Professional Documents
Culture Documents
OPTICAL IMAGI G
A D ABERRATIONS
PART II
SECON DEDITION
ExP
II
R- - - - - - r
y
VIRENDRA N. MAHAJAN
OPTICAL IMAGING
AND ABERRATIONS
PART II
Mahajan, Virendra N.
Optical imaging and aberrations / Virendra N. Mahajan. -- 2nd ed.
v. cm. -- (Press monograph ; 209)
Contents: -- pt. 2. Wave diffraction optics.
Includes bibliographical references and index.
ISBN 978-0-8194-8699-8 (pt. 2)
1. Aberration. 2. Imaging systems. 3. Geometrical optics. I. Title.
QC671.M36 2011
621.36--dc23
2011020047
Published by
SPIE
P.O. Box 10
Bellingham, Washington 98227-0010 USA
Phone: +1 360.676.3290
Fax: +1 360.647.1445
Email: Books@spie.org
Web: http://spie.org
All rights reserved. No part of this publication may be reproduced or distributed in any
form or by any means without written permission of the publisher.
The content of this book reflects the work and thought of the author(s). Every effort has
been made to publish reliable and accurate information herein, but the publisher is not
responsible for the validity of the information or for any outcomes resulting from reliance
thereon.
Three years ago Vini Mahajan published a book on the subject of Geometrical
Images in the presence of aberrations. In this book, Mahajan extends this work to include
the effect of wave optics. He continues his thorough tutorial on image formation with a
detailed look at the approaches to calculating the form of images. Anyone interested in
understanding the methods of predicting the light distribution to be expected in real
imaging situations will find this book of interest.
The book begins with an exhaustive development of the basics of diffraction image
formation. Mahajan covers the issues associated with the calculation of point-spread
functions and discusses the accuracy of such calculations. He introduces the Optical
Transfer Function as the Transform of the Point Spread Function and reviews the
procedures involved in calculating the OTF. Asymptotic and approximate evaluations of
the OTF are included, as are several examples throughout the book.
These approaches are then applied to some real examples of circular and annular
apertures. In this discussion, Mahajan carries out in detail many of the classical
computations for various image descriptors. This is a topic that is generally treated only
lightly in most texts on the subject. Such issues as edge response and line spread function
and encircled power are carefully considered. A good discussion of optimal balancing of
aberrations is also provided. The treatment of aberration balancing and tolerances in
annular pupils is unique in its completeness in this book.
vii
TABLE OF CONTENTS
ix
1.9.2 Approximate Expressions for Strehl Ratio............................................................. 56
1.9.3 Determination of Strehl Ratio ................................................................................ 58
1.10 Hopkins Ratio ....................................................................................................................... 59
1.11 Line- and Edge-Spread Functions (LSF and ESF) ........................................................... 61
1.11.1 Line-Spread Function ............................................................................................. 61
1.11.2 Edge-Spread Function ............................................................................................ 64
1.11.3 LSF and ESF in Terms of OTF .............................................................................. 65
1.12 Shift-Invariant Imaging of a Coherent Object .................................................................. 67
1.12.1 Coherent Point-Spread Function ............................................................................ 67
1.12.2 Coherent Transfer Function ................................................................................... 70
1.13 Summary of Theorems......................................................................................................... 71
Appendix A: Fourier Transform Definitions ............................................................................... 74
Appendix B: Some Frequently Used Integrals ............................................................................. 75
References ........................................................................................................................................ 76
Problems........................................................................................................................................... 78
x
2.6.3 Gaussian Approximation...................................................................................... 134
2.6.4 Summary of Results ............................................................................................. 135
2.7 Symmetry Properties of an Aberrated PSF ..................................................................... 136
2.7.1 General Theory ..................................................................................................... 137
2.7.2 Symmetry About the Gaussian Image Plane ........................................................ 138
2.7.3 Symmetry of Axial Irradiance.............................................................................. 141
2.7.4 Symmetry in Sign of Aberration Coefficient ....................................................... 141
2.8 PSFs for Primary Aberrations .......................................................................................... 142
2.8.1 Defocus ................................................................................................................. 142
2.8.2 Spherical Aberration Combined with Defocus..................................................... 142
2.8.3 Astigmatism Combined with Defocus ................................................................. 144
2.8.4 Coma..................................................................................................................... 148
2.8.5 2D PSFs ................................................................................................................ 150
2.8.6 Comparison of Diffraction and Geometrical PSFs .............................................. 157
2.9 Line of Sight of an Aberrated System .............................................................................. 159
2.9.1 PSF and Its Centroid............................................................................................. 159
2.9.2 Numerical Results ................................................................................................ 162
2.9.2.1 Wavefront Tilt ...................................................................................... 162
2.9.2.2 Primary Coma ...................................................................................... 162
2.9.2.3 Secondary Coma .................................................................................. 165
2.9.3 Comments ............................................................................................................. 168
2.10 Diffraction OTF for Primary Aberrations....................................................................... 169
2.10.1 General Relations ................................................................................................. 169
2.10.2 Defocus ................................................................................................................. 172
2.10.3 Spherical Aberration............................................................................................. 174
2.10.4 Astigmatism ......................................................................................................... 174
2.10.5 Coma..................................................................................................................... 176
2.11 Hopkins Ratio ..................................................................................................................... 182
2.11.1 Tolerance for Primary Aberrations ...................................................................... 182
2.11.2 Defocus ................................................................................................................. 182
2.11.3 Hopkins Ratio in Terms of Variance of Aberration Difference
Function ................................................................................................................ 185
2.11.4 Variance of Aberration Difference Function for Primary Aberrations ................ 186
2.12 Geometrical OTF................................................................................................................ 187
2.12.1 General Relations ................................................................................................. 188
2.12.2 Radially Symmetric Aberration............................................................................ 189
2.12.3 Defocus ................................................................................................................. 189
2.12.4 Spherical Aberration Combined with Defocus..................................................... 190
2.12.5 Astigmatism Combined with Defocus ................................................................. 190
2.12.6 Coma..................................................................................................................... 191
2.13 Incoherent Line- and Edge-Spread Functions ................................................................. 191
2.13.1 Theory .................................................................................................................. 192
xi
2.13.1.1 LSF from PSF ...................................................................................... 192
2.13.1.2 LSF from Pupil Function ..................................................................... 192
2.13.1.3 Struve Ratio and Aberration Tolerances .............................................. 193
2.13.1.3.1 Defocus ............................................................................. 194
2.13.1.3.2 Astigmatism Combined with Defocus .............................. 195
2.13.1.4 LSF from OTF...................................................................................... 196
2.13.1.5 ESF from OTF...................................................................................... 198
2.13.2 Numerical Results ................................................................................................ 199
2.14 Miscellaneous Topics ......................................................................................................... 205
2.14.1 Polychromatic PSF ............................................................................................... 205
2.14.2 Polychromatic OTF .............................................................................................. 208
2.14.3 Image of an Incoherent Disc................................................................................. 209
2.14.3.1 Gaussian Image .................................................................................... 210
2.14.3.2 Diffraction Image ................................................................................. 210
2.14.3.3 Numerical Results ................................................................................ 213
2.14.4 Pinhole Camera .................................................................................................... 218
2.15 Coherent Imaging............................................................................................................... 222
2.15.1 Coherent Spread Function .................................................................................... 222
2.15.2 Coherent Transfer Function ................................................................................. 223
2.15.3 Coherent LSF ....................................................................................................... 224
2.15.4 Coherent ESF ....................................................................................................... 229
2.15.5 Image of a Coherent Disc ..................................................................................... 234
2.15.5.1 Diffraction Image ................................................................................. 234
2.15.5.2 Numerical Results ................................................................................ 235
2.15.6 Use of a Lens for Obtaining Fourier Transforms ................................................. 238
2.15.7 Comparison of Coherent and Incoherent Imaging ............................................... 241
2.15.7.1 Frequency Spectra of Images ............................................................... 241
2.15.7.2 Two-Point Resolution .......................................................................... 245
2.16 Summary ............................................................................................................................. 253
References ...................................................................................................................................... 258
Problems......................................................................................................................................... 262
xii
3.3 Strehl Ratio and Aberration Tolerance............................................................................ 287
3.3.1 Strehl Ratio........................................................................................................... 288
3.3.2 Primary Aberrations ............................................................................................. 289
3.3.3 Balanced Primary Aberrations ............................................................................. 289
3.3.4 Comparison of Approximate and Exact Results .................................................. 290
3.4 Balanced Aberrations and Zernike Annular Polynomials ............................................. 297
3.5 Defocused System ............................................................................................................... 304
3.5.1 Point-Spread Function.......................................................................................... 304
3.5.2 Focused Beam ...................................................................................................... 305
3.5.3 Collimated Beam.................................................................................................. 309
3.6 Symmetry Properties of an Aberrated PSF ..................................................................... 311
3.7 PSFs and Axial Irradiance for Primary Aberrations ..................................................... 314
3.8 2D PSFs ............................................................................................................................... 317
3.9 Line of Sight of an Aberrated System .............................................................................. 328
3.9.1 PSF and Its Centroid............................................................................................. 328
3.9.2 Numerical Results ................................................................................................ 329
3.9.2.1 Wavefront Tilt ...................................................................................... 329
3.9.2.2 Primary Coma ...................................................................................... 330
3.9.2.3 Secondary Coma .................................................................................. 333
3.10 Summary ............................................................................................................................. 336
References ...................................................................................................................................... 339
Problems......................................................................................................................................... 340
xiii
4.3.4.1 Pupil Function ...................................................................................... 372
4.3.4.2 PSF ....................................................................................................... 373
4.3.4.3 Axial Irradiance .................................................................................... 374
4.3.4.3.1 Focused Beam ..................................................................... 374
4.3.4.3.2 Depth of Focus .................................................................... 379
4.3.4.3.3 Diffraction Focus ................................................................. 379
4.3.4.3.4 Collimated Beam ................................................................. 382
4.3.4.4 OTF ...................................................................................................... 385
4.3.5 Balancing of Defocus Aberration with Spherical Aberration or
Astigmatism ......................................................................................................... 387
4.3.5.1 Focused Beam ...................................................................................... 387
4.3.5.2 Collimated Beam.................................................................................. 392
4.3.6 Aberrated System ................................................................................................. 395
4.3.6.1 Spherical Aberration............................................................................. 395
4.3.6.2 Symmetry Properties of an Aberrated PSF .......................................... 398
4.3.7 Weakly Truncated Gaussian Pupils and Beams ................................................... 401
4.3.7.1 Pupil Function ...................................................................................... 401
4.3.7.2 PSF ....................................................................................................... 402
4.3.7.3 Radius of Curvature of the Propagating Wavefront ............................. 407
4.3.7.4 Collimated beam .................................................................................. 408
4.3.7.5 Beam Focusing and Waist Imaging by a Lens ..................................... 410
4.3.7.6 OTF ...................................................................................................... 414
4.3.7.7 Strehl Ratio, Aberration Balancing, and Zernike-Gauss Circle
Polynomials .......................................................................................... 415
4.3.7.8 Beam Characterization and Measurement............................................ 419
4.4 Systems with Annular Pupils............................................................................................. 421
4.4.1 Pupil Irradiance .................................................................................................... 422
4.4.2 Aberration-Free System ....................................................................................... 423
4.4.3 Strehl Ratio and Aberration Tolerance ................................................................. 425
4.4.4 Balanced Aberrations and Zernike-Gauss Annular Polynomials ......................... 425
4.4.5 Defocused System ................................................................................................ 429
4.4.5.1 PSF ....................................................................................................... 429
4.4.5.2 Axial Irradiance .................................................................................... 429
4.4.5.3 Collimated Beam.................................................................................. 432
4.4.6 Symmetry Properties of an Aberrated PSF .......................................................... 433
4.5 Line of Sight of an Aberrated System .............................................................................. 434
4.5.1 PSF and Its Centroid............................................................................................. 434
4.5.2 Numerical Results ................................................................................................ 435
4.5.2.1 Wavefront Tilt ...................................................................................... 435
4.5.2.2 Primary Coma ...................................................................................... 435
4.5.2.3 Secondary Coma .................................................................................. 436
xiv
4.6 Summary ............................................................................................................................. 438
References ...................................................................................................................................... 441
Problems......................................................................................................................................... 444
xv
5.3.8.3 Far-Field Imaging ................................................................................. 518
5.3.8.4 Simulated Star Images.......................................................................... 521
5.3.9 Lucky Imaging and Adaptive Optics.................................................................... 525
5.4 Summary ............................................................................................................................. 527
Appendix: Fourier Transform of Zernike Circle Polynomials................................................ 530
References ...................................................................................................................................... 532
Problems......................................................................................................................................... 535
xvi
PREFACE TO THE FIRST EDITION
In Part I of this book, we discussed imaging based on ray geometrical optics. The
aberration-free image of an object according to it is the exact replica of the object, except
for its magnification. The aberration-free image of a point object is also a point. In reality,
however, the image obtained is not a point. Because of diffraction of the object wave at
the aperture stop, or equivalently, the exit pupil of the imaging system, the actual
aberration-free image for a circular exit pupil is a light patch surrounded by dark and light
rings. Geometrical optics is still assumed to hold from the object to the exit pupil in that
rays are traced through the system to determine the shape of the pupil and the aberration
across it. The determination of the characteristics of the diffraction image of an object
formed by an aberrated system is the subject discussed in Part II. The emphasis of this
part is on the numerical results on the effects of aberrations on a diffraction image and not
on the formalism, exposition, or a critique of the variety of diffraction theories proposed
over the years. It is a compilation of the works of masters like Nijboer, Hopkins, Barakat,
and Fried with a sprinkling of my own work.
Chapter 2 on systems with circular pupils starts with the aberration-free PSF and its
encircled and ensquared powers. The effects of primary aberrations on its Strehl ratio are
discussed, and aberration tolerances are obtained. Balanced primary aberrations are
xvii
considered and identified with Zernike circle polynomials. Focused and collimated beams
are discussed, and the concept of near- and far-field distances is introduced. Aberrated
PSFs and their symmetry properties are discussed, and a brief comparison is made with
their counterparts in geometrical optics. The line of sight of an aberrated system,
identified with the centroid of the PSF, is determined for primary aberrations. The OTF
for these aberrations is also discussed, phase contrast reversal is explained, and aberration
tolerances for a certain value of Hopkins ratio are given. Expressions for the geometrical
OTF for primary aberrations are also given. Both incoherent and coherent line- and edge-
spread functions are discussed, and aberration tolerances for a certain value of the Struve
ratio are given. A brief comparison of incoherent and coherent imaging is given with
special reference to the Rayleigh criterion of resolution. The Fourier-transforming
property of wave propagation is illustrated in altering the image of an object by spatial
filtering in the Fourier-transform plane.
Systems with annular pupils are given a cursory look at best in books where imaging
is discussed. Our Chapter 3 is written in a manner similar to Chapter 2 where, for
example, the balanced aberrations are identified with the Zernike annular polynomials.
Although the propagation of Gaussian beams is discussed in books on lasers, their
treatment is generally limited to the weakly truncated aberration-free beams. In Chapter
4, we consider the effect of arbitrary truncation of beams with and without aberrations.
The balanced aberrations in this case are identified with the Zernike-Gauss polynomials.
It is shown, for example, that the pupil radius must be at least three times the beam radius
in order to neglect beam truncation without significant error.
Finally, random aberrations are considered in Chapter 5. The effect of random image
motion is considered first, and expressions and numerical values of time-averaged Strehl
ratio, PSF, and encircled power are given for systems with circular and annular pupils.
The random aberrations introduced by atmospheric turbulence when a wave propagates
through it, as in astronomical observations by a ground-based telescope, are discussed;
and expressions for time-averaged PSF and OTF are obtained. The aberration function for
Kolmogorov turbulence is expanded in terms of the Zernike polynomials, and auto-
correlation and cross-correlation of the expansion coefficients are given. The atmospheric
coherence length is defined, and it is shown that the resolution of a telescope cannot
exceed that of an aberration-free telescope of this diameter. Both the short- and long-
exposure images are considered.
As in Part I, each chapter ends with a set of problems. It is hoped that they will
acquaint the reader with application of the theory in terms of some practical examples.
References for addtional reading are given after the Bibliography. These references
represent the author’s collection as the editor of Milestone Series 74 entitled Effects of
Aberrations in Optical Imaging, published by SPIE Press in 1993.
xviii
ACKNOWLEDGMENTS
It is a great pleasure to acknowledge the generous support I have received over the
years from my employer, The Aerospace Corporation, in preparing this book. My special
thanks go to the senior vice president Mr. John Parsons, for his continuous interest and
encouragement in this endeavor. I thank Mr. John Hoyem for preparing the figures and
Mr. Victor Onouye for some figures as well as the final composition. My thanks also go
to Dr. Rich Boucher for computer generating the 2D PSFs and Mr. Yunsong Huang for
numerical analysis and computer plotting. The Sanskrit verse on p. xxv was provided by
Professor Sally Sutherland of the University of California at Berkeley.
I thank my sister Pushpa and brother Devinder for encouraging me to study phyiscs. I
can not say enough about the constant support I have received from my wife Shashi over
the many years it has taken me to complete this two-part book. I dedicate this book to my
departed parents who brought and nourished me in this world.
Finally, I would like to thank SPIE Press Senior Editor Dara Burrows for
meticulously editing the book and SPIE Press Manager Tim Lamkins for facilitating the
publication of this Second Edition. It has been a pleasure to work with them, especially
because of their cooperative spirit and quality support.
xix
PREFACE TO THE SECOND EDITION
Ten years have passed since the publication of the first edition in April 2001. Many
of the typographical errors were corrected in the Second Printing that took place after 3
years in 2004. Only a small amount of new material, approximately 11 pages, was added
at that time. It included Appendix B in Chapter 1, Gaussian OTF in Chapter 4, and an
extension of the discussion of the short-exposure image in Chapter 5. Additional
corrections were made in the e-version of the book in 2009, and the discussion of Zernike
circle, annular, and Gauss polynomials was streamlined utilizing an abbreviated notation
with emphasis on their orthonormal form. However, a considerable amount of new
material amounting to another 88 pages has been added in this Second Edition. Besides
correction of some residual typographical errors, a Summary section has been included in
Chapters 2, 3, and 5 for consistency with Chapters 1 and 4. Any compound references
have been split into single ones. The new material is primarily in Chapters 4 and 5.
In Chapter 5, the effect of random transverse image motion on the PSF has been
supplanted by a discussion of the effect of random longitudinal defocus. The coherence
length of atmospheric turbulence is calculated for the Hufnagel-Valley model of the
refractive index structure parameter, for both looking up and down through the
atmosphere. The angle of arrival of the light wave propagating through turbulence is
discussed for both the Zernike tilt as well as the centroid of the PSF, showing that they
are nearly equal. A brief discussion of lucky imaging is also given, where better quality
short-exposure images are selected, aligned, and added to obtain a high-quality image.
xxi
SYMBOLS AND NOTATION
xxiii
Anantaratnaprabhavasya yasya himam
. na saubhagyavilopi jatam
Eko hi doso
. gunasannipate
. ˙ .
nimajjatindoh. kiranesvivankah
.
The snow does not diminish the beauty of the Himalayan mountains
which are the source of countless gems. Indeed, one flaw is lost
among a host of virtues, as the moon’s dark spot is lost among its rays.
xxv
CHAPTER 1
IMAGE FORMATION
1
2 IMAGE FORMATION
Geometrical optics is assumed to hold from the point object to the exit pupil in that
rays are traced through the system to determine the shape of the pupil and the aberration
across it for the point object under consideration, as discussed in Section 3.2 of Part I.
The amplitude associated with a ray is obtained by consideration of the transmission and
reflection characteristics of the imaging elements of the system. In many applications, the
amplitude variation across the pupil is negligible. There are instances, however, when the
transmission across the pupil is varied to alter certain characteristics of the diffraction
3
4 IMAGE FORMATION
PSF. The pupil is said to be apodized in these cases. Another important example of
variable amplitude across the pupil is that of laser beams for which the amplitude has a
Gaussian form. The irradiance distribution of the diffraction image of an incoherent
object is obtained by adding the irradiance distribution of the images of its infinitesimal
elements. In particular, we show that the diffraction image of an isoplanatic object, i.e.,
one for which the pupil function of the system is independent of the position of a point on
it, is equal to the convolution of its Gaussian image and the (diffraction) PSF of the
system.
We also discuss the optical transfer function of an imaging system, which is equal to
a Fourier transform of its PSF. It describes the amplitude and phase of the sinusoidal
image of a sinusoidal object as a function of its spatial frequency. Thus, it relates the
spatial frequency spectrum of the object to that of the image. We show, for example, that
the slope of its real part at the origin is independent of aberrations. We also compare the
diffraction OTF with the corresponding OTF based on geometrical optics. The asymptotic
behavior and centroid of an aberrated point-spread function are related to the slope of the
real and imaginary parts of the OTF at the origin, respectively. It is shown that the
centroid of the diffraction PSF is identical with that of the geometrical PSF.
Next, image quality based on the Strehl ratio of an aberrated imaging system is
discussed, and approximate expressions in terms of its aberration variance are obtained. A
brief discussion of how the Strehl ratio of a system may be determined is also given. The
image quality based on Hopkins ratio in the spatial frequency domain is also discussed.
Whereas the Strehl ratio represents the ratio of the central value of the irradiance
distribution of the aberrated image of a point object and the corresponding aberration-free
value, the Hopkins ratio represents the ratio of the contrasts of the aberrated and
aberration-free images of a sinusoidal object of a certain spatial frequency. The Strehl and
Hopkins ratios provide simple means of determining the aberration tolerances. The line-
spread function (i.e., the image of a line object) and the edge-spread function (i.e., the
image of an edge or a step object) are discussed in terms of the PSF and related to the
OTF. The Struve ratio of a system, representing the ratio of the aberrated central value of
the LSF and the corresponding aberration-free value, is also considered, and aberration
tolerances are derived from it.
coherent spread function, and its incoherent transfer function is equal to the
autocorrelation of its coherent transfer function.
The fundamental relations derived in this chapter are stated as theorems and used in
the succeeding chapters to obtain some practical results for imaging systems with
circular, annular, and Gaussian pupils.
2 2 2
2 = 2 + 2 + . (1-2)
x y z 2
r r
In Figure 1-1, r is used in the z = 0 plane to distinguish it from the r in the z plane, as
the two planes become the source and observation planes, respectively.
6 IMAGE FORMATION
-
P' (r' ; 0)
P (r; z)
r'
O,~/O
z=OPlane ~
z~
~o
z - Plane
Figure 1-1. Geometry of wave propagation for determining the complex amplitude
V(r; z) in a z plane from its knowledge V(r'; 0) in the z = 0 plane. O'P = [0'
P'P=[ and O'O=z.
To relate V(r; z) to V(r; 0), we decompose both of them into plane waves. Thus,
for example, we write
where A(v; 0) is the amplitude of a plane wave propagating with direction cosines
(a, ~, y) such that the spatial frequency v is given by
(1-4)
and
(1-5)
A(v; 0) is also referred to as the spectral component of V(r; 0) with a spatial frequency
V. Written in the form A( a/A, ~/A; 0), it is also called the angular spectrum of
V(r; 0). [Equation (1-4) may be obtained by comparing the exponential in Eq. (1-3) with
the standard form of a plane wave exp[i( wt - k . r)], where wand k are its angular
frequency and propagation vector, and t is time.]
Multiplying both sides ofEq. (1-3) by exp(2nir· v') and integrating, we obtain
JU(r; 0)exp(2nir ·v')dr = JA(v; O)dv Jexp[2nir -(v' - v)]dr
r
where () is a Dirac delta function. It is clear from Eqs. (1-3) and (1-6) that U ( r ; 0) and
r r
A( v ; 0) form a 2D Fourier-transform pair; U ( r ; 0) is the inverse Fourier transform of
r r r
A( v ; 0) and A( v ; 0) is the Fourier transform of U ( r ; 0) . (For a definition of the Fourier
transform, see the Appendix.) Similarly, we may write
r r r r r
U ( r ; z ) = A( v ; z ) exp (
2ir v ) d v , (1-7)
r r
where A( v ; z ) is the amplitude of a plane-wave component of spatial frequency v in the
z plane given by
r r r r r
A( v ; z ) = U ( r ; z ) exp (2ir v ) dr . (1-8)
r
Since U ( r ; z ) satisfies the wave equation, which is linear, each component of the
plane waves that comprise it must also satisfy it. Hence, substituting
r r r
A( v ; z ) exp (
2ir v ) for it into Eq. (1-1), we obtain
r
r 2 A( v ; z )
k A( v ; z ) +
2 2
= 0 . (1-9)
z 2
so that
r r r r
exp (ik z ) = h( r ; z ) exp (2 ir v ) dr . (1-13)
where we have used Eq. (1-3). By change of variables, Eq. (1-14) containing the
convolution integral can also be written
r r r r r
U ( r ; z ) = h( r
r ; z ) U ( r ; 0) dr . (1-15)
r
The integrals in Eqs. (1-14) and (1-15) giving the wave field U ( r ; z ) at the z plane
8 IMAGE FORMATION
r r
represent a convolution of h( r ; z ) and the wave field U ( r ; 0) in the z = 0 plane.
r
If a point source of unit amplitude is placed at a point ro in the z = 0 plane, i.e., if
r r r
U ( r ; 0) = ( r
ro ) , (1-16)
r
then Eq. (1-15) shows that the complex amplitude at a point r in the z plane is given by
r r r
U ( r ; z ) = h( r
ro ; z ) . (1-17)
r r
Thus, h( r ; z ) represents the complex amplitude at a point r in the z plane due to a point
r
source of unit amplitude located at r = 0 in the z = 0 plane. It is called the complex
amplitude point-spread function (or impulse response) of free space. Carrying out the
integration in Eq. (1-12), it can be shown that5
r 1 1 z exp (ikl0 )
h( r ; z ) =
i , (1-18)
kl0 l0 l0
where
( )
1/ 2
l0 = z 2 + r 2 (1-19)
is the distance OP between the origin in the z = 0 plane and the observation point
r
( r ; z) . Equation (1-18) is a mathematical expression for a Huygens’ spherical wavelet
diverging from the point source, and it describes the complex amplitude point-spread
r r
function of free space. We note from Eq. (1-12) that h( r ; z ) (r ) as z 0 , i.e., it
becomes the point source, as expected. Accordingly, Eq. (1-15) is a mathematical
description of Huygens-Fresnel principle, namely, that the complex amplitude in the z
r r
plane is a linear superposition of Huygens’ secondary spherical wavelets h( r
r ; z )
r
weighted by the amplitudes U ( r ; 0) of the wave where they originate (Theorem 1). The
diffracted wave field described by Eq. (1-15) is shift invariant (or isoplanatic) in that a
r r r r
spherical wavelet at a point r due to a source point at r depends on r
r , i.e., the
form of a spherical wavelet is independent of the location of its origin in the z = 0 plane,
except for a shift in the center of the distribution.
[ r r 2
]
12
l = z2 + r
r (1-21a)
r r 1 r r 4
= z+
1 2
2z
( )
r + r 2
2 r r
3 r
r + ...
8z
(1-21b)
1.2 Rayleigh Sommerfeld Theory of Diffraction and Huygens Fresnel Principle 9
r r
is the distance P P between a source point (r ; 0) and the observation point (r ; z ) .
Equation (1-20) is the Rayleigh-Sommerfeld formula describing propagation of a wave
from one plane to another.
r
i r2 ikr 2 r 2 i r r r
U (r ; z) = exp ik z + exp U ( r ; 0) exp
z r r dr ,
z
2 z
2 z
r r 4
for z 3 >> k r
r max 8 ( Fresnel) , (1-22a)
and
r
i r2 r 2 i r r r
U (r ; z) = exp ik z + U ( r ; 0) exp
r r dr ,
z
2 z
z
Thus, depending on the value of the distance z relative to the extent of the regions of the
source field and observation, the complex amplitude in a z plane is proportional to the
inverse Fourier transform of the complex amplitude in the z = 0 plane with or without
modification by a quadratic phase factor kr 2 / 2 z (Theorem 2). Equations (1-22a) and
(1-23a) represent diffraction integrals in the Fresnel and Fraunhofer approximations,
respectively. It should be noted that the Fresnel condition of large distance z in Eq. (1-
22a) is a sufficient but not a necessary condition. What is necessary is that the sum of the
neglected terms be small so that their contribution to the diffraction integral is negligible.
The integrals in Eqs. (1-22a) and (1-23a) are referred to as representing the Fresnel
r
and Fraunhofer diffraction patterns of the distribution U ( r ; 0) . The region of space
satisfying the condition z >> krmax
2 / 2 is called the Fraunhofer or the far-field region of
diffraction. The condition itself is called the Fraunhofer or the far-field condition. The
r r 4
region of space satisfying the Fresnel condition z 3 >> k r
r max / 8 but not the
Fraunhofer condition is often referred to as the region of Fresnel or near-field diffraction.
The region of very small z values is referred to as the Rayleigh-Sommerfeld region of
diffraction. It should be evident, though, that the Rayleigh-Sommerfeld integral in Eq. (1-
20) will yield accurate results (within the range of its validity) regardless of the value of
z. Similarly, the Fresnel integral will yield accurate results in the Fraunhofer region as
well. However, calculations in the Fraunhofer region are simpler and more common in
imaging applications (due to cancellation of the quadratic phase factor of free-space
propagation by the focusing quadratic phase factor provided by the imaging system).
10 IMAGE FORMATION
Once z is large enough to satisfy the Fraunhofer condition, a larger z value changes only
the scale of the irradiance distribution without changing the distribution.
r r
r r
( )
2
r 2 r r r 2
r r
l = l0 +
+ ... . (1-21c)
2l0 l0 8l0 3
Accordingly, Eqs. (1-22a) and (1-23a) representing the Fresnel and Fraunhofer
approximations are replaced by
r
i z exp(ikl0 ) ikr 2 r 2 i r r r
U (r ; z) = exp U ( r ; 0) exp
r r dr (1-22b)
l0 l0
2l0
l0
and
r
i z exp(ikl0 ) r 2 i r r r
U (r ; z) = U ( r ; 0) exp
l r r dr . (1-23b)
l0 l0
0
It should be evident that exp(ikl0 ) l0 represents a spherical wave originating at the origin
r
of the z = 0 plane, and ( r ; z ) l0 represents the direction cosines of the position vector of
the observation point. Since l0 varies with the point of observation, a uniform array of
points in the observation plane becomes nonuniform when divided by l0 , thereby making
the numerical calculations of the Fourier transform cumbersome. In Eq. (1-22b), there is
an additional difficulty due to the quadratic phase factor under the integral. These
difficulties may be overcome by making observations on a hemisphere of radius l0 , since
the points in that case are uniformly distributed in the direction cosine space. The
obliquity factor z l0 reduces to unity in that case. In imaging, however, the observations
are made in a plane. Also, the fields of view of high-quality imaging systems are
generally small, and Eqs. (1-22a) and (1-23a) representing diffraction at small angles
suffice. Moreover, imaging systems generally have aberrations that increase rapidly with
the field of view and impact the diffracted field a lot more significantly than any error
made in assuming small angles of diffraction.
