ( n l )incident at an angle 4, which is refracted
at an angle 4, in a film ( n 2 ) d thick on a substrate ( n 3 ) . Multiple substrate
reflections interfere with front surface reflections to produce minima and
maxima in the reflectance depending on the incidence angle, film thickness,
and material constants.23
The phase shift added to a beam after traveling through the film thickness
once compared to a beam traveling in the medium is,
cos 4,
phase thickness = /3 = Wr =2 m 2 d -. (3.23)
A
The reflected and transmitted beams are written in terms of Fresnel's
reflection, r m , n , and transmission, r m , n , where the coefficients with the
subscripts m and n denote the boundary media, with m being the incident
medium. For the reflected beam, the components in order of intensity are:
e-'J@,.. . .
e-''@, t12r23r21r23t21
r I 2 ,r12r23r21
The total reflection coefficient is:
e-2Jp+ r21r:3r12r:1
rs,p= r 1 2 +r23rlzr21 e-'J@+ - .
*
The reflection coefficient, rs.p, can be rewritten with relations between the
transmission and reflection coefficients.
+
- r I 2 r2, e-J2@ (3.24)
'S.P - 1 + r12r23 e - i 2 ~ .
This is written for complex P or S coefficients and is a description for an

absorbing substrate and film. Beam splitters, overcoated mirrors, interfer-
ence filters, etc., have variations in polarization due to thin films which
need to be considered.
Rewriting the reflection coefficient between a dielectric film and a metal
substrate in complex exponents produces a single equation for the
reflectance.
r23= pZ3eJ'23,
(3.25)
FIG. 13. Multiple reflections in a thin film.
The reflectance thus varies periodically with thickness with extrema given
by d R / d d =O. This occurs when
sin(S2,+2P)=0, S2,+2P=rn7r, rn=0,*1,*2 , . . . ,
The phase term (S2,+2P) can be adjusted with /3 either by film thickness
. ~ ~ 14 is a plot of
or by the incidence angle for a specific ~ a v e l e n g t hFigure
r, and rp versus film thickness for two incidence angles showing the periodic
nature due to interference. The maximum and minimum values are deter-
mined by the reflection coefficients and the material. For a dielectric film
and substrate, the maxima and minima occur at phase thickness equal to
odd and even multiples of a quarter wave. Figure 15 is a plot of the reflection
coefficients versus the incidence angle for two film thicknesses. In Fig. 16
the incidence angle and the film thickness are held constant, making the S
and P reflections vary periodically with wavelength.
3.5.1.2. Linear Polarizers and Phase Retarders. Writing the ellip-
sometric equation r p / r s=tan Ic/ e J Afor the general case of an absorbing film
and absorbing substrate gives an equation nicely mapped in the complex
plane. The points where tan II, equals zero or infinity are possible designs
for a P or S extinguishing polarizer, while points tan Ic/ = 1 and A = *77/2
produce a circular polarizer. This suggests that the phase difference can be
controlled between r,* and r2, via the incident angle, refractive indices, and
film thickness to make P or S reflective polarizers, linear partial polarizers
(which differentially attenuate the P and S components without shifting
the phase between the components), and wave retarders.25-28
The P or S extinguishing polarizers with a single dielectric film on a
metal substrate require large incidence angles for the higher reflecting
THIN F I L M S 127
OA la
Film thickness, d
1.0
FIG. 14. rr and r,, for reflections from air to glass film to glass substrate showing the effect
on r, and r,, as a function of film thickness, n , = 1.0, n, = 2.0, and n, = 1.5.
1.0 r I
....
Incidence angle, (h
Ir, I
-.-.-.- Irpl
0.u 90.0-
Incidence angle,
F I G . 15. r\ and rp for reflections from air to glass film to glass substrate with a 2h and a
Oh film, showing the film's effect as a function of incidence angle.
-1
4
-
L-
p = 30'
0.0
5000 A 5500 A
Wavelength, h
1.0
t = 60'
FIci. 16. rs and rp for reflections from air to glass film to glass substrate showing the film's
effect as a function of wavelength about a central wavelength A o .
substrates, since the reflectance required from the dielectric's surface to

match the substrate's reflectance means large incident angles (> 80°).2y~'0
Also, the slope for the reflectance versus the incidence angle for the dielectric
is very steep, thus making the angular acceptance very sensitive. A gold
substrate with its lower reflectance can be used with a high-index film to
produce a reasonable P or S extinguishing polarizer which has high extinc-
tion, smaller incidence angle, and better angular sensitivity. Another solution
is to add some absorption to the film (this cannot be used in high-energy
beams due to the absorption) which produces multiple incidence angles
(75'-85') at which the P or S component is extinguished and the sensitivity
to angular acceptance is d e ~ r e a s e d . There
~' is a trade-off of throughput for
angular insensitivity and the unwanted component's intensity. These are
simple, single-film, wavelength-dependent reflection polarizers which are
useful when used with low-power densities. There also exist multilayer
dielectric films on a transparent substrate which are used as reflection/trans-
mission polarizers for higher-power densities.
3.5.2. Multilayer Films

3.5.2.1.Polarization Effects. Another way to make thin-film polarizers
is to use periodic multilayer dielectric films of high and low in dice^.^*-^^
THIN FILMS 129
These are suitable for beams with high-energy densities. The time depen-
dence of the wave equation, wt, is assumed fixed, and the equations are
expressed as second-order differential equations in space for the magnetic
and electric wave propagation. The multilayer’s effect on the incident beam’s
magnetic and electric field is expressed as a 2 x 2 unimodular transformation
matrix. The new beam produced from the first layer is applied to the second
layer, producing a modified beam for the third layer, etc. The total thin-film
stack can thus be expressed as a single unimodular transformation matrix.
To see the effect of the dielectric multilayer on the S and P polarization,
first consider the mth thickness z, at normal incidence. The matrix transfor-
mation of the electric and magnetic fields is
The diagonal and cross-diagonal terms are pure real cosine and pure
imaginary sine terms with arguments which are the phase thickness,
2rn,z,
A,=-. (3.28)
A0
To apply this at oblique incidence angles, the effective phase thickness and
effective index (a, p m )are used to describe the uniform plane wave at an
oblique incidence as a nonuniform plane wave at normal incidence, via
6, = A, cos 4, and
where 4, is the refracted angle through the layer. The last two relations
were introduced in Fresnel’s equations earlier and are substituted into
normal incidence Fresnel equations to produce nonnormal incidence
equations.
Periodic multilayer dielectric thin films have a transformation matrix,
also known as a Herpin matrix, which is equivalent to a two-film combination
for a particular wavelength. A symmetric three-layer film is a special case
which simplifies to an equivalent single-film matrix for all wavelengths. The
single film is described by an equivalent index, N, and an equivalent phase
thickness, r.
(3.29)
A symmetric three-film combination with an effective phase thickness (Y + /3

can be made by sandwiching one layer between the halves of the other
layer. Three films with index layers n,,, n p , n, and phase thicknesses a / 2 ,
p / 2 , a / 2 result. Equating the three-film transformation matrix to the single-
film equivalent matrix gives an equation for the equivalent phase thickness,
cosh r = cos (Y cos /3 (3.30)
For values of lcosh rl> 1, the effective index N is complex in order to

preserve the pure imaginary cross diagonal terms for the equivalent single
film. This means that there is attenuation of the beam propagating through
the medium (a stop band), while for lcosh I'l< 1, N is real and the beam
propagates through the medium (pass band). The number of periodic layers
increases the attenuation and improves performance. Since the effective
index for the S and P components are functionally different in 4, an
incidence angle and phase thickness can be found for a given wavelength
which will give lcosh rl> 1 for the S component and lcosh rl< 1 for the P
component, thus making an S-reflecting, P-transmitting polarizer. If it is
possible, N is adjusted to the medium surrounding the film, or else an
antireflection coating is needed. In practice, there is more to multilayer
polarizer design, such as antireflection coatings to index match the multilayer
to the air and/or glass interface, or to design the multilayer to take into
account the electric field distribution through the stack to reduce laser
damage.37-41
3.5.2.2. Linear Polarizers and Phase Retarders. Mahlein has published
articles on quarter-wave multilayer dielectric stacks for laser mirror
His graphs represent various thin-film combinations of n, versus
np ranging from 1.3 to 3.0 with various incidence angles for P or S
extinguishing reflection/transmission polarizers. He also shows possible
designs for nonpolarizing beam splitters and polarizing interference filters.
As for reflection quarter-wave retarders, Southwell has designed a dielec-
tric multilayer on a silver substrate with a reflectance equal to 99'/0.~~His
design starts with a quarter-wave-thick periodic multilayer stack, and the
film thicknesses is adjusted according to a merit function which maximizes
the desired characteristics-90" retardance *3' over a 5% range about the
design wavelength and a reflectivity of 99.8%. These coatings can be used
to reduce polarization effects in corner cube reflectors.
3.5.2.3. MacNeille Polarizing Prisms. Another application of thin films
as a polarizer is in beam-splitting glass cubes first described by
~ ~ cube is cut along a diagonal, and a multilayer stack
M a ~ N e i l l e . ~A' *glass
is deposited such that the Brewster angle condition is met at all the diagonal
interfaces, as drawn in Fig. 17. If the effective indices for the high- and
THIN FILMS 131
FIG. 17. A P-polarizing MacNeille prism using multiple reflections at Brewster's angle &.
low-index films are set equal, the parallel Fresnel-reflection coefficient is

zero. This condition with the Brewster angle condition produces an equation
in terms of the refractive indices and the incident angle in the prism.
nG sin cPG = n H sin 4H= nL sin 4L,
(3.31)
Commercially-made, broad-spectral-band, thin-film-polarizing cubes

(4000 A to 7000 A) are available with an extinction ratio of 100 to 1, and
narrow-band cubes are available with an extinction ratio of 1000 to 1.
Optically coupling the prism halves for high-power densities is difficult,
and residual birefringence exists in the glass. Liquid prisms with immersed
interference-coated substrates have been demonstrated as practical to over-
come these problems and can have a wide spectral range from the UV
through the ~ i s i b l e . ~ '
3.5.2.4. Total Internal Reflection Devices. Total internal reflection
phase-retarding devices, such as the Fresnel rhomb mentioned earlier,
control the phase difference introduced on reflection to make quarter- and
half-wave retarder^.^"'^ The phase difference between the P and S com-
ponents, A, for most glasses in air does not occur at the maximum, A,,,,
in the plot A versus the incidence angle (Fig. 6). Usually the larger incidence
angle is used to make the rhomb, since its angular sensitivity is not as large
as the smaller incidence angle. For the least angular sensitivity, Amax would
equal the desired phase retardation. Increasing the refractive index increases
A,,,, so a film with a smaller refractive index is deposited, and its thickness
is adjusted to give A,, = 45". For a *3" incidence angle variation over the
132 O P T I C A L POLARIZATION
visible spectrum, the retardation variation is 0.4" for a coated rhomb and
5.0" for an uncoated version.
Imperfections in total internal reflection devices also are due to strain
birefringence which is inversely proportional to wavelength, and for applica-
tions in the ultraviolet it becomes critical, especially if the retarder is to be
achromatic. Also, it is reported that polishing the faces that are not used
reduces strain 10-15 mm inside the glass by a factor of two over that for a
fine-ground face. Attention should be given to mounting the retarder, and
silicone rubber sealant as an adhesive does not introduce any measurable
strain.
Another error source is the refractive index of the glass and thin film.
For the glass, the surface is slightly different in phase retardation, about
0.5" more than the theoretical values, because surface shaping and polishing
changes the refractive index. Also, generally for most materials, the refractive
index increases with a decrease in wavelength, except for polyvinyl alcohol.
Retardation error sources can be corrected by adjusting the reflected beam's
phase difference experimentally with film thickness. The film can increase
achromatism, because the increase in refractive index with decrease in
wavelength can be balanced by the thin film's increased phase thickness
due to p. King has made Fresnel rhombs with either 144A magnesium
fluoride films on each reflecting surface or a 268A film on one surface
leaving the other uncoated, which produce a retardation of 90" f 0.2" from
3440 A to 6680 A compared to 90"*2.5" for an uncoated rhomb.
In Fig. 18, various other retardation devices are shown. They have about
the same chromaticity but various other advantages and disadvantages with
regard to beam displacement, physical size, and angular acceptance. A
Fresnel rhomb displaces the beam; and, if not optimized by thin films, it
has a 9" variation in retardation with a *7" incidence angle. Achromatic
device one, ADl, is self-compensating in incidence angle, because the first
two reflections on the minus side of the optimum angle will have two
reflections on the positive side, therefore almost canceling the off -angle
effects. It is a very long prism, 175 mm long for a 10-mm beam, and it would
be hard to meet theoretical predictions due to strain birefringence. It has
a 0.7" retardance variation for a *7" incidence-angle variation. The Mooney
rhomb is made of two dense flint 60" prisms and is self-compensating in
incidence angle with a performance similar to AD1. Also, its physical size
is small. Achromatic device two, AD2, is physically small and does not
deviate or displace the beam, but its incidence angle sensitivity is terrible
with a 13.2" retardation variation for a *3" incidence-angle variation.''
Half-wave or full-wave rhombs have also been designed to deviate the light
beam 90" without changing the polarization. A simple design is the dove
prism.
WAVE PROPAGATION I N A N I S O T R O P I C M E D I A 133
FIG. 18. Total internal reflection devices. (a) Fresnel rhomb, ( b ) Mooney rhomb, (c)
achromatic device 1 ( A D l ) , and (d) achromatic device 2 (AD2).
3.6. Wave Propagation in Anisotropic Media
3.6.1. Birefringence in Crystals

3.6.1.1. Dielectric Tensor and Energy Distribution. Another way to
separate orthogonal polarizations is by dielectrics with anisotropic optical
constants either naturally occurring in crystals or induced by an electromag-
netic field or stress and compression. This produces different velocities for
orthogonal polarizations depending on the travel direction in the crystal.”
The medium can be described in terms of its dielectric constants as a
nine-element tensor.
(3.32)
where D is now not necessarily parallel to E as for an isotropic media, and

the angle difference is between the wave front’s normal, s, and Poynting’s
energy propagation vector, S. This means that there is a wave-front phase
velocity and an energy ray velocity as shown in Fig. 19. The phase velocity
is a projection of the ray velocity onto the wave front’s normal. Maxwell’s
equations and the material equations with the dielectric tensor give an
equation relating the phase velocity to the wave normal direction and the
material constants.
Considering the functional form of the energy density, the dielectric
tensor is seen to be symmetric, and at most six independent values need to
be determined. A coordinate system, termed the principal axes, can be fixed
to the tensor’s diagonal. In this coordinate system, the material equations
simplify to the diagonal terms.
E,,+Ex, Eyy+Ey, EZZ+EZ,
i = x, y , x.
Assume the media is homogeneous, nonconductive, nonmagnetic, and
electrically anisotropic. A plane wave, Eq. (3.4) simplifies the derivatives
in Maxwell’s equations.
V x H =j w D = -jks x H
V x E = -jpwH = -jks x E
(3.33)
:);( = phase velocity = V,
(kY2 =velocity in i direction =
(3.34)
where s is the unit vector for propagation direction and k defines equal
phase planes A apart. D and H are perpendicular to s, but E is not. E is
perpendicular to the Poynting energy flow vector, S.
WAVE PROPAGATION I N ANISOTROPIC M E D I A 135
S
E
FIG. 19. The orientation of the electromagnetic vectors H, D, and E, the wave normal s,
and the energy propagation vector S in an anisotropic medium.
The Ei can be eliminated, leaving an equation of the material constants,

the propagation direction, and the phase velocity.
(3.35)
The result is known as Fresnel’s equation of wave normals. V ; has two

values and implies with Eq. (3.34) that the medium has two linearly polarized
beams with different phase velocities. It can also be shown that the two
electric displacement vectors are perpendicular.
3.6.1.2. Crystal Types by Optical Properties. The dielectric tensor is a
second-order symmetric tensor, and therefore it needs six parameters to
describe it. These parameters can be interpreted as the principal axes’ lengths
and angular orientations with respect to the crystallographic axes. The
number of parameters to describe the tensor is dependent upon their optical
anisotropies which are subdivided according to crystallographic structure.”
(1) Optically isotropic crystals are called cubic crystals. Three crystal-
lographically, mutually orthogonal axes can be found, and the diagonal
terms of the dielectric tensor are equal. Only one parameter, the refractive
index, is needed to describe the dielectric tensor.
(2) Uniaxial crystals have a plane with equal optic constants and a
direction normal to the plane with a different optic constant called the optic
axis. Two parameters, the lengths of the two principal axes, are needed to
describe the dielectric tensor. A set of crystallographically equivalent direc-
tions are in the plane about an axis of symmetry aligned with the optic
axis. Crystals can have three, four, or six crystallographically equivalent
directions in a plane or three-, four-, or sixfold symmetry about the optic axis.
Trigonal, tetragonal, and hexagonal crystals are in this group. Calcite is
an example. It is a rhombohedra1 crystal, and the optic axis is a threefold
136 0 PTlC AL POLAR1 ZATIO N
symmetry axis directed into the rhomb’s corner. A surface cut perpendicular
to the optic axis on the corner accurately determines the crystal’s principal
axes.’
Unpolarized light incident on a uniaxial crystal produces two linearly
polarized beams determined by the optic constants for the plane and the
optic axis. For all field orientations with respect to the optic axis, an electric
vector parallel to the plane exists and behaves as for an isotropic medium.
This is termed the ordinary beam, 0, and is described by a spherical velocity
surface. The wave perpendicular to the 0 beam depends on the incident
beam’s orientation and varies from the optic constants for the plane to the
optic constants perpendicular to the plane. This is the extraordinary beam,
E, and its phase velocity as a function of propagation direction is a
spheroidal surface.
The phase velocity surfaces for a uniaxial crystal in terms of the principal
axes are found by applying Fresnel’s equation of wave normals. The z axis
is made parallel to the optic axis, and the angle from the wave normal to
z is 9.
v, = v, = v,, v, = v,,
s: + s: = sin2 9, s t = cos2 8,
( V ; - vf)[(V ; - ~ f sin’
) 9 + ( V : - v;)cos2 e] = 0, (3.36)
v;,= vf,
v;*= vf sin2 e + v’,cos2 e.
Figure 20 is a plot of the two phase velocity surfaces and an arbitrary wave
normal, s, with its two phase velocities. When s is in the optic axis direction,
the phase velocities are equal, and when s is perpendicular, the difference
between phase velocities is maximum.
z O.A. z O.A.
FIG. 20. Cross sections of phase-velocity surfaces for a negative and a positive uniaxial
crystal predicted by the Fresnel equation for wave normals. The phase velocities are determined
by the wave normal’s orientation.
WAVE PROPAGATION IN ANISOTROPIC MEDIA 137
Birefringence is the difference between the 0 and E refractive indexes.

birefringence = nE - no (3.37)
The sign means that the ordinary velocity surface is ahead of or behind the
extraordinary surface.
(3) Crystals with principal axes as a function of three dielectric constants
and angle orientation with the crystal axes are called biaxial crystals. TE
waves are refracted at an angle depending on orientation with respect to
the crystal's axes and wavelength. Both waves perpendicular to s act as
extraordinary waves, and the velocity surfaces are general ellipsoids.
Orthorhombic crystals have their principal axes aligned with the crystallo-
graphic axes, and so only three parameters, namely, the lengths of the three
principal axes, are needed. Monoclinic crystals have a fixed principal axis
parallel to a crystal axis and need four parameters for the dielectric tensor.
Triclinic crystals need all six parameters to describe the dielectric tensor.
Applying the Fresnel equation of wave normals to biaxial crystals has
three dimensions that can vary. Three cross-sections, each about a different
principal axis, simplify the problem. Figure 21 shows three cross sections
for a biaxial crystal. One cross section has the curves touching at four
points, and the two lines connecting the origin with the points are the
directions of the optic axes. With the wave normal in these directions, the
two phase velocities 'are equal.
Quartz and calcite are the most commonly used crystals for crystal-
retarding plates and prism polarizers and beamsplitters. Quartz is a biaxial
crystal. A beam propagating along the optic axis separates into L and R
circular polarizations due to a weak circular birefringence.
nL- n R = *0.000071 for A = 5892 A.
Separating L and R polarization in phase is equivalent to rotating the
PoincarC sphere about the L and R poles, and the resulting elliptical
2 Z Y
vx > vy > vz
FIG. 21. Three cross sections of phase-velocity surfaces described by the Fresnel wave
normal equation for a biaxial crystal.
polarization azimuth is rotated according to the phase shift. This

phenomenon is termed optical activity.
The beam’s ellipticity changes quickly to linear polarization for angles
not far from the optic axis, and quartz behaves almost as a uniaxial crystal.
Quartz’s weak linear birefringence makes it useful for retardation plates.
nE- no = 0.0091 for A = 5892 A.
Calcite is a negative uniaxial crystal. Prism polarizers and beamsplitters

are usually made from calcite, because the transmission and birefringence
are large, from 2000 to 12,000 A.
nE = 1.486, no = 1.658 for A = 5892 A.
The literature has references with material constants given for the more
commonly used crystals and wavelength ranges.
3.6.2.Crystal Polarizers
3.6.2.1. Prism Beamsplitters. Beam-splitting polarizers, such as the
Rochon, SCnarmont, and Wollaston are biprisms which separate the 0 and
E beams by introducing a discontinuity in the refractive index for either
the E beam (Rochon and Sknarmont) or for both the E and 0 beams
(Wollaston). Figure 22 shows the E and 0 rays for a Rochon and a
Wollaston. The Rochon and SCnarmont are basically the same with their
first prisms’ optic axis perpendicular to the entrance face.
The Rochon gives a slightly larger E-beam deviation than the SCnarmont,
because the beam sees the maximum birefringence. These prisms can be
used backwards, but the E-beam deviation will be slightly less. Snell’s law
at the prism discontinuity, now from nE to no, reverses the incident beam’s
refracted angle from greater to smaller deviation. The beam deviation and
FIG.22. ( a ) Rochon and ( b ) Wollaston polarization beam splitters. The optic axis direction
is oriented parallel I or perpendicular 0 to the page.
WAVE PROPAGATION IN ANISOTROPIC MEDIA 139
accepted field angle are a p p r ~ x i m a t e l y ~ ~
(3.38)
A Wollaston polarizing beamsplitter has the optic axis of its first prism
parallel to the face to produce an E and an 0 beam which sees the second
prism's optic axis rotated 90" about the beam axis to produce a maximum
birefringence and deviation for both beams. Wollaston prisms can produce
angular separations up to 20" with a slight asymmetry, and triple-element
Wollastons produce larger angular separations. At large angles through
crystal polarizers, image distortion occurs due to nE varying in angle with
respect to the optic axis.
Since the birefringence of Rochon's first prism is not used, it can be
substituted with strain-free glass of either index, nE or no. Glass eliminates
image distortion due to the effects of the birefringent crystal and is easy to
manufacture cheaply with a good tolerance. The residual birefringence in
glass needs to be c o n ~ i d e r e d . ~ ~
3.6.2.2.Prism Polarizers. Polarizing prisms are birefringent crystals cut
into two pieces with an interface of an index which totally internally reflects
the higher-index beam, for calcite the ordinary beam.' A Glan-Thompson
prism is illustrated in Fig. 23. A Nicol prism was the first polarizing prism
design, and it is essentially a calcite rhomb cut along a diagonal and
cemented together. Glan-type prism polarizers arrange the optic axis to give
i"
i
i
i
I
i
i
,-I _._.-.- 0
- - - - - - -. E
FIG. 23. Clan-Thompson prism drawn with normal and extreme rays. The optic axis is
perpendicular to the page, 0.
maximum use of birefringence, thus increasing acceptance angle. Beam

distortion and polarization uniformity are better controlled for Glan prisms.
The prism’s angular acceptance is limited by either the critical angle of
the larger-index O-rays, or by the lower-index transmitted E-ray being
refracted along the prism’s cut angle. The refractive index discontinuity,
n,, at an angle S in Fig. 23 has its normal rotated ad angle ( 9 0 ” - S ) with
respect to the entrance normal. Applying Snell’s law to the 0 beam at
surface 1-2 gives an equation relating the maximum incidence angle for
total internal reflection to the angle S and the cement’s refractive index, n,,
n, sin di= ( n i - n:)”, cos s - n, sin S. (3.39)
For the E beam’s limit, the total internal reflection case n E > n, can be
described with the 0 beam’s equation and the appropriate substitutions,
n , sin di= n2 sin s - ( n i- n:)”, cos S. ( 3.40)
To make the maximum usable acceptance angle for both 0 and E beams,
the two incidence angle equations are made equal in magnitude by adjusting
the cut angle and cement. This gives,
( n h- n:)”,
tan S = (3.41)
2n2
The acceptance angle reaches a maximum for n2 = n,.
For n E < n,, the maximum accepted incident angle for the transmitted
beam is refracted along the cut.
n, sin d
,, = n2 sin S.
By equating the maximum incident angle for the transmitted beam to the
totally reflected beam gives an equation relating the cut angle and the
(3.42)
Considering the totally reflected beam, the maximum incident angle to be

reflected increases with the smallest cut angle which is achieved with n, = n E .
The tan S equations reduce to the same form for n2 = nE. Glan-type prisms
are sold with the length-to-aperture ratio assuming a square aperture, and
this gives an idea of its acceptance angle.
L 1
(3.43)
A tans’
Since optical cements have absorption, air-spaced calcite prism polarizers
are made for use up to calcite’s ultraviolet limit and for high-energy densities.
WAVE PROPAGATION I N ANISOTROPIC MED IA 141
The usable maximum (symmetric about the prism's normal) incident angle
or semifield angle is about 8" in the ultraviolet, decreasing for longer
wavelengths, and is limited by the transmitted E beam being totally inter-
nally reflected at the cut. Large prism apertures can be made, since most
of the crystal is used.
An air-spaced Glan-Thompson is termed a Clan- Foucault The optic
axis is parallel to the cut surface, so that the transmitted beam has its
electromagnetic vibration oriented parallel to the cut surface, which has
the higher Fresnel reflection coefficient, r,. If the optic axis is rotated 90"
about the beam axis, the reflection coefficient is now rp,which is significantly
reduced. The transmission is increased in the ultraviolet by 10% for this
design, which is called a Clan- Taylor. The optic axis orientation decreases
the useful birefringence which reduces the semifield angle. The cement n,
is unity, and the angular acceptance primarily depends on the cut angle
and the crystal's wavelength-dependent birefringence. The L / A ratio is
about equal to one.
To increase the semifield angle or to decrease the L / A ratio for a given
maximum incident angle, double prisms are constructed.' Examples are the
Ahrens prism (two Glan-Thompson prisms next to each other), the Grosse
prism (an air-spaced Ahrens), and the Marple-Hess prism (a double Glan-
Taylor back to back). Since the number of components increases in double
prisms, the manufacturing difficulty is increased, and extinction perform-
ance is about an order of magnitude worse than for single polarizing prisms.
The Ahrens and Grosse prisms have half the L / A ratio of their Glan
single-prism counterparts. For example, an Ahrens prism with an L / A of
1.8 has been made with a 26" semifield angle. The Grosse prism has an L / A
about equal to one-half and is useful for tight physical limitations. Marple-
Hess prisms have definite advantages due to the cut angles S and -S. 0-rays
below the axis will be totally reflected by the first surface, and 0-rays above
the axis will be totally reflected by the second surface. The axial 0-ray and
its total internal reflection due to n , ( A ) decreasing in A is the design
parameter for the ordinary beam. The prism's limiting field angle is deter-
mined by the E-ray's total reflection, and the limiting wavelength depends
on the axial 0-ray's transmission. These prisms have a lower transmission
in the ultraviolet because of their increased L / A ratio.
Glass-calcite Glan-type prism polarizers are possible, because the second
prism is used to transmit the E-ray, and its birefringence is not used. The
second crystal prism can be replaced with an index n, glass prism. Or, the
first prism can be replaced with an no index glass, and an n, cement used
to assemble the prism. The E-ray is totally reflected in this case. These
prisms are usually not as good as all-crystal polarizers, because the greater
L/ A ratio and optical cement enhance the resulting residual strain.s3
As with prism beamsplitters, image distortion in prism polarizers occurs

for off-axis beams. Other contributions to degradation in extinction ratio
is scattered light and the matter of eliminating the unwanted ray. One
solution is to blacken the crystal’s sides. Another method is to cut and
polish the prism side to make the unwanted beam exit normal to the exit
face. The unwanted beam is saved. Prism performance studies show an
order of magnitude increase in extinction ratio when the detector is moved
farther from the prism, and entrance and exit apertures are used.s4 The
crystal grades, optical finish, strain birefringence in contacting the two
prisms, and optical alignment affect the prism’s p e r f ~ r m a n c e . ~ ~
3.6.3. Retarding Waveplates

3.6.3.1. Crystal Retardation Plates. A quartz or calcite plate with its
optic axis parallel to the entrance surface introduces a phase shift between
linear orthogonal polarizations,
d
6 = 274 n E - n,) -. (3.44)
A
For n E > no, linearly polarized light travels faster parallel to the optic axis,
and for nE < no, the fast direction is perpendicular to the optic axis. A fast
(F) and a slow (S) coordinate system oriented with the optic axis describes
the retarder’s orientation.
The PoincarC sphere is useful to describe retardation effects. Incident
and resultant elliptical polarizations are points on the sphere which are
transformed due to phase shifts. A retarder’s phase shift between F and S
rotates the incident point about the point marking F’s orientation. A positive
phase shift is a clockwise rotation about F equal to the phase shift. For
example, a quarter-wave retarder with its fast axis at +45” will pass H as
R and V as L polarizations.
The spectrum viewed with a retarding plate between crossed polarizers
is fringed with the polarizations’ cyclical wavelength retardation. The
wavelength interval between equal polarization states decreases for the
thicker plates and can be used to filter or select a wavelength by selecting
the corresponding polarization ellipticity. Lyot birefringent filters use this
method to make typically a 1/4-A pass band isolating the fringe from its
neighbors by cascading plates with double the fringe spacing in the spectrum
or half the previous element’s thickness until the final element’s fringe
spacing is 128 or 256 A.
Retardation plates are usually made of quartz, mica, or plastic for the
near ultraviolet to the near The spectral range is expanded
by using other birefringent crystals. Calcite’s birefringence is very high, thus
enhancing its dependence on thickness and temperature. A quarter-wave

plate for a particular wavelength would be more difficult to obtain and
maintain than for quartz. The thickness of a retardation plate can be a
single-order retardation period (which can be achieved with split mica or
quartz in the infrared) or multiple orders thick.
d = mh. (3.45)
Thick multiple-order plates are sensitive to changes in the retardation

equation. Wavelength, birefringence, refracted angle, and temperature vari-
ations are factors in determining the resulting retardation. The sensitivity
can be decreased by using thick plates by orienting one plate’s optic axis
perpendicular to the other plate and by adjusting their thickness difference
to a single-order thick. This combination acts as a single-order plate and
is called a jirst-order plate.
3.6.3.2.Compensating Retardation Plates. All simple retardation
plates still depend on wavelength. The retardation’s dependence on
wavelength can be corrected for either with variable thickness plates called
compensators, or with composite plates to produce the desired retardation
characteristics.
Compensators change the net thickness experiencing a birefringence
which directly changes the phase difference between the two orthogonal
polarizations, as demonstrated by Poincare’s sphere. The Babinet com-
pensator is a fixed and a movable quartz wedge with the optic axis parallel
to the entrance and exit surfaces, but crossed to each other. The field’s
center is marked with a reference line on the fixed wedge, and the relation
between the movable-wedge displacement and retardation can be empiri-
cally determined by using crossed polarizers as a polarizer and an analyzer.
White light viewed with a Babinet between crossed polarizers produces
colored bands and extinction indicating changing retardation orders. Zero
phase difference at the center will produce a dark line with the reference
line. Introducing a retarding plate will shift the fringes, and a correction
opposite the retarding plate’s phase shift is needed to extinguish the center
again. The distance moved by the wedge determines the introduced phase
shift.
To increase the useful field, a Soleil compensator is used. It is a fixed
retarding plate and a variable-thickness plate with their optic axis crossed.
The variable thickness plate introduces a uniform variable phase shift across
the viewing aperture. If white light is viewed with a Soleil between crossed
polarizers, the field is tinted with phase-shifted wavelengths. Zero net
thickness between the plates results in extinction. A retarding plate intro-
duced between the polarizers will transmit a particular wavelength, and a
phase correction made by the variable-thickness plate will again extinguish

that wavelength.
The Senarmont compensator is an elliptical polarization analyzer consist-
ing of a quarter-wave plate and a polarizer. Elliptical polarization is incident
with its major axis aligned with the wave plate's fast and slow axes. The
wave plate will retard either the major or minor axis by a quarter wave,
resulting in a linear polarization with an azimuth determined by the ellip-
ticity. An elliptical polarization is selected with the azimuth of a following
linear polarization analyzer with respect to the wave plate's axes. If the
elliptical polarization analyzer is turned around with light now incident on
the linear polarizer, then a known elliptical polarization with its axes aligned
with the wave plate is produced. For example, consider how circular
polarization is produced. The linear polarizer oriented parallel to the wave
plate's axes produces linear polarization. The linear polarizer oriented 45"
to the wave plate will produce either L or R circular polarization.
3.6.3.3.Achromatic Wave Plates. Composite achromatic wave plates
can be divided into composites of the same or into composites
of positive and negative birefringent crystals similar to focus correction for
lenses.60 Possible positive and negative birefringent wave-plate composites
for visible applications are magnesium fluoride and ammonium deuterium
phosphate or magnesium fluoride and potassium deuterium phosphate
designs with a 0.5% deviation; a calcite, quartz, and magnesium fluoride
triplet with a 0.5% deviation; a magnesium fluoride and quartz doublet
with a 5% deviation over the spectral region.
Composite wave plates with the same material can be made with zero
first-order variations in retardation and wave plate axis. This is demonstrated
using Jones' matrix techniques. Three- and nine-plate combinations act as
pure wave plates, and four- and ten-plate combinations behave as a wave
plate plus a rotator.
A three-element achromatic wave plate made of the same material with
the first and third elements identical leads to the consequence that they are
parallel. Various achromatic combinations of three-element wave plates for
a desired retardation can be made using these assumptions, but the retarda-
tion axis may vary to first order in wavelength. Constraining the retardation
axis leads to the middle plate being a half-wave plate, which is rotated by
an angle 0 with respect to the outer elements. The total retardation is 26.
COS~O=-T~~
cos 6 +-
( 3 = dlsin(26,), (3.46)
where *A6 is the tolerance as a function of wavelength.

For three-plate combinations, some chromaticity is left, due to a quadratic

term. A central-design wavelength needs to be chosen, and the allowable
deviation from the desired retardation gives the useful spectral range. The
correct retardation occurs at a wavelength above and below the central
wavelength for all but the half-wave combination which varies cubicly.
Combinations for the nine-element retarder (a threefold three-element
retarder) are possible to make a retarder's retardation vary about 1% from
3514 to 10,000 A.
For example, the three-element quarter-wave plate consists of two outer
plates with a 115.5" retardation and a half-wave plate. A half-wave plate is
made of three half-wave plates. A small additional rotation of the central
half-wave plate introduces a first-order term which cancels the third-order
term, improving the retarder's performance.
If a three-element half-wave plate is split in the middle and the halves
are then rotated, retardation is equal to twice the complement of the angle
rotated. Also, the polarization is rotated by the complement of the separation
angle. This combination of quarter-wave and half-wave plates is a four-
element variable-retardation plate which rotates the polarization axis. Nine-
element half-wave combinations split in the middle give a ten-element
variable wave plate with a performance similar to the nine-element one and
in addition, it rotates the polarization as the four-element wave plate does.
Two-element composite retarding plates have been mentioned in the
literature, but upon considering the Jones' matrix calculation for two wave
plates, no degree of achromaticity can be obtained, and the plates need to
be tuned for the wavelength.
Wave plates are usually made of quartz, mica, or stretched polyvinyl.
alcohol. Since the birefringence of mica and polyvinyl alcohol is variable,
wave plates need to be selected for the application. Some of the plastic
laminates have a curl to them, and it is best to leave the curl alone.
Mica sheets, usually muscovite, can be split into large apertures to the
desired retardation by trial and error to account for the birefringence
differences for different crystal samples.6' This can be accomplished with
adhesive tape, which can remove layers that are 5-micron-thick or less and
can remove high spots in large apertures. Since the wave plate is thin, little
beam deviation and wave-front distortion occurs, but multiple coherent
reflections can occur which can be eliminated by cementing thin films into
an optically thick medium. Ultraviolet light curing cement is suitable for
this purpose, taking care that the illumination is uniform across the aperture
to minimize stress.
3.6.3.4. Measuring Wave Plates. To determine the optic axis for a wave
plate, it is placed between crossed polarizers and rotated for maximum
intensity. The entrance polarizer and wave plate act as an elliptical polarizer
with the resultant polarization ellipse axes aligned 45" to the wave plate's
extraordinary and ordinary axes. The analyzer is then oriented 45" to the
ellipse axes. Rotating the wave plate about the optic axis direction has no
effect on the birefringence. The light-path length in the crystal and retarda-
tion will change, but the analyzer will transmit. Rotating the wave-plate
surface perpendicular to the optic axis changes the birefringence from a
maximum to a zero value for a 90" rotation, and so extinction occurs.62
The wave plate's retardation is related to the intensity of the elliptical
polarization's major and minor axes, as seen earlier. It is a matter of
measuring the maximum intensity of the above crossed polarizers and
wave-plate arrangement and rotating the analyzer 90" to measure the
minimum or minor axis intensity,
(3.47)
The error in measuring the wave plate's retardation is determined by the

intensity's error measurement and the crossed polarizer's extinction ratio.
A 0.001YO intensity uncertainty produces less than 0.1" retardation measure-
ment error.63
3.6.4. Induced Birefringence and Phase Modulation

Anisotropies can also be induced in a material's dielectric tensor by elastic
stresses, electric fields, and magnetic fields, divided into elasto-optic, electro-
optic, and magneto-optic effect^.^^'^^ Linear electro-optic and elasto-optic
effects are due to nonlinearities between the material polarization vector P
and the electric field. The material effects are described in terms of the
impermeability tensor, B = E - ' . Small changes in the tensor due to linear
changes in the electric field ( E ) and strain (S) is
A B, = rzkEk + pfklsk[= A B z -k A B: . (3.48)
where rllk is the electro-optic coefficient, and P Y k l is the elasto-optic
coefficient. The superscript implies holding that variable constant while
determining the other variable. Summation is implied for indices which
appear in both terms of a product of matrix elements. For example, the
term r,,kEk means to sum the elements varying k to account for all k
contributions. Also, symmetric tensors can be expressed in a compressed
matrix notation to simplify the matrix element numbering from row and
column to an ordering as follows:
r I 3= r3, = r s , r,* = rz, = r,.

WAVE PROPAGATION IN ANISOTROPIC M E D I A 147
Crystals with a linear electro-optic effect are also piezoelectric. Strained

piezoelectric crystals produce an electric displacement related to the strain,
so that the electro-optic and elasto-optic coefficients are defined while
holding the other effect constant. The variable that is held constant, either
E or S, is denoted by a superscript. The effect on a strained piezoelectric
crystal is,
(3.50)
where C:k/ is the elastic tensor element, ekliis the piezoelectric tensor element,
and Tj is the strain tensor element. Conversely, an applied electric field in
a properly cut piezoelectric crystal can generate shear or longitudinal waves
which are used for transducers.
The change in refractive index can be found by considering how the
impermeability matrix changes for either effect. Under the assumption that
the index change is small to the material’s initial index, the index change is
nips
Ans = --
2 ’
(3.51)
where p and r are chosen for a particular crystal orientation or application,

and p S and rE are typically less than lop4.
3.6.4.1. Electro-optic Effect. Electro-optic modulators use either the
linear Pockels electro-optic effect or the quadratic Kerr electro-optic effect,
and the light beam travels either parallel to (longitudinal mode) or perpen-
dicular to (transverse mode) the applied electric field. The retardation across
an electro-optic modulator is limited by the beam’s maximum incident
angle, electric field uniformity, scattered light, and refractive index variation.
These modulators work best in collimated light, and less than 5% retardation
variation across the aperture can be attained. The aperture should be checked
between crossed polarizers in the on and off state. Extinction ratios between
the on and off state are determined by the crossed polarizers extinction and
the modulator’s retardation uniformity; extinction values range from 100 : 1
to better than 1000: 1.
Some useful materials for Pockels cells are ammonium dihydrogen
phosphate (ADP), potassium dihydrogen phosphate (KDP), potassium
dideuterium phosphate (KD*P), lithium tantalate, lithium niobate,
hexamine, and lead lanthanum zirconate titanate (PLZT) ceramic.66-68Spec-
tral transmission, chemical stability, optical constants, and modulation
frequency determine the material’s application. The natural birefringence
can be either offset with a bias voltage or can be eliminated by using crossed
crystal pairs to cancel each other’s birefringence (this also reduces tem-
perature-birefringence effects). The voltage applied to produce half-wave
retardation is in the kilovolts range. The drive voltages for modulators
depend also on the path length through the crystal and on electrode
configuration. With many passes through a crystal or with cascaded crystal
pairs, the drive voltage is reduced.
High-frequency modulation is limited by capacitance, mechanical stress,
and heating; gigahertz frequencies can be achieved. Low-frequency modula-
tion on the order of a second or longer has the problem of current leaking
across the crystal. The charge needs to be replaced. This is accomplished
with a slightly conducting grease for contacting the electrodes to the crystal.69
Longitudinal Pockels cells have either transparent electrodes or ring
electrodes on the entrance and exit surfaces. The phase shift is
(3.52)
where I is the distance through the crystal, and E is the electric field. A
specification for longitudinal . modulators is the voltage for half-wave
retardation, VAI2, which is needed to produce maximum transmission
between crossed polarizers. Large apertures can be made.
Transverse Pockels cells have electrodes on the crystal’s sides, and the
light beam passes perpendicular to the applied field. Retardation increases
with an increased path length through the crystal and with a decreased
entrance aperture or electrode gap. Capacitance increases with increased
crystal length so that the frequency bandwidth decreases. The retardation is
(3.53)
where I is the crystal length, a is the electrode gap, and is the

half-wave field distance product defined for a particular crystal, its orienta-
tion, wavelength, and temperature.
The quadratic Kerr effect aligns the material’s molecular polarization
vectors by an applied electric field. The time to orient and induce birefrin-
gence with a high voltage and to disorient it again determines the operating
frequency for an electro-optic Kerr cell. The retardation is

S = 2.rrKE21,
K = Kerr constant (3.54)
I = length
Electro-optic Kerr cells use a variety of gases, or liquids, or PLZT. For
liquids and gases, the drive voltages are typically tens to hundreds of
kilovolts, and the switching times are nanoseconds. For PLZT, the drive
voltages are hundreds of volts and the switching speeds are microseconds.
Also, liquid and gas Kerr cells do not have any residual natural birefringence.
Subpicosecond laser pulses and high-power densities can be used to
induce birefringence in Kerr materials, This is the optical Kerr efect.
The optical Kerr cell can be either transverse or longitudinal; the type
determines the modulated beam’s direction with respect to the gating laser
pulse. The transverse cell has a long interaction path with the modulated
laser beam, and self-focusing problems can be induced for crystal lengths
greater than one or two centimeters thick. Usually, the optical Kerr cell is
a longitudinal cell with the gating laser pulse colinear with the light beam
to be modulated. The limiting switching speed is the time to randomize the
molecular polarization after a gating pulse has passed. For nitrobenzene,
the relaxation time is 32 ps and for carbon disulfide about 2 ps. The induced
birefringence due to the electric field E ( t ) from the laser is
6nll- Sn, = n28 I:,E 2 ( t ) exp[-(t - r’)]

dt’
-,
T
(3.55)
where n is the refractive index, n28 is the ac Kerr coefficient, Sn, and Snll
are the refractive index changes perpendicular and parallel to the E field,
T is the relaxation time, t is the gating pulse length, I?( t ) is the time average
over one period, and P ( t ) is the power density. For t > T, the birefringence
, n,,E2( t ) .
~ n ,-,~ n = (3.56)
The sample beam’s retardation is
I
6 = 27rn,,P( t ) ~ (3.57)
(Acne,) ‘
The on/off extinction ratio is typically about 1000: 1, determined mostly
by the extinction ratio of the entrance and exit polarizer, and the aperture
is the gating beam diameter.
3.6.4.2. Acousto-optic Effect. There are several ways to use the elasto-
optic effect depending on incident beam diameter and frequency modula-
t i ~ nA. uniaxial
~ ~ static strain deforms the isotropic media’s spherical phase
velocity surface into a spheroidal phase velocity surface with an optic axis
oriented along the strain direction with a birefringence,
Anli= - n o3- PS-