To get an idea of the distances of the Fresnel and Fraunhofer regions from a
diffracting aperture, we consider a circular aperture with a diameter of 1 cm (so that
rmax = 0.5 cm ) illuminated by a collimated beam of light with a wavelength of 0.5 mm.
r4
The Fresnel condition z 3 > k rmax
/ 8 for on-axis diffraction is satisfied for z > 10 cm. Of
course, the larger the distance, the better the approximation. However, for the Fraunhofer
r2
condition z > k rmax
/ 2 to be satisfied, observations must be made at z > 157 m. In the
nomenclature of optical aberrations, a phase factor varying as r 4 is referred to as a
primary spherical aberration. Similarly, a phase factor varying as r 2 is referred to as a
defocus aberration. They are discussed in detail in Section 1.4.2. At z = 10 m, for
1.2 Rayleigh Sommerfeld Theory of Diffraction and Huygens Fresnel Principle 11
example, the amount of defocus wave aberration rmax 2 2 z = 2.5 , but the spherical
4 8z 3 is only on the order of 10-7 l. Even at z = 1 m, the defocus aberration
aberration rmax
is 25 l but spherical aberration is only on the order of 10-4 l and, therefore, negligibly
small. At shorter distances, it increases rapidly and the Fresnel approximation does not
hold. Since the only difference between the diffraction integrals represented by Eqs. (1-
22a) and (1-23a) is the presence of the quadratic phase factor in the former, Fresnel
diffraction may be referred to as the defocused Fraunhofer diffraction.
In the Fresnel approximation, Eq. (1-18) for the point-spread function reduces to
r i r2
h( r ; z ) =
exp ikz 1 + 2 . (1-24)
z
2 z
Here, kz represents the phase delay of all components in propagating from one plane to
another a distance z apart. The factor of
i indicates that the diffracted wave is out of
phase with the incident wave by / 2 . The inverse dependence on z represents the
inverse-square law of wave intensities. Equation (1-24) shows that a Huygens’ spherical
wavelet is replaced by a parabolic one in the Fresnel approximation. The parabolic
surface shifts laterally as the source of the secondary wavelet shifts in the z = 0 plane,
and the diffracted field is shift-invariant. The shift-invariant property may also be seen by
r r r
letting U ( r ; 0) equal a Dirac delta function ( r
r0 ) in Eq. (1-22a).
r
If we consider a Huygens’ wavelet centered at r , Eq. (1-24) is replaced by
r r i ( r r 2
r
r )
h( r
r ; z ) =
exp ikz 1 + ! . (1-25)
z 2 z 2
"
In the Fraunhofer approximation, the term in r 2 in the exponent on the right-hand side
of Eq. (1-25) is neglected so that it reduces to
rr r r
i r 2
2r r
h( r ; r ; z ) =
exp ikz 1 + . (1-26)
z
2z 2
If we ignore the quadratic phase factor varying as r 2 in Eq. (1-26) (because it is small for
large z and it does not impact the irradiance distribution of the Fraunhofer pattern), Eq.
(1-26) shows that in effect the Huygens’ spherical wavelet has been replaced by a planar
one. The direction cosines of its surface normal are given by (#, $, ) , where
r
(#, $) = (r / z) . Physically, this is understandable since a small portion of a large
spherical surface can be approximated by a plane. As the source of the secondary wavelet
r
shifts in the z = 0 plane, thereby changing r , the plane surface representing it is tilted.
Hence, the shift-invariant property of the diffracted field is lost. This may also be seen by
r r r
letting U ( r ; 0) equal a Dirac delta function ( r
r0 ) in Eq. (1-23a). However, this loss
is only in a mathematical sense. In reality, the quadratic phase factor k r ¢ 2 2z (which is
12 IMAGE FORMATION
negligibly small in the Fraunhofer region) is nonzero, and the diffracted wave field
remains shift invariant.
[ (
H (, ; z ) = exp ik z 1 2 + 2 )]
1/ 2
( Rayleigh Sommerfeld) . (1-27)
For small values of the direction cosines a and b, i.e., for small angles of diffraction with
z axis, Eq. (1-27) reduces to
2
(
H (, ; z ) = exp ik z 1 2 + 2
1
)
( Fresnel and Fraunhofer ) . (1-28)
It is easy to show that it represents the Fourier transform of the point-spread function in
the Fresnel approximation given by Eq. (1-24) [see Problem 1]. Hence, the Fresnel
approximation is synonymous with diffraction in the paraxial approximation. It should be
noted that, mathematically, there is no transfer function associated with Fraunhofer
diffraction since the shift-invariant property of the diffracted field is destroyed in the
Fraunhofer approximation. However, as explained above following Eq. (1-26), the
Fresnel transfer function can be used in this approximation as well.
Theorem 3. The Gaussian image of a small object lying at a large distance from an
imaging system is an exact replica of the object, except for its magnification.
The object and the image lie in mutually parallel planes that are perpendicular to the
optical axis of the system. The aperture stop of the system and its images in the object
and image spaces, namely, the entrance and exit pupils, respectively, also lie in planes
that are parallel to the object and image planes.
1.3 Gaussian Image 13
r
Consider, as illustrated in Figure 1-2, an object of radiance B ( ro ) (in W m 2 sr ),
r
where ro is the position vector of a point in the object plane lying at a distance zo from
the plane of the entrance pupil. The position of the Gaussian image of a point object
r
located at ro is given by
r r
rg = M ro (1-29)
in the Gaussian image plane, where M is its transverse magnification. An object area
r r
element ro centered at ro is imaged as an area element
r r
rg = M 2 ro (1-30)
r
centered at rg . We assume that the object lies at a very small angle from the optical axis
of the system so that its cosine is approximately equal to unity. We also assume that the
size of the entrance pupil is small compared to zo , so that all points on it lie at
approximately the same distance zo from the object element. Accordingly, the solid angle
subtended by an entrance pupil of an area Sen at the object element is Sen zo2 . The power
entering the pupil from the object element is given by
r r
( )
Pen = Sen zo2 B ( ro ) ro . (1-31)
If is the transmission factor of the system representing the fraction of the incident
power transmitted by it, the corresponding power exiting from the exit pupil is given by
Pex = Pen
r r
( )
= Sen zo2 B ( ro ) ro . (1-32)
EnP ExP
Po( ro) Pg( rg)
Oen OA Oex
Figure 1-2. Schematic of Gaussian imaging by an optical system with its optical axis
OA, entrance pupil EnP, and exit pupil ExP. A point object lies at Po with a position
r
vector ro in the object plane at a (numerically negative) distance zo from the
r
entrance pupil. Its Gaussian image lies at Pg with a position vector rg in the
Gaussian image plane at a distance z g from the exit pupil.
14 IMAGE FORMATION
r
(r )
This power is contained in the image element rg . If Ig rg is the irradiance (in W m 2 )
of this image element, then
(r )
Pex = Ig rg rg
r
. (1-33)
(r ) r
( ) r
Ig rg rg = Sen zo2 B ( ro ) ro
r
. (1-34)
(r ) (
Ig rg = Sen zo2 M 2 B rg M) (r ) . (1-35)
Equation (1-35) describes the irradiance distribution of the Gaussian image, i.e., the
aberration-free image according to geometrical optics. It shows that, except for its
magnification, the Gaussian image of a small object lying at a large distance from the
imaging system is an exact replica of the object (Theorem 3). For systems with large
fields of view, the irradiance actually decreases according to the fourth power of the
cosine of the chief ray angle (i.e., the angle of the ray passing through the center of the
aperture stop, and therefore through the centers of the entrance and exit pupils), as
discussed in Section 2.6 of Part I. There is an additional decrease beyond a certain angle
due to vignetting of the rays by one or more elements of the system, resulting in an
effective aperture with a reduced area compared to that for an on-axis point object (see
Section 2.2.4 of Part I). Although the aberrated image according to geometrical optics can
be calculated, we will not do so, since what is observed in practice is determined by
diffraction, which we discuss next.
surface S passing through the center of the exit pupil with its center of curvature at the
Gaussian image point Pg is called the Gaussian reference sphere. A spherical wave
emerging from the exit pupil implies that the object rays transmitted by the system travel
equal optical path lengths when propagating from the object point to the reference sphere.
It should be evident that the reason for considering diffraction of the object radiation
at the exit pupil is that it represents the limiting aperture of the ray bundle converging to
the image. However, for point objects with sufficiently large field angles, different
portions of a ray bundle may be limited by different apertures (including imaging
elements) of the system. In such cases, the complex amplitude in the image plane may be
determined by considering diffraction at each of the limiting apertures. In practice,
however, the shape (perimeter) of the exit pupil is determined in the plane of the axial
exit pupil and the image is determined by considering diffraction at it.
If the optical path lengths of the rays are not equal to each other, a distorted spherical
wave emerges from the exit pupil and the system is said to be aberrated. If rays from a
point object are traced through the system so that they travel an optical path length equal
to that of the chief ray up to the exit pupil, the surface passing through their end points is
called the system wavefront for that point object. The optical deviation (i.e., the
ExP
Pi (ri; zi)
Pig
Pg(rg; zg)
R
O l s
Pp (rp; 0) W
S
Defocused Gaussian
image plane image plane
zi
zg
geometrical deviation times the refractive index ni of the image space) of the wavefront
from the Gaussian reference sphere along a ray is called the wave aberration. It
represents the difference between the optical path lengths of the ray under consideration
and the chief ray (passing through the center of the exit pupil) in propagating from the
point object to the reference sphere. Thus, the wave aberration of a ray is numerically
positive if it travels a longer optical path length compared to the chief ray in reaching the
reference sphere. Since the optical path lengths of the rays from the reference sphere to
the Gaussian image point are equal, the wave aberration of a ray is also equal to the
difference between its optical path length and that of the chief ray in propagating from the
point object Po to the Gaussian image point Pg .
r (r r )
For an aberrated system, let W rp ; ro be the wave aberration of a ray passing
r
through a point rp in the plane of the exit pupil. The complex amplitude at rp due to an
r r
object element ro centered at ro may be written
(r r ) (r r )
Uex rp ; ro = P rp ; ro exp ( iks) , (1-36)
where
xp
Pp(rp)
p xi
rp
Pi (ri)
Op
Pig ri i xg
an il
pl up
e
P
zi
R Pg(rg)
yp
Oi g
rg
e ion
an at
pl erv
zg
bs
O
Og
yi
z
pl n
e
e sia
an
ag us
im Ga
yg
Figure 1-4. Right-hand coordinate system in the pupil plane, and defocused and
Gaussian image planes. The optical axis of the system is along the z axis.
pg = p g and pi = p i .
1.4 Diffraction Image 17
r r2 12
s = zg2 + rp rg
(1-37a)
1 r r 2 1 r r 4
= zg + rp rg rp rg + .... (1-37b)
2 zg 8zg3
1 1 2
= zg + ( )
r 2 + rg2 - 2r p rg cos q pg - 3 r p2 + rg2 - 2r p rg cos q pg
2z g p 8z g
( ) + ... (1-37c)
is the distance between the pupil point Pp and the Gaussian image point Pg (see Figure
1-4) and
(r r ) (r r ) [
P rp ; ro = A rp ; ro exp ikW rp ; ro (r r )] , inside the exit pupil
= 0 , outside the exit pupil , (1-38)
r r
is called the pupil function of the system. The amplitude function A rp ; ro , called the ( )
apodization function, represents the variation of amplitude of the wave across the exit
pupil and accounts for any nonuniform transmission of the system. The inverse-square-
law dependence of irradiance on the distance s is contained in the amplitude function. It
r r
should be evident that rp = rp , rg = rg , and pg is the angle between the position
r r r
( )
12
vectors rp and rg . Note that s = R = zg2 + rg2 when rp = 0 .
r
The irradiance at a point rp in the plane of the exit pupil due to the object element
under consideration is given by
(r r ) (r r )
2
I p rp ; ro = Uex rp ; ro (1-39a)
(r r )
2
= P rp ; ro (1-39b)
(r r )
= A 2 rp ; ro . (1-39c)
The corresponding power in the exit pupil is obtained by integrating the irradiance across
the pupil, i.e.,
r r r r
P (r ; r )
2
Pex ( ro ) = p o d rp (1-40a)
r r r
=
A ( rp ; ro ) d rp
2
. (1-40b)
r r r r r
Ui ( ri ; ro ; zi ) =
i
zi
( ) [
P rp ; ro exp ik (l s) d rp ] , (1-41)
where
r r 2 1 2
l = zi2 + rp ri (1-42a)
1 r r 2 1 r r 4
= zi + rp ri rp ri +K (1-42b)
2 zi 8zi3
1 1 2
= zi +
2z i
(
r p2 + ri2 - 2r p ri cos q pi - 3
8z i
) (r 2
p + ri2 - 2r p ri cos q pi ) + ... (1-42c)
r
is the distance of the observation point Pi from a pupil point Pp , ri = ri and pi is the
r r
angle between the position vectors rp and ri (see Figure 1-4). We may add that if the
obliquity factor is not close to unity, it may be replaced by z R for a better
approximation.
1Ê 1 1 ˆ 4 1 3 Ê ri rg ˆ
- Á - ˜ r p + r p Á cos q pi - cos q pg ˜
8 Ë z i3 z g3 ¯ 2 Ë z i3 z g3 ¯
1 2 Ê r2
i 2
rg2 2
ˆ 1 2 Ê r 2 rg2 ˆ
- r Á 3 cos q - cos q pg ˜ - r p Á i3 - 3 ˜
2 p z
Ë i
pi
z 3
g ¯ 4 Ë zi zg ¯
1 Ê r3
i
rg3 ˆ
+ r Á 3 cos q - 3 cos q pg ˜ + ... . (1-43)
2 p Ë zi
pi
zg ¯
Each term on the right-hand side of Eq. (1-43) depends on the location of the
observation and Gaussian image planes and represents a phase term when multiplied by k.
The first term represents a piston aberration. It is simply the distance between the two
planes. Similarly, the second term represents a quadratic phase factor. These two terms do
not affect the irradiance distribution in the defocused image plane. Hence, they have no
impact on the image of an incoherent extended object since it is linear in irradiance. The
third term, when exponentiated, forms a Fourier kernel in Eq. (1-41). The quantity
r
( )
zi zg rg in this term represents the position vector of a point Pig where the line joining
the center Op of the exit pupil and the Gaussian image point Pg intersects the defocused
image plane (see Figure 1-4). The fourth term represents the usual defocus aberration due
to the observation plane being different from the Gaussian image plane. The fifth term is
1.4 Diffraction Image 19
a higher-order piston aberration, which also does not affect the irradiance distribution in
the observation plane. The sixth term has the form of a primary spherical aberration, as
may be seen by comparing it with Eq. (3-34) of Part I. The next four terms have two
components each, but only one of them represents a classical aberration. The other
component in each case is only deceptively similar to a classical aberration, but it
depends on the coordinates of the observation point (rather than the Gaussian image
point). For example, the component varying as rp3ri cos pi may be called axial coma in a
manner somewhat similar to the axial coma of a perturbed system (see Section 7.2.1 of
Part I), and the component varying as rp3rg cos pg may be called Seidel coma.
The quantity
r r r 1 r r z r
( ) zi
Wd rp ; ri ; rg = l s + rp ri i rg
zg
(1-44)
is the aberration of the diffracted wave.6,7 It should be noted, however, that this aberration
is related to the propagation of the wave and has nothing to do with the characteristics of
the imaging system (other than its dependence on the location of the Gaussian image
point). It would be convenient to call the last five terms of Eq. (1-43) spherical aberration,
coma, astigmatism, field curvature, and distortion of the diffracted wave, respectively.
The aberrations that are radially symmetric in pupil coordinates vanish when the
observation plane coincides with the Gaussian image plane, i.e., when zi = zg . Hence,
they are aberrations of the diffracted wave associated with defocus. Only the first four
terms are retained in the Fresnel approximation, and only the first three are retained in the
Fraunhofer approximation. These approximations limit the region of observation space
that accurately satisfies the Fourier-transform relationship to near the optical axis and
away from the pupil. When the observation plane coincides with the Gaussian image
plane, the Fraunhofer region of diffraction reduces to the Gaussian image plane.
In practice, the aberrations of the diffracted waves are negligibly small. 8,9 For
example, the depth of focus will be determined by the amount of defocus aberration that
can be tolerated. For systems with large Fresnel numbers, e.g., photographic systems, the
depth of focus for 8 defocus aberration tolerance is F 2 . Here, F = zg D is the focal
ratio of the image-forming light cone, where D is the diameter of the exit pupil. In that
case, the corresponding spherical aberration term in Eq. (1-43) is approximately equal to
( )
3 128 F 2 , which is negligibly small. The spherical aberration is also small for a
system with a small Fresnel number. This may be seen by considering, as discussed in
Section 2.5, a laser transmitter with an aperture diameter of 25 cm focusing a beam of
wavelength 10.6 mm at a distance of 1.47 km so that its Fresnel number is unity. Based on
the defocus aberration, a Strehl ratio of 0.8 is obtained at a distance of 3 km, showing a
large depth of focus. The magnitude of the spherical aberration term in this case is only
about 10 -2 m . For the angle-dependent terms, such as coma, it is important to note that
the region of interest in the image of a point object is where there is a significant
r
( )
illumination. This region corresponds to ri , pi values that lie, say, within the Airy disc
20 IMAGE FORMATION
(discussed in Section 2.1) that is centered at the Gaussian image point. Thus, even for
rg >> F , we have ri
rg and pi differs from gi by no more than approximately D.
Hence, the coma term for a field angle = rg zg is approximately equal to 16 F 2 , ( )
which is also negligibly small. The value of the axial coma is equal to 16 F 2 .
Similarly, the astigmatism, field curvature, and distortion terms are approximately equal
( ) ( ) ( )
to 2 8F , 3 2 16 , and 3 4 , respectively. Their axial values are extremely
( ) ( )
small and equal to 8zg , 16zg , and ( 2 D) . Thus, the aberrations of a
2
diffracted wave are negligibly small. The expressions for their values are summarized in
Table 1-1. It should be noted that in high-quality imaging systems, the value of is often
quite small (few degrees or less) since it is limited by the aberrations of the system itself.
We may neglect all terms in Eq. (1-43) beyond the defocus term if
1 1 4
3 3 a < 8 , (1-45)
zi zg
where a is the maximum value of rp , i.e., if the peak aberration contributed by the
primary spherical aberration is < / 8 . We will see, for example, in Section 2.3 that such
an aberration has a negligible effect on the quality of an image. Hence, Eq. (1-41) for the
complex amplitude may be written
r r 1 r 2 rg
2
Ui ( ri ; ro ; zi ) =
i
zi
( )
exp ik zi zg + i
2 zi zg
r r 2 i r r z r r
× ( )
P rp ; ro ; zi exp
zi
rp ri i rg d rp
zg
, (1-46)
where
Table 1-1. Aberrations of a diffracted wave for an on-axis and an off-axis point
object. The field angle of the image point is indicated by the angle .
Defocus /8 /8
Spherical (3 128 F ) 2
(3 128 F ) 2
Coma 16 F 2 ( 16 F ) 2
Distortion ( 2 D)2 ( 4)
3
1.4 Diffraction Image 21
r r r r ik 1 1
(
P rp ; ro ; zi ) ( )
= P rp ; ro exp rp2
2 zi zg
(1-47)
is the defocused pupil function signifying that the image is observed in a plane other than
the Gaussian image plane, i.e., zi ! zg .
So far we have been concerned with the phase variations across the exit pupil for a
wave from a given point object and ignored the optical path length of the chief ray from
the object point to its Gaussian image point. There is an additional phase factor that needs
to be considered when comparing the waves from different point objects. It represents the
difference in the optical path lengths of their chief rays from the object points to their
Gaussian image points. The difference in the optical path lengths of a chief ray for an
r
object point located at ro at a distance zo from the entrance pupil and that for an axial
point object, up to the quadratic approximation (see Figure 1-2), is given by
(r r )
W ro ; rg = ( Po Oen + z ) Oex Pg zg ( )
2 r2
~ ro + g , (1-48)
2 zo 2 zg
where zo is numerically negative according to our sign convention given in Section 1.3.2
of Part I. It is assumed that the optical path lengths of the chief rays from the entrance
pupil to the exit pupil are the same. This must, of course, be verified for the system under
consideration by ray tracing, or the phase must be adjusted if the assumption is not valid.
Equation (1-48) is evidently valid in the case of a thin lens when its aperture stop and,
therefore, its entrance and exit pupils are located at the lens. Adding the phase
contribution from Eq. (1-48), Eq. (1-46) is modified to
r r 1 r 2 r 2
Ui ( ri ; ro ; zi ) =
i
zi
( )
exp ik zi zg + i o
2 zi zo
r r 2 i r r z r r
× (
P rp ; ro ; zi exp
)z
rp ri i rg d rp
zg
. (1-49)
i
The additional phase factor does not impact imaging of an incoherent object, but does
impact coherent imaging, which is discussed in Section 1.12.
In obtaining Eq. (1-49), we have in effect propagated the wave from the exit pupil to
the observation plane in the Fresnel approximation, as may be seen by substituting Eq. (1-
36) into Eq. (1-22a) and retaining only the first two terms on the right-hand side of Eq.
(1-37b). Indeed, for a thin lens of focal length f, Eq. (1-49) can also be obtained by
considering Fresnel propagation of the spherical wave from a point object to the entrance
pupil, its modification by the lens, i.e., introduction of the quadratic phase factor
( )
exp ikrp2 2 f , and Fresnel propagation to an observation plane (see Problem 4).
22 IMAGE FORMATION
r r r
= Pex (ro ) PSF ( ri ; ro ; zi ) , (1-50)
where
2
rr r r 2 i r r z r r
PSF ( ri ; ro ; zi ) = r
1
Pex ( ro )2 zi2
( )
P rp ; ro ; zi exp
zi
rp ri i M ro d rp
zg
(1-51)
is called the incoherent point-spread function of the imaging system, as we shall see in
the next section. It is proportional to the modulus square of the inverse Fourier transform
r
of the defocused pupil function of the system (Theorem 4). Hence, substituting for Pex (ro )
from Eq. (1-32) into Eq. (1-50) and integrating over the object, we obtain
r rr
Ii ( ri ; zi ) = " I ( ri ; ro ; zi )
r r r r
(
= # Sen zo2 ) " B ( ro ) PSF ( ri ; ro ; zi ) d ro
object
, (1-52)
where we have neglected the dependence of Sen and the solid angle it subtends at an
object element on the location of the element. Thus, the object is assumed to be small, far
1.4 Diffraction Image 23
from the system, and at a small angle from the axis of the system so that the cosine-to-
the-fourth-power dependence on the chief ray angle in the object space is negligible, and
vignetting of its pupil for the off-axis object elements (discussed in Section 2.6 of Part I)
is either zero or negligible.
r (r r
)
The pupil function P rp ; ro ; zi represents the actual complex amplitude at a point
r r
rp in the plane of the exit pupil due to an object element ro centered at ro . Evidently, it
r
depends on the location ro of the object element. We now introduce a relative pupil
r r
( )
function G rp ; ro ; zi that represents the complex amplitude per unit amplitude of the
object element. The two pupil functions are related to each other according to
(r r
P rp ; ro ; zi ) r
(r
= U ( ro ) G rp ; ro ; zi
r
) , (1-53)
r r
where U ( ro ) is the amplitude at a point ro . It should be evident that the two pupil
r
functions have identical dependence on rp . Substituting Eq. (1-53) into Eq. (1-40a), we
obtain
r r r r r
G (r ; r ; z )
2
Pex ( ro ) = U ( ro )
2
p o i d rp . (1-54)
For a small object, we may neglect any changes in the nonuniform transmission or
r r r
aberrations introduced by the system as a function of ro . Hence, G rp ; ro ; zi may be
r ( )
assumed to be independent of the position vector ro . [The assumptions of a small field of
view and a small object are the same as those made in obtaining the Gaussian image
distribution of Eq. (1-35).] An object of small extent for which the relative pupil function
r r
( )
G rp ; ro ; zi may be considered independent of the position of a point on it is called
isoplanatic (or the system is said to be isoplanatic for the small object under
r r r
( )
consideration). Of course, the pupil function P rp ; ro ; zi will vary as U ( ro ) varies with
r
ro for such an object.
Substituting Eqs. (1-53) and (1-54) into Eq. (1-51), neglecting the dependence of
r r r
( )
G rp ; ro ; zi on ro , and letting zi ~ zg ~ R , for an isoplanatic object imaged by a system
with a small field of view, we obtain
24 IMAGE FORMATION
r 2 i r r r r
2
r r
( )
G rp ; zi exp R rp
( ri M ro ) d rp
PSF ( ri ; ro ; zi ) = r r . (1-55)
( )
2
2 R 2 G rp ; zi d rp
We note that the integral in the numerator on the right-hand side of Eq. (1-55) depends on
r r r
the position vectors ri and rg (which is equal to M ro ) of the observation and Gaussian
r r
image points, respectively, through their difference ri rg only, and we may replace
r r r r
PSF ( ri ; ro ; zi ) by PSF ( ri M ro ; zi ) . Thus, the PSF is shift invariant in the sense that its
form does not change as the object point is shifted; only its location changes by virtue of
r
it being centered at rg . Accordingly, Eq. (1-52) for the irradiance distribution of the image
of an isoplanatic incoherent object may be written
r r r r r
(
Ii ( ri ; zi ) = Sen zo2 ) B ( ro ) PSF ( ri M ro ; zi ) d ro
object
, (1-56a)
(
= Sen zo2 M 2 ) B (rr M ) PSF (rr rr ; z ) d rr
g i g i g (1-56b)
r r r r
= Ig ( rg ) PSF ( ri )
rg ; zi d rg , (1-56c)
where we have made use of Eq. (1-35). Thus, the diffraction image of an isoplanatic
incoherent object is equal to the convolution of its Gaussian image (which is a scaled
replica of the object) and the PSF of the imaging system (Theorem 5). It should be
evident that the magnification M of the image is assumed to be constant across the
isoplanatic region of the object. Of course, a nonuniform magnification will result in a
distortion of the image also.
Multiplying the numerator and denominator of the right-hand side of Eq. (1-55) by
r
U ( ro ) , we may also write the PSF in the form
2
2
r Û r Ê 2pi r r ˆ r
PSF (ri ; z i ) =
1
2 2 Ù P rp ; z i exp Á -
Pex l R ı Ë lR
( )
rp ri ˜ d rp
¯
◊ , (1-57)
r r
where we have suppressed any explicit dependence on ro , and the position vector ri is
r
written with respect to rg . The irradiance distribution of the image of a particular object
element is obtained by multiplying the PSF by Pex for that element.
r r
where ( ) is a Dirac delta function. Its Gaussian image point is located at rg = M rj .
1.5 Physical Significance of Incoherent PSF 25
Substituting Eq. (1-58) into Eq. (1-52), we obtain the irradiance distribution of its image,
namely,
(r r
Ii ri ; rj ; zi ) (r
= Pex PSF ri ; rj ; zi
r
) , (1-59a)
where
(
Pex = Sen zo2 B j ) (1-59b)
represents the total power from the point object in the exit pupil of the imaging system.
Substituting Eq. (1-51) into Eq. (1-59a), we obtain
2
rr r r
2 i r r z r
r
(
Ii ri ; rj ; zi ) 1
= 2 2
zi
(
P rp ; rj ; zi exp
zi
)
rp ri i M rj d rp
zg
. (1-60)
For an isoplanatic object, we substitute Eq. (1-58) into Eq. (1-56a) and obtain
(r r
Ii ri ; rj ; zi ) (r
= Pex PSF ri M rj ; zi
r
) , (1-61)
rr 1 r
2 i r r r
r
2
(
Ii ri ; rj ; zi ) R
(
= 2 2 P rp ; zi exp
R
rp ) ( ri M rj d rp
) . (1-62)
It is evident from Eq. (1-62) that the images of two point objects lying within an
isoplanatic patch are identical except that they are centered at their respective Gaussian
image points. Such an imaging is referred to as shift-invariant imaging, and comes about
because of the invariance of the relative pupil function with the position of an object
point on the isoplanatic object. Integrating both sides of Eq. (1-62), we find that
r r r 1 r r
2 i r r r r
2
(
i i j i R
)
I r ; r ; z d ri = 2 2 d ri P rp ; zi exp
R
rp ( ) ( )
ri M rj d rp
r r r r r r
= d rp P ( rp ; zi ) P * ( rp) ( rp rp ) d rp
r r
P (r ; z )
2
= p i d rp
= Pex . (1-63)
We obtain the same result if we integrate both sides of Eq. (1-61). It represents the
conservation of power; i.e., the total power in the image is equal to the total power in the
r r r
exit pupil. It also shows that the integral of PSF ri ; rj ; zi over all values of ri in the ( )
image plane is unity. Hence, the PSF given by Eq. (1-51) or Eq. (1-57) represents the
r
irradiance distribution of the defocused image of a point object located at ro per total
26 IMAGE FORMATION
power in the image. It is called the diffraction point-spread function (PSF) of the system.
Note that its dimensions are m 2 .
r 2 i r r
r
2
Ii ( ri ; zi ) =
I0
2 R 2
r
[ ( )]
exp i kW rp ; zi exp
R
rp ri d rp
, (1-64)
where I0 = A02 is the irradiance at the exit pupil. If, in addition, the pupil is aberration
free and the observation is made in the Gaussian image plane so that there is no defocus
r
( )
aberration, i.e., if W rp ; zg = 0, then Eq. (1-64) reduces to
r 2 i r r
r
2
Ii ( 0 )
Ii ( ri ) = 2 exp R rp ri d rp
, (1-65)
Sex
where
Pex Sex
Ii ( 0 ) = (1-66)
2 R 2
r
is the aberration-free irradiance at the Gaussian image point ri = 0 . Of course,
Pex = I0 Sex in this case. The effect of aberrations and or amplitude variations on the
central irradiance is discussed in Section 1.9.
yi yg
Pi Pi
ri ri
Pig Pg
xi xg
Oi Og
r
Figure 1-5. Redefinition of ri as the position vector of an image point Pi in the (a)
defocused image plane with respect to Pig and the (b) Gaussian image plane with
respect to Pg .
1.6 Optical Transfer Function (OTF) 27
Theorem 7. The OTF of an imaging system is also equal to the autocorrelation of its
pupil function.
Theorem 8. The PSF and the OTF of a system with a radially symmetric pupil
function form a zero-order Hankel transform pair.
A relationship between the spatial-frequency spectra of the object and image may be
obtained by taking a Fourier transform of both sides of Eq. (1-56c). Let
r r r r r
B˜ ( vo ) = B ( ro ) exp (2 i vo ro ) d ro (1-67)
and
r r r r r
I˜i ( vi ) = Ii ( ri ) exp (2 i vi ri ) d ri (1-68)
be the spectral components of the object and image corresponding to the object and image
r r
spatial frequencies vo and vi , respectively, where
r r
vi = vo M . (1-69)
28 IMAGE FORMATION
Substituting Eq. (1-56c) into Eq. (1-68) and suppressing the zi dependence of the PSF
and the OTF, we obtain
r r r r r r r r
I˜i ( vi ) = d ri exp (2 i vi ri ) Ig ( rg ) PSF ( ri rg ) d rg
r r r r r r r r r r
= d rg Ig (rg ) exp (2i vi rg ) PSF (ri rg ) exp [2 vi (ri rg )] d ri
r r
= I˜g ( vi ) ( vi ) , (1-70)
r
where Ĩg ( vi ) is the spectral component of the Gaussian image corresponding to a spatial
r
frequency vi and
r r r r r
( vi ) = PSF ( ri ) exp (2 i vi ri ) d ri (1-71)
is called the optical transfer function (OTF) of the imaging system at this frequency.