- SC. (3.58)
2
C = stress-optic constant
A useful wave plate can be made with a stressed plate and light traveling
perpendicular to the stress. The equation for phase induced by stress is,
2 IT A nl 2 TSCl
6=--- - (3.59)
A A ’
where 1 is the distance through the crystal.
For quartz, the stress needed to produce quarter-wave retardation at
5500 A is an order of magnitude less than needed to destroy the plate.
If the stress is a sound wave with a wavelength, A, much larger than the
beam diameter, the phase modulation varies sinusoidally with the stress.
s = s,,, e-i(kr-RO - s,,, COS(Rf) (3.60)
for A > than the beam diameter
6 = cos(Rr),
where R is the angular frequency, and k is the wave number for sound. If
the sound wavelength is comparable to the beam diameter, the phase
modulation or birefringence will vary across the aperture, and the intensity
modulation contrast between crossed polarizers will not be as large. This
limits the maximum frequency for resonating elasto-optic modulators to
less than 1 MHz.
Standing-wave photo-elastic modulators can drive in a self-resonating
mode and can be made with zinc selenide, calcium flouride, KRS-5 infrared
glass, germanium, and fused quartz. There are two types of construction,
either the bar type73or an isometric-contour element.74 The bar type is the
first design which is made of either one or two transducers connected to
either side of the optical element to be modulated. The transducers and the
optical element all resonate in a standing-wave mode with nodes at the
joints. The elements can be cemented together with silicone cement to reduce
strain birefringence. Bar-type modulators have two error sources which are
eliminated by the isometric-contour-type modulator. The errors are, ( 1 ) a
nonlinear, anharmonic dc birefringence proportional to the driving ampli-

tude squared (an error of about l % ) , and (2) the reflectance at the modu-
lator’s surfaces is modulated by the modulated refractive index adding a
small variation on the order of 0.001%.
The isometric-contour-type modulator is a rectangular optical element
supported at its corners along the length and modulated across the width.
The light beam travels the long direction through the bar’s center. The
usable entrance aperture is a little larger for these types. For photo-elastic
modulators in general, their entrance apertures can be large, and the
incidence angle can be as large as 50”.
Modulators can be driven by transducers at sound frequencies with many
sound wavelengths across the beam diameter.72s75 The compressions moving
at sound’s velocity will appear to the light beam as an almost stationary
phase grating with a grating spacing equal to the sound’s wavelength. If
light’s interaction with the sound waves is small, it is diffracted into orders,
N, with angle, + N , according to the thin phase grating equation,
NA
sin+N-sin+,,=-, N=0,*1,*2 ,.... (3.61)
A
The amplitude of the various orders is given by Bessel functions and the
maximum phase shifts, 6, in the medium due to the sound wave.
(3.62)
where I,, is the incident intensity. Changing the sound frequency changes
the orders’ angles, and changing the sound amplitude modulates the
intensity in other orders. These effects can be used in a feedback loop as
an active element for beam steering and modulating.
Since the phase grating is traveling, the diffracted light in the N t h order
will be frequency-shifted by an amount,
(3.63)
where V, is the material’s sound velocity, and & is the light frequency.
For high sound frequencies or for a long interaction path with light, the
behavior changes from a thin phase grating operating in the Raman- Nath
regime to the thick phase grating behavior similar to X-ray diffraction called
the Bragg regime. The light is diffracted into two orders, because the sound
wave is a sinusoidally varying density function needing one frequency term
to describe it, while for X-ray diffraction in particle distributions, higher
spatial harmonics are needed to describe the particles’ positions producing
a range of diffracted angles.
Light incident at the angle OB determined by the Bragg conditions will

be diffracted into a zero-order beam and either the plus or minus first-order
beam, depending on the incident angle.
A
sin Be = - (3.64)
2n0A’
where no is the material’s refractive index.
Modulators are usually Bragg-type modulators, and the on/off contrast
in the first-order beams are limited by scattering. Extinction ratios of 1000 : 1
are possible in the first-order beams. But, the modulator is not able to
diffract all of the energy out of the zero-order beam, because rediffraction
of higher orders and the sound wave’s refractive effects add light to the
zero-order beam. Contrast ratios of 10: 1 are typical. To improve the on/off
contrast, spatial or polarization filtering is used to separate the zero-order
beam. Both filtering methods use two modulators. Spatial filtering uses
aperture stops. The first modulator has an aperture placed far behind it to
isolate the zero order; the beam is then passed through a second modulator
and aperture combination like the first. Contrast ratios of about 100: 1 are
produced. Polarization filtering uses a shear-driven modulator which acts
as a half-wave retardation for the first-order beams. The zero-order beam
is selected by an entrance polarizer parallel to the transducer surface and
a parallel exit polarizer. To improve contrast, the beam is passed through
a second combination or reflected through the first combination. The con-
trast ratio is again 100 : 1 .76
The efficiency to drive an acousto-optic modulator is a measure of a
material’s usefulness as a modulator and its frequency b a n d ~ i d t h . ~The
’
stress retardation for an elasto-optic material can be rewritten by using a
relation between stress, velocity v, and mass density p ; this gives an
expression for the strain now replaced with an expression from the sound
power density P, of energy propagating at sound velocity under strain.
2 Ps1n:p2 ‘
’= -[ (pv3hA2)]
(3.65)
where M is the acousto-optic figure of merit and is related to the power

needed to produce a given retardation, h is the height perpendicular to the
light and sound waves. Other figures of merit take into account refinements
in the theory such as the bandwidth.
Acousto-optic modulators recently have been used for spectral filters

tuned by sound frequency and observing the first-order output as the sound
frequency is swept. Rapid scans, of usually less than a second, can cover
large spectral regions, and applications in the infrared are common.78
Another use for acousto-optic modulators is to impress light-phase informa-
tion on electrical signals controlling a modulator in a phase-locked loop to
measure without ambiguity light-phase shifts. Other uses are beam steering,
light-frequency shifting, and m o d ~ l a t o r s . ~
The
~*~useful
~ bandwidth is
usually hundreds of MHz, and the maximum useful frequency modulation
is limited by the number of sound waves across the beam diameter, usually
tens of MHz.
3.6.4.3. Magneto-optic Effect. Magneto-optic effects are the Faraday,
Kerr, and Cotton-Mouton effect; the Faraday effect is the most prominent
effect and most often used. A helical electric current coiling along a material
demonstrating the Faraday effect will rotate the plane of polarization along
the current direction due to the phase retardation between left and right
circular polarization. The retardation introduced is
S = VHI. (3.66)
V = Verdet constant
H =field strength
1 = light path length
This also means that multiple reflections will show an accumulation of

multiple retardations and rotations. Faraday rotating glasses are available,
too.
3.6.4.4. Depolarizers and Polarization Scramblers. Polarization can be
scrambled such that the detection system sees a net zero polarization created
by either a time average, a spectral average, or a spatial average across the
beam, or by all three effects. Incoherent multiple scattering randomizes the
polarization to produce an unpolarized beam.7 The Lambertian sphere, an
opal glass diffusing plate, powder-liquid mixtures, hot-pressed polycrystal-
line aluminum oxide or thorium oxide, cold-pressed magnesium oxide, and
cloudy fused quartz are examples. The powder-liquid mixtures can
effectivelydepolarize the light for very thin films. But, scattering depolarizers
destroy the image and usually strongly attenuate the beam.
The preservation of the image and intensity suggests using birefringence
to phase average incident polarization either in wavelength, across the
aperture, or in time.'' A birefringent plate with its optic axis oriented 45"
to the azimuth of the incident polarization produces a cyclically varying
retardation in wavelength increasing in frequency with plate thickness. A
154 OPT1C A L PO LA R I ZATIO N
thick plate with many retardation cycles in a small wavelength range

produces a fairly unpolarized beam for a system with poor spectral
resolution. To allow for a varying azimuth incident on a thick-plate
pseudodepolarizing element, a Lyot pseudodepolarizer is made of two plates
with a 2 : 1 thickness ratio and a 45" angle between their optic axes.82 The
45" angle is most critical to the depolarizer's performance, and the thickness
ratio only needs to be different from unity. Lyot depolarizers also limit
spectral resolution.
Birefringent crystal wedges can be made to introduce a variable phase
shift in one direction across the beam diameter for a particular wavelength
and is a monochromatic depolarizer if the optical system's final image does
not spatially resolve the phase variation. Two wedges with the optic axes
oriented 45" to e8ch other are independent of the azimuth of the incident
polarization. Double images and beam deviation occur for these depolarizer
types.
A varying driving function with a single optic modulator aligned with its
axes horizontal and vertical produces a polarization intensity modulation
expressed in four intensities, for example total intensity, horizontal linear,
+45" linear, and right circular polarization intensities modulated by the
driving function.
driving function (3.67)
Sb = so total intensity polarized + unpolarized
S : = S, cos S - S3 sin 6 horizontal linear polarization
s; = s, +45" linear polarization
S ; = S, sin 6 + S3 cos S circular polarization,
where S' and S are the transformed and incident polarization vectors. The
phase-modulated beam will always have one linear component unmodu-
lated. If the modulator is rotated 45", then the modulated exiting linear
polarization changes from horizontal to 45" linear.
s; = s,,
S ; = S, cos S - S, sin 6.
If the driving function is a sinusoidally varying function, then the intensity

phase modulation can be expressed with Bessel function expansions for
the sine and cosine intensity modulation.
S = S,,, sin wt, (3.68)

u)
sin 6 =sin(S,,, sin w t ) = 2 C sin[(2n + l)ot],

J2n+l(Smax)
n=O
m
cos 6 =cos(&,,, sin wt)=Jo(6,,,)+2 C JZn(Smax) sin(2nwt).
fl=l
The time averages for the sine terms are zero, and the remaining term is
J,( S,,,) = 0 for S,,, = 2.405, the first zero of J, . This means zero polarization
can be produced in either horizontal linear and circular polarizations, or
in 45" linear and circular polarizations, and can be used to calibrate instru-
mental polarization to an accuracy of 0.01 to o . o o ~ ~ ~ . ~ ~
If an optic modulator were driven by a sawtooth function such that the
phase modulation varied linearly from 0 to 2rr, the time-averaged output
light would have only a linear polarized component. If a second modulator
is added with its optic axis 45" to the first modulator and driven at an
integral multiple of the first frequency nw, then the resulting polarization
is zero.
Another method to produce depolarized light makes use of the time-
average effects for rotating retardation plates.83
s= plate retardation
st,= s, total intensity polarized + unpolarized
- 2( 1 + cos 6)Sl
S' -I horizontal linear polarization
S ; = $(1 + cos 6)S2 +45" linear polarization
s; = s3 cos 6 circular polarization
A rotating achromatic quarter-wave plate will pass only unpolarized and
linearly but not circularly polarized light. A rotating achromatic half-wave
plate will pass only unpolarized and circularly but not linearly polarized
light, and the circular polarization's handedness is reversed. A combination
of two wave plates rotating in opposite directions produces an element
passing only unpolarized light.
3.6.5. Liquid Crystals

Liquid crystals (LCD) are organic molecules in a state between liquid
and solid and show electromagnetic anisotropies. They are structured in
three basic geometries which are easily disrupted by electric fields, magnetic
fields, or temperature, depending on the design. The liquid crystals can be
chosen such that the net molecular polarization lines up parallel or perpen-
dicular to the crystal's long axis; this corresponds to positive or negative
-
birefringence, A n 0.2. For example, tunable birefringence can be produced
by treating the container's walls such that negative nematic crystals line up
parallel to the light beam in the off state. An electric field applied to the
walls will flip the liquid crystals perpendicular to the electric field and light
beam and produce a retardation proportional to the voltage up to a threshold
voltage. Or, the cell's interior wall can be treated to align the positive
nematic crystal structure perpendicular to the beam; this is done by rubbing
the glass to form parallel microgrooves and then treating the surface with
a high-energy surfactant. Then, a 90" twist is introduced between the entrance
and exit surface, rotating the beam 90". A voltage of 5-10 volts will com-
pletely disrupt the rotation and extinguish the beam between crossed
polarizers. Liquid crystals are low-current, low-voltage, high-contrast, thin-
display p o l a r i z e r ~ . ~ ~
3.6.6. Dichroic Polarizers

Linear dichroic molecules and crystals absorb polarization with different
absorption coefficients K, and K y . To reduce scattering effects, molecules
of small dimensions are used, usually needle-shaped iodine molecules. To
align the molecules, they are either applied to a burnished glass surface
with parallel microgrooves and allowed to settle, or are placed on a polyvinyl
alcohol or nitrocellulose base and stretched, which also introduces unwanted
birefringence. Better, controllable performance is obtained with the plastic-
sheet p ~ l a r i z e r . ~ ~
Extinction ratios and transmission vary with the concentration of dichroic
molecules; and concentration of molecules, standard densities, and extinc-
tion ratios are commercially available. Heavily dyed polarizers have a
parallel transmission of 12-14% and a crossed transmission of 0.001%. As
the dye is decreased, the short-wavelength extinction or optical density
decreases, thus decreasing the spectral region with high uniform optic
densities for crossed polarizers. Their parallel transmission is 30% or more.
To extend the plastic's usefulness into the infrared, the polyvinyl alcohol
sheet is catalytically dehydrated to polyvinylene. It is useful in the infrared
from about 8000 A-20,OOO A, with a large ttansmission of the unwanted
component at 9200 A. For near ultraviolet use, the polyvinyl alcohol/iodine
polarizer can be purified. Its range is down to 2000A. The thin Polaroid
sheets are usually cemented in glass or plastic laminates. King has reported
on dichroic polarizers and crystal prism polarizers. The dichroic polarizer
type has its place according to degree of p e r f o r m a n ~ e . ~ ~
SLITS, GRATINGS, A N D METAL GRID WLARIZERS 157
3.7. Slits, Gratings, and Metal Grid Polarizers
An ideal slit is an infinite conductor and is very thin. Approximate

examples are scratches on thin silver-coated mirrors. The light transmitted
by narrow slits of width d satisfying A / d > 2 are completely S polarized
and perpendicular to the slit. This is the electromagnetic theory region, and
the effect described is Hertzian polarization. Either finite conductivity or a
thick slit will decrease the slit’s transmission. When A / d < O S , then the
polarization effect for P or S polarizations is negligible. This is called the
scalar theory region described by infinite conductivity. For the region 2 >
h / d > 0.5, the transmitted polarization is predominantly P polarization due
to the slit’s finite conductivity and thicknesss6
Wire grids with a spacing of A / d > 0.5 are made to S polarize light.
Successful wire grids are either gratings with their ridges metal-coated at a
high-incidence angle or metal whiskers grown by a metal beam directed at
a substrate at a high-incidence angle.87Grid spacings as small as visible-light
wavelengths can be made for polarizers for wavelengths greater than 8000 A.
Extinction ratios are greater than 100: 1 and improve with wavelength. Also,
wire-grid polarizers can have a large acceptance angle and aperture.’
Diffraction gratings polarize the light according to the gratings material
and overcoat, the groove shape and spacing d, the wavelength A, and the
order N. As for slits, the P polarization is predicted fairly well with a scalar
theory down into the ultraviolet.** The S polarization perpendicular to the
slit is not as well behaved, and polarization anomalies in different orders
and wavelengths are more pronounced than for P polarizations. Energy
can be either diffracted into or out of the observed order, so the anomalies
can appear either bright or dark. The influence of a dielectric coating is to
broaden the anomaly and to shift it to longer wavelengths. For h / d < 0.2,
the polarization effects become negligible, and scalar theory applies to both
the P and S components. The grating’s efficiency in wavelength and diffrac-
ted angle is strongly influenced by anomalous behavior.
The grating anomalies are either Wood’s anomalies or resonance
a n ~ m a l i e s . *Wood’s
~ * ~ ~ anomalies occur in a grating order at a wavelength
which is at the same time being diffracted at *90° in another order, passing
onto or off the grating. Wood’s anomalies occur at Rayleigh’s wavelengths
given by the grating equation
nh = a(sin i * I ) , (3.69)
where n is the passing-off order.“
Resonance anomalies are leaky surface waves supported by the grating
and depending on the grating surface structure. A grating with shallow
grooves and a slightly modulated surface has little reactance and little
158 0 PTIC A L PO LA RI ZATlO N
anomalous effect. The resonance and Wood's anomalies are close together
for this type of surface. For a strongly modulated surface, the reactance
increases and the anomalies separate.
Blazed gratings in the Littrow configuration (the incident and. diffracted
beams have the same angle), with right triangle-shaped grooves and the
90" corner at the apex, can be described by the blaze angle only. First-order
diffraction is most strongly influenced by anomalies which decrease for
higher orders because A / d decreases. Higher-order anomalies and diffracted
angle deviations can be deduced in terms of the first-order Littrow configur-
atiom9*
First-order gratings with a blaze angle of less than 5" behave close to
scalar theory and show few anomalies. The P and S polarizations have a
peak in their efficiency at the blaze angle equal to 100%. For gratings blazed
between 5" to 18", Wood's anomalies at the Rayleigh wavelength for the
-1 and + 2 order at A/d = 2 / 3 , about 19.5", decrease in strength with the
blaze angle. The S grating efficiency is 100% at the blaze angle, and the P
efficiency peak is decreasing. For gratings blazed at 18" to 22", the anomalies
are suppressed because the Rayleigh pass-off anomaly occurs in the same
direction as the blaze angle. The S efficiency becomes more uniformly high
for diffraction angles greater than about 19". The P efficiency is a minimum
of about 80%. For blaze angles greater than 22", the S efficiency for
diffraction angles greater than 19" is high up to 60" and is 100% at the blaze
angle. The P efficiency peak is 100% at the Rayleigh wavelength A / d = 2 / 3 .
Gratings with an angular deviation between the incident and reflected beams
have lower P and S efficiencies, and the anomalies and peak efficiencies
shift in wavelength. Gratings with known P and S efficiencies can be used
as a polarizing element in an instrument.
3.8. Light Source and Detector Polarizations
Light sources can be polarized for a number of reasons. Lasers are usually
linearly polarized coherent emission sources. Incandescent and fluorescent
emissions are polarized because the emitted radiation is refracted at the
surface; and radiation refracted at a grazing angle to the surface is strongly
P polarized.93 Metal filaments, ribbons, and Nernst glowers have demon-
strated strong polarization effects for radiation from the highly angled
cylindrical edges. Also, deuterium lamps have been studied, which demon-
strate strong polarization effects for old lamps with a thin metal film
deposited on the glass envelope.94 New lamps are not as strongly polarized
across the light spot, but the polarization is still nonuniform away from the
central light spot, which is possibly due to polarized reflections from the
POLARIZATION DETERMINATION A N D MATHEMATICAL DESCRIPTION 159
envelope. The envelope’s polarization effects have been demonstrated for

tungsten lamps. Limiting light to the light spot reduces polarization
effect^.^^'^^ Depolarizers can be used to remove the polarization effects if
the effects are found to be significant enough.
Detectors are also polarization sensitive for various reasons which some-
times cannot be neglected. For example, photomultipliers or photocathodes
are sensitive to polarization due to their intrinsic characteristics. Also
detectors and arrays can have windows with varying static birefringence.
The polarization effects need to be considered for the particular application
and can be removed, if necessary, by either calibration or a depolarizer.
3.9. Pola rizat ion Deter mi nation and Mat hemat i ca I Descri ption
Polarization is either coherent and totally polarized or incoherent with

an unpolarized and a polarized component. In either case, the PoincarC
spherical surface description is applicable. An elliptical polarization is
marked on the sphere and transformed to a new polarization after passing
through an optical element. Coherent polarized light is known in terms of
phase shift between orthogonal polarizations and their intensities, for
example linear or circular polarizations. Incoherent polarized light is
described by Cartesian coordinates as the amount of linear polarization in
the x and +45” axes, and the amount of L or R polarization. This coordinate
system plus intensity is known as Stokes’ parameters.
Two- or four-dimensional transformation matrices describe the polariz-
ation vector and the optical element’s effects; and the cumulative effect is
a matrix multiplication of all the effects in their order, resulting in a single
transformation matrix describing the system. Coherent light can be written
as a two-dimensional vector multiplied by a complex phase term and
transformed with 2 x 2 matrices known as Jones’ matrices. Incoherent radi-
ation can be described in real numbered Stokes’ parameters and transformed
with 4 x 4 matrices known as Mueller’s matrices.
Light represented in terms of Jones’ matrices is the vector sum of two
linear orthogonal amplitudes E , and E,, and a phase shift 6. The fact that
they have the same frequency can be implied and removed from the
calculations. The modified form for an incident beam E and output beam
E’ is,
E = E x e x+ EyeY=
(3 ,
(3.70)
Clear isotropic materials, linear polarizers, retarders, reflections, and co-

ordinate rotations transform E to E’.
E‘=M*E
transparent isotropic medium in x - y
M = ( kx
0
0
k,
), 1 Eb
kx
k, E,
Ex )
partial linear polarizer in x - y
6 phase retardation in x - y
reflection in x - y
Ex cos a + E, sin a
-sina cosa -Ex sin (Y + E, cos a
rotation of an angle a counterclockwise looking into the beam
1 0 cos2 a cos a sin a
general orientation of linear polarizer
M =R(7Y)( ‘6“ cos2a e i r r I 4

e P * , 4 ) R ( a ) = ( j ( 2 ) ’ / 2cos a sin a
j ( 2 ) ” 2 cos a sin a
sin2a ePJrrI4
general orientation of quarter-wave plate (3.71)
Light represented in terms of Stokes’ vectors and Mueller’s matrices are
related to the Jones’ matrices. The Stokes’ vector has several labeling
schemes,
(I Q u V)
(I M C S)
(So s, s
2 SJ
POLARIZATION DETERMINATION A N D MATHEMATICAL DESCRIPTION 161
The following relations hold between the Stokes' vectors and the time-
averaged electromagnetic vectors,
I:=(E:)+(E:)+I~,
2 Q 2 + U 2 + V2,
0 = ( E 3 -W;L
U = 2( E,E,,) cos S = Re( I?$,,),
V = 2( ExE,,)sin 6 = Im( ExEy). (3.72)
The Stokes' vectors can also be written in terms of the light's polarization
azimuth a,ellipticity e, and the percent polarization P.
( Q 2 + U 2 + V2)'/2
P= 3
IT
a = 0.5 arctan( :),

e = 0.5 arcsin
Q = ITP cos(2e) sin(2a),

U = ITP cos(2e) cos(2a),
V = ITP sin(2e). (3.73)
The Mueller transformation matrices are similar to the Jones' matrices

and are determined by how they transform the incident light to output light.
1 1 0 0
0" linear polarization
-1 0 0
0 0 0
90" linear polarization
1 0 1 0
45” linear polarization
right circular polarization

1 0 0 - 1
-1 0 0
left circular polarization
0 0
0
rotation of an angle cy counterclockwise looking into the beam.
The 4 x 4 transformation matrices can also be determined experimentally
to take out an instrument’s polarization. A fairly complete list of matrices
is given by Shurcliff,2’ and recent articles go into details for error analysis
with Jones’ and Mueller’s matrices for polarization measurement^.^'-^^'
References
1. J. Strong, Concepts of Classical Optics, p. 110, W. H. Freeman, San Francisco, 1958.

2. M. Born and E. Wolf, Principles ofOprics, p. xxi, Pergamon Press, New York, 1964.
3. R. H. Muller, “Definitions and Conventions in Ellipsometry,” Surface Sci. 16, 14 (1969).
4. J. D. Jackson, Classical Electrodynamics, p. 269, Wiley, ,New York, 1975.
5. Ramachendran and Ramaseshan, Encyclopedia of Physics, Vol. 2 5 , Part 1, S, p. 1 (Flugge,
ed.), Springer-Verlag, 1961.
6. R. M. A. Azzam and N. M. Bashara, Ellipsometry and Polarized Light, p. 270, North-
Holland Publ. Amsterdam, 1977.
7. J. M. Bennett and H. E. Bennett, “Polarization,” in Handbook ofoptics, Chap. 10, ( W .
G. Driscoll and W. Vaughan, eds.), McGraw-Hill, New York, 1972.
REFERENCES 163
8. R. M. A. Azzam, “Consequences of Light Reflection at the Interface Between Two

Transparent Media Such that the Angle of Refraction is 45”;’ J. Opt. SOC.Am. 68, 1613
(1978).
9. R. M. A. Azzam and N. M. Bashara, Ellipsometry and Polarized Light, p. 288, North-
Holland Publ., Amsterdam, 1977.
10. R. M. A. Azzam, “Direct Relation Between Fresnel’s Interface Reflection Coefficients
for the Parallel and Perpendicular Polarizations,” J. Opt. SOC.Am. 69, 1007 (1979).
11. R. M. A. Azzam, “Transformation of Fresnel’s Interface Reflection and Transmission
Coefficients Between Normal and Oblique Incidence,” J. Opt SOC.Am. 69, 590 (1979).
12. R. M. A. Azzam, “Mapping of Fresnel’s Interface Reflection Coefficients Between Normal
and Oblique Incidence: Results for the Parallel and Perpendicular Polarizations at Several
Angles of Incidence,” Appl. Opt. 19, 3361 (1980).
13. M. V. R. K. Murty and R. P. Shukla, “Extinction of Light in Brewster Polarizers: a
Shadowlike Phenomenon,” Appl. Opt. 22, 1094 (1983).
14. G. Hass and W. R. Hunter, “Reflection Polarizers for the Vacuum Ultraviolet Using
Al+MgF2 Mirrors and an MgF2 Plate,” Appl. Opt. 17, 76 (1978).
15. H. A. Van Hoof, “Imaging Polarizer for the VUV with Spherical Mirrors,” Appl. Opt.
19, 189 (1980).
16. G. Hass, “Reflectance and Preparation of Front-surface Mirrors for use at Various Angles
of Incidence from the Ultraviolet to the Far Infrared,” J. Opt. SOC.Am. 72, 27 (1980).
17. J. 0. Stenflo, H. Biverot, and L. Stenmark, “Ultraviolet Polarimeter to Record Resonance-
line Polarization in the Solar Spectrum Around 130-150 pm,” Appl. Opt. 15, 1188 (1976).
18. J . T. Cox and G. Hass, “Highly Efficient Reflection-type Polarizers for 10.6 p m CO, laser
Radiation using Aluminum Oxide Coated Aluminum Mirrors,” Appl. Opt. 17,1657 (1978).
19. W. R. Hunter, “Design Criteria for Reflection Polarizers and Analyzers in the Vacuum
Ultraviolet,” Appl. Opt. 17, 1259 (1978).
20. D. T. Rampton and P. W. Grow, “Economic Infrared Polarizer Utilizing Interference
Effects in Films of Polyethylene Kitchen Wrap,” Appl. Opt. 15, 1034 (1976).
21. W. W. Shurcliff, Polarized Lighf,p. 78, Harvard University Press, Cambridge, Mass., 1964.
22. T. J. McIlrath, “Circular Polarizer for Lyman-Alpha Flux,” J. Opt. SOC.Am. 58,506 (1968).
23. R. M. A. Azzam and N. M. Bashara, EIlipsornefry and Polarized Light, p. 283, North-
Holland Publ., Amsterdam, 1977.
24. M. Ruiz-Urbieta, E. M. Sparrow, and E. R. G. Eckert, “Thickness and Optical Constants
of Films,” J. Opt. SOC.Am. 61, 351 (1971).
25. R. M. A. Azzam, A.-R. M. Zaghloul, and B. M. Bashara, “Ellipsometric Function of a
Film-substrate System: Applications to the Design of Reflection-type Optical Devices
and to Ellipsometry,” J. Opt. SOC.Am. 65 252 (1975).
26. R. M. A. Azzam, A.-R. M. Zaghloul, and B. M. Bashara, “Design of Film-substrate
Single-reflection Linear Partial Polarizers,” J. Opt. SOC.Am. 65, 1472 (1975).
27. A.-R. M. Zaghloul, R. M. A. Azzam, and N. M. Bashara, “Designs of Film-substrate
Single-reflection Retarders,” J. Opt. SOC.Am. 65, 1043 (1975).
28. S. Kawabata and M. Suzuki, “MgF,-Ag Tunable Reflection Retarder,” Appl. Opt. 19,
484 (1980).
29. M. Ruiz-Urbieta and E. M. Sparrow, “Reflection Polarization by a Transparent Film-
absorbing Substrate System,” J. Opt. SOC.Am. 62, 1188 (1972).
30. M. Ruiz-Urbieta and E. M. Sparrow, “Wavelength Bandwidth and Other Design Aspects
for Film-substrate Reflection Polarizers,” Appl. Opt. 12, 590 (1973).
31. M. Ruiz-Urbieta and E. M. Sparrow, “Effect of Film Absorption on a Film-substrate
Reflection Polarizer,” J. Opt. SOC.Am. 63, 194 (1973).
32. N. Born and E. Wolf, cf. Ref. 2, Chap. 1.
33. A. Herpin, “Optique E1ectromagnttique.-Calcul du Pouvoir Dttlecteur d’un Systtme

Stratifit Quelconque,” Comptes Rend. 225, 182 (1947).
34. F. Abeles, “Thesis,” Ann. Phys. (Paris) 5, 596 and 706 (1950).
35. L. 1. Epstein, “Design of Optical Filters,” J. Opt. SOC.Am. 42, 806 (1952).
36. W. W. Buchmann, S. J. Holmes, and F. J. Woodberry, “Single-Wavelength Thin-Film
Polarizers,” 1. Opt. SOC.A m . 61, 1604 (1971).
37. A. Thelen, “Nonpolarizing Interference Films Inside a Glass Cube,” Appl. Opt. 15, 2983
(1976).
38. V. R. Costitch, “Reduction of Polarization Effects in Interference Coatings,” Appl. Opt.
9, 866 (1970).
39. 0. Arnon and P. Baumeister, “Electric Field Distribution and the Reduction of Laser
Damage in Multilayers,” Appl. Opt. 19, 1853 (1980).
40. A. Thelen, Physics of Thin Films, Vol. 5 , p. 47 (G. Hass and R. Thun, eds.), Academic
Press, New York, 1969.
41. H. M. Liddell, Computer Aided Techniques f o r the Design of Multilayer Filters, Adam
Hilger, Bristol, 1981.
42. H. F. Mahlein, “Properties of Laser Mirrors at Non-normal Incidence,” Opt. Acta 20,
687 (1973).
43. H. F. Mahlein, “Generalized Brewster-Angle Conditions for Quarter-Wave Multilayers
at Non-Normal Incidence,” J. O p t SOC.Am. 64, 647 (1974).
44. W. H.Southwell, “Multilayer Coating Design Achieving a Broadband 90” Phase Shift,”
Appl. Opt. 19, 2688 (1980).
45. M. Banning, “Practical Methods of Making and Using Multilayer Filters,” J. Opt. SOC.
Am. 37, 792 (1947).
46. P. B. Clapham, M. Downs, and R. King, “Some Applications of Thin Films to Polarization
Devices,” Appl. Opt. 8, 1965 (1969).
47. J. A. Dobrowolski and A. Waldorf, “High Performance Thin Film Polarizer for the UV
and Visible Spectral Regions,” Appl. O p t 20, 111 (1981).
48. R. J. King, “Quarter-Wave Retardation System Based on the Fresnel Rhomb Principle,”
J. Sci. Instr. 43, 617 (1966).
49. R. W. Anderson, “Polarization Conserving Light Bending Prisms and Optimized Fresnel
Rhombs,” Appl. Opt. 13, 1110 (1974).
50. J. M. Bennett, “A Critical Evaluation of Rhomb-Type Quarter-wave Retarders,” Appl.
Opt. 9, 2123 (1970).
51. M. Born, and E. Wolf, c$ Ref. 2, Chap. 14.
52. D. L. Steinmetz, W. G. Phillips, M. Wirick, and F. F. Forbes, “A Polarizer for the VUV,”
Appl. Opt. 6, 1001 (1967).
53. E. 0. Ammann and G. A. Massey, “Modified Forms for Clan Thompson and Rochon
Prisms,” J. Opt. SOC.A m . 58, 1427 (1968).
54. J. F. Archard, “Performance and Testing of Polarizing Prisms,” J. Scr. Instr. 26, 188
(1949).
55. R. J. King and S. P. Talim, “Some Aspects of Polarizer Performance,” J. Phys. E 4, 93
(1971).
56. W. C. Davis, “Simplified Wave Plate Measurements,” Appl. Opt. 20, 2879 (1981).
57. E. L. Gieszelman, S. F. Jacobs, and H. E. Morrow, “Simple Quartz Birefringent Quarter-
Wave Plate for Use at 3.39km,” J. Opt. SOC.A m . 59, 1381 (1969).
58. A. M. Title and W. J. Rosenberg, “Achromatic Retardation Plates,” in Polarizers and
Applications ( G . B. Trapani, ed.), Proc. SOC.Phot. Opt. Inst. Eng. 307, 120 (1981).
59. A. M. Title, “Improvement of Birefringent Filters 2: Achromatic Waveplates,” Appl. Opt.
14, 229 (1975).
60. J. M. Beckers, “Achromatic Linear Retarders,” Appl. Opt. 10, 973 (1971).
REFERENCES 165
61. S. Chu, R. Conti, P. Bucksbaum, and E. Commins, “Making Mica Retardation Plates:
A Simple Technique,” Appl. Opt. 18, 1138 (1979).
62. J. Strong, Procedures in Experimental Physics, p. 388, Prentice-Hall, New York, 1952.
63. W. C. Davis, “Simplified Wave Plate Measurements,” Appl. Opt. 20, 2879 (1981).
64. E. Hartfield and B. J. Thompson, “Optical Modulators,” in Handbook ofOptics, Chap.
17 (W. G. Driscoll and W. Voughan eds.), McGraw-Hill, New York, 1972.
65. E. K. Sittig, “Elastooptic Light Modulation and Deflection,” in Progress in Optics, Vol.
10, Chap. 6, p. 231 (E. Wolf, ed.), North-Holland Publ., Amsterdam, 1972.
66. C. Forno and 0. C. Jones, “Hexamine Electro-Optic Light Modulators,” J. Phys. E 7,
101 (1971).
67. B. P. Blumich and R. Germer, “Electro-Optic Shutters with PLZT Ceramics for Infrared
Applications,” J. Phys. E 12, 770 (1979).
68. E. R. Kocher and R. P. Novak, “Fast Electro-Optical Shutter with Programable Trans-
mission Control,” 1. Phys. E 13, 542 (1980).
69. T. G. Baur, D. E. Elmore, R. H. Lee, C. W. Querfeld, and S. R. Rogers, “Stokes 11. A
New Polarimeter for Solar Observation,” Solar Physics 70, 395 (1981).
70. J. Etchepare, G. Grillon, R. Muller, and A. Orszag, “Kinetics of Optical Kerr Effect
Induced by Picosecond Laser Pulses,” Opt. Comm. 34, 269 (1980).
71. M. A. Duguay, “The Ultrafast Optical Kerr Shutter,” in Progress in Optics, Vol. 14, Chap.
4, p. 163 (E. Wolf, ed.), North-Holland Publ., Amsterdam, 1976.
72. A. Korpel, Applied Optics and Optical Engineering, Vol. 6, p. 89 (R. Kingslake and B. J.
Thompson, eds.), Academic Press, New York, 1980.
73. J. C. Cheng, L. A. Nafie, S. D. Allen, and A. I. Braunstein, “Photoelastic Modulator for
the 0.55-13 pm Range,” Appl. Opt. 15, 1960 (1976).
74. J. C. Kemp, “Photoelastic-Modulator Polarimeters in Astronomy,” in Polarizers and
Applications (G. B. Trapani, ed.), Proc. Soc. Phot. Opt. Inst. Eng. 307, 80 (1981).
75. A. Korpel, Electro Optics-Principles and Applications (B. Thompson and J. B. DeVelis,
eds.), Proc. SOC.Phof. Opt. Inst. Eng. 38, 3 (1973).
76. G. B. Brandt, M. Gottlieb, and J. J. Conroy, “Improved Contrast for Acousto-Optical
Modulators Using Spatial Filtering,” J. Opt. Soc. A m . 67, 1269 (1977).
77. A. Korpel, “Acousto-Optics,” in Applied Oprics and Optical Engineering, Vol. 6, Chap. 4
(R. Kingslake and B. J. Thompson, eds.), Academic Press, New York, 1980.
78. 1. C. Chang, “Acousto-Optic Tunable Filters,” Opt. Eng. 20, 824 (1981). ,
79. J. B. Houston, Jr., M. Gottlieb, Shi-Kay Yao, I. C. Chang, J. Tracy, L. M. Smithline, and
G. J. Wolga, “The Potential for Acousto-Optics in Instrumentation: An Overview for the
1980s.” Opt. Eng. 20,712 (1981).
80. M. Gottlieb and G. W. Roland, “Infrared Acousto-Optic Materials: Applications,
Requirements and Crystal Development,” Opt. Eng. 19,901 (1980).
81. Shan-Ling Lu and A. P. Loeber, “Depolarization of White Light b y a Birefringent Crystal,”
J. Opt. SOC.A m . 65, 248 (1975).
82. A. P. Loeber, “Depolarization of White Light by a Birefringent Crystal. 11. The Lyot
Depolarizer,” J. Opt. SOC.A m . 72,650 (1982).
83. B. H . Billings, “Monochromatic Depolarizer,” J. Opt. SOC.A m . 41,966 (1951).
84. R. A. Saref, in Electro Optics-Principles and Applications (B. J. Thompson and J. B.
DeVelis, eds.), Proc. SOC. Phof. Opt. Inst. Eng. 38,23 (1973).
85. E. H. Land, “Some Aspects of the Development of Sheet Polarizers,” J. Opt. SOC.A m .
41, 957 (1951).
86. R. V. Jones and J. C. S. Richards, “The Polarization of Light by Narrow Slits,” Proc.
Roy. SOC.(London), A225, 122 (1954).
87. R. E. Slocum, “Evaporative Thin Metal Films as Polarizers,” in Polarizers and Applicafions
(G. B. Trapani ed.), Proc. SOC.Phot. Opt. Inst. Eng. 307,25 (1981).
88. A. J . Caruso, G. H. Mount, and B. E. Woodgate, “Absolute S- and P-Plane Polarization

Efficiencies for High Frequency Holographic Gratings in the VUV,” Appl. Opt. 20, 1764
(1981 ).
89. C. H . Palmer, “Parallel Diffraction Grating Anomalies,” J. Opt. SOC.Am. 42, 269 (1952).
90. C. H. Palmer and F. W. Phelps, Jr., “Grating Anomalies as a Local Phenomenon,” J. Opt.
SOC. Am. 58, 1184 (1968).
91. A. Hessel and A. A. Oliner, “A New Theory of Wood’s Anomalies on Optical Gratings,”
Appl. Opt. 4, 1275 (1965).
92. E. G. Loewen, M. Neviere, and D. Maystre, “Grating Efficiency Theory as it Applies to
Blazed and Holographic Gratings,” Appl. Opt. 16, 271 1 (1977).
93. 0. Sandus, “A Review of Emission Polarization,” Appl. Opt. 4, 1634 (1965).
94. H . P. Lengkeek and P. Winsemus, “Polarization of Light Sources: a Photographic
Method,” Appl. Opt. 14, 2766 (1975).
95. R. K. Kostuk, “Polarization Characteristics of DXW-Type Filament Lamp,” Appl. Opt.
19, 2274 (1980).
96. D. L. Spooner, “Polarization Pattern of a High Intensity Incandescent Lamp,” Appl. Opt.
11, 2984 (1972).
97. M. A. Bouchiat and L. Pottier, “A High-Purity Circular Polarization Modulator: Applica-
tion to Birefringence and Circular Dichroism Measurements on Multidielectric Mirrors,”
Opt. Comm. 37, 229 (1981).
98. H. G. Berry, G . Gabrielse, and A. E. Livingston, “Measurement of the Stokes Parameters
of Light,’’ Appl. Opt. 16, 3200 (1977).
99. R. M. A. Azzam, “Division of Amplitude Photopolarimeter (DOAP) for the Simultaneous
Measurement of all Four Stokes Parameters of Light,” Opt. Acta 29, 685 (1982).
100. V. M. Bermudez and V. H. Ritz, “Wavelength-Scanning Polarization Modulation Ellip-
sometry: Some Practical Considerations,” Appl. Opt. 17, 542 (1978).
101. P. S. Hauge, ”Techniques of Measurement of the Polarization-Altering Properties of
Linear Optical Systems,” in Optical Polarimetry (R. M. A. Azzam and D. L. Coffeen,
eds.), Roc. SOC.Phot. Opr. Inst. Eng. 112, 2 (1977).
4. HOLOGRAPHY
R. D. Bahuguna*
Rose-Hulman Institute of Technology
5500 Wabash Avenue
Terre Haute, Indiana 47803
D. Malacara
Centro de lnvestigaciones en Optica. A X
Apartado Postal No. 948
37000 Leon, Gto. Mexico
4.1. Introduction
The invention of the laser has had a great impact on the science of optics.
In particular, the high coherence of the laser light has completely revitalized
those methods of optics which use interference as the basic phenomenon.
Among these is the method of wave-front reconstruction. Also called
holography, this method was invented by Denis Gabor in 1948'*2.3 to improve
the resolution in the image obtained with the electron microscope. Although
the initial goal did not prove successful, holography in itself became an
important subject of investigation. However, due to the lack of a coherent
light source, it remained in obscurity for more than a decade.
In the 1960s, with the availability of the laser, a major advance was made
at the University of Michigan Institute of Science and Technology. Leith
and Upatnieks in 1962495*6 applied the laser light to holography producing
excellent three-dimensional images which astonished all those who saw
them. In the same year, Denisyuk' made a remarkable contribution to
holography by combining it with the volume-recording process, invented
by Lippmann' in 1891, to make color photographs. Using a thick emulsion
to record the interference fringes, a method was produced for color hologra-
phy in which reconstruction could be carried out with a point source of
white light. His hologram is, in effect, a hologram and a narrow-band
spectral filter, combined into a single structure.
*Work was done while the author was at Centro de lnvestigaciones en Optica, A.C.,
Apartado Postal No. 948, 37000 Leon, Gto. Mexico.
167
METHODS OF t X P E R I M t N T A L PHYSICS Copyright 01988 by Academic Press. Inc.
Vol ?b All rights of reproduction in an) form reserred.
I S B N 0-12-475971-8
168 HOLOGRAPHY
Stephen Bentong of the Polaroid Corporation made another significant

contribution in display holography. The method, known as rainbow hologra-
phy or Benton holography, makes it possible to view the hologram with an
extended white light source. Cross” combined the rainbow holograms with
a technique known as composite or multiplex holography to produce cylin-
drical holograms.
Holography, although applicable to all waves, i.e., optical waves, micro-
waves, X-rays, electon waves, seismic waves, and acoustic waves, has
progressed exceptionally fast in the first category. This is essentially because
of the availability of a highly coherent source: the laser. Research continued
throughout the sixties and seventies. To date, hundreds of scientific papers
and many books like those by Stroke,’’ Smith,” Collier et al.” Francon,14
Caulfield,” Abramson,I6 and Hariharan” have been published about
holography and its applications to various other areas.
Holography is a relatively new process that is similar to photography in
some respects, but is nonetheless fundamentally different. Some of the basic
differences are the following:
(a) Photography records the image of an object formed by the lens of a
camera. The image is recorded as two-dimensional and contains information
about the object as viewed from a single particular angle. In holography,
the object wave front is reconstructed, giving a three-dimensional view. It
is as if one is viewing the object itself, possessing depth and parallax
properties.
(b) The photograph is in itself a direct record of the scene and can be
viewed by focussing one’s eyes onto its flat surface. The hologram is the
interference record of the object wave with a reference phase-related wave.
To reconstruct the image, the record is illuminated by the reference beam.
The image is reconstructed at the same position as the object (or its image)
during recording, and thus, not necessarily on the holographic plate itself.
(c) Photography can be done in ordinary light, whereas holography
requires some amount of coherence.
(d) In photography the intensity is a quantity averaged over all the phases
of the light wave, and so the phase information about the object is lost. On
the other hand, in holography the amplitude and phase are encoded with
the help of the reference wave and can be decoded in the reconstruction step.
4.2. Theory of Holography
The of-axis holography4 is the best-known form of holography and is,

conceptually, the simplest to treat from the theoretical point of view. In
this section, we describe the theory of this method for plane holograms and
THEORY OF HOLOGRAPHY 169
later see the effect of thick emulsions. Finally, we discuss the nonlinear
effects and aberrations involved in the process.
4.2.1. Basic Theory of Plane Holograms

By a plane or thin hologram we mean one in which the recording material
is thin compared to the spacing of the highest spatial frequency of the
exposure variations, i.e., the recording medium is essentially two-
dimensional.
Consider the arrangement shown in Fig. 1 , where 0 is the object wavefront
and R is the reference field. They both interfere on the photographic
emulsion H. The total field on H is O + R.
The irradiance I ( x ) on the photographic plate is given by:
I ( x ) = ( O + R ) ( O +R ) *
= I 0l2+ IR 1' + OR * + O* R. (4.1)
Assuming that we are in the linear region of the amplitude transmittance
f ( x ) versus the exposure E ( x ) curve, we have:
t(x)=to+BE(x), (4.2)
where to is a constant background transmittance and B is a parameter
determined by the emulsion and the processing conditions. Since E ( x ) for
a given exposure time is directly proportional to Z(x), we may write using
eqns. (4.1) and (4.2):
t(x)= t,+B'(1012+lR12+OR*+O*R) (4.3)
where B'= B T ; t is the exposure time.
FIG. 1. Basic arrangement when recording a hologram.

170 HOLOGRAPHY
The developed plate which is now a hologram is illuminated by the

reference wave R . The transmitted field X ( x ) at the hologram plane is then
given by:
X ( x )= R ( x ) t ( x )
= R~,,+B’(R~O~*+RIR~~+ I R ~*o+R*o*) . (4.4)
Now, if the reference beam is sufficiently uniform so that IRI2 can be
treated as constant, then the fourth term of Eq. (4.4)is merely a constant
times the object wave, showing that the object is reconstructed as such. The
hologram is then like a window through which a virtual image of the object
is seen. The last term represents the conjugate of the object wave times R’.
(C 1 (d)
FIG. 2. Method to determine the conjugate image: Join the reference point source R and
the object point 0, by a straight line. Draw a perpendicular R P to the plate H through R.
Measure the angle RPO = a say. Draw a line passing through P at an angle ( - a ) and let it
meet O , R at 0;. 0;is then the conjugate image. ( a ) , (b) and (c) are different possible
recording geometries; (d) is similar to ( a ) except that one is viewing the image in reflection.
Reflection images of 0, and 0:are seen as 0,and 0;respectively.
T H E O R Y OF HOLOGRAPHY 171
The conjugate image might be a real or a virtual image, depending on the

particular geometry used to form the hologram. When the image is real, it
is spatially located in front of the hologram. A simple method to determine
the type of the conjugate image and its position is described in the book
by AbramsonI6 and is shown in Fig. 2.
The paths of the light in a hologram are reversible. Thus, if the reconstruct-
ing beam reaches the hologram from the opposite side and in an antiparallel
direction to the original reference beam, the two images are formed at the
same places. The only difference is then that a virtual image becomes real,
and vice versa.
If, when making a hologram, the object is real on the same side as the
collimated reference beam, we have the following property. The real images
produced by holograms, unlike those produced by lenses, are pseudoscopic,
that is, they are “inside out.” Concave surfaces appear as convex, and vice
versa. However, an orthoscopic image may be obtained if the pseudoscopic
real image of another hologram is used as the object when taking the
hologram. Figure 3 shows two alternative ways to reproduce these holo-
graphic images.
FIG. 3. Basic arrangement to reproduce the images of a hologram: (a) recording set up;
(b) reconstruction with a beam identical to the reference beam; (c) reconstruction with a beam
antiparallel to the reference beam.
172 HOLOGRAPHY
4.2.2. Nonlinear Effects and Aberrations

The effect of film nonlinearities on holograms has been studied by many
authors, like Friezem and Zelenka,'' Goodman and Knight," Bryngdahl
and Lohmann,20 and Kozma."
When the visibility of the fringes is high, the variations on the exposure
E ( x ) are large, and hence the linear approximation in Eq. (4.2) might not
be valid. Then, in general:
t ( x ) = t o + B l E ( x ) + B 2 E 2 ( x ) + B 3 E 3 ( ...
x) . (4.5)
The third- and higher-order terms produce higher-order images. The
contrast of the object is reduced and some ghost images appear. If the
object is a set of discrete points, false images may be produced.
This analysis explains nonlinear effects in flat transmission holograms,
but not in volume or phase holograms. Besides film nonlinearities, intrinsic-
phase nonlinear effects also appear in phase holograms.
The images of a hologram may be reproduced by using any of the two
basic configurations described in the former section. However, the images
may also be obtained after changing one or more of the following para-
meters: ( a ) position of the reconstructing light source, (b) wavelength, ( c )
size of the hologram, etc. The changing of these parameters in general
affects the position of the images, the type of these images, and their size.
However, it is very important to point out that these changes also affect the
quality of the images, because some aberrations are also introduced, as
pointed out by Meier.22
4.2.3. Effects of Thickness: The Bragg Effect

If the photographic emulsion is somewhat thicker than the period of the
interference fringes, the hologram is called a volume hologram. This means
that the interference pattern extends not only across the surface of the
emulsion but, since the emulsion is quite transparent, throughout its depth
as well. A volume hologram has a number of properties not shared by the
plane hologram.
Volume holograms can be made to produce an image either by light
transmitted through them (transmission holograms) or b y light reflected
from their surfaces (reflection holograms).
We can schematically represent the formation of a volume-transmission
hologram, as in Fig. 4, where R and 0 are the reference and object plane
waves. (The results of plane waves can be generalized for any object, as
the object wave can always be decomposed into plane waves). The waves
R and 0 interfere with an angle 4 between them and 4 / 2 with respect to
the normal to produce a set of parallel fringes perpendicular to the emulsion.
T H E O R Y O F HOLOGRAPHY 173
3 Emulsion
FIG. 4. Arrangement to form a volume-transmission hologram.
The spacing is easily seen to be given by:

A
d= (4.6)
(2 sin(4/2))'
Let us now illuminate the hologram at an angle 8, which is, in general,
not equal to 4/2 at recording. In order to obtain an image reconstructed
from the plane object wave, it is necessary that all the rays add in phase
after reflection from the semireflecting planes. The following relation must
then be satisfied:
+A
sin 0 =-, (4.7)
2d
which is the familar Bragg condition.
On comparing Eqs. (4.6) and (4.7), we have the following four solutions:
and
e = * ( n - 4/21.
We will consider each case separately as follows:
(a) When the illuminating beam is identical to the reference beam ( 0 =
4/2), the object wave is reconstructed, and the virtual image formed. No
conjugate image is observed.
(b) When the illuminating beam is in the opposite direction to the initial
reference wave 6 = - ( n - 4 / 2 ) , a real image is reconstructed. This real
image is pseudoscopic, and no virtual image is formed.
174 HOLOGRAPHY
(c) When the illuminating beam is in the direction of the object beam
( 0 = - 4 / 2 ) , only an imperfect real image, with aberrations, is formed. The
complete object cannot be seen, becasue the Bragg condition is not satisfied
for all waves from the object.
(d) When the illuminating beam is in an opposite direction to the object
wave ( 0 = T - 4 / 2 ) , a partial, imperfect virtual image, with aberrations, is
reconstructed. The complete image is not observed for the same reason
given before.
Summarizing, we can say that in thick holograms only one image is
reconstructed, at,the same position where the object was when the hologram
was formed. If the reconstructing beam is parallel to the reference beam,
the image is orthoscopic. However, if the reconstructing beam is antiparallel
to the reference beam, the image is pseudoscopic.
So far, we have discussed the transmission volume holograms. The reflec-
tion type can be made as shown in Fig. 5 . The object wave 0 and the
reference wave R come from opposite directions. The interference fringes
thus recorded are approximately planes parallel to the emulsion. The dis-
tance between the planes is h / 2 . If such a hologram is illuminated with a
beam identical to the reference beam, it will reconstruct the virtual image
of the object. The reconstruction is viewed in reflection rather than trans-
mission.
The major feature of the refection holograms is that a single-color wave
front may be reconstructed with white-light illumination. The Bragg planes
act as an interference filter and have a high reflectivity only for those
wavelengths that produced them. However, in practice, such a hologram
recorded at one wavelength may reconstruct in a slightly shorter wavelength.
Object wave 0
i-
' Reference wave R
F I G . 5. Arrangement to form a volume-reflection hologram.