Equation (1-71) states that the OTF, which is dimensionless, is the Fourier transform of
the PSF. From Eq. (1-35), it is evident that
r r
(
I˜g ( vi ) = Sen zo2 B˜ ( M vi ) ) (1-72a)
r
(
= Sen zo2 B˜ ( vo ) . ) (1-72b)
r
Equation (1-70) states that the spatial-frequency spectrum Ĩi ( vi ) of the diffraction image
r
of an isoplanatic incoherent object is equal to the product of the spectrum Ĩg ( vi ) of its
Gaussian image (which in turn is equal to a scaled spectrum of the object) and the OTF
r
( vi ) of the imaging system (Theorem 6). It is a consequence of the fact that if a function
is equal to the convolution of two functions, as in Eq. (1-56c), its Fourier transform is
equal to the product of their Fourier transforms.10
Substituting Eq. (1-57) into Eq. (1-71), the OTF can also be written
r r r r r 2i r r
r
2
( vi ) =
1
Pex 2 R 2
( )
d ri exp (2 i vi ri ) P rp exp
R
rp ri d rp
r r
=
1
Pex 2 R 2
d rp P rp ( ) d rrp P * (rrp) d rri exp
2Ri rri (rrp rrp + R vri )
r
= Pex 1 d rp P rp (r ) d rrp P * (rrp) (rrp rrp + R vri )
r r r r
= Pex 1 P ( rp ) P ( rp R vi ) d rp (1-73a)
r r r r r 2 r
P ( rp ) P * ( rp R vi ) d rp P ( rp ) d rp
=
(1-73b)
r r r
( ) (
= Pex 1 A rp A rp R vi exp iQ rp ; zi d rp ) [ (r )] r
, (1-73c)
1.6 Optical Transfer Function (OTF) 29
(r )
where the pupil function P rp is given by Eq. (1-38) when the image is observed in the
Gaussian image plane and by Eq. (1-47) when it is observed in a defocused image plane,
and
(r r
Q rp ; vi ; zi ) (r ) (r
= rp ; zi rp R vi ; zi
r
) (1-74)
is a phase aberration difference function defined over the region of overlap of two pupils,
r r r
one centered at rp = 0 and the other at rp = Rvi . The phase aberration is related to
the wave aberration W according to
= (2 ) W . (1-75)
Equation (1-73b) shows that the OTF of an incoherent imaging system is also equal to the
normalized autocorrelation of its pupil function (Theorem 7). [The autocorrelation of a
function f ( x ) is equal to its convolution with the function f * ( x ) .] The region of
integration in the numerator of the right-hand side of this equation is the overlap region of
r
the two pupils. We may note here that the region of overlap is maximum for vi = 0,
giving a value of unity for (0) . Because of the finite size of the pupil, the overlap region
r
reduces to zero at some frequency vc , called the cutoff frequency, and stays zero for
r r r
larger frequencies; i.e., ( vi ) = 0 for vi vc . If the image is observed in a defocused
image plane such that zi is significantly different from R, then R in Eq. (1-73) is replaced
by zi , as may be seen by using the PSF obtained from Eq. (1-60). Problem 4.2 is an
example of this type, where the defocused OTF of a weakly truncated Gaussian pupil is
considered.
r r
which shows simply that the functions PSF( ri ) and ( vi ) form a 2D Fourier transform
pair. The value of the PSF at the origin may be written
r r
PSF (0) = ( vi ) d vi (1-77a)
r r
= Re ( vi ) d vi , (1-77b)
r r
where Re ( vi ) is the real part of ( vi ) . Since the PSF at any point is a real quantity,
only the real part of the complex OTF contributes to the integral. The integral of its
imaginary part must be zero.
Consider a system with a radially symmetric pupil function, i.e., one for which
30 IMAGE FORMATION
(r )
P rp = P rp ( ) . (1-78)
If we let
r
( )
rp = rp cos p , sin p , 0 p < 2 , (1-79a)
and
r
ri = ri (cos i , sin i ) , 0 i < 2 , (1-79b)
Noting that11
2
exp (ix cos ) d = 2 J0 ( x ) , (1-81)
0
where J 0 ( ) is the zero-order Bessel function of the first kind, Eq. (1-80) reduces to
( ) P ( r ) J (2 r r )
2
PSF(ri ) = 4 2 Pex 2 R 2 p 0 p i R rp drp . (1-82)
Thus, the PSF of a system with a radially symmetric pupil function is also radially
symmetric. Multiplying both sides of Eq. (1-82) by Pex , we obtain the radially symmetric
irradiance distribution of the image of a point object. Similarly, Eqs. (1-71) and (1-76)
reduce to
and
r
where vi = vi . Thus, the PSF and the OTF of a system with a radially symmetric pupil
function form a zero-order Hankel transform pair (Theorem 8). This result is a
consequence of the fact that the Fourier transform of a radially symmetric function is a
zero-order Hankel transform. From Eq. (1-83), we note that, since the PSF is a real
function, the radially symmetric OTF is also a real function.
By multiplying PSF(ri ) with Pex in Eqs. (1-83) and (1-84) we obtain relationships
between the irradiance distribution of the image of a point object and the OTF of a system
with a radially symmetric pupil function. For example, the irradiance distribution may be
obtained from the OTF according to
1.6 Optical Transfer Function (OTF) 31
The encircled power of the image, i.e., the power contained in a circle of radius rc
centered at the Gaussian image point, is given by
r r
Pi (rc ) = I ( r ) d ri
rr r i i
. (1-86)
i c
Or, substituting Eq. (1-85) into Eq. (1-87) and noting that11
z0
where J1 ( ) is a first-order Bessel function of the first kind, we obtain
Equation (1-89) gives the encircled power of the image in terms of the radially symmetric
OTF of the system.
To understand the physical significance of the OTF defined by Eq. (1-71), we derive
the following theorem.
Bo(1+ m)
1:.0
iil
t BO(1-m)
0
15
_0>
t 10(1-m)
0
Ic: »1 -- Xj
1/1vj I ~W~
1 ~ 10(1+m l't Il
t 10(1- m l't Il
Figure 1-6. Image of a sinusoidal object shown along the x axis. (a) Object, (b)
Gaussian image, and (c) Diffraction image.
(1-91)
where
(1-92)
is the average irradiance of the image. We note that the image is also sinusoidal but with
a frequency vi = Va / M, where M is its magnification. Its modulation and phase are the
same as that of the object, respectively.
Substituting Eq. (1-90) into Eq. (1-56a), we obtain the irradiance distribution of the
diffraction image, which may be written
(
= 1 Pex 2 R 2 ) d rr P (rr ) d rr P * (rr )
p p p p
r
2 i r r
× d ro exp
R
rp rp ( ) (rri M rro )
(
= 1 Pex M 2 ) d rr P (rr ) d rr P(rr ) (rr rr )
p p p p p p
( ) P (rr ) r
2
= 1 Pex M 2 p d rp
= 1 M2 . (1-94)
Following Eq. (1-71), the second integral on the right-hand side of Eq. (1-93) may be
written
r r r r r r r
PSF ( ri M ro ) exp [2 i vi ( ri M ro )] d ro = M 2 ( vi ) . (1-95)
Thus, like the Gaussian image, the diffraction image of a sinusoidal object of spatial
r r r
frequency vo is also sinusoidal with a spatial frequency vi = vo M , and its average
r
irradiance is Io . However, the modulation of this image is m ( vi ) ; i.e., its modulation
r
is different from that of the object or its Gaussian image by the MTF factor ( vi ) . It is
r
shown later [see Eq. (1-106)] that ( vi ) 1. Thus, the modulation of the diffraction
image is always less than or equal to that of the object at any spatial frequency. The phase
of the sinusoidal image is also different from that of the object or its Gaussian image by
r r
the PTF ( vi ) . Hence, the OTF ( vi ) of an incoherent imaging system corresponding to
r
an image spatial frequency vi is a complex function whose modulus when multiplied by
r
the modulation of the corresponding sinusoidal object (of spatial frequency vo ) gives the
modulation of the sinusoidal image, and whose phase gives the phase of the diffraction
image relative to that of the object or its Gaussian image, as illustrated in Figure 1-6
r r r r
(Theorem 9). Note that vo ro = vi ri .
1.6.3 Properties of OTF
Now we derive certain properties of the OTF from its definition.
Theorem 10. (a) The OTF is a complex symmetric function with a value of unity at
the origin. (b) Its magnitude, the MTF, is unity or less at any other spatial frequency. (c)
The aberrated MTF at any frequency is less than or equal to the corresponding
aberration-free value.
34 IMAGE FORMATION
From Eq. (1-73b), we note that its value at the origin is unity, i.e.,
(0) = 1 . (1-98)
as may be seen from Eqs. (1-61) and (1-63). Equation (1-98) represents the fact that the
OTF at zero spatial frequency is unity, i.e., the contrast of an image is zero for an object
of zero contrast. From Eq. (1-71), we also note that
r r
( vi ) = * ( vi ) , (1-100)
i.e., the OTF is complex symmetric or Hermitian (Theorem 10a). Therefore, its real part
is even and its imaginary part is odd; i.e.,
r r
Re ( vi ) = Re ( vi ) ,
(1-101)
and
r r
Im ( vi ) = Im ( vi ) . (1-102)
r r
For two arbitrary but well-behaved functions f ( r ) and g( r ) , Hölder’s inequality12
states that
r r r r r 1n
r n ( n 1) r ( n 1) n
f ( r ) g( r ) d r
f ( r ) d r
g( r ) dr
n
. (1-104)
r r r r r r 12
( vi ) Pex 1
( ) ( ) d rp
2 2
P rp d rp P rp R vi . (1-105)
Thus, the MTF at any spatial frequency is less than or equal to unity (Theorem 10b).
(r ) (r ) (r
Using Hölder’s inequality with f rp = P rp P * rp R vi , g( r ) = 1 , and n = 1,
r
) r
Eq. (1-103) yields
1.6 Optical Transfer Function (OTF) 35
r r r r r
( vi ) Pex 1 P (rp )P * (rp R vi ) d rp
r r r r
= Pex 1 P ( rp ) P * ( rp R vi ) d rp
r r r r
= Pex 1 A( rp ) A( rp Rvi ) d rp
r
= ( vi ) w 0
; (1-107)
i.e., the aberrated MTF at any spatial frequency is less than or equal to the corresponding
aberration-free MTF (Theorem 10c). The ratio of the aberrated and aberration-free MTFs
at a certain spatial frequency is called the Hopkins ratio for that frequency, as discussed
later in Section 1.10. Equation (1-107) shows that the Hopkins ratio is less than or equal
to one.
The MTF of an aberration-free system with a nonuniform amplitude across its exit
pupil can be higher or lower than that for a uniform amplitude, depending on the spatial
frequency and the nautre of the amplitude distribution. For example, as will be discussed
in Section 4.3.3.2, the MTF of a Gaussian pupil is higher for low frequencies and lower
for high frequencies. However, if the amplitude increases from the center of the pupil
toward its edge, then the MTF is lower (higher) for low (high) frequencies.
We now summarize the results represented by Eqs. (1-98), (1-100), (1-106), and (1-
107) (Theorem 10).
i. OTF at the origin is unity.
ii. OTF is complex symmetric or Hermitian.
iii. MTF at any spatial frequency is less than or equal to unity.
iv. Aberrated MTF at any spatial frequency is less than or equal to the
corresponding aberration-free MTF; i.e., the Hopkins ratio is less than or
equal to one.
( )
12
vi = 2 + 2 (1-108)
and
= tan 1 ( ) . (1-109)
Thus, vi is the magnitude of the spatial frequency of a periodic image and gives its
orientation, corresponding to rectangular frequency components and . It should be
evident that the OTF corresponding to a 2D periodic image of spatial frequencies and
36 IMAGE FORMATION
Following Eq. (1-73a), we write the OTF corresponding to a spatial frequency (, )
in the form
( ) (
(, ) = Pex 1 P x p , y p P * x p R, y p R dx p dy p ) , (1-110)
r
( )
where x p , y p are the rectangular coordinates of a pupil point with a position vector rp .
To determine the OTF slope at the origin, we write the integral in Eq. (1-110) in a (U, V )
(
coordinate system whose origin is shifted from the x p , y p coordinates system by )
( )
( R 2)(, ) but whose axes are parallel to the x p , y p axes, as indicated in Figure 1-7.
Thus, letting
xp = U + R 2 (1-111a)
and
yp = V + R 2 , (1-111b)
Equation (1-38) [or Eq. (1-47) with zi dependence suppressed] for the pupil function may
be written
yi
v
q
i
i 1 p
(R/2)(,)
yp
xi
u
xp
(–R/2)(,)
)
-m
I o(1 R
)
-m
I o(1
(a) (b)
Figure 1-7. (a) Sinusoidal object yielding image of spatial frequency (vi , ) or (, ) ,
mean irradiance I0 , and modulation m . (b) Geometry for evaluating the OTF of
an optical imaging system. The centers of the two pupils are located at (0, 0) and
( )
R ( , ) in the x p , y p coordinate system, m ( R 2) (, ) in the (U , V ) coordinate
system, and m ( R 2) (vi , 0) in the ( p, q ) coordinate system, where vi = 2 + 2 ( )
12
and = tan 1 ( ) . The shaded area is the overlap area of the two pupils.
1.6 Optical Transfer Function (OTF) 37
( ) ( ) [ (
P x p , y p = A x p , y p exp i x p , y p )] ,
(1-113)
[
P(U + R 2, V + R 2) = A(U + R 2, V + R 2) exp i(U + R 2, V + R 2) ] .
(1-114)
Substituting Eq. (1-114) into Eq. (1-112), the real part of the OTF may be written
[ ]
Re (, ) = Pex 1 I (U , V ; , ) cos Q(U , V ; , ) dU d V , (1-115)
where
and
Q(U , V ; , ) = (U + R 2 , V + R 2) (U R 2 , V R 2) . (1-117)
The region of integration in Eq. (1-115) is the overlap area of two pupils centered at
m ( R 2)(, ) in the (U , V ) coordinate system. Expanding (U ± R 2 , V ± R 2) in
a Taylor series about the point (U , V ) , e.g.,
(U , V ) (U , V )
(U + R 2 , V + R 2) = (U, V ) + ( R 2) +
U V
1
2 (U , V ) 2 (U , V ) 2 (U , V )
2
(1-118)
+ ( R 2 ) 2 2 + 2 + + ... ,
2! U 2
U V V 2
we find that
(U , V ) (U , V )
Q(U , V ; , ) = R +
U V
3
( R 2)3
+
(U, V ) + ... .
1
+
3! U V (1-119)
x2 x4
cos x = 1 + K , (1-120)
2! 4!
we may write
[ ( )]
Re (, ) = Pex 1 I (U , V ; , ) 1 + O 2 , 2 , dU d V . (1-121)
Thus, Re (, ) does not consist of any aberration-dependent terms that depend linearly
on or alone. Hence, we obtain
38 IMAGE FORMATION
Re (, ) = Pex 1 I (U , V ; , ) dU d V
, (1-122)
0+ 0+
and a similar equation for the derivative with respect to h. Thus, the derivative of the real
part of the OTF with respect to a spatial frequency component x or h evaluated at the
origin is independent of the aberration, irrespective of the shape of the pupil. It does,
however, depend on the amplitude variations across the pupil. If the amplitude is
uniform, then the integral on the right-hand side reduces to the power contained in the
overlap region of the two pupils whose centers are separated by R(, ) from each
other. When divided by the total power Pex , it yields the fractional overlap area of the
two pupils.
Re (, ) = Re ( , ) . (1-123)
(, ) [
= Re (, )] + [Im (, )]
2 2 2
(1-124)
[
P( p, q ) = A( p, q ) exp i( p, q ) ] , (1-126)
r
[ ]
cos Q( p, q; vi ) = 1
1
2
[
( Rvi )2 ( p, q) ]2 + O (vi4 ) . (1-131)
Taking the derivative of both sides of Eq. (1-132) with respect to vi and evaluating at
vi = 0, we obtain
r
Re (vi , ) = Pex 1 I ( p, q; vi ) dp dq . (1-133)
iv vi 0
iv vi 0
That is, the derivative of the real part of the transfer function with respect to the
radial frequency v i evaluated at the origin is independent of the aberration, irrespective
of the shape of the pupil. It does, however, depend on the amplitude variations across the
pupil. This result is to be expected from the component derivatives given by Eq. (1-122).
As discussed later in Section 1.7, the asymptotic behavior of the irradiance distribution of
the image of a point object depends on the value of this derivative. Once again, the
imaginary part of the transfer function depends on the sine of the aberration difference
40 IMAGE FORMATION
which, when expanded in a power series, will contain a linear term in vi . Consequently,
its derivative with respect to vi evaluated at vi = 0 will depend on the aberration.
However, the derivative of the modulus of the transfer function evaluated at vi = 0 is also
independent of the aberration. Indeed, the derivatives of the real part and the modulus of
the transfer function evaluated at vi = 0 are equal. This may be seen by evaluating the
derivative of both sides of = ( Re )2 + (Im )2 at vi = 0 , and noting from Eq. (1-
2
We summarize the results of this section in that the derivative of the real part of the
OTF of a system with respect to a spatial frequency component or or a radial
frequency vi evaluated at the origin is independent of its aberration irrespective of the
shape of its exit pupil. The value of this derivative is equal to the corresponding
derivative of the MTF of the system.
Theorem 12. The diffraction OTF approaches the geometrical OTF as the
wavelength approaches zero.
Substituting Eq. (1-114) into Eq. (1-112), the diffraction OTF can be written
{ }
(, ) = Pex 1 I (U , V ; , ) exp iQ(U , V ; , ) dU d V , (1-134)
where the region of integration is the overlap area of two pupils centered at
m( R 2)(, ) . Substituting Eq. (1-75) into Eq. (1-119), we obtain
W (U , V ) W (U , V )
Q(U , V ; , ) = 2 R
U
+
V
+ O 2 ( ) , (1-135)
( )
where O 2 consists of terms with third and higher odd-order derivatives of W (U , V ) .
Substituting Eq. (1-135) into Eq. (1-134) and letting 0, we obtain
W (U , V ) W (U , V )
!
[ (, )] 0 = Pex 1 A 2 (U , V ) exp 2 iR
U
+
V #
" dU d V , (1-136a)
or
[ (, )] 0 (
= Pex 1 I x p , y p )
$
W x p , y p
× exp 2i R +
( )
W x p , y p( )
!$ dx
" p d yp , (1-136b)
$ x p y p $
#
where
1.6 Optical Transfer Function (OTF) 41
( ) (
I x p , y p = A2 x p , y p ) (1-136c)
( )
is the irradiance at a pupil point x p , y p , and the region of integration is the area of the
exit pupil. Equation (1-136b) may also be obtained directly from Eq. (1-110) in a similar
manner. However, the equation corresponding to Eq. (1-135) in that case will consist of
terms with odd and even-order derivatives of W x p , y p . ( )
We show in the next section that the OTF obtained according to geometrical optics is
also given by Eq. (1-136b). Hence, it yields the result that the diffraction image reduces
to the geometrical image as the wavelength approaches zero (Theorem 12). It should be
noted that, in practice, the limit of zero wavelength is equivalent to a very large
aberration. This is why the diffraction PSF may be expected to be similar to the
geometrical PSF for large aberrations.
The geometrical PSF (not the Gaussian image) is given by [see Eq. (4-10) of Part I]
1
( x, y)
(
I g ( x, y) = I x p , y p ) , (1-137)
(
x p , yp )
( )
where x p , y p are the coordinates of a ray in the plane of the pupil and ( x, y) are its
coordinates in the image plane with respect to the Gaussian image point. Its Fourier
transform for unit total power gives the geometrical OTF:
[
g (, ) = Pex 1 Ig ( x, y) exp 2 i ( x + y) dx d y ] (1-138a)
1
( x, y)
= Pex 1 (
I x p , yp ) [ ]
exp 2 i ( x + y) dx d y
(
x p , yp )
( ) [
= Pex 1 I x p , y p exp 2 i ( x + y) dx p d y p ] . (1-138b)
Note that it is properly normalized since g (0, 0) = 1. Like the diffraction OTF, it is also
Hermitian.
Substituting for the ray aberrations ( x, y) in terms of the slope of the aberration
function [see Eq. (3-11) of Part I], namely,
42 IMAGE FORMATION
( x, y) = R
( ,
)
W x p , y p W x p , y p ( )
, (1-139)
x p y p
we obtain
$
W x p , y p ( )
W x p , y p ( )
!$ dx
g (, ) =
( )
Pex 1 I x p , y p exp 2 i R
x p
+
y p
"
$
p d yp ,
$ #
(1-140)
which is the same as Eq. (1-136b). The refractive index of the image space is assumed to
be unity here, which is generally the case in practice. If this is not true, then the right-
hand side of Eq. (1-139) is divided by the refractive index. Also in that case, the
wavelength of the object radiation is divided by the refractive index in all of the
diffraction equations.
To obtain an approximate expression for the geometrical OTF for small spatial
frequencies, we write Eq. (1-138b) in terms of the centroid ( xc , yc ) of the aberrated PSF
discussed later [see Eq. (1-177)] in the form
[
g (, ) = Pex 1 exp 2 i( xc + yc ) ]
( ) { [
× I x p , y p exp 2i ( x xc ) + ( y yc ) dx p d y p ]} . (1-141)
Expanding the exponential under the integral in a power series and retaining only the first
three terms for small values of (, ) (the second term vanishes upon integration), we
obtain
(
× I x p , yp ) {1 2 2 [( x xc ) + ( y yc )] 2 } dx p d y p . (1-142)
~ 1 2
2
g (, ) I ( x p , y p ) [( x xc ) + ( y yc )]
2
dx p d y p (1-144a)
Pex
and
We note that the geometrical PTF depends on the pupil irradiance only through the
centroid of the PSF. The PTF for a symmetric aberration is zero, as expected, since the
centroid for such an aberration lies at the origin.
It is evident from Eq. (1-144a) that the slope of the geometrical MTF at the origin is
zero:
g (, )
g (, )
= 0 = . (1-145)
0 0
This result can also be obtained directly from Eq. (1-138b). As shown in Section 1.6.4,
the slope of the diffraction MTF at the origin, which is equal to the corresponding slope
of the real part of the OTF, is nonzero.
If we write the wave and ray aberrations in the ( p, q ) coordinate system, Eq. (1-
138b) reduces to
( )
g (vi , ) = Pex 1 I ( p, q) exp 2 ivi x dp d q , (1-146a)
where
W ( p, q )
x = R (1-146b)
p
is the component of the ray aberration along the p axis. It may also be obtained from the
ray aberration ( x, y) according to
( ) [ (
g (vi , ) = Pex 1 exp 2 ivi x I ( p, q ) exp 2 ivi x x )] dp d q , (1-148a)
where
x = Pex 1 I ( p, q ) x dp d q (1-148b)
is the mean value of x . For small values of the radial spatial frequency vi , we expand
the exponential under the integral in Eq. (1-146a) and retain only the first three terms (the
second term vanishes upon integration) and obtain
( )
2
g (vi , ) ~ 1 2 2 vi2 x x (1-149a)
and
44 IMAGE FORMATION
(vi , ) ~ 2 vi x . (1-149b)
Thus, the variance of the ray aberration along the p axis determines the reduction of the
MTF along that axis, and its mean value determines the PTF. It is evident from Eq. (1-
149a) that the slope of the MTF g (vi , ) with respect to the radial frequency vi
evaluated at the origin is zero.
It should be noted that the diffraction OTF depends strongly on the wavelength of the
object radiation. The cutoff frequency, for example, is inversely proportional to the
wavelength. Since the wave aberration of a ray as an optical path length error in units of
wavelength will be different for different wavelengths, the aberrated OTF will also vary
with the wavelength. However, the geometrical OTF depends on the ray aberrations,
which are independent of the wavelength as long as the optical path length errors are
independent of it. If the path length errors of a system vary with the wavelength, i.e., if it
has chromatic aberration, then the ray aberrations will also vary with it. Thus, for
example, the diffraction OTF of a system consisting of mirrors will vary with the
wavelength, but the geometrical OTF will not. In a system using lenses and suffering
from chromatic aberration, there will be some variation of the geometrical OTF as well
with the wavelength.
1.6 Optical Transfer Function (OTF) 45
Theorem 13. The PSF of a system with a radially symmetric pupil function behaves
asymptotically as the inverse cube of the distance from its center, independent of the
aberration.
For an imaging system with a radially symmetric pupil function, its PSF may be
obtained from its OTF according to [see Eq. (1-84)]
where (vi ) is a real function. Willis19 has shown that for a function f ( x ) whose
derivatives exist
1 f (0) 3 f iv (0)
m f ( x ) J 0 ( mx ) dx ~ f (0)
2 + 2 K (1-151)
2 m 2 2! m 4
0
[
f ( n ) (0) = n f ( x ) x n ] x 0
. (1-152)
Letting x = vi , f (vi ) = vi (vi ) and m = 2 ri , Eq. (1-151) can be written for large values
of ri :
( 0 ) 3 (0)
PSF (ri ) ~
2 3 +
... . (1-153)
4 ri 32 4 ri5
Equation (1-153) gives an asymptotic representation of the PSF in terms of the properties
of its Fourier transform, the OTF, at the origin.
PSF (ri ) ~
(0) 4 2 ri3 . (1-154)
However, as shown in Section 1.6.4, (0) is independent of aberration. Hence, for large
values of ri , the PSF is independent of an aberration and varies with ri as ri
3 (Theorem
13). Tatian’s results for circular pupils show that Eq. (1-154) holds even for rotationally
nonsymmetric aberration.20 This result is plausible, because the effect of an aberration is
to reduce the irradiance at the center of the image and increase it at nearby points. This
does not imply that the minimum value of ri above which the PSF is (approximately)
described by an ri
3 dependence is the same whether or not the imaging system is
aberrated, but that for large enough values of ri , the point-spread function depends on ri
through ri
3 even when the system is aberrated.
1.7 Asymptotic Behavior of PSF 47
Theorem 14. The encircled power for large circles is independent of the aberration.
where I (ri , i ) is the irradiance at a point (ri , i ) in the image plane. Following Eqs. (1-
61) and (1-76), we may substitute for I (ri , i ) in terms of the corresponding transfer
function (vi , ) , and thus write Eq. (1-155) in the form
rc 2 2
Pi (rc ) = Pex ri dri d i vi dvi
0 0
(vi , ) exp [
2 iri vi cos (i
)] d . (1-156)
0
Changing the order of integration, the integration over i gives 2 J 0 (2 ri vi) , and noting
that
x0
J ( mx ) x dx = x 0 J ( mx ) ,
0 1 0 (1-157)
m
0
the integration over ri gives (rc vi ) J1 (2 rc vi ) , where J1 ( ) is the first-order Bessel
function of the first kind. Thus, Eq. (1-156) reduces to
2
Pi (rc ) = Pex rc d (vi , ) J1 (2 rc vi ) dvi . (1-158)
0
Since the left-hand side is real, the integral over the imaginary part of (vi , ) must be
zero. Hence, we may consider only the real part of (vi , ) in Eq. (1-158) and write it in
the form
2
Pi (rc ) = Pex rc d Re (vi , ) J1 (2 rc vi ) dvi .
0 (1-159)
for large positive values of m. Therefore, for large values of rc , Eq. (1-159) can be
written
2
Re (0, )
Pi (rc ) ~ Pex d +
Re (0, )
[ ] [
Re (0, )
+ ...
,
] (1-161)
2 4 2 rc 32 3r 3c
0
is the nth radial derivative of Re (vi , ) evaluated at the origin. Equation (1-162) gives
an asymptotic representation of the encircled power in terms of the radial derivatives of
the real part of the OTF at the origin. Since the transfer function is normalized to unity at
the origin, the first term on the right-hand side is Pex . Note that the equation is properly
normalized since Pi (rc ) = Pex . For large values of rc , we may neglect higher-order
terms and write
2
Pi (rc ) (
~ Pex 1 + 1 4 2 rc
) [Re (0, )] d . (1-163)
0
Since the integrand in Eq. (1-163) is independent of aberration [see Eq. (1-133)], we
conclude that, up to the first order, the encircled power for large circles is also
independent of an aberration (Theorem 14).
(r )
We now show that, for an unapodized system [i.e., one for which A rp = A0 across
its exit pupil], the integral on the right-hand side of Eq. (1-163), and therefore the asymp-
totic behavior of the encircled power, is related simply to the ratio of the perimeter length
and the transmitting area of the exit pupil.21 From Eq. (1-73b), we note that the OTF of an
r
unapodized aberration-free system corresponding to a spatial frequency vi is equal to the
r
fractional area of overlap of two pupils whose centers are separated by R vi . Thus, from
Figure 1-8, we note that the OTF may be written
r
( vi ) = 1
( R 2 Sex ) vi cos ! ds , (1-164)
s
where ds is a differential length segment along the perimeter of the pupil and ! is the
r
angle between vi and the normal to the segment. The factor of 2 accounts for the fact that
the nonoverlap area is counted twice in the integral in Eq. (1-164). For very small values
of vi (since we are interested in the slope of the OTF at the origin), it may be considered
constant and brought outside the integral. Hence, for such values, we may write
r
( vi ) = 1
( Rvi 2 Sex ) cos ! ds . (1-165)
s
1.7 Asymptotic Behavior of PSF 49
ds cos!
n^
O
R" i
O
Figure 1-8. Geometry for calculating the OTF of a system for very small spatial
r
frequencies. The displaced pupil is centered at O , whose position vector is Rvi
making an angle with respect to the pupil centered at O. Note that n̂ is a unit
vector along the normal to the pupil at the point where the line joining O and O
intersects it. The overlap area of the two pupils is shown shaded.
Since d! = d , we obtain
2 2
( vi , ) d = 2
( RLvi 2 Sex ) cos! d!
0 0
= 2
2 RLvi Sex , (1-166)
where L is the length of the perimeter of the exit pupil. Taking the derivative of both sides
with respect to vi , we obtain for very small values of vi and in the limit vi 0 :
2
2 RL
(0, ) d =
S . (1-167)
0 ex
Since the slope of the real part of the OTF at the origin is independent of aberration, Eq.
(1-167) is valid for aberrated but unapodized systems as well. Substituting this equation
into Eq. (1-163), we obtain
(
P (rc ) ~ Pex 1
RL 2 2 rc Sex ) . (1-168)
Thus, we see that the asymptotic behavior of encircled power depends on the ratio of the
perimeter length L of the exit pupil and its transmitting area Sex (Theorem 15). The
effect of an aberration is to increase the value of rc for which Eq. (1-168) is valid. The
larger the aberration, the larger the value of rc required for the validity of Eq. (1-168).
50 IMAGE FORMATION
Theorem 16. (a) The centroid of the diffraction PSF is given by the slope of the
imaginary part of the its diffraction OTF at the origin. (b) It is the same as the centroid of
the ray geometrical PSF. (c) The centroid of an aberration-free system lies at the
Gaussian image point regardless of the amplitude variations across its exit pupil. (d) For
a system with an aberrated but unapodized exit pupil, the centroid can be obtained from
its aberration only along its perimeter.
R
R
) (
x p xi + y p yi dx p dyp
) , (1-169)
where ( xi , yi ) are the coordinates of a point in the image plane with respect to the
( )
Gaussian image point, and x p , y p are the coordinates of a point in the plane of the exit
pupil. By dividing both sides of Eq. (1-169) by Pex we obtain the function PSF ( xi , yi )
for the system.
xi = Pex
1 xi Ii ( xi , yi ) dxi dyi (1-170a)
and
yi = Pex
1 yi Ii ( xi , yi ) dxi dyi . (1-170b)
From the definition of the PSF of an imaging system, it should be evident that its centroid
is synonymous with that of the irradiance distribution of the image of a point object.
(, ) = Pex
1 Ii ( xi , yi ) exp [2 i ( xi + yi )] dxi dyi , (1-171)
r
where (, ) are the rectangular components of a spatial frequency vector v in the image
plane. Differentiating both sides of Eq. (1-171) with respect to x and evaluating the result
at = = 0 , we find that
1 # &
xi = % * . (1-172a)
2 i $ ' 0
1.8 PSF Centroid 51
Similarly,
1 # &
yi = % * . (1-172b)
2 i $ ' 0
However, since xi and yi are real, only the slope of the imaginary part of the OTF at
the origin contributes to the centroid. Thus, we may write
1 # Im &
xi = % * (1-173a)
2 $ ' 0
and
1 # Im &
yi = % * . (1-173b)
2 $ ' 0
Thus, the centroid of the PSF of an optical system is given by the slope of the imaginary
part of its OTF at the origin (Theorem 16a). It can not, for example, be obtained from a
knowledge of only the MTF of the system.
( ) (
(, ) = Pex
1 P x p , y p P * x p
R , y p
R dx p dy p ) . (1-174)
R P * x p , y p ( )
xi =
Im P x p , y p
2 Pex
( x p
) dx p dy p (1-175)
and a similar equation for yi . The pupil function given by Eq. (1-38) or Eq. (1-47) may
be written in the form
( ) ( ) [
P x p , y p = A x p , y p exp (2 i ) W x p , y p ( )] . (1-176)
W x p , y p ( )
xi =
R
Pex
(
Ip x p , yp )
x p
dx p dy p , (1-177a)
( ) ( )
where I p x p , y p = A 2 x p , y p is the irradiance at a pupil point x p , y p . Similarly, ( )
W x p , y p ( )
yi =
R
Pex
I x ,y
p p p ( y p
) dx p dy p . (1-177b)
52 IMAGE FORMATION
( ) ( )
From Eq. (1-142), R W x p and R W y p represent the ray aberrations, i.e., the
( )
image-plane coordinates of a ray passing through the pupil point x p , y p . Hence, Eqs.
(1-177a) and (1-177b) show that the centroid of the PSF according to wave diffraction
optics is identical with that according to ray geometrical optics (Theorem 16b).
From Eqs. (1-177a) and (1-177b), we also note that amplitude variations across the
pupil affect the centroid only if it is aberrated. In the absence of aberrations, the PSF
centroid lies at (0, 0); i.e., it lies at the Gaussian image point where the center of
curvature of the spherical wavefront lies, regardless of the shape of the pupil and/or the
amplitude distribution across it (Theorem 16c). This may also be seen from Eqs. (1-169)
( )
and (1-176). We note from these equations that if W x p , y p = 0 , then Ii ( xi , yi ) =
Ii (
xi ,
yi ) . Hence, the symmetry of the aberration-free PSF yields its centroid at the
Gaussian image point. Similarly, since in that case the aberration-free OTF is real [see
Eq. (1-174)], Eqs. (1-173a) and (1-173b) also give the centroid at the Gaussian image
point.
Equations (1-170), (1-173a), (1-173b), (1-177a), and (1-177b) give the centroid in
terms of the PSF, OTF, and the aberration function, respectively. In practice, given an
imaging system, the most convenient expression to use would be Eqs. (1-170), since the
PSF can be measured by using a photodetector array. In optical design and analysis, the
simplest way to obtain the centroid would be to use Eqs. (1-177a) and (1-177b) since the
aberrations must be calculated even if the other two expressions were used. Thus, one
may trace rays all the way up to the image plane and determine the centroid of the ray
( )
distribution in this plane with appropriate weighting I p x p , y p of each ray.
( )
A x p , y p = A0 (1-178)
( )
I p x p , y p = A02
(1-179)
= Pex Sex ,
=
R
(
W x p , y p )
xi dx p dy p (1-180a)
Sex x p
and
1.9 Strehl Ratio 53
=
R
(
W x p , y p )
yi dx p dy p . (1-180b)
Sex x p
Using Stokes’ theorem, 23 the surface integrals in Eqs. (1-180a) and (1-180b) involving
the derivative of the aberration function can be written in terms of its line integral along
the curve bounding the surface. Thus, we may write
r
W ( x , y ) xˆ d s
R
xi = p p p (1-181a)
Sex
and
r
yi
( )
= ( R Sex ) W x p , y p yˆ p d s , (1-181b)
r
where x̂ p and ŷ p are unit vectors along the x p and y p axes, respectively, and d s
represents an element of arc length vector along the perimeter of the pupil. It is evident
from Eqs. (1-181a) and (1-181b) that, in the case of an aberrated but an unapodized
pupil, the centroid of the PSF can be obtained from the value of the aberration function
only along the perimeter of the pupil (Theorem 16d). Accordingly, in that case, to
calculate the centroid, the knowledge of the aberration across the interior of the pupil is
not needed.
Theorem 17. (a) The central irradiance for an apodized-aberrated system is less
than or equal to the corresponding value for an unapodized-unaberrated system. (b) For
a given total power, any amplitude variations reduce the central irradiance and any
54 IMAGE FORMATION
phase variations further reduce it. (c) The peak value of an unaberrated PSF lies at the
center, regardless of the amplitude variations across the pupil.
The irradiance distribution of the (defocused) image of a point object is given by Eq.
r
(1-60). For simplicity of notation, we let ri be the position vector of the observation point
r
( )
with respect to the point zi zg rg lying on the line joining the center of the exit pupil
and the Gaussian image point in the observation plane at a distance zi . Thus, the
irradiance distribution given by Eq. (1-60) may be written
2
r 1 r
2 i r r
r
I ( ri ; zi ) = 2 2 P rp ; zi exp
zi
(
zi
rp r d rp
) , (1-182)
(r )
where P rp ; zi is the defocused pupil function given by Eq. (1-47). The irradiance at
the center of the distribution for an apodized and aberrated pupil is obtained by letting
r
ri = 0 . Thus,
r r
P ( rp ; zi ) d rp
1 2
Iaa (0; zi ) = . (1-183)
2 zi2
Similarly, for the same total power Pex in the image, the central irradiance at the
Gaussian image point for an unapodized and unaberrated system may be obtained from
r r r
( )
Eq. (1-183) by letting zi zg , A rp = A0 and W rp = 0 , so that P rp = A0 . Thus, ( ) ( )
we may write
(
Iuu 0; zg = ) Pex Sex
2 zg2
(1-184)
r r
P (r ; z )
Sex 2
= p i d rp . (1-185)
2 zg2
The ratio of the central irradiance in the defocused plane for an apodized and
aberrated pupil, and in the Gaussian image plane for the unapodized and unaberrated
pupil may be written
Iaa (0; zi )
2
zg
= Saa , (1-186)
(
Iuu 0; zg ) zi
where
r r
P ( rp ) d rp
2
Saa = r r (1-187)
( )
2
Sex P rp d rp
is the corresponding Strehl ratio. It represents the ratio of the central irradiances in the
defocused and Gaussian image planes, except for the inverse-square-law effect
( )
2
represented by the factor zg zi . This factor is unity if the aberrated irradiance is also
1.9 Strehl Ratio 55
observed in the Gaussian image plane. It is practically equal to unity in imaging systems
because of their small depth of focus. However, it can be significantly different from
r r r
( ) ( ) ( )
unity in laser transmitters. Letting f rp = P rp , g rp = 1, and n = 2 in Hölder’s
inequality (1-104), we find that
Saa 1 . (1-188)
(r )
P rp = A rp (r ) , (1-189)
( ) ( ) [ A (rr ) d rr ]
2
Iau 0; zg = 1 2 zg 2 p p . (1-190)
(
Sau = Iau 0; zg ) Iuu (0; zg )
[ A (rrp ) rd rrp ]r .
2
= (1-191)
Sex A 2 ( rp ) d rp
Thus, any amplitude variations across the pupil of an aberration-free system reduce the
central irradiance. For example, as discussed in Section 4.3, a pupil with a Gaussian
illumination across it yields a smaller value of the central irradiance than a pupil with a
uniform illumination, the total power being the same in the two cases.
Comparing Eq. (1-191) with Eq. (1-187), and using Hölder’s inequality (1-104) with
r r r
( ) ( ) ( )
f rp = P rp , g rp = 1, and n = 1, we find that
r r 2
Saa P ( rp ) d rp
=
[ A (rrp ) d rrp ]
2 (1-193a)
Sau
1 . (1-193b)
56 IMAGE FORMATION
From Eqs. (1-192) and (1-193b), we note that amplitude variations reduce the central
irradiance, and phase variations (i.e., aberrations) further reduce it. Note, however, that
an irradiance reduced by phase variations alone does not necessarily reduce any further
if amplitude variations are also introduced (Theorem 17b). In fact, amplitude variations
can even increase this irradiance. The maximum value of central irradiance is obtained
when the system is unapodized and unaberrated.25 Hence, the maximum value of
irradiance at a certain point in the image space is obtained when a beam exiting from the
system is focused at it with uniform amplitude and phase at its exit pupil. Any variation
in the amplitude or phase across the exit pupil reduces the value at the chosen point.
The peak value of the aberrated irradiance distribution of the image of a point object
does not necessarily occur at the center of the reference sphere. However, the peak value
of its unaberrated image does occur at this point regardless of the apodization. The
Huygens’ spherical wavelets emanating from the spherical wavefront are equidistant from
this point. Hence, they interfere constructively, producing a maximum possible value at
r r
( ) ( )
this point. Mathematically, this may be seen by letting zi = zg , and P rp = A rp in
r r
Eq. (1-182), and comparing the irradiance at a certain point ri with that at ri = 0 . Thus,
we may write
r r r
f ( rp ) d rp
2
Iau ( ri )
=
Iau (0)
[ (r ) r
]
2
f rp d rp
1 , (1-194)
where
r r
2 i r r
( ) ( )
f rp = A rp exp
zg
rp ri .
(1-195)
Hence, the peak value of an unaberrated image occurs at the Gaussian image point
regardless of the apodization of the system (Theorem 17c).
Theorem 18. For small aberrations, the Strehl ratio of an image is determined by
the variance of the phase aberration across the exit pupil.
The ratio of the central irradiance at a distance zi with aberration and that at the
Gaussian image point without aberration is given by
1.9 Strehl Ratio 57
Iaa (0; zi )
2
zg
= S , (1-196)
(
Iau 0; zg ) zi
where
r r r
A (rp ) exp [ i (rp )] d rp
2
S =
[ (r ) r (1-197)
]
2
A rp d rp
is the Strehl ratio of the image. The right-hand sides of Eqs. (1-193a) and (1-197) are
identical. Hence,
S 1 . (1-198)
This Strehl ratio gives a measure of the irradiance reduction due to aberration in the
system and/or due to defocus. It can be written in an abbreviated form
2
S = exp (i ) , (1-199)
where the angular brackets L indicate a spatial average over the amplitude-weighted
pupil, e.g.,
r r r
=
A ( rp ) ( rp ) d rp
[ A (rrp ) d rrp ]
2 . (1-200)
r
Since is independent of rp , Eq. (1-199) can be written
[ )]
2
S = exp i (
= cos ( ) + sin ( )
2 2
cos ( )
2 (1-201)
,
equality holding when is zero across the pupil, in which case S = 1. For small
aberrations, expanding the cosine function in a power series and retaining the first two
terms, we obtain the Maréchal result generalized for an apodized pupil
(1 2)
2
S >~ 2
, (1-202)
where
2 = ( )2 (1-203)
For small values of , three approximate expressions have been used in the
literature:
~ (1 2 2)
2
S1 , (1-204)
S2 ~ 1 2 , (1-205)
and
S3 ~ exp ( 2 ) . (1-206)
The first is the Maréchal formula, 26 the second is the commonly used expression obtained
when the term in 4
in the first is neglected,27,28 and the third is an empirical expression
giving a better fit to the actual numerical results for various aberrations,29 as we shall see
in Sections 2.2.4 and 3.2.4. The simplest expression to use is, of course, S2 , according to
which 2 gives the drop in the Strehl ratio. We note that the Strehl ratio for a small
aberration does not depend on its type but only on its variance across the apodized pupil
(Theorem 18). For a high-quality imaging system, a typical value of the Strehl ratio
desired is 0.8, corresponding to a wave aberration standard deviation of w = 14 .
Second, the Strehl ratio may be determined from the PSF of the system. We calculate
its unaberrated PSF from its pupil shape and amplitude distribution, and normalize it so
that its central value is unity. Next, we integrate it in its plane to give the total power.
Finally, we measure the aberrated PSF and normalize it so that its integral has the same
value as for the aberration-free case. This step insures that the total power in the aberrated
image is the same as in the aberration-free image. The central value of the aberrated PSF
normalized in this manner gives the Strehl ratio.
Third, the Strehl ratio may be determined from the OTF of the system. Since its PSF
and OTF form a Fourier transform pair [see Eq. (1-77)], its Strehl ratio may be written
integrating the real part of the measured aberrated OTF over all spatial frequencies and
dividing it by a similar integral of the calculated unaberrated OTF. In any of these three
approaches for determining the Strehl ratio, one must take into account the effect of the
inverse-square law, i.e., the effect of zi being different from zg , unless they are
practically equal to each other (see Section 2.5).
Theorem 19. (a) For small values of the phase aberration difference function, the
r
Hopkins ratio for a certain spatial frequency vi is determined by the variance of the
difference function across the overlap area of two pupils displaced with respect to each
r
other by Rvi . (b) The mean value of the phase aberration difference function represents
the corresponding PTF.
r
Following Hopkins,30 we define a modulation ratio H ( vi ) as the ratio of the MTFs
r r r
( vi ) and u ( vi ) of a system at a spatial frequency vi with and without aberration, i.e.,
r r r
H ( vi ) = ( vi ) u ( vi ) . (1-209)
r
For obvious reasons, we call H ( vi ) the Hopkins modulation (or contrast) ratio. From Eq.
(1-107), this ratio is less than one. From Eq.(1-73c) it can be written
r r r
r A(r ) A(r
p p ) {[ (r ) (r
Rvi exp i rp rp R vi
r
)] } d rrp
H ( vi ) = r r r r . (1-210)
( ) (
A rp A rp Rvi d rp )
In the rotated ( p, q ) coordinate system shown in Figure 1-4, Eq. (1-209) may be written
r r
r
H ( vi ) =
I ( p, q; vi ) exp [iQ ( p, q; vi )] dp dq
r , (1-211)
I ( p, q; vi ) dp dq
r
where the mutual irradiance function I ( p, q; vi ) and the phase aberration difference
r
function Q( p, q; vi ) are given by Eqs. (1-128a) and (1-128b). As in Eq. (1-125), a
function in the ( p, q ) coordinate system is obtained from the corresponding function in
( )
the x p , y p coordinate system by replacing x p with p cos q sin and y p with
p sin + q cos . As discussed in Section 3.5 of Part I, the aberration function
60 IMAGE FORMATION
( )
x p , y p for a rotationally symmetric imaging system depends on x p and y p through
r
x 2p + y 2p and x p , where x p = rp cos , y p = rp sin , and is the angle the vector rp
makes with the x p axis (which lies in the tangential plane containing the optical axis and
( )
the object point). Hence, ( p, q ) is obtained from x p , y p by replacing x 2p + y 2p by
p 2 + q 2 and x p by p cos q sin .
where the angular brackets indicate an average across the overlap region of the two
apodized pupils. For example,
r r
Q =
I ( p, q; vi ) Q(rp, q; vi ) dp dq . (1-213)
I ( p, q; vi ) dp dq
For small values of Q Q , we may retain only the first three terms in the expansion of
the exponential in Eq. (1-212), one of which reduces to zero upon averaging. Thus, we
obtain the result
r ~ 1 2
H ( vi ) 1 Q , (1-214)
2
where
2
Q2 = Q 2 Q (1-215)
is the variance of the phase aberration difference function across the overlap region of the
r
two apodized pupils. Multiplying both sides of Eq. (1-208) by exp i
( vi ) , we obtain [ ]
[see Eq. (1-96)]
r r r r
[ ]
H ( vi ) exp i
( vi ) = ( vi ) u ( vi )
{
= exp i [Q Q ]} exp (i Q )
Ê 1 ˆ
~ ÁË 1 - s Q2 ˜¯ exp (i Q ) . (1-216)
2
We noted in Section 1.9.2 that, for small aberrations, the Strehl ratio of an aberrated
system depends on the variance of the aberration function across its pupil and not on the
type of the aberration. Similarly, we note from Eq. (1-214) that, for small values of the
phase aberration difference function, the Hopkins ratio for a certain spatial frequency
1.11 Line and Edge Spread Functions (LSF and ESF) 61
depends on the variance of the difference function across the overlap region of two
displaced pupils (displacement depending on the spatial frequency) and not on the type of
the aberration (Theorem 19a). Similarly, Eq. (2-217) shows that the mean value of the
phase aberration difference function represents the PTF (Theorem 19b).
r ~ Ê 1 ˆ
H (v i ) exp Á - s Q2 ˜ . (1-218)
Ë 2 ¯
Theorem 20. (a) The aberration-free line-spread function is symmetric about its
center, and the aberrations reduce its central value. (b) The line-spread function is equal
to the derivative of the edge-spread function.
= PSF( xi , yi )dyi . (1-219a)
For a line object along the yo axis, the LSF depends only on xi , i.e., its variation with xi
is the same regardless of the value of yi , as expected for an isoplanatic line object. Since
r
( ) ( )
its Gaussian image Ig rg = x g , its diffraction image can also be obtained by applying
Theorem 5 as the convolution of its Gaussian image and the PSF, i.e.,
( )
LSF( xi ) = x g PSF( xi x g , yi yg ) dx g dyg . (1-219b)
It should be evident that the LSF can be obtained from the PSF by scanning it with a long
and narrow slit. If the slit is parallel to the yi axis, the LSF thus obtained is LSF( xi ) .
62 IMAGE FORMATION
The LSF of a system can be written in terms of its pupil function by substituting for
its PSF from Eq. (1-57) into Eq. (1-219a) . Thus
• ÛÛ È 2pi ˘
LSF ( x i ) =
1
(
2 2 Ú dy i Ù Ù P x p , y p exp Í -
Pex l R • ı ı Î lR
)
x i x p + y i y p ˙ d x p dy p
˚
( )
ÛÛ È 2pi ˘
(
¥ Ù Ù P * x ¢p , y ¢p exp Í
ıı
)
Î lR
(
xi x ¢p + yi y ¢p ˙ d x ¢p dy ¢p
˚
)
ÛÛ Ê 2pi ˆ ÛÛ Ê 2pi ˆ
ıı
( )
= Ù Ù P x p , y p exp Á -
Ë lR
x i x p ˜ d x p dy p Ù Ù P * x ¢p , y ¢p exp Á
¯ ıı Ë lR
( )
x i x ¢p ˜ d x ¢p dy ¢p
¯
•
Û È 2 pi ˘
¥
1
2 2 Ù exp Í
Pex l R ı ÎlR
(
yi y ¢p - y p ˙ dyi
˚
)
•
2
Û Ê 2pi ˆ
=
1
Pex l R ı
(
Ú dy p Ù P x p , y p exp Á -
Ë lR
)xi x p ˜ d x p
¯
, (1-220)
where in the last step we have used the fact that the integral over yi is equal to
( )
R y p y p and thereby carried out the integration over y p . Hence, the LSF
normalized by its aberration-free value at the origin is given by
2
Û Ê 2pi ˆ
ı
(
Ú dy p Ù P x p , y p exp Á -
Ë
)l R
xi x p ˜ d x p
¯
LSF ( x i ) = . (1-221a)
[Ú ( ) ]
2
Ú dy p P xp, yp d xp
The aberration-free LSF is symmetric about xi = 0 . The central value of the LSF
normalized by its aberration-free value is called the Struve ratio. It is given by
( )
2
dy p P x p , y p d x p
LSF(0) = (1-221b)
dy p [ P( x p , y p ) d x p ]
2
1 , (1-221c)
where in the last step we have used Hölder’s inequality (1-104) with f x p = P x p , ( ) ( )
( )
g x p = 1, and n = 2 . Thus, aberrations reduce the central value of the LSF, as in the case
of the PSF.
If the line object lies along a yo axis (or parallel to it), making an angle with the
yo axis, as illustrated in Figure 1-9, then the LSF( xi) is obtained by writing the pupil
( )
function in the coordinate system x p , y p by replacing x p by x p cos y p sin and y p
by x p sin + y p cos .
1.11 Line and Edge Spread Functions (LSF and ESF) 63
yo
yo
xo
xo
yp
xp
xp
yi
xi
xi
Figure 1-9. Coordinate system in various planes for determining the LSF. (a) Object
plane. (b) Pupil plane. (c) Image plane. The line object lies along the yo axis or
parallel to it, making an angle with the yo axis.
64 IMAGE FORMATION
xi
= dx PSF ( x , y ) dy , (1-222b)
where in the last step we have substituted for the LSF from Eq. (1-219a). It is evident
from Eq. (1-222a) that the LSF is the derivative of the ESF, i.e.,
d
LSF( xi ) = ESF( xi ) . (1-223)
dxi
Moreover, it is evident from Eq. (1-222b) that ESF( ) = 1, as may be seen by the use of
Eqs. (1-61) and (1-63). It may also be seen by noting that the integral for xi = is equal
to
(0, 0) , which, in turn, is equal to unity.
Introducing the step or the edge function representing the edge object,
H ( x ) = 1 for x 0
(1-224)
= 0 for x < 0 ,
= PSF( x, y) H ( xi x ) dx dy . (1-225b)
PSF. The PSF may also be measured using a 2D array of small photodetectors. But,
again, precise measurements of the PSF outside the bright spot may be difficult owing to
their low values. The LSF can be obtained by scanning a corresponding PSF with a long
and narrow slit. The difficulty of a low signal can be overcome by using a line object and
measuring the LSF by scanning its image with a long and narrow slit. The slit is parallel
to the line object and its scan direction is normal to it. By removing one side of the slit
used as a line source, the ESF can be measured in a similar manner.
Theorem 21. (a) The LSF and the OTF form a 1D Fourier pair. (b) Similarly, the
ESF and the OTF divided by a spatial frequency are related to each other by a 1 D
Fourier transform.
PSF ( x, y) =
(,
) exp [ 2 i ( x +
y)] d d
, (1-226)
where, for simplicity, we have dropped the subscript on the image coordinates ( xi , yi ) ,
Eq. (1-219) may be written
[ ]
LSF ( x ) = dy
(,
) exp 2 i ( x +
y) d d
=
(,
) exp ( 2 i x ) d d
exp ( 2 i
y) dy
=
(,
) exp ( 2 i x ) (
) d d
=
(, 0) exp ( 2 i x ) d . (1-227a)
Thus, LSF( x ) , representing the image of a line object lying along the yo axis, is the 1D
inverse Fourier transform of
(, 0) (Theorem 21a). Since LSF( xi ) is a real function,
only the real part of the integrand contributes to the integral; the imaginary part yields
zero upon integration. Hence, Eq. (1-227a) may be written
LSF( x ) = Re
(, 0) cos(2 x )d Im
(, 0) sin(2 x )d . (1-227b)
Similarly, by considering a line object along the x axis, we can show that its image
LSF( y) is the inverse Fourier transform of
(0,
) , i.e.,
LSF( y) =
(0,
) exp ( 2 i
y) d
. (1-228 )
The LSF for a line object making an angle with the yo axis is given by
66 IMAGE FORMATION
LSF( x ) =
( , 0) exp ( 2 i x ) d , (1-229)
where x and are parallel to each other and make an angle with the x axis in the
image plane, and
(, 0) is simply the OTF along the direction . It should be clear
from Eqs. (1-227a) through (1-229) that the LSF represents the variation of irradiance in
the image of a line object along a direction that is normal to the direction of the object.
Now we consider an edge object that is parallel to the yo axis. To relate the ESF to
the OTF, we Fourier transform both sides of Eqs. (1-225a) and (1-225b). Since the
integrals on their right-hand sides represent convolutions, their Fourier transforms are
equal to the product of the Fourier transforms of the convolving functions. From Eq (1-
226), the Fourier transform of LSF( x ) is given by
LSF( x ) exp (2 i x )d x =
(, 0) . (1-230)
H ( x ) exp (2 i x )d x = () 2 (1 2 i) . (1-231)
ESF( x ) exp (2 i x )d x = () 2
(, 0) 2 i , (1-232)
since
(, 0)() =
(0, 0) = 1 . Fourier transformation of Eq. (1-225b) also yields Eq. (1-
232). The left-hand side of Eq. (1-232) represents the spatial-frequency spectrum of the
image of an edge object lying parallel to the yo axis, and its right-hand side represents the
product of the spatial-frequency spectrum of the edge object (or its Gaussian image) and
the system OTF for
= 0 , as expected for an isoplanatic object according to Eq. (1-70).
Inverse Fourier transforming Eq. (1-232), we obtain the ESF in terms of the OTF:
1 1
( , 0 )
ESF( x ) = exp ( 2 i x ) d . (1-233)
2 2 i
For an incoherent object, ESF( x ) represents the irradiance distribution of the image of an
edge object. It is, therefore, a real function. Hence, only the imaginary part of the
integrand in Eq. (1-233) contributes to the integral; the real part yields zero upon
integration. Accordingly, Eq. (1-233) may be written
1 sin(2 x ) cos(2 x )
ESF( x ) = + Re
(, 0) d Im
(, 0) d . (1-234)
2 2 2
1 Im
(, 0)
ESF(0) = d , (1-235)
2 2
1.12 Shift Invariant Imaging of a Coherent Object 67
i.e., the value of the ESF at the origin is determined solely by the imaginary part of the
OTF
(, 0) . If the OTF is real, e.g., for a symmetric aberration, then ESF(0) = 1 2 ,
independent of the aberration. If the edge object makes an angle with the yo axis, then
the ESF is given by an equation similar to Eq. (1-233), except that x and axes are
replaced by x and axes as in Eq. (1-229).
Equations (1-227a), (1-227b), and (1-233) for the LSF and ESF of a system in terms
of its OTF are also applicable to coherent imaging, provided the OTF is replaced by the
corresponding coherent transfer function (CTF) discussed in Section 1.12.2. This is done
in Section 2.15 for systems with circular pupils.
Theorem 23. (a) The complex amplitude image of an isoplanatic coherent object is
equal to the convolution of its Gaussian amplitude image and its coherent spread
function. (b) Accordingly, the spatial frequency spectrum of the diffraction amplitude
image is given by the product of the spectrum of the Gaussian amplitude image and the
coherent transfer function.
i r2 r r r2
= exp ik zi zg + i d ro Uo ( ro ) exp ik o
zi 2 zi 2 zo
r r 2 i r r z r r
(
× G rp ; ro ; zi exp
zi
) zg
rp ri i rg d rp , (1-236)
r r r r
where Uo ( ro ) is the amplitude at the object point ro , rg = M ro is the position vector of
the Gaussian image point, zg is the distance of the Gaussian image plane, and
r r r r ik 1 1
(
G rp ; ro ; zi ) ( )
= G rp ; ro exp rp2
2 zi zg
(1-237)
r
Uo ( ro ) = U j ro rj (r r
) , (1-238)
(r r
Ui ri ; rj ; zi ) = U j CSF ri ; rj ; zi(r r
) , (1-239)
where
r r " 1 r 2 rj $"
2
(
CSF ri ; rj ; zi ) =
i
zi
( )
exp !ik zi zg + i + %
"# 2 zi zo "
&
r r 2 i r r z r r
(
× G rp ; rj ; zi exp
zi
) zg
rp ri i M rj d rp
(1-240)
determining the defocus aberration, which varies as rp2 . For systems with small fields of
view, we may replace zg by R, the radius of curvature of the Gaussian reference sphere
with respect to which the aberration is defined.
For coherent imaging, we must consider another approximation, namely, that the
r
variation of the quadratic phase factor depending on ro is negligible.33 We will see that
the magnitude of the CSF is significant only in a small region (on the order of the Airy
disc discussed in Section 2.2) surrounding the Gaussian image point. Therefore, at a
r
particular point of observation ( ri ; zi ) , the contributions of the quadratic phase factor will
r r r
come from a small region of the object. Hence, we may replace ro = rg M by ri M in
the object-dependent quadratic phase factor. Any phase factors that do not depend on the
location of the object point can be dropped if it is the irradiance that is measured in the
observation plane.
(r r
CSF ri ; rj ; zi ) =
i
R
r
( 2 i r r
G rp ; zi exp )
R
r r
(
rp ri M rj d rp
) . (1-241)
r r
We note that the integral on the right-hand side depends on the position vectors ri and rg
of the observation and Gaussian image points, respectively, through their difference
r r r r r r
( ) ( )
ri rg only. Hence, we may replace CSF ri ; rj ; zi by CSF ri M rj ; zi . Thus, the CSF
is shift invariant in that its form does not depend on the location of the object point; the
r r
whole distribution shifts by virtue of being centered at rg = M rj . Accordingly, for an
isoplanatic coherent object, Eq. (1-236) for the amplitude distribution of its image may be
written
r
(r ) (r
Ui ( ri ; zi ) = M 1 Uo rg M CSF ri rg ; zi d rg
r
) r
r r r r
= Ug ( rg ) CSF( ri rg ; zi ) d rg , (1-242)
where
r i r 2 i r r r
CSF( ri ; zi ) =
R
( )
G rp ; zi exp
R
rp ri d rp
(1-243)
and
(r )
U g rg = M 1 Uo rg M (r ) (1-244)
is the amplitude distribution of the Gaussian image. Thus, the coherent spread function is
proportional to a Fourier transform of the pupil function (Theorem 22), and the image
amplitude distribution for an isoplanatic coherent object is equal to the convolution of the
Gaussian amplitude image (which is an exact replica of the object amplitude except for its
magnification M) and the coherent spread function (Theorem 23a). Comparing Eq. (1-55)
70 IMAGE FORMATION
with Eq. (1-242), we find that the incoherent and coherent point-spread functions are
related to each other according to
r
CSF( ri )
2
r
PSF( ri ) = r r . (1-245)
( )
2
G rp ; zi d rp
where
r r r r r
U˜ g ( vi ) = Ug ( rg ) exp (2i vi rg ) d rg (1-247)
r
is the coherent transfer function of the system corresponding to a spatial frequency vi in
the image plane. Substituting Eq. (1-243) into Eq. (1-248) and dropping the constant
phase factor i, we obtain
r 1 r r r r 2 i r r r
CTF( vi ) =
R
d ri exp (2 vi ri ) G rp exp
R
( )
rp ri d rp
r
r r r r r rp
=
1
R
d rp G rp ( )
d ri exp 2 i ri vi
R
r
r r r rp
=
1
R ( )
d rp G rp vi
R
r
= G( Rvi ) . (1-249)
Thus, the coherent transfer function is simply a scaled version of the relative pupil
function (Theorem 24). If the system is diffraction limited, i.e., if it is aberration free, and
r
( )
its transmission across the pupil is uniform so that G rp is equal to a constant across the
pupil (whose value varies inversely with the distance zo of the entrance pupil from the
r
object), then the normalized CTF is unity for those frequencies for which Rvi lies
inside the pupil.