DIFFERENT TYPES O F HOLOGRAMS 175
For example, the recording in red color may reconstruct in green. This is
because of emulsion shrinkage during photographic processing. The Bragg
planes pull together so that the final spacing is smaller than the one at
recording, leading to reconstruction at a different wavelength. It is worth
pointing out that white-light illumination results in a loss of image resolution,
since the hologram acts as a spectral filter with a bandwidth of about 100 A,
which is too large for high-resolution imagery.
4.3. Different Types of Holograms
There are a number of hologram forming configurations depending upon

the position and form of the object and the reference beams. The simplest
one, invented by Gobor, is the in-line hologram. In this method, the object
is a transparency illuminated by a reference beam. The scattered light
contains information about the object, and the unscattered light acts as a
reference wave. The source is the green line of the mercury arc lamp,
Although successful in reconstructing the transparency, this hologram
suffered from a serious drawback: the twin-image problem. The primary
image and the conjugate image are in the same direction and so are
inseparable. The problem was overcome successfully by Leith and
Upatnieks4 using their off-axis configuration, and later by Bryngdahl and
Lohmann” and by Meier24using single-side band holography.
In the following section, we describe Fresnel, Fraunhofer, Fourier trans-
form, image-plane, color, composite holographic stereogram, rainbow, and
computer (synthetic) holograms.
4.3.1. Fresnel Holograms

When the recording plate is placed in the Fresnel diffraction region of
the object and at an arbitrary distance from the reference source, we obtain
a Fresnel hologram. A practical arrangement is shown in Fig. 6 . It is the
simplest and most convenient configuration, requires no lenses either in
formation or in reconstruction, and produces images of three-dimensional
objects as shown in Fig. 7.
4.3.2. Fraunhofer Holograms

A Fraunhofer hologram can be defined as one which records interference
of the far-field diffraction pattern of the object, with a reference light source
not coplanar with the object.
176 HOLOGRAPHY
L S
FIG.6. Recording a Fresnel Hologram.
FIG.7. Reconstructed image from a Fresnel Hologram.

\
d
-
.
$
-4
H
0
FIG.8. Recording of a Fraunhofer hologram with a free space configuration.
The following are the two geometries with which to record such a
hologram.
(1) This configuration is a free-space geometry (see Fig. 8). The subject
(a small transparency) is illuminated by a broad plane-parallel beam. The
recording plate is kept in the far field of the subject. The light scattered
from the subject interferes with the uniform background to form the
hologram.
This geometry was utilized by Thompson et aLZ5for measuring the size
and shape of dynamic aerosol particles. The twin-image problem in this
geometry is apparently removed.
(2) In this configuration (see Fig. 9), the subject is placed close to the
lens and is illuminated by a plane-parallel beam. The recording plate is
FIG. 9. Recording of a Fraunhofer hologram with a lens.

178 HOLOGRAPHY
kept at the back focal plane of the lens, and the reference wave is incident
obliquely from the side.
4.3.3. Fourier Transform Holograms

A Fourier transform hologram records the interference pattern of the
Fourier transform fields of both the object and the reference source. A
practical arrangement is shown in Fig. 10.
Such holograms are useful as spatial filters for pattern recognition, where
the properties of the Fourier transform provide the basis for the recognition
process, as described by Vander Lugt.26
It is worth noting that when the Fourier transform hologram is formed
with a plane-reference wave, the generated image remains stationary with
respect to the translation of the hologram. For example, a moving reel of
such holograms would project stationary images.
By a variation of the above geometry, one obtains the so-called quasi-
Fourier transform h01ogram.l~ In this case, the subject is adjacent to the
Fourier transforming lens with a coplanar reference source. As is well
known, parallel illumination of the transparency produces its Fourier trans-
form at the back focal plane together with a phase factor, as shown by
G ~ o d m a n . ~The
’ coplanar reference source removes the phase factor, and
thus we obtain the same properties as obtained with the Fourier transform
hologram.
Another variation of the geometry leads to the lensless Fourier transform
hologram by virtue of the absence of a lens.13 Here the Fresnel diffraction
field of the transparency is recorded together with the field from a coplanar
reference source. The phase factor is again removed, and the hologram
L, H
FIG. 10. Recording of a Fourier transform hologram.

formed with this arrangement has a transmittance resembling that of the

Fourier transform hologram.
4.3.4. Image-Plane Holograms

When the image of an object is formed on the plane of the recording
plate and a hologram is formed with the help of a reference wave, the
hologram is known as an image-plane hologram. A practical arrangement
is shown in Fig. 11.
The image reconstructed with the reference beam in the original direction
is similar to that obtained with a Fresnel hologram, except that the field of
view is limited by the aperture of the lens. The field of view can be increased
significantly by using the conjugate illumination. Then, the illuminating
beam is on the opposite side and in opposite direction to the original
reference beam, as described by Brandt28and R o ~ e n s . * ~
Since the conjugate illumination is preferred, the image in the reconstruc-
tion is pseudoscopic, or “inside out.” The image becomes orthoscopic if a
conventional hologram is used instead of the lens to form a real image over
the holographic plate.
In image-plane holograms, the requirements for the illuminating source
are less stringent. The effect of the extension of the source on the resolution
or image blur As of the reconstructed image can be seen from the following
relations.
As = (2) Ar, (4.8)
I
I
1
p,
L
Q2
FIG. 1 1 . Arrangement to form an image plane hologram.

180 HOLOGRAPHY
where Ar is the extent of the source, Z, its distance from the hologram,
and Z1the distance of a point on the object from the hologram.
Further, the equation describing the effect of the finite bandwidth Ah on
the resolution Au of the reconstructed image is given by the following
relation:
AU = z1sin 0, (y) , (4.9)
where Or is the angle of the reference beam with the normal to the hologram
plane (the subject and image are considered to be on-axis).
Equations (4.8) and (4.9) reveal that for small 2,,extension of the source
and finite bandwidth have little effect. Thus reconstructions from image-
plane holograms are sharp and achromatic even with white-light-extended
sources. However, for image points which are far outside the plane of the
hologram, the reconstruction shows color dispersion and blur.
In conclusion, image-plane holography is particularly suitable whenever
display of holograms, double-exposure interferometric holograms, or vibra-
tion-analysis holograms are desired in white light.
4.3.5.Color Holograms
Multicolor holography was first suggested by Leith and Upatnieks6 The
basic idea of color holography is to use three-color object and reference
beams to record a hologram on a single photographic plate. If the object
and the reference beams contain the three primary colors, the hologram
would, when illuminated with a beam identical to the reference beam,
reconstruct the object wave in full color.
There are basically two main techniques to record color holograms ( a )
holograms recorded as planar diffraction gratings and (b) holograms recor-
ded as volume diffraction gratings. The difference in response of the two
grating types is made evident by considering the grating equations:
d(sin i + sin 0) = A (planar grating), (4.10)
2d sin i =A (volume grating). (4.11)
The planar grating equation reveals that for a given A, diffracted light
would always be observed for any incidence. Thus the response is that
broad-band and special techniques are required to separate or eliminate
the spurious images which result from light of wavelength A l , diffracting
from the component hologram recorded with the wavelength A2.
The volume grating equation, on the other hand, states that the incident
and diffraction angles are equal, and thus only wavelengths satisfying Eq.
(4.11) would be significantly diffracted. Thus the spurious images are sup-
pressed by the filtering action of the volume grating. Further, there is no
loss of resolution, contrast, or spatial frequency bandwidth.
We now describe the various techniques used to record color holograms.
4.3.5.1. Plane Hologram Techniques. There are several basic tech-
niques to make color plane holograms, as will be described next.
In the three-reference-beam method due to Leith and U p a t n i e k ~ ,the
~
object beam is a mixture of three primary colors. The reference beam is a
combination of three separate beams in the three different primary colors.
Each reference beam is inclined at a different angle to the recording plate.
In this way, the spurious or the cross-talk images resulting from diffraction
of light of wavelength A , and from the component hologram recorded at
wavelength A 2 are spatially separated.
This method, however, is not very practical because the field of view is
very much restricted. Furthermore, during reconstruction, the illuminating
beam must be accurately aligned so as to be identical to the reference beam.
The spatial multiplexing method is essentially due to Collier and
Penningt01-1.~'Here, we spatially multiplex, in a nonoverlapping manner,
the several holograms corresponding to the several colors used, i.e., the
various colored interference patterns are not allowed to overlap on the
recording medium. During reconstruction, the colored illuminating light is
incident in such a way that each component hologram is illuminated with
only that color light used to form it.
There are several experimental arrangements which could be used to
form such a hologram. One possible scheme is shown in Fig. 12. A mask
F consisting of strips of red, green, and blue filters is imaged onto the
hologram plane by a lens with the multiwavelength reference beam. The
image, which consists of discrete areas of single-wavelength light, acts as
FIG. 12. Recording a color hologram by spatial multiplexing.

182 HOLOGRAPHY
the reference wave front. The light in each strip then interferes with the
light of the same color coming from the object, giving nonoverlapping
multiplicity of single-color holograms. The reconstruction step consists of
placing the hologram in register and illuminating through the same filter
mask.
Some resolution is sacrificed in this method, since each component
hologram will be smaller than the total recording. Also, this scheme results
in the production of some noise, because in a A ,-component hologram there
will also be light present of A2 and A 3 from the object, thereby decreasing
the diffraction efficiency of the hologram ( A , , A 2 , A 3 are the three primary
colors).
The coded-reference-beam method, also due to Collier and Pennington,30
involves coding of the reference beam. The amplitude and phase of the
reference wave are made to vary across the hologram plane in a significantly
different manner for each wavelength. During reconstruction the developed
plate has to be replaced precisely in the same position that it occupied
during the formation of the hologram.
A possible arrangement for recording such a hologram is shown in Fig.
13. The ground-glass screen D serves as the reference-beam coding plate.
Each colored component of the reference beam, when passing through the
diffuser, produces a complicated speckle pattern on the hologram plane.
The component speckle patterns are randomly different from each other
with respect to phase and intensity at a point, meaning that each colored
reference wave is uniquely coded. Collier and Pennington have noted that
when only light of wavelength 0.514 km was used, an excellent image was
obtained. However, when the wavelength was changed by only 0.0128 km
and the coding plate was illuminated in exactly the same way, only a uniform
noise was observed.
FIG. 13. Recording a color hologram using a coded reference beam.

Successful holograms have been recorded using this method, although

the noise problem still remains. The geometry of the code plate plays a
major part in determining the degree to which the noise affects the recon-
struction. The noise, which is principally due to the interaction of the
wavelengths A , and A 2 , with the hologram formed with A 3 , is more uniformly
distributed over the image area by increasing the solid angle subtended by
the diffuse screen at the hologram.
4.3.5.2. Volume-Hologram Techniques. There are two kinds of volume
hologram techniques, namely, transmission and reflection volume
holograms, as will be described in this section.
The transmission volume hologram technique was used to record the first
mulitcolor hologram by Pennington and Lin3' in Kodak 649F photographic
emulsion. Light of wavelength 6328 8, from a He-Ne laser was mixed with
light of wavelength 4880 8, from an argon laser, using a beam splitter. Of
the two doubly colored beams so produced, one acts as a reference and the
other illuminates a color transparency. The reference and the subject beams
then combine together at an angle of 90" with respect to each other at the
hologram plane. As noted before, the interference is throughout the depth
of the thick emulsion.
An arrangement to record a three-color hologram devised by Friesem
and Fedorowicz3* is shown in Fig. 14. The two three-colored wave fronts
are generated in a similar manner as in the earlier r e f e r e n ~ e . ~A' better
color fidelity is, of course, obtained in the reconstruction than with two-color
wave fronts.
Another technique uses the properties of the reflection holograms already
discussed in Section 4.2.3. The hologram can be reconstructed with a
multicolor wave front identical to the reference, or with white light as noted
earlier.
L2 0
FIG. 14. Recording a color volume-transmission hologram.

184 HOLOGRAPHY
4.3.6. Composite Holographic Stereograms

An interesting application of holography is the three-dimensional display
of an object by the use of two-dimensional pictures or projections. This
field, known as composite holography, started in the late sixties with
McCrikerd and George,33 George et Redman," and De B i t e t t ~ . ~ ~
Composite holography consists of producing a multiple hologram formed
by many vertical strip holograms, each with a different paralax. The observer
then obtains the stereo effect by looking at a collection of very thin vertical
holographic windows. A different way to synthetize three-dimensional
holograms from two-dimensional pictures has been described by Sopori
and Chang3' and Haig.38 They multiplexed many images in a single plate
by changing the angle of incidence of the object or the reference beams
when making the hologram.
Kasahara et ~ l . Groh
, ~ ~and Koch? and Grossman4' applied composite
holography to the three-dimensional display of internal human organs from
ordinary X-ray projections.
4.3.7. Rainbow Holograms

White-light illumination of display holograms has proven to be so much
more practical than laser or arc-lamp illumination, that most of the
holograms that we see today are of some white-light type. One of these is
the rainbow or Benton hologram. Rainbow holography was first reported
by Benton' as a two-step process, but recent developments have simplified
the procedure to one step with some limitations.
Another development is the cylindrical holographic stereogram derived
from the rainbow holograms by Cross." Below, we describe these three
methods one by one.
4.3.7.1. Two-step Method. A conventional hologram H , is first recor-
ded as shown in Fig. 15(a). The real image is then reconstructed by a beam
that is antiparallel to the original reference beam. However, during recon-
struction, instead of illuminating the whole plate, only a horizontal slot
about 10mm wide is illuminated. A second hologram plate H 2 is placed
near the real image, and a new hologram is recorded with a reference beam
coming from the same side as the rays producing the real image [see Fig.
15(a)l.
After development, the second hologram H2is illuminated by a white-light
beam in a direction antiparallel to the reference beam. The object rays are
retraced in the direction they came from, back to the position of the slit.
In other words, the light is diffracted only towards the real image of the
thin horizontal slot of the master hologram. Further, because of dispersion,
we obtain many colored slit images one over the other (see Fig. 1%).
FIG. 15. Recording a rainbow hologram by the two-step method.
The presence of the slit makes a significant reduction in the information

content in the second hologram, allowing a relaxation in the coherence
requirements for the reconstructing light source. When the eye is placed at
one of the real images of the slit, one sees the complete virtual image of
the object in one color, depending on through which colored slit image one
is watching. The holographic image of the object is almost monochromatic,
and there is almost no blurring effect due to dispersion.
Instead of a point light source, an incoherent vertical line source may be
used to produce a continuum of vertically displaced horizontal strip images.
Another advantage is that the reconstruction efficiency is very high, since
most of the light is diffracted towards the observer’s eyes. A disadvantage
is that the vertical parallax is sacrificed. This is not a serious drawback, as
most of the time the horizontal parallax is sufficient, and it is the one that
is necessary for stereoscopic vision.
As noted earlier, in a Lippman-Denisyuk hologram, the colors are separ-
ated by an interference filter effect, and one single color is selected by the
illumination angle. On the other hand, in rainbow holography, the colors
are separated by diffraction, and only one single color is selected by the
observation angle.
186 HOLOGRAPHY
4.3.7.2. One-Step Method. The one-step method was first introduced

by Chen and Yu4* and was further developed by Chen.4'.44 Review papers
on this subject have been presented by Chen et al.45and by Benton et al.46
The orthoscopic recording scheme is shown in Fig. 16(a). A horizontal slit
A is placed on the lens which images the object near the recording plate
P2. The reference-wave point source SLis located in the plane of the lens
and is centered above the slit.
The finished hologram, when reconstructed with the same reference wave,
produces an orthoscopic image. But, the observer can only view a part of
this image directly in front of the slit image; the rest is vignetted out and
one must move about to sample it. Illumination with a white light source
S,, however, produces a dispersed image of the slit, allowing the observer
to see more of the image but in a spectrum of colors (Fig. 16(b)]. Chen
and Yu42 have shown that by locating the slit between the object and the
FIG. 16. One-step orthoscopic recording scheme of a rainbow hologram ( a ) Recording;

(b) Observation.
front focal plane of the lens, a monochromatic image can be obtained. This
is because the image of the slit is formed on the same side of the plate as
the observer, thus allowing the observer to place his eyes directly at the
position of the slit image. But the size of the image becomes limited due
to vignetting by the lens aperture.
The pseudoscopic configuration is shown in Fig. 17(a). The geometry is
almost similar to the one used for orthoscopic configuration, with the only
difference that the reference wave is convergent instead of divergent. In the
reconstruction step [see Fig. 17(b)], the conjugate illumination with white
\
. \
'\
\
\
\
FIG. 17. One-step pseudoscopic recording scheme of a rainbow hologram. ( a ) Recording;

(b) Observation.
188 HOLOGRAPHY
light is used to give real images of the slit through which the observer can
see the image of the object. Unfortunately, the real image that is obtained
positions the front of the object closest to the lens, farthest from the viewer.
Thus the viewer is presented with the familiar “inside-out” or reversed-depth
experience of a pseudoscopic image. However, there is an advantage that
the image is bright and sharp and is undistorted and unvignetted by the
lens aperture. Therefore, for applications where the depth perception is not
important, the pseudoscopic configuration would be more desirable.
The astigmatic recording scheme is yet another method for producing
rainbow holograms in one step and produces a good orthoscopic image
with large field of view in the horizontal direction. The recording is illustrated
in Fig. 18. The cylindrical lens L, images the object 0 in the vertical direction
near the hologram plate H. The slit S is also imaged but beyond the recording
plate. A divergent beam is used as the reference wave front. The image of
the object can be reconstructed with a white-light point source in the same
position as the reference source. Since in the vertical direction the hologram
is an image plane hologram, the image in this direction is sharp (see Section
4.3.4).
It is worth pointing out that in the one-step rainbow process using
conventional spherical lenses, the field of view is narrow in the horizontal
direction, which is a major drawback. This is primarily because of the
difflculty in getting a large aberration-free spherical lens. The astigmatic
recording has removed this obstacle, since long convex cylindrical lenses
are easily available.
The lensless method of Bahuguna and Santoyo4’ is particularly suitable
for transparencies. In this method, a master hologram of a diffuse slit is
recorded. By using conjugate illumination a wave converging to a slit is
reconstructed. This reconstructed wave is utilized to illuminate the object
transparency for creating a rainbow hologram. Since the master hologram
can be used for making a rainbow hologram of an unlimited number of
transparencies, this method is essentially a one step method. A variation
of the above method has been described by B a h ~ g u n a ~ ~ .
FIG. 18. Astigmatic recording scheme of a rainbow hologram.

DIFFERENT TYPES OF HOLOGRAMS 189
4.3.7.3. Rainbow Cylindrical Holograms. Lloyd Cross” invented the

cylindrical holographic stereograms, utilizing the earlier work in rainbow
holography invented by Benton,q and composite holography.
Composite holography consists of making a motion picture film of a
subject, placed on a slowly rotating table, with a cine camera. The individual
frames of the film are hologramed on strips of about one millimeter wide
and twenty cemtimeters long. The holographic film is then processed and
bent into a cylindrical shape. When the observer views the film, each eye
sees a photograph of the object taken from a different angle (just as with
a stereo viewer), and consequently a three-dimensional image is perceived
at the center of the cylinder as the observer walks around the hologram.
An ordinary tungsten lamp with a transparent glass envelope is used as a
light source.
A machine to make this kind of hologram using anamorphic optics (plastic
and holographic cylindrical lenses) was described by Fuesk and Huff .49
4.3.8. Computer Holograms

A computer-generated hologram (CGH), also called synthetic hologram,
is produced with a graphical output from a digital computer. The hologram
is calculated from the knowledge of a wave front or the object to be
represented. The actual hologram is made by photoreducing the computer
graphical output on a photographic film or glass plate.
The first CGHs were developed by Brown and Lohmann” with the
purpose of making spatial filters for optical-data processors. Since then,
the number of applications of CGHs have grown, and they may now be
divided into the following five’areas:
a. three-dimensional image display
b. optical-data processing
c. interferometry
d. optical memories
e. laser-beam scanning
All of these applications have been described in a very complete review
by Lee.s’
Computer-generated holograms may be of three kinds:
1. Binary holograms. Most computer plotters may produce only high-
contrast prints without gray tones. When the print is copied and minified
by means of a high-contrast film, the resulting hologram is like a black
screen with many small holes and is called a binary hologram, as described
by Lohmann and Paris.” Efficient holograms have a total transparent area
of about 50%.
190 HOLOGRAPHY
Binary holograms are analogous to normal holograms recorded on a

very-high-contrast and nonlinear photographic film. Then, high-order
images should be expected, but they do not overlap if the sampling theorem
is obeyed.
There are many applications for binary holograms, but their most interest-
ing applications are probably in the field of optical testing, as shown by
W ~ a n andt ~ loo
~ mi^.'^ Computer-generated holograms for this application
are used to produce reference aspheric wave fronts in order to test an
aspheric optical surface or lens. This kind of hologram looks like an
interferogram, as shown in Fig. 19.
2. Half-tone holograms. The computer hologram may also be obtained
by displaying the results on a suitable half-tone printout or on a high-
resolution television screen. This kind of representation approaches a real
hologram.
3. Phase holograms (kinoforms). A kinoform, as described by Lesem et
~ ’a phase hologram obtained by the following method. A variable-
a ~ , is
density display is calculated in such a way that after bleaching, the profile
of the emulsion gives a phase hologram which reconstructs only one image.
FIG. 19. Binary computed holograms to test aspherical surfaces (from Ref. 53).
S O M E APPLICATIONS OF HOLOGRAPHY 191
Then, the picture is taken and bleached. Such a hologram is of interest,

because the entire flux is concentrated in this image.
4.4. Some Applications of Holography
There are numerous applications for holography. Most of them utilize

and exploit the fundamental difference between holography and photogra-
phy, namely that holography records the whole wave front, while photogra-
phy is a recording of the irradiance distribution in the image. In this section,
we deal with the more important applications.
4.4.1. Holographic Interferometry

Interferometry is a laboratory technique by which extremely small move-
ments of objects are detected and measured by exploiting the interference
properties of light. It can detect changes as small as a fraction of the
wavelength of light. Conventional interferometry is mainly useful for making
measurements on highly polished surfaces of relatively simple shape. Holo-
graphic interferometry extends the range to include three-dimensional sur-
faces of arbitray shape and surface condition. It is concerned with the
formation and interpretation of the fringe pattern which appears when a
wave generated at some earlier time and stored in a hologram is later
reconstructed and made to interfere with a comparison wave.
Holographic interferometry has been applied with success to test and
measure aspheric optical surfaces, for example, by Broder and M a l a ~ a r a , ’ ~
FouCrC and M a la ~a ra ,~’ Malacara and M a l l i ~ k The
. ~ ~ time-delay aspect
makes the holographic method a very useful tool in many fields. For
example, a machine part can be tested before and after stress. Further, due
to the basic three-dimensional structure of the holographic reconstruction,
a complex object can be examined interferometrically from many different
perspectives: a single interferometric hologram is equivalent to many
observations with a conventional optical interferometer. We shall discuss
below the three basic forms of holographic interferometry.
4.4.1.l. Real-Time Holographic Interferometry. The hologram of the
test object is first recorded in the usual way. The hologram plate is processed
and put back in exact register on the plate holder, with the object and the
reference beam undisturbed. The illumination can be adjusted so as to
equalize the absolute values of the original and reconstructed object-wave
amplitudes. Under the ideal conditions, the two wave fronts must exactly
cancel, and the object should not be visible. In practice, at least one broad
192 HOLOGRAPHY
bright fringe is usually observed, because the processing of the hologram

inevitably distorts the emulsion to some small degree.
In the next step, the object may be placed under stress and caused to
deform, or it may be released from stress and allowed to flow and creep or
expand. As a result, the actual object wave differs at several points with
the reconstructed wave, and the field of view is now crossed with fringes.
Mathematically we can write the object wave in the initial and final state
as Oo(x)exp(i+o(x)) and Oo(x) exp(+;(x)) respectively. The location of
the fringes would then be defined by the locus of all points for which
&(x) - 40(x)= const. (4.12)
A dark fringe is produced whenever
r r=;O , 1,2,....
4 & ( ~ ) - 4 ~ ( ~ ) = 2 ( m + l ) .m (4.13)
Many authors, like Aleksandrov and B o n ~ h - B r u e v i c hEnnos,60
,~~ Froehly
et a1.,6' Tsuruta et S0llid,6~Molin and Stetson: and Hecht et ~ 1 1 . 6 ~
have worked on the problem of relating the observed fringes to the actual
object deformation. According to Abramson,""-" the interpretation of the
fringes is simple in most of the cases. His holodiagram allows measurement
of deformation even for relatively complex and large objects. This technique
and other methods, for example the method of Hecht et al.,6' have made
holographic interferometry a very useful metrological tool for routine
measurements.
A spherical form of the single-exposure interferometer is shown and
described in Fig. 20. First a hologram of a diffuser is taken [Fig. 2O(a)].
The hologram is then put back in the plate holder in exact register so that
both the actual and the reconstructed waves superpose. If a test flat T is
inserted in the object beam [Fig. 20(b)], interference fringes appear which
D D T
P H P H
(a) (b)
FIG. 20. Instrument to perform single-exposure holographic interferometry

SOME APPLICATIONS OF HOLOGRAPHY 193
are related to the optical inhomogeneities, wedge and thickness variations

present in the test plate.
4.4.1.2. Double-Exposure Holographic Interferometry. When it is
sufficient to form a permanent record of the relative surface displacement
occurring after a fixed interval of time, a simpler technique, the double-
exposure method is employed. In this method, two exposures are taken,
one for the initial state and the other for the final state of the object, on
the same hologram plate. The interferogram so formed, yields the necessary
fringes on illumination by the reference beam. Figure 21 shows an inter-
ferogram of an aluminum block tilted between the exposures.
This method is similar to the single-exposure method in most respects.
Unlike the previous method, exact manual registration of the reconstructed
wave with the original one is not required. Further, distortion due to
emulsion shrinkage is identical for both reconstructed waves and therefore
does not affect the spacing of the fringes formed by the interference of the
two waves.
FIG. 21. Holographic interferogram of an aluminum block, tilted between two exposures.
194 HOLOGRAPHY
4.4.13. Time-Averaged Holographic Interferometry. The multiple

exposure interferometry can be extended to the limiting case of continuum
of exposures. The resulting technique, known as time-averaged holographic
interferometry, is used to study how surfaces vibrate. It was first reported
by Powell and Stetson.”
Suppose the object is vibrating in one of its normal modes. The time-
averaged hologram of such an object gives what is effectively a double
exposure of the vibrating surface seen at either extreme of its oscillation.
Thus, the object would appear crossed with fringes.
Figure 22 shows some of the time-average interferograms of a 35-mm
film can at its second resonance. Here the line across the middle of the can
is a node of vibration of the can, and the contours to either side are loci
of constant amplitude of vibration. Each bright area represents an amplitude
of vibration of approximately a multiple of a half wavelength.
4.4.1.4. Sandwich Holography. The fringes in interferometric hologra-
phy appear because the object either moved or changed its shape between
the first and the second exposure. Many times, however, it is not easy to
eliminate the effect of the movement of the object as a rigid body and to
determine the object deformations due to loads or stresses.
Once the second exposure is taken, the effect of the movement of the
object as a rigid body can only be compensated if the two holograms were
recorded with two different and separated reference waves. However, an
elegant solution is the sandwich hologram invented by Abram~on.’*-’~
In sandwich holography, a pair of holographic plates are simultaneously
exposed by placing them in the same holder, with their emulsions towards
the object and with the antihalation backing removed from the front plate.
Two sandwiches at least must be exposed, one without any stress in the
object and the second after the stress is introduced, as shown in Fig. 23.
Upon reconstruction, both sandwich holograms will reconstruct the object
FIG. 22. Time-average interferograms of a 35-mm film can at its second resonance (from
Ref. 12).
SOME APPLICATIONS OF HOLOGRAPHY 195
A , A2 OBJECT
OBJECT
FIG. 23. Sandwich holography.
without any fringes. However, if the sandwiches are now formed by the
pairs of plates A l with B2 or by BI with A * , the reconstructed image of the
object will show the fringes.
The tilting of a hologram laterally moves the image of the object when
reconstructing the image of an object. However, if the sandwich pair is
tilted, the images of both holograms move almost together, producing only
a very small tilt of one image with respect to the other. This property of
sandwich holograms permits compensation or simulation of small rigid
movements of the body by means of small tiltings of the sandwich. In this
manner, the fringe pattern due to the object stresses is easily isolated from
any possible object rigid motion fringes.
4.4.2. Holographic Microscopy

The application of holography to microscopy dates back to the origin of
holography itself. Gabor had intended to change the wavelength between
formation of the hologram and reconstruction of the same in order to obtain
a magnification in the ratio of the wavelengths. The magnification can also
be obtained by scaling down the hologram, or by decreasing the curvature
196 HOLOGRAPHY
of the reconstructing wave as compared to the reference wave. The general

formula for magnification is given by
mL
M= (4.14)
I*---
m2Zo z,'
PZC Z R
where the upper sign is for the primary image and the lower sign is for the
conjugate image. Here m is the linear magnification of the hologram, is
the ratio of the two wavelengths, and Z,, Z,, and 2, are the z-coordinates
of the object-point, the reference-point source and the illuminating point
source respectively.
The above principles have been effectively utilized to form holographic
microscopes. For details of the theory, construction, and working, the reader
is referred to the literature, specially the papers by E l - S t ~ m El-Sum
,~~ and
Kirkpatri~kB , ~ a~ e ~ Van
, ~ ~Ligten and O ~ t e r b e r g and
, ~ ~ Thompson et al.25
4.4.3. Imagery through Diffusing Media

The recovery of the image of an object obscured by a distorting medium,
for example the distorting atmosphere, has been a subject of investigation
for several years. One possible use for holography is to solve such problems.
Let the object be obscured by a diffuser. To image the object faithfully,
we proceed as described by Leith and UpatnieksBOA hologram of the object
wave through the diffuser is taken with the help of a reference wave. The
processed plate is put back in register, with the diffuser still in place. The
hologram is now illuminated by a conjugate wave. This reconstructs the
complex conjugate of the signal plus diffuser which, after passing through
the diffuser, produces the conjugate real image, with the noisy part due to
the diffuser having been filtered out.
The explanation of the phenomena is as follows. The diffuser can be
mathematically represented by the function exp[ i e ( x , y )]. During recon-
struction, the conjugate real image of the diffuser, given by exp[-iO(x, y ) ] ,
is superposed on the diffuser, mutually cancelling the phase factors. It is
as if the diffuser was not there, and we observe the real image of the object.
In case the disturbance is time varying, the above method does not help.
The possibility of a solution exists if the reference and the object wave are
made to traverse essentially through the same part of the intervening
medium, as explained by Goodman." Then, at a first approximation, the
phase difference between the two waves at the hologram will be largely
independent of the intervening medium, implying that the hologram is
practically unaffected by the perturbing medium.
SOME APPLICATIONS O F HOLOGRAPHY 197
4.4.4. Matched Filtering and Character Recognition

An optical matched filter is a photographic recording of the Fourier
transform of the amplitude distribution of the object. Since the amplitude
as well as the phase are recorded, this is essentially a Fourier transform
hologram of the object. The primary use of such a filter is character
recognition, as shown by Vander LugtZ6and Vander Lugt er a/.'* The
procedure is as follows. The filter, corresponding to the character to be
read, is placed in the filter plane of a 4f-filtering system. The filtering process
performs a cross-correlation between the object and hologram filter. In the
image plane, we get a bright spot corresponding to the detected character.
In case a set of characters are to be stored, the hologram filter consists of
a number of holograms, corresponding to these characters, stored in a single
master hologram.
4.4.5. Holographic Optical Elements and Aberration Corrections

Conventional optical elements like lenses and prisms may be replaced
by holographic optical elements in many applications, with a lower cost,
as described by Close." Computer-generated holograms may be produced
to simulate and substitute aspheric lenses that produce complicated wave
fronts.
A holographic optical element may even be analyzed by conventional
lens methods, like ray tracing. It has been shown by Sweattx4that the analogy
is so close that a holographic lens may be considered as a normal lens with
an infinite index of refraction.
A hologram may also be used as an extra optical element in an optical
system in order to correct some aberrations, as shown by Upatnieks et a/.,*'
and to improve its performance. An obvious problem, however, is that the
light efficiency of a hologram is, in general, quite low. A second problem
is that monochromatic light must be used due to the color dispersion of
the hologram.
4.4.6. Acoustical Holography

Most of the principles pertaining to optical holography are also applicable
to acoustical holography. However, the recording and the readout tech-
niques so far applied have been rather complex and difficult, and the results
have been of a poor quality. But, due to the large number of interesting
applications, it is hoped that new and simpler techniques will evolve in the
near future.
198 HOLOGRAPHY
There are principally four methods for recording an acoustic hologram:

a. recording by the surface of a liquid
b. recording by a matrix of detectors
c. hologram is obtained by scanning, pointwise
d. hologram is obtained by direct interaction of light and acoustic waves
In this section, we only deal with the first method. For details and other
methods, the reader is referred to the literature like the papers by Mueller
and Sheridon,86 Young and Wolf,87 and Preston.88
Figure 24 illustrates the geometry for recording and reading such a
hologram. S , and S2 are the two coherent ultrasonic sources. S , produces
the reference wave and S2 illuminates the object A. The reference wave and
the scattered wave from the object interfere on the surface of the liquid.
The information about the object is in the form of ripples on the surface
of the liquid.
In the reconstruction step, the distorted surface is illuminated by a parallel
beam of monochromatic light, and the scattered light wave is collected by
the lens, as shown in Fig. 24. The image of the light source is blocked by
a small stop. The two reconstructed images A' and A" appear very near the
focal plane of the lens.
The large differencebetween the recording ultrasonic wavelength and the
illuminating light has two deteriorating effects: ( a ) the images are formed
far from the surface, ( b ) the images are greatly aberrated. In order to
s, s,
FIG. 24. Recording a reconstruction of an acoustic hologram
EXPERIMENTAL PROCEDURES I N HOLOGRAPHY 199
eliminate these effects, the photographic record of the distorted liquid

surface is taken with a lens of sufficiently small aperture and later reduced
by the ratio of the two wavelength. This photograph can now serve as a
hologram to obtain an optical reconstruction of the acoustically illuminated
object.’“
Acoustical holography is finding several applications in various areas as
described by Green,’g and by Hildebrand and Brenden.” Since many objects
apaque to light waves are transparent to ultrasonic waves, they can be
optically examined through ultrasonic holography. Because of this feature,
it is also applicable to nondestructive testing, as shown by Byron and
Collins,” and to medical imaging, as shown by Malacara.’’
4.5. Experimental Procedures in Holography
The basic requirement in holography is to record a stable fringe pattern

due to the interference between two light beams. In order to achieve this,
certain procedures are to be followed in general. In what follows, we discuss
these together with the necessary equipment required to produce a good
hologram.
4.5.1. Light Sources for Holography

The requirements placed on the light sources to be used in hologram
formation depend on the subject and the experimental setup. Specifically,
the light disturbances reaching any point on the hologram plane (via
reflections from the subject surface or reference mirrors) must be mutually
coherent. In other words, the maximum optical path difference for light
rays between the source and the hologram via the above-mentioned path
must be less than the coherence length. Besides, the object and the reference
waves must be spatially coherent. Poor coherence would effect the fringe
contrast and thus the diffraction efficiency. Fortunately, with the availability
of the laser, the above requirements are easily achieved, and thus making
a hologram has become a routine task in the laboratory. It is, however,
worth noting that Leith and Upatnieks,’3 by using an achromatic fringe
system, have produced holograms with a high-pressure mercury arc,
although of a lower quality than those made with a laser source.
During reconstruction, the restrictions on the source depend on the type
of the hologram. For example, an image-plane hologram, a rainbow
hologram, or a volume hologram can be reconstructed with a n extended
white-light source. For best results, however, the reconstructing wave must
be identical to the one used during formation of the hologram. In general,
200 HOLOGRAPHY
deviation from this condition and poor coherence of the source affect the
resolution in the image.
Normally, for a small object of a few centimeters in size, a 2-mW He-Ne
laser is quite adequate in terms of the required power. For larger objects,
one can use a higher-power He-Ne laser. If still higher power is needed,
one may want to use an argon laser which gives output in the range of a
few watts. To study transient phenomena, Brooks er have shown that
a pulsed laser is a must.
4.5.2. Mechanical Stability

There are essentially two stability conditions that must be met to record
a fringe pattern. We discuss them in the following two sections.
4.5.2.1. Stability of the Recording Plate. The recording plate must be
stable relative to the fringe pattern in order to record the pattern faithfully.
Its stability requirements depend on the highest frequency content in the
fringe pattern. In most holograms, the minimum fringe spacing is approxi-
mately equal to the wavelength of the forming light. So the recording medium
must not move more than a small fraction of a wavelength during exposure.
4.5.2.2. Stability of the Fringe Pattern. Special efforts have to be made
to ensure the stability of the fringe pattern. The whole optical setup must
be isolated from external vibrations. An inexpensive method is to mount
the optical setup on a massive slab of iron or concrete (weighting about
0.5 tons) supported on partially inflated motorcycle tubes. The resonance
frequency must be much lower (about 1 Hz) than the frequency of the
building vibrations. Sophisticated ready-made vibration-free tables are
available from the Newport Corporation.
Acoustic and thermal disturbances are more troublesome. The best way
is to eliminate the sources of such disturbances themselves, for example
switching off air-conditioning units, fans, and heaters, etc. Reducing the
exposure time is also advantageous, but at the expense of having a more
powerful laser. One must also reduce, as far as possible, the optical path
lengths, especially those between the beam splitter and the recording
medium.
4.5.3. Beam Expansion

The laser-output cross-section is typically 1 mm in diameter to illuminate
the subject. While making a hologram, one needs a larger cross-section,
and hence the need of expanding the beam arises. The beam can be expanded
by one or more lenses or spherical mirrors.
EXPERIMENTAL PROCEDURES IN HOLOGRAPHY 201
Because of the diffraction effects at the walls of the laser cavity, the
TEMoo mode (the usual mode of operation) may be modulated, giving a
speckled appearance to the expanded beam. This is especially the case when
the laser resonant cavity is formed with two large-radius mirrors. One can
get rid of this modulation by focusing the laser output through a microscopic
objective and putting a spatial filter (a fine pinhole) of diameter between
10 to 40 pm, depending on the magnification of the microscope objective
being used, at the focus of the beam. The output is a clean expanded beam.
4.5.4. Holographic Materials

The subject of holographic materials is so important that complete books
have been written about it, for example, by Smith.95Good chapters in books
covering this topic have been written by Hariharan” and by Gladden and
L e i g h t ~ Additional
.~~ general references are given in this last chapter, an
interesting reference being that of Bartolini et aL9’
4.5.4.1. Photographic Emulsion. Silver halide photographic emulsions
are by far the most popular of all recording media. They are available in
film or glass plates, mainly from Kodak or from Agfa-Gevaert. Their spectral
sensitivities are of different types, as shown in Fig. 25 and Fig. 26. With
Y
a
5 a;
1000
e m -
Moo-
swo 4000 4MK) 1100 eooo 0100 Too0 7100
WAVELENGTH OF LIGHT
F i c . 25. Spectral sensitivity curves for some Kodak holographic silver halide emulsions.
202 HOLOGRAPHY
WAVELENGTH OF LIGHT
FIG. 26. Spectral sensitivity curves for some Agfa-Gevaert holographic silver halide
emulsions.
them it is possible to match most laser wavelengths. Table I summarizes

some of the most important characteristics of these photographic materials.
For ordinary practical purposes, the optical density versus log E
(exposure) is the characteristic normally specified. For holographic pur-
poses, the transmittance t versus the exposure E was originally suggested;
however, from the point of view of diffraction efficiency, it is more important
to plot the transmittance t versus the logarithm of the exposure. These
curves for some holographic photographic emulsions are given in Fig. 27.
Glass photographic plates are available with and without antihalation
backing. If desired, the antihalation backing may be removed with alcohol.
4.5.4.2. Dichromated Gelatin. The most important feature of hardened
gelatin is that it is not dissolved in water, but it only swells three to four
times its original thickness.
If a gelatin film contains a small amount of dichromate((NH.&CrzO,),
when exposed to blue light, the Crh' ion is reduced to Cr3+,making the
gelatin even harder, as described by ChangY8and Curran and Shankoff."
If the plate is adequately processed, as shown for example by Hariharan,"
a modulation of the thickness as well as of the refractive index is obtained,
producing in this manner a highly efficient phase hologram.
EX PER1 M E N T A L P R O C E D U R E S I N H O L O G R A P H Y 203
TABLE1. Silver Halide Films and Plates for Holography
TYPe Resolving Emulsion

Spectral Power in Thickness
Plates Films Sensitivity I/mm. in pm.
Kodak
649 F 649 F Panchromatic 2000+ 17/16

120-01/02 SO-I73 Blue and Red 2000+ 6
125-01/02 SO-424 Blue-Green 1250 7/3
131-01/02 SO-453 Blue-Green 1250 9
Agfd-Gevaert
10E56 10E56 Blue-Green 1500f 7

10E75 10E75 Panchromatic 1500+ 7
8E56 H D 8E56 H D Blue-Green 2500+ 7
8E75 H D 8E75 H D Blue and Red 2500+ 7
EXPOSURE IN ERGS/cm'
FIG. 27. Curves of amplitude transmittance versus exposure, in a logarithmic scale, for
some holographic silver halide emulsions, using a wavelength of 6328 A.
204 H 0 LOG RAPH Y
4.5.4.3. Photoresists. The photoresists are resins that, like dichromated

gelatin, are hardened by the action of the light. Their use in holography
has been described in several articles by Bartolini100-'02and by Norman
and Singh.Io3 Blue or green light produces chemical changes in the resin
that make it more difficult to dissolve with a solvent. Phase-diffraction
gratings or holograms may be produced with photoresists, but unlike those
produced with dichromated gelatin, no modulation of the refractive index
is obtained, but only of the thickness.
The developer is a suitable solvent that dissolves the unexposed region
faster than the exposed ones, or vice versa, depending upon whether the
photoresist is negative or positive, respectively.
A positive photoresist should be used for holographic purposes. Shipley
AZ-1350 has been the most widely used photoresist, but recently it has been
replaced by Shipley Microposit 1350 J photoresist. Since the sensitivity of
this photoresist is greater for shorter wavelengths, blue light for a He-Cd
laser or green light from an argon laser should be used.
If linearity is desired, the appropriate development procedure should be
performed, using a developer different from that used for the normal uses
of the photoresist, as described by Norman and Singh."'
4.5.4.4. Other Materials. Besides the materials just described, there are
various other recording materials, for example: photochromics,'04~'09 photo-
polymers,' ' O J ' ~ thermoplastics,' l 4 electro-optic crystaIs,"5-'20 magneto-
optic materials,'2"'22 metal films,123and ferroelectric m a t e ~ i a 1 s . I ~ ~
4.5.5. Recording, Processing, and Bleaching

These three basic steps in hologram-making using photographic emulsions
have been described in detail by Benton'" and will now be considered.
4.5.5.1. Recording. There is no universal recording material that is
suitable for all of the varied applications for holography. However, the
photographic emulsion is still the principal recording medium, as it comes
close to having ideal properties for a wide range of requirements. The most
essential characteristic of the recording medium is that it should be capable
of recording the generally high frequencies encountered in holography. For
example, for an angle of 30" between the object and the reference beams,
the frequency is about 1000 cycles/mm, which is beyond the resolution
limit of most of the recording materials. Besides this, the medium must be
reasonably sensitive to the particular radiation involved.
4.5.5.2. Processing. The following procedure, not very different from
processing of the high-resolution photographic image, is recommended for
the exposed holographic plate:
EXPERIMENTAL P R O C E D U R E S IN HOLOGRAPHY 205
1. Development: 5 minutes continuous agitation in Kodak D-19

developer at 20°C, followed immediately by a 1-minute rinse in water.
2. Fixation: 5 minutes continuous agitation in Kodak rapid fixer, followed
by a 1-minute rinse in water.
3. Removal of residual fixer: I f minutes in Kodak hypoclearing agent.
4. Wash: 5 minutes in flowing water.
5 . Removal of dye-sensitizer: 5 minutes in methanol, followed by a
l-minute rinse in water.
6. Drying: Emulsion shrinkage is a real problem in holography. A one-
wavelength shrinkage can cause a 180O-phasereversal of the reconstruc-
ted wave front, a disadvantage for interferometric applications. Also,
there is a color shift in reflection holograms when reconstructed in
white light (see Section 11.2.2). Some special measures are to be taken
in order to avoid emulsion ~ h r i n k a g e . ' ~ . ' ~
4.5.5.3. Bleaching. An absorption hologram recorded in a photographic

emulsion is converted into a phase hologram by bleaching out the silver
image, thus obtaining higher diffraction efficiencies. There are essentially
two methods, as described in detail by Benton'" which are: ( a ) Direct
bleaching and (6) Reversal bleaching.
In the direct bleaching method, the film or plate is developed and fixed
as usual. It is then put in a bleach bath which converts the metallic silver
into a transparent insoluble salt having a refractive index significantly higher
than that of the gelatin. One of such agents is potassium ferricyanide, as
described by Burckhardt and Doherty.'" The hologram efficiency of this
and other similar processes is analyzed in a paper by Upatnieks and
Leonard,'27 showing that it is between 25% and 50%. Higher efficiencies,
up to 70%, may be obtained with a wet bromine process described by
Lehmann er al.,'" or with a dry process using bromine vapor, as shown by
G r a ~ b e .Unfortunately,
'~~ holograms bleached by these methods are associ-
ated with flare light due to the enhancement of the low spatial frequencies.
This is due to the fact that the resulting relief images augment the index
images. The noise can be substantially reduced by substituting Kodak D-76
for the usual D-19 developer. It can be further reduced by immersing the
hologram in an index-matching liquid gate or by coating the hologram
surface with an index-matching film.
The reversal bleach method described by Smith'30 overcomes the noise
problem associated with the previous method by producing relief images
which tend to cancel the index variation at low spatial frequencies. Here,
the plate is developed but is not fixed. It is then put in the R-9 bleach bath,
which converts the metallic silver to a soluble salt and removes it entirely
from the emulsion. The result is that the remaining silver halide appears
206 HOLOGRAPHY
reversed with respect to the case of the direct bleach method, the relief
however remaining at regions where development took place. Thus the relief
image tends to cancel the phase variations due to the silver halide, and
consequently the speckle noise is reduced.
4.5.6. Embossing of Holograms

Phase photoresist holograms may be replicated by a process similar to
the one used to make photographic records, as described by Bartolini ef
C I ~ . , ’ ~ ’ Clay
, ” ~ and Gore,’33and by Burns.’34 The process consists first of
coating the photoresist hologram with a thin layer of nickel or silver either
by vacuum evaporation or by an electroless method in order to make it
conductive. The second step is to deposit on top of this metallic film a
thicker layer of nickel by electrolysis to form a metal stamper master.
The next step is to emboss this master into a transparent flexible thermo-
plastic like PVC or polycarbonate that softens at a temperature far from
that of thermal decomposition, as shown by Clay,’35to produce the desired
replicas. Thousands of holograms can be produced with a single nickel
master, at a very low cost for each copy.
Acknowledgments
The authors want to thank Prof. N. Abramson for his very valuable comments, Mr. Fernando
Mendoza for his help with the laboratory work, and Mr. Krishna Morales for the drawings
in this chapter.
References
1. D. Gabor, “A New Microscopic Principle,” Nature 161, 777 (1948).
2. D. Gabor, “Microscopy by Reconstructed Wavefronts,” !‘roc. Roy. Soc. A 197,454 (1949).
3. I). Gabor, “Microscopy by Reconstructed Wavefronts: 11,” Roc. Phys. Soc. B 64, 449
(1951).
4. E. N. Leith and J. Upatnieks, “Reconstructed Wavefronts and Communication Theory,”
’
J. Opt. Soc. Am. 52, 1123 (1962).
5. E. N. Leith and J. Upatnieks, “Wavefront Reconstruction with Continuous Tone Objects,”
J. Opt. Soc. Am. 53, 1377 (1963).
6. E. N. Leith and J. Upatnieks, “Wavefront Reconstruction with Diffused Illumination
and Three-dimensional Objects,” J. Opt. Soc. Am. 54, 1295 (1964).
7. Y. N. Denisyuk, “Photographic Reconstruction of the Optical Properties of an Object
in its Own Scattered Radiation Field,” Sou. Phys. Dokl. 7 , 543 (1962).
8. G. Lippmann, “La Photographie de Couleurs,” Compr. Rend. Acad. Sci. Paris 112, 2 7 4
(1891).
9. S. A. Benton, “Hologram Reconstruction with Extended Incoherent Sources,” J. Opt.
Soc. Am. 59, 1545A (1969).
REFERENCES 207
10. L. Cross, “Multiplex Technique for Cylindrical Stereograms,” in Three-Dimensional

Imaging, SOC.Phot. Opt. Instr. Eng. 120, 25 (1977).
11. G . N. Stroke, An Introduction to Coherent Optics and Holography, Academic Press, New
York, 1969.
12. H. M. Smith, Principles of Holography, Wiley, New York, 1975.
13. R. J. Collier, C. B. Burkahardt, and L. H. Lin, Optical Holography, Academic Press, New
York, 1971.
14. F. Francon, Holography, Academic Press, New York, 1974.
15. H. J. Caulfield, ed., Handbook of Optical Holography, Academic Press, New York, 1979.
16. N. Abramson, The Making and Evaluation of Holograms, Academic Press, New York, 1981.
17. P. Hariharan, Optical Holography, Cambridge University Press, Cambridge, 1982.
18. A. A. Friesem and J. S. Zelenka, “Effects of Nonlinearities in Holography,” Appl. Opt.
6, 1755 (1967).
19. J. W. Goodman and G. R. Knight, “Etfects of Film Nonlinearities o n Wavefront-
Reconstruction Images of Diffuse Objects,” J. Opt. SOC.A m . 58, 1276 (1968).
20. 0. Bryngdahl and A. Lohmann, “Nonlinear Effects in Holography,” 1. Opt. SOC.A m . 58,
1325 (1968).
21. A. Kozma, “Analysis of Film Nonlinearities in Hologram Recording,” Oprica Acra 15,
527 (1968).
22. R. W. Meier, “Cardinal Points and the Novel Imaging Properties of a Holographic
System,” J. Opt. SOC.A m . 56, 219 (1966).
23. 0. Bryngdahl and A. Lohmann, “Single Sideband Holography,” J. Opt. SOC.A m . 58, 620
(1968).
24. R. W. Meier, “Twin Image Elimination in Holography using Single Sideband Waves,”
J. Opt. SOC.A m . 59, 358 (1969).
25. B. J. Thompson, J. H. Ward, and W. R. Zinky, “Application of the Hologram Techniques
for Particle Size Analysis,’’ Appl. Opr. 6, 519 (1967).
26. A. Vander Lugt, “Signal Detection by Complex Spatial Filter,” IEEE Trans. Inform.
Theory IT-10, 2 (1964).
27. J. W . Goodman, Introduction to Fourier Optics, McGraw-Hill, New York, 1968.
28. G. B. Brandt, “Image Plane Holography,” Appl. Opt. 8, 1421 (1969).
29. L. Rosens, “Focused Image Holography with Extended Sources,” Appl. Phys. 9, 337
(1966).
30. R. J. Collier and K. S. Pennington, “Multicolor Imaging from Holograms Formed on a
Two-dimensional Media,” Appl. Opt. 6, 1091 (1967).
31. K. S. Pennington and L. H. Lin, “Multicolor Wavefront Reconstruction,” Appl. Phys.
Lett. 7, 56 (1965).
32. A. A. Friesem and R. J. Fedorowics, “Multicolor Wavefront Reconstruction,” Appl. Opt.
6, 529 (1967).
33. J. T. McCrikerd and N. George, “Holographic Stereogram from Sequential Component
Photographs,’’ Appl. Phys. Lett. 12, 10 (1968).
34. N. Geoge, J. T. McCrickerd, and M. M. T. Chang, “Scaling and Resolution of Scenic
Holographic Stereograms,” in Seminar-In-Depth Holography, Proc. SOC.Phot. Instr. Eng.,
117 (1968).
35. J. D. Redman, “The Three-Dimensional Reconstruction of People and Outdoor Scenes
using Holographic Multiplexing,’’ in Seminar-in-Depth Holography, Proc. SOC.Phot. Opt.
Instr. Eng. 161 (1968).
36. J. D. De Biteto, “Holographic Panoramic Stereograms Synthesized from White Light
Recordings,” Appl. Opt. 8, 1740 (1969).
208 HOLOGRAPHY
37. B. L. Sopori and W. S. C. Chang, “3-D Hologram Synthesis from 2-D Pictures,” Appl.
Opt. 12, 419 (1973).
38. N. D. Haig, “Three-Dimensional Holograms by Rotational Multiplexing of Two-
Dimensional Films,” Appl. Opt. 12, 419 (1973).
39. T. Kasahara et al., “Stereo-Radiography using Holographic Techniques,” Jap. J. Appl.
Phys. 8, 124 (1969).
40. G. Groh, and M. Koch, “3-D Display of X-ray Images by Means of Holography,” Appl.
Opt. 9, 775 (1970).
41, M. Grossman, P. Meyrucys, and J. Fontaine, “Three-Dimensional Holographic Synthesis
of X-ray Pictures”, in Holography in Medicine and Biology (G. Von Bally, ed.), Springer-
Verlag. Berlin, 1979.
42. H. Chen and F. T. S. Yu,“One-Step Rainbow Hologram,” Opfics Left. 2, 85 (1978).
43. H. Chen, “Color Blur of the Rainbow Hologram,” Appl. Opt. 17, 3290 (1978).
44. H. Chen, “Astigmatic One-Step Rainbow Hologram Process,” Appl. Opt. 18, 3728
(1979).
45. H. Chen, F. T. S. Yu, and A. M. Tai, “One-Step Rainbow Holography Old and New,”
in Recent Advances in Holography, Proc. SOC.Phot. Opt. Instr. Eng. 215, 162 (1980).
46. S. A. Benton, H. S. Mingace, Jr., a n d W. R. Walter, “One-Step White-Light Transmission
Holography,” in Recent Advances in Holography, Proc. SOC. Phot. Opt. Instr. Eng. 215,
156 (1980).
47. R. D. Bahuguna and F. M. Santoyo, “Simple Rainbow Holographic Technique for Two
Dimensional Transparencies,” Optics Lett. 9 381 ( 1984).
48. R. D. Bahuguna, “Simple Lensless One-step Rainbow Holographic Technique for 2-D
Transparencies, Applied Opt. 24, 61 1 (1985)
49. R. L. Fusek and L. Huff, “Use of a Holographic Lens for Producing Holographic
Stereograms,” in Recent Advances in Holography, Proc. SOC. Phot. Opr. Instr. Eng. 215,
32 (1980).
50. B. R. Brown a n d A. W. Lohmann, “Complex Spatial Filtering with Binary Mask,” Appl.
O p f 5, 967 (1966).
51. W. H. Lee, “Computer Generated Holograms: Techniques a n d Applications,” in Progress
in Optics (E. Wolf, ed.), Vol. 16, North Holland, Amsterdam, 1978.
52. A. W. Lohmann and D. P. Paris, “Binary Fraunhofer Holograms, Generated by Com-
puter,’’ Appl. Opt. 6, 1739 (1967).
53. J. Wyant, “Holographic and Moire Techniques,” in Optical Shop Tesfing (D. Malacara,
ed.), Wiley, New York, 1978.
54. J. S. Loomis, “Computer Generated Holography a n d Optical Testing,” in Recent Advances
in Holography, Proc. Soc. Phot. Opt. Instr. Eng. 215, 59 (1980).
55. L. B. Lesem, P. M. Hirsch, and J. A. Jordan Jr., “The Kinoform: A New Wavefront
Reconstruction Device,” IBM J. Res. Develop. 13, 150 (1969).
56. F. Broder and D. Malacara, “Holographic Interferometer to Test Optical Surfaces,” Appl.
O p f . 14, 2280 (1975).
57. J. C. Fouere a n d D. Malacara, “Holographic Radial Shear Interferometer,” Appl. Opt.
13, 2035 (1974).
58. D. Malacara and S. Mallick, “Holographic Lateral Shear Interferometer,” Appl. Opt. 15,
2695 (1976).
59. E. B. Aleksandrov and A. M. Bonch-Bruevich, “Investigation of Surface Strains by the
Hologram Technique,” Sou. Phys.-Tech. Phys. 12, 258 (1967).
60. A. E. Ennos, “Measurements of In-Plane Surface Strain by Hologram Interferometry,”
J. Sci. Instrum., Ser. 11, 1, 731 (1968).
REFERENCES 209
61. C. Froehly, J . Monnerer, J. Pasteur, and J. Ch. Vienot, “Etude des Faibles DCplacements
d’Objects Opaques et de la Distorsion Optique dans les Lasers a Solide par InterfCrometrie
Holographique,” Optica Acta 16, 343 (1969).
62. T. Tsuruta, N. Shiotake, and Y. Itoh, “Formation and Localization of Holographically
Produced Interference Fringes,” Optica Acta 16, 723 (1969).
63. J . E. Sollid, “Holographic Interferometry Applied to Meassurements of Small Static
Displacements of Diffusely Reflecting Surfaces,” Appl. Opt. 8, 1587 (1969).
64. N. E. Molin and K. A. Stetson, “Fringe Localization in Hologram Interferometry of
Mutually Independent and Dependent Rotations Around Orthogonal, Non-Intersecting
Axes,” Optik 33, 399 (1971).
65. N. L. Hecht, J. E. Minardi, D. Lewis, and R. L. Fusek, “Quantitative Theory for Predicting
Fringe Pattern Formation in Holographic Interferometry,” Appl. Opt. 12, 2665 (1973).
66. N. Abramson, “The Holo-Diagram: A Practical Device for Making and Evaluating
Hologams,” Appl. Opt. 8, 1235 (1969).
67. N. Abramson, “The Holo-Diagram. 11: A Practical Device for Information Retrieval in
Hologram Interferometry,” Appl. Opt. 9, 97 (1970).
68. N. Abramson, “The Holo-Diagram. 111: A Practical Device for Predicting Fringe Patterns
in Hologram Interferometry,” Appl. Opt. 9, 231 1 (1970).
69. N. Abramson, “The Holo-Diagram. IV: A Practical Device for Simulating Fringe Patterns
in Hologram Interferometry,” Appl. Opt. 10, 2155 (1971).
70. N. Abramson, “The Holo-Diagram. V: A Device for Practical Interpreting Hologram
Interference Fringes,” Appl. Opt. 11, 1143 (1972).
71. R. L. Powell and K. A. Stetson, “lnterferometric Vibration Analysis of Wavefront
Reconstruction,” J. Opt. SOC.Am. 55, 1593 (1965).
72. N. Abramson, “Sandwich Hologram Interferometry: A New Dimension in Holographic
Comparison,” Appl. Opt. 13, 2019 (1974).
73. N. Abramson, “Sandwich Hologram Interferometry: 2. Some Practical Calculations,”
Appl. Opt. 14, 981 (1975).
74. N. Abramson, “Sandwich Hologram Interferometry: 3. Contouring,” Appl. Opt. 15, 200
(1976).
75. N. Abramson, “Sandwich Hologram Interferometry: 4. Holographic Studies of Two
Milling Machines,’’ Appl. Opt. 16, 2521 (1978).
76. H. M. A. El-Sum, Reconstructed Waoefront Microscopy, Ph.D. Thesis, Stanford University,
November 1952.
77. H. M. A. El-Sum and P. Kirkpatrick, “Microscopy by Reconstructed Wavefronts,” Phys.
Rev. 85, 763 (1952).
78. A. V. Baez, “A study in Diffraction Microscopy with Special References to X-Rays,’’ 1.
Opt. SOC.Am. 42, 756 (1952).
79. R. F. Van Ligten and H. Osterberg, “Holographic Microscopy,” Nature 211, 282 (1966).
80. E. N. Leith and J. Upatnieks, “Holographic Imagery Through Diffusing Media,” J. Opt.
Sac. Am. 56, 523 (1966).
81. J. W. Goodman, “Wavefront Reconstruction Imaging Through Random Media,” Appl.
Phys. Lett. 8, 31 1 (1966).
82. A. Vander Lugt, F. B. Rotz, and A. Klooster, Jr., “Character Reading by Optical Spatial
Filtering,’’ in Optical and Electro-Optical Information Processing (J. T. Tipett el a/, eds.),
p. 125, MIT Press, Cambridge, Mass., 1965.
83. D. H. Close, “Holographic Optical Elements,” Opt. Eng. 14, 408 (1975).
84. W. C. Sweatt, “Describing Holographic Optical Elements as Lenses,” J. Opt. Sac. Am.
67, 803 (1977).
210 HOLOGRAPHY
85. J. Upatnieks, A. Vander Lugt, and E. N. Leith, “Correction of Lens Aberrations by Means
of Holograms,” Appl. Opt. 5, 589 (1966).
86. R. V. Mueller and N. K. Sheridon, “Sound Holograms and Optical Reconstruction,”
Appl. Phys. Lett. 11, 328 (1966).
87. J. D. Young and J. E. Wolf, “A New Technique for Acoustic Holography,” Appl. Phys.
Lett. 11, 294 (1967).
88. K. Preston, Jr. and J. L. Krenzer, “Ultrasonic Imaging Using a Synthetic Holographic
Technique,” Appl. Phys. Lett. 10, 150 (1967).
89. P. S. Green, ed., Acoustical Holography, Plenum Press, New York, 1974.
90. B. P. Hildebrand and B. B. Brenden, A n Introduction to Acoustical Holography, Plenum
Press, New York, 1974.
91. 8. B. Byron and H. D. Collins, “Acoustical Holography with Scanned Hologram Systems”
in Holographic Nondestructive Testing (R. K. Erfd, ed.), p. 405 Academic Press, New
York, 1974.
92. D. Malacara, “Optical and Electronic Processing of Medical Images,” in Progress in
Optics, Vol. 22, Chap. 1 (E. Wolf, ed.), North-Holland Publ., Amsterdam, 1985.
93. E. N. Leith and J. Upatnieks, “Holography with Achromatic Fringe Systems,” J. Opt.
SOC.A m . 57, 975 (1967).
94. R. E. Brooks, L. 0. Heflinger, and R. F. Wuerker, “Pulsed Laser Holograms,” IEEE J.
Quant. Electr. QE-2, 275 (1966).
95. H. M. Smith, ed., Holographic Recording Materials, Springer-Verlag. Berlin, 1977.
96. J. W. Gladden and R. D. Leighty, “Recording Media,” in Handbook ofoptical Holography
(H. J. Caulfield, ed.), Academic Press, New York, 1979.
97. R. A. Bartolini, H. A. Weakliem, and B. F. Williams, “Review and Analysis of Optical
Recording Media,” Opt. Eng. 15, 99 (1976).
98. M. Chang, “Dichromated Gelatin of Improved Optical Quality,” Appl. Opt. 10, 2551
(1971).
99. R. K. Curran and T. A. Shankoff, “The Mechanism of Hologram Formation in Dichro-
mated Gelatin,” Appl. Opt. 9, 1651 (1970).
100. R. A. Bartolini, “Characteristics of Relief Phase Holograms Recorded in Photoresists,”
Appl. Opt. 13, 129 (1974).
101. R. A. Bartolini, “Improved Development for Holograms Recorded in Photoresists,” Appl.
Opt. 11, 1275 (1972).
102. R. A. Bartolini, “Photoresists,” in Holographic Recording Materials, ( H . M. Smith, ed.),
Springer-Verlag, Berlin, 1977.
103. S. L. Norman and M. P. Singh, “Spectral Sensitivity and Linearity of Shipley AZ-1350
J Photoresist,” Appl. Opt. 14, 818 (1975).
104. J. P. Kirk, “Hologram in Photochromic Glass,” Appl. Opt. 5, 1684 (1966).
105. R. J. Araujo, “Recent Advances in Display Media,” in N A S A Symposium Proceedings,
p. 63, Cambridge, Mass., 1967.
106. D. R. Bosomworth and H. J. Gerritzen, “Thick Holograms in Photochromic Materials,”
Appl. Opt. 7, 95 (1968).
107. A. L. Mikaeliane, A. P. Axenchikov, V. 1. Brobinev, E. H. Gulaniane, and V. V. Shatun,
“Holograms on Photochromic Films,” IEEE J. Quant. Elect. QE-4, 757 (1968).
108. A. S. Mackin, “Holographic Recording on Electron Beam Colored Sodium Chloride
Crystals,” Appl. Opt. 9, 1658 (1970).
109. D. L. Ross, “Photochromic Indigoids. 111: A Photochromic Element Based on the cis-trans
Photoisomerization of a Thioindigo Dye,” Appl. Opt. 10, 571 (1971).
110. J. A. Jenney, “Holographic Recording with Photopolymers,” J. Opt. SOC.A m . 60, 1155
(1970).
REFERENCES 21 1
111. W. S. Colburn and K. A. Haines, “Volume Hologram Formation in Photopolymer

Materials,” Appl. Opt. 10, 1636 (1971).
112. J. Kosar, Light Sensitiue Systems, Wiley, New York, 1965.
113. J. C. Urbach and R. W. Meier, “Thermoplastic Xerographic Holography,” Appl. Opt. 5,
666 (1966).
114. T. C. Lee, “Holographic Recording on Thermoplastic Films,” Appl. Opt. 13, 888 (1974).
115. F. S. Chen, J. T. La Macchia, and D. B. Fraser, “Holographic Storage in Lithium Niobate,”
Appl. Phys. Lett. 13, 223 (1968).
116. J. B. Thaxter, “Electrical Control of Holographic Storage in Strontium-Barium Niobate,”
Appl. Phys. Lett. 15, 210 (1969).
117. F. S. Chen, “Optically Induced Change of Refractive Indices in Lithium Niobate and
Lithium Tantalate,” J. Appl. Phys. 40, 3389 (1969).
118. L. H. Lin, “Holographic Measurements of Optically Induced Refractive Index
lnhomogeneities in Bismuth Titanate,” Proc. IEEE 57, 252 (1969).
119. J. J. Amodei and D. L. Staebler, “Holographic Pattern Sizing in Electro-Optic Crystals,”
Appl. Phys. Lett. 18, 540 (1971).
120. J. J. Amodei, D. L. Staebler, and A. W. Stephens, “Holographic Storage in Doped Barium
Sodium Niobate,” Appl. Phys. Lett., 18, 507 (1971).
121. R. S. Mezrich, “Curie Point Writing of Magnetic Holograms on Mn Bi,” Appl. Phys. Lett.
14, 132 (1969).
122. G. Fan, K. Pennington, and J. H. Greiner, “Magneto Optics Hologram,” J. Appl. Phys.
40, 974 (1969).
123. J. J. Amodei and R. S. Mezrich, “Holograms in Thin Bismuth Films,’’ Appl. Phys. Lett.
15, 45 (1969).
124. S. A. Keneman, A. Miller, and G. W. Taylor, “Phase Holograms in a Ferroelectric
Photoconductor Device,” Appl. Opt. 9, 2279 (1970).
125. S. A. Benton, “Photographic Materials and their Handling,” in Handbook of Optical
Holography, (H. J. Caulfield, ed.), Academic Press, New York, 1979.
126. C. B. Burckhardt and E. T. Doherty, “A Bleach Process for High Efficiency Low-Noise
Holograms,” Appl. Opt. 8, 2479 (1969).
127. J. Upatnieks and C. Leonard, “Diffraction Efficiency of Bleached, Photographically
Recorded Interference Patterns,” Appl. Opt. 8, 85 (1969).
128. M. Lehmann, J. P. Lauer, and J. W. Goodman, “High Efficiencies, Low Noise, and
Suppression of Photochromic Effects in Bleached Silver Halide Holography,” Appl. Opt.
9, 1948 (1970).
129. A. Graube, “Advances in Bleaching Methods for Photographically Recorded Holograms,”
Appl. Opt. 13, 2942 (1974).
130. H. M. Smith, “Photographic Relief Images,” J. Opt. SOC.Am. 58, 533 (1968).
131. R. A. Bartolini, W. J. Hannan, D. Karlsons, and M. J. Lurie, “Embossed Hologram
Motion Pictures for Television Playback,” Appl. Opt. 9, 2283 (1970).
132. R. A. Bartolini, N. Feldstein, and R. J. Ryan, “Replication of Relief-Phase Holograms
for Prerecorded Video,” J. Elecrrochem. SOC.120, 1408 (1973).
133. B. R. Clay and D. A. Gore, “Holographic Moving Map Display,” in Coherenr Oprics in
Mapping, Proc. SOC.Phot. Opt. Instr. Eng. 45, 149 (1974).
134. J. R. Burns, “Large Format Embossed Holograms,” in Applications of Holography, Proc.
SOC.Phot. Opt. Instr. Eng. 523, 7 (1985).
135. 8 . E. Clay, “Two-Dimensional Displays,” in Handbook of Optical Holography (H. J.
Caulfield, ed.), Academic Press, New York, 1979.
5. PHOTOMETRY AND RADIOMETRY
William L. Wolfe
Optical Sciences Center
University of Arizona
Tucson, Arizona 85721
5.1. Introduction
The practice of radiometry deals with the calculation of radiation from

objects of interest, in optical systems, and on detectors, and with the
measurement of various radiometric quantities on both a relative and
absolute basis. The calculation of radiative transfer alone has been the
subject of several monographs (e.g., Chandrasekhar’), and the measurement
of radiation has likewise been the subject of entire books (e.g., Bauer,’ and
Smith et d 3 )This
. can only serve as an introduction to the field in general
and descriptions of several topics of special interest.
All equilibrium radiation problems can begin with a consideration of the
famous blackbody distribution first described by P l a n ~ k .The~ angular,
spatial, and areal characteristics of real bodies can then be expressed in
terms of an efficiency factor, the emissivity, that multiplies a blackbody
function to obtain the true flux characteristics. Several important laws of
radiation distribution and transfer apply to the blackbody idealization, and
most experimental situations can be modeled with only minor variations
from these laws. Accordingly, the first portion of the chapter is devoted to
blackbody laws, partly because many of the variations and forms of this
law are not widely known.
An area of radiation transfer that is becoming increasingly important is
transfer in partially transparent media. The new techniques of sounding the
atmosphere from satellites to infer temperature and of probing media like
molten glass depend upon this theory. It also gives some insight concerning
the nature of Lambertian radiators.
Temperature measurements made with multispectral radiometers have
received new emphasis from the satellite techniques of reconnaissance with
“crude spectrometers.” The techniques have application in areas other than
those of assessing crops and drainage patterns and may prove to be a most
powerful and practical radiometric technique. The section on multicolor
213
METHODS OF EXPERIMENTAL PHYSICS Copyright 01988 by Academic Press. Inc.
ISBN 0-12-475971-8
214 PHOTOMETRY A N D RADIOMETRY
techniques provides some new results or perhaps an enhanced under-

standing of old techniques applied to new problems.
Last and maybe least is the question of nomenclature. There has been
confusion in the field of radiometry. Some of it is due to the many different
fields in which radiometric measurements are made and discussed, and
some of it is apparently due to the fact that photometry preceded radiometry
just as the awareness of light preceded that of infrared and ultraviolet
radiations (Herschel,’ and Ritter6).
5.2. Symbols, Units, and Nomenclature
Table I presents a list of all the symbols used in this chapter along with
their definitions and units associated with them. The units are given not
just to define the dimensionality, but they are intended to help clarify the
meaning. SI units are generally used, although frequently used units like
centimeters and micrometers are indicated rather than meters where it is
appropriate. The symbols are listed alphabetically, but the ensuing dis-
cussion treats them topically.
Perhaps the simplest of the variables to be discussed is the independent
spectral variable. Most of us think in terms of the optical wavelength A of
the radiation to be measured. It is also useful to think in terms of the
frequency which can be expressed in several ways. The wave number i j is
just the reciprocal of the wavelength. The frequency v is the speed of light
times that ( v = c/A), and the radian frequency w is 27rv. Similarly, k, the
radian wavenumber is 27rl or 2 7 r l A . The nondimensional frequency x is
c2/AT (where c2 is the second radiation constant defined below, and T is
the temperature), a convenient independent variable for many calculations.
The spectral distribution of blackbody radiation can be given in terms of
any of these variables. In order to treat this somewhat generally, the symbol
y is used to represent any of these spectral variables.
Radiation transfer is ultimately expressed in terms of a quantity like
power, the time rate of energy. An equivalent concept is the time rate of
the transfer of a number of photons or of the amount of power which evokes
a visual response. Each of these is a flux and is denoted by the symbol P.
To distinguish between them, the subscripts u, q, and u respectively, are
recommended by the International Illumination Society.’ I have not used
a subscript when the symbol can refer to any of these or when it refers
specifically to energy transfer. I believe, no confusion will result. The rest
of the symbols which are used to describe radiative transfer are L, M , E,
I, and S. These are, respectively, the radiance or flux per unit normal area
and solid angle; the irradiance or incidance or incident flux per unit area;
S Y M B O L S , UNITS, A N D NOMENCLATURE 215
the emittance or exitance (emitted flux per unit area); the intensity or flux
emitted per solid angle-generally from a subresolution source; and the
radiance per unit path length in a medium which scatters or is a volume
emitter. This is also called the volume radiance. In several sections I have
dealt with irradiance and electric field strength in the same equation and
have used U for the electric field. These equations have also involved
m U U * for which I have used 9.I would prefer using I and E for these,
and J and H for intensity and irradiance, but that would be contrary to
the recommendations of the International Commission on Illumination
(CIE). Several writers have proposed nomenclature schemes to make these
definitions logical, consistent, and easy. Table I1 is a summary of these.
The fluometry scheme of Jones has as its main feature the unifying concept
that these flux terms are really all geometric concepts. He therefore used
sterance, incidance, exitance, intensity, and sterisent as generic terms to be
modified by “radiant,” “photon,” and ‘‘luminous.’’ Nicodemus,’ largely in
an effort to overcome the ambiguity of “intensity” (it is variously used to
mean E, M , I, and L by different scientists), has coined the geometric terms
of areance, and pointance for E or M and I, respectively. Worthing,” years
ago proposed that “-ancy” be used to mean “per unit area.” He went further
and proposed that the transmission, absorption, and reflection properties
be specified with “-ance” endings when a sample is specified, and “-ivity”
endings when a substance is specified.” This, of course, makes emittance
an ambiguous term if one uses the “conventional” terms. The nomenclature
system has not yet been proposed. The concept of path radiance may be
somewhat foreign to some readers. In a scattering medium or volume
radiator the radiance is not constant. It can increase with path length,
because each element of path contributes more radiation to the total. It
thus has units of W cm-* sr-’ cm-I, where the final cm-’ is the (reciprocal
of the) unit of path length. There is no recommended symbol for this. I
have chosen S for this quantity because it agrees with “sterisent,” and many
other appropriate letters have been preempted. I have also replaced with
P (to stand for phlux) for ease of typing. Finally, to indicate any of these
radiometric quantities, I have used the symbol R.
The atomic constants are listed in a conventional way: cI and c2 are the
first and second radiation constants, equal to 27rc2h and hc/ k g , respectively.
The speed of light in vacuo is c; Planck’s constant, h ; the Stefan-Boltzmann
constant a ; that for photon distributions, aq;and the Boltzmann constant
k g (because k is the radian wave number).
Some of the common symbols are x, y, and z for general distances. These
do not get confused with the spectral variables. Angles are denoted by 6
and 4 ; solid angles by R; areas by A; time by 1 ; temperature by T ; integers
by m, P, 4.
216 PHOTOMETRY A N D R A D I O M E T R Y
TABLE1. Symbols, Units and Nomenclature
Symbol Definition Units
a, b, c constants -
A area mZ
B effective bandwidth Hz
BRDF Bidirectional Reflectance Distribution Function sr '
b base of a prism m
C constant -
C speed of light in a vacuum m s-'
CI first radiation constant w cm2 s-I
c2 second radiation constant cm K
D diameter m
D* specific detectivity cm Hz'/'W I
E irradiance or incidance w cm-:

F focal ratio -
f focal length o r area factor m or -

constant ( 2 or 4 ) -
g
h Planck's constant or height w s2
I intensity W sr-'
K luminous efficacy ImW '
Kt, radiation interaction factor -
k radian wavenumber J K-'
k, Boltzmann's constant -
L radiance Wcm 'sr-'
I length m
M radiant emittance or exitance Wcm '
m. P, 9 integer constants -
N number of photons -
NEFD noise equivalent flux density W cm-'
NEAL noise equivalent change in radiance W cm-'sr-'
n refractive index -
P power or flux W
R general radiometric variable or range various
9 responsivity VW-' or AW-'
r radius m
S volume radiance W cm-'sr-'
SNR signal-to-noise ratio -
T temperature K
I time S
U electromagnetic field quantity ( E o r H) v m-' or v
U normalized nondimensional frequency -
V volume or voltage or voltage ratio m' or v or -
V(A) visibility curve -
u velocity m SCI
W width m
X nondimensional frequency or coordinate length - or m
Y general spectral variable or coordinate length various or m
I coordinate length m
SYMBOLS, UNITS, A N D NOMENCLATURE 217
TABLEI. (coni.)
Symbol Definition Units
a absorptivity or absorptance
Y absorption coefficient
E emissivity
1) quantum efficiency
1)" luminous efficiency
8 angle
A wavelength
v frequency
v wavenumber
P reflectance or reflectivity
U Stefan Boltzmann constant
u(A) slit function
T transmittance or transmissivity
@ angle
R solid angle
w radian frequency
Subscripts
a, b indicating different spatial points
a atmosphere
C collector or optical element
d detector
e emitted
h hemispherical
I incident
q photon
S source
U energy
V visible
1.2.i first, second ith of a group
W infinite (number of reflection)
Superscripts
BB blackbody
h hemispherical
m maximum, molecular
0 optics
P projected
time rate
- weighted average
TABLE11. Radiometric Nomenclature
~~ ~ ~
Field
Quantity Jones Nicodemus Worthing Theorists Meteorologists
Flux Density [Wm-*] Areance Intensity Flux Density

>
Incident Incidance Incident Areance Radiancy - - z
Exiting Exitance Exitent Areance Irradiancy - - 0
Flux per solid angle [w sr-'] Intensity Pointance - - -
Flux density per solid

angle [w sr-' m-*] Sterance Sterance - - Specific Intensity
Volume Radiance Sterisent - - - - rn
-I
P
<
FORMULAS FOR BLACKBODY RADIATION 219
5.3. Formulas for Blackbody Radiation
The derivation of the Planck equation can be based on the product of

the function which gives the density of states available for photons and the
average number of photons in a state. The expression is derived in terms
of the radian wave number k ( = 2 r / h ) . The number of photons nk per
unit volume with wave number between k and k + d k is given by
nkdk = NkdkV-' = k 2 K 2 ( e "- l ) - ' d k [m-3], (5.1)
where
V = volume
x = c2/hT
c2 = hc/ k ,
h = Planck's constant
c =velocity of light
k , = Boltzmann's constant
Notice that the subscript notation here indicates a partial derivative. The
quantity nk is the photons per unit volume per unir radian wave number.
Other forms of the Planck equation are obtained in subsequent sections.
5.3.1. Geometric Distributions

It can be shown12 that the radiant exitance and radiance are related to
nk by the following:
Mqk = (f) nk
(5.3)
These are purely geometric relationships and apply to other spectral distribu-
tions. One important and very useful characteristic of the radiation is that
the geometrical characteristics are independent of the spectral distributions.
There are basically four spectral distributions: power vs frequency, power
vs wavelength, photons vs frequency, and photons vs wavelength.
5.3.2. Generalized Spectral Distributions

The spectral distributions can be written in terms of the concentration
of flux and the sterance. The different spectral forms of the basic four are
PA= 2 r ~ ~ h A A - ~-( 1)-'

e" [W CLm-lI, (5.4)
(5.5)
(5.7)
Other variations involve different arrangements of the atomic constants,

and different powers of a frequency or wavelength variable. They can be
written in general as
R , = Cy*'"(ex - l)-', (5.8)
where
R is any radiometric quantity
y is any spectral variable
rn is the proper power of the spectral variable
C is a constant made up of atomic constants.
In this formulation, when y represents a frequency variable, the positive
sign applies, and when y represents a wavelength variable, the negative
sign does. Then expressions for the maximum can be derived for the general
case and applied to specific problems. The result is that the basis for finding
the maximum of any spectral distribution is to find the value x, for which
(5.9)
The usual expression for the distribution of power vs wavelength can then
be found from
c2 hc hu
x=-=---
AT A k R T - kBT'
and
c 1.43883
AmT = L = - - - cm K =2898 [ p m K]. (511)
Xm Xrn
This is the common expression for the Wien displacement law." The
constants for the other forms are given in Table 111.
The distribution can also be expressed usefully in terms of the dimension-
less variable x by applying the technique described above:
R, = 2ch - x'(e" - 1)-' (5.12)

( C 3
FORMULAS FOR BLACKBODY RADIATION 22 1
TABLE111. Maxima of the Dependent and Independent Variables for Different

Isothermal Planck Spectral Distributions
Function
Dependent Independent
Variable Variable m Xmax Rmax
Photons Y 2 1.593624260 0.6476

Power Y 3 2.82 1439372 1.4214
Photons A 4 3.920690395 4.7796
Power A 5 4.965 11423 21.2036
Power contrast A 6 5.96940917 115.9359
The relative value can be written as

R,
-- x3(err- l ) - ' (ex"'- 1)
- = u3 (5.13)
Rx,,, x ~ ( e x m - l ) - ' (e'-l) '
Any of the distributions will be of this form but with a different exponent
on the relative dimensionless frequency u (=x/x,).
The radiance can change either as a result of a change in temperature or
a change in the spectral variable.
Here it can be seen that the factor xex(e' - 1)-' is the factor for spectral
changes. Each variation individually can be viewed as an exponential
relation, as pointed out in part by
R,. = T P , (5.14)
where
p = xe'(e'- I)-'
and
R, = y 4 ,
where
* + xe '( e* - 1
q = rn )-I.
The temperature-variation portion of this development is described and

discussed by DreyfusI4 and several applications are given.
The reader can find descriptions of the properties of blackbody radiation
in several texts (Hudson"; Holter et al."; Bramson18; Kruse et al."; and
Kingston").
A general treatment follows for the integrals of these functions in terms
of the dimensionless variable x.
5.3.3.Generalized Integrals
The flux and change in flux over a finite spectral band are found by
integrating the appropriate monochromatic function over the spectral band.
The functions can be the radiance, photon sterance, or their respective
changes with temperature:
(5.15)
(5.16)
(5.18)
The finite integrals of the Planck function can be found by expanding

+
(e" - 1)-' or (e" - 2 Kx)-'in a series. The results of these expansions are
given below.
rn-4ee-m"[(rn~)3+3(rnx)2+6rnx+6],
e~mx[(rn~)2+2m~+2],
=-(-)
2c2h T 1 rn~4e-"A[(rn~)4+4(rn~)3+12(rnx)2
T c2 m;l
+24rnx+24],
C rn-' e-mx[(mx)3+3(rnx)2+6rnx+6].
5.3.4.Infinite Integrals
The value of the integral of a blackbody function is of course independent
of the spectral variable, and the simplest variable to use is the non-
dimensional frequency x. Then the total radiance is
L = 2c2h (3'x 4 ( e x - I)-' dx
(5.19)
SIMPLE RADIATIVE T R A N S F E R 223
For photons, one has
L, = 2 c ( z)3 x3(ex- 1)-' dx
2kiT' T'
=2.4041-- - 1.5202 x 10" -
c2h3 ?l
u T'
--.
- (5.20)
?l
Both of these are solved by resourse to zeta functions. This latter expression
can also be written as
L (5.21)
' = &
2.15 k B T '
This would indicate that the average value of the energy of a photon is
2.15 k B T .
5.3.5.S o m e Useful Blackbody Curves

Tabulations of various blackbody functions are presented in various
texts.20These generally include the distribution of power and photons with
respect to wavelength and the integrals of these functions from A = 0 to the
tabulated wavelength. Some also give tabulations of the change in the
radiant emittance with respect to temperature. None of those I have found
do the same for photon distributions. Computers make easy the calculation
of the appropriate quantity of Sections 9.3.2 and 9.3.3, and they make
tabulations of these values essentially unnecessary. In the spirit of providing
approximate values for ready access, curves are given in Figs. 1-8. These
include: the relative spectral distribution of photons and flux, their relative
cumulative distributions, the relative distribution and relative cumulative
distributions of their temperature, all as a function of the normalized
nondimensional frequency.
5.4. Simple Radiative Transfer
This section deals with the radiative transfer in nonscattering media. The
primary purpose is to describe the characteristics of the radiometric variables
rather than the physical processes which give rise to absorption and emission.
The first part of the description then applies to transfer in nonabsorbing,
nonscattering, nonemitting media, and the second to absorbing and emitting
but nonscattering media.
8.9
0.8
0.7
8.6
8.5
0.4
0.3
8.2
0.1
0 1
0 1 2 3 4 5
NORNALIZED NONDIMENSIONAL FREQUENCY
FIG. 1. Normalized blackbody energy distribution as a function of normalized non-
dimensional frequency of the radiation.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
NORMALIZED NONDIMENSIONQL FREQUENCY
FIG. 2. Normalized blackbody photon distribution as a function of normalized non-
S I M P L E RADIATIVE TRANSFER 225
1 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2
0.1
0
NORFlALIZED NONDIMENSIONAL FREQUENCY
FIG.3. Normalized blackbody energy contrast distribution as a function of normalized
nondimensional frequency o f the radiation.
0.9
0.8
0.7
0.6
e. 5
0.4
8.3
0.2
0.1
0
NORNALIZED NONDIIIEHSIONAL FREQUENCY
FIG. 4. Normalized blackbody photon contrast distribution as a function o f normalized
nondimensional frequency o f the radiation.
226 P H o r o M E r u Y A N D RADIOMETRY
8.9
9.8
0.7
0.6
0.5
0.4
0.3
8.2
0.1
NOREIALIZED NONDIHENSIONAL FREQUENCY

FIG.5. Normalized blackbody energy integral as a function of normalized nondimensional
frequency of the radiation.
0.9
0.8
0.7
8.6
0.5
0.4
0.3
8.2
8.1
8
0 1 2 3 4 5 6 7 8 9 18
NORMALIZED tiONDIMENSIONAL FREQUENCY
FIG. 6 . Normalized blackbody photon integral as a function of normalized nondimensional
frequency of the radiation.
0.9
0.8
0.7
8.6
6.5
0.4
0.3
1:: / * , . . . . .
0 ~
1 -
0.9 .
0.8 -
8.7 -
0.6 -
0.5 -
0.4 -
0.3 -
:::t/
0
B 1 2 3 4 5
. . . . . . .
6 7 8 9 10
,
NORFIALIZED NONDIMENSIONAL FREQUENCY

FIG.8. Normalized blackbody photon contrast integral as a function of normalized non-
5.4.1. Transfer in a Vacuum

The basic equation of radiation transfer involves the radiance of the
source, the areas of the emitter and receiver, and the distance between them.
Two general surfaces, A , and A 2 ,shown in Fig. 9, have on them differential
elements dA, and dA2.The line-of-centers distance between these elements
is x, and the angles made between the normals and x are 8 , and 02. Then
the amount of flux dPI2radiated from dA, and received by dA2 is given by
L dA, cos 0 , dA2 cos O2
dPI2= =L dA, cos 0, d R 2 ,
X2
The expression on the right was written four different ways to illustrate the
different interpretations which have been applied to it. The first can be
viewed as a variation of the definition of radiance: since radiance is power
per projected area and solid angle, then power is radiance times projected
area and solid angle. The second form shows that the solid angle can be
written as a projected area divided by a distance squared. This form gives
a physical and symmetrical expression for the transfer: the product of two
projected areas divided by the square of the distance between them. The
third form just indicates that in this chapter (and other places in the
literature) the projected area is indicated with a subscript. The final form
uses the “p” to indicate a “projected solid angle.” This is a nonphysical
quantity, which when multiplied by an area and a radiance gives the flux
transfer.
The expression can be used to obtain the relation between radiance and
flux density for Lambertian bodies and the very useful configuration factor
that is found in extensive tables; see Sparrow and Cess2’ for more on this.
The relationship between Land M (regardless of whether photons, power,
or any other quantity is being transferred) is M = TL. One can imagine that
a differential area of a radiator which has the same radiance in all directions
FIG. 9. The geometry of radiative transfer. Radiation travels from a differential element
dA, of surface A , along R , , to dA, of A,. The angles between R , , and the surface normals
are 0, and 0,.
S I M P L E RADIATIVE T R A N S F E R 229
is at the center of a sphere. The flux per unit area d A , of this source radiated
into a hemisphere is given by
L d A , cos 8, dA2 cos e2
M=-= (5.23)
dA I x’dA,
The radiance is independent of angle, and O2 is zero, so that
M =L lo2=cos 8 sin e de d 4
= TL. (5.24)
It should be noted that the roles of emitter and receiver can be reversed
either by use of Helmholtz’s reciprocity theorem or by direct integration.
This relationship can be used in connection with the flux transfer theorem
to obtain a configuration factor (also called angle factor and view factor).
On a differential basis, one has
L I 2dA, cos 8, dA, cos O2
d R I 2= (5.25)
X-
For a Lambertian source MI2= 7rLI2.Then

dA, cos 8, dA2 cos tI2
d P l z =M (5.26)
TX2
The configuration factor K I 2is the fraction of radiation emitted by the first
surface that is received by the second. Thus,
dP12- ( MI2dA, cos 8, dA2 cos O2)/x2

&Al - dA2 - 9
dP, MI2 dA,
(5.27)
The configuration factor is a function of the “tilt” of the radiating surface

and the solid angle subtended by the receiving surface.
View factors for surfaces of a finite size are more complicated. The
definition is still the same, but integration over the source, the receiver, or
both must be performed. The configuration factor for a differential element
dA, and an extended area A 2 is obtained as follows:
Extended sources, of course, are not the only type of radiator one
encounters in practice. The second major class is generally called “point
sources,” although in theory there is no such.thing (at least one that radiates
appreciably). What is usually meant is a subresolution source, one which
is small compared to the image of the detector on the source. For such a
source, radiance is not a good descriptor, but the integral of radiance over
the area of source is appropriate. Such a quantity is usually called intensity
and labeled I by radiometrists.
It can be defined in two ways:
(5.29)
Then the transfer is written as

I , ? dA, cos O2
dP,, = (5.30)
XZ
The radiance is a function of both the direction and the location on the
surface, whereas intensity is a function only of direction.
Before considering transfer in media that absorb but have inappreciable
emission, it is useful to consider two brief calculations involving flux on
the detector of a radiometer. The first calculation involves an extended
source, the second a subresolution one.
Assume that a radiometer views an extended source of area A , at a
distance x with radiance L. The radiometer has a collector area A , and
focal length f, detector area A , optical speed or Flnumber F, and trans-
mission r. Then, if all solid angles can be approximated by areas divided
by distances, the flux is given by
(5.31)
This can be reformulated in terms of solid angles and Flnumbers:
(5.32)
The total flux o n the detector must be obtained from one of the last two
expressions. The last expression on the right is in terms of the collector
area and the field of view of the detector. To keep a constant flux on the
detector, these two quantities must be varied inversely. Usually one wants
as small a detector field as one can get, but this must be accompanied by
a concomitant increase in collector area (even without diffraction consider-
ations). The penultimate expression shows that the flux concentration on
S I M P L E RADIATIVE T R A N S F E R 23 1
the detector is independent of the diameter of the optics and is a function

of transmission and F l n u m b e r . The total flux increases with detector size
but only until the source becomes subresolution.
5.4.2. Transfer in Transparent, Nonscattering Media

The basis of the relationship that extends the transfer of flux from source
to receiver a n d then t o detector is that radiance through a n optical system
is constant (Nicodemus,” Rainwater”). The simplest extension of the
transfer theory from a vacuum to real media is the consideration of a
medium which absorbs, has a refractive index different from 1.0, a n d has
n o measurable emission. The first change is that the radiance is n o longer
conserved. Rather, it is L / n 2 , the radiance divided by the refractive index
squared, which is a constant. So the product of L K 2 a n d n 2 A R , which
gives the flux, is still the same a n d still constant-except for any transmission
losses. These transmission values can be applied to the optical elements,
the overall system, a n d the intervening atmosphere T,,. In fact, the flux on
the aperture of the radiometer in such a situation is given by
P, = T,LA,R,, = T,LA,R,, . (5.33)
The flux on the detector can be written as
(5.34)
Careful practice results in accurate determination of T,,, A d , a n d a,,
because these are all properties of the radiometer a n d are readily accessible
(but sometimes hard to measure). In many cases, the atmospheric trans-
mission is very hard to measure. Consequently, many investigators report
not only their estimate of L but their estimate of T,L, which can be
determined much more accurately. The quantity T,L is dubbed the apparent
radiance.
5.4.3. Properties of Partially Transparent Materials

Partially transparent materials are bulk radiators. Every portion of the
volume of the material contributes in a significant way to the radiation
measured at a distance from the surface. Understanding of the phenomena
is important for certain technical reasons as well a s for a n improvement in
the understanding of what are usually regarded as surface radiators. Some
of the main contributors to this work have been G a r d ~ n McMahon,2’
,~~
and Czerny a n d Genzel.26 This discussion is based on the treatment of
Gardon.
FIG. 10. Geometry of radiative transfer in a volume radiator, shown as a side view. An
annular volume element radiates a distance x/cosO to the surface.
A partially transparent, semi-infinite body at a uniform temperature T

and with an absorption coefficient y, and a spectral path radiance S, sterisent
[W m-3 sr-' pm-'] or just S is the subject of consideration. The radiance
from the surface of a medium with a uniform absorption can be obtained
from the path radiance by treating a differential element of volume as shown
in Fig. 10. The result is that the surface radiance L of a slab of infinite
lateral extent is given in terms of the volume or path radiance S as
S
L=-. (5.35)
Y
The basic assumptions in this derivation are that the exponential law of
absorption is valid and that the volume emission is the same in every
direction (an isotropic radiator). Integration over the hemispherical projec-
ted solid angle yields
7TS
M=-. (5.36)
Y
From the earlier discussion of radiance in media of any refractive index n,