1.13 Summary of Theorems 71
r
Fourier transforming both sides of Eq. (1-245), substituting for CSF( ri ) from Eq. (1-
243), and proceeding as we did in obtaining Eq. (1-73b), we obtain
r r r r r r r
G (r ) G * (r ) G (r )
2
( vi ) = p p R vi d rp p d rp . (1-250)
Thus, the incoherent transfer function is the (normalized) autocorrelation of the coherent
transfer function. Since the pupil functions P and G are proportional to each other, the
OTF given by Eq. (1-250) is the same as that given by Eq. (1-73b).
where the spectra of the amplitude diffraction and Gaussian images are given by
r r r r r
U˜ i ( vi ) = Ui ( ri ) exp (2 vi ri ) d r (1-252)
and
r r r r r
U˜ g ( vi ) = Ug ( ri ) exp (2 vi ri ) d r , (1-253)
respectively. Thus, the spectrum of the diffraction image is given by the product of the
spectrum of the Gaussian image and the CTF (Theorem 23b).
Theorem 3. The Gaussian image of a small object lying at a large distance from an
imaging system is an exact replica of the object, except for its magnification.
Theorem 7. The OTF of an imaging system is also equal to the autocorrelation of its
pupil function.
Theorem 8. The PSF and the OTF of a system with a radially symmetric pupil
function form a zero-order Hankel transform pair.
Theorem 10. (a) The OTF is a complex symmetric function with a value of unity at
the origin. (b) Its magnitude, the MTF, is unity or less at any other spatial frequency. (c)
The aberrated MTF is less than or equal to the corresponding aberration-free value.
Theorem 11. The slope of the real part of the OTF at the origin is independent of the
aberration, and it is equal to the corresponding slope of its MTF.
Theorem 12. The diffraction OTF approaches the geometrical OTF as the
wavelength approaches zero.
Theorem 13. The PSF of a system with a radially symmetric pupil function behaves
asymptotically as the inverse cube of the distance from its center, independent of the
aberration.
Theorem 14. The encircled power for large circles is independent of the aberration.
Theorem 16. (a) The centroid of the diffraction PSF is given by the slope of the
imaginary part of its diffraction OTF at the origin. (b) It is the same as the centroid of the
ray geometrical PSF. (c) The centroid of an aberration-free system lies at the Gaussian
image point regardless of the amplitude variations across its exit pupil. (d) For a system
with an aberrated but unapodized exit pupil, the centroid can be obtained from its
aberration only along its perimeter.
Theorem 17. (a) The central irradiance for an apodized-aberrated system is less
than or equal to the corresponding value for an unapodized-unaberrated system. (b) For
a given total power, any amplitude variations reduce the central irradiance, and any
phase variations further reduce it. (c) The peak value of an unaberrated PSF lies at the
center, regardless of the amplitude variations across the pupil.
Theorem 18. For small aberrations, the Strehl ratio of an image is determined by
the variance of the phase aberration across the exit pupil.
1.13 Summary of Theorems 73
Theorem 19. (a) For small values of the phase aberration difference function, the
r
Hopkins ratio for a certain spatial frequency vi is determined by the variance of the
difference function across the overlap area of two pupils displaced with respect to each
r
other by Rvi . (b) The mean value of the phase aberration difference function represents
the corresponding PTF.
Theorem 20. (a) The line-spread function is symmetric about its center and the
aberrations reduce its central value. (b) The line-spread function is equal to the
derivative of the edge-spread function.
Theorem 21. (a) The LSF and the OTF form a 1D Fourier pair. (b) Similarly, the
ESF and the OTF divided by a spatial frequency are related to each other by a 1 D
Fourier transform.
Theorem 23. (a) The complex amplitude image of an isoplanatic coherent object is
equal to the convolution of its Gaussian amplitude image and its coherent spread
function. (b) Accordingly, the spatial frequency spectrum of the diffraction amplitude
image is given by the product of the spectrum of the Gaussian amplitude image and the
coherent transfer function.
Thus, we use a + i in the Fourier transform integral and a i in the inverse Fourier
transform integral. Accordingly, since there is a i in the exponent in Eq. (1-51), the
incoherent PSF is proportional to the modulus square of the inverse Fourier transform of
the pupil function. Similarly, the OTF, which by definition is the Fourier transform of the
PSF, has a + i in the exponent in Eq. (1-71). It is also equal to the autocorrelation of the
pupil function, as in Eq. (1-73b).
Our definition of a Fourier transform is similar to that by Born and Wolf2 , but is
different from that by the others. For example, Hopkins 30 uses a + i in the PSF integral,
and a i in the OTF definition, which he calls an inverse Fourier transform. Welford34
uses a i in the PSF and a + i in the OTF, just as we have in Eqs. (1-51) and (1-71).
However, he refers to OTF as the inverse Fourier transform of PSF. Goodman3 and
Gaskill35 both use i in both the PSF and the OTF. According to them, PSF is related to
the Fourier transform of the pupil function, and OTF is the Fourier transform of PSF. The
OTF according to them, as the autocorrelation of the pupil function, is the complex
conjugate of the corresponding integral given in Eq. (1-73b). Goodman uses + i for the
Fourier transform in another book. 36
r r
This ambiguity does not arise if f (r ) is radially symmetric, i.e., if f (r ) = f (r ) ,
r
where r = r , since in that case its Fourier transform is real and reduces to a zero-order
Hankel transform:
2
F(') = f (r ) exp(2 ir' cos *) rdrd*
0 0
= 2 f (r ) J 0 (2 r') rdr , (A3)
0
f (r ) = 2 F(') J 0 (2 r') ' d' . (A4)
0
According to our sign convention, as discussed in Part I of this book, the wave
aberration associated with an object ray is numerically positive if it travels an extra
optical path length to reach the reference sphere compared to the corresponding chief ray.
Appendix B: Some Frequently Used Integrals 75
Born and Wolf2 and Goodman3 have the same sign convention, but Welford’s34 is
opposite to ours. A change in the sign of the wave aberration changes the sign of the
imaginary part of the pupil function, and the coherent as well as the incoherent transfer
functions. It also rotates the incoherent PSF by p as, for example, in the case of coma.
It is quite common in the optics literature to consider a point object lying along the y
axis. When using polar coordinates of a point in the plane of the exit pupil, the polar
angle q in that case is defined as the angle made by the position vector of the point with
the y axis, contrary to the standard convention as the angle with the x axis. We choose a
point object along the x axis so that, for example, coma aberration is expressed as
( ) ( )
x x 2 + y 2 and not as y x 2 + y 2 . A positive value of our coma aberration yields a PSF
that is symmetric about the x axis (or symmetric in y) with its peak and centroid shifted to
a positive value of x with respect to the Gaussian image point.
f ( x ) ( x ) dx = f (0) , (B2)
1
( ax ) = ( x ) . (B3)
a
•
f ( x) ƒ g * ( - x) = Ú f ( x ¢ ) g * ( x ¢ - x ) dx ¢ . (B5)
•
Autocorrelation of a function f ( x ) :
•
f ( x) ƒ f * ( - x) = Ú f ( x ¢ ) f * ( x ¢ - x ) dx ¢ , (B6)
•
Zero-order Bessel function J 0 (◊) of the first kind and its radial integral:
2
exp(ix cos *) d* = 2 J 0 ( x ) . (B7)
0
r0
r0
J 0 ( mx ) rdr = J1 ( mr0 ) . (B8)
0 m
76 IMAGE FORMATION
REFERENCES
1. V. N. Mahajan, Optical Imaging and Aberrations, Part I: Ray Geometrical Optics
(SPIE Press, Bellingham, WA, 1998; Second Printing 2001).
2. M. Born and E. Wolf, Principles of Optics, 7th Ed. (Cambridge University Press,
New York, 1999).
3. J. W. Goodman, Introduction to Fourier Optics, 2nd ed. (McGraw-Hill, New
York, 1996).
4. A. Sommerfeld, Optics, Vol 4 (Academic, New York, 1972); also J. W.
Goodman, Introduction to Fourier Optics, 2nd ed. (McGraw-Hill, New York,
1996).
5. E. Lalor, “Conditions for the validity of angular spectrum of plane waves,” J. Opt.
Soc. Am. 58, 1235–1237 (1968).
6. J. E. Harvey and R. V. Shack, “Aberrations of diffracted wave fields,” Appl. Opt.
17, 3003–3009 (1978).
7. J. E. Harvey, “Fourier treatment of near-field scalar diffraction theory,” Am.
J. Phys. 47, 974–980 (1979).
8. V. N. Mahajan, “Aberrations of diffracted wave fields I. Optical Imaging,” J. Opt.
Soc. Am. 17, 2216–2222 (2000).
9. V. N. Mahajan, “Aberrations of diffracted wave fields II. Diffraction Gratings,” J.
Opt. Soc. Am. 17, 2223–2228 (2000).
10. R. Bracewell, The Fourier Transform and Its Applications (McGraw-Hill, New
York, 1965).
11. For properties of Bessel functions, the reader may refer to G. N. Watson, Treatise
on the Theory of Bessel Functions (Cambridge University Press, New York,
1944).
12. G. A. Korn and T. M. Korn, Mathematical Handbook for Scientists and Engineers
(McGraw-Hill, New York, 1968), p. 118.
13. H. H. Hopkins, “Geometrical optical treatment of frequency response,” Proc.
Phys. Soc. (London) B 70, 1162–1172 (1957).
14. K. Miyamoto, “On a comparison between wave optics and geometrical optics by
using Fourier analysis. I. General Theory,” J. Opt. Soc. Am. 48, 57–63 (1958).
15. K. Miyamoto, “On a comparison between wave optics and geometrical optics by
using Fourier analysis. II. Astigmatism, coma, spherical aberration,” J. Opt. Soc.
Am. 48, 567–575 (1958).
16. K. Miyamoto, On a comparison between wave optics and geometrical optics by
using Fourier analysis. “III. Image evaluation by spot diagram,” J. Opt. Soc. Am.
49, 35–40 (1959).
17. K. Miyamoto, “Wave optics and geometrical optics in optical design,” Progress in
Optics, Vol. 1, 31–65 (1960).
References 77
PROBLEMS
1. Free-Space PSF and Transfer Function: Show that the transfer function of free-
space propagation in the Fresnel approximation given by Eq. (1-28) is the Fourier
transform of the corresponding PSF given by Eq. (1-24).
5. Symmetric Pupil Function: Show that the PSF is symmetric and the OTF is real for
r r
r r r *r ( ) ( )
a system with a symmetric pupil function, i.e., show that if P rp = P rp , then
PSF ( ri ) = PSF ( ri ) and
( v ) =
( v ) .
79
80 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Aberrated PSFs for rotationally symmetric aberrations are considered, and numerical
results are obtained that show that the size of their central spot is practically independent
of the aberration. Symmetry properties of the aberrated PSFs are considered next. Full
PSFs for primary aberrations are shown graphically and pictorially. A brief comparison
of the diffraction and geometrical PSFs is also given. The line of sight of an optical
system is identified with the centroid of its PSF and discussed for aberrated systems. It is
shown that only coma-type aberrations change the line of sight. The OTFs for primary
aberrations are discussed, and the concept of contrast reversal is described. The OTF
aberrated by coma is complex, and we give its real and imaginary parts as well as its
modulus and phase. Aberration tolerances are given for a certain value of the Hopkins
ratio. The geometrical OTF is also discussed and approximate expressions for it are
obtained for the primary aberrations. It gives reasonably accurate results in the region of
very low spatial frequencies, a region of practical interest when the MTF of an aberrated
system at high frequencies is practically negligible. It is shown that the MTF averaged
over all angular orientations of a spatial frequency vector is maximum when the standard
deviation of the ray aberration [i.e., the root mean square (rms) radius of the spot diagram
with respect to its centroid] is minimum. The diffraction OTF is shown to approach the
geometrical OTF in the limit of zero wavelength. The differences between the diffraction
and geometrical OTFs are summarized. The incoherent line- and edge-spread functions
81
82 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
are discussed, and numerical results for both the aberration-free and the aberrated systems
are given. The aberration tolerances for a certain value of the Struve ratio are also given.
A few miscellaneous topics are considered next. The changes in the monochromatic
PSF as the spectral bandwidth increases are discussed. It is shown that the diffraction
rings do not disappear until the relative bandwidth approaches almost unity. The
aberration-free polychromatic OTF is also considered, and it is shown that it decreases at
every spatial frequency as the spectral bandwidth increases. Next, the image of a
uniformly radiating or illuminated disc is considered as an example of an extended
incoherent object. As a corollary, we show that an object whose Gaussian image size is
less than or equal to one quarter of the Airy disc can be treated as a point, thereby
establishing a criterion for the size of a pinhole. A pinhole camera is also considered, and
a relationship between the radius of the pinhole and camera length is obtained from
different points of view.
Finally, coherent imaging is discussed briefly, and expressions for the point-spread,
transfer, line-spread, and edge-spread functions are given. Both aberration-free and
aberrated systems are considered. The image of a coherently illuminated disc is
considered as an example of an extended coherent object. It is shown that the image
obtained is distorted compared to that of an incoherent disc. The Fourier-transforming
property of wave propagation is considered, and it is shown how the image of an object
illuminated by a plane wave can be altered by spatial filtering of its spectrum obtained in
the focal plane of the imaging lens. The chapter ends with a brief comparison of coherent
and incoherent imaging. In particular, the Rayleigh criterion of two-point resolution is
discussed.
Xi
R -------:;~
--~~~-=~======~~--------~~~------~z
Yi
Yg
- - - - - - - - - Zg - - - - - - - - - - ;... ~I
Figure 2-1. Geometry of imaging indicating an optical system with a circular exit
pupil of radius a forming an image in a plane at a distance Zg ~ R, where R is the
radius of curvature of the spherical wave emerging from the exit pupil with its
center of curvature at the Gaussian image point Pg having a position vector rg • The
position vector of the observation point in the Gaussian image plane is ~ with
respect to Pg • (For a complete schematic of imaging, see Figure 1-2.)
located at a point Pg with a position vector r;" = M r" in the image plane, where M is the
magnification of the image. The solid angle subtended by the entrance pupil at the point
object is given by Sell! dli .
Let 11 be the transmission factor of the system for light
propagation from its entrance pupil to its exit pupil.
Assuming that the exit pupil is uniformly illuminated and aberration free, a spherical
wave originating at the point object and incident on the system emerges from its exit
pupil as a spherical wave converging to the Gaussian image point Pg . The irradiance at a
point ~ in the Gaussian image plane with respect to the Gaussian image point r;" is given
by Eq. (1-65):
(2-1)
where
(2-2a)
is the total power in the exit pupil, and, therefore, in the image; R is the radius of
curvature of the spherical wavefront at the exit pupil passing through its center and
centered at the Gaussian image point;
Sex = p a 2 (2-4)
r
is the area of the exit pupil over which the integration in Eq. (2-1) is carried out; and rp is
r
the position vector of a point in its plane such that rp £ a . The quantity F in Eq. (2-2b)
is given by
F = R D (2-5)
and represents the focal ratio (f-number) of the image-forming light cone exiting from the
exit pupil. The aberration-free properties of an optical system, such as the irradiance
distribution given by Eq. (2-1), are often referred to as its diffraction-limited properties. A
uniformly illuminated pupil is called an unapodized pupil. If we let
r
(
r p = r p cos q p , sin q p ) , 0 £ rp £ a , 0 £ q p £ 2p , (2-6a)
and
r (2-6b)
ri = ri (cos q i , sin q i ) , 0 £ q i < 2 p ,
(
= r cos q p , sin q p ) , (2-8b)
r r
r = ri l F (2-9a)
and
r r
I (r ) = Ii (ri ) Ii (0) , (2-10)
1 2p 2
I ( r, q i ) = p 2 Û Û
[ (
Ù Ù exp - pir r cos qp - q i r dr d qp
ı ı
)] . (2-11)
0 0
Noting that1
2p
Û
Ù exp (ix cos a ) da = 2 p J0 ( x ) , (2-12)
ı
0
◊
where J 0 ( ) is the zero-order Bessel function of the first kind, Eq. (2-11) reduces to
1 2
ÈÛ ˘
I (r ) = 4 ÍÙ J 0 ( p r r) r dr˙ . (2-13)
ÍÎı ˙˚
0
◊
where J1 ( ) is the first-order Bessel function of the first kind, Eq. (2-13) becomes
[
I ( r ) = 2J1 (p r ) p r ]2 . (2-15)
We note that the irradiance distribution is radially symmetric about the Gaussian image
point r = 0 ; i.e., the irradiance at a point (r, q i ) in the image plane is independent of q i .
This is a consequence of the fact that the pupil function for the aberration-free circular
exit pupil is radially symmetric. Indeed, it is uniform across the pupil. From Eq. (1-59a),
we note that
r r
PSF(ri ) = Ii (ri ) Pex . (2-16)
Figure 2-2a shows the radially symmetric irradiance distribution in the image plane
as described by Eq. (2-15). It is also shown in a pictorial form in Figure 2-2b. This figure
is referred to as the Airy diffraction pattern for a circular aperture.2 We note that the
principal maximum of the irradiance distribution occurs at the Gaussian image point r =
0. Since the system is aberration free, all of the Huygens’ spherical wavelets originating
at the spherical wavefront at the exit pupil arrive in phase at this point. Therefore, they
interfere constructively, thereby yielding a principal maximum at this point. Note that
È 2 J ( x) ˘
Limit Í 1 ˙ = 1 . (2-17)
x Æ 0 Î x ˚
The minima of the irradiance distribution are zero at the positions given by the roots of
86 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0
0.8
P
I(r), P(rc)
0.6
0.4
0.2 I
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
r, rc
(a)
(b)
Figure 2-2. (a) Irradiance and encircled power distributions for an aberration-free
system with a circular pupil. (b) 2D PSF, called the Airy pattern.
2.2 Aberration Free System 87
J1 ( p r ) = 0, r π 0 . (2-18)
Noting that
d È J1 ( x ) ˘ J (x)
= - 2 , (2-19)
dx ÍÎ x ˙˚ x
J2 ( p r ) = 0, r π 0 , (2-20)
◊
where J2 ( ) is the second-order Bessel function of the first kind. The positions of several
minima and maxima of the irradiance distribution are given in Table 2-1. For increasing
values of r, the separation between two successive minima or two successive maxima
approaches a value of unity.
Substituting Eqs. (2-9a) and (2-10) into Eq. (2-21a) and defining a normalized or
fractional encircled power
Table 2-1. Irradiance and encircled power corresponding to the maxima and
minima of the PSF. The irradiance is normalized by the central value
Ii (0) = Pex Sex l 2 R 2 , and the encircled power is normalized by the total power Pex
in the exit pupil and, therefore, in the image. r and rc are in units of lF.
Max/Min r, rc I(r ) P( rc )
Max 0 1 0
Min 1.22 0 0.838
Max 1.64 0.0175 0.867
Min 2.23 0 0.910
Max 2.68 0.0042 0.922
Min 3.24 0 0.938
Max 3.70 0.0016 0.944
Min 4.24 0 0.952
Max 4.71 0.0008 0.957
88 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
we obtain
rc l F
(
P (rc ) = p 2 2 ) Û
Ù I (r ) rdr .
ı
(2-22a)
0
P (rc ) ( Û
)
= p 2 Ù I (r ) r dr .
2
ı
(2-22b)
0
we obtain3
Since, according to Eq. (2-18), the dark rings (minima of zero irradiance) correspond to
J1 ( p r ) = 0 , we note that the powers inside and outside an m-th dark ring of radius rm (in
units of l F ) are given by
respectively.
Figure 2-2a also shows how the encircled power P (rc ) varies with rc . The relative
irradiance and encircled power corresponding to maxima and minima of the irradiance
distribution are given in Table 2-1. We note that the radius of the first dark ring is 1.22 (in
units of lF ), and it contains 83.8% of the total power. The central bright spot of radius
1.22 is called the Airy disc of the Airy pattern. The first bright ring contains 7.2%, the
second bright ring 2.8%, and the third bright ring 1.4% of the total power.
ÛÛ
Pi (rs ) = ÙÙ Ii (ri ) ri dri d q i , (2-26)
ıı
det
where the integration is carried over the square region of the detector. Integration over q i
yields
2 rs
Û
Pi (rs ) = Ù Ii (ri ) q i (ri ) dri , (2-27)
ı
0
where q i (ri ) is the angle subtended by detector points lying at a distance ri from the
center. For ri £ rs , it is evident that q i (ri ) = 2 p , as in the case of a circular detector.
However, as indicated in Figure 2-3, only some of the points for which rs < ri £ 2 rs lie
on the square detector. For a given value of ri , the angle subtended by these points is
made up of four equal but discontinuous parts, one in each quadrant. Each part
contributes an amount ( p 2) - 2 cos 1 (rs ri ) to the angle q i ( ri ) . Hence, this angular
function may be written
q i (ri ) = 2 p , 0 £ ri £ rs (2-28a)
= 2 p - 8 cos 1
(rs ri ) , rs < ri £ 2 rs . (2-28b)
yi
ri
qi(ri)/4
xi
rs
2 rs 2 rs
Û Û
Pi (rs ) = 2 p Ù Ii (ri )ri dri - 8 Ù Ii (ri ) cos 1
(rs ri ) ri dri . (2-29)
ı ı
0 rs
The first term on the right-hand side of Eq. (2-29) represents the image power contained
in a circle of radius of 2 rs , and the second term gives the power contained between a
circle of radius 2 rs and a square of half-width rs . Substituting Eqs. (2-9), (2-10), and
(2-15) into Eq. (2-29), the ensquared power may be written
Ps (rs ) = Pc ( )
8Û
2 rs - Ù J12 ( p rs u) cos
pı
1
(1 u)
du
u
, (2-30)
1
where
and
Xi = Pex - Pi , (2-32)
where Pi is the corresponding included power. Equation (2-25b) gives the excluded
power for an m-th dark ring of the Airy pattern. For arbitrary sizes, the excluded power
can be calculated quite accurately in a closed form, if the included area is large enough so
that Xi £ 0.1 Pex .
For large arguments, we can use the asymptotic expression for Bessel functions,
namely, that1
J1 ( z ) ~ (2 pz )1 2 sin ( z - p 4) . (2-33)
1.0
I(r)
Ps
Pc
(r) Pc (rc) Ps (rs) 0.8
0.6
10(Ps – Pc)
0.4
0.2
0.0
0 1 2 3 4
r, rc, rs
Figure 2-4. Encircled and ensquared power distributions for a circular pupil. The
irradiance distribution and the difference between ensquared and encircled power
distributions are also shown.
Noting that the average value of a sine square is half, the average irradiance (indicated
by a bar) for large values of r is given by
I (r ) ~ 4 p4r 3 . (2-35)
For a large circular detector of radius rc , following Eq. (2-22b), the excluded power
in units of Pex may be written
Xc (rc ) ~ ( Û
)
p 2 Ù I (r ) r dr
2
ı
rc
= 2 p 2 rc . (2-36)
Similarly, for a large square detector of half-width rs , the excluded power is given by
92 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Table 2-2. Irradiance and encircled and ensquared powers for a circular pupil. r is
the radial distance of a point in the image plane from the Gaussian image point in
units of lF. rc and rs represent in units of lF the radius and half-width of a
circular and square detector, respectively, centered in the image plane at the
Gaussian image point.4
r , rc , rs I(r ) Pc ( rc ) Ps ( rs )
Xs (rs ) ~ Û Û
Ù dx Ù dyI (r )
ı ı
x > rs y > rs
= 4 2 p 3rs , (2-37)
( )
12
where r = x 2 + y 2 . From Eqs. (2-36) and (2-37) we note that
When rc or rs is greater than 1.6, Eqs. (2-36) and (2-37) give the excluded power
accurate to within 1% of the total power, as may be seen by comparing the results given
in Table 2-2.
The approximate result of Eq. (2-35) and those that follow from it, although obtained
for the aberration-free case, are valid even when aberrations are present in the system.
[This may be seen by substituting Eq. (2-46) given later into Eq. (1-154) and considering
the normalizations used in these equations.] It should be noted, however, that the value of
r for which Eq. (2-35) is valid increases as aberrations are introduced into the system.
Moreover, if the aberration is rotationally nonsymmetric, the irradiance distribution in an
image plane is not radially symmetric, but the first term of its asymptotic expansion is
symmetric and given by Eq. (2-35). Similarly, Eq. (2-36) may be used to obtain the
excluded power even when aberrations are present. However, the value of rc for which it
is valid increases as aberrations are introduced into the system.
r r r r r
Û
( ) (
t (vi ) = Pex 1 Ù A rp A rp - l R vi d rp
ı
) , (2-39)
where
(r )
A rp = ( Pex Sex )
12
,
r
rp £ a ,
= 0 , otherwise . (2-40)
It is evident that the OTF is radially symmetric and represents the fractional area of
overlap of two circles, each of radius a, separated by a distance l Rvi .
From Figure 2-5, we note that the area of overlap is given by four times the
difference between the area of a sector of radius a and cone angle b and the area of the
triangle OAB, i.e.
94 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
a
b
O
A O¢
lRni
Figure 2-5. Aberration-free OTF as the fractional area of overlap of two circles
whose centers are separated by a distance lRvi .
r r r r Ê b ˆ
Û
ı
( ) ( )
Ù A rp A rp - l Rvi d rp = 4( Pex Sex ) ÁË
2 p
1
p a 2 - OA ◊ AB˜
2 ¯
where
cosb = OA OB
= l R vi 2 a
= vi (1 l F ) . (2-42)
Let
v = vi (1 l F ) (2-43)
be a normalized radial spatial frequency. Thus, cos b = v . Hence, substituting Eq. (2-41)
into Eq. (2-39) and writing b in terms of v, we obtain
(
t (v) = (2 p) Ècos 1 v - v 1 - v 2 )
1 2˘
, 0 £ v £1 ,
ÍÎ ˙˚
= 0 , otherwise . (2-44a)
1.0
0.8
0.6
t
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
n
v t (v )
0 1.000
0.1 0.873
0.2 0.747
0.3 0.624
0.4 0.505
0.5 0.391
0.6 0.285
0.7 0.188
0.8 0.104
0.9 0.037
1.0 0.000
According to Eqs. (1-83) and (1-84), the PSF and the OTF of an optical system with a
radially symmetric pupil function are related to each other by a zero-order Hankel
transform. It can be shown that the PSF given by Eq. (2-15) and the OTF given by Eq. (2-
44) for a system with a circular pupil are related to each other by such a transform. We
also note that
1
Û
Ù t (v) v dv = 1 8 . (2-45)
ı
0
t ¢( 0 ) = - 4 p . (2-46)
As pointed out in Section 1.6.4, the slope given by Eq. (2-46), although obtained from the
aberration-free OTF given by Eq. (2-44) is aberration independent. Equation (2-46) can
also be obtained from Eq. (1-167) by noting that t ¢(0, f) is independent of f , v is in
units of 1 l F , L = 2 p a , and Sex = p a 2 .
where I ( r ) is given by Eq. (2-15). The corresponding OTF is given by Eq. (2-44). To
relate the two, we write, for example, Eq. (1-83) in terms of the irradiance distribution.
Thus,
Û
t (vi ) = (2 p Pex ) Ù Ii (ri ) J 0 (2 pvi ri ) ri dri . (2-48)
ı
Substituting Eqs. (2-9a), (2-10), and (2-43), Eq. (2-48) may be written
Û
t (v) = ( p 4) 2 p Ù I (r ) J 0 (2 pvr ) rdr . (2-49a)
ı
Similarly,
Û
I (r ) = ( 4 p) 2 p Ù t (v) J 0 (2 pvr )v dv . (2-49b)
ı
The functions I (r ) and t (v) given by Eqs. (2-15) and (2-44) are related to each other,
according to Eqs. (2-49a) and (2-49b), by slightly modified zero-order Hankel transforms.
These modifications consist of the factors of p 4 and 4 p on the right-hand side of these
equations because of the particular normalizations used; namely, I (0) = 1 and t (0) = 1 .
Û
= 2 p rc Ù t (v) J1 (2 prc v) dv . (2-50)
ı
Equations (2-49a), (2-49b), and (2-50) also hold for a radially symmetric aberration, as
may be seen by comparing with Eqs. (1-83), (1-85), and (1-89), respectively.
2.3 Strehl Ratio and Aberration Tolerance 97
1 2p 2
I (r , q i ) = p 2 Û Û
[ (
Ù Ù exp i F r, qp
ı ı
)] exp[- pi rr cos (q p )]
- q i r dr d q p , (2-51)
0 0
( ) (
where F r, q p is the phase aberration of the system at a point r, qp in the plane of its )
exit pupil. By definition, the Strehl ratio of the image or the system is given by the ratio
of the irradiances at the center r = 0 with and without aberrations. The aberration-free
central irradiance is unity in units of Pex Sex l2 R 2 [as may be seen from Eqs. (2-2a) and
(2-51)]. Hence, the central irradiance obtained from Eq. (2-51) is the Strehl ratio, i.e.,
1 2p 2
Û Û
S = p 2
ı ı
[
Ù Ù exp i F(r, q) r dr dq ] , (2-52)
0 0
where we have dropped the subscript p on the angle q p for simplicity. Approximate
expressions for the Strehl ratio when the aberration is small are given by Eqs. (1-204)
through (1-206), i.e.,
~ (1 - s 2F 2)
2
S1 , (2-53a)
S2 ~ 1 - s 2F , (2-53b)
and
S3 ~ exp (- s 2F ) , (2-53c)
where
is the variance of the phase aberration across the uniformly illuminated pupil. The mean
and the mean square values of the aberration are obtained from the expression
1 2p
Û 1Û
< F > = p Ù Ù F n (r, q) r dr dq ,
n
(2-55)
ı ı
0 0
Table 2-4 gives the form as well as the standard deviation s F of a primary (or a
Seidel) aberration, where an aberration coefficient Ai represents the peak value of the
aberration. It also lists the tolerance, i.e., the value of the aberration coefficient Ai , for a
Strehl ratio of 0.8. This tolerance has been obtained by using the Strehl ratio expression
S2 , according to which the standard deviation for a Strehl ratio of 0.8 is given by
sF = 0.2 (2-56a)
or
where s w is the standard deviation of the wave aberration. The aberration tolerance
listed in Table 2-4 is for the wave (as opposed to the phase) aberration coefficient, as is
customary in optics. We have used the symbol Ad for the coefficient of field curvature
aberration, which varies quadratically with the angle a point object makes with the optical
axis of the system. However, to avoid confusion, we will use the symbol Bd for
representing the defocus wave aberration, which is independent of the field angle but has
the same dependence on pupil coordinates as field curvature. Similarly, we have used the
symbol At for distortion, which varies as the cube of the field angle. However, we will
use the symbol Bt to represent wavefront tilt, which is independent of the field angle but
has the same dependence on pupil coordinates as distortion.