SIMPLE RADIATIVE TRANSFER 233
it is clear that (with a little algebraic manipulation)
(5.37)
This gives the volume radiance in terms of the absorption coefficient y and
refractive index of the medium and the “standard” surface radiance. The
external radiation will be in a solid angle that is larger by n 2 than the
internal angle and is reduced by the surface reflection losses. Further,
radiation cannot exit at angles which exceed the critical angle.
If the partially transparent medium is of a finite rather than semi-infinite
thickness denoted by X , then the integration yields a radiance given by
L = n2~BB(1- e - ~ . ~ / ~ ~ s @ ) . (5.38)
As Gardon points out, the radiance from the semi-infinite plate follows
Lambert’s law, whereas that for the finite plate does not.
Several interesting characteristics of radiation in partially transparent
media can be developed from this beginning: the expression for radiation
from plane parallel plates, for directional and hemispherical emissivities,
Bouguer’s law for finite radiators, and the technique of inversion applied
to the atmosphere and to radiance solids like glass.
Integration of the radiance over the hemispherical projected solid angle
yields the following:
M = n 2 ~ n B [ 2 - e - y ‘ ( 1- y r ) - ( y r ) * ~ i ( - y t ) I , (5.39)
where
This exponential integral is to diffuse radiation what the simple exponential

function is to directional radiation.
Gardon shows further that the directional and hemispherical emissivities
are given respectively by
and
Eh*:I= &(ei)sin 6, cos ee de,, (5.41)
where
ei = angle of incidence
ee= angle of emission
The quantities r and p are the directional transmittances and reflectances

of a single surface; the subscripts indicate the state of polarization (s
perpendicular and p parallel); the derivation is (as they say) straightforward
but tedious, making repetitive use of Fresnel’s equation at each surface for
reflecting the volume radiance contribution. This result is then compared
to the radiation from a blackbody.
The transmission, reflection, and absorption of a plane parallel plate can
be done in various degrees of detail that can take into account polarization,
self-emission, scattering, etc. Several of these situations are discussed below;
any special situations can be obtained by extensions of these results.
The first situation encompasses plane radiation incident normally on a
plane parallel plate of material which absorbs but whose reemission is
negligible. Then the total reflectivity, taking into account an infinite number
of multiple reflections, can be written
pe = p + ( l - p ) ’ p e-2yx+ ( I - p ) ’ p 3 e - 4 Y X + . . . . (5.42)
This is an infinite geometric series that can be written as
p a ? = p + ( l - p ) ’ p e - * Y X ( 1 - p 2 e-2y.r)-’. (5.43)
The nomenclature is somewhat simpler if the single-pass internal transmitt-
ance e-”‘ is written as 7.Then
pm=p+(1-p)2p?2(1-p 2 72 ) - 1 . (5.44)
Notice that when r = 1 ,
(5.45)
Similarly the transmittance can be written

7,= (1 - p ) 2 7 + ( 1 - p ) 2 r 3 p 2 + ...
= ( l - p ) ’ T ( l - p 2 72 ) - I .
(5.46)
Noticethatwhenr= 1, ~ ~ = ( l - p ) / ( l + p ) = 2 n / n ’ + l .Inthesimplestcase,
this is a collimated beam both into and out of the sample. Therefore, power
in the beam must be conserved. That which is not reflected or transmitted
must be absorbed:
(5.47)
The proof that a , = E, for a partially transparent material requires the

calculation of E ~ the
, emittance of the partially transparent plane parallel
plate divided by that of a blackbody.
SIMPLE RADIATIVE TRAN SFE R 235
One can apply the same techniques of integration, but for the sake of
variety it will be done in terms of flux density rather than radiance. In a
small solid angle dR, the flux density is (1/47r)SdxdR in a small element
dx. This quantity multiplied by the transmission is integrated to the surface
--SdR
- (1-T). (5.48)
47rY
This is the radiant exitance just inside the surface; outside, it is
(5.49)
If the body is opaque, then T = 0, and Kirchhoff's law provides the following
result:
1dR MHH
M =- (1 - p ) =- ( 1 - p ) dR. (5.50)
4ffY 7r
From this, one can write that S / y = 47rM BB. Therefore, the nonreflected
volume radiation gives rise to one component of radiant exitance:
M I = ( 1 - p ) ( 1 - T ) M B B r - 'd R . (5.51)
That which is reflected from the top and the bottom and emerges again is
M2= P ~ T ~ M , . (5.52)
After one more trip, one has
M 3 = p27'M2 = p 4 r 4 M I . (5.53)
The sum of the infinite geometric series is
00
C M , = M,7r-'dR = ( 1 - p ) ( 1 - T ) M ' ~ T - 'dR/(1 - P ~ T ~ ) . (5.54)

i=O
The emissivity is the same as the absorptivity, and Kirchhoff's law applies
to semitransparent substances.
There is one difference between this development and McMahon's: the
radiant exitance is derived from a consideration of radiance flowing into a
projected hemisphere. This changes the relationship between volume radi-
ance and radiant exitance by a factor of 2.
5.4.4. Path Radiance

A special application of the radiometry of partially transparent media is
the flux one obtains at the top of the atmosphere in a relatively well-
collimated beam. If L, is the radiance of the ith element of the atmosphere,
and L"" is the blackbody radiation of that element, then
L, = &,LY". (5.55)
If T , is the transmission of the ith element, then the radiation at the top of
the atmosphere is
BB
&,L, T,-17,-2 ' ' * 71.
If the atmosphere does not scatter, then

&, = (Y, = 1 - 7,. (5.56)
The difference in transmission between the top and the bottom of the ith
element is
d T , = ( l - ~ , ) 7 , - , ~ , T- ~~ . .. . (5.57)
Therefore the contribution of the ith element is
LY" dr,.
The total beam radiation then is
IhoL"" dT.
This is the path integral, emphasizing the variation of transmission along

the path but with a simplified geometry.
5.4.5. Atmospheric Sounding Techniques

There are two techniques by which a radiative inversion can be accom-
plished: nadir and slant path. The radiative transfer calculation in a partially
transparent medium consists of determining the amount of radiation at
some point for known temperatures and quantities of the constituents along
the path. The inverse calculation is to determine the temperature and
quantity of the constituents along the path from the radiance at the end of
the path, usually in several spectral bands. These techniques have been
reviewed by Wark and Fleming,27by Hanel and Conrath,28and by Gille.29
A short description of the techniques is given here to illustrate the radio-
metric principles involved.
The temperature profile can be determined by measuring the irradiance
at the aperture of a satellite instrument in several narrow spectral bands in
the absorption region of a gas whose mixing ratio is known. Two atmospheric
gases meet this requirement: C O , and 0 2 .Kaplan3" proposed using the
667 cm-' (15 pm) band of C 0 2 for this purpose, and experiments have been
performed based on that ~uggestion.'~
Once the temperature profile has been established, other spectral bands
can be used for determining the profiles of other gases. The radiative transfer
equation for monochromatic radiation can be written in terms of the radiance
at a given frequency F and zenith angle 8 as'
dL(F, e)={-L(F, e ) + LBB[V,T(z)])yrn(V,z) p( z) secedz, (5.58)
where
T (z) = atmospheric temperature at z

p ( z ) = atmospheric density at z
z =height at local vertical.
The equation can be recognized as the one derived earlier. In this case,
however, the path differential is sece dz and the absorption is p y m , because
the product of the gas density and molecular absorption coefficient gives
the overall absorption coefficient. The assumptions include a plane parallel
atmosphere in which there is no scattering or refraction and which is in
local thermodynamic equilibrium. The term L(V, 6 ) is the radiance at the
lower boundary; the minus sign arises from the integration convention. If
the instrument views a narrow cone, vertical to the surface of the earth,
then sece = 1. The independent variable can be converted from altitude to
the pressure at that altitude:
L(F,O)= E(F)LBB(F, T ) r ( F ) -
I,' d(F' )'
LBB(F' T , d(1og P)
The radiance at the top of the atmosphere L(F,O) is found from the
d(log P ) . (5.59)
contribution oftheground E ( P ) r ( F)LBB(F,T), where r ( P) is the transmission

of the entire path from the ground to the top of the atmosphere, and the
path radiance expressed by the integral. The limits specify that the integra-
tion is from the top of the atmosphere down to a height at which the pressure
is p. It is worth writing this one more time and somewhat more compactly.
For each wavelength or wave number, the transfer equation is
Lo=&rL-
I L
dr
d(logP)
d(1og P ) .
An iterative inversion approach is then used. One tries a reasonable T(z)

(5.60)
in the expression for each of the L( V ) until a best fit is found to the measured
values of Lo( F). Various techniques have been used to obtain convergence."
A second way to accomplish atmospheric sounding is to use a long horizon

path, and, in fact, a series of paths none of which intersect the earth itself.
Inversions are done on each of these paths, and an entire distribution of
temperatures and constituents can be generated. For this horizon case, there
is no earth contribution, so the radiative-transfer equation is
T(Y, P)
L( a ) = LRR(a, T ) dT (a, P ) . (5.61)
The integration proceeds from a no-atmosphere case (7= 1) to the full-

atmosphere transmission case. Instruments used for this technique need to
have narrow fields of view to attain good vertical resolution. consequently,
they need relatively wide spectral bands to establish sufficient signal-to-noise
Inversion
R I d iorondr
Temperature (OK)
FIG. 1 1 . Temperature profile derived from IRIS data by inversion of the radiative transfer
equation; data were recorded near Brownsville, Texas, 22 April 1969, at 1937 GMT. Data
from a radiosonde ascent at 1800 GMT are shown for comparison.
RADIOMETRIC TEMPERATURE MEASUREMENTS 239
ratio. Accordingly, an integration using the radiometer spectral responsivity

is required.
The inversion procedure most often discussed which uses this technique
is the "onion peel" approach. Inversion is carried out for the highest layer;
then successive layers are added, and each time a new inversion is accom-
plished. Thus, inversions for a series of almost horizontal lines are obtained.
The data need only to be rearranged to get the vertical profiles.
The radiometric considerations involved in this technique include the
choice of the right gas, a set of spectral lines that are narrow enough and
far enough part for good inversion but wide enough to encompass enough
photons-and, of course, the design of an instrument which provides plenty
of throughput. One result of the vertical sounding techniques is shown in
Fig. 11.
5.5. R ad iomet r ic Te m perat ure Measure ment s
A radiometer can be used to measure the temperature of remote objects

in several different ways: total radiation, radiation in a narrow spectral
band, and by fitting the blackboy curve to two or more points. This section
includes discussions of some of the accuracies and techniques by which
these different measurements can be accomplished.
5.5.1. Radiation Temperature

One measures the radiation in as wide a spectral band as is instrumentally
possible with a detector which is as spectrally flat as possible in order to
obtain a radiation temperature. The output voltage is given by
V= lorn
~ ( A ) E ( A ) T ( A ) c , A - ' ( ~-" I)-' dA. (5.62)
We assume that 9,the responsivity, includes in it any necessary constant

geometrical factors. The usual and simplest analysis includes assumptions
that & ( A ) = T ( A ) = 1 and % ( A ) = 9.Then
V = aT4%.
Therefore,
T=(&)"4. (5.63)
The radiation temperature is the temperature of the blackbody which gives

the same flux as the body being measured. We can now deal with successively
more complicated situations: a constant emissivity and infinite spectral
band, constant emissivity but finite spectral band, etc. For constant
emissivity and infinite spectral band, the expression is
V=9 ~ a T
T=(&)"' (5.64)
One is usually in possession of the values for 9 and V but must assume
that E is 1. Then the relative error is
1 - &-1/4
The error is different for different values of E and is negative for positive
values of E. This means that the error is such that a higher temperature is
inferred than is really the case.
The next complication is that 9 ( A ) = T ( A ) = & ( A ) = 1, but that the band-
width is finite. Then the relative error is
1-
IA, c,A-'(e" - l)-' dA
(5.65)
aT4
A further complication is that % ( A ) and T are constant. One can usually
measure 9(A ) to sufficient accuracy. Then
V= J A.4
~ "l)-' dA.
~ ( A ) C , A - ~ (- (5.66)
In principle, the temperature can be found because this is an integral of

the product of two functions which are known. The only errors will arise
from inaccuracies in the determination of % ( A ) and the measurement of
the voltage. If, in addition, & ( A ) and T( A ) are not constants and are unknown,
then the error is
(5.67)
The effects of E and T can be combined into a single product spectral

function for most purposes. Then the error is
1 - (E)-I (5.68)
where is the effective epsilon-tau product.
RADIOMETRIC TEMPERATURE M E A S U R E M E N T S 24 1
5.5.2. Brightness (Radiance) Temperature

This simple measure infers temperature from a radiance value in a narrow
spectral band:
LA dA = & ( A ) c I r - 1 A - 5 ( e X- 1)-' dA. (5.69)
Such a measurement is made in a narrow spectral band but not in an
infinitesimal or square one. Errors arise from the fact that the narrow band
has a shape, that & ( A ) is not known, and that there is some uncertainty in
picking dA. If the Wien approximation ( x >> 1) can be used and & ( A ) = 1, then
L, = c,T-'A-' e-x, (5.70)
T = cz/ A In( c, r - ' h -'). (5.71)
If the line is chosen in a region in which the Planck law must be used, then
(5.72)
The error is given by

ln(9&clr - l / A V+ 1)
1- (5.73)
ln(9c,.rr-'/ASV+ 1) .
This expression is not in an easily used form and requires specific details
for evaluation.
5.5.3. Ratio Temperature

The simplest form of a two-color or two-wavelength or multispectral
technique is just a measurement of the ratio of the outputs in two narrow
spectral bands. This voltage ratio V is given by (in the region where the
Wien law holds):
(5.74)
then
(5.75)
In the region for which the Planck function must be used, the temperature
cannot be found explicitly. The voltage ratio is
(5.76)
Generally one chooses the two spectral bands such that = E ? , but only a
knowledge of the ratio is necessary. The measurements are usually made
through an atmosphere of transmission T, so that an ET product in each
band should be used instead of just E . Then, since xI= c2/A I T = x2 = c2/A2T,
there is only one value of T which satisfies the set of conditions.
The errors resulting from an inaccurate assessment of the ratio of the ET
product in the two bands can be made by assuming the ratio is 1 and by
calculating the T values for other values of the ratio. The error can be
evaluated by calculating the value of T, using the proper ET ratio, then
calculating T for an ET ratio of 1. The relative difference between the two
temperatures is the error. As usual, the expression can be formulated in
terms of the dimensionless variable x. The solution, however, is a function
of both x1 and x2, not the ratio or the sum.
The analysis for finite spectral bands proceeds in almost the same way
as for the monochromatic analysis. The radiation received in each of the
spectral bands is a combination of the measured spot and any background
in the field of view. Now, however, for the radiances, one must write integral
expressions. Thus, for a spot of area A
V=
A [ E T ( A ) A - ~ ( ~ "l)-'
- dA + ( 1 - A )
I,, & ~ ( h ) A - ~ ( el ') - ' dA
A f'
Ah2
E T ( A ) A - ~ ( ~-" 1)-' dA +(1 - A )
I,,,' E T ( A ) A - ~ ( ~-' I ) - ' dA '
(5.77)
where
V = voltage ratio
Ah I = first spectral band
Ah2 = second spectral band
XI = C 2 / A / TI
XI = CI/A/TZ
It is assumed that a quantum detector is used so that photon flux is expressed.

The emissivity & ( A ) and the transmittance can be removed from the integrals
by the use of effective emissivities as follows:
-
[ ~ ( h ) A - ~ ( e " - l )dA
-'
&T = ( &T),R = (5.78)
[ K 4 ( e "- l ) - ' dh '
In general, these effective or weighted average emissivities are functions of

the spectral band, the temperature, and of course & ( A ) . In this case, we can
RADIOMETRIC TEMPERATURE MEASUREMENTS 243
include the transmission E T ( A ) . The rest of the procedure involves appropri-

ate regions of the spectrum, in which the transmission is the same in both
regions. the detailed calculations cannot be included here, but it can be
shown as a representative example that temperatures of about 500°K can
be measured to within 5% to 10% through the atmosphere by using the
spectral region from 3 pm to 15 Figures 12 and 13 show some
examples.
5.5.4. Ratio Temperature Difference

In some applications of scanning radiometers, the temperature difference
from one point in space to another is the main concern. Medical thermo-
graphy is one important application. The clinician who, for example, is
screening women for the presence of breast cancer, is interested in regions
of increased temperature, an almost sure sign of pathology. He is not
necessarily interested in a variation in skin emissivity. In fact, medical
thermography to date has used radiation patterns as a diagnostic tool usually
with the tacit or explicit assumption that the emissivity is constant-or even
f
V
--
0
U
.-
> 0 3 -
2
0 2 -
01 -
Emissivity- transrni?lance ratio
FIG. 12. Error in temperature measurement by a two-color technique as a function of the

ernissivity-transmittance ratio for different ratios of background area.
0 3
a 05 I
Background temperature 0 s percentape of cargef temperalure Mortmum
error occurs when Dockground I S 7 5 % of torpet
FIG. 13. Error in temperature measurement by a two-color technique as a function of percent

of background area for different emissivity-transmittance ratios.
1. The technique of using the ratio of radiation in two spectral bands can
be used to measure temperature differences over the surface.33
To emphasize the main point and to avoid getting lost in the details, we
shall use the calculations based on very narrow bands and shall ignore the
contributions of the background. Then the voltage ratio from the two spectral
measurements at any scanned point is given by (where responsivities have
been cancelled)
(5.79)
where
c2
x, =-
A~ T’
x --. c2
2 - ~ Z ~
RADIOMETRIC TEMPERATURE M E A S U R E M E N T S 245
The difference in this ratio when the system looks at two different spatial
points a and b can be written
(5.80)
etc.
In this difference expression, A l and A2 are both known. If the emissivity
ratio E ~ / Eis ~the same for both spatial positions, then one has
(5.81)
Only T, and Tb are unknown.

For the temperature and wavelengths peculiar to medical thermographic
applications, ex >> 1, so that the 1s can be dropped with only a few percent
error. Then
The temperature difference can be found uniquely, provided that the ratio
of emissivities is spatially invariant. Dereniak has described this technique
in more detail.34
It seems important to emphasize the difference between two measurement
situations. The ratio of radiances can be used to measure temperature if
the body is gray-or at least if the emissivity is the same in two different
spectral bands. However, the ratio of radiances can be used to measure the
difference in temperature of two spatially separated points if the ratio of
emissivity in the two spectral bands is spatially invariant-even if the
emissivities are different in the two spectral bands.
How does one choose the spectral bands? They should be narrow enough
that the theory applies, broad enough to get satisfactory signal-to-noise
ratio, far enough apart to reduce the errors, and close enough together that
the emissivities are highly correlated.
5.6. Radiometric Instruments
Devices used to measure the amount of radiant flux are generally called
radiometers, but there are radiometers of various sorts and kinds, and even
with different names. This section describes some of these different types
and their salient characteristics. Devices called radiometers by most people
measure the amount of energetic flux in a relatively wide spectral band and
with a flat response. Photometers measure the flux in the visible spectral
region with a response that matches that of the eye. Pyrometers are devices
which measure the temperature of relatively hot bodies by virtue of some
radiation property of the body. Pyrheliometers measure the radiation from
the sun. Some measure incident flux, and others measure net flux. Radio-
meters which have a very narrow spectral bandpass are generally called
spectroradiometers, although some are also inconsistently called spec-
trophotometers. In addition to these classifications, which are generally
spectral in nature, radiometers can be classified as to whether they measure
the spatial and/or temporal distribution of radiation. Spatial radiometers
can be line scanners or can scan in two dimensions, creating the information
necessary for constructing an image. Whether an image is created or not,
these are classified as imaging radiometers. Another distinction which can
be made is whether the instrument measures the average or dc value of the
flux or whether it measures the high-frequency fluctuations. The latter is
often called a temporal or fluctuation radiometer. The most recent types of
radiometers are those which detect coherent radiation, or at least radiation
which is in some way coherent with a specified source, and those which
might be called power and energy meters, which are spectrally uniform and
relatively slow.
5.6.1. "Standard" Radiometers

These devices, which collect all the flux over a designated and relatively
wide spectral band, are designed with or without radiation choppers. The
radiation is incident upon the entrance aperture and brought to a focus at
the position of a chopper if there is one. Good design practice then calls
for a lens or mirror to collect all the radiation and put it onto the detector
by imaging the entrance pupil on the detector, a so-called field lens. This
makes the detector an exit pupil and the chopper the field stop. The chopper
can be either absorbing (black) or reflective. If the latter, then the surface
should shine a reference source onto the detector; if the chopper is absorb-
ing, then the radiation reference level of the radiometer is that of the
chopper. The detector senses the radiation level on its surface, first that of
the field and then that of the chopper or reference. Two main electronic
RADIOMETRIC INSTRUMENTS 247
processing schemes can be employed. One is to synchronize a rectifier with

the chopper and record only the signals that occur during positive periods.
Another is to record the difference between the reference signal and the
signal from the field. The second scheme appears to have the advantage of
an internal reference and no need for external calibration. In most cases,
however, the internal reference can only serve as a transfer standard. These
chopper radiometers are by far the most frequently occurring instruments
for general laboratory measurements. They usually incorporate either a
thermistor bolometer or thermopile as their sensing element. Their noise-
equivalent flux density can be estimated as described below.
The signal-to-noise ratio (SNR) can be written in terms of the specific
detectivity D* of the detector of area Ad and the power Pd on its surface:
D*Pd dA
SNR=
I,, m .
(5.83)
Both D* and Pd are in general spectral-varying quantities, but for a good

radiometer, the D* will be quite constant over the spectral band AA. The
power on the detector is related to the irradiance E at the entrance pupil
as described before:
Pd = TOEA,. (5.84)
where A,, is the area of the entrance pupil. Then the noise equivalent
irradiance or noise equivalent flux density (NEFD) is given by
E m
NEFD =---. (5.85)
sN R - D* T O ~ O
Values of D* for the thermal detectors used in most instruments range from
lo8 to109 cm Hzl” W-’., th ey are typically 0.1 cm on a side; and 7,) is about
50%. Therefore, the NEFD is about 5 x lo-“’ W cm-* for a 1-Hz bandwidth
and 1-cm2 entrance pupil. The time constants are about 1 ms. A typical
radiometer of this type is shown in Fig. 14. The operation is as follows.75
Light from a distant source is focused onto the detector through the chopper
mirror, filter, and field stop. The detector views either the field of view when
the open part of the butterfly-shaped chopping mirror is in position or the
light from the blackbody which is brought to the detector in the same cone
as the main radiation by the conformal mirror. The position of the chopper
is monitored by a small detector which views the light from the lamp. In
this instrument, the “back” side of the chopper is used. A sighting scope,
a light stop, and rudimentary baffle are also shown.
OPTICAL DIAGRAM FIELD
PRIMARY STOP
CAVl T Y
BLACKBODY
PREC I S ION APERTURE , n
CONFORMAL MIRROR
FILTER
FIELD STOP
DETECTOR
SECONDARY
MIRROR
PELTIER COOLER cFCCE+
OR H E A T E R
CHOPPER MOTOR
REFLECTIVE
CHOPPER BLADE
SYNCHRONOUS
PIC K O F F
LAMP RETICLE
VIEWING
OPT1 CS LLIGHT
STOP
FIG. 14. The schematic of a typical chopper radiometer

RADIOMETRIC I N S T R U M E N T S 249
5.6.2. Astronomical Radiometers

Radiometers of this general type have also been used in infrared
astronomy.36 Larger apertures, long observation times, and laboratory-like
installations characterize their configurations and environments. The iadi-
ation from the atmosphere has usually been the limiting noise and, as is so
often the case, the problem was seeing the small source in a large back-
ground. The technique employed by most was the wobbling secondary. The
secondary of a Cassegrain system, for example, is tilted back and forth so
that the instrumental field of view included alternately the star or planet
under observation plus some sky, and then only the sky. If the sky is
sufficiently uniform, then this background-cancellation technique works
well. Most of the detectors for these applications are described by their
noise equivalent power (NEP) values, because they do not follow a square-
root-of-area law. The ratio of the radiometer NEFD to that of the detector
FIG. 15. Optical train of the balloon-experiment: 1 m primary ( I ) ; wobbling secondary (2);
field stop (3); off-axis mirrors (4,6); lamellar grating (5); bolometer (7); instrument box (8).
is Ad/rOAc. Since NEFD values are typically lo-' and detector areas are
about 0.01 cm, it is easy to see how NEFD values of W cm-* are
obtainable with reasonable instruments. Better (lower) detector values,
larger entrance pupils, and smaller detectors can move these values down
another decade or two with great care. Then the limitations are radiation
fluctuations, emissions from the optics, and inhomogeneities in the atmos-
phere. Figure 15 shows one of these astronomical radiometers.
5.6.3. Space-defense Radiometers

Infrared radiometers have been used as part of missile detection systems.
They have been at least proposed for detecting the radiation from rockets
in their powered or boost phase, in their midcourse or coast phase, and
when they reenter the atmosphere. The radiometric aspects of each of these
applications are considered below.
When the rocket is in its boost phase, it emits very large amounts of
radiation coming from a point (or subresolution) source. The problem is
to design an instrument that can cover a large area all the time and still be
quite sensitive. Because it keeps a selected area of the earth continuously
in its field, a synchronous satellite is a very desirable platform. A field of
view of about 5000 km2 is also desirable and probably not excessive or
unreasonable. The signal-to-noise ratio can be calculated using the same
equation for SNR. The problem is one of detection rather than measurement,
and so the most sensitive detector will be used. Further, one can expect
most of the radiation to be in the emission bands of water at 2.7 pm and
of carbon dioxide at 4.3 pm (because these are the main combustion prod-
ucts of hydrocarbon fuels). Detectors for these regions can have D* values
of l o i 2 to l o L 3andI2 are constant over the relatively narrow spectral bands.
Thus,
SNR = l O I 3 (5.86)
Since this is a narrow band, the power can be written
Pd dA = Pd AA = roPAAA. (5.87)
Then the NEP of the system is given by

NEP = low3m / r o [ WI. (5.88)
The spectral NEP is
NEP= 1 0 - ' 3 m / A A 7 0 [ W pm-'I. (5.89)
RADIOMETRIC I N S T R U M E N T S 25 1
The power on the entrance pupil can be found in terms of the target radiance
L,, even though this is a subresolution target. In the spectral band AA, it is
A
P = L,A, 7 = L,A,R. (5.90)
d
Here L,A, represents the watts per steradian radiated on the average from
the entire target, and L? is the solid angle that the entrance pupil subtends
at the target. The range is approximately 33 x lo6 m, so that for the square
meter entrance pupil, the power is
P = L,A, x 10-15. (5.91)
Thus the signal-to-noise ratio is
(5.92)
If the optical transmission is 50%, the detector area is (50 pm)', and the
bandwidth is lo6 Hz,then a target intensity of lo3 W sr-' gives a signal-to-
noise ratio of 1. The system must be designed so that photon noise is low
enough for this D* of l O I 3 to be attainable. The problem now becomes one
of scanning and numbers of detectors. The smaller the instantaneous field
of view, the better the chance of seeing the difference between the target
and background. The faster the scan rate, the more looks obtained but the
larger the bandwidth. The smaller the detector, the better the signal-to-noise
ratio, and (other things being equal) the smaller the instantaneous field and
the bigger the bandwidth required. Suffice it to say that these radiometer
designs use many detectors and various ingenious optical and scanning
systems. They can also take on a variety of forms, because there are so
many parameters that can be varied.
The midcourse detection problem is a different one. Here we can assume
that there is no thruster power; the only signal is that from the ballistic
projectile. Its radiant intensity is then calculable if certain assumptions are
made. The radiant intensity in the forward hemisphere is &ALBB(A, T). Its
emissivity is between 0 and 1. Thus an estimate of 0.5 is reasonable. A better
estimate reiuires knowledge of the material, its state of aging, exposure,
etc. For estimation purposes 0.5 is pretty good. The projected areas of most
missiles and satellites range from 10m2 to 1OOOm' but are mostly about
100 m2. They sit on the launch pad, where they come into equilibrium with
their surroundings at about 300" K. So a midcourse object has the intensity
of a 300" K blackbody with an emissivity-area product of (cross section)
50 m2. It peaks at 9.6 pm, where its spectral radiant intensity is approximately
350 W sr-' pm-'. The SNR is again given by the same equation. The power
on the detector is obtained from
(5.93)
where
T, = optics transmission (0.5)
EALT-’ =target radiant intensity (350 W sr-’ km-’)
A,/ R 2 = solid angle of sensor at the target
The noise equivalent range can be calculated by assuming Ad = 50 km,
B = lo6 Hz,and by setting S N R equal to 1. Then the noise equivalent range
is
NER = 6 x 105AA12.
The units are those of A;/’. If A;’’ is 10cm for instance, then the noise
equivalent range for these conditions is 60 km. We can see that this is a
more difficult problem that should be solved by the use of a larger aperture
and more detectors.
The radiometric problems of reentry are trivial. The distances are hun-
dreds of kilometers, and the bodies have become very hot from aerodynamic
friction. Clearly the calculations above show that detection is easy. The
problems relate to search rates and discrimination. The multicolor tech-
niques described above may be appropriate.
5.6.4. Scanning and Imaging Radiometers

Instruments which belong to this category utilize a mirror or mechanical
mechanism to cause the image of the detector to scan a larger field of view.
The detector angular subtense is called the instantaneous field of view; the
larger angular subtense is the total field of view. These devices come in
single- and multiple-detector versions and have been used for medical
investigations, nondestructive testing, industrial inspection, and for military
reconnaissance and night-vision applications.
Imaging radiometers provide a spatial map of the flux distribution over
an area. The usual procedure is to couple the detector to the electronics
through a capacitor. In this way, only changes are recorded; these changes,
as with the standard radiometer using synchronous rectification, are then
coupled to an artificial average level or pedestal. Much of the technology
of modern television is applicable to the design of these systems. The
equations for signal-to-noise ratio and related measures are derived here.
As usual, the starting place is the SNR equation. Radiation from extended
targets is measured; the detector power is therefore given by (for a distant

scene)
(5.94)
where 7 0 , is the optics transmission, L is the radiance, Ad is the area of the

detector, and A is the area of the optics.
If the optical aperture is circular, then
where
SNR =
I T ~ ~ ~ L
dhrD~AB/4f2B”’,
D * (5.95)
Do= optics diameter

f/ Do= focal ratio
Estimates can be made for the performance of such a system, based on
values chosen for the optics and detector. The instruments are AC-coupled;
they measure difference in signal level from point40 point in the field. Thus
the important parameter for assessing their performance is the change in
radiance that is equal to detector noise. This is the change in signal-to-noise
ratio, and it can be written as
D A9
r,rOD* A L d h r - (5.96)
4f B1/’*
If r,, T ~ and
, D* are constant over the spectral band, then
D A0
ASNR= ToTaD*AL -
r (5.97)
4fB1l2‘
This is a relatively easy (simplified) equation with which the performance
can be evaluated. Typically, To = 0.3, = 0.8, D* = 10”. Then
DAB
h s N R - 2 ~10’OAL- (5.98)
fB’/*‘
The noise-equivalent radiance difference is ALISNR:
fB1”
NEAL= 5~ 10-94 - (5.99)
d AB’
Although there are many nuances in the design of these imaging radiometers,
the main factors are displayed in this equation: make the Flnumber as
small as possible and use a minimum bandwidth. The limits of performance
are based on what the optics and the detector can do. The Rayleigh criterion
indicates that DAB is approximately equal to the wavelength of the center

of the spectral band. The D* of the detector, when the detector is photon-
limited, is given by’*
(5.100)
where
g = 2 or 4 for a photovoltaic or photoco,nductive detector
7 = detector quantum efficiency
When the system is limited by photon noise, the SNR is independent of

the focal ratio of the optics. This condition is often met by systems which
view the ambient background of the earth. The value is about lo”, as given
above. Therefore, an approximate value of the performance of such systems
is
NEAL = 5 x 10-6B1’2. (5.101)
The bandwidth can be calculated on the basis that there are about 500 lines
in a TV frame and the frame rate is 30 sec-’. Thus there are 7 x 10 picture
elements scanned every second. If only a single detector is used, it must
have a time constant less than 1 ps, and the NEAL will be about
lo-* W cm-’ sr-’. Such a radiance change corresponds to a change of about
10°K or 10% in emissivity, neither of which is very interesting for most
applications. Most such instruments therefore use 100 or more detectors to
reduce the bandwidth (or signal average in some other way). Since the
bandwidth can be reduced in proportion to the number of detectors used,
the NEAL is proportional to the square root of the number,
An important question to be asked about such scanning radiometers is
how, and how accurately, can they be calibrated. The calibration schemes
vary, based on the intended use. Almost all are ac-coupled instruments; in
temporal frequency space, there is a low-frequency cuton determined gen-
erally by an RC pi network in series with an amplifier. This means that the
average or dc value is lost. It must be returned as a pedestal level on which
small variations ride, and this combined signal is recorded on a cathode
ray tube (CRT), fi1m;magnetic tape, or other appropriate medium. If there
is little concern about the true radiometric level, then averaging the signal
level over some appropriate length of time (e.g., one line or one frame)
provides an excellent reference. If the scene does not change much, then
the small signals are displayed well and d o not get lost in one end of the
dynamic range. If, on the other hand, one is really interested in measuring
radiometrically, then on-board calibration is required. Probably the simplest
R A D l O M ETRIC I N S T R U M E N T S 255
way to d o this is with a small source that periodically floods the detector
(array). If this source is not collimated, it will not illuminate the detector
in the same way as the distant objects, but it can be calibrated. If the source
does not shine through all the optical elements, then its calibration gradually
becomes invalid as the elements deteriorate. Obviously this source should
be stable or measurable. The most accurate way seems to be to use a small
heated cavity at the focus of a well-protected collimating system. The
resultant beam shines at’the detector system through all the optical elements
from just outside the active field of view. Ideally, two or more such sources
are used, each at a different temperature and with their temperatures
monitored. Diode emitters are a reasonable second choice. They are not as
good, because they do not match the spectral input and because the stability
of infrared diodes (8-14 pm) has not yet been established. In the visible,
a stable silicon diode could be used, but any spectrally selective deterioration
of optical elements would alter the calibration of the instrument.
There are several ways to calibrate these instruments in the laboratory
and in the field. The usual specifications of performance are the NETD
and the MRT-the noise-equivalent temperature difference and the
minimum resolvable temperature. The first of these is defined as the
difference in temperature between two blackbodies which both fill the field
of view of the radiometric scanner that gives a peak-to-peak signal difference
equal to the rms noise. The second is the minimum temperature difference
that can be resolved by an observer at a specified spatial frequency. To
make the first measurement, one needs two large-area blackbodies (almost
a contradiction in terms) that can be carefully controlled in temperature.
The usual sources are large blackened tanks of water with stirrers and
thermometers of some type. They are set so that they are different in
temperature by some 10°K or 20°K and so that a large SNR is obtained.
The temperature difference is reduced until the signal and noise are about
equal. A curve of ATISNR is plotted. Each point is a measure of the
NETD-almost. Two types of errors occur: one because the sources are
blackbodies, and one because they are not. Let us see how this comes about.
The difference in signal-to-noise ratio is given by
ASNRKA 7(h)D*(A)L(A, T ) dA. (5.102)
The terms that would make the proportionality an equality are all geometric
and temporal bandwidth factors. The ~ ( h factor
) is the product of the
spectral transmission of the optics and the atmosphere. The scanning radio-
meter views first one body and then the other. It is assumed that they are
close together so that there is no difference in atmospheric transmission,
optical transmission, or D* between the two measurements. Thus,
T ( A ) D * ( A AL(A,
) T ) dA. (5.103)
Only the radiance changes. Let us consider only that term. The radiance
projected toward the sensor is the sum of the radiation emitted from the
source and the background radiation reflected from the source to the sensor.
AL(A, T ) = A [ & ( A ) L R B ( AT ,) + p b E ( A , T ) ] , (5.104)
ALL - AE E+ -L1-aadTLT + A ( y )
- _E+ - -xex
A AT E
+-Apb+-.
PbAE
- (5.105)
E ex-1 T L L
The first term is typically about 1%, a change in emissivity from about 0.98
to 0.97 of these near-blackbody sources. The second term is about 5 AT/ T
or 0.017 AT The next two terms represent background contributions. In
fact, each would really be the sum of all the variegated components of the
background. The first one is the contribution of the variation in reflectivity
with a constant background. The second one represents the contributions
of different background terms if the reflection geometry with the two samples
is slightly different. These effects have been considered in more detail.I2
The analytical results there show that errors of several percent are easily
attained. One message is clear: the radiometric differences of such test
objects should be measured radiometrically.
When really large patterns for field work are to be used, the use of water
baths and other actively heated structures gets cumbersome. Bartell” has
described a test pattern design that can be used for NETD and MRTD
measurements. As Lloyd38shows, an MRTD test involves several important
psychophysical factors. These include the temporal and spatial integration
properties of the eye and its tendency to be a matched filter. Accordingly,
the test pattern must have groups of four bars with a 7 to 1 aspect ratio. If
one can accept that the effects of the psychophysical factors can be calcu-
lated, then simpler means, including edge gradients and line spread func-
tions, are a~ailable.~’
5.6.5. Spectroradiometers
These devices differ from standard radiometers principally in that they
use some technique for selecting narrow portion of the spectrum, and only
radiation in that band is measured. The radiometric calculations of the
responsivity and SNR are in principle identical to those of the standard

radiometers. The user must pay particular attention to the definition of the
selected spectral band (the passband) and to the rejection of the rest.
The passband is usually characterized by the spectral width between the
wavelengths at which the transmission is one half maximum. Older literature
generally calls this the half-width, while most modern writers use the phrase
“full width at half maximum.” The symbol is Ah. Sometimes this spectral
band is normalized to the central wavelength Ao, and the reciprocal of the
ratio is the resolving power RP.
RP=l--l AA -I
-I
.
There are several direct methods for filtering, and there are some that
might be called exotic or indirect. Certainly the use of a prism, grating, and
filter must be considered direct and straightforward-even if they are not
always simple.
5.6.6. Filter Radiometers

The simplest filter in a conceptual sense is surely the absorption filter.
One can define a passband by the transmission cuton of a material like
germanium and the cutoff by the detector spectral response roll-off. Many
combinations exist, especially with the advent of mixed crystal materials.
For example, one could pick a detector made of either HgCdTe, PbSnTe,
or InAsSb and obtain a sharp long-wavelength cutoff which occurs anywhere
from 1 pm to 15 km. One could combine this with absorption filters made
of germanium, silicon, zinc selenide, sodium chloride, fused silica,
potassium bromide, and others. The combinations, although not endless,
are certainly too large to enumerate here. The user should be aware of the
possibilities. One should be aware, too, that the spectral response must be
calibrated at the temperature of use. All of these materials have transmission
characteristics that are strong functions of temperature.
In many instruments, interference filters are utilized either to define the
passband or to provide an antireflective coating on other optical elements.
These filters or coatings should usually be put in a collimated beam, because
the essence of their operation is the interference of forward and backward
light waves. This, in turn, depends upon the optical phase difference
k n d cos8. For a given wavelength (and refractive index), the phase
difference and therefore conditions for maximum and minimum of the filter
transmission depend upon the cosine of the angle between the normal and
the direction of the ray. If the beam is not collimated, the filter transmission
curve is a weighted average over all these angles. For an F / 1 cone of light,
FIG. 16. Rays on a thin film illustrating the reflection from the first surface plus two from
the second.
for instance, there is a phase shift variation with angle of 50% between
central and marginal rays. This is equivalent to moving the central
wavelength from 2 pm to 1 pm or from 10 pm to 5 pm. When the filter is
used in a collimated beam, it can be tipped to change the passband position;
such a procedure entails somewhat complicated polarization effects which
must be calibrated. An interference filter in general has what might be called
side lobes or higher-order passbands. For the simplest case of obtaining
good transmission at Ao, the total reflected radiation should destructively
interfere. The first ray has n o phase shift. The once-reflected ray undergoes
a phase shift given by 2knd cos 0 + rr. The additive rr comes from the phase
shift which arises from reflection when light goes from a dense to a rare
material as it exits from the film. The third ray which undergoes two
reflections is just twice this, etc., which is shown schematically in Fig. 16.
For destructive interference, each of these rays should have a phase shift
with the wave reflected from the front surface that is exactly rr, or an odd
multiple of T. The design equation then is
mA
ndcos @=--, even m. (5.106)
4
Although more complicated designs exist, they generally have the charac-
teristics that a variety of combinations of m and A satisfy the conditions
for maximum t r a n s m i s ~ i o n For
. ~ ~ example, a filter designed for one-quarter
wavelength thickness at 10 pm will have a peak at 10 pm, 5 pm, 2.5 pm,
etc. The radiation in these higher-frequency passband regions is usually
absorbed by the substrate. Caution should be exercised by the investigator
when using such a filter in that this absorption can be highly temperature-
dependent-the substrate is an absorption filter and has the properties
described above.
Radiometers often use these filters in a fixed position or in a filter wheel,
a set of different filters which can be introduced into an appropriate part
RADlOM ETRIC INSTRUMENTS 259
of the optical train one by one. The device is usually circular (a wheel) but
can take on other forms, linear or cylindrical for example.
One very useful device for defining a passband may be thought of as a
continuously varying interference filter; its most popular form is the cir-
cularly variable interference filter or CVF.4' Its center wavelength (or
frequency) is determined by position on the filter, which consists of a circular
substrate on which are evaporated interference layers whose thicknesses
vary with circumferential position around the circle. The calibrations, cau-
tions, and limitations which apply to interference filters apply also to CVFs.
In addition, positional calibration must be considered. The bandpass is an
average over the area of the filter that is viewed, and the center wavelength
is a function of the position of that area-which is usually a slit. Half of a
CVF typically covers one octave of the spectrum, from A,, to 2Ao. Thus, one
generally has a bandpass of about A,/ 180".The RP is 180 per degree of the
CVF that is viewed.
The concepts of throughput and the optical invariant are required for
further useful discussions and comparisons of these instruments. The optical
invariant when applied to pupil and image planes is just that ny sin 6 is a
constant, where n is the refractive index of the medium, y is the height of
the image or pupil, and 6 is the angle the next image or pupil subtends at
the one being considered, as shown in Fig. 17. The throughput is the square
of this optical invariant.
Suppose that an optical system with area A . and focal length f focuses
radiation onto a CVF over a width x and length y (as defined by a mask
or slit in front of the CVF). Then the throughput is A o x y / f 2 .If the slit
extends from r / 2 to r (where r is the radius of the CVF) and is x wide at
the circumference of the CVF, then the slit area is ry/4. The resolving power
is 47r/x (using the average width of the slit). Therefore, one increases the
resolving power at the expense of throughput, which is directly related to
flux. There is a linear and inverse relationship between these two. The loss
in throughput resulting from reduction of slit width can be compensated
for by making the slit longer, thereby requiring the radius of the CVF to
be larger.
FIG.17. The Laprange invariant for pupils. The product of n , y , sin 0, is equal to n z y , sin 0 2 .
In this example, n , = n z .
This problem can be avoided to a first approximation by using the

individual absorption or interference filters just discussed. It has already
been pointed out that an interference filter should be used in a collimated
beam to obtain a narrower spectral band. If the filter is made the size of
the entrance pupil and placed there, all is well, but economics drives one
toward using smaller filters in smaller collimated beams. Then the optical
invariant says that rays will diverge more or less as the size of the beam is
decreased or increased. Over relatively small angles, and the operation
region of most spectroradiometers, there is a linear proportionality between
angle and diameter and a cosine relationship between resolving power
and the linear dimensions of the aperture. If, however, an absorption
filter is used, the relationship does not apply. Systems which use
absorption filters are generally so crude spectrally that resolving power is
hardly meaningful.
5.6.7. Prism Disperser Radiometer

Prism systems can have resolving powers considerably better than the
filter systems discussed above. A collimated beam is directed toward a prism
which has length 1, width w, and base b, as shown in Fig. 18. The prism
refracts the light at different angles according to the value of the refractive
index at each wavelength. The beam is focused to a plane where a single
slit or many can be placed to select radiation of different frequencies.
Its resolving power is
A dn
RP=-=b-. (5.107)
AA dA
The linear dispersion dx/dA of a prism is given by42
-g=f (&.b dn
(5.108)
The throughput Th can be calculated as the area of the beam on the prism
times the area of the slit divided by the square of the focal length:
h dx
T h = ( h w c o s 0)-. (5.109)
f’
If the above expression is used for the linear dispersion dx/dA, then
dA
Th = RP h2 cos 8 -= the throughput. (5.110)
f
RADlOM ETRlC INSTRUMENTS 26 1
,
FIG. 18. Prism geometry.
5.6.8. Grating Disperser Radiometer

Grating systems make use of the interference of many diffracted waves
arising from the n slits of the grating. The grating equation can be written
mh = d(sin 4 +sin 0), (5.111)
where
m = order of the maxima
d = line spacing
4 = angle of incidence
0 = angle of diffraction
The resolving power is obtained by differentiating with respect to A :
m = d cos 0 (s)
d0 - dD - m
dh dh d cos 0
d cos d D
dh =
m
A
RP=---=d(sin 4 + s i n 8)
(5.1 12)
dh d cos 0 d D
5.6.9. lnterferometric Radiometer43

A third major type of spectroradiometer is the interferometer. This is an
indirect device which first forms the Fourier transform of the radiation; the
262 PHOTOMETRY AND RADIOMETRY
spectrum is obtained by the use of a computer to transform the interferogram.

It is appropriate to review the basic operation of an interferometer to obtain
some of its radiometric characteristics. A simple Michelson interferometer
is taken as the prototype. When quasi-monochromatic light illuminates the
instrument, the center of the focal plane will have a maximum when there
is no phase difference between the two arms. If one arm is moved at a
constant rate, the phase difference changes at a constant rate, thereby causing
a sinusoidal modulation on a detector placed at the center of the focal
plane. If a second monochromatic beam of different frequency and different
wavelength is also part of the input radiation, its frequency of modulation
at the output will be different, and by superposition the output will be the
sum of these two signals. The output flux density for light of a single
wavelength (traversing paths a and b ) is given by
E = uu* = u, &‘ + u,,&(““+~)12, (5.113)
E=E,+E,+2ERbcos
:[ -ndcos8
1.
The optical path difference nd cos 8 is just the difference in the length of
(5.1 14)
the two arms, assuming that a compensator plate has been inserted in the
horizontal arm. The length of the horizontal arm then is
l=tu(t)+Io.
The output flux density for a monochromatic input is given by
(5.115)
For a polychromatic input, the output intensity as a function of time (where

it is assumed the beamsplitter is 50% so that E, = Eb) is
E ( t ) =2
I
EA( t ) 4-EA COS[ ( t u ( t ) -4- 1.3)- (5.1 16)
If EA is a constant, then E ( t ) and EA are cosine transforms. This means

that the spectrum of the radiation can be found this way as long as the
intensity of any spectral component remains constant. The effects of changes
in intensity with time are interesting, but their discussion will be postponed
until the resolving power and resolution characteristics have been described.
The resolution of an interferometer of this type is usually specified in wave
numbers and is equal to the reciprocal of the total path difference. The
shape of the bandpass is a sinc function, and the bandwidth is specified as
the interval between the first zeroes. Many systems use electronic processing
to alter this sinc-shaped spectral band in order to reduce the contributions
MEASUREMENTS 263
from the side lobes. The resolving power can be written as nd cos B / A ; it
is the optical path difference measured in wavelengths.
The throughput of an interferometer like the Michelson is the area of the
mirrors times the area of the detector divided by the square of the focal
length of the final beam. If a T w y m a n - G ~ e e nversion
~~ is used, it is easy
to see that this is (2.44A/D)*(.rrD2/4)=A2.
The use of an interferometer like the Michelson for radiometric purposes
requires some special considerations. The most important of these are that
there normally is no dc reference available, and that the source intensity
must remain constant during the time of the scan. The lack of a d c reference,
of course, means that the spectrum can be obtained, but only on a relative
basis. A fairly obvious solution to this is to measure the total radiation with
an auxiliary radiometer and use it to set the radiation scale. Variations of
the source output intensity generate much more subtle results. It can be
seen that the spectrum is obtained by a cosine transform:
co
/-. Z cos(kvr) dr.
If the intensity varies, then the interferogram is affected according to the

nature of that variation, and the effect ofthis variation is different in different
parts of the radiation spectrum.
5.7. Measurements
Radiometric and photometric measurements are not difficult in principle.
They are very hard to accomplish with a precision or accuracy better than
about 1YO. Most field measurements which are made with portable equip-
ment and distant (in both space and time) from a primary calibration have
errors of more like 20%. The measurement of the spectral radiant intensity
of rocket plumes has been considered satisfactory when the error was less
than an order of magnitude! These errors can usually be attributed to the
fact that the measurement of the unknown was different from the calibration.
Accordingly, rule 1 of radiometric measurement is to calibrate the radio-
meter in a configuration that is as similar to the measurement situation as
possible.
The responsivity of a radiometer depends upon very many factors. It can
be a function of wavelength position, angle, polarization, flux level, time,
humidity, temperature, and many other things. Phase of the moon and the
name of the operator have also been proposed as factors influencing the
measurement. This great variability leads to rule 2, which is to think of
everything.
These two rules are very general. The rest of this section is intended to
make them a little more concrete by discussions of examples. The discussion
starts with standards, the primordial element of every radiometric
measurement.
5.7.1. Standards
Any radiometric measurement must be considered in the light of the
calibration against known standards. It can be no better than the standard.
The National Bureau of Standards (NBS) maintains and provides radio-
metric standards in the form of 1000 W DXW-type tungsten-halogen lamps.
They are described in detail elsewhere.I2 The lamps are given a primary
calibration and their irradiance is fit to the following curve:
E ( A , t ) = A - 5 e(o,-4.7+~f+bflA) (5.117)
The values of the constants ah, 6, and c are found from least squares fitting
for different values of the wavelength A and time t. Four lamps are used
and the proper 3a error found from among these, as shown in Table IV.45
Table IV summarizes the uncertainties assigned by NBS to these standards
at different wavelengths.* Item l ( a ) is the error in primary calibrations;
l ( b ) represents long-term reproducibility rather than error with respect to
SI units. Item 2(a) represents the transfer to an irradiance scale, while 2(b)
is the error in reproducibility of the four standard lamps. Item 2(c) represents
the errors involved in the fitting procedure described above. Item 3 represents
the error in transferring the value of the four standard lamps to the working
standard being calibrated.
The total uncertainty with respect to SI units then is found from the rss
of l(a), 2(b), 2(c), and 3. Reproducibility is found from l(b), 2(a), 2(b),
2(c), and 3. Note that these are all three standard deviation errors.
It should be observed that the errors increase rapidly in going toward
250nm largely because the flux falls off so rapidly. It is enlightening to
calculate the ratio of the total 2800°K blackbody to the amount of radiation
at each wavelength (the approximate temperature of these sources). These
sources are limited in their spectral extent at the short wavelength end by
lack of flux and at the longer wavelength end by the absorption of their
glass or quartz. These are the certified standards available from NBS.
Instructions for their use come with them and have been published else-
where.'*
A very encouraging development in radiometric standards is that of the
electrically calibrated radiometer ( ECR).46 The receiver is blackened and
* I am indebted to Karl Kessler for providing a detailed explanation of this table.

TABLEIV. NBS Radiometric Standard Uncertainties ( % )
1. NBS Spectral Radiance Scale 250 350 450 555 654.6 800 1300 1600
nm nm nm nm nm nm nm nm
(a) Absolute error (with respect to SI units) 1.66 1.20 0.93 0.65 0.65 0.46 0.28 0.27
(b) NBS long term reproducibility 1.06 0.76 0.61 0.30 0.26 0.22 0.12 0.18
3
rn
2. NBS Spectral lrradiance Scale
(a) Systematic errors 0.16 0.16 0.10 0.10 0.08 0.08 0.07 0.07
R
c
(b) Random errors (estimated reproducibility of the mean 0.72 0.55 0.53 0.56 0.54 0.51 0.80 0.80
of standards I, - I,, 3 - u) I
(c) Model error 1.35 0.80 0.73 0.73 0.78 0.76 0.77 0.82
4
v1
3. Transfer Calibration of a Test Lamp
(a) Random errors (3u precision) 1.35 0.65 0.50 0.48 0.41 0.51 0.39 0.36
4. Uncertainty of Reported Values

(a) With respect to SI units 2.6 1.7 1.4 1.2 1.2 1.2 1.2 1.2
(b) With respect to NBS (long term reproducibility) 2.3 1.4 1.2 1.1 1.1 1.1 1.2 1.2
shaped to be as black as possible; its surface also contains windings of

heating coils. This cavity interacts optically with a detector (a pyroelectric
one in the current realization). The detector senses first the cavity which is
heated by the radiation from the field of view and then from the heating
coils. The development of such an ECR as a standard with accuracy of 1Yo
requires excruciating care and attention. It appears that this receiver will
become the standard over a wide spectral range because it is more accurate
than the current source standards and because it can cover a wider spectral
band.
A more recently developed, more accurate standard has also been
developed by G e i ~ t , a~ 'device he calls a self-calibrating detector. Presently
it is a silicon diode and is therefore useful only from about 0.3 pm to about
1 pm. He estimates inaccuracies of approximately 0.1%, more than an order
of magnitude better than previous calibration capabilities. A silicon diode
is used. First the reflectivity is measured by one of the methods described
later. These are all basically ratio methods, so that inaccuracies of less than
0.1% can be attained. Then light illuminates the diode, which is used in a
back-bias mode. The back-bias is gradually increased as the output is
monitored. When there is no longer any increase in output with increase
in back-bias, the internal quantum efficiency is 1. The external or total
quantum efficiency is the product of 1 minus the reflectivity and the internal
quantum efficiency. Careful attention to the design of the diode, especially
its thickness and the geometry of the junction region, assures that there is
very little error in the assumptions about the internal quantum efficiency
reaching unity. This is a very exciting development in high-precision
radiometry.
5.7.2. The Accuracy and Precision of Field Measurements

Experienced radiometrists will often state that a radiometric measurement
made in the field (top of a mountain, floor of a desert, instrument bay of
an airplane, or a satellite) is good if the error is 50% or less. This section
is an attempt to explain why the errors are so large when the (present) NBS
standards are about 2%.
Radiometric errors in the ultraviolet have been discussed in two articles
in Optical Radiation N e ~ s . ~The' , ~discussion
~ is given in the context of a
spectroradiometric measurement in the 0.25-pm to 0.35-pm region. The
errors in general are a function of the complexity of the radiation as well
as the type, duration, and intervals of measurement. The uncertainties one
should add to those of the standards described in Section 9.7.1 are discussed
below.
MEASUREMENTS 267
If the radiant power is uniformly distributed with respect to angle, area,

and polarization, if it does not vary much with wavelength, if it is in the
middle of the dynamic range, and if the irradiance is about equal to the
calibration irradiance, then the additional uncertainty is about 0.1 YO to
0.5%. If these conditions do not obtain, then the additional errors are as
follows: position, direction, and responsivity variations add 1‘/o ;wavelength
variation and wavelength calibration errors add 3% ; dynamic range, non-
linearities, 0.5% ; polarization responsivity, 0.5%. If the irradiance level is
too low or the time of response of the instrument is inadequate to obtain
a signal-to-noise ratio above 10, one should probably not make the measure-
ment. Otherwise, important errors relating to random noise in the instrument
will exist. The errors listed above, which apply to relatively good measure-
ment conditions add in quadrature to values of 3.3%. Note that the angular
and spectral contributions are dominant.
Among the contributions listed for unfavorable conditions of measure-
ment are the following: uncorrected instrument temperature variations
(0-lo%), old standards (2%), time interval since last calibration YO),
variation in system alignment (0-5%).
The authors of the first article48 conclude, “thus under very favorable
measurement conditions, the state of the art of spectral irradiance measure-
ments in the 250-nm to 350-nm region varies from 3% to 9% . . . .” They
emphasize that his can be obtained only with the best instruments and
techniques. I believe one can make measurements with an uncertainty of
2% to 3% for well-behaved radiation and 3% to 5% for more complex
radiation. This, of course, is not a major disagreement; my reasoning is as
follows: the uncertainty in the irradiance standard is given as 2.4% at
0.25 pm. If the radiation is well behaved, an additional 0.1% to 0.5% is
likely to be encountered with great care and excellent instrumentation. This
gives a total (rms) error of 2% and 3%.
In the second article, the authors conclude that for performing terrestrial
solar irradiance measurements in the UV-B region, an ultra low scatter,
double monochromator with 1-nm spectral bandpass should be used, that
the response should be deconvoluted, and that “it should be possible to
limit the measurement uncertainty to about 15% at 295nm and to about
5% at 320 nm and longer wavelengths.” The errors they discuss are instruc-
tive. The first relates to the slit function or spectral response of the
spectroradiometer. The output voltage (or current) V is given by
m
V= E,,(h)%!(h)dh. (5.118)
-m
If the overall responsivity takes into account the slit function as well as the
instrument responsivity by expressing % ( A ) as % r ( A ) a ( A -Ao), where it is

assumed that r is independent of Ao, then
V=% I--oo
a-
E,,(A)r(A)cT(A-Ao) dA. (5.119)
If the slit can be described accurately enough as a delta function, the integral
unfolds immediately:
v= BEA(AO)r(AO) = %(AO)EA(AO). (5.120)
The spectral response at A. can be found by the use of any sufficiently
well-known standard source. In practice, of course, there are no delta
functions. Slit functions peak and have long tails that are relatively wide.
The long tails which reach longer wavelengths at values of to lo-' are
the source of error in measurement-at least until they are corrected. The
interesting features are the means by which such corrections can be made-
and the limitations in making them. The slit function can be measured by
the use of a device like the ECR and a source (like a laser) with a line
width smaller than the spectral slit width. The spectral slit function can be
taken as known at least to the resolution of the best laser in the spectral
region of interest. But even though the function is known, the integral is
not unfolded simply. The output voltage is given by
V=% I-m
a-
E , ( A ) r ( h ) a ( A-Ao) dh.
We can assume r ( A ) is constant over the band in order to arrive at the

essence of the problem. Then
V=% I m
-m
E , ( A ) a ( A - A o ) dh. (5.121)
It would appear that one cannot obtain E A ( A ) from this, because it is an

unknown and arbitrary function. Many different distributions of EA(A ) for
a given a ( A - Ao) can give the same value of V. But a spectrometer makes
many measurements at different values of Ao, obtaining different values of
V. Thus, for known values of % ( A ) and o ( A - A i ) , the appropriate
expression is
vi= I_,
00
% ( A )EA(A - hi) dA. (5.122)
The more individual measurements that can be made, the more determined
EA(A)will become. The problem is one of estimating E A ( A )from a set of
samples with known sampling shape. One can argue that a(A- A i ) can
change as hi changes. This is inconsequential as long as o ( A - A i ) is known.
MEASUREMENTS 269
Another way to view this procedure is to recognize that the right side of
the equation is a convolution. Therefore, the Fourier transform of the voltage
is equal to the product of the transforms of the slit function and irradiance.
The procedure for obtaining the irradiance is then straightforward but
involves much calculation. It might be added in passing that the use of this
approach to measuring spectra leads to the possibility that the smallest
possible slit width is not the best. An interesting tradeoff between slit width,
signal-to-noise, and processing procedures arises. Too wide a bandpass
leads to processing errors which are too large; too small a slit width leads
to a signal-to-noise ratio which is too low.
The authors of the NBS article then provide a figure which shows the
true spectrum and the inferred one. The major portion of the error is at the
short wavelengths where there is very little radiation and where the longer-
wavelength scattered radiation predominates.
5.7.3. The Measurement of Emissivity

We have already the properties of an ideal radiator. Now we
can imagine a body in thermodynamic equilibrium at temperature T. The
radiation it emits into a hemisphere is specified as M ( A ) and that of a
blackbody MBB(A). Then the hemispherical, spectral emissivity of a material
& h ( A ) is given by
(5.123)
The spectral emissivity of the material may also be described in terms of

photon radiation
(5.124)
The total hemispherical emissivities are given by
Iorn MA
M
Iom
&h =
--
- (5.126)
aT4'
M :B dA
low--.
M'-lA "
-
M
(5.127)
Eqh= [" Mz: dA u q 7-4
Jo .
Any of these ratios is a function not only of the properties of the

material but also of the properties of its surface. In general, a rougher
surface causes the ratios to be higher and the emission to be more
uniform.
In addition to spectral and total values of emissivity, there exist the values
of emissivity averaged over some spectral band, i.e.,
& ( h ) M ” ” ( hT, ) dA/

I,,
k f ” ’ ( A , T ) dh. (5.128)
It can be seen that E ~ is ,the ~ weighted

~ average of the emissivity over
the spectral band, where the weighting function is the radiant emittance
of a blackbody at temperature T. Whereas the spectral emissivity is a
function of temperature through the variation of material properties and
the average emissivity is a function of temperature through the variation
of material properties, the average emissivity also depends upon the
temperature via the weighting function M ( h , T ) . These are the spectral
categories of emissivity: spectral, total, weighted average, power, and
quantum.
Emissivities can also be categorized in terms of the geometric relation-
ships. We have already dealt with the hemispherical reflectivities,
&h = M / M ~ ’ in their various spectral guises. Directional emissivities
are important for samples which are not close to Lambertian. These
emissivities are defined in terms of radiances. The spectral directional
emissivity then is
(5.129)
It is true that a blackbody is Lambertian so that LYB(A)is not a function

of 8, and 4.
At one extreme of the measurements of emissivity is the total hemi-
spherical emissivity. The essence of these measurements is the determination
of the temperature of the sample and the total flux emitted by the sample.
Three notable procedures are: the hot-filament method, the temperature-
decay method, and the radiometric method.
In the first method, current is caused to flow through a long strip of the
(metallic) material in question. Typical four-terminal measurements are
made of the current and voltage in the strip. The temperature is measured
by a thermocouple, a platinum resistance thermometer, or an optical
pyrometer. This choice is made based on the temperature range and the
personal preferences of the investigator. The inaccuracy of this method is
MEASUREMENTS 27 1
from 2% to 20%. The contributions to this total error include unaccounted

flux losses by conduction along the wire, uncertainty in the temperature
measurement (usually due to the fact that thermometer does not measure
the temperature of the sample), uncertainty in the net flux transfer from
the walls (because it is the net flux that counts), and any convection losses
if the system is not evacuated.
In the temperature-decay method, a sample of small mass and large
surface area is heated inside a cooled, evacuated chamber. It can be heated
in any of the standard ways-resistively, inductively, optically-so that
conducting or insulating samples can be measured. The temperature is
measured as before, and the rate of cooling is noted. The inaccuracy of this
method is also from 2% to 20% with most of the methods yielding about
5%. The errors are due to the same phenomena as before: temperature
errors and nonradiative flux transfers.
The radiometric method utilizes a large specimen with a small detector
next to it. Sometimes a reflector or light pipe is used to help define this
hemisphere (or almost hemisphere). Calibration of the detector against a
blackbody completes the technique. The uncertainty in this method is 5%
to 10%.
Total normal emissivity, a special case of total directional emissivity, is
also done in three different ways: emissometer, comparative radiometry,
and by integrating spectral normal emissivity. The emissometer is an evacu-
ated chamber which has cooled, blackened walls. The detector, a thermopile,
views a cooled wall from the back and a sample through a KRS-5 window
from the front. The field of view is a fairly large cone. The technique is
comparison of the sample flux density to that from a reference-when the
two are at the same temperature. This is more an inspection technique than
a precise radiometric measurement. The method of comparative radiometry
uses a black detector which views a blackbody at a known temperature and
then views the sample. The sample temperature is measured as before. The
sample can also be heated to any appropriate temperature, although there
is no requirement for heating rate, temperature level, etc. The measurement
is made at the temperature required. The measurement uncertainty depends
upon the temperature required. The measurement uncertainty also depends
upon the wavelength and temperature values involved. The radiation fields
of the sample and the blackbody are being measured; all the caveats and
techniques of field measurements apply. The final technique is integrating
the spectral normal emissivity, although that is really the hard way. These
measurements have an uncertainty ranging from 2% to 3% for high-
temperature samples. The measurement of directional emissivity can be
made by a geometrical adjustment of the comparative radiometric method
for normal emissivity.
5.7.4. The Measurement of Reflectivity

The reflection of light from a sample is a function of the spectral interval
and geometry and depends upon the nature of the source, the sample, and
its surface. The spectral considerations are the same as for emissivity and
will not be discussed here. The different geometric concepts apply for
monochromatic, total, and band-limited radiation. These are the spectral,
total, and weighted average values. The main difference between emission
and reflection is that the latter is a function of the source geometry as well
as the receiver geometry.
The well-known law of reflection states that an optically smooth surface
reflects the light in the plane of incidence and at an angle equal to the angle
of incidence. Thus, the ratio of the flux density in a collimated beam to the
incident flux density also in a collimated beam is the specular reflectivity;
it is a function of the angle of incidence for any material. If all the above
conditions are met, then Fresnel's equations apply. Note that the specular
reflectivities are functions of the state of polarization and the refractive
index of the material.
If the reflected radiance is not a function of angle, i.e., if the reflected
radiance is uniform, then the surface is said to be a uniform or Lambertian
reflector. There is no such thing as a true Lambertian reflector which
generates uniform reflected radiation, no matter what the angle of incidence.
Some materials, like magnesium oxide, barium sulfate, and sodium chloride
are Lambertian out to about 70" or 80" and for angles of incidence smaller
than about 30". Many real objects like paint, sand, dust, and fabric are also
Lambertian to the extent just described.
Any general surface consists of both specular and diffuse components.
The reflectivity must be described in a more general way. J ~ d has d ~ ~
described nine different types of reflectivity. The beam can be collimated,
conical, or hemispherical. The illuminating and collecting beams can be
any of these. The combination comes to nine. The most general is conical-
conical, and the others are limiting cases. In every case, the reflectivity is
the ratio of two flux densities. A more useful concept is the bidirectional
(or bistatic) reflectance distribution function (BRDF). This is a complete
description of the radiance reflected in some direction 8, and 9,as a result
of the flux density incident on the surface from some incident direction Bi, 9i:
(5.130)
In most radiative transfer problems, the irradiance on a surface can be

calculated from the source radiance and the geometry. The next step is to
MEASUREMENTS 273
use the reflected radiance as the source for the next transfer. The BRDF is
just the right function for this.
Usually the spectral nature of the measurements is determined by the
source and the detector. Only one system (Gier-Dunkle) uses a different
method. The spectral region is determined by such difficulties as ambient
fluxes, gradients, etc.
The main methods for the measurement of normal specular reflectivity
are those of Strong” and Bennett and Koehler.’l These are illustrated in
Figs. 19 and 20. In the Strong method, light shines into the sample area
that has no sample. The beam reflects once from the folding mirror. Then
the mirror is rotated to its sample position and the sample is inserted. Now
the beam reflects once off the folding mirror and twice off the sample. The
geometry of the apparatus prevents true normal-incidence measurements.
They certainly are directional. Strong’s apparatus operated between 0.3 pm
and 10 pm, with the spectral region limited primarily by the integrating
sphere used for putting uniform flux on the detector. The inaccuracy is
about 0.1%. The diagram of the Bennett-Koehler apparatus shows how it
works. The beam with and without the sample travels along the following
FIG. 19. Sketch of the Strong-type reflectometer used in the visible. The sample reflectance
R is obtained from the square-root of the ratio of two oscillograph galvanometer deflections
G , and G, which correspond to SAMPLE-IN and SAMPLE-OUT.
FIG.20. The Bennett-Koehler reflectometer.
sets of mirrors: MI to M 6 , and MI3 to Mzl.These are the input and output
groups, respectively. When the sample is in, as shown, the light follows the
path M 6 - M , - S - M 8 - S - M 9 - M 1 3 .There are two reflections off the sample
as with the Strong reflectometer. When the sample is out, the path is
M6-M,-Ml1-M9-Ml3.The reflection of the mirror M I , replaces that of
two sample reflections and M 8 . The ratio is the square of the sample
reflectance, except for the ratio of M , to MI . To correct for this, the sample
as well as mirrors M6 and MI, are rotated 180". Then the light follows the
dashed lines with the sample in, i.e., M13-Ml,-S-M,l-S-M12- MI3.With
the sample out, the path is Ml3-MIO-M8-MI2-Ml3. This ratio is then
S-MII-S to M 8 . Thus the ratio from the first two measurements would be
S 2 M 8 / M , , ,and that from the second would be S 2 M l , / M 8 The . ratio of
these ratios gives the fourth power of the sample reflectance. The sample
labeled T is a transparent one, the transmission of which can be measured
with a sample-in sample-out technique. Although this technique does require
the extra measurements for its ultimate accuracy (about 0.001), it does not
require the use of an integrating sphere as does the Strong system (or it
would be very sensitive to small misalignments of the sample).
5.7.5. The Measurement of Directional Reflectance

The bidirectional reflectance distribution function is, as the name implies,
a distribution. This means that true measurement is that of an infinitesimal
and can never be carried out perfectly. Differentials must replace derivatives,
M EASUREM ENTS 215
and the requirements for satisfactory signal-to-noise ratio dictate that these
quantities are sometimes larger than one would really desire.
The geometry of the measurement is shown in Fig. 21. The angles of
incidence are Oi and +i; those of reflection are Or and 4,. The sample is
placed with its normal along the z axis and is assumed to be flat (on the
average). Two geometries have been used in measurement^.^^'^^ The first
emulates the diagram just discussed; the second places the beam and
detector in a horizontal plane and tilts and rotates the sample. The first
geometry requires no coordinate transformation to describe the defining
geometry, but it is not capable of measuring with detectors that are cooled
in dewar flasks with liquid refrigerants.
In both cases, the azimuthal angle of incidence is chosen to be zero. This
is a free choice for the origin of the coordinate system. The incident flux
is almost always a laser beam which has been expanded and clipped by an
appropriate pinhole. It is then aimed at the sample as a slowly convergent
beam that is focused at the detector position. The beam could be collimated,
but this would eliminate one of the two methods of calibration described
below. The function of the beam is to illuminate a portion of the sample
with a uniform irradiance. The sample then scatters light to a detector-
receiver assembly-a form of radiometer. The sizes of these can be deter-
mined to some extent by writing the definition of BRDF in its limiting form:
Ai A n d
BRDF= lim -, (5.131)
L\A,-O
an-0
ai
-
AAi
where AAi is the illuminated area and Andis the solid angle of the detector
as subtended at the sample (=Ad/d2). The flux on the detector is an
integration over the sample area and the detector area. It is said that the
BRDF is measured at angle Or, but it is really an average from O,-AO, to
O,+AO, and over all points on the sample. Reasonable values are d =
1 m, = 1 mm. Then the averaging range is 1 mr. For most purposes, this
angular resolution is finer than necessary. The illuminated area should be
large enough to adequately represent the sample unless several measure-
ments are to be made of several portions of the same sample. Rough samples
typically have roughness zones as large as 500 pm. Thus a sample illumina-
tion of 5 mm is about right to characterize these. An F/100 beam will
encompass a 1-cm sample and focus at a detector 1 m away from the sample.
But the beam may not be uniform.
For specular reflection methods, a ratio technique may be used for
calibration. Somewhat similar approaches have been tried for these direc-
FIG. 21. BRDF angles. The incidence angle is Bi. The angle of specular reflection is Bo.
The arbitrary polar angle of reflection is 8,. The azimuthal angle of reflection is br.
tional methods. The two main calibration techniques are what we have
called the “reference method” and the “no-sample method”. We can gain
an understanding of them from an analysis of the detector output voltage.
It is found to be
%LAdCOS 8dA, COS 8,
v=%!Pd= (5.132)
d2
The reflected radiance L is by definition equal to the sample irradiance
times its BRDF. Thus,
v, = % E , P h A d COS8dA, COS os
(5.133)
d2
The detector is always arranged to be perpendicular to the radiometer line
of sight, so cos8d is 1 . If there is no sample, then all of the flux at the sample
area gets to the detector, and
VNs= %!E,Ai.
If we take as the sample area the illuminated portion only, then A, = Ai.
Therefore,
(5.134)
The BRDF is given by the product of the voltage ratio and the reciprocal
of the projected solid angle of the detector.
MEASUREMENTS 277
Alternately, one can use a reference whose BRDF is known. Then
(5.135)
where VK is the voltage from the detector when the known sample with
BRDF pK is in place. This is simple in concept and calculation but requires
a reference whose BRDF is known for all conditions and comparison
measurements made at all positions. The obvious question then is: “How
did you measure the reference?” The procedure which seems to devolve
from this is to use a Lambertian reference. It has a BRDF which is Phr-’,
where Ph is the hemispherical reflectivity. Then
(5.136)
But there is no such thing as .a true Lambertian surface. Some materials,

as was discussed earlier, are very good approximations when they are
illuminated normally. Then
(5.137)
where V(0) is the voltage resulting from the reference illuminated normally.
This approach has the advantage that the instrument parameters are not
involved; however, a separate measurement must be made of Ph.
Measurements of specular, hemispherical, and directional transmittance
follow those of reflection with an obvious change in geometry.
5.7.6. The Measurement of Laser Power

There are basically two ways to measure laser power: with an appropriate
detector or with a calorimeter. The latter technique has come into consider-
able use, because lasers usually have enough power to generate a useful
signal and because a calorimeter is color-blind; it responds in the same way
to laser radiation of any wavelength. One example of a calorimeter has
been given by NBS. Figure 22 is a schematic diagram of the instrument. A
laser beam enters the baffle section from the left (as usual) and is absorbed
in the convoluted shaped cavity-a cone plus cornucopia shape. The water
of the reservoir is pumped through the walls of this cavity and its temperature
is raised. The increase in temperature is related to the input energy. The
relationship is determined by heating the cavity electrically and monitoring
FIG. 22. Schematic diagram of an NBS laser calorimeter. Explanation in the text.
the current, voltage, and time. Nitrogen is flowed through the cavity to
reduce scattering, and temperature sensors are used to record any stray
absorption in the baffle structure. They claim 13% single-measurement
uncertainty for power levels of about 100 kw in the wavelength range from
1 to 11 pm (13 kW). The unit is 1.2 m high, 1.5 m long and 0.65 m wide and
weighs about 400 kg.
Although they do not discuss the sources of uncertainty, one can attribute
them in general to differences between the measurement and the calibration.
It is unlikely that the calibration source is 100 kW; it is probably considerably
less. The errors then can arise from different amounts of conduction and
convection of heat away from the cavity. Although the baffle structure is
monitored, some errors will arise from spillover onto it. The relative contri-
bution of heating from the pump will be different, and no cavity is a 100°/~
absorber.
Calorimeters exist for measuring laser powers from milliwatts to hun-
dreds of kilowatts. The uncertainties are lowest, of course, for the middle
of the range where the calibration can be very similar to the measurement
and where there is sufficient energy that it causes an appreciable temperature
rise. Of course, for fluxes of less than 100 kW, the instruments use smaller
reservoirs, sometimes just the heating of an appropriate detector mass. At
the lower end of the scale, direct detection by pyroelectric detectors is very
competitive with these calorimetric techniques. When the measurements are
made in the visible, silicon detectors can be used.
MEASUREMENTS 279
5.7.7. Synchronous Detection

When the investigator has control over the source of the radiation, he
should use the technique of synchronous or lock-in detection. Some older
radiometers of this type were said to have synchronous rectifier amplifiers
(or SRAs). A small source of radiation is placed close to one side of the
blades of the chopper, and a small detector is placed facing it on the other
side. In this way, both the frequency and the phase of the chopper is
measured. I f the optical frequency is labeled vo, the electrical frequency
ver and the chopper frequency v,, then the electrical signal will be given by
V( t ) =
I I'%%
% cos 2 ~ ~ , t P ( ~ 0 dvo
) dt.
The optical frequencies are so much higher than the electrical frequencies
(5.138)
that very many optical cycles are integrated, and P ( v o ) may therefore be
considered a constant, Po. We are also assuming that the only temporal
variations in the power are in the modulator-which is assumed to be
cosinusoidal. .
V( t ) =
I,%Po cos 27~v,tdt. (5.139)
One nonsynchronous way to handle this signal is to consider the frequency

domain version of P ( t ) which is just a delta function at v,. Then a very
narrow bandpass filter centered around v, may be used. This is not syn-
chronous detection; it is simply narrow-band detection by which much
noise is eliminated. Establishing a narrow bandpass at high frequencies is
difficult (10% of v, is good!), and so a heterodyne approach is used in the
synchronous system. An electronic bipolar switch is used. This can be
represented as a rectangle function. Then the output is given by
V( t ) =
I, %Po cos 27rv,t rect( v,t - 4 ) dt. (5.140)
The phase of the electronic switch is set relative to the zero phase of the
chopper. If it is zero, then the output V ( t ) is just
V( t ) =
I, %P0lcos 2 ~ v , t ldt. (5.141)
Nonzero phase shifts will provide different values for V( t ) . The final output
voltage is obtained by integrating over the time constant 7 which is set by
a low-pass filter. It is worth considering all of this more generally and in
the frequency domain.
The output is given by the product of the Fourier transforms of the

modulation function, the chopper function, and the low-pass filter. The
latter is usually adjustable on the console of the amplifier to values of
integration times ranging from 0.01 s to 1 s (f=0.16 to 0.60 Hz).This filter
can be modeled reasonably well by a single time constant filter whose
frequency characteristic is (1 + j w T ) - ' . The switch can be modeled as a
rectangle function with values of +1 and -1. The modeling of the temporal
waveform of the flux is a much more complicated matter. If the flux is all
concentrated in a spot that is small with respect to the opening of the
chopper, it can be modeled as a rectangle function or square wave as well.
The modulation and flux waveforms can then both be represented as square
waves, and their Fourier spectra are sinc functions. Now we can see for
these assumption that the phase relationship is indeed important. There is
a direct and linear relationship between the percentage of error in period
matching and in the flux that is sensed. Of course, the flux waveform will
not be a perfect square wave as indicated, but it will be the convolution of
the shape of the spot with the chopper. Detailed considerations of the
various geometries are beyond the scope of this treatment, but the principles
have been established. The investigator must consider the power that is
converted to frequencies above the cutoff of the low pass as well as errors
in phase or frequency adjustment, and as a result of any undue spreading
of the size of the spot. Some lock-in systems now use quadrature addition;
they eliminate errors due to phase shifts.
5.7.8. Integrating Spheres

In many radiometric measurements, one wishes to provide uniform radi-
ation on a detector or a sample. The so-called integrating sphere is one of
the methods for doing this. The basic concept is that light entering a port
in the sphere is reflected diffusely many times until it becomes well random-
ized. It then exits a second port as diffuse illumination. These results are
derived in two steps. The first is the very simple theory described many
years ago. The second is a more useful, more complete treatment.56
The theory is based on some of the geometric properties of the sphere.
Figure 23 shows radiative transfer between two elements of the sphere. The
general expression for radiative transfer is
,
@ = L dA cos 0, dA2 cos 0,/ d '. (5.142)
In this case, the angles 0, and O2 are equal (0, = O2 = 0). Further, the
center-to-center distance is d = 2R cos 0. Thus the flux on dA2from dA, is
MEASUREMENTS 28 1
FIG.23. Integrating sphere geometry.
given by
@ = L dA, dA2COS’ 8/4R2 COS’ 8
= L dA, dA2/4R2. (5.143)
Since the points were chosen arbitrarily, this shows that dA, will irradiate
all elements dA, uniformly. Now if Ed is the irradiance on element dA,
caused by a source shining through the entrance port, then for a Lambertian
surface the total diffuse irradiation is
Ed,-=pEd+p’Ed+p’Ed+. * .
When the direct irradiation is added to this, the total becomes
Ed
E = ~ ~ ( i + ~ + .)=-
~ * + . (5.144)
(1 -PI’
The resultant light is diffused and is an inverse function of the absorptivity
of the wall. The hemispherical reflectivity p has a maximum value of
-
T-’ 0.3. Thus the ratio of total irradiance to direct irradiance is at most
about 1.4.
A more detailed story described by Goebel and others is given be lo^.^',^^
Imagine a sphere divided up into n spherical caps, each of which rep-
resents an entrance port, exit port, or some other discontinuity in the wall
of the sphere. It can be seen from Fig. 2 that the fractional area of any of
these caps is given by
2Rh
f = -4.nR2-
- - - 2R-
h
2R
=”
R - J R ’ - ~ ~1 -
2
Jq] = f ( l -C0S2 0)
= sin’ 8. (5.145)
The radiation entering is labeled O0and reflects from area a, an amount
p0@, uniformly in all directions. I f f ; = ai/A,, where a, is the ith special
area and A, is the spherical area, then the average wall reflectivity or Pw is
n
Pw = c Pf; + P w ( 1 -EL). (5.146)
This is just the area-weighted reflectivity. The unabsorbed flux after the
second reflection is pwpoQo.This is uniform and is distributed over the wall
and the special areas so that the flux after the third reflection is just p ~ p o @ o .
The total unabsorbed flux after an infinite number of reflections then is
po - po/ ( 1 - p w ) . The flux on aperture a, then is given by
f;P O Q O
-@) + c 4 ’
(5.147)
= [ 1- P w ( 1
and the ratio of the flux out of ai to the input flux Q, is the cavity efficiency
E =&Po[ 1 - P w ( 1 -if;) (5.148)
A simple two-port integrating sphere where a; is the illuminated portion

of the wall so that po= p w ; the entrance port has area a , ( p , = 0); and the
Fractional Hole Area
FIG.24. Sphere efficiency as a function of the fractional hole area for different numbers of
holes.
MEASUREMENTS 283
8.6 -
0.5 -
0. 4 -
8.3 -
0.2 -
0.1 -
Fractional hole area
FIG. 25. Sphere efficiency as a function of the fractional hole area for different wall
reflectivities.
exit port has area a&, = 0). Then
-@)] .
-I
& =hw[
1-Pw( 1 (5.149)
In this form, we can see that the simple model above is modified by J p ,
in the numerator. This represents the amount of flux getting out the ith
hole. It is also modified by 1 -ZJ; to indicate that some of the wall surface
is not available for reflection. Curves of sphere efficiency for different area
ratios and coatings are shown in Figs. 24 and 25. The message is clear.
Keep the reflectance high; make the ports as large as the uniformity consider-
ations allow. Another message is apparent: there is a strong dependence
on the value of p w . Any spectral variations in the coating are magnified in
the sphere. The relative error in efficiency is
(5.150)
Notice that a 1% error in p for a p of 0.9 gives a sphere error of about

10%. Other uses of the sphere can be analyzed in a similar way.
Unfortunately, there is no such thing as a truly Lambertian source, so
the theory described above is subject to question. Fortunately, the manifold
reflections serve to help the diffuseness of the result.
Although it is beyond the scope of this treatment, we can mention the
more general treatment and refer the reader to it.59
5.8. Photometry: Radiometry of Visible Light
The visible region has a very special significance in the entire electromag-
netic spectrum. Only visible light can be detected in a useful way by humans,
and the process of seeing is important to work, leisure, recreation, and
everyday living. Accordingly, a special branch of radiometry, photometry,
deals with flux levels of radiation as they generate a response in the eye.
A typical problem is the calculation of the proper light source to illuminate
a work surface for a particular function. Psychologists have measured the
appropriate light levels for performing many different tasks. Light sources
with different spectral distributions but the same total radiation elicit
different amounts of response from the eye according to how the spectrum
of the source matches that of the eye. This can be expressed in terms of
response of the eye at its maximum K,the spectral distribution of the eye’s
response V ( A ) ,and the spectral distribution of power:
P,= K
jom V ( A ) P ( A )dA [lm]. (5.151)
The visible flux P, can be thought of as “light watts” or “eyeball watts.”

It is measured in lumens. The efficiency with which any source provides
lumens can be written as
PV
K jOm V ( A ) P ( A )dA
0v=-= (5.152)
pe j9mP(A)dA
Therefore, for any source, the luminous output is
P,= qv
jOWdA
P(A) = vvP. (5.153)
All the geometric relations which apply to flux in general apply to luminous
flux in particular. The main differences and special features rest in the
PHOTOMETRY: RADIOMETRY O F VISIBLE LIGHT 285
handling of the spectral normalization and the many units and names for
them which have grown like topsy with this area of application.
5.8.1. Normalization and its Use

In the previous section, it was shown that the luminous flux can be found
from the product of the luminous efficiency and the flux of energy, the
power. The response of the eye can then be obtained directly. We show
that this is true by writing the responsivity of the eye as a function of
wavelength as %(A). Then the output V is
V= [a
3
-W
%(A)Pe(A)dA. (5.154)
Each of the quantities in the integral can be written as a constant times a

normalized distribution:
P ( A 1 = @(A ),
%(A) = %r(A).
Then the output is
V = BP, [m
-m
r(A)p(A) d. (5.155)
But r ( A ) is the same as V(A), so that
j-,
m
V = %P, V(A)p(A) dA. (5.156)
The constant % is proportional to K, the maximum of the luminous efficiency

curve, so that
V=(const) K [ W
-a
V(A)@(A) dh
(5.157)
The constant depends upon one’s choice of what constitutes a barely

perceptible constant level, a flicker detection, enough detection for the
performance of detailed tasks, or some other criterion. This is a matter left
for the applications and the psychophysicists.
The effects of spectral distribution are, however, a different matter. They
arise from careful considerations of the response of the eye and calibration
of sources. It is known that the scotopic response of the eye (under low
levels of illumination) is redder than the photopic response (under normal
levels). This means, of course, that r ( h ) is changed. Another, more subtle

effect relates to the distribution of the source used for calibration. Many
photoelectric detectors are calibrated with a tungsten lamp source operated
at a temperature of about 2800°K. These lamps are almost blackbodies and
even closer to gray bodies. Accordingly, the output can be written
m
V = j p m s l ( A ) P ( A )dh. (5.158)
The luminous efficiency is written
vv= Im
-m
K V ( A ) P ( A )dh Im -m
P ( h ) dh
=K I m
-m
V ( h ) p ( A )dh I*-m
p ( h ) dh. (5.159)
Thus for a calibration source with spectral distribution of flux given by Pc,
one has for the luminous responsivity
slv= V I P c =
9
I r ( A ) P c ( A )dh
(5.160)
K
I V ( h ) P c ( A dh
)
The flux distribution from some other source P , ( h ) evokes an output

given by
V=
I s l ( h ) P , ( h )dA
(5.161)
= P, r ( h ) p , ( A ) dA.
By substitution for 3,one can obtain
I V ( A ) p c ( h )dA
- ps I
V=%,K
I ( A ) p c ( A 1 dh
r ( A ) p , ( h ) dh.
Divide both sides by Pv[ = P,K V ( A ) p , ( A )d h ] .Then the output is given by
v = slvPv
I V(A)pc(A)dh J r ( h ) p s ( A ) dh
(5.162)
REFERENCES 287
This result deserves discussion. The output is not always the luminous flux
on the receiver times the luminous response of that receiver. It is true if
the integral factor is equal to 1. For that to be true, either r ( A ) must equal
V(A), or p c ( A ) must equal p , ( A ) . If one calibrates a n eye with p c a n d then
uses the eye, such is the case. If one calibrates a TV tube with a tungsten
lamp a n d then uses the same kind of tungsten lamp t o illuminate the televised
scene, it is also true. It is not generally true, n o matter how logical it sounds.
This should be borne in mind in the photometric calibration of TV a n d
other kinds of tubes, which d o not usually have reponses that match the eye.
5.8.2. Photometric Units

In radiometry, in which watts are the unit of flux, the units of sterance
and concentration are W m-’sr-* a n d W m-’ (or some appropriate vari-
ation). I n photometry these units become, respectively, Im m-’sr-’ a n d
Im m-’. These are surely true, but history a n d usage have provided other
units. Other references are available for a discussion of these complicated
units.60
References
1. S. Chandresekhar, Radiatiue Transfer, Dover, New York, 1960.
2. G. Bauer, Measurements of Opfical Radiation, Focal Press, London, 1965.
3. R. A. Smith, F. E. Jones, and R. P. Chasmar, The Defection and Measurement oflnfrared
Radiation, Clarendon Press, Oxford, 1968.
4. M. Planck, “Zur Theorie des Gesetzes der Energieverteilung im Normalspectrum,” Ver-
handlungen der Deutschen Physikalischen Gesellschafr 2, 237 ( 1901).
5. W. Herschel, “Experiments on the Refrangibility of the Invisible Rays of the Sun,” Phil.
Trans. Roy. SOC.90, 255 (1800).
6. 1. W. Ritter, as described in Optical Anecdotes, p. 55, by D. J. Lovell, SPIE, Bellingham,
MA, 1981.
7. U S A Standard: Nomenclature and Dejinitions for Illuminating Engineering RP- 16. Unfted
States of America Standards Institute.
8. R. C. Jones, “Terminology in Photometry and Radiometry,” J. Opt. SOC.Am. 53,1314 (1963).
9. F. E. Nicodemus, as reported by 1. Spiro in Opt. Eng. 13, G183, (1974), and C. L. Wyatt,
Radiometric Calibration: Theory and Methods, Academic Press, New York, 1978.
10. A. G. Worthing and D. Halliday, Heat, Wiley, New York, 1948.
11. A. G. Worthing, in Temperature: Its Measurement and Control in Science and Industry,
Reinhold, New York, 1941.
12. W. L. Wolfe and G. J. Zissis, The InJrared Handbook, Office of Naval Research, Washington,
D. C., 1978.
13. J. Strong, Concepts of Classical Optics, Freeman, San Francisco, 1958.
14. M. G. Dreyfus, “Spectral Variation of Blackbody Radiation,” Appl. O p t 2, 1113 (1963).
15. R. D. Hudson, Infrared System Engineering, Wiley, New York, 1969.
16. M. R. Holter, S. Nudelman, G. H. Suits, W. L. Wolfe, and G. J. Zissis, Fundamentals of
InJrared Technology, Macmillan, New York, 1962.
17. P. Kruse, McLaughlin, and McQuistan, Elements of Infrared Technology, Wiley, New York,
1962.
18. M. A. Bramson, Infrared: A Handbook for Applications, Plenum, New York, 1968.
19. R. H. Kingston, Detection of Optical and Infrared Radiation, Springer-Verlag, Berlin, 1978.
20. J. N. Howard, Book review of Seven Place Tables of the Planck Function for the Visible
Spectrum, by D. Hahn, J. Metzdorf, B. Schleef and J. Seich, Academic Press, 1964, in
Appl. Opt. 4, 808 (1965); other volumes are mentioned.
21. E. M. Sparrow and R. D. Cess, Radiation Heat Transfer, Brooke-Cole, Belmont, Calif. 1966.
22. F. E. Nicodemus, “Radiance,” Am. 1. Phys. 31, 368 (1963).
23. J. Rainwater, “Generalization of the Abbe Sine Law in Geometric Optics,” Am. J. Phys.
32, 626 (1964).
24. R. Gardon, “The Emissivity of Transparent Materials,” J. Am. Ceram. SOC.39,278 (1956).
25. H. 0. McMahon, “Thermal Radiation from Partially Transparent Reflecting Bodies,”
1. Opt. SOC.Am. 40,376 (1950).
26. M. Czerny and L. Genzel, “Uber die Eindringtiefe raumlich diffuser Strahlung in Glas,”
Glastech. Ber. 25, 134 (1952).
27. D. Q. Wark and H. E. Fleming, Monthly Weather Reo. 94, 351 (1966).
28. H a n d and Conrath, Science 165, 1258 (1969).
29. J. C. Gille, “Inversion of Radiometric Measurements,” Bull. Am. Mer. SOC.49, 903 (1968).
30. L. D. Kaplan, “Inference of Atmospheric Structure from Remote Radiation Measure-
ments,” J. Opt. SOC.Am. 49, 1004 (1959).
31. S. Twomey, J. Geo. Res. 66, 2153 (1961); 1. Assoc. Comp. Mach. 10, 97 (1963); J. Frank
Inst. 279, 95 (1965); Monthly Weather Reo. 91, 659 (1963).
32. W. L. Wolfe and H. P. Stahl, “Some Calculational Results Using Multicolor Radiation
Inversion,” IR Phys. 20, 293 (1980).
33. W. L. Wolfe and E. L. Dereniak, in Imagingfor Medicine, Vol. I Chap. 13 (S. Nudelman
and D. D. Patton, eds.), Plenum, New York, 1980.
34. E. L. Dereniak, Ratio Temperature Thermography, Ph. D. Thesis, University of Arizona,
Tucson, 1976.
35. Electro-Optical Industries Brochure.
36. R. Hofmann, S. Drapitz, and K. W. Michel, “Lamellar Grating Fourier Spectrometer for
a Balloon-borne Telescope,” IR Phys. 17, 451 (1977).
37. F. 0. Bartell and W. L. Wolfe, “Cavity Radiators: an Ecumenical Theory,” Appl. Opt. 15,
84 (1976).
38. J. M. Lloyd, Thermal Imaging Systems, Plenum, New York, 1975.
39. J. D. Gaskill, Linear Systems, Fourier Transforms and Optics, Wiley, New York, 1978.
40. W. G. Driscoll and W. Vaughan, Handbook of Optics, McGraw-Hill, New York, 1978.
41. R. A. Sawyer, Practical Spectroscopy, Dover, New York, 1963.
42. R. J. Bell, Introductory Fourier Transform Spectroscopy, Academic Press, New York, 1972.
43. A. C. Candler, Modern Interferometers, Hilger and Watts, London, 1951.
44. R. D. Saunders and J. B. Shumaker, Optical Radiation Measurements: The 1973 Scale of
Spectral Irradiance, U. S . Government Printing Office, 1977.
45. J. Geist and W. R.Blevin, “Chopper Stabilized Null Radiometer Based upon an Electrically
Calibrated Polyelectric Detector,” Appl. Opt. 12, 2532 (1973); W. M. Doyle, B. McIntosh,
and J. Geist, Roc. SOC.Phot. Inst. Eng. 62, 166 (1975); and R. J. Phelan and A. R. Cook,
“Electrically Calibrated Pyroelectric Optical Radiation Detector,” Appl. Opt. 12, 2494
(1973).
46. J. Geist, E. F. Zalewski, and A. ‘R. Schafer, “Spectral Response Self-Calibration and
Interpolation of Silicon Photodiodes,” Appl. Opt. 19, 3795 (1980).
47. National Bureau of Standards, Optical Radiation News, 20 April (1977) and 24 April (1978).
REFERENCES 289
48. F. E. Nicodemus, J. C. Richmond, 1. W. Ginsberg, and T. Limpens, Geometrical Consider-

ations and Nomenclature for ReJlectance, NBS Monograph 160, U.S. Dept. of Commerce,
1977.
49. D. B. Judd, “Terms, Definitions and Symbols in Reflectometry,” J. Opt. SOC.Am. 57, 445
(1967).
50. J. Strong, Procedures in Experimental Physics, Prentice-Hall, New Jersey, 1938.
51. H. E. Bennett and W. F. Koehler, “Precision Measurements of Absolute Specular Reflect-
ance with Minimized Systematic Errors,’’ J. Opt. SOC.Am. 50, 1 (1960).
52. J. W. Harvey, Light-scattering Characteristics of Optical Surfnces, Ph. D. Thesis, Optical
Sciences Center, University of Arizona, Tucson, 1976.
53. F. 0. Bartell, E. L. Dereniak, and W. L. Wolfe, “The Theory and Measurement of
Bidirectional Reflectance Distribution Function (BRDF) and Bidirectional Transmission
Distribution Function (BTDF),” Proc. SOC.Phot. Instr. Eng. 257, 157 (1980).
54. NBS Dimensions, April, 1977.
55. J. A. Jacques and H. F. Kuppenheim, J. Opt. SOC.Am. 45, 460 (1955).
56. D. G. Goebel, “Generalized Integrating-Sphere Theory,” Appl. Opt. 6, 125 (1967).
57. B. J. Hisdal, “Perfect Samples in Integrating Sphere,” J. Opt. SOC.Am. 55, 1122 (1965)
and “Non Perfect Surfaces in Integrating Sphere,” J. Opt. SOC.Am. 55, 1255 (1965).
58. F. 0. Bartell, Blackbody Simulator, Cavity Radiation Theory, Ph. D. Thesis, University of
Arizona, Tucson, 1978.
59. J. Meyer-Arendt, “Radiometry and Photometry: Units and Conversion Factors,” Appl.
Opt. 7 , 2081 (1968).
6. DETECTORS
T. 0. Poehler
Applied Physics Laboratory
Johns Hopkins Road
Laurel, Maryland 20707
6.1. Introduction
Quantitative measurements of optical radiation are usually accomplished

by converting the radiant energy into electrical signals. The electrical output
consists of the response of the detector to the radiation, which is designated
the signal and random fluctuations known as noise. Detectors may be
characterized by the physical mechanisms by which they convert optical to
electrical energy, the spectral range over which they respond, the rate at
which they respond, the minimum radiative power that they can detect, and
the amplitude of the electrical signal generated for a unit of incident
radiation.
Optical radiation may be detected by either thermal detectors where an
increase in temperature of the sensing element exposed to radiation is
observed, or by quantum detectors where photons are directly converted
into the electrons or mobile charge carriers in the detector. Many physical
effects have been used to detect optical radiation. Devices based on either
a photoexcitation or photoemission of electrons have produced detectors
with the highest performance. Devices in this general category, which on
the one hand include photomultiplier tubes and semiconductor photodetec-
tors constructed from such elements as silicon or germanium primarily in
a photodiode configuration, have been used to detect most radiation in the
shorter wavelength regimes. On the other hand, certain simple photoconduc-
tive detectors have found a widespread application in the long-wavelength
portion of the spectrum. Detectors based on the conversion of photons due
to a thermal effect within the detector medium have also found wide
application. The primary practical modern device of the thermal variety
currently being utilized is the pyroelectric detector, which is used for both
energy and power measurements over a range of the spectrum extending
from visible to far-infrared wavelengths.
29 1
METHODS OF EXPERIMENTAL PHYSICS Copyright 0 1988 by Academic Press Inc.
ISBN 0-12-475971-8
292 DETECTORS
6.2. Figures of Merit
Since there are a wide range of physical mechanisms employed in radi-

ation detectors, several figures of merit are used to quantitatively characterize
performance, independent of the physical basis of their operation.' A simple
criterion that can be applied to typify the response is the responsivity R ",
generally defined as the ratio of the rms value of the signal voltage, V, to
the rms power, P, incident on the detector and specified in units of volts/watt.
The total power incident on a detector of area A is associated with an
irradiance H (watts/cm2), so that the responsivity is
vs vs
R - -= -(volts/ watt).
"-P HA
The responsivity is usually a valid description of the detector performance
over the three to six orders of magnitude of radiant power that is typical
of the range over which detectors respond with a linear transfer function.
The responsivity may be specified either as R( Tf),which represents the
output voltage resulting from input radiation from a blackbody at a tem-
perature T with a signal derived from a source modulated at a frequency
or R(A,f), which is the responsivity signal associated with a monochro-
matic source at a wavelength A with modulation frequency f:
The limiting sensitivity of any detector is determined by random fluctu-
ations which exist in its output, and hence it limits detection to a certain
minimum detectable power for any system. The power required to generate
an output signal equal to the noise is frequently referred to as noise-
equivalent power (NEP). To determine the noise-equivalent power requires
measurement of the amount of radiative power from a blackbody source
which must fall on a detector to yield an rms signal V, equal to that
generated by the detector noise VN when it is shielded from the source.
The noise-equivalent power must be specified for a particular source tem-
perature, signal modulation frequency, and amplifier-system bandwidth.
The NEP for detection under a particular set of operating conditions
including a 500°K blackbody radiation source modulated at 900 Hz with a
1-Hz bandwidth is written as
NEP( 500"K,900 Hz, 1 Hz) = -= --