2.3 Strehl Ratio and Aberration Tolerance 99
Table 2-4. Standard deviation and aberration tolerance for primary aberrations.
As r 4 2 As As l 4.19
Spherical =
3 5 3.35
Ac r3 cos q Ac Ac l 4.96
Coma =
2 2 2.83
Ad r2 Ad Ad l 4.06
Field Curvature =
(defocus) 2 3 3.46
Based on diffraction, the best image for small aberrations is one for which the Strehl
ratio is maximum. Since, according to each of Eqs. (2-53a), (2-53b), and (2-53c), the
Strehl ratio is maximum when the aberration variance is minimum, the best image plane
is one that minimizes the variance of the aberration. Thus, for example, we balance
spherical aberration with defocus and write it as
F(r) = As r 4 + Bd r2 . (2-57)
∂ s 2F
= 0 . (2-58)
∂ Bd
Astigmatism and coma aberrations can be treated similarly. Table 2-5 lists the
form of a balanced primary aberration, its standard deviation, and its tolerance for a
Strehl ratio of 0.8 according to Eq. (2-53b). Also listed in the table is the location of the
diffraction focus, i.e., the point with respect to which the aberration variance is minimum
so that the Strehl ratio is maximum at it. We note that in the case of coma, the balancing
aberration is a wavefront tilt with a coefficient that is minus two-thirds of the coma
coefficient. Thus, the maximum Strehl ratio is obtained at a point that is displaced from
the Gaussian image point by 4 FAc 3 but lies in the Gaussian image plane. It is shown in
Section 2.9.2.2 that this is valid for Ac < ~ 0.7 l . For larger values of Ac , the peak
irradiance lies at a point that is closer to the Gaussian image point than that given by
4 FAc 3 . The standard deviation for balanced coma is 1/3 of its corresponding value
when not balanced. In the case of astigmatism, the best Strehl ratio is obtained in the
plane of the circle of least confusion, i.e., at a distance 4 F 2 Aa from the Gaussian image
plane. The standard deviation of balanced astigmatism is only 1/1.225 of its
corresponding value for classical astigmatism (i.e., without any balancing).
1.00
0.75 W(r)
= r4 +(Bd /As)r2
As
0.50
Bd
=0
0.25 As
W(r)
As
0.00
–1
– 0.25
–1.5
– 0.50
–2
– 0.75
– 1.00
0.0 0.2 0.4 0.6 0.8 1.0
r
Figure 2-7. Variation of spherical aberration with r when balanced with various
amounts of defocus.
2.3 Strehl Ratio and Aberration Tolerance 101
Spherical (
As r 4 - r2 ) (0, 0, 8F A )
2
s
As
6 5
0.955l
( )
Ac
Coma Ac r3 - 2r 3 cos q (4 FAc 3, 0, 0 ) 0.604l
6 2
Astigmatism (
Aa r2 cos 2 q - 1 2 ) (0 , 0 , 4 F A )
2
a
Aa
2 6
0.349l
= ( Aa 2) r2 cos 2q
*The diffraction focus coordinates are relative to the Gaussian image point.
Figure 2-8 shows how the Strehl ratio of a primary aberration varies with its
standard deviation. Approximate as well as exact results are shown in this figure. The
curves for a given aberration and for the corresponding balanced aberration can be
distinguished from each other by their behavior for large s w values (near 0.25 l ). For
example, coma is shown by the evenly dashed curves; the lower curve being for balanced
coma. The same holds true for astigmatism. The curves for spherical and balanced
spherical aberrations are identical since the Strehl ratio for a given value of s w is the
same for the two aberrations. The following observations may be made from Figure 2-8.
i. For small values of s w , the Strehl ratio is independent of the type of aberration. It
depends only on the value of its variance, as illustrated by Eqs. (2-53a) through (2-
53c).
ii. The expressions for S1 and S2 underestimate the true Strehl ratio.
iii. The expression for S3 underestimates the true Strehl ratio only for coma and
astigmatism; it overestimates for the other aberrations. The error, defined as
100 (1 - S3 S ) , is £ 10% as long as S > 0.3 .
102 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Table 2-6. Primary aberrations, their peak and peak-to-valley values, and standard
deviations; and the exact expression for the Strehl ratio and values of the aberration
coefficient, peak absolute aberration, and peak-to-valley aberration for a Strehl
ratio of 0.8.
S = 0.8
( b ) + S ( b )˘˙˚
2 As 1È 2 †
Spherical As r 4 C 2 As As 0.25 0.25 0.25
3 5 b ÍÎ
( b ) + S ( b )˘˙˚
Balanced As 1È 2 †
spherical As (r 4 – r 2 ) C 2 As / 4 A s / 4 1 0.25 0.25
6 5 b ÍÎ
Ú ( ) ˚˙
2
Ac È J 2 2s x dx ˘
1 3/ 2
Coma Ac r 3 cos q Ac 2Ac 0.21 0.21 0.42
2 2 ÎÍ 0 0 F
2
Balanced Ac ÏÔ 1 Ê 3/2 2 1/2 ˆ ¸Ô
Ac (r 3 – 2r / 3) cos q Ì Ú0 J 0 [6 2s F Á x – x ˜ ]dx ˝ Ac / 3 2 Ac / 3 0.63 0.21 0.42
coma 6 2 ÔÓ Ë 3 ¯ Ô˛
b b
† b = 3 5 s F / p , C (b) = Ú cos(p x
0
2 / 2)dx, S (b) = Ú sin(p x
0
2 / 2) dx
iv. S3 gives a better approximation for the true Strehl ratio than S1 and S2 . The reason is
that, for small values of s w , it is larger than S1 by approximately s F
4
4 . Of course,
S1 is larger than S2 by s F 4 .
4
v. The Strehl ratio depends strongly on the standard deviation of an aberration but
weakly on its detailed distribution over a wide range of Strehl ratio values and not
just for large values of it.
1.0
0.8
0.6
S
0.4
S3
0.2
S2
S1
0.0
0.00 0.05 0.10 0.15 0.20 0.25
sw
Figure 2-8. Strehl ratio for primary aberrations as a function of their standard
deviation sw in units of optical wavelength l . For large values of sw , coma and
astigmatism give a higher Strehl ratio than the corresponding balanced aberration.
The Strehl ratios for spherical and balanced spherical aberrations for the same
value of sw are identical. sF = (2 p l) s w . Spherical or balanced spherical....,
Coma– – – –, Astigmatism __.__.__.
{[
S = ( p 2 As ) C(b+ ) + C(b )]
2
[
+ S(b+ ) + S(b )]
2
} , (2-59a)
where
104 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
b± = ( As ± D ) (2 p As )
12
, D = As + Bd , (2-59b)
◊ ◊
As and Bd are in radians, and C( ) and S( ) are the Fresnel integrals defined in Table 2-6.
It is evident that the Strehl ratio is independent of the sign of D, the deviation of
defocus from its optimum (in the sense of minimum variance) value. Thus the axial
irradiance of a focused wave aberrated by spherical aberration is symmetric about the
axial point with respect to which the aberration variance is minimum. Figure 2-9 shows
how S varies with D in the case of circular pupils for several typical values of As . It is
seen that, for large Strehl ratios, S is maximum when D = 0 , i.e., a minimum variance
leads to a maximum of Strehl ratio. For small Strehl ratios, however, minimum variance
gives a minimum of Strehl ratio. The Strehl ratio is maximum for a nonoptimally
balanced aberration. For example, when As = 3l , the optimum amount of defocus is
Bd = - 3l , but the Strehl ratio is a minimum and equal to 0.12. The Strehl ratio is
maximum and equal to 0.26 for Bd ~ - 4l or - 2l . For As < ~ 2.3l , the axial irradiance
is maximum at a point with respect to which the aberration variance is minimum.
0.8
0.6
S
As=1l
0.4
2l
0.2
3l 4l
0.0
0 1 2 3 4
D
Figure 2-9. Strehl ratio for circular pupils aberrated by spherical aberration as a
function of the deviation of focus from its optimum balancing value ( D = As + Bd )
for several values of the aberration coefficient As . The curves are symmetric about
the origin.
2.4 Balanced Aberrations and Zernike Circle Polynomials 105
where Bt is the peak value of the tilt aberration. We note from Table 2-6 that
Bt = - 2 Ac 3 for minimum variance. Thus, the point with respect to which the aberration
variance is minimum lies in the image plane at a distance of 4 FAc 3 from the origin,
where Ac is in length units. As discussed later in Section 2.9.2.2, we find that maximum
irradiance occurs at this point only if Ac < ~ 0.7l , which in turn corresponds to
S >
~ 0 . 76 . For larger values of Ac , the distance of the point of maximum irradiance does
not increase linearly with its value, and even fluctuates in some regions. Thus, only for
large Strehl ratios is the irradiance maximum at the point associated with minimum
aberration variance. Moreover, it is found that for Ac > 2.3l , the Seidel coma gives a
larger Strehl ratio than the balanced coma; i.e., the irradiance at the origin is larger than at
the point with respect to which the aberration variance is minimum. An examination of
the axial irradiance for astigmatism (discussed later in Section 2.8.3) shows that the peak
irradiance lies at the point Bd = - Aa 2 corresponding to minimum aberration variance
only for Aa < l .
When secondary spherical aberration and secondary coma are balanced with lower-
order aberrations to minimize their variance, it is found that a maximum of Strehl ratio is
obtained only if its value comes out to be greater than about 0.5. Otherwise, a mixture of
aberrations yielding a larger-than-minimum possible variance gives a higher Strehl ratio
than the one provided by a minimum variance mixture.8
• n
F(r, q) = Â Â c nm Z nm (r, q) , 0 £ r £ 1 , 0 £ q £ 2p , (2-61a)
n =0 m =0
where cnm are the orthonormal expansion coefficients that depend on the field angle of
the object, n and m are positive integers including zero, n – m ≥ 0 and even, and
[
Z nm (r, q) = 2( n + 1) (1 + d m 0 ) ]1/ 2Rnm (r) cos mq . (2-61b)
(n m) / 2
(-1)s (n - s)!
Rnm (r) = Â n+m ˆ Ên-m ˆ
rn 2s
(2-61c)
s=0 s! Ê -s ! -s !
Ë 2 ¯ Ë 2 ¯
1 1
Ú Rn (r) Rn ¢ (r) r dr = d
m m
. (2-61d)
0 2(n+ 1) nn ¢
1 1 2p m
Ú Ú Z (r, q)Z n ¢ (r, q) r dr d q = d nn ¢ d mm ¢
m¢
. (2-61f)
p0 0 n
as may be seen by substituting Eq. (2-61a) and utilizing the orthonormality of the
polynomials.
The Zernike circle polynomials are unique in that they are the only polynomials in
two variables r and q, which (a) are orthogonal over a circle, (b) are invariant in form
with respect to rotation of the coordinate axes about the origin, and (c) include a
polynomial for each permissible pair of n and m values.12
The orthonormal Zernike polynomials and the names associated with some of them
when identified with aberrations are listed in Table 2-7 for n £ 8. The polynomials
independent of q are the spherical aberrations, those varying as cos q are the coma
aberrations, and those varying as cos 2q are the astigmatism aberrations. The number of
Zernike (or orthogonal) aberration terms in the expansion of an aberration function
through a certain order n is given by
2
Nn = Ê + 1ˆ
n
for even n , (2-63a)
Ë2 ¯
= (n + 1)(n + 3) 4 for odd n . (2-63b)
n (r, q) = c nm Z n (r, q) .
Fm m
(2-64)
= 0 , n π 0, m π 0 . (2-65a)
For m = 0 , this may be seen with the help of Eq. (2-61d) and the fact that R00 (r) = 1 is a
member of the polynomial set. The orthogonality Eq. (2-61f) yields the result that the
2.4 Balanced Aberrations and Zernike Circle Polynomials 107
0 0 1 Piston
2 0 (
3 2r 2 - 1 ) Field curvature (defocus)
2 2 6 r cos 2q
2
Primary astigmatism
3 1 (
8 3r3 - 2r cos q ) Primary coma
3 3 8 r3 cos 3q
4 0 (
5 6r 4 - 6r2 + 1 ) Primary spherical
4 2 (
10 4r - 3r 4 2
) cos 2q Secondary astigmatism
4 4 10 r cos 4q 4
5 1 (
12 10r5 - 12r3 + 3r cos q ) Secondary coma
5 3 12 (5r 5
- 4r 3
) cos 3q
5 5 12 r5 cos 5q
6 0 (
7 20r6 - 30r 4 + 12r2 - 1 ) Secondary spherical
6 2 (
14 15r6 - 20r 4 + 6r2 cos 2q ) Tertiary astigmatism
6 4 14 (6r 6
- 5r 4 cos 4q)
6 6 14 r6 cos 6q
7 1 ( )
4 35r7 - 60r5 + 30r3 - 4r cos q Tertiary coma
7 3 4 (21r - 30r + 10r ) cos 3q
7 5 3
7 5 4 (7r - 6r ) cos 5q
7 5
7 7 4 r7 cos 7q
8 0 (
3 70r8 - 140r6 + 90r 4 - 20r2 + 1 ) Tertiary spherical
*The words “orthonormal Zernike” are to be associated with these names, e.g., orthonormal Zernike primary
astigmatism.
108 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
mean value of Rn0 (r) is zero. This indeed is the purpose of the constant term in Rn0 (r) ; it
makes the mean value zero. When m π 0 , the average value of cos mq is zero. Similarly,
the mean square value of the aberration is given by
1 2p 1 2p
[ (r, q)] Û Û
[ ] Û Û
2 2
F nm = Ù Ù F nm (r, q) r dr d q Ù Ù r dr d q
ı ı ı ı
0 0 0 0
= cnm
2
. (2-65b)
(F mn ) 2
2
s 2nm = - Fm
n
= cnm
2
, n π 0, m π 0 . (2-66)
Thus, each orthonormal expansion coefficient, with the exception of c00 , represents the
standard deviation of the corresponding aberration term. The variance of the aberration
function is accordingly given by
• n
s 2F = F 2 (r, q) - F(r, q)
2
= Â Â cnm
2
. (2-67)
n =1 m = 0
Unless the mean value of the aberration < F > = 0, s F π F rms , where F rms = < F 2 >1 / 2
is the root-mean-square (rms) value of the aberration. Substituting Eq. (2-67) into any of
Eqs. (2-53a) through (2-53c) yields the Strehl ratio for small aberrations.
= (
5c40 6r 4 - 6r2 + 1 ) . (2-68)
Comparing this with the balanced spherical aberration given in Table 2-5,
F bs (r) = As r 4 - r2 ( ) , (2-69)
s bs = c40
= As 6 5 , (2-70)
(
= 2 2 c31 3r3 - 2r cos q . ) (2-71a)
F bc (r, q) = Ac Ê r3 - rˆ cos q
2
(2-71b)
Ë 3 ¯
s bc = c31 = Ac 6 2 . (2-71c)
(
= 2 6 c22 r2 cos 2 q - 1/2 ) . (2-72a)
(
F ba (r, q) = Aa r2 cos 2 q - 1 2 ) (2-72b)
s ba = c22 = Aa 2 6 . (2-72c)
= (
3c20 2r2 - 1 ) . (2-73a)
F d (r) = Bd r2 (2-73b)
s d = c20 = Bd 2 3 . (2-73c)
110 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
s t = c11
= Bt 2 . (2-74c)
which represents a uniform (piston) aberration. Obviously, its standard deviation is zero
and has no effect on the Strehl ratio of a system with a single exit pupil. (This will not be
true in the case of a multiexit pupil system, as in a phased array system.)
Thus, we see that Zernike polynomials can be identified with balanced aberrations;
that, in fact, is their advantage. Here we have discussed only the primary aberrations. In
general, the aberration function of an optical system may consist of higher-order
aberrations. Moreover, in a system without an axis of rotational symmetry, the aberration
function will consist of terms not only in cos mq but in sin mq as well. This is true in the
case of aberrations introduced by atmospheric turbulence (see Chapter 5).
We have already seen that a quarter wave of defocus aberration yields a Strehl ratio
of 0.8 (see Table 2-4). How this defocus aberration translates into longitudinal defocus, or
depth of focus, depends on the Fresnel number of the focused image or beam. In this
section, we show that the depth of focus for a system with a large Fresnel number, such
as a photographic camera, is very small. However, it is quite large for a system with a
small Fresnel number, e.g., a laser transmitter focusing a beam on a distant target. We
also show that for such a system, maximum central irradiance is not obtained at the
geometrical focus but at a defocused point that is closer to the system. However, this does
not mean that a beam should be defocused to obtain maximum central irradiance on a
target at a given distance. Maximum central irradiance on a target is still obtained when a
2.5 Defocused System 111
beam is focused on it, even though a larger irradiance is obtained at a defocused point
that is closer to the focusing system. A far-field distance of the system is defined such
that a beam focused beyond it behaves practically like a collimated beam.
( )
Wd rp =
1 Ê1
2Ëz
- ˆ rp2 + Ê 3 - 3 ˆ rp4 + º ,
1
R¯
1 1
8Ëz
1
R ¯
(2-76)
where the first term on the right-hand side is a defocus aberration and the second term is
fourth-order spherical aberration. For small f-numbers, a better approximation is obtained
by dividing the sag difference by the cosine of the angle of the marginal ray in the image
space. For a circular exit pupil of radius a, if z is large enough that
Ê 1 - 1 ˆ a4 £ l ;
Ë z 3 R3 ¯
i.e., if
1 1 l
3 £ 3 + 4 , (2-77)
z R a
then the term in rp4 and those of higher order in Eq. (2-76) may be neglected. A spherical
aberration of l 8 gives a Strehl ratio of 0.946. When this aberration is balanced with an
equal and opposite amount of defocus, the Strehl ratio increases to 0.996. Assuming the
validity of Eq. (2-77), Eq. (2-76) may be written
( )
Wd rp =
1 Ê1
2Ëz
- ˆ rp2
1
R¯
. (2-78)
Moreover, the quantity l0 in the Huygens’ spherical wavelet of Eq. (1-18) or l in Eq.
(1-20) is now approximately equal to z . Hence, the irradiance distribution in a defocused
image plane at a distance z from the pupil plane may be written
112 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
r r Ê 2pi r r ˆ r
2
I i (ri ) = (Pex Sex l z
2 2
) Û
{[ ( )
Ù exp i F r p + F d r p
ı
( )] } exp Á - ◊
r r ˜dr
Ë lz p i ¯ p
, (2-79)
r
where Ii (ri ) represents the irradiance at a point whose position vector in the observation
r
plane is ri with respect to the center of curvature of the new reference sphere, and
( )
F d rp = (2p l )Wd rp ( ) (2-80)
(r )
is the defocus phase aberration. It should be clear that F rp is the phase aberration at a
r
point rp in the plane of the exit pupil with respect to a reference sphere centered at the
r r
Gaussian image point. Dividing Ii (ri ) by Pex , we obtain PSF(ri ) . Equation (2-79) may
r
also be obtained from Eq. (1-51) by regarding ri as stated above and multiplying both
sides by Pex .
Following the procedure used to obtain Eq. (2-11), we find that Eq. (2-79) for the
defocused irradiance distribution may be written for an otherwise aberration-free
r
[( ) ]
F rp = 0 system
1 2
I ( r; z ) = (2 R z )
2 Û
ı
( 2
)
Ù exp i Bd r J 0 ( p rr R z ) r dr , (2-81a)
0
where
Bd ( z ) = p N Ê - 1ˆ
R
(2-81b)
Ëz ¯
N = a2 l R (2-81c)
is the Fresnel number of the exit pupil as observed from the focal point. As in Eq. (2-11),
r
r = a 1 rp , r is in units of l F , and the irradiance is in units of the aberration-free central
irradiance Pex Sex l2 R 2 . In Eq. (2-81a), we have already carried out the angle integration
following Eq. (2-12).
From Eq. (2-81a), we note that the irradiance distribution is asymmetric about the
Gaussian image plane; i.e., the irradiance distributions in observation planes located at
z = R ± D , where D is a longitudinal defocus, are not identical. There are three reasons
for this asymmetry. First, the inverse square law dependence on z increases I (r ) for
z < R and decreases it for z > R . Second, Bd is asymmetric since
Bd ( R + D ) π - Bd ( R - D ) . (2-82)
However, if N is very large (>> 10), Bd becomes large even for very small
differences in z and R. In that case, the defocus tolerance dictates that z be
approximately equal to R. Hence, Eq. (2-81b) may be written
( )
Bd = p 4l F 2 ( R - z ) , (2-83a)
Bd ( R + D ) = - Bd ( R - D ) . (2-83b)
Now, according to Eq. (2-81a), the irradiance distribution is independent of the sign of
Bd . Hence, for z ~ R , the distribution is symmetric about the Gaussian image plane.
For small N (£ 10) , z can be much different from R for Bd to achieve a significant
value. In this case, therefore, all three factors contribute to the asymmetry of the
irradiance distribution about the Gaussian image plane. One consequence of this is that
the irradiance on and near the axis can be higher for z < R than that for z > R .
I (0; z ) = ( R z ) S ,
2
(2-84a)
or
( )
Ii (0; z ) = Pex Sex l2 z 2 S , (2-84b)
where
[
S = sin ( Bd 2) ( Bd 2) ]
2
(2-84c)
is a Strehl ratio defined as the ratio of the axial irradiance at a distance z when the beam is
focused at a distance R to that when it is focused at a distance z . The axial irradiance is
minimum and equal to zero when Bd is an integral multiple of 2p; i.e., when the peak
defocus wave aberration is an integral multiple of l. The corresponding z values are
given by
R z = 1 + 2n N , (2-85)
The maxima of axial irradiance can be determined by equating to zero the derivative
of Ii (0; z ) of Eq. (2-84b) with respect to z . We find that they occur at z values given by
the solutions of the transcendental equation
tan ( Bd 2) = ( R z ) Bd 2 , z π R . (2-86)
We note from Eqs. (2-84a) and (2-84c) that, since S £ 1, I (0; z ) £ 1 for z ≥ R .
However, for z < R , an increase in axial irradiance due to the inverse square law
competes with a decrease due to the defocus aberration (nonconstructive interference of
Huygens’ spherical wavelets). Which effect dominates depends on the Fresnel number N.
When N is very large, a slight amount of longitudinal defocus produces a large amount of
defocus aberration Bd and correspondingly a small Strehl ratio S . Hence, for z < R , an
increase in axial irradiance due to the inverse square law is overcome by the loss due to
the destructive interference of Huygens’ spherical wavelets. Accordingly, in such cases,
I (0; z ) < 1 even when z < R . When N is very small, z can be much different from R
before Bd becomes significant. Hence, over a considerable range of z < R , ( R z ) > S 1
2
so that I (0; z ) > 1 . For example, S = 0.8 is obtained at R z = 1.52 when N = 1, giving
I (0; z ) = 1.84 at this z value. When N = 10 , S = 0.8 is obtained at R z = 1.05 so that
I (0; z ) = 0.88 . Similarly, when N = 100 , R z = 1.005 for S = 0.8 , and, accordingly,
I (0; z ) is practically the same as S.
The range of distance z over which I (0; z ) > I (0; R) depends on the Fresnel number.
Let z0 be the distance at which I (0; z0 ) = I (0; R) , so that for z0 < z < R ,
I (0; z ) > I (0; R) . It is evident that z0 lies between the focus and the first minimum to the
left of it; i.e., 1 < R z0 < 1 + 2 N . As N increases, z0 Æ R ; i.e., the range of the distance
over which I (0; z ) > I (0; R) reduces to zero. For N < 0.25 , i.e., for R > D2 l , a focused
beam propagates very much like a collimated beam, as discussed later in Section 2.5.3.
1.0 10
zp/R
0.8 8
z0/R
0.6 6
z0 /R zp /R
(zp)/ (R)
0.4 4
0.2 2
I(zp)/I(R)
0.0 0
0 2 4 6 8 10
N
( )
Figure 2-10. Variation of z 0 , z p , and I z p with N. z 0 is the minimum distance of an
axial point from the aperture so that I ( z 0 ) = I ( R) = p 2 N 2 I0 , where I0 is the pupil
irradiance. z p is the distance of the location of the principal maximum. I z p is the ( )
irradiance at this point.
to the focal-point irradiance over a significant range of z values when N < ~ 3 . We note
that z0 R <~ 0 . 92 when N <
~ 5. When N <
~ 3 , z 0 R <
~ 0 . 8 . As N Æ 0 , corresponding
( )
to a collimated beam, R Æ • . Therefore, I 0; z p I (0; R) Æ • , z0 R Æ 0 , and
zp R Æ 0 .
Figure 2-11 shows how the axial irradiance varies for beams with N = 1, 10, and
100. We note that it is highly asymmetric about the focal point when N = 1, but it
becomes more and more symmetric as N increases. Figure 2-12, which shows the axial
irradiance for N = 1 labeled as S ¢ , also shows how the peak defocus aberration
Wm = (l 2 p) Bd ( z ) varies with z. It is evident that Wm , shown in units of l , is not
symmetric about the focal point. In particular, we note that when N = 1, the principal
maximum of axial irradiance lies at z = 0.6 R and not at the focal point z = R . The value
of Wm for this point is only l 3 and reduces the irradiance to 68% of the focal-point
value. However, the inverse square law increases the irradiance by a factor of
approximately 2.78. Hence, the net irradiance at this point is 1.9 times the focal-point
irradiance Pex Sex l2 R 2 .
The dotted curve in Figure 2-12 shows how the axial irradiance at a distance z varies
with z if the beam is focused at this distance. Considering that the irradiance in this figure
is in units of Pex Sex l2 R 2 , this curve represents the variation of ( R z ) as a function of
2
z R. The ratio of the solid and dotted curves gives the Strehl ratio S shown in this figure
as a function of z R. It should be evident that S = 1 when z = R .
116 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
lg
0.5 0.5 0.5
2.0 1.00
S¢
S
1.5 0.75
S¢ ÔWmÔ (R/z)2
1.0 0.50
(R/z)2
0.5 0.25
ÔWmÔ
0.0 0.00
0.0 0.5 1.0 1.5 2.0
z/R
Now, we ask the question: Given a target at a distance z , how should a beam be
focused on it so that the central irradiance on it is maximum? Thus, we would like to
know the optimum value of R. The answer is that we should choose R = z , as may be
seen by differentiating Eq. (2-84b) with respect to R and equating the result to zero. It is
evident that the central irradiance on a target when the beam is focused on it is
Pex Sex l2 z 2 , which is 1.47 times the corresponding irradiance when the beam is focused
at a distance R = 0.6 z .
Figure 2-13 illustrates how the central irradiance on a target at a fixed distance z
varies when the beam is focused at various distances R along its axis. The irradiance in
this figure is in units of Pex Sex l2 z 2 . The quantity
Nz = a2 l z (2-87)
represents the Fresnel number of the circular exit pupil as observed from the target. We
note that as N z increases, the curves become symmetric about the point R = z . The
encircled power in a defocused image plane is given by
rc
( Û
)
P(rc ; z ) = p 2 Ù I (r; z ) rdr .
ı
2
(2-88)
0
Numerical calculations of integrals in Eqs. (2-81) and (2-88) are discussed in Section 2.6.
Figure 2-14a shows for N = 1 the irradiance distribution and encircled power in the
focal plane ( z = R) and in the planes z = (1 ± 0.4) R . We note that the power is
concentrated most in the plane z = 0.6 R . However, for 0.65 < rc < 1.10, the power in this
plane is smaller than the corresponding value in the focal plane. The irradiance and
Figure 2-13. Central irradiance (in units of Pex Sex l 2 z 2 ) at a distance z from the
plane of the exit pupil when a beam is focused at various distances R. The quantity
N z = a 2 l z represents the Fresnel number of the exit pupil as observed from the
target. The dashed curves are for a Gaussian beam.
118 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
2.0 1.00
1.5 0.75
P(rc)
l(r)
1.0 0.50
z=R
z = 0.6 R
z = 1.4 R
0.5 0.25
0.0 0.00
0.0 0.5 1.0 1.5 2.0
r, rc
(a)
1.00
rc = 1.5
0.75 1.0
0.8
P(z rc)
0.50
0.6
0.25
0.4
0.2
0.00
0.4 0.6 0.8 1.0 1.2 1.4
z /R
(b)
Figure 2-14. (a) Focused and defocused irradiance distributions and corresponding
encircled powers for a circular focused beam with N = 1 . The irradiance is
normalized by the focal-point irradiance and encircled power is normalized by the
total power Pex . The units of r and rc are lF. (b) Encircled power in a circle of fixed
radius rc for a circular focused beam as a function of the axial distance z from the
exit pupil.
2.5 Defocused System 119
power distributions when the beam is focused at a distance z = 0.6 R is the same as for
z = R , except that the horizontal scale is changed by a factor of 0.6 so that r and rc are in
units of 0.6 l F . Thus, 84% of the power is contained in a radius of 0.6 l F instead of
l F , i.e., the power is much more concentrated.
Figure 2-14b shows how the power in a spot of certain radius rc (in units of lF )
varies with the axial distance z in the vicinity of the geometrical focus. Although the axial
irradiance has a principal maximum at z = 0.6 R , where its value is nearly twice the
focal-point irradiance, the encircled power is maximum at this z value only for small spot
sizes. For moderate spot sizes (0.3 £ rc £ 1), the maximum of encircled power occurs at a
z value that varies with rc but lies between the point of principal maximum and the focal
point. It is evident from this example that where the diffraction focus of a converging
spherical wave lies depends on the criterion used to define it. Is it the maximum of
irradiance or maximum of encircled power? Maximum encircled power on a target at a
given distance is obtained when the beam is focused on it. Thus, for a moving target, the
beam must be actively focused on it to obtain the maximum possible encircled power.
However, if the encircled power is adequate when the beam is focused on a target at a
certain distance so that the Fresnel number of the beam aperture as observed from the
focus is small < ( )
~ 5 , it is more than adequate over a considerable range of the target
distance without any active focusing of the beam.
The results for a collimated beam can be obtained from those for a focused beam by
letting R Æ • . Thus, for example, the peak defocus aberration given by Eq. (2-82)
becomes
Bd = p a 2 l z
= Sex l z . (2-89)
It represents the peak phase aberration of a plane wavefront with respect to a reference
sphere centered at a distance z from the exit pupil and passing through its center.
Equation (2-84b) for axial irradiance reduces to
(
I (0; z ) = 4 I0 sin 2 p a 2 2 l z ) , (2-90)
where
is the irradiance at the exit pupil. The axial irradiance is maximum and equal to 4 I0 at z
120 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
values given by
z = a 2 l (2 n + 1) , n = 0 , 1 , 2 , K . (2-92)
z = a 2 2 l (n + 1) , n = 0 , 1 , 2 , K . (2-93)
These z values for the location of maxima and minima correspond to those axial positions
at which the circular exit pupil subtends an odd or an even number of Fresnel’s half-
wave zones, respectively.
z ≥ D2 l , (2-94)
distance z in this figure is in units of D2 l , which is called the far-field distance of the
exit pupil. It is evident that a collimated beam yields practically the same irradiance on a
target lying in the far field of the exit pupil as a beam focused on it; in other words, beam
focusing does not significantly increase the power concentration on the target.
4.0 1.00
3.0 0.75
(0 z) (p/4z)2
I(z)
2.0 0.50
(p/4z)2
1.0 0.25
0.0 0.00
0.00 0.25 0.50 0.75 1.00
z
Figure 2-15. Axial irradiance of a collimated circular beam normalized by the exit
pupil irradiance I0 . The axial distance z is in units of the far-field distance D 2 l.
2.6 PSFs for Rotationally Symmetric Aberrations 121
The irradiance distribution in a plane at a distance z can be obtained from Eq. (2-81)
by letting R Æ • and noting that the units of irradiance in this equation are Pex Sex l2 R 2
and those of r are l F = l R D . Thus, for a collimated beam, we may write
1 2
Û
( )
I (r; z ) = 4 Ù exp i Bd r2 J 0 ( p rr) r dr
ı
, (2-95)
0
where the units of irradiance are Pex Sex l2 z 2 and those of r are lz D. The irradiance
distribution given by Eq. (2-95) is called the Fresnel diffraction pattern of a circular exit
pupil. For z ≥ D2 l , since the aberration is negligibly small ( Bd £ p 4 or l 8), Eq. (2-
95) reduces to the aberration-free result
I ( r; z ) ~ [2 J1 (p r ) p r]2 , z ≥ D2 l . (2-96)
Equation (2-94) is the far-field condition for a circular exit pupil and Eq. (2-96)
represents its far-field or Fraunhofer diffraction pattern. Except for the units of irradiance
and r, Eq. (2-96) is the same as Eq. (2-15).
( )
The irradiance distribution for small z values z < D2 l , called the near-field
diffraction pattern, is equivalent to a defocused Fraunhofer distribution. Therefore, using
our approximate model discussed in Section 2.5, we can quickly estimate the encircled
power for small z values. Thus, for example, for z > a 2 l , corresponding to S > 0.4 , the
encircled power for rc < 0.8 (in units of lz / D ) is within 8% of S times the encircled
power for a beam focused at a distance z , i.e., it is given by
2.6.1 Theory
Consider an optical imaging system with an exit pupil of radius a forming a Gaussian
image in a plane at a distance R from the plane of its exit pupil. The irradiance
distribution in a plane at a distance z is given by Eq. (2-81). The effect of defocus as an
aberration can be separated from the scaling effects due to z being different from R, by
comparing the irradiance distributions (or PSFs) observed at a distance z when the image
is formed at a distance R and when it is formed at a distance z . Thus, if we measure r in
units of lz D (instead of l R D ) and I in units of Pex Sex l2 z 2 , Eqs. (2-81) become
2
1
Û
ı
[ ]
I (r ) = 4 Ù exp iF 2 (r) J 0 ( p r r) r dr , (2-98)
0
where
F 2 (r) = Bd r2 (2-99a)
Sex Ê 1
- ˆ r2
1
= (2-99b)
l Ëz R¯
represents the defocus phase aberration. Equation (2-98) does not change if we replace
F 2 (r) by the orthonormal Zernike circle polynomial representing defocus, i.e., if we let
= ( )
3 2r 2 - 1 s F , (2-100b)
where
Sex 1 1
sF = - (2-101)
2 3l z R
In a similar manner, we can show that for a system aberrated by a rotationally (or
radially) symmetric phase aberration F(r) , Eq. (1-82) giving the irradiance distribution
of the image of a point object, formed in a plane at a distance R from the plane of its exit
pupil, reduces to
1 2
Û
ı
[ ]
I (r ) = 4 Ù exp iF(r) J 0 ( pr r) r dr , (2-102)
0
According to this equation, the aberration-free central value of irradiance I (0) is unity.
For numerical analysis we now write the right-hand side of this equation as a product of
two complex conjugate integrals
1 1
Û Û
ı
[ ı
]
I (r ) = 4 Ù exp iF(r) J 0 ( p r r) r dr Ù exp [ - iF (s)] J 0 ( p rs) sds . (2-103)
0 0
Combining the two exponentials and noting that, since I (r ) is a real quantity, the
imaginary part on the right-hand side of Eq. (2-103) must be zero, we may write it in the
form
1 1
Û Û
[ ]
I (r ) = 4 Ù Ù cos F(r) - F (s) J 0 ( p r r) J 0 ( p rs) rs dr ds .
ı ı
(2-104)
0 0
The Strehl ratio of the image is the central value I (0) ; i.e.,
1 1
Û Û
S = 4 Ù Ù cos F(r) - F(s) rs dr ds .
ı ı
[ ] (2-105)
0 0
The encircled power, i.e., fraction of total power in a circle of radius rc (in units of
lz D for the defocused image and l R D for the aberrated images) is given by [see Eq.
(2-22b)]
rc
P(rc ) ( Û
= p 2 2 Ù I (r ) rdr .
ı
) (2-106)
0
where
rc
Û
Q(r, s; rc ) = Ù J 0 ( p rr) J 0 ( p rs) rdr
ı
0
( )[
= rc2 2 J 02 ( p rrc ) + J12 ( p rrc ) ] if r = s , (2-108a)
[ (
= rc p r2 - s 2 )] [r J (p rr ) J (p sr ) - sJ (p sr ) J (p sr )] if r π s .(2-108b)
1 c 0 c 1 c 0 c
124 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
P(rc ) ~ (p rc 2) 2 S , (2-109)
which simply amounts to assuming that the average irradiance in the image plane across a
circle of small radius rc is approximately uniform and equal to its central value S.
F 2 (r) = ( )
3 2r 2 - 1 s F , (2-110a)
F 4 (r) = ( )
5 6r 4 - 6r2 + 1 s F , (2-110b)
F 6 (r) = (
7 20r6 - 30r 4 + 12r2 - 1 s F ) , (2-110c)
and
(
F 8 (r) = 3 70r8 - 140r6 + 90r 4 - 20r2 + 1 s F ) . (2-110d)
and, therefore,
Thus, the coefficient s F of each aberration represents its standard deviation as well as its
root mean square value. How these aberrations vary with r is shown in Figure 2-16 for
s F = 1. Since 0 £ r £ 1, it is evident that n 2 is the number of roots of the aberration
F n (r) ; i.e., F n has a value of zero at n 2 different values of r.
The integrals in Eqs. (2-104), (2-105), and (2-107) for the aberrations given above
may be evaluated by the Gauss quadrature formula,18 according to which, for a function
f (r, s) ,
2.6 PSFs for Rotationally Symmetric Aberrations 125
3.0
F8
2.0
F2
1.0
F(r)
0.0
1.0
F4
2.0
F6
3.0
0.0 0.2 0.4 0.6 0.8 1.0
r
1 1
Û Û M M
Ê 1 + xi 1 + x j ˆ
Ù Ù f (r, s) dr ds = (1 4) Â Â w iwj f Á
Ë 2
, ˜ , (2-112)
ı ı i =1 j =1 2 ¯
0 0
where M is the number of points of a 1D quadrature, x i is the i-th zero of the M-th order
Legendre polynomial PM ( x ) , and w i are the weight factors given by
2 dPM ( xi )
2
wi = . (2-113)
1 - xi2 dx
Since the integrands in these equations are symmetric in r and s, Eq. (2-112) reduces for
our application to
1 1
Û Û 2 Ê 1 + xi 1 + xi ˆ
M
Ù Ù f (r, s) dr ds = (1 4) Â w i f Ë 2 , 2 ¯
ı ı i =1
0 0
M i 1
Ê 1 + xi 1 + x j ˆ
+ (1 2) Â Â w iw i f Á , ˜ . (2-114)
i =1 j =1 Ë 2 2 ¯
xi are the zeros of Legendre polynomials and wi are the weight factors
±xi wi ±xi wi
n =2 n =8
0.18343 6424 5650 0.36268 7833 8362
0.57735 2691 9626 0.00000 0000 0000 0.52553 4099 6329 0.31370 6458 7887
0.79666 4774 3627 0.22238 0344 3374
n =3 0.96028 8564 7536 0.10122 5362 0376
0.00000 0000 0000 0.88888 8888 8889
0.77459 6692 1483 0.55555 5555 5556
n= 4 0.00000 0000 0000 0.33023 3550 1260
0.33998 0435 4856 0.65214 1548 2546 0.32425 4234 3809 0.31234 0770 0003
0.86113 3115 4053 0.34785 8451 7654 0.61337 4327 0590 0.26061 6964 2935
0.83603 1073 6636 0.18064 1606 4857
n =5 0.96816 2395 7626 0.08127 3883 1574
0.00000 0000 0000 0.56888 8888 8889 n = 10
0.53846 3101 5683 0.47862 6704 9366 0.14887 3389 1631 0.29552 2247 4753
0.90617 8459 8664 0.23692 8850 6189 0.43339 3941 9247 0.29626 7193 9996
n =6 0.67940 5682 9024 0.21908 3625 5982
0.23861 1860 3197 0.46791 9345 2691 0.86506 3666 8985 0.14945 3491 0581
0.66120 3864 6265 0.36076 5730 8139 0.97390 5285 7172 0.06667 3443 8688
0.93246 5142 3152 0.17132 4923 9170 n = 12
0.12523 4085 1469 0.24914 0458 3403
n =7 0.36783 4989 8180 0.23349 5365 8355
0.00000 0000 0000 0.41795 1836 3469 0.58731 9542 6617 0.20316 4267 3066
0.40584 1513 7397 0.38183 0505 5119 0.76990 6741 4305 0.16007 3285 3346
0.74153 1855 9394 0.27970 3914 9277 0.90411 2563 0475 0.10693 3259 5318
0.94910 9123 2759 0.12948 9661 8870 0.98156 6342 6719 0.04717 3363 6512
±xi wi
n = 16
0.09501 25098 37637 440185 0.18945 06104 55068 496285
0.28160 35507 79258 913230 0.18260 34150 44923 588867
0.45801 67776 57227 386342 0.16915 65193 95002 538189
0.61787 62444 02643 748447 0.14959 59888 16576 732081
0.75540 44083 55003 033895 0.12462 89712 55533 872052
0.86563 12023 87831 743880 0.09515 85116 82492 784810
0.94457 50230 73232 576078 0.06225 35239 38647 892863
0.98940 09349 91649 932596 0.02715 24594 11754 094852
n = 20
0.07652 65211 33497 333755 0.15275 33871 30725 850698
0.22778 58511 41645 078080 0.14917 29864 72603 746788
0.37370 60887 15419 560673 0.14209 61093 18382 051329
0.51086 70019 50827 098004 0.13168 86384 49176 626898
0.63605 36807 36515 025453 0.11819 45319 61518 417312
0.74633 19064 60150 792614 0.01093 01198 17240 435037
n = 24
0.06405 68928 62605 626085 0.12793 81953 46752 156974
0.19111 88674 73616 309159 0.12583 74563 46828 296121
0.31504 26796 96163 374387 0.12167 04729 27803 391204
0.43379 35076 26045 138487 0.11550 56680 53725 601353
0.54542 14713 88839 535658 0.10744 42701 15965 634783
0.64809 36519 36975 569252 0.09761 86521 04113 888270
0.74012 41915 78554 364244 0.08619 01615 31953 275917
0.82000 19859 73902 921954 0.07334 64814 11080 305734
0.88641 55270 04401 034213 0.05929 85849 15436 780746
0.93827 45520 02732 758524 0.04427 74388 17419 806169
0.97472 85559 71309 498198 0.02853 13886 28933 663181
0.99518 72199 97021 360180 0.01234 12297 99987 199547
2.6 PSFs for Rotationally Symmetric Aberrations 127
Substituting Eqs. (2-110a) through (2-210d) into Eq. (2-105), we can calculate the
Strehl ratio for the aberrations under consideration. In the case of defocus and fourth-
order spherical aberration, the Strehl ratio can be obtained analytically from Eq. (2-102)
by letting r = 0 and carrying out the integration. The result obtained is
[(sin ) ]
2
S2 = 3 sF 3 sF (2-115)
and
(
S4 = C 2 b + S 2 b b , ) (2-116)
where
b = 3 5s F p , (2-117a)
◊ ◊
and C( ) and S( ) are the Fresnel integrals given by
b
Û
C(b) = Ù cos p x 2 2 dx
ı
( ) (2-117b)
0
and
b
Û
(
S(b) = Ù sin p x 2 2 dx .
ı
) (2-117c)
0
Note that S2 gives the ratio of central irradiances at a distance z when the focused image
lies at a distance R and when it lies at a distance z. The Strehl ratio S4 for fourth-order
spherical aberration was considered in Section 2.3.4 also; see Table 2-6.
Figure 2-17 shows how the Strehl ratio varies with the standard deviation s w (in
units of the optical wavelength l ) of the various aberrations under consideration. The
wave and phase aberrations are related to each according to s w = (l 2 p) s F . S2 , S4 , S6 ,
and S8 represent the actual Strehl ratios corresponding to aberrations F 2 , F 4 , F 6 , and
F 8 , respectively. For comparison, the Strehl ratios obtained by using the approximate
( ) ( )
2
expressions 1 - s 2F 2 – Maréchal formula, exp - s 2F – Gaussian approximation Sg ,
( )
are also plotted in this figure. It is evident that exp - s 2F approximates the Strehl ratio
very well and better than the other two expressions. We note that S2 deviates the most
from Sg . However, the difference between the two is less than 10% for S2 > 0.38. The
values of s F , accurate to the fourth decimal place, for Strehl ratios of 0.1 (0.1) 1 are
given in Table 2-9. We note that, except for very small Strehl ratios, their difference for
various aberrations for a given Strehl ratio is negligible. However, we used these exact
values to calculate and assess the aberrated PSFs and encircled powers.
128 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0
0.8
0.6
S
0.4
Sg
0.2 1 sF2
Sm
0.0
0 0.05 0.10 0.15 0.20 0.25
sw
sF (Radians)
S F2 F4 F6 F8
1.0 0 0 0 0
0.9 0.3228 0.3233 0.3236 0.3238
0.8 0.4671 0.4687 0.4695 0.4702
0.7 0.5865 0.5899 0.5913 0.5928
0.6 0.6964 0.7025 0.7044 0.7071
0.5 0.8034 0.8139 0.8159 0.8203
0.4 0.9129 0.9303 0.9314 0.9383
0.3 1.0302 1.0598 1.0570 1.0680
0.2 1.1646 1.2203 1.2037 1.2216
0.1 1.3385 1.5014 1.3986 1.4312
2.6 PSFs for Rotationally Symmetric Aberrations 129
Substituting Eqs. (2-110a) through (2-110d) into Eqs. (2-104) and (2-107), we can
calculate the defocused PSFs and encircled powers. The PSF and encircled power curves
for S ∫ S2 = 0.8, 0.6, 0.4, 0.2, and 0.1 are shown in Figure 2-18. Comparing the actual
aberrated encircled power with the aberration-free encircled power scaled by the actual
Strehl ratio, we find that for S ≥ 0.4, rc £ 0.8, the difference between the two is < 8%. For
S ≥ 0.1 and rc £ 0.5, the difference is 5%. Note that a given value of sF and, therefore, a
given value of S2 is obtained for two values of z, one for which z < R and the other for
which z > R . Since r and rc in Figure 2-18 are in units of lz D, the spread of the
irradiance distribution for a given value of S2 depends on whether z < R or z > R . It
should be clear, however, that the central irradiance and the concentration of power on a
target at a distance z are maximum when the beam is focused on it, i.e., when R = z .
The aberrated PSFs corresponding to Strehl ratios of 0.8, 0.6, 0.4, 0.2, and 0.1 for the
fourth-order spherical aberration are shown in Figure 2-19. For comparison, the
aberration-free PSF (corresponding to a Strehl ratio of 1) is also included in this figure.
The corresponding curves for encircled power are also shown in this figure. As expected,
the irradiance distribution decreases inside the Airy disc (of radius 1.22) and increases in
the bright ring around it. However, the location of the first minimum remains practically
unchanged. Moreover, its value remains close to zero. Thus, the Airy disc remains
distinct and practically unchanged in size as a fourth-order spherical aberration is
introduced into an otherwise aberration-free optical system.
1.0
Pg(rc)
0.8
F2
l(r)/S P(rc)/S
0.6
0.4
0.2
lg(r)
0.0
0.0 0.5 1.0 1.5 2.0
r, rc
Figure 2-18. Defocused PSFs normalized to unity at the center and encircled power
corresponding to various values of S. Aberration-free ( S2 = 1) curves are included
for comparison. A Gaussian approximation of the PSFs is also included. S2 = 1, ___;
S 2 = 0.8, _ _ _ S2 = 0.6, _. _. _; S2 = 0.4; - - -; S2 = 0.2, ...; S2 = 0.1.
130 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0
Pg S=1
0.8
0.8
0.6
l(r) P(rc)
0.6
0.4
0.4
lg(r)
0.2 0.2
0.1 F4
0.0
0.0 0.5 1.0 1.5 2.0
r, rc
Figure 2-19. Aberrated PSFs and encircled power for F4 corresponding to Strehl
ratios of 0.8, 0.6, 0.4, 0.2 and 0.1. Aberration-free ( S = 1) curves are included for
comparison. The PSF approximated by a Gaussian function is also included. The
parameters r and rc are in units of lF , where F = R D is the focal ratio of the
image-forming light cone. The Airy disc radius is 1.22 in these units.
An even more dramatic result is obtained if the aberrated PSFs of Figure 2-19 are all
normalized to unity at the center as in Figure 2-20. It is evident from this figure that,
within the Airy disc, the normalized aberrated PSFs are practically identical with the
aberration-free PSF. Thus, the aberration reduces the irradiance distribution quite
uniformly by the Strehl ratio at points in the region, say, r <
~ 1. The encircled power in
this region is also correspondingly scaled by the Strehl ratio, as may be seen from the
normalized encircled power curves shown in Figure 2-20. Note that the total normalized
power is given by S 1 instead of unity. Therefore, the normalized power curves should
be taken seriously only for rc <
~ 1. Similar results are obtained for the sixth- and eighth-
order spherical aberrations, as shown in Figures 2-21 and 2-22, respectively. For these
two aberrations, only the normalized curves are shown.
È 2 J ( pr ) ˘
2
I (r ) = S Í 1 ˙ , 0 £ r £1 . (2-118)
Î pr ˚
If the aberration is not represented by a Zernike polynomial, e.g., if the phase aberration
is given by
2.6 PSFs for Rotationally Symmetric Aberrations 131
1.0
Pg(rc)
0.8
l(r)/S P(rc)/S F4
0.6
0.4
lg(r)
0.2
0.0
0.0 0.5 1.0 1.5 2.0
r, rc
Figure 2-20. Aberrated PSFs for F4 normalized to unity at the center and
corresponding normalized encircled power. The curves correspond to Strehl ratios
indicated in Figure 2-19. The Gaussian PSF and encircled power are also included.
1.0
Pg(rc)
0.8
l(r)/S P(rc)/S
0.6
F6
0.4
0.2
lg(r)
0.0
0.0 0.5 1.0 1.5 2.0
r, rc
Figure 2-21. Aberrated PSFs for F6 normalized to unity at the center and
corresponding normalized encircled power. The curves correspond to Strehl ratios
indicated in Figure 2-19. The Gaussian PSF and encircled power are also included.
132 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0
Pg(rc)
0.8
l(r)/S P(rc)/S
0.6
F8
0.4
lg(r)
0.2
0.0
0.0 0.5 1.0 1.5 2.0
r, rc
Figure 2-22. Aberrated PSFs for F8 normalized to unity at the center and
corresponding normalized encircled power. The curves correspond to Strehl ratios
indicated in Figure 2-19. The Gaussian PSF and encircled power are also included.
then the central portion of the PSF spreads somewhat more than the Airy disc.
Szapiel17 has shown that a better approximation in such cases is obtained if the right-
hand side of Eq. (2-118) is multiplied by a factor
{
f (r ) = exp [ g1 F(r )] - g2 F(r )
2
} , (2-120)
where
[
F(r ) = 1 - (1.22 r )
2
] , (2-121a)
and
where
b0 = a 02 , b1 = 2a 0 a1 , b2 = 2a 0 a 2 + a12 , (2-122a)
and
Êa a a ˆ
a0 = - 1 + 2 + 3 + ... , (2-123a)
Ë2 3 4 ¯
so that
F =0 (2-123b)
and
s 2F = F 2
b1 b2 b3
= b0 + + + + ... . (2-123c)
2 3 4
i si
1 0.272 443
2 0.247 985
3 0.209 272
4 0.177 535
5 0.153 009
6 0133 952
7 0.118 879
8 0.106 727
9 0.096 753
10 0.088 439
134 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
The error is < 10% as long as S > ~ 0.3 . The percent error is defined to be 100 1 - Sg S . ( )
Note that this error is negative for the aberrations considered.
Now we approximate the PSF with a Gaussian function. The exact aberration-free
PSF and encircled power are given by Eqs. (2-15) and (2-24), respectively. According to
these equations, the PSF has a central value of unity and its total power is also unity. A
2D Gaussian function with a central value of unity and a total power of unity is given by
[
Ig (r ) = exp – ( p r 2)
2
] . (2-125)
The corresponding encircled power, obtained by substituting Eq. (2-125) into Eq.
(2-106), is given by
[
Pg (rc ) = 1 - exp ( - p rc 2) ]
2
. (2-126)
Since, for small values of r, the aberrated PSF and the corresponding encircled power
are simply scaled by the Strehl ratio, their Gaussian approximations may be written
[
Ig (r; S ) = S exp ( - pr 2)
2
] (2-127a)
and
{ [
Pg (rc ; S ) = S 1 - exp - ( p rc 2)
2
]} , (2-127b)
respectively. For very small values of rc , Eq. (2-127b) reduces to Eq. (2-109).
Substituting for S from Eq. (2-124) into Eqs. (2-127), we obtain
( ) [
Ig (r; s F ) = exp - s 2F exp - ( p r 2)
2
] (2-128a)
and
( ){
Pg (rc ; s F ) = exp - s 2F 1 - exp - ( p rc 2) [ 2
]} , (2-128b)
respectively. It is evident, as expected, that Eqs. (2-128) are not valid for large values of r
( )
or rc . For example, although the total power is unity, Pg (rc ; s F ) Æ exp - s 2F as rc Æ • .
Equations (2-124) and (2-128) give a complete description of our Gaussian model for
2.6 PSFs for Rotationally Symmetric Aberrations 135
aberrated systems. Of course, Eqs. (2-124) and (2-128b) can be obtained from Eq. (2-
128a). For example,
Thus, Eq. (2-128a) is the basic equation for the Gaussian model.
Figure 2-23 shows how the aberration-free exact results given by Eqs. (2-15) and (2-
24) compare with their corresponding Gaussian approximations given by Eqs. (2-125)
and (2-126). We note that the two sets of results are practically identical for r and
rc <
~ 0.5 . As r and rc increase, the differences between the two also increase.
The Gaussian PSF and encircled power for an aberrated system are shown in Figures
2-18 through 2-22. It is evident that the results for the aberration-free case scaled by the
Strehl ratio approximate the results for an aberrated case better and over a wider range
than the corresponding results obtained by the Gaussian approximation.
1.0
Pg(rc)
0.8
2 2
1 – J0(prc) – J1 (prc)
0.6
(r) P(rc)
0.4
2
[2J1 (pr)/pr]
0.2
lg(r)
0.0
0.0 0.5 1.0 1.5 2.0
r, rc
aberrations reduce the irradiance distribution quite uniformly within the Airy disc scaled
simply by the Strehl ratio of the system. Accordingly, the calculation of encircled power
(
for small circles rc < )
~ 1 reduces to a knowledge of the Strehl ratio. In the case of
spherical aberrations, for S ≥ 0.4 and rc £ 1, the aberration-free encircled power scaled
by the exact Strehl ratio is in error by < 2%. For S ≥ 0.1, and rc £ 1, the error is <~ 10%.
In the case of defocus, for S ≥ 0.4 and rc £ 0.8 the error is < 8%. For S ≥ 0.1 and
rc £ 0.5 , the error is <
~ 5%. The error is generally larger for a smaller Strehl ratio and a
larger value of rc .
To evaluate the integrals involved in the expressions for the Strehl ratio, PSF, and
encircled power, adequate accuracy is obtained even with an 8-point Gauss quadrature.
For example, for S ≥ 0.4 , r £ 1, the encircled power is in error by at most a few percent
(<
~ 5%). Using a 12-point quadrature, the error is reduced to < 2% for Strehl ratios as
small as 0.2 and rc as large as 3. The error generally increases as S decreases and as rc
and the aberration order increase. If the aberration is not represented by a Zernike
polynomial, the PSF in its central region spreads more than the Airy disc, and a simple
correction can be applied to account for it
plane unless the Fresnel number N of the exit pupil as observed from the Gaussian image
point is very large. In this section, we discuss the symmetry of the irradiance distribution
of the aberrated image of a point object about the axis of the image-forming light cone as
well as in and about the Gaussian image plane.
where Anm is the aberration coefficient. We will refer to the plane in which the center of
the reference sphere lies as the Gaussian image plane. If the image is observed in another
plane that lies at a distance z from the exit pupil, then the aberration becomes
For an unapodized system, Eq. (2-79) for the irradiance distribution of the image of a
point object in a defocused plane at a distance z from the plane of the pupil may be
written
1 2p 2
2
ÏR¸ Û Û È ˘
Ó pz ˛ ı ı
[ Î
R
z
]
I (r , q i ; z ) = Ì ˝ Ù Ù exp i F(r, q) exp Í - pi r r cos (q - q i )˙ r dr dq
˚
, (2-131)
0 0
we may write
Ï ¸
exp Ìi ÈÍF (r, q) - p r r cos (q - q i )˘˙˝
R
ÓÎ z ˚˛
) [ ]
• •
(
= 4 Â ¢ Â ¢ i s ( -i ) s ¢ exp iBd r2 Js Anm Rnm (r)
s = 0 s¢ = 0
where a prime on the summation sign indicates, for example, that terms with s = 0 are to
be taken with a factor of 1/2. Substituting Eq. (2-133) into Eq. (2-131) and noting that
138 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
2p
Û
ı
[ ]
Ù cos ( msq) cos s ¢(q - q i ) dq = p cos (s ¢q i ) d ms, s ¢ , (2-134)
0
•
( m 1) s
I ( r, q i ; z ) = ( 4 R z ) Â ¢( - i ) cos ( msq i )
2
s=0
1
2
Û
( ) [ ]
¥ Ù exp i Bd r2 J s Anm Rnm (r) J ms ( p r rR z ) r dr
ı
. (2-135)
0
The irradiance distribution for the aberration-free case can be obtained by letting
Anm = 0 . Since Js ( 0 ) = d s, 0 , only the s = 0 term in Eq. (2-135) is nonzero. This term is
associated with a factor of 1/2. Hence, Eq. (2-135) reduces to Eq. (2-81) when the system
is aberration free. Since
È 2p ˆ ˘
cos Íms Ê q i + j = cos ( msq i ) , j = 1, 2, K, m ,
Î Ë m ¯ ˙˚
Eq. (2-135) shows that the irradiance distribution in any observation plane (normal to the
z-axis) is m-fold symmetric about the z axis. It also shows that the tangential plane z x and
all planes containing the z axis and making angles of pj m with the tangential plane are
planes of symmetry. For m = 0 the irradiance distribution does not depend on q i . This is
to be expected; for a radially symmetric aberration, the irradiance distribution in any
observation plane is also radially symmetric. These symmetry properties are also
possessed by the aberration given by Eq. (2-130a), and they are not affected by the
inclusion of the defocus aberration as in Eq. (2-130b). It is shown in Section 2.7.2 that the
distribution in the Gaussian image plane is 2m-fold symmetric when m is even, although
the aberration is only m-fold symmetric.
• ( m 1) s
I (r, q i ; z ) = 16 Â ¢( - i ) cos ( ms q i )
s=0
1
2
Û
( ) [
¥Ù exp iBd r2 J s Anm Rnm (r) Jms ( p r r) r dr
ı
] , (2-136)
0
2.7 Symmetry Properties of an Aberrated PSF 139
where
pa2
Bd = ( R - z) . (2-137)
l R2
1
2
Û
ı
{[
= 4 Ù exp i An 0 Rn0 (r) + Bd r2 ] } J (p r r) rdr
0 . (2-138b)
0
Two planes located symmetrically about the Gaussian image plane at distances z = R + D
and z = R - D from the exit pupil, where D is a small distance along the z axis,
correspond to Bd = - D 8 F 2 and Bd = D 8 F 2 , respectively. It is evident from Eq. (2-
138b) that if we change the sign of Bd , we obtain a different distribution. Hence, the
irradiance distribution is not symmetric about the Gaussian image plane. It is shown in
Section 2.7.3 that the axial irradiance is symmetric about this plane. However, we note
from Eq. (2-138b) that the irradiance distribution does not change if the sign of An0 is
also changed along with that of Bd . In that case, the total aberration function of Eq. (2-
130b) (with m = 0 ) changes sign. Thus, the irradiance distribution in two symmetrically
defocused image planes are identical provided they are for spherical aberrations of equal
magnitude but opposite signs.
( m 1) s
When m is an odd integer (e.g., for coma), the quantity ( - i ) in Eq. (2-136) is
real; it is either equal to 1 or –1. Since Bd changes sign as we go from one defocused
plane to another located symmetrically about the Gaussian image plane, the values of the
integral in Eq. (2-136) for the two defocused planes are complex conjugates of each
other. Hence, the irradiance distributions in these two planes are equal; i.e., the
distribution is symmetric about the Gaussian image plane. Accordingly, the axial
irradiance is also symmetric about this plane. Note that the total aberration function
according to Eq. (2-130b) is different for the two symmetrically defocused planes.
( - i )( m 1) s
= ± 1 when s is even
= ± i when s is odd
and
[ ]
cos ms(q i + p m) = cos ( msq i ) when s is even
Therefore,
Hence, the irradiance at a point in a defocused plane is equal to the irradiance at a point
resulting from a reflection in the Gaussian image plane and a rotation through an angle
p m about the z axis. Since the axial irradiance is unaffected by this rotation, it is
symmetric about the Gaussian image plane. The total aberration functions corresponding
to two symmetric planes, of which one is rotated with respect to the other by p m , have
equal magnitude but opposite signs. It follows from Eq. (2-136) that the irradiance
distribution in the Gaussian image plane is 2m-fold symmetric about the z axis when m is
even, although the aberration function is only m-fold symmetric. This, of course, is true
for small values of N also.