P HA
-HA
S / N S / N - v.1V N
(watts), ( 6 . 2 )
where P is the detected power, H the irradiance, A the detector area, S

and N are the signal and noise measured, respectively. This latter ratio
may be expressed as either a voltage or current ratio. Often the reciprocal
of the NEP designated as the detectivity is used to characterize some
detectors and is specified in units of watts-'. To make comparison between
T H E R M A L DETECTORS 293
detectors, their respective performance must be reduced to representative

conditions, so the detectivity is often normalized to an amplifier bandwidth
of 1 Hz and a detector area of 1 cm2. This figure of merit’ called D* is
D * = Ll VN ( A )‘I2 =- VS/vN(AfA)’”(cm(Hz)’~’/W)
(6.3)
H P
where Af is the electrical bandwidth in Hz.A detector whose performance
is specified at a particular wavelength A from a monochromatic source is
designated as D,*(A,f;Af).
The performance of low-noise detectors may also be limited by radiative
noise arriving at the detector from the background environment. When
operating under these conditions, a detector is said to be background-limited
in performance or in the BLIP mode. Since this measure of the detectivity
is dependent upon the half angle 6 from which the detector can see
background radiation, a normalization including this dependence on field
of view is used and is designated D**, where
D** = D* sin 6, (6.4)
where the half angle of the field of view is 6.
6.3.. Thermal Detectors
6.3.1. Thermocouple Calorimeters

Thermocouples have been employed to detect radiation since the early
nineteenth century3 subsequent to the discovery of the Seebeck effect. The
thermocouple radiation detector is simply composed of a circuit of two
metal conductors and the junctions between them, where one of the two
junctions is exposed to radiation, thus raising its temperature above the
unilluminated cold junction. A potential difference will be generated by the
Seebeck effect4 between the two junctions, which can then be observed by
a simple voltage measurement. .A number of thermocouples connected in
series are called a thermopile andhave the advantage of a larger amplitude
output signal than that generated by a single thermocouple under ordinary
conditions.
Thermocouple detectors are often made from fine wires attached to a
thin-film radiation absorber whose temperature rise in sensed by the ther-
mocouple.’ The pairs of metals used in these devices include bismuth-silver
and copper-constantan. The sensitivity of such a device is enhanced if it
is placed in a vacuum enclosure with a window that is transparent in the
spectral range of interest; commercially available units are supplied in
evacuated mounts. Devices such as these are usually used for detection of
294 DETECTORS
0.75 mm
4 b -
I Radiation
Bismuth Silver contacts
502
Silicon
FIG. 1. Thin-film thermocouple, fabricated by vacuum evaporation. (After Contreras and

Caddy.”)
infrared radiation in the 1- to 20-pm range where the “black” film is highly
absorbing, but are characterized by low-power-handling capability and long
response times. Thin-film thermocouples fabricated by vacuum evaporation5
of the components often exhibit improved properties in comparison to bulk
devices. These detectors, made by precision photolithographic techniques
with small dimensions and low mass, can yield lower time constants and
impedances better matched to electronic circuitry (Fig. 1). Evaporated arrays
of thermocouples composed of antimony and bismuth are used in commer-
cial thermopiles.
While thermopiles are satisfactory for many spectroscopic applications,
their low-power-handling capability and long time constant make them
inappropriate for many laser measurements. In particular, measurements
of CW laser power in the multiwatt range or pulsed lasers with pulse energies
of several joules require calorimeters capable of withstanding these inputs
without damage. These requirements have stimulated the construction of a
multitude of calorimeters for high-laser-power measurements. Typically,
these instruments use a disc6 or cone’ coated with an absorbing layer to
absorb the laser radiation, while the temperature of the absorber can be
monitored with a thermocouple to obtain an estimate of the energy
deposited. While absorbing discs are common, radiation trapping can be
increased by directing the radiation into a cone which is coated with
absorbing material. Multiple reflections of the incident beam will cause
nearly complete absorption and will also tend to increase the damage
threshold of the detector. A diagram of a typical absorbing cone calorimeter
typical of those used for power and energy measurements is illustrated in
Fig. 2. Commercial calorimeters are available which are capable of measur-
ing power of up to several kilowatts and short-time-duration pulses of
greater than 1000joules, using these principles. With the absorbing coatings
generally used on these calorimeters, the spectral range which can be
Black surface
I
Thermal
sensing
element
FIG. 2. Absorbing cone calorimeter, used for power and energy measurements.
measured is typically from the visible spectrum through approximately

30-km wavelength in the infrared (IR).
6.3.2. Bolometer
The heating effect of absorbed radiation can be used to cause a change
in the resistance of a metal or semiconductor from its normal resistance at
room temperature in devices which are classified as bolometer detectors.
This change in resistance can be measured when the resistive bolometer
element is incorporated in a bridge circuit in combination with a matching
bolometer element not exposed to radiation to compensate for temperature
changes in the environment. Initially, bolometers were constructed of evap-
orated metal films, but these devices have been noisy and less sensitive than
thermocouples. Thermistor bolometers composed of semiconductor alloys
have been used in modern infrared work.899
A number of specialized bolometer detectors have been applied to
measurements of incident laser pulses at intensities much greater than those
of interest in other applications. Early work on laser-power measurements
utilized a tangled mass of enamel-insulated wire contained in an enclosure."
These "rats-nest" calorimeters effectively absorbed all the incident optical
energy after a sufficient number of reflections from the wire, and the change
in resistance of the wire was used directly as a measure of incident power.
The system could be calibrated by passing a known current pulse through
the conductor. The upper limit on the optical energy before the thin insulat-
ing coating on the wires was damaged was typically several joules per square
centimeter. Several thin-film bolometers have also been used as room-
temperature detectors of high-power-pulsed 10.6-km radiation.'' The low
296 DETECTORS
mass of these devices permits response times of less than 1 psec., and at
least one device has been reported capable of withstanding 100-5 input
energy.I2
Two important quantities necessary in characterizing the performance of
a sensitive bolometer are the specific heat capacity C and the normalized
temperature coefficient of resistance a, where
1 dR
a=--
R dT'
Here R is the bolometer resistance (ohms) and T is the temperature of the
element (OK). The responsivity of R of an element with a bias voltage V
applied is
R=aV/G, (6.6)
where G is the thermal conductance to a heat sink or bath. A high-
responsivity bolometer thus requires a large temperature coefficient of
resistance, as is observed in most semiconductors at cryogenic temperatures.
Hence, extremely sensitive bolometers can be obtained by using materials
such as carbon,I3 g e r r n a n i ~ m , or
' ~ ~ i l i c o n at
' ~ temperatures below 20°K.
Operation at low temperature also improves the response time T,
75-
C
G'
At low temperatures, the specific heat capacity is, of course, proportional
to T 3so that the speed of response can be made faster according to (T/ Td)3,
far below the Debye temperature, Td.Since both the responsivity and time
constant of a bolometer depend inversely on the thermal conductance,
bolometer performance can be tailored to a particular application by the
degree of thermal contact with the bath. This parameter is usually varied
by altering either the length or the composition of the lead wires to the
element. Carbon bolometers with R L- 1 x lo4V/W at 2°K have been con-
structed, while Ge and Si bolometers with R = 1 x lo5V/ W and response
times of 10 psec have been demonstrated. Since the specific heat capacity
of silicon at low temperatures is eight times smaller than that of germanium,
silicon has a faster response and is used in commercial bolometers. Since
these are thermal detectors, the spectral response is uniform throughout
much of the infrared spectrum and usually limited by the transmission of
windows or radiation shields used to admit radiation to the vacuum-enclosed
bolometer element. Doped-germanium bolometer elements have been used
effectively by Lowl4 for infrared astronomical measurements. Metal
bolometers cooled to near their superconducting transition temperature
where very large temperature coefficients of resistance occur have also been
FIG.3. Cooled Si semiconductor balometer arrangement.
suggested. Devices of this type using thin tin films are sensitive infrared
detectors,16 but require a high degree of temperature stability compared to
cooled semiconductor bolometers.
A typical cooled Si semiconductor bolometer arrangement is shown in
Fig. 3.” The bolometer is mounted in a vacuum rather than immersed in
the cryogenic fluid bath to allow the sample temperature to change in
response to incident radiation. The bolometer is connected to wires in good
contact with the temperature bath. Room-temperature radiation is prevented
from heating the bolometer by cooled filters, usually quartz and high-density
black polyethylene. The bolometer is often mounted at the focal point of
a spherical or conical cavity to enhance the absorption of radiation. A
simple dc bias circuit is often used in combination with a low-noise pream-
plifier and a signal-averaging system if the signal is repetitive. The maximum
bias current is limited by dc ohmic heating which will raise the equilibrium
detector temperature, or by excess noise generation in the bolometer as the
bias field is increased. This excess noise occurs at relatively low applied
electric fields preliminary to breakdown effects associated with impact
ionization phenomena, which typically occur at field strengths of a few
volts per centimeter in Ge and Si.
6.3.3.Golay Cell
The Golay cell is a thermal detector with a nearly uniform energy response
from the ultraviolet to the microwave region that has been widely used in
298 DETECTORS
infrared work." The detector consists of a gas-filled cell enclosed by two

thin membranes. The incident radiation is absorbed by an aluminum layer
deposited on one membrane, with the metal film thickness chosen for a
surface impedance yielding maximum absorption of radiation virtually
independent of spectrum. The absorbed radiation heats the gas in the cell
with an attendant pressure increase which distorts the second flexible
membrane. The distortion is sensed by a photodetector cell, which detects
a light beam reflected from the membrane. The intensity of the light incident
on the photodetector depends on the deflection of the beam caused by the
membrane distortion. This effect is amplified by using two fine grids arranged
to have the image of the lines of one grid incident on the spaces of the
second grid, permitting no light transmission at equilibrium. When the
membrane is distorted, the image of the first grid shifts, allowing significantly
more light on the photodetector. This grid system is, in effect, a mechanical
means of enhancing the effect of small displacements of the membrane
(Fig. 4).
The Golay cell has a linear response for continuous power levels of
1x to 5 x lo-' W a n d can be used with caution up to the milliwatt level.
While it has an NEP of 5 x lo-" W, the limiting sensitivity is usually fixed
by the bias and amplifier electronics. In particular, a small light source
which illuminates the optical system in the cell must be energized by a
carefully isolated dc supply or batteries to attain low-noise performance.
Since the Golay cell is operated at room temperature and has a wide spectral
response, it has been used widely in many applications, particularly in
infrared spectral measurement^.'^ The spectral response depends primarily
on the input window, usually quartz, diamond, or some infrared-transmitting
crystal. For many applications, however, the cell is undesirable because of
its rather long time constant. The response time is determined by the rate
Cell Line grid
Incident
radiation
Window
Absorber Flexible mirror
FIG. 4. Golay cell and optical system to measure small displacements of the membrane.
THERMAL DETECTORS 299
of gas expansion and physical motion of the membrane, so that a minimum

time constant of about 1 msec is observed. The number of commercial
suppliers for this historically important detector has been sharply curtailed.
6.3.4. Pyroelectric Detector

The pyroelectric detector is a thermal detector which responds to a
temperature rise induced by absorbed radiation through a net change in
surface The pyroelectric effect is observed in noncentrosym-
metric crystals having a unique axis along which there is exhibited spon-
taneous electric polarization. In insulating crystals of this type, net charge
will be induced on the surfaces by the internal dipole moment until it is
neutralized by extrinsic surface charge. The extrinsic surface charge is not
mobile, so that if there is an abrupt change in the internal dipole moment
resulting from a temperature change, the induced charge will produce a net
transient electrical field for a substantial time period. This effect has been
widely used as a room-temperature method of detecting infrared radiation
by detecting the charge or voltage on a capacitor constructed using a
pyroelectric material and a pair of metallic electrodes. The simple geometry
of a pyroelectric detector is shown in Fig. 5(a); the radiation can be detected
,Electrodes,
Pyroelectric Low noise

detector amDlifier
FIG. 5 . Geometry of a pyroelectric detector and electrical equivalent, considering the

detector as a voltage source.
300 DETECTORS
using partially transparent electrodes on the faces of the pyroelectric ele-

ment, or electrodes may be attached to the edges of the sensor.
The pyroelectric effect can occur in crystals whose polarization is either
reversible or nonreversible in an applied electric field. The former class,
known as ferroelectric crystals, contains most of the important pyroelectric
materials. Application of an electric field significantly greater than the
coercive field in the direction of the polarization axis to convert a multi-
domain crystal to a single domain is called poling. The pyroelectric effect
occurs only below the Curie temperature in poled ferroelectric materials.
If the temperature is allowed to exceed the Curie point, the pyroelectric
crystal must be repoled to reestablish the single domain structure.
Pyroelectric detectors can be used to detect radiation modulated at a
constant frequency w ; to observe transient radiation when operated in
conjunction with an appropriate matching amplifier; or as an energy-
measuring device by storing the total charge generated by an input signal.
A pyroelectric detector responding to radiation modulated at a frequency
w can be considered as a parallel plate capacitor in parallel with a resistance
while the alternating charge generated by the radiation is equivalent to a
current generator in parallel with the passive components. If the pyroelectric
element is connected to an amplifier, the effective input impedance of the
amplifier can be considered the resulting combined amplifier and detector
impedances. Radiation modulated at a frequency w will generate a varying
nonequilibrium temperature T, in the pyroelectric element, giving rise to
an alternating charge PAT,,,; where p is the pyroelectric coefficient and A
is the detector area. If the detector is considered as a voltage source [Fig.
5(b)] operating into a high-input impedance voltage amplifier, it will be
characterized by a voltage responsivity R v (volts/watt) of2’
Rv = 7 (wpAR/ G ) (1 + w 2 7 ; ) 1 +O ~ T ~ ) - ” * , (6.8)
where 7 is the emissivity, R is the equivalent circuit resistance, and G is
the thermal conductance, while T~ and T are the electrical and thermal time
constants of the system, respectively. Since the T € and T are on the order
of approximately 1 sec in most cases, the high-frequency limit of this
equation is usually applicable, yielding
(6.9)
where c’ is the volume specific heat and E is the dielectric constant of the
material.
Alternately, if the detector is a current source operating into a low-input
impedance current amplifier, it will have a current responsivity R,
THERMAL DETECTORS 301
(amperes/watt) at high frequency,
R.=-77P (6.10)
' c'h'
where h is the detector thickness. The operating regime is usually in the
high-frequency region where the current responsivity is independent of
frequency, while the voltage responsivity decreases at 6 db/octave. If the
detector is used as a voltage source, the ratio of the pyroelectric coefficient
to the dielectric constant is the important parameter in achieving a high
responsivity, while as a current source, the pyroelectric coefficient alone is
the significant factor.
A good pyroelectric material is typified by a large pyroelectric coefficient,
a small dielectric constant, a low thermal conductivity and heat capacity,
and a high absorption coefficient in the spectral region of interest.22Table
I shows properties of some of the pyroelectric materials of greatest interest.
In addition to a number of inorganic insulating crystals, such as strontium
barrium niobate (SBN)23 with pyroelectric coefficients as high as 1 . 2 ~
lo-' C c K 2 K-', useful pyroelectric effects are observed in polymers such
as polyvinylidene fluoride (PVF2).24*25 From Eqs. (6.8) and (6.9), it is
apparent that the proper choice of a pyroelectric detector material depends
on the required frequency response and mode of operation. In low frequency
applications, SBN with its large pyroelectric coefficient can yield
responsivities as high as 2.5 x lo3V/ W, but its high-frequency performance
TABLEI. Properties of Pyroelectric Materials
Pyroelectric Curie Specific

Coefficient Dielectric Temperature Heat
P Constant T, C
Material ( C cm-'K-') E ( W (J gm-'K-')
Ba(Ti0,) 2 x lo-" 610 (ell) 399 0.5

4100(e,)
Triglycine Sulphate 4x 50 322 0.97
(TGS)
Li,SO,. H,O 1.ox 10 0.4
LiNbO, 4x 10-~ 30 (el,) 1573
75 ( E L )
LiTaO, 6 x lo-' 58
NaNO, 1.2 x lo-" 8 427 0.96
Strontium Barrium 1.2 x lo-' 1400 333
Niobate (SBN)
Polyvinylidene 4x 15
Fluoride (PVF,)
302 DETECTORS
as a voltage source is strongly reduced by the high dielectric constant. At

high frequencies, TGS'" and LiTa0" 27 with much smaller dielectric
coefficients are preferred in the voltage source mode. While TGS has been
widely used as a pyroelectric element, its low Curie temperature makes
frequent repolarization by application of a poling electric field a necessity,
and other more stable materials, such as lithium tantalate, have been widely
adopted for general use.
The fundamental pyroelectric detector assembly consists of the pyroelec-
tric element, a load resistor, and a field effect transistor (FET), all contained
in an evacuated housing with a transmitting window. The pyroelectric charge
developed in the thin crystal wafer is collected on metal electrodes in either
face-electrode or edge-electrode configurations, with the face electrodes
being the more common. Since the spectral response of pyroelectric detectors
is dependent on the absorption of the radiant energy by the sensing element,
the response is inherently large only when the fundamental absorption
coefficient of the pyroelectric material is large, usually in the infrared region.
For detection in the visible and near infrared, the front surface of the
element must be coated with an absorbing layer to obtain adequate response.
Because the detector elements are commonly mounted i n a hermetically
sealed or evacuated package to minimize heat transfer, the spectral response
is also restricted by the window transmission spectrum. Materials commonly
adopted for windows include quartz, silicon, germanium, and irtran. Exclus-
ive of the window properties, pyroelectric detectors with transparent elec-
trodes have flat response from 2 to 30 microns, with substantial sensitivity
throughout the far infrared.
6.3.5. Liquid-Crystal and Luminescent Detectors

Where the radiation intensity is sufficiently great ( > 5 mW/cm2), several
methods are available for direct visualization of the radiation pattern in
spectral regions where photographic film is not sensitive. Liquid crystals
are cholesteric compounds whose structure varies with temperature, often
in a small temperature range. Absorption of radiant energy by a thin
liquid-crystal layer will raise its temperature and cause a change in the
wavelength of light reflected from the crystal surface. Observation of this
change provides a means of monitoring the spatial and temporal behavior
of a radiant source, often an infrared laser. Liquid-crystal displays have
been used to monitor laser mode patterns and infrared interferograms at
wavelengths where conventional photographic plates fail to respond.28
Another widely used method of observing the spatial variation of infrared
radiation is through thermal quenching of visible l u m i n e ~ c e n c eIf. ~a~ZnCdS
phosphor is excited by an ultraviolet lamp, visible yellow fluorescence is
PHOTON DETECTORS 303
TABLE11. Thermal Detectors
Operating
Temperature R" D* T
Detector ( O K ) NEP(Wj (V/W) (cm Hz'"/W) (secj
~
Thermocouple 3 00 lo-"'
2.5~ 1 3x10' 3x
BoI ome t e r 3 00 I x 10' 2 x IOU 1x lo-'
Carbon Bolometer L Ix 10-l' 2 x 104 4.5 x io10 1x
Ge Bolometer 2 5 x lo-'' 1 x 10' 8 x 10"1.ox lo-'
Si Bolometer 2 5 x 10-I) 1 x 10' 2x
TGS Pyroelectric 3 00 X
I 1x10~ 1 x loy 1x lo-)
SBN Pyroelectric 3 00 5x10" 1x lo-'
Golay Cell 3 00 5x10-I' 2x10' 1.6~10~ 1x
induced which can in turn be quenched wherever the phosphor is locally

heated by radiant energy. Thin plastic films can be coated by a lacquer
bearing the phosphor, after which the film is mounted on an aluminum
heat sink to provide units with sensitivities of approximately 100 mW/cm*
and a response time of 0.1 sec.3"
6.3.6.Summary
The performance of a number of different thermal detectors is summarized
by the characteristics shown in Table 11. The room-temperature thermal
detectors, such as thermocouples and bolometers, are relatively slow and
exhibit moderately good detection sensitivity as measured by NEP and D*.
The cooled bolometers are significantly more sensitive, have better response
times, but are vastly more difficut to use. The pyroelectric detectors exhibit
a sensitivity similar to the other uncooled thermal detectors and are becom-
ing the standard room-temperature sensor for energy and power measure-
ments.
6.4. Photon Detectors

Optical radiation in the visible portion of the spectrum is most frequently
detected by a photon detector where the incident photons directly interact
with electrons in the detector medium to produce an electrical signal
proportional to the intensity of the radiation. There are three basic processes
which may be used to categorize the majority of practical photon detectors-
the photoemissive effect, the photovoltaic effect, and the photoconductive
effect. Since these processes directly convert radiation into electrical signals
by a quantum (rather than thermal) process, detectors of this class are
304 DETECTORS
typically characterized by high sensitivity, short response times, and strong

spectral variations in the response to radiation.
6.4.1. Photomultiplier
One of the most important direct photon effects used to detect radiation
is the photoemissive effect, also known sometimes as the external photo-
effect. Photoemissive devices are particularly well suited to detection of
low-intensity radiation and offer major advantages in detection of fast
low-level signals. In a photoemissive detector, the action' of incident radi-
ation is to induce emission of an electron from the surface of a photocathode
into space, where it is subsequently collected by an anode. There are a
number of devices in which the photoemissive effect is applied for sensing
in a simple photomultiplier or for imaging radiation in image intensifiers
or image tubes, primarily in the visible spectrum. The most commonly
employed device is the photomultiplier, where the photoemitted electrons
impinge upon a sequence of electrodes called dynodes, which act as secon-
dary electron Each electron which is incident on a dynode
induces emission of one or more additional electrons, so that an
amplification or gain mechanism exists for the production of large numbers
of electrons.
The spectral properties of the photomultiplier are primarily determined
by the photocathode. Photocathodes are generally thought to be of two
types, conventional photocathode^^^ and so-called negative electron affinity
cathodes.33734The two types of cathodes differ from each other in the
magnitude of the photoelectric work function associated with the energy
required to remove an electron from its equilibrium thermal level inside of
a solid to the energy level above the potential barrier at the solid surface
which allows the electron to escape from the solid. The minimum energy
which a photon must possess to induce photoemission equals the work
function, 4, of the metal [Fig. 6(a)]. Within the category of conventional
photocathodes, both metals and semiconductors are used as electrode
materials. The energy band diagrams associated with the photoemissive
properties in these types of materials are illustrated in Fig. 6(a) and (b),
respectively. In the conventional semiconductor electrode, the minimum
energy required for a photon to induce photoemission is still that quantity
which will raise the electron to a level higher than the potential barrier at
the surface of the material [Fig. 6(b)]. Semiconductor photocathodes with
a positive electron affinity generally require less energy than metallic photo-
cathodes and hence have spectral responses which extend to somewhat
longer wavelengths than conventional metallic photocathodes. In either
case, these photocathodes can respond to radiation of wavelengths extend-
Metal Vacuum
(a)
Semiconductor Vacuum
Photoexcitation-! \
Conduction
band
Fermi level
Valence band
FIG. 6. Energy band diagrams associated with the photoemissive properties of photo-
cathodes. ( a ) For metals, and (b) for semiconductors.
ing only from the visible spectrum into the near infrared. These material
limitations, which have been encountered over most of the history of the
use of photomultiplier tubes, have permitted only very marginal response
at wavelengths as long as 1 p. In the high-energy ultraviolet part of the
spectrum, the tubes often do not respond to wavelengths shorter than 0.2 p
as a result of the transmission spectra of typical window materials used in
their construction. There have been some extensions of the response to near
0.1 p by using rather delicate materials, such as lithium fluoride, and even
further extension of their response into the X-ray spectrum by using very
thin wafers and material such as gallium arsenide. An alternative procedure
that is used to produce some ultraviolet response in photomultiplier tubes
is to coat the entrance window of the tube with an ultraviolet-sensitive
material which is capable of fluorescing in a region where the tube does
respond sensitively. Materials such as sodium silicate have been used to
produce relatively short wavelength response in conventional tubes by using
this approach.
In recent years, many advances have been made in increasing the sensitiv-
ity in the long-wavelength portion of the spectrum and extending that
response to the wavelengths as long as 1.5 p.33-36 Prior to this work, photo-
306
Semiconductor Vacuum
Conduction band
Photoexcitation +
Fermi leve
Valence band
FIG. 7. Energy-band diagram associated with the photoemissive properties of photocathodes

with sensitivity in the infrared. (After K r ~ s e . ~ ’ )
emissive detectors were not available for this portion of the spectrum. The
extension of spectral response of photomultiplier tubes into the infrared
has resulted from the discovery of the negative electronic affinity photo-
cathode. It was discovered that by overcoating the surface of certain p-type
semiconductors with evaporated layers of low-work-function materials, the
resulting structure has a negative electronic affinity. This is illustrated in
Fig. 7. Here, in contrast to the conventional photocathodes previously
described, the incident photon must have an energy which only equals or
exceeds the energy gap, E,, of a semiconductor to allow photoemission.
Hence, by judicious choice of the semiconductor and associated coating,
it is feasible to construct photocathodes where the spectral response does
extend well into the near infrared. Even here, however, the quantum
efficiency does diminish at increasing wavelengths, so that there is yet a
relatively severe limitation on use of photomultiplier systems at wavelengths
exceeding 1.2 p.
The conventional photomultiplier tubes are usually based on a photo-
emissive structure where a thin evaporated layer composed of a compound
including some alkali metal (with the predominate example being that of
cesium), metallic elements from Group V of the periodic table, such as
antimony and other species, are combined in a ternary or similar alloy.
Examples of the spectral response of representative conventional photo-
cathode?’ are shown in Fig. 8. These show some of the conventional
materials designated by the typical nomenclature referring to the photo-
cathodes as S1, S17, and S20. For example, the S1, which is silver (Ag)
on cesium oxide (CsO),has some response to wavelengths as long as 1.1 p.
The conventional photocathode with the most sensitivity in the red portion
of the spectrum is the so-called S 2 0 , composed of cesium-sodium-
potassium-antimony [ (Cs)Na2KSb].This system is able to respond, at least
weakly, to almost 0.8 p. Some extended response versions of the S 2 0
1 1 1 1 I 1 I
0.4 0.6 0.8 1
Wavelength (rm)
FIG. 8. Spectral response of representative conventional photocathodes. (After K r ~ s e . ~ ’ )
photocathode are able to respond sensitively to wavelengths as long as 0.9 p

and are therefore useful in sensing red radiation from sources such as the
helium-neon and ruby lasers and even the gallium arsenide semiconductor
laser at 0.83 micron.
The so-called negative electron a f f i n i t ~class
~ ~ ,of
~ ~photoemissive devices
utilizes photoconductive semiconductors, such as gallium arsenide, with a
very thin surface coating that is most commonly cesium oxide. The produc-
tion of this surface layer on the semiconductor causes bending of the energy
bands of the semiconductor at its surface, as was shown in Fig. 7. The effect
of this band-bending is to produce a condition where the energy level of
the electron at the conduction-band minimum in the bulk of the semiconduc-
tor crystal is higher than the energy level of an electron in the vacuum.
Therefore, an electron in the conduction band in the bulk of the material
will energetically fall out of the photocathode semiconductor into the
vacuum if it is able to reach the activated surface region without recombin-
ing. Thus the conduction-band electrons require no energy above that of
3 08 DETECTORS
I
I 1
I I
I
1
0.4 0.6 0.8 1

Wavelength (pm)
FIG. 9. Spectral response of a typical GaAs and InAs, P,-, photocathode.
E, to escape from the photocathode, while in conventional photocathodes

substantially higher energy is required for the electrons to escape into the
vacuum [see Fig. 6(b)]. The existence of a region in the negative electron
affinity photocathode, where the energy at the bottom of the conduction
band exceeds the vacuum potential, is peculiar to this class of photocathodes
and does not occur in the conventional photoemitters. The spectral response
of a typical GaAs and an InAs,P,-, photocathode is illustrated in Fig. 9.
While the photocathode is the most crucial part of the photomultiplier,
since it converts the incident radiation to electronic current and hence
determines the spectral characteristics of the detector and the ultimate
sensitivity, the remainder of the structure is important in achieving the high
amplification and noise figure as well as the fast time response associated
with the photomultiplier. A simple schematic of a photomultiplier-type
structure is illustrated in Fig. 10. The electrons that are emitted from the
photocathode are electrostatically focused and accelerated toward a
Dynodes
II I
I
I
Focusing Photocathode
electrode (semitransparent)
FIG. 10. Schematic structure of a photomultiplier.
sequence of electrodes termed dynodes. The electrons from the photocathode

arrive at the first dynode with a kinetic energy of from 1 to 400 eV. Secondary
emission from the surface of the dynode causes a multiplication of the
initial current. Using dynode materials which are chosen primarily for high
secondary emission ratios, the secondary emission ratio can be from 5 to
10 per electrode, depending on the primary electron energy. Conventional
materials such as beryllium oxide (BeO), magnesium oxide (MgO), and
cesium antimonide (Cs,Sb) are used as dynode materials in most photo-
multiplier wavelength systems. The gallium arsenide or gallium phosphide
negative electronic affinity secondary emitters are also used in multipliers,
since they have high secondary emission ratios. Secondary emission ratios
of 5 to 50 are obtained over a primary electron energy range of from 100
to 600 eV. The process of secondary emission from a succession of dynode
surfaces causes a large multiplication of the initial current. The process is
repeated at a sufficient number of dynodes, so that the initial current emitted
from the photocathode will be multiplied by a large factor. For a secondary
emission ratio of 5 at each dynode, the multiplication between cathode and
anode is approximately lo7 for a series of 10 dynodes. The last electrode
in this chain is a collector anode without gain.
The structure of the electron multiplier chain can take on a variety of
forms, each with special characteristics as illustrated in Fig. 11. A circular
arrangement provides a structure which is both simple and compact with
relatively high-bandwidth and electron-collection effi~iency.~’ A variety of
other ladder or venetian-blind structures have been constructed which are
compact, exhibit stable gain characteristics, and allow a relatively large
number of stages, but with rather poor time response. The relationship
between response and electron-collection efficiency is an inverse one, since
the time response is largely determined by a spread in the time of flight of
310 DETECTORS
FIG. 11. Circular arrangement of an electron-multiplier chain. (After Sommer.”)
high-velocity and low-velocity electrons. The avoidance of the large spread

in time of flight implies a focusing structure where electrons with unequal
paths are either forced into compensating long and short pathlengths,
respectively, or rejected as they progress between electrodes. High collection
efficiency requires that all electrons be collected independent of their velocity
or path.
The necessity of minimizing time-of-flight variations between the photo-
emissive surface and successive electrodes has given rise to photomultiplier
designs such as the crossed-field photomultiplier. Here the magnetic field
is used together with an electrostatic field to produce varying pathlengths
depending on the electron energy, that is, high-energy electrons are forced
into long pathlengths while low-energy electrons are induced into corre-
spondingly short paths. This design reduces time-of-flight variations and
produces rise times at the anode of about 0.1 ns as compared to several
nanoseconds for various other photomultiplier designs. Another rather
different photomultiplier is the so-called channel p h o t ~ m u l t i p l i e r .In
~ ~this
device, large numbers of parallel channels, each channel being a small
diameter cylinder of material exhibiting electron gain, are used to form
continuous dynode surfaces. Voltage is applied along the channel length,
and multiple reflections of electrons traveling down the channel produce
extremely high gain. Devices of this type, where channels with density as
high as a million channels per square inch exist in parallel, are used in
imaging, and with a high channel density, they can retain a high degree of
image resolution.
Because of the high current amplification and the noise, the photomulti-
plier is one of the most sensitive instruments for detection of visible and
near infrared radiation. The photomultipliers have been used to detect
optical power levels as low as lo-'' W. As a detector of an optical signal
modulated at some low frequency, the limiting factor in sensitivity is the
so-called dark current of the photomultiplier. The random fluctuations
observed in a photomultiplier output are due to either cathode or dynode
shot noise, Johnson and 1/ f noise. The cathode shot noise is composed of
a fluctuating current emitted by the photocathode due to incident signal
power and another component observed in the absence of radiation, which
is the so-called dark current arising from random thermal excitation of
electrons. All electrons in the cathode which are thermally excited to energies
greater than the work function or electron affinity are emitted, and each
emitted electron is then effectively replaced by another electron from the
cathode. The minimum detectable optical power is determined by the
magnitude of this thermally generated dark current and typically leads to
a minimum detectable power on the order of W for simple video
detection. An obvious means of reducing the thermal dark current is to cool
the photo~athode.~'*~" If the temperature of a conventional photoemitter is
lowered to at least -2O"C, the thermal dark current becomes negligible. For
systems with lower thermal activation energy, such as the negative electron
affinity emitters, correspondingly lower temperatures are required. Cooling
of this type has traditionally been achieved by using coolants such as dry
ice to reduce the temperatures to a point where thermal carrier generation
is reduced to negligible levels. More recently, thermoelectric cooling has
been found to be more convenient and able to achieve temperatures which
are precisely those required for optimum cooling, without reducing the tube
temperature to a point where factors such as moisture and condensation
become troublesome.
6.4.2. Photoconductive Detectors

For radiation with wavelength in both the visible4' and near infrared42
spectrum, photoconductive and photovoltaic semiconductor detectors are
generally employed. A semiconductor photodetector is a simple device with
high detectivity. Incident photons excite charge carriers from bound states
into free states, increasing the number of mobile charge carriers, and these
photoexcited carriers change the electrical properties of the material. The
physical realization of this effect depends on the form and properties of
the material. In a photoconductive sample, radiation is detected through
an increase in electrical conductivity, while in a photovoltaic device it is
detected as a photon-generated voltage. Excitation of carriers in a semicon-
312 D E r ECrO RS
hu 2 Ed
\G Donor level
hv > Eg
FIG. 12. Energy-band diagram for a photoconductive detector.
ductor may take place from states near the top of the valence band across
the energy gap into states near the bottom of the conduction band, producing
excess electron-hole pairs, or alternately from impurity levels creating either
excess electrons or holes (Fig. 12). The former are designated intrinsic
photoconductors, while the latter, in which the photoexcitation is from
impurity levels, are termed extrinsic. Extrinsic detectors may incorporate
either donor or acceptor impurities, leading to sensors where the radiatively
induced carriers are either electrons or holes, respectively. Extremely long
wavelength detection may be obtained by electron intraband excitations in
materials such as InSb, yielding response at wavelengths in the 100-km to
1-mm range.
Intrinsic photoconductors respond to photons of sufficient energy to cause
an interband transition on being absorbed, thus creating excess electron-
hole pairs in the semiconductor. The semiconductor will absorb only
photons with energy greater than the energy gap, E B ,so that the maximum
wavelength A which can be detected is
(6.11)
where h is the Planck’s constant and c is the speed of light. The photodetector
can respond to all radiation with wavelengths shorter than A,,,, with
maximum sensitivity for photon energies slightly greater than the gap energy.
The photoresponse of several common infrared photoconductors is shown
in Fig. 13. The variation in spectral sensitivity for photon energies greater
FIG. 13. Spectral response of several common infrared photoconductors.
than the gap arises from the strong absorption of the radiation which
produces excitation only near the surface. The location of the maximum in
the spectral response is dependent on the thickness of the crystal and the
carrier recombination parameters.
The spectral regions in which intrinsic semiconductor photoconductors
are capable of responding depend on the energy band gaps in elemental
and compound semiconductors which have been developed. As shown in
Table 111, the range from the visible spectrum to approximately 8 pm is
well covered by a number of common semiconductors. While the points
shown are for a fixed temperature, the band gaps of most semiconductors
vary strongly with temperature, so that the maximum sensitivity for a given
material will change if the detector is cooled. For example, when InSb is
cooled from 300°K to 77"K, the energy gap increases from 0.18 eV to 0.23 eV
causing a reduction in A,, from 7 pm to 5.5 pm. Materials such as PbS
314 DETECTORS
TABLEI 1 I . Energy Band Gaps for Photoconductive Semiconductors
Band Gap Cutoff Wavelength

Material (ev) (wm)
~
Si 1.10 1.12
Ge 0.68 1.82
lnAs 0.33 3.75
InSb 0.18 6.88
PbS 0.4 3.1
PbSe 0.25 4.95
PbTe 0.31 4.0
GaAs 1.4 0.88
CdSe 1.74 0.71
CdS 2.4 0.52
Hg,-,Cd,Te( x = 0.2) 0.09 14
Pb,-,Sn .Te(x = 0.2) 0.083 15
and PbSe have a positive temperature coefficient describing the band-gap

variation and respond to longer wavelengths at lower temperatures. The
spectral variation in sensitivity of PbS and PbSe detectors as a function of
temperature is illustrated in Fig. 13. Cooling is usually required in infrared
photodetectors to reduce competition between the carriers generated by
photons and thermally generated carriers.
Beyond about 8 pm, there are no simple intrinsic elemental or compound
detectors available. The range of intrinsic detectors has been extended
beyond this point by alloying two compounds of different bandgaps to yield
materials of a desired energy gap. When a compound such as CdTe with
a band gap of 1.5 eV (300°K) is combined with HgTe, which is a semimetal
in which the valence band and conduction band effectively overlap by
0.15 eV, the resultant band gap is dependent on the fractional composition
of the compounds (Fig. 14). In the alloy Hg,-,Cd,Te, the gap varies nearly
linearly with x between the two extreme values, so that it goes through zero
at x = 0.10 at 300°K.42*43
For x = 0.2 at 77”K, the gap is approximately 0.1 eV,
yielding an intrinsic detector in the 8- to 14-pm range where there is an
important “window” in atmospheric transmission. Development efforts in
this material have centered on this region, although narrow-gap Hg, -,Cd,Te
detectors have been made with response beyond 40 pm.44
The binary, semiconductors Pb,_,Sn,Te and Pb, -,Sn,Se also have com-
position-dependent energy gaps which can be made small corresponding
to detection at wavelengths beyond 10 p.m.45.46 Development of the
Pb, - .Sn,Te alloy system has received significant emphasis in recent years
particularly as photovoltaic detectors. The limiting wavelength response of
I I I I
1.6 -
1.4 -
1.2 -
-
m
-
-
-
-
-0.4
0 0.2 0.4 0.6 0.8 1.0
HgTe CdTe
FIG. 14. Dependence of the resulting band gap on the fractional composition of the
compounds in Hg,-,Cd,Te.
a number of the elemental, compound, and alloy semiconductor photocon-

ductors is listed in Table 111. It illustrates that the intrinsic semiconductors
cover the visible and infrared spectrum to about 15 pm with a readily
available set of photodetectors.
In practical use, the PbS and PbSe detectors are widely used in the
infrared from 1 to 5 pm, with a reasonable sensitivity at room temperature,
although an increase in responsivity and decrease in noise is obtained by
operating at low temperature. The intrinsic photoconductors InSb and InAs
can be operated at room temperature, but the large number of thermally
excited carriers at room temperature limit their response, so that they are
customarily cooled. The narrow-gap alloys Hg, -,Cd,Te and Pb, -,Sn,Te,
used even further in the infrared, must always be cooled to about 77°K to
adequately reduce the number of thermally generated carriers.
As shown in Fig. 13, a number of detectors have enhanced sensitivity
and extended cutoff wavelengths when operated below 300°K. In addition,
detectors used in the wavelength range greater than 10 pm must be cooled
to prevent a large number of thermally excited carriers from competing
with the photoexcited carriers. These factors have led to the widespread
use of cooled photon detectors.
When a detector is to be cooled, it is frequently mounted in an insulated
dewar flask, illustrated in Fig. 15. The walls of the dewar can be made of
glass, Pyrex, or metal and may either be permanently sealed on manufacture
316 DETECTORS
Pump out
PO rt
Electrical leads
FIG. 15. Cooling of a detector mounted in an insulated dewar flask.
or provide for periodic evacuation. Most infrared detector systems are

usually cooled to liquid-nitrogen temperature (-77°K). The semiconductor
sensor is usually cemented to the wall of the coolant container and connected
by small-diameter wires to an external electrical connector. A window
admitting radiation to the detector is selected for a spectral band pass
matching the useful range of the detector and may be either removable or
permanently mounted to the dewar.
Several other methods are also used to cool detectors, in addition to the
common cryogenic fluids such as nitrogen. Several open- and closed-cycle
refrigeration systems have been used, including Joule-Thomson and Sterling
refrigerators. These systems are generally capable of achieving operating
temperatures of 70-80°K, with a refrigeration capacity of 10 to 20 W. Units
of this type have been used to cool detectors in remote-sensing satellite
systems where other methods were not feasible.
Finally, there has been a resurgence of interest in solid-state refrigerators
using the thermoelectric effect. Following the discovery of the thermoelectric
effect by Seebeck, Peltier discovered that when current flows in a circuit of
two dissimilar metals, heat is absorbed at one junction and released at the
other. The practical use of materials for thermoelectric devices can be
evaluated by the thermoelectric figures of merit:
(6.12)
where a is the thermoelectric coefficient (V/"K), p the resistivity ( a - c m ) ,

and k is the thermal conductivity (W/"K cm). Semiconductors generally
have the highest thermoelectric figure of merit, with the commonly used
bismuth telluride having 2 = 3 x OK-'. With a value of Z such as
described, a temperature differential of 60-70°K can be obtained with a

semiconductor detector as a heat load.
6.3.3.Extrinsic Photoconductors
The requirement for quantum detectors beyond wavelengths of approxi-
mately 10 pm frequently makes it necessary to use extrinsic semiconductors.
Impurity or defect states in a semiconductor create discrete energy levels
in the band gap of a semiconductor with relatively low activation energies.
Carriers can be photoexcited from these levels near the conduction or
valence band edges, creating a decrease in the detector’s electrical resistance
which can then be measured, as was previously shown in Fig. 12. Impurity
levels near the conduction band edge are called donor levels, since photoex-
citation of carriers will cause extra electrons to be added to the number
already in the band at thermal equilibrium. Levels adjacent to the valence
band edge are called acceptor levels, since excitation of carriers into these
levels will leave excess holes in the band. The nature of the impurity level
depends on the valence of the impurity relative to the host semiconductor.
Since these are states with very low activation energies, detectors based on
photoexcitation from these levels must be cooled to prevent thermal excita-
tion from masking the photoeffect. The technique is limited to semiconduc-
tors of a high degree of purity and crystal perfection where the level of
incidental impurities and defects is small.
In materials which d o meet these requirements, the ionization energy of
isolated impurities can be approximately treated by a description similar
to an isolated hydrogen atom. The ionization energy in a semiconductor
with a dielectric constant K and an effective carrier mass m* is given by
m*
E = 13.6 -
m K 2 eV9 (6.13)
where m is the free electron mass.

The ionization energy of an impurity state is much smaller than that of
a hydrogen atom, due to the difference in the dielectric constant and the
effective carrier mass in the solid. For example, the donor ionization energy
in GaAs with K = 13.5 and m* = 0.072 m is approximately, 0.005 eV,
equivalent to an excitation wavelength of 200 pm. This far-infrared transi-
tion is observed in pure GaAs epitaxial films and is used as the basis for a
far-infrared photoconductive detector.
In materials such as the elemental Group IV semiconductors, Ge and Si,
with complicated energy band structure, application of the simple expression
for ionization energy yields only an approximate result. Accurate calcula-
318
Conduction band
- cu Egap = 0.75 eV
Au
0 2
-
0.33
Zn
- Ga
0.02
025 0.04
-
0.03
Valence band
FIG. 16. Typical acceptor impurity levels in germanium.
tions using correct anisotropic masses predict donor ionization energies of

about 0.01 eV for germanium and 0.03 eV for silicon. The ionization energies
predicted are valid only for elements removed by the valence group of the
host element by 1. For example, in Ge, elements of Group Ill are impurities
with a single acceptor level, and elements of Group V are impurities with
a single donor level with an ionization energy of approximately 0.01 eV.
The typical acceptor impurity levels in germanium are illustrated in Fig.
16. Elements further removed from Group IV have more than a single level
with substantially different ionization energies, as shown in Fig. 16.47*48*49
Impurity ionization energies in Ge that lead to extrinsic detectors, which
are of practical importance, yield photoconductive response as shown in
Fig. 17. The small values of most donor and acceptor ionization energies
are comparable with thermal energy (0.025 eV) at room temperatures, so
that the extrinsic photoconductors must be cooled to reduce thermal ioniz-
ation.
For wavelengths shorter than the maximum, the radiation is strongly
absorbed and produces excitation only near the surface. Since it is common
that the surface imperfections are characterized by rapid nonradiative re-
combination, the excitation of photoconductivity and luminescence is less
when limited to the surface. The location of the maximum in the spectral
response is dependent on the thickness of the crystal and the recombination
parameters. For wavelengths longer than the maximum, the excitation
decreases simply because the absorption producing free carriers is
decreasing.
Solay cell
\ Ge:Au
I I 1 1 1 1 l 1 I I I I I 1 Ill I I 1 1 1 1 1 1
1.o 10 100 1000
Wavelength ( r m )
FIG. 17. Spectral response of some extrinsic photoconductors.
6.4.4. Photoconductive Response

A photoconductor is operated using a simple electrical circuit consisting
of the detector, an external bias supply, and a load resistance. The photosig-
nal generated when radiation is incident on the detector is usually measured
across the load resistance, as shown in Fig. 18.
When a photoconductor with a bias voltage V, responds to radiation by
a change in conductivity us,a signal voltage V, is produced, where
Vs=(:)V", (6.14)
with (T being the total detector conductivity. For an absorbed power per
unit volume P,/Ax, the signal photoconductivity induced by the radiation
320 DETECTORS
Photo
voltage
Bias
SUPPlY
Load resistor
Radiation
-/ 1 1 1
/-
/
/ Semiconductor /
contacts
FIG. 18. Measurement of the photosignal generated when radiation is incident on a photo-
conductive detector.
is
(6.15)
where q is the quantum efficiency, A is the wavelength, and p is the carrier

mobility. The detector is assumed to have a thickness x and an illuminated
area A. The spectral responsivity RA will then be given by the expression
RA=-qA"0' (volts/watt), (6.16)

hcnxA
where n is the carrier concentration. The response of a photoconductive
detector depends on having a material with a low thermal equilibrium
carrier concentration, small volume, high quantum efficiency, and a long
carrier lifetime. Since the excess carrier lifetime is the parameter which
determines the response time of the detector, the requirement for a long
lifetime for high responsivity conflicts with a short photoconductive response
time. Typical response times in intrinsic photoconductors range from
lO-'sec to sec, making these detectors very superior in this respect to
thermal detectors with equivalent sensitivity. The responsivity also has a
linear bias field dependence which is, of course, limited in practice by either
Joule heating or other high field effects. Operation of these detectors at low
temperatures will typically reduce the equilibrium carrier concentration and
increase the excess carrier lifetime, leading to a significant increase in R,.

The spectral detectivity 0:can be written as
(6.17)
where Af is the noise bandwidth and vT is the total noise voltage. Reducing
the detector temperature will reduce the thermal noise component, so that
the noise will be dominated by generation-recombination noise. The total
effect of cooling is a significant increase in detectivity, as was shown for
several intrinsic photoconductors in Fig. 13.
In addition to specification of responsivity and detectivity, the perform-
ance of either extrinsic or intrinsic photoconductors under steady-state
incident radiation can be described in terms of the short-circuit photocurrent
or the open-circuit photovoltage modes. The short-circuit photocurrent Z,,
is typically given by Z,, = TqNg, where 7 is the quantum efficiency, q is the
electronic charge, N is the number of photons absorbed in the sample per
unit time, and g is the gain of the photodetector. If the photoconductive
gain and photon-absorption rate are expressed in terms of fundamental
materials parameters, the short-circuit photocurrent can then be written as
(6.18)
Here P is the absorbed power at a particular wavelength, A, V, is the bias

voltage applied to the sample, 1 is the distance between electrodes on the
sample, p is the majority carrier mobility, and T is the majority carrier
lifetime. From this expression it can be seen that the low-frequency photo-
current is directly proportional to the absorbed power and also depends
directly on the wavelength for values of the wavelength below the long-wave
cutoff limit associated with each individual semiconductor. An important
observation is that the photoconductivity depends on the lifetime and
mobility of the majority carriers, since photoconductivity is in fact a major-
ity-carrier phenomenon. In a photoconductive device, minority carriers,
which typically have lifetimes significantly shorter than the majority carriers,
contribute little to the change in photocurrent and hence can be ignored to
a first approximation.
If a detector is inserted into an electronic circuit with some effective load
resistance, the photocurrent in the total circuit is modified. The photovoltage
which appears across such a load resistance can be simply expressed in
terms of the product of the short-circuit current and the equivalent load
resistance. If the load resistance, R , , is large with respect to the detector
resistance, Rdr over its range of operation, the photovoltage is essentially
322 DETECTORS
the open-circuit photovoltage, Voc.When this is the case, the photovoltage

V,, can then be written as follows
(6.19)
Here the detector resistance was expressed in terms of its majority carrier
concentration n, mobility p, and respective length, width, and thickness
1, w, and x. As in the case of the short-circuit current, the open-circuit voltage
is then dependent on the incident power, the wavelength, and bias and is
inversely proportional to the majority-carrier concentration in the device.
These expressions for short-circuit current and open-circuit voltage, are
low-frequency or steady-state descriptions, and do not express information
about the response time of the device which is crucial in many applications.
Since a photoconductor is, in fact, a majority-carrier device, the frequency
response of the open-circuit voltage V, is then given by the expression
where w is the angular frequenccy of the modulation of wavelength of

radiation of wavelength A. At low frequencies where the product W T is much
less than 1, the device response is essentially independent of frequency,
while at high frequencies where OT becomes much larger than 1, the response
is inversely proportional to frequency and depends explicitly on the major-
ity-carrier lifetime, which is thus the primary limit on high-frequency perfor-
mance of photoconductive devices.
6.4.5. Photovoltaic Devices

A mode of detection of great practical significance is that using the
photovoltaic eff e ~ tWhile ~ effect relies on direct conversion of photons
. ~ this
to carriers in the material, as does photoconductivity, it places additional
requirements on the nature of the detector in that some form of potential
barrier with an internal electric field is required to cause a separation of
the photoexcited electron-hole pair. Photovoltaic detectors are constructed
from materials where the intrinsic photovoltaic effect arising from band-to-
band transitions generates electron-hole pairs which are subsequently separ-
ated by an internal field and separately collected at opposite electrodes of
the device. The photovoltaic effect is essentially limited to materials where
intrinsic photodetection occurs, although there have been a few limited
reports of extrinsic photovoltaic detectors. The simplest and classical struc-
ture for separation of the photoexcited electron-hole pairs occurs in the
simple p-n junction; however, a number of other structures have recently

been vigorously investigated.
By far the most common form of photovoltaic effect is that which occurs
in a p-n junction fabricated in a common semiconductor, such as silicon,
by doping or formation of epitaxial layers at one surface. A simple p-n
junction photodiode is illustrated in Figure 19, where two electrical contacts
are formed and a thin region at the surface has been prepared to allow the
incident radiation to be absorbed and to generate electron-hole pairs. The
photoexcited electrons and holes are separated by the high field in the space
Incident
radiation
Top surface Top contact
P-n junction n-type
I 1
Contact
N-region P-region
Conduction band
Valence band
FIG. 19. (a) A simple p-n junction photodiode. (b) Associated energy-band diagram.
324 DETECTORS
Current
Saturation
current
Dark \
current Ill Voltage
Photovoltage (open circuit)
Photocurrent
Photocurrent (short circuit)
Bias -
IT-$ Radiation
Photodiode
+~
source+ '6
I+ RL I Load
resistor Photovoltage
FIG. 20. (a) Observation of photovoltage and photocurrent in a photovoltaic device. (b)
Circuit to observe the photovoltage.
charge region between the p and n regions. This charge separation creates
an open-circuit voltage across the junction and leads to a voltage which
can be detected that is proportional to the absorbed photon density. The
direct observation of this photovoltage is the simplest and most straightfor-
ward mode of operation of a photovoltaic detector [Fig. 20(a)]. Junction
detectors are often operated as well under some form of reverse bias voltage
which leads to observation of a photocurrent [Fig. 20(a)] rather than a
photovoltage and are said to be operated in a photoconductive mode, despite
the fact that it is not equivalent to the photoconductive mode previously
described.
The short-circuit current associated with the photovoltaic detector is
identical in form to that for a photoconductive detector, with the exception
that the gain is unity, that is, the short-circuit current is given by
nqPA
I,=-, (6.20)
hc
The open-circuit voltage, which is the commonly observed photovoltaic
signal, is obtained as the product of the short-circuit current with the
dynamic resistance of the junction in the absence of bias. Hence the
open-circuit photovoltage V,, for a photovoltaic detector is approximately
(6.21)
Here I,,, is the saturation current of the junction in the absence of a photo
signal, k is Boltzmann’s constant, and T is the absolute temperature. Just
as in the case of the photoconductive detector, the photovoltaic signal
depends on the incident power and wavelength, but it does not depend on
the majority carrier lifetime. The photovoltaic effect, in fact, is essentially
independent of the majority-carrier lifetime, since in creating an electron-
hole pair, the determining factor is the minority-carrier lifetime, which is
significantly shorter. Since the frequency response is determined by the
product of the modulation frequency of the radiation w and the minority-
carrier lifetime 7, the response is essentially independent of frequency for
values of W T < 1. The point at which W T becomes unity is at a significantly
higher frequency for a photovoltaic device; therefore, photovoltaic detectors
generally display a higher frequency response or a faster time response than
photoconductive detectors of exactly the same material.
The most widely used photovoltaic detectors for the visible spectrum rely
on the common elemental semiconductors Ge and Si with cutoff wavelengths
of 1.5 pm and 1.1 pm, r e s p e c t i ~ e l yFor
. ~ ~applications at longer wavelengths,
photovoltaic structures have been developed using 111-V compounds or the
alloy systems Pb, -,Sn,Te and Hg, -,Cd,Te. For optical communications
systems, detectors utilizing Ge”,52 or 111-V alloys, such as
InxGal-xAsyP1-y,53-55 have been emphasized with peak response in the
1.1- 1.7-pm range. For longer-wavelength infrared extending to wavelengths
as long as 14 pm, the Hg,-,Cd,Te and Pb,-,Sn,Te alloy systems have been
vigorously de~eloped.’~-’~
While the discussion of photovoltaic detection has centered on the
classical p-n junction, there are a number of structures which have been
employed for photovoltaic detection. The most direct extension of the simple
p-n junction is the p-i-n photodiode which incorporates an intrinsic region
between the p and n portions of the traditional junction. If either the p or
n side of the junction is made very thin compared to the optical absorption
length in the medium, the incident optical radiation will readily penetrate
into the intrinsic region. In the intrinsic region, the absorption will produce
a large number of electron-hole pairs. As these electron-hole pairs are
produced in a region of high electric field, the carriers will rapidly be drifted
out of that region into the adjacent p and n sides of the junction. This rapid
carrier separation leads to an improved frequency response and efficiency
in the p-i-n diode as compared to that of a pn diode of a comparable
material.
In addition to being used to enhance detection, p-i-n configurations have
also been employed in so-called avalanche photodiodes.""."' An avalanche
photodiode is a structure with an internal gain achieved through the so-called
avalanche breakdown effect, which occurs in the high field regions of a p-n
or p-i-n junction under high reverse bias. The carriers generated by photo-
excitation are accelerated to velocities which are sufficiently high so that
upon collision with atoms in the lattice, additional free electrons are gener-
ated by the impact ionization process. These free electrons are also acceler-
ated and undergo more collisions and hence will also lead to the generation
of more free electrons. The so-called avalanche effect, which occurs within
a high field region in a reverse bias p-n or p-i-n junction, can lead to a
significant internal gain."2 Hence, a larger photosignal is produced in a
material with the same incident radiation power than would be produced
under conventional diode operation. While the photogenerated current for
a p-i-n diode is as previously described by Eq. (6.20), for conventional
photovoltaic devices the value of current, I,", in an avalanche diode is
nqPA
I,, = -M , (6.22)
hc
where M is the avalanche gain. The gain depends on device design and
bias conditions and may be as high as 100. The avalanche process is a
statistical one, where M fluctuates for each chain of events. Hence the
amplification process in these devices is subject to an additional noise-
generation process.
In many semiconductors, particularly those of significance for infrared
detection, production of both n- and p-type materials of high quality is not
always possible within the current state of the art. A method for forming
photovoltaic structures in materials where p-n junctions cannot be formed
is through creation of a Schottky barrier p h ~ t o d i o d e . " ' .A
~ ~Schottky barrier
is a junction formed at the interface of a metal semiconductor structure. In
a manner analogous to that described for the p-n junction, a potential
barrier is created at the metal-semiconductor interface, which will effectively
separate electron-hole pairs generated by incident optical radiation. A
diagram of a typical Schottky barrier junction is shown in Fig. 21. In most
EF -
Metal
Semiconductor
FIG. 21. Diagram of a typical Schottky barrier function.
cases, the Schottky barrier is formed by producing a thin film of metal on

the surface of the semiconductor, where the thickness of the metal is
sufficiently small so that it is partially transparent to incident radiation at
the wavelengths of interest. Incident photons can generate electron-hole
pairs within the semiconductor or in the vicinity of the potential barrier at
the metal-semiconductor interface. Depending on the specific structure, the
Schottky barrier photodiode may be illuminated either from the metal side
of the junction or it may, in fact, be illuminated through the semiconductor
if it is made sufficiently thin so that carriers can reach the interfacial region.
6.4.6. Photon-Drag Detector

While most semiconductor quantum detectors respond to radiation by a
photo-induced change in free carrier concentration, the photon-drag detec-
tor operates through a change in carrier momentum as photon energy is
absorbed.65 The simplest description of photon drag attributes the
phenomenon to radiation pressure which transfers momentum to the mobile
carriers in the semiconductor. As the radiation propagates through the
semiconductor, it acts upon mobile carriers, affecting carrier transport in
the direction of light propagation. A voltage called the “photon-drag vol-
tage” is generated between a contact at the face of the sample where the
radiation enters and another contact where it leaves the sample. The photon-
drag effect has been exploited for use as a high-speed room-temperature
infrared detector, with wide use for high-power CO, lasers at 10.6 pm.
While the simple description of the effect is accurate for C 0 2lasers operating
328 DETECTORS
around 10.6 pm with the photon-drag detector at room temperature,

observations of the frequency and temperature dependence indicate that
for wider use, this description is inadequate. Gibson ef al.66.67have shown
that in germanium, a more accurate representation must take into account
the differences in energies associated with interband transitions of electrons
originating in quantum states of the same energy but different wavevectors
when the transitions also involve momentum transfer from the photon filed
to the electrons. Based on this model using an accurate band structure for
p-type germanium, the frequency and temperature dependence of the effect
can be explained.
While photon drag has been observed in compounds such as InSb and
GaAs. Ge has been found to be the most efficient detector material and is
used in commercially produced detectors. The responsivities in typical Ge
detectors are on the order of 1 x lop6V/W. While having a very low
responsivity, the photon-drag detector does have a linear response at
extremely high incident intensities and a very high frequency response.
Most detectors are linear up to incident power density of greater than
10 MW/cm2 with response times of less than 1 nsec. It has been proposed
that the response time is limited only by the transit time of photons through
the detector.
6.5. Noise
A fundamental limit exists in the minimum radiation power that can be

detected by all forms of detectors, arising from the noise that is present in
addition to the Noise may arise from within the detector, may be
associated with the source of radiation, or may appear in the electronic
system used to enhance the electrical output signal of the detector. The
latter element can be largely eliminated through attention to the proper
electronic design and may, in fact, include certain low-noise amplifiers that
can reduce electronic contribution to the noise to a very small level. The
noise which actually arises due to mechanisms internal to the detector
generally falls into several distinct categories.
Many sources of noise are common to or appear in most forms of radiation
detectors. In particular, almost all forms of noise are present, to at least
some degree, in the solid state detectors used to detect most forms of visible
and infrared radiation. In the absence of electrical bias, there is a low-level
source of noise usually described as Johnson noise. This type of noise is
generated by the random motion of charge carriers in any resistive medium.
The Johnson noise is present at a power level which is determined by the
temperature of the material and the bandwidth of the detection system. In
addition to these factors which determine the noise power for Johnson
NOISE 329
noise, the noise voltage and current also depend on the resistance value of
the detector. The mean square Johnson noise power, which then appears
even in the absence of electrical bias as a fluctuating current or voltage, is
given by P, = kTB, where k is the Boltzmann's constant, T is the absolute
temperature of the detector, and B is the measurement of bandwidth. The
open-circuit noise voltage V, is then given by
V, = (4kTRB)'/2, (6.23)
where R is now the resistance of the detector. In most solid state photodetec-
tors, the Johnson noise is present at all frequencies, but it is not the dominant
noise, except at high frequencies. At lower frequencies, several other sources
of noise dominate. This is illustrated in Figure 22, which illustrates an ideal
noise spectrum for a typical solid-state photodetector in the absence of
radiation. At low frequencies, another noise source, called l/f noise, domi-
nates, while at intermediate frequencies generation-recombination noise
dominates. The specific ranges over which each of these noise sources is
predominant varies according to the details of the composition and structure
of the detector, but typically below 1 kHz is the range for l/f; which is the
principal noise source. Between 1 kHz and as high as 1 MHz, generation-
recombination noise would be the dominant noise, while above that
frequency, the Johnson noise would be the predominant noise source.
At the lowest frequencies, the source of noise found in most solid-state
detectors is l/f noise characterized by a spectrum where the noise power
I I I I 1 1 I I 1 I
- -
-
Generation-recombination noise
-
Johnson
-
noise
I I 1 I I I I I I I
2 4 6 8 10
FIG. 22. Different types of noise of a solid state photodetector.

330 DETECTORS
varies approximately inversely with frequency. Most solid-state detectors

usually exhibit l/f noise at low frequencies. Although at higher frequencies,
the amplitude drops below that one or below the other types of frequency-
independent noise. The mechanism for l/f noise is not well understood. It
has been attributed to the presence of potential barriers at contacts or at
the surface of the detector. The understanding and reduction of l/f noise
is then a process which depends on the procedures used in preparation of
the contacts and surfaces of the detector and is very much an art. The noise
current associated with l/f noise is often expressed as
I,= ( c1;- , (6.24)
where C is a constant, Ib is the bias current, B is the bandwidth, and f is

the frequency. There are often exponents associated with the frequency
given here by a, where a is approximately, although not always, precisely
equal to unity.
A third and very important form of noise found in most solid-state
detectors is the so-called generation-recombination noise. Generation-
recombination noise is caused by fluctuation in the rate of thermal generation
and recombination of charge carriers in a solid, giving rise to a fluctuation
in the average charge-carrier density. Because of this variation in carrier
density, the resistivity will also fluctuate, and this will, of course, give rise
to a fluctuating voltage at the terminals of the detector when a fixed bias
current is flowing through the solid. In many solid-state detectors, the
fluctuations in charge-carrier concentration will arise either from intrinsic
carrier generation, that is, the creation of free electron-hole pairs by the
thermal energy of the lattice, or by extrinsic carrier generation where
fluctuation in the number of carriers can arise from the generation and
recombination rates associated with an impurity level. For a simple intrinsic
photoconductor with a single type of thermally generated carrier, usually
electrons, the noise current will be given as
I =(%)([*]””
+ 1 w2r2 (6.25)
where p is the total number of free holes in the sample, r is the free-carrier
lifetime, B is the measurement bandwidth, and n is the total number of
electrons in the material. It is clear that when the frequency becomes large
with respect to the free-carrier lifetime, the noise power would diminish,
which will limit the upper frequency limit over which generation and
recombination noise is an important factor.
OPTICAL WINDOW MATERlAL 33 1
A final form of noise which is found in photo-emissive devices is that of

shot noise. The shot noise in a photoemissive detector is equivalent to the
generation-recombination noise in a semiconductor in that it arises from
randomness in the emission of electrons. This noise is generated even at
low temperature where thermal carrier generation can be neglected. The
shot noise in a photomultiplier includes both the cathode shot noise due
to random current fluctuations in the average current emitted by the photo-
cathode due to the incident signal power, and the dynode shot noise due
to random secondary emission at the dynodes. Since the current originating
at a dynode does not see the full gain contribution of the tube, the dynode
shot noise is a small fraction of the cathode shot noise.
6.6. Optical Window Material
One of the limitations in the spectral response in many detector systems

is associated with the transmission spectrum of optical window material
used to protect the detector element. There are approximately one hundred
optical materials with a useful transmission band for application as windows.
A variety of physical properties can be used to select a fraction of these as
technologically significant. The most significant physical properties in win-
dow selections are spectral transmittance and the index of refraction of the
material. In addition, a number of mechanical and thermal properties are
often of importance, including: hardness, thermal conductivity and
TABLEIV. Optical Window Materials
Cutoff Wavelengths (pm)* Index of

Material Lower Upper Refraction
Borosilicate Glass 0.3 2.8 1.49

Fused Silica 0.18 4.3 1.44
Sapphire 0.15 6.5 1.73
Magnesium Fluoride (IRTRAN-1) 0.45 9.5 1.38
Magnesium Oxide (IRTRAN-5) 0.40 9.5 1.68
Calcium Fluoride (IRTRAN-3) 0.15 12. 1.40
Zinc Sulfide ( I RTRAN-2) 0.5 14.5 2.23
Sodium Chloride 0.2 2.0 1.51
Zinc Selenide ( I RTRAN-4) 0.5 22. 2.41
Cadmium Telluride (IRTRAN-6) 0.9 30. 2.65
Silicon*t 1.2 55. 3.46
Germaniumt 1.8 55. 4.01
KRS-5 0.5 55. 2.35
* Transmission of 10% through 2-mm specimen thickness.

t Lattice bands cause discontinuous transmission spectrum.
332 DETECTORS
expansion coefficients, specific heat, elastic moduli, and variations of all of

these with temperature.
There are about 10 to 12 materials that experienced optical-system design-
ers incorporate in most practical systems. These are described in Table IV,
which indicates the useful transmission range of each material and its index
at the center of this range. Those materials useful as windows in the near
ultraviolet and visible part of the spectrum include borosilicate glass, fused
silica, sapphire, and calcium fluoride. In the infrared region, there are a
number of elemental and compound semiconductors with useful passbands
including Ge, Si, CdTe, ZnSe, ZnS, and MgO. As can be seen from Table
IV, most of these semiconductors have a high index of refraction so that
antireflection coatings are necessary for efficient transmission. A number
of the latter group are available commercially prepared by a powder hot
pressing process under the trade name IRTRAN.
References
1. R. C. Jones, “Noise in Radiation Detectors,” Proc. Inst. Rad. Eng. 47, 1495 (1959).
2. R. A. Smith, F. E. Jones, and R. P. Chasmar, The Defection and Measurement of Infrared
Radiation, Oxford Univ. Press, London, 1968.
3. E. S. Barr, “The Infrared Pioneers, JI; Macedonio Melloni,” Infrared Phys. 2, 67 (1962).
4. A. H. Wilson, The Theory of Metals, Cambridge Univ. Press, London, 1954. Also H. P. R.
Frederikse, V. A. Johnson, and W. W. Scanlon, in Methods of Experimental Physics, Vol.
6B, Chap. 7 (L. Marton, ed.), Academic Press, New York, 1959.
5. G. W. Day, 0. L. Caddy, and R. J. Iversen, “Detection of Fast Infrared Laser Pulses with
Thin Film Thermocouples,” Appl. PHys. Lett 13, 289 (1968).
6. D. A. Jennings and E. D. West, “A Laser Power Meter for Large Beams,” Rev. Sci. Insfr.
41 565 (1970).
7. J. W. Stearn, “A Radiometer for Continuous Wave Laser Radiation,” J. Sci. Instr. 44, 218
(1967).
8. R. De Waard and E. M. Wormser, “Description and Properties of Various Thermal
Detectors,” Proc. Insf. Rad. Eng. 47, 1508 (1959).
9. G. Koren, Y. Yacoby, H. Lotem, M. Kosower, and G. Greenwald, “Thin Film Calorimeter
for Low-Energy Laser Pulse Measurements,” Appl. Phys. Left. 23, 73 (1973).
10. A. J. Schmidt and R. C. Greenhow, “A Practical Laser Output Meter,” J. Phys. E 2, 438
(1969).
11. B. Contreras and 0. L. Caddy, “Nanosecond Response Time Room-Temperature Infrared
Detection with Thin Film Bolometers,” Appl. Phys. Left. 17, 450 (1970).
12. J. Stricker and J. Rom, “A Fast Response Thin Film Wattmeter for Measurements of CO,
Laser Power,” Rev. Sci. Instr. 43, 1168 (1972).
13. W. S. Boyle and K. F. Rodgers, “Performance Characteristics of a New Low-Temperature
Bolometer,” J. Opr. SOC.Am. 49, 66 (1959).
14. F. J. Low, “Low Temperature Germanium Bolometer,” J. Opt. SOC.Am. 51, 1300 (1961).
15. R. Bachmann, H. C. Kirsch, and T. H. Geballe, “Low Temperature Silicon Termometer
and Bolometer,” Rev. Sci. Insrr. 41, 549 (1970).
REFERENCES 333
16. D. Bloor, T. J. Dean, J. 0.Jones, D. H. Martin, P. A. Mauer, and C. R. Perry, “Spectroscopy

of Extreme Infrared Wavelengths,” Proc. Roy. Sac. A (GB), 260, 510 (1961).
17. M. A. Kinch, “Compensated Silicon-Impurities Conduction Bolometer,” J. Appl. Phys.
42, 5861 (1971).
18. M. J. E. Golay, “A Pneumatic Infrared Detector,” Rev. Sci. Insrr. 18, 357 (1947).
19. J. R. Hickey and D. B. Daniels, “Modified Optical System for the Golay Detector,” Rev.
Sci. Insrr. 40 732 (1969).
20. E. H. Putley, in Semiconductors and Sernimerals, Vol. 12, p. 441, (R. K. Willardson and
A. C. Beer, eds.), Academic Press, New York, 1977.
21. E. H. Putley, in Semiconductors and Semimerals, Vol. 5, p. 259 (R. K. Willardson and A.
C. Beer, eds.), Academic Press, New York, 1970.
22. S. T. Liu and D. Long, “Pyroelectric Detector and Materials,” Proc. IEEE 66, 14 (1978).
23. A. M. Glass, “Investigation of the Electrical Properties of Sr,-,Ba,Nb20, with Special
References to Pyroelectric Detection,” J. Appl. Phys. 40, 4699 (1969).
24. A. M. Glass, J. H. McFee, and J. G. Bergman, Jr., “Pyroelectric Properties of Polyvinylidene
Fluoride and its Use for Infrared Detection,” J. Appl. Phys. 42, 5219 (1971).
25. M. G. Broadhurst, W. P. Harris, F. 1. Mopsik, and C. C. Malinberg, Polymer Preprints,
14, 820 (1973).
26. A. Hadni, Y. Henninger, R. Thomas, P. Vergnat, and B. Wynche, J. Phys. 26, 345 (1965).
27. A. M. Glass, “Dielectric Thermal and Pyroelectric Properties of Ferroelectric LiTaO,,”
Phys. Rev. 172, 564 (1968).
28. F. Keilmann, “Infrared Interferometry with a C 0 2 Laser Source and Liquid Crystal
Detection,” Appl. Opr. 9, 1319 (1970).
29. G. A. Condas and D. W. Brown, “Variable Sensitivity Vision Displayer for Infrared Laser
Beams,” Rev. Sci. Instr. 41, 888 (1970).
30. T. J. Bridges and E. G. Burkhardt, “Observation of the Output of a C 0 2 Laser by
High-Resolution Thermographic Screen,” IEEE J. Quanr. Elecrron. QE-3, 168 (1967).
31. A. H. Sommer, Photoemissive Materials, p. 222, Wiley, New York, 1968.
32. RCA Phoromulriplier Manual, RCA, Electronic Components, p. 3, Hamson, New Jersey,
1970.
33. R. U. Martinelli and D. G. Fisher,. Proc. IEEE, 62, 1339 (1974).
34. R. L. Bell, Negarive Electron Aflniry Devices, Clarendon Press, Oxford, 1973.
35. D. G. Fisher, R. E. Enstrom, J. S. Escher, and B. F. Williams, J. Appl. Phys. 43, 3815 (1972).
36. A. H. Sommer, R C A Rev. 34,95 (1973). Also published in A. H. Sommer, Gallium Arsenide
and Related Compounds, p. 143, in 1972 Proceedings of rhe Fourth Inrernarional Conference,
Con$ Ser. No. 17, Institute of Physics, London, 1973.
37. P. KNse in Topics in Applied Physics, Vol. 19, p. 5, Oprical and Infrared Detecrors (R. J.
Keyes, ed.), Springer-Verlag, New York, 1980.
38. G. W. Goodrich and W. C. Wiley, Rev. Sci. Insrr. 33, 761 (1962).
39. A. H. Sommer, Photoemissive Materials, p. 68, p. 229, Wiley, New York, 1968.
40. R. B. Murray and J. J. Manning, IRE Trans. Nuc. Sci. NS-7, 80 (1960).
41. D. H. Seib and L. W. Aukerman, Advances in Electronics and Elecrron Physics 34, 95-221,
(L. Marton ed.) Academic Press, New York, 1973.
42. D. Long, in Topics in Applied Physics, Vol. 19, p. 101, Optical and Infrared Derecrors (R.
J. Keyes, ed.), Springer-Verlag, New York, 1980.
43. D. Long and J. L. Schmit, Semiconductors and Semimetals 5, pp. 175-255, (R. K. Willardson
and A. C. Beer, eds.) Academic Press, New York, 1970.
44. D. Long, “Properties of Semiconductors Useful for Sensors,” IEEE Trans. Elecrron Devices
ED-16, 836 (1969).
334 DETECTORS
45. T. C. Harman and 1. Melngailis, Applied Solid State Science, p. 1, 4th ed. (R. Wolfe, ed.),
Academic Press, New York, 1974. Also: I. Melngailis, J. Luminescence, 7, 505 (1973).
46. 1. Melngailis and T. C. Harman, in Semiconductors and Semimetals, p. 11 1, 5th ed. (R. K.
Willardson and A. C. Beer, eds.), Academic Press, New York, 1970.
47. R. B. Emmons, “The Frequency Response of Extrinsic Photoconductors with Reduced
Background,” Infrared Phys. 10, 63 (1970).
48. P. R. Bratt, in Semiconductors and Semimetals, Vol. 12, p. 39 (R. K. Willardson and A. C.
Beer, eds.), Academic Press, New York, 1977.
49. H. Levinstein and J. Mudar, “Infrared Detectors in Remote Sensing,’’ Proc. IEEE 63, 6
(1975).
50. R. G. Smith, “Photodetectors for Fiber Transmission Systems,” Proc. IEEE 68,1247 (1980).
51. J. Conradi, “Planar Germanium Photodiodes,” Appl. Opt. 14, 1948 (1945).
52. H. Melchior and W. T. Lynch, “Signal and Noise Response of the High Speed Germanium
Avalanche Photodiodes,” IEEE Trans. Electron Devices, ED-13, 829 (1966).
53. M. Feng, J. D. Oberstar, T. H. Windhorn, L. W. Cook, G. E. Stillman, and G. E. Streetman,
“Be-Implanted 1.3 pm InGaAsP Avalanche Photodetectors,” Appl. Phys. Lett. 34, 591
(1979).
54. T. P. Lee, C. A. Burrus, Jr., and A. G. Dentai, “lnGaAsP/InP Photodiodes: Microplasma-
Limited Avalanche Multiplication at 1-1.3 k m Wavelength,” IEEE J. Quantum Electron,
QE-15, 30 (1979).
55. M. A. Washington, R. E. Nahory, M. A. Pollack, and E. D. Beebe, “High Efficiency
In,~.,Ga,As,.P,_,/lnP Photodetectors with Selective Wavelength Response Between 0.9
and 1.7 km,” Appl. Phys. Lett. 33, 854 (1978).
56. W. H. Rolls and D. V. Eddolls, “High Detectivity Pb,Sn,-,Te Photovoltaic Diodes,”
Infrared Phys. 13, 143 (1973).
57. C. C. Wang and S. R. Hampton, “Lead Teluride-Lead Tin Teluride Heterojunction Diode
Array,” Solid State Electron. 18, 121 (1975).
58. A. M. Andrews, J. T. Longo, J. E. Clarke, and E. R. Gertner, “Backside-Illuminated
Pb, - ,Sn,Te Heterojunction Photodiode,” Appl. Phys. Lett. 26, 439 (1975).
59. D. Eger, M. Oron, A. Zussman, and Z. Zemel, “The Spectral Response of PbTe/ Pb, -,Sn,Te
Heterostructure Diodes at Low Temperatures,” Infrared Phys. 23, 69 (1983).
60. G. E. Stillman and C. M. Wolfe, in Semiconductors and Semimetals, Vol. 12 (R. K.
Willardson and A. C. Beer, eds.), Academic Press, New York, 1977.
61. P. P. Webb, R. J. Mclntyre, and J. Conradi, “Properties of Avalanche Photodiodes,” R C A
Reu. 35, 234 (1974).
62. A. G. Chynoweth, in Semiconductors and Semimetals, Vol. 4 (R. K. Willardson and A. C.
Beer, eds.), Academic Press, New York, 1983.
63. S. M. Sze, Physics ofSemiconductor Deoices, Wiley, New York, 1983.
64. E. M. Logothetis, H. Holloway, A. J. Varga, and E. Wilkes, “Infrared Detection by Schottky
Barriers in Epitaxial PbTe,” Appl. Phys. Left. 19, 318 (1971).
65. A. F. Gibson, M. F. Kimmitt, and A. C. Walker, “Photon Drag in Germanium,” Appl.
Phys. Lett 17, 75 (1970).
66. A. F. Gibson, C. A. Rosito, C. A. Raffo, and M. F. Kimmit, “An Optical Bridge for the
Assessment of Modern Locked C 0 2 Laser,” J. Phys. D 5, 1800 (1972).
67. P. J. Bishop and A. F. Gibson, “Absorption Coefficient of Germanium at 10.6 pm,” Appl.
Opt. 12, 2549 (1973).
68. W. B. Davenport and W. L. Root, A n Introduction to the Theory of Random Signals and
Noise, McGraw-Hill, New York, 1958.
A Brewster angle, 116, 121, 122
Abbe theory of the microscope, 86 brightness temperature, 241
aberration correction by holography, 197 Burch interferometer, 20
aberrations, 83
in holography, 172
Achromatic
devices 1 and 2, 133 C
wave plates, 144 channeled spectrum, 28
acoustical holography, 197 character recognition, 197
acousto-optic effect, 150 chromatic dispersion in gratings, 72
Airy disc, 67, 80 circular
angle measurements, 41 aperture, diffracting, 58, 66
angular resolution of telescopes, 81 polarization, 110
anisotropic media, 133 polarizers, 124
antireflection coatings, 33 coded reference beam for color holograms,
astronomical radiometers, 249 182
atmospheric sounding techniques, 236 coherence length, 6, 41
color holograms, 180
common-path interferometers, 19
compensated interferometer, 11
B compensating
plate, 11
Babinet
retardation plate, 143
compensator, 143
compensators, 143
principle, 68
complex plane representation of
band-pass filters, 33
polarization, 118
beam expansion in holography, 200
composite holography, 184
beam splitters, 33
computer hologram, 189
Bennett-Koehler reflectometer, 274
conjugate image in holography, 170
Bessel function, first order, 66
Cornu spiral, 56
binary hologram, 189
Cotton-Mouton effect, 153
birefringence
crystal
in crystals, 133
polarizers, 138
induced, 146
retardation plates, 142
blackbody
types, 135
curves, 223
cubic crystals, 135
radiation, 219
blaze angle in gratings, 78
blazed gratings, 78, 158
bleaching of holograms, 204
0
bolometers, 295
Bragg depolarizers, 153
effect in holograms, 172 detection, synchronous, 279
regime, 151 detector polarization, 158
335
336 INDE X
detector electromagnetic description of light, 108

Golay cell, 298 elliptical polarization, 113
photon-drag, 327 embossing of holograms, 206
pyroelectric, 299 energy
quantum, 303 distribution in gratings, 75
detectors, 291 theorem, 70
figures of merit, 292 equal-thickness fringes, 8
liquid crystal, 302 etalon, 41
luminescent, 302 experimental procedures in holography, 199
noise, 328 extraordinary beam, 136
photoconductive, 3 I 1 extrinsic photoconductors, 3 17
photon, 303
thermal, 293, 303
dichroic F
mirrors, 33 Fabry-Perot
polarizers, 156 etalon, 31
dichroism, 33 interferometer, 25
dichromated gelatin for holography, 202 Faraday effect, 153
dielectric tensor and energy distribution, field
133 equations, 108
diffracting measurements, 266
circular aperture, 58, 66 figures of merit for detectors, 292
rectangular aperture, 65 filter radiometers, 257
slit, 57, 65 finesse of interference fringes, 27
straight edge, 57 first order plate, 143
two-dimensional array of holes, 80 Fizeau interferometer, 8
diffraction, 49 four mirror polarizer, 124
images with aberrations, 83 Fourier transform, 62, 69
dipole radiation pattern, 98 holograms, 178
directional reflectance, measurement of, 274 Fraunhofer
division diffraction, 62
of amplitude, 4 hologram, 175
of wave front, 1 Fresnel
Doppler biprism, 3
effect, 42 diffraction, 53
interferometry, 42 equations, I13
double exposure holographic hologram, 175
interferometry, 193 integrals, 56
double-pass interferometer, 37 rhomb, 133
zone plate, 59
fringes
E
of equal chromatic order, 28
echelles, 79 of equal inclination, 5
echellettes, 79 of equal thickness, 4
echelon, 31
edge waves, 52
Edser-Butler fringes, 28 G
electro-optic Gabor plate, 61
effect, 147 ghost lines in gratings, 75
modulators, 147 Clan-Foucault prism, 141
INDEX 337
Glan-Taylor prism, 141 Huygens wavelet, 49

Clan-Thompson prism, 139 Huygens- Fresnel
Golay cell, 297 principle, 49, 54
grating theory, 5 1
angular distribution of light, 70
disperser radiometer, 261
polarizers, 157 I
gratings image-plane hologram, 179
blazed, 158 imagery through diffusing media, 196
chromatic dispersion, 72 imaging radiometers, 252
resolving power, 72 impermeability tensor, 146
gyroscope, optical, 43 inclination factor, 52, 56
induced birefringence, 146
H integrating spheres, 280
interference, 1
Haidinger fringes, 11 filters, 31
half-tone hologram, 190 fringes, 1
Helmholtz-Kirchhoff theorem, 50 interferometric radiometer, 261
Herpin matrix, 129 interferometry applications, 38
hertzian polarization, 157 irradiance interferometer, 24
hologram embossing, 206 isotropic
hologram, Fourier transform, 178 crystals, 135
Fraunhofer, 175 media, 113
Fresnel, 175
Lippman-Denisyuk, 185
binary, 189 J
color, 180 Jamin interferometer, 12
computer, 189 Jones’ matrices, 159
half tone, 190
image plane, 179 K
phase, 190
Kerr effect, 147, 153
rainbow cylindrical, 189
kinoforms, 190
volume reflection, 174
Kirchhoffs diffraction theory, 50, 5 1
holograms, effects of thickness in, 172
knife-edge test, 39
rainbow, 184
reflection, 172
transmission, 172 L
volume, 172 lambertian sphere, 153
holographic laser power, measurement of, 277
bleaching, 204 lateral shearing interferometer, 17
interferometry, 191 length measurements, 41
materials, 201 light source polarization, 158
microscopy, 195 light sources
optical elements, 197 for holography, 199
processing, 204 for interferometers, 6
recording, 204 linear polarization, 110
stereograms, 184 linear polarizers, 126, 130
holography, 167 Lippman-Denisyuk hologram, 185
applications, 191 liquid crystals, 155
composite, 184 liquid-crystal detectors, 302
338 INDEX
Littrow configuration, 158 N

Lloyd’s mirror, 1
Newton
localized fringes, I2
luminescent detectors, 302 interferometer, 7
Lummer-Gehrke plate, 30 rings, 4, 5
Lyman ghosts, 76 Nicol prism, 139
noise in detectors, 328
nonlinear effect in holography, I72
M
MacNeille polarizing prisms, 130
0
Mach-Zehnder interferometer, 12, 17
magneto-optic effect, 153 off-axis holography, 168
matched filtering. 197 one-step method for rainbow holograms.
materials for optical windows, 331 186
Maxwell equations, 108, 134 optic axis of a crystal, 135
measurement optical
of directional reflectance, 274 activity, 138
of emissivity, 269 elements, holographic, 197
of laser power, 277 polarization, 107
of reflectivity, 272 transfer function, 82, 84
measuring wave plates, 145 window material, 331
mechanical stability in holography, 200 ordinary beam, 136
metal grid polarizers, 157 orthoscopic recording in holography, 186
metal mirrors protection, 33
metals, wave propagation for, 120
meter definition, 41
Michelson P
fringes, 12 P-polarization, 114
interferometer, 4, 9 Parseval’s theorem, 69
stellar interferometer, 22 partial coherence, 89
Michelson- Morley experiment, 38 partially transparent materials, 231
microscopy, 42 path radiance, 236
microtopography. 29 phase
Miyamoto- Wolf theory, 53 contrast microscope, 90
modulation transfer function, 85 difference in gratings, 72
Mooney rhomb, 133 holograms, 190
Mueller’s matrices, 159 modulation, 146
multilayer films, 35, 128 plate, 91
multiple reflection interferometer, 30 retarders, 126, 130
multiple reflections in a thin film, 126 shift upon reflection, 115
multiple-beam interferometer, 25 velocities in a crystal, 136
multiple-pass photocathodes, spectral response of, 307
Fizeau interferometer, 37 photoconductive
Twyman-Green interferometer, 37 detectors, 3 1 1
interferometers, 37 response, 3 19
multiple-reflection photoconductors, extrinsic, 317
Fizeau interferometer, 28 photographic emulsions for holography, 201
interferometers, 25 photometers, 246
lNDEX 339
photometric pyroelectric
and radiometric detector, 299
nomenclature, 214 materials, 301
symbols, 214 pyroheliometers, 246
units, 214
measurements, 263
standards, 264 Q
units, 286 quantum detector, 303
photometry, 213 quarter-wave stack, 35
photomultiplier, 304, 309
photon detectors, 303
photon-drag detector, 327 R
photoresists for holography, 204 radial shear interferometer, 18
photovoltaic devices, 322 radiance temperature, 241
pinhole camera, 61 radiation temperature, 239
plane hologram techniques, 181 radiative transfer, 223
plane-holograms, 169 in a vacuum, 228
Pockels electro-optic effect, 147 in transparent media, 231
PoincarC sphere, 112, 142 radiometer,
point spread function, 85 interferometric, 261
polarization prism disperser, 260, 261
circular, 110 radiometers, 246
determination, 159 astronomical, 249
effects, 128 filter, 257
in metals, 120 imaging, 252
in thin films, 125 scanning, 252
elliptical, 113 space defense, 250
linear, 110 standard, 246
mathematical description, 159 radiometric
optical, 107 instruments, 246
scramblers, 153 measurements, 263
polarizer standards, 264
made of four mirrors, 124 temperature measurements, 239
made of three mirrors, 123 radiometry, 213
polarizers, of visible light, 284
grating, 157 rainbow
linear, 126, 130 cylindrical hologram, 189
metal grid, 157 holograms, 184
slit, 157 Raman-Nath regime, 151
polarizing prisms, MacNeille, 130 ratio temperature, 241
principal angle, 122 difference, 243
prism Rayleigh
beam splitters, 138 interferometer, 22
disperser radiometer, 260 scattering, 99
polarizers, 139 real-time holographic interferometry, 191
processing of holograms, 204 recording of holograms, 204
pseudoscopic configuration in holography, rectangular aperture, 65
187 reference beam in holography, 170
pupil function, 68 reflectance, directional measurement of. 274
340 INDEX
reflection spectral response of photocathodes, 307

circular polarizer, 124 spectroradiometers, 246, 256
diffraction gratings, 78 spectroscopy, 43
holograms, 172 spherical Fabry-Perot interferometer, 29
polarizers, 122 standard
reflectivity, measurement of, 272 lamp, 6
reflectometer, Bennett-Koehler, 274 radiometers, 246
refractometry, 39 standards, radiometric and photometric, 264
relativity measurements, 38 star diameter measurements, 39
resolving power stereograms, holographic, 184
in gratings, 72 Stokes parameters, 159
of optical instruments, 79 straight edge, diffracting, 57
resonance anomalies, 157 synchronous detection, 279
retarding waveplates, 142
reversal shear interferometers, 19
T
ring interferometer, 42
Rochon prism, 138 temperature measurements, 213
Ronchi-test, 39 radiometric, 239
rotational shear interferometers, 19 temporal coherence, 6
Rowland ghosts, 76 thermal detectors, 293, 303
thermocouple calorimeters, 293
S
thermography, medical, 243
S-polarization, 114 thin films, 32, 125
Sagnac interferometer, 42 matrix, 129
saltus problem, 52 polarization effects, 125
sandwich holography, 194 Thompson scattering, 101
scalar theory region, 157 three mirror polarizer, 123
scanning radiometers, 252 three-reference beam method for color
scatter-plate interferometer, 20 holograms, 181
scattering, 49, 92 time averaged holographic interferometry,
by large objects, 94 194
by small particles, 95 total internal reflection devices, 131
cross section, 100 transfer function
schlieren techniques, 39 of lenses, 40
second Brewster angle, 121 optical, 82, 84
senarmount compensator, 144 transmission
shearing interferometers, 17 diffraction gratings, 78
signal-to-noise ratio, 247 holograms, 172
single layer films, 125 polarizers, 122
single-layer coatings, 33 volume color holograms, 183
single-slit diffraction, 53 two dimensional array of holes, diffracting,
slit polarizers, 157 80
slit, diffracting, 57, 65 two-beam interferometers, 7, 22
solei1 compensator, 143 two-step method for rainbow holograms,
space defense radiometers, 250 184
spatial Twyman-Green interferometers, 12
filtering of images, 91
frequencies, 84
multiplexing method for color holograms, U
181 uniaxial crystals, 135
INDE X 34 1
V wave-front topography, 40
white-light compensation, 6
visible light, radiometry of, 284
Wien displacement law, 220
volume hologram, 172
Wollaston prism, 138
techniques for color holograms, 183
Wood anomalies, 157
volume-reflection hologram, 174
W
wave
equations, 108 Y
plates Young's experiment, 1
achromatic, 144
measuring, 145
propagation
for metals, 120
in anisotropic media, 133 2
in isotropic media, 113 zone plate, Fresnel, 59

(Methods in Experimental Physics 26) Daniel Malacara (Eds.) - Physical Optics and Light Measurements-Academic Press (1989) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Methods in Experimental Physics 26) Daniel Malacara (Eds.) - Physical Optics and Light Measurements-Academic Press (1989) PDF

Uploaded by

Copyright:

Available Formats

Methods of

PHYSICAL OPTICS AND LIGHT MEASUREMENTS

Robert Celotta and Judah Levine, Editors-in-Chief

Physical Optics and

ACADEMIC PRESS, INC.

Boston San Diego New York

ACADEMIC PRESS, INC.

United Kingdom Edifion publid~edbv

Library of- Congress Cbraloging-in-Publication Data

Printed in the United States of America

2 . Diffraction and Scattering

2.1. Diffraction ..................... 49

References ..................... 102

3.1. Introduction . . . . . . . . . . . . . . . . . . . . . 107

4.1. Introduction . . . . . . . . . . . . . . . . . . . . . 167

5 . Photometry and Radiometry

5.1. Introduction ..................... 213

5.4. Simple Radiative Transfer . . . . . . . . . . . . . . . 223

6.1. Introduction ........... . ......... 291

Numbers in parentheses indicate pages on which the authors’ contributions begin.

RAMENDRADEOBAHUGUNA(167), Centro de Znvestigaciones en Optica,

Volume 1. Classical Methods

Volume 2. Electronic Methods, Second Edition (in two parts)

Volume 3. Molecular Physics, Second Edition (in two parts)

Volume 4. Atomic and Electron Physics-Part A: Atomic Sources and

Volume 5. Nuclear Physics (in two parts)

Volume 6. Solid State Physics-Part A: Preparation, Structure,

Volume 7. Atomic and Electron Physics-Atomic Interactions (in two

Volume 8. Problems and Solutions for Students

Volume 9. Plasma Physics (in two parts)

Volume 10. Physical Principles of Far-Infrared Radiation

Volume 11. Solid State Physics

Volume 12. Astrophysics-Part A: Optical and Infrared Astronomy

The luminous phenomenon called interference is a direct consequence

1.1 .l. Methods to Obtain Interference Fringes

the interference pattern would be observed at infinity, but lens L serves to

7T-1 sin{ ?sin e]

FIG.2. ( a ) Lloyd's mirror and (b) Fresnel biprism.

1.1.2. Classification of Interference Fringes

FIG. 3. Arrangement to observe Newton fringes.

FIG. 4. Arrangement to produce fringes of equal inclination.

1 .I .3. Light Sources for Interferometers

Howlett”), or even several hundreds of meters for a single-frequency gas

1.2. Two-Beam Interferometers

Another possible classification for interferometers takes into account the

1.2.1. The Newton Interferometer

FIG.S . Newton interferometer.

This instrument forms equal-thickness fringes; and a bright fringe is

2nd = mA, (1.3)

1.2.2. The Fizeau Interferometer

OPD = 2nd, ( 1.4)

't2 urf ace

F I G .6. Testing of spherical surfaces with Newton interferometer.

1.2.3. The Michelson Interferometer

FIG. 7. Typical interference patterns with Newton interferometer.

FIG. 8. Fizeau interferometer.

Fic;. 10. Three kinds of Michelson fringes.

1.2.4. Modifications of t h e Michelson Interferometer

FIG. 11. Jamin interferometer.

F I G . 12. Mach-Zehnder interferometer

FIG. 14. Testing optical components with a Twyman-Green interferometer.

mu t be monochromatic. Many different kinds of optical components ca

FIG. 15. Some Twyman-Green interferograrns.