In the Gaussian image plane, z = R and Bd = 0 , and Eq. (2-136) for the irradiance
distribution reduces to
2
[ (r)]J ms (prr)rdr
• ( m 1) s 1
I (r, q i ; R) = 16 Â ¢( - i ) cos ( ms q i ) Ú J s Anm Rnm . (2-139)
s=0 0
An interesting property on the effects of aberrations in this plane can be obtained from
◊
the above equation by expanding J s ( ) in a power series according to
• ( x / 2) s + 2 n
J s ( x ) = Â ( -1) s .
n=0 n ! ( n + s )!
( )
1 1
I (r, q i ; R) = 4 Ú J 0 ( prr)rdr - 2(i ) m 1
cos ( m q i ) Anm Ú Rnm (r) J m ( prr)rdr + O Anm
2
.
0 0
(2-140)
È 2 J ( pr ) ˘
2
I ( r , q i , R) = Í 1 ˙ + O ( Anm ) , m odd (2-141a)
Î pr ˚
È 2 J ( pr ) ˘
2
= Í 1
Î p r ˙ + O Anm
˚
2
( ) , m even . (2-141b)
Thus, for example, a first-order amount of aberration produces a first-order change in the
irradiance distribution of the aberration-free image in the case of coma, but a second-
order change in the case of astigmatism. The ring structure of the Airy pattern near its
center is affected much more strongly in the case of coma than in the case of astigmatism.
2.7 Symmetry Properties of an Aberrated PSF 141
It is evident from Eq. (2-142) that the axial irradiance does not depend on the sign of Bd ;
i.e., it is symmetric about the Gaussian image plane.
Since
(
R20n (r) = Pn 2r2 - 1 ) , (2-144)
◊
where Pn ( ) is a Legendre polynomial of degree n, then letting
2r 2 - 1 = x ,
1 2
I (0; z ) = (1 4)
Û
ı
[ ]
exp (i Bd 2) Ù exp (i Bd x 2) exp i An 0 Pn 2 ( x ) dx . (2-145)
1
J s ( – z ) = ( –1) s J s ( z ) , (2-146a)
[ ] [ ] J [- A (r)]
*
( - i )( m 1) s
J s Anm Rnm (r) = ( - i )( m 1) s
s
m
nm Rn . (2-146b)
Hence, we note from Eq. (2-139) that a change in the sign of the aberration coefficient
Anm for even values of m (including zero) has no effect on the irradiance distribution in
the Gaussian image plane. (This is true for systems with small values of N as well.) This
may be seen more easily from Eq. (2-141b) when m = 0. Thus, the sign of such an
aberration cannot be determined from the Gaussian image-plane distribution. In a
defocused plane, the total aberration function changes when the sign of the aberration is
changed, and so the distribution also changes. However, for even m, distributions in two
symmetrically defocused planes are the same, provided the aberration coefficient for one
has a sign opposite to that for the other. Similarly, for odd m, the distributions in any
observation plane for aberration coefficients with opposite signs are the same, provided
one is rotated with respect to the other by an angle p. In each case, the total aberration
functions are the same except for a difference in their signs.
From the general discussion given above, it is easy to obtain the symmetry properties
of the PSFs aberrated by primary aberrations. These are discussed in the next section.
2.8.1 Defocus
As explained in Section 2.5.1, the irradiance distribution is symmetric about the
Gaussian image plane when the Fresnel number is large. Figure 2-24 shows the defocused
PSFs for several values of the peak defocus aberration Bd , such that the aberration-free
central value is unity. As Bd increases, the central value, i.e., the Strehl ratio, decreases.
The minima of the aberration-free PSF are no longer zero. The central value is zero when
Bd is an integral number of waves, as may be seen from Eq. (2-84c). Since the
corresponding PSFs have very low values, they have been multiplied by a factor of 10 in
the figure. The PSF for Bd = 0.64l is included here because, as will be discussed in
Section 2.10, the OTF is negative for certain spatial frequency bands for values of Bd
larger than this.
1.00
Bd = 0 Defocus
W(r) = Bdr2
0.75
0.25
(r; Bd)
0.50 1
0.5
10 ¥
0.25
3 10 ¥
0.64
2
10 ¥
0.00
0 1 2 3
r
Figure 2-24. PSFs of a defocused system. Bd represents the peak value of defocus
aberration in units of l. The curves for Bd = 1, 2, and 3 have been multiplied by ten.
Figure 2-25 shows the axial irradiance for As = 1l, 2 l, and 3 l . It is evident that
the axial irradiance is symmetric about the defocused point corresponding to Bd = - As ;
i.e., about the plane with respect to which the aberration variance is minimum.
1.0
W(r) = Asr4
1
0.8
As = 0
0.6
(0; Bd)
2
0.4
3
0.2
0.0
6 5 4 3 2 1 0 1 2 3
Bd
Figure 2-26 shows the PSFs for spherical aberration combined with defocus. Figure
2-26a shows the PSFs in defocused image planes Bd = - As corresponding to minimum
aberration variance. The PSF for As = 2.2 l is also included here because, as discussed
later in Section 2.10, the OTF is negative for certain spatial frequency bands for values of
As larger than this. We note that the radius of the central bright spot does not change as
the aberration is increased. Figure 2-26b shows the PSFs for As = 1 l and Bd As = 0 , –1,
–1.5, and –2, corresponding to the Gaussian, minimum-aberration-variance, circle-of-
least-confusion, and marginal image planes, respectively. The PSFs in the Gaussian and
marginal image planes have been multiplied by a factor of 10 in this figure because of
their low values. We note that the central irradiances in these two planes are equal to each
other, showing that the axial irradiance is symmetric about the Bd = - As plane, as
pointed out in Table 2-11, where the symmetry properties are summarized. It is also clear
from the figure that the best image is obtained in the plane corresponding to minimum
aberration variance. The PSFs in other defocused image planes illustrating symmetry of
axial irradiance are shown in Figure 2-26c.
Figure 2-29 shows the PSFs for Aa = 1l and Bd Aa = 0, –1/2, and –1,
corresponding to sagittal-line, minimum-aberration-variance [circle-of-least (astigmatic)-
confusion], and tangential-line image planes, respectively. (As discussed in Section 4.3.3
Balanced Spherical
W(p) = As(pL p2)
0.75
«w
..s 0.50 (a)
t 2
". , \
"-
0.25 2.'
2').... \
3
. "-' . ~
\
\
"
,,~ \
-.~.
0.00 ~-- --
0 2 3
_ r
1.00
t ..'. .'.
" ..... 10 x.······· ..
.......
: .......
'
0.12
- 0.5 (C)
0.08
0.2 0.04
_r 2 3 4
0.00
Figure 2-26. PSFs for spherical aberration. (a) Various values of As and defocused
image planes Bd = -As corresponding to minimum aberration variance. (b) and (c)
Fixed value of As and various image planes. Bd and A., are in units of A, and r is in
units of AF. The curves for Bd = 0 and - 2 have been multiplied by 10 in (b).
Similarly, they have been multiplied by 5 in (c), or ignore the factor of 5, but use the
right-hand scale for these curves.
146 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0 ,---,---,-----,------,--,---.,.------,------;-------,
0.8
;;r 0.6
t 0.4
0.2
- 4 - 3 - 2 -1 o 2 3
- Bd
1.0 0 . . . . . - - - - - - - - - - , - - - - - - - , - - - - - - - - - - - - - ,
Balanced Astigmatism
W(p,9) = Aa(p2cos29- p2/2)
0.75 9;= 0
:;:m
.s 0.50 (a)
f
0.25 2 10 x
2 3
_ r
1.00 ....- - - - - - - - , - - - - - - - , - - - - - - - - - - ,
Balanced Astigmatism
W(p ,9) = Aa(p2cos29- p2/2)
9;= 1t/4
.s 0.50 (b)
f
0.25 - .. __ 2
3
'- -"
.. - - _ .. __ 10 x
.. __ .... - .. .. -
0.00 L _ _ _ _ _..J:::s··...
0 2 3
·....-_-"""'-
_r
=!-;;,,;i!i.t;o:..::=--_ _ _-.J
Figure 2-28. PSFs for various amounts of astigmatism in defocused image planes
corresponding to minimum aberration variance along the directions. (a) 8 i = 0 and
(b) It/ 4. Au represents the peak value of astigmatism in units of A, and a defocused
image plane is represented by Bd = - Aa / 2.
2.8 PSFs for Primary Aberrations 147
– 0.5
0.1
–1
0.0
0 1 2 3
r
(a)
Bd = – 0.5
0.1
0, –1
0.0
0 1 2 3
r
(b)
of Part I), the two line-image planes are those in which the astigmatic focal lines based on
geometrical optics are obtained.) We note from Figure 2-29a that the central irradiance of
the line images are equal to each other, showing that the axial irradiance is symmetric
about the Bd = - Aa 2 plane, as pointed out in Table 2-11. The distributions of the two
line images are rotated with respect to each other by p 2. Accordingly, the solid curve in
Figure 2-29b corresponds to both Bd Aa = 0 and – 1.
148 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
2.8.4 Coma
Figure 2-30 shows the axial irradiance for coma with Ac = 1l . The axial irradiance
for coma without any tilt is along the z axis. For coma optimally balanced with tilt, the
axial irradiance shown is along an axis that is parallel to the z axis but passing through the
diffraction focus (4/3, 0). The irradiance is symmetric about the Gaussian image plane. It
is evident from the figure that the central value of the PSF is much larger at the
diffraction focus than at the Gaussian focus.
The PSF for coma is symmetric about the tangential plane z x . Thus, it has a line
symmetry in any observation plane, the line lying in the tangential plane, i.e., the PSF is
symmetric about the xi axis. The distribution in two observation planes located
symmetrically about the z = R plane are identical. The PSF is also symmetric about the
Gaussian image plane. A change in the sign of the aberration coefficient Ac produces a
rotation of the distributions by p about the axis. The centroid of the PSF does not lie at
the Gaussian image point due to its asymmetric distribution. Its location is discussed in
Section 2.9.
Figure 2-31 shows the PSFs for coma along the directions for q i = 0, p 4 , p 2,
where q i = 0 corresponds to the xi axis. Similarly, q i = p 2 corresponds to the yi axis.
The diagonal corresponding to q i = p 4 is indicated by xi¢ in Figure 2-31b. We note that,
as pointed out in Section 2.3.3, the peak value of an aberrated PSF does not lie at the
Gaussian image point (r = 0) . Its location is discussed in Section 2.9.2.2. Figure 2-31c
shows the symmetry of the PSFs in yi , i.e., about the xi axis.
0.6
W(r,q) = Acr3cosq
0.5 Ac = 1
0.4 Through
(4/3, 0)
I(Bd)
0.3
0.2
On axis
0.1
(0, 0)
0.0
–4 –3 –2 –1 0 1 2 3 4
Bd
Figure 2-30. Axial irradiance through the Gaussian image point and the diffraction
focus for coma with Ac = 1 .
2.8 PSFs for Primary Aberrations 149
1.00
Ac = 0 0.25 Coma
W(r,q) = Acr3 cosq
qi = 0
0.75 0.5
(xi; Ac)
1
0.50 (a)
0.25
2
3
0.00
2.5 0.0 2.5 5.0
xi
1.00
Ac = 0 Coma
W(r,q) = Acr3 cosq
0.75 qi = p /4
0.25
(x⬘i ; Ac)
0.50 (b)
0.5
0.25
3
1
2
0.00
2.5 0.0 2.5 5.0
xi⬘
1.00
Ac = 0 Coma
W(r,q) = Acr3 cosq
0.75 0.25 qi = p /2
(yi; Ac)
0.50 (c)
0.5
0.25
3 2
1
0.00
2.5 0.0 2.5 5.0
yi
Figure 2-31. PSFs for coma along the directions (a) qi = 0, (b) qi = p 4 , and
(c) qi = p 2 . Ac is the peak value of coma aberration in units of l . Symmetry of the
PSF about the xi axis is evident in (c).
150 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
2.8.5 2D PSFs
The PSF Figures 2-24 through 2-31 are useful for quantitative assessment of the
effect of aberrations. However, they do not lend themselves easily to what the PSFs may
look like when observed in practice. This is especially true when the PSF is not radially
symmetric. For this reason, we give computer-generated 2D pictures of the PSFs in this
section for various values of a primary aberration. The emphasis of these pictures is on
the structure of a PSF, i.e., on the distribution of its bright and dark regions, and not on its
irradiance distribution. Some of the symmetry properties of the aberrated PSFs discussed
above are evident from these pictures. It should be clear, however, that a random mixture
of various aberrations will only lead to a complex PSF.
Figure 2-32 shows how the aberration-free PSF, or the Airy pattern, appears when
observed in a defocused image plane. The central irradiance approaches zero when a
defocus aberration of integral number of waves is introduced; hence the dark spot in the
center when Bd = 1, 2, or 3 l . The PSFs aberrated by one wave of spherical aberration
( As r 4 with As = 1l) are shown in Figure 2-33 when observed in defocused image planes
that are often considered in geometrical optics (see Figure 4-3 of Part I). These planes are
the Gaussian ( Bd = 0 ), minimum variance ( Bd = - As ), circle of least confusion
( Bd = - 1.5 As ), and marginal ( Bd = - 2 As ). The PSFs for an increasing amount of
spherical aberration observed in a corresponding defocused image plane Bd = - As , so
that the variance of the aberration is minimum, are shown in Figure 2-34. There is no
dark spot at the center of the PSFs aberrated by spherical aberration.
The PSFs aberrated by coma are shown in Figure 2-37. Their symmetry about the xi
axis is evident. Moreover, it is easy to see that their centroid does not lie at the Gaussian
image point. As the aberration increases, they begin to take the conical shape of the
geometrical PSFs. The locations of peak irradiance and centroid are discussed in Section
2.9.2.2.
2.8 PSFs for Primary Aberrations 151
Bd = 0 Bd = 0.5
Bd = 1 Bd = 1.5
Bd = 2 Bd = 3
Figure 2-32. Defocused PSFs. Bd represents the peak value of defocus wave
aberration in units of l . The central value of the PSF is zero when Bd is equal to an
integral number of wavelengths.
152 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
(a) Bd = 0 (b) Bd = - 1
As = 0.25 As = 0.5
As = 1 As = 2
As = 3
[ ( )]
Figure 2-34. PSFs for balanced spherical aberration As r4 - r2 . Thus, a PSF is
observed in a defocused image plane corresponding to Bd = - As . The aberration
coefficient As is in units of l .
154 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Aa = 1, Bd = 0 Aa = 1, Bd = - 0.5
Aa = 1, Bd = - 1 Aa = 1, Bd = 0.5
Aa = 3, Bd = 0 Aa = 3, Bd = 1.5
Aa = 0.25 Aa = 0.5
Aa = 1 Aa = 2
Aa = 3
( )
Figure 2-36. PSFs for balanced astigmatism Aa r2 cos 2 q - r2 2 . Thus, Bd = - Aa 2,
and the PSFs are 4-fold symmetric. The aberration coefficient Aa is in units of l .
156 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Ac = 0.25 Ac = 0.5
Ac = 1 Ac = 2
Ac = 3
( )
Figure 2-37. PSFs aberrated by increasing amount of coma Acr3 cosq . They are
symmetric about the horizontal ( xi ) axis. The peak and the centroid of the PSFs do
not lie at the Gaussian image point. The aberration coefficient Ac is in units of l .
2.8 PSFs for Primary Aberrations 157
For an aberration-free system, the wavefront exiting from the exit pupil is spherical,
and all of the object rays transmitted by the system converge to its center of curvature
where the Gaussian image point lies. Accordingly, the irradiance or the density of rays is
infinity at this point and zero elsewhere in the Gaussian image plane. Thus, the
aberration-free geometrical PSF is a Dirac delta function. The corresponding diffraction
PSF is the Airy pattern consisting of a bright central Airy disc of radius 1.22l F
containing 83.8% of the total power P surrounded by alternating dark and bright rings.
The central irradiance is finite and given by p P 4l2 F 2 . The radius of the (aberration-
free) central bright disc increases linearly with the wavelength l , and the central
irradiance decreases as l 2 . The aberrated geometrical PSFs increase linearly in size as
the aberration increases. They are completely independent of the wavelength except for
any dependence of the aberration on it. The diffraction PSFs, on the other hand, undergo
dramatic changes as the aberration increases.
The aberrations reduce the central irradiance of the Airy pattern and increase it at the
other points. In the case of spherical aberration, the radius of the central bright spot does
not change significantly as the aberration increases. The diffraction PSFs are not bounded
in size as the geometrical PSFs are. Moreover, the diffraction PSF is zero at certain points
compared to the geometrical PSF, which is nonzero everywhere inside the image spot.
The presence of zeros in a diffraction PSF is due to the destructive interference of
Huygens’ spherical wavelets exiting from the exit pupil, a phenomenon that is absent in
geometrical optics. Note also that the irradiance in a diffraction PSF is never infinity
158 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
unlike at some points in a geometrical PSF. For small aberrations, the central irradiance
( )
of the diffraction PSF varies approximately as exp - s 2F , where s 2F is the variance of
the phase aberration across the exit pupil.
In the case of coma, the irradiance of the geometrical PSF is infinite along the lines
of a (bounded) cone of semiangle 30 ˚ with its vertex at the Gaussian image point. The
peak value of the diffraction PSF does not occur at the Gaussian image point, unlike the
geometrical PSF, which has a value of infinity at that point. For small aberrations, the
peak value occurs at the point [( 4 / 3) FAc , 0] , where Ac is the peak value of the coma
wave aberration. For large aberrations there is no simple relationship for obtaining the
peak-value point. The centroid of the diffraction PSF is identical with that of the
geometrical PSF. The two PSFs begin to resemble each other qualitatively as the
aberration increases (Compare Figure 2-33 with Figure 4-8 of Part I).
The diffraction PSFs for astigmatism are completely different from the
corresponding uniform geometrical PSFs. The general elliptical spot image in the case of
astigmatism also has a uniform irradiance. The sagittal and tangential line images, which
are special cases of the elliptical image, have infinite irradiance. As the aberration
increases, the central bright spot of the Airy pattern is replaced by a thick line image
corresponding to the sagittal and tangential line images of the geometrical PSFs.
Measuring the centroid of a PSF with respect to a point other than the Gaussian
image point is equivalent to introducing a wavefront tilt in the aberration function. Thus,
the balanced coma aberration that yields the minimum spot sigma is given by
( )
Ac r3 - r cos q . The balanced aberration for minimum variance is
[ ]
Ac r3 - (2 / 3) r cos q , similar to the Zernike circle polynomial R31 (r) cos q , consistent
with the peak value of the PSF being at the point [( 4 / 3) FAc , 0] for small aberrations. In
the case of astigmatism, the smallest geometrical spot (the circle of least confusion) and
minimum aberration variance are obtained when Bd = - Aa 2; i.e., the balanced
(
aberration based on both geometrical and diffraction optics is Aa r2 cos 2 q - r2 2 , )
similar to the Zernike polynomial R22 (r)cos 2q.
It is a common practice in lens design to look at the spot diagrams in the early stages
of a design, in spite of the fact that they do not represent reality. Just as in the diffraction
treatment an optical system is considered practically diffraction limited if it yields an
image with a Strehl ratio of 0.8, or the peak (or peak-to-valley) aberration is less than
2.9 Line of Sight of an Aberrated System 159
l 4 (Rayleigh's quarter-wave rule), or the standard deviation of the aberration across the
exit pupil is less than l 14 (Maréchal’s criterion), similarly there is a golden rule in
optical design that a system is close to its diffraction limit if the ray spot radius is less
than or equal to the radius of the Airy disc. We note, for example, that this holds for
spherical aberration in the Gaussian image plane if As £ l 6 .6 , although a larger value of
As is obtained in the other image planes. Considering that the long dimension of the
coma spot is 6FAc and the line image for astigmatism is 8 FAa long, the aberration
tolerance for the spot size to be smaller than the Airy disc is Ac < 0.4 l and Aa < 0.3 l ,
respectively. The aberration tolerances based on the spot size are summarized in Table 2-
12. These tolerances are roughly consistent with Rayleigh’s quarter-wave rule. Hence, it
is reasonable to use the size of the spot diagrams as a qualitative measure of quality of the
design until it becomes smaller than the Airy disc. Thus, as the design improves, a
designer strives for spot diagrams of a size smaller than or equal to that of the Airy disc,
and then analyzes the system by its aberration variance and diffraction characteristics
such as the PSF or the MTF, since only a diffraction PSF represents the actual image
quality of a system. Key results of the comparison of the geometrical and diffraction
PSFs are summarized in Table 2-13.
In this section, we discuss the line of sight (LOS) of an aberrated system in terms of
the centroid of its PSF. It is shown that only coma terms of an aberration function affect
the LOS. Moreover, terms with cosq dependence affect the x-LOS and those with the sinq
dependence affect the y-LOS. Hence, in considering the tolerances for coma terms, their
impact on the LOS of the system must be taken into account. Numerical results are given
showing that it may be advantageous to define the LOS in terms of the peak of the PSF
rather than its centroid.
r r 2 pi r r ˆ r
2
Ii (ri ) = ( Û
ı
) [ ( )]
Pex Sex l2 R 2 Ù exp i F rp exp Ê - r r dr
Ë lR p i ¯ p ◊ . (2-147)
The PSF is obtained from Eq. (2-147) by dividing both sides by Pex .
The LOS of an aberration-free optical system coincides with the center of its PSF.
For an aberrated system, let us define its LOS as the centroid of its aberrated PSF. Let the
aberration function of the system in terms of Zernike circle polynomials be given by
• 12
[ ]
n
W (r, q) = Â Â 2 (n + 1) (1 + d m 0 ) Rnm (r) (cnm cosmq + snm sinmq) , (2-148)
n=0 m=0
where cnm and snm are the orthonormal Zernike aberration coefficients representing
standard deviation of a corresponding aberration term, with the exception of c00 and s00 .
For a system with a uniformly illuminated circular exit pupil of radius a, the centroid of
its aberrated PSF, following Eq. (1-181), is given by
2p
Û
xi , yi = ( R Sex ) Ù W ( a, q) (cosq, sinq) a dq , (2-149)
ı
0
where W ( a, q) is obtained from Eq. (2-148) by letting r = 1, and we have used the
relations
2.9 Line of Sight of an Aberrated System 161
and
r
(
d s = x p , y p dq . ) (2-150b)
where a prime indicates a summation over odd integral values of n. We note that only
those aberrations contribute to the LOS that vary with q as cosq and sinq. Aberrations
varying as cosq contribute to xi , and those varying as sinq contribute to yi . This
may also be seen from the symmetry properties of the aberrations. Since Rn1 (r) consists
of terms in rn , rn 2 , K, and r, therefore Rn1 (r) cos q , for example, is symmetric in y p
but not in x p . Hence, the PSF is symmetric in yi , as may be seen from Eq. (2-147).
Accordingly, yi = 0 for this aberration. For a given value of cn1 or sn1 , an aberration
of a higher order gives a larger LOS error. Thus, two Zernike aberrations with m = 1 but
different values of n having the same standard deviation give different LOS errors, even
though they give (approximately) the same Strehl ratio. A higher-order aberration, i.e.,
one with a larger value of n, yields a larger LOS error. Hence, in tolerancing an optical
system, one should be careful in allocating equal standard deviation to two aberration
terms that also contribute to the LOS error.
where Wn is in units of l, and use normalized quantities of Eqs. (2-8) through (2-10), Eq.
(2-147) may be written
1 2p 2
Û Û Ï È ¢ ˘¸
I (r , q i ) = p 2
Ù Ù exp Ì p i Í Â 2W n r cos q - rr cos(q - q i )˙ ˝ r dr dq
n
. (2-153)
ı ı Ó În ˚˛
0 0
where
2
Ê ˆ
B2 = Á ¢ 2Wnrn - rr cos q i ˜ + (rr sin q i )
2
(2-155)
Ën ¯
and
162 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Ê ˆ
tan y = - rr sin q i Á ¢ 2Wn rn - rr cos q i ˜ . (2-156)
Ën ¯
Thus, Eq. (2-153) reduces to
2
1 2p
Û Û
I ( r, q i ) = p 2
[ ]
Ù Ù exp p i B cos (q - y ) r dr dq
ı ı
,
0 0
2
È1 ˘
ÍÛ
= 4 Ù J 0 ( p B) r dr˙ , (2-157)
Íı ˙
Î0 ˚
where we have used Eq. (2-12).
Substituting Eq. (2-152) into Eq. (2-149), we obtain the centroid of the aberrated
PSF,
xi = 2 F Â ¢ Wn , (2-158)
n
where it is understood that yi = 0 . Thus, the LOS error depends on the value of the
peak aberration Wn but not on n. We note that for n = 1, the aberration is a tilt and for
n = 3 it is primary coma, but they both give the same LOS error if W1 = W3 , even though
the corresponding PSFs are completely different. The reason for the same LOS error is
that for a uniform circular pupil, the centroid depends only on the aberration along the
perimeter of the pupil, which depends on Wn but not on n. From Eq. (2-153), it is evident
that the centroid of a PSF aberrated by balanced coma represented by a Zernike
polynomial Rn1 (r) cos q does not lie at the origin. However, if the aberration is balanced
for minimum spot radius (see Table 4-3 of Part I), then the centroid lies at the origin; in
other words, such an aberration minimizes the spot radius with respect to the centroid.
The PSF simply shifts such that its peak and centroid locations move from (0, 0) to
The aberrated PSF may be obtained from Eq. (2-157) by substituting Eq. (2-156) into it
with n = 3. If we let q i = 0 , then y = 0 ; and Eq. (2-157) giving the PSF along the x axis
may be written
2
È1 ˘
Û
I ( x ) = ÍÙ J 0 (p B) dt ˙ , (2-162)
Íı ˙
ÍÎ 0 ˙˚
where
t = r2 (2-163)
and
B = (2tW3 - x ) t 1 2 . (2-164)
Figure 2-38 shows how I ( x ) varies with x for several typical values of W3 ∫ Ac (in
units of l ) varying from 0 to 3. Equation (2-158) shows that the centroid of the PSF is
given by
1.0
W3 = 0
0.8 0.5
0.6
(x 0)
0.4
1.5
0.2
2
0.0
–3 –2 –1 0 1 2 3
x
For small values of W3 , the peak value of the aberrated PSF occurs at a point such that, if
the aberration is measured with respect to a reference sphere centered at this point, the
variance of the aberration across the pupil is minimum. From the properties of Zernike
polynomials, we find that the polynomial R31 (r) cos q gives the optimum combination of
r3 cos q and r cos q terms leading to a minimum variance. Since
we note that, for small values of W3 , the peak value of the aberrated PSF occurs at
x m = ( 4 3) W3 , (2-167)
where the subscript m refers to the point corresponding to minimum aberration variance.
From the form of the aberration, it is evident that ym = 0 . Thus, primary coma shifts the
centroid and the peak of the PSF by different amounts, the movement of the peak being
two-thirds of the shift of the centroid.
Figure 2-39a shows how the irradiance I m at x m , the peak irradiance I p , and the
irradiance Ic at < x > vary with W3 . Figure 2-39b shows how x m , x p (the point at which
the peak irradiance occurs) and xi vary with W 3 . Several typical values of x m , I p , and
< x > and the corresponding irradiances Im , I p , and Ic are noted in Table 2-14. The
aberrated central irradiance I(0) is also given in this table. The irradiance values I(0) and
I m are the Strehl ratios for primary and balanced primary coma, respectively. We note
that the peak of the PSF lies approximately at x m , the point corresponding to
W1 = (2 3)W3 only when W3 < ~ 0.7l . For larger values of W3 , the peak occurs closer to
1.0
0.8
10
8
0.6
l
6
<x>
x
0.4
xm
4
0.2 xp
lp
2
lm
lc
0.0 0
0 1 2 3 4 5 0 1 2 3 4 5
W3 W3
(a) (b)
Table 2-14. Typical values of xn , x p , and < x > and corresponding irradiances Im ,
I p , and Ic for circular pupils aberrated by primary coma.
W3 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
0.5 0.67 0.66 1.00 0.8712 0.8712 0.6535 0.3175
1.0 1.33 1.30 2.00 0.5708 0.5717 0.1445 0.0791
1.5 2.00 1.80 3.00 0.2715 0.2844 0.0004 0.0618
2.0 2.67 1.57 4.00 0.0864 0.1978 0.0061 0.0341
the origin than the point corresponding to minimum aberration variance. For W3 > ~ 1.6l ,
the distance of the peak from the origin does not increase monotonically, but fluctuates as
W3 increases. Since, according to Eq. (2-165), the distance of the centroid increases
linearly with W3 , it is clear that the separation between the locations of the centroid and
the peak increases as W3 increases.
Its standard deviation is given by s W = A5 2 3 . Following Eq. (2-157), we find that the
aberrated PSF along the x axis is given by Eq. (2-162) where
( )
B = 2t 2 W5 - x t 1 2 . (2-169)
Figure 2-40a shows how I(x) varies with x for several typical values of W5 . According to
Eq. (2-158), the centroid of the PSF is given by
From Eqs. (2-165) and (2-170), we note that the primary and secondary coma with the
same peak value yield the same centroid < x > , although the corresponding aberrated
PSFs are different. This may be seen from Figure 2-41, where the PSFs for W3 = 5 = W5
are shown.
The standard deviation of the aberration given by Eq. (2-168) is reduced by a factor
of 2 if an aberration W1r cos q , where W1 = - 0.5 W5 , is introduced. Accordingly, for
small values of W5 , the peak of the corresponding aberrated PSF occurs at x m = W5 ,
although its centroid lies at < x > = 2W5 . The values of x m , x p , and < x > and the
corresponding irradiances Im , I p , and Ic for the values of W5 considered in Figure 2-40
are noted in Table 2-15.
166 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0
W5 = 0
0.8
0.5
0.6
l(x)
0.4 1
1.5
0.2
2
0.0
3 2 1 0 1 2 3
x
(a)
1.0
W5 = 0
1
W3 = 1.2 W5
xm = 0.6 W5 2
0.8
0.6
l(x)
0.4
0.2
0.0
3 2 1 0 1 2 3
x xm
(b)
Figure 2-40. (a) PSF for several typical values of secondary coma W5 in units of
l.(b) PSF for combined primary and secondary coma aberration given by Eq. (2-
172). Note that the horizontal coordinate is x - x m .
2.9 Line of Sight of an Aberrated System 167
(a) (b)
Figure 2-41. PSF aberrated by (a) primary coma with W3 = 5 l and (b) secondary
coma with W5 = 5 l .
Table 2-15. Typical values of x m , x p , and < x > and corresponding irradiances Im ,
I p , and Ic for circular pupils aberrated by secondary coma given by Eq. (2-168).
W5 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
0.5 0.50 0.49 1.00 0.8150 0.8153 0.4114 0.4955
1.0 1.00 0.83 2.00 0.4464 0.4664 0.0025 0.2332
1.5 1.50 0.81 3.00 0.1685 0.3237 0.0098 0.1873
2.0 2.00 1.11 4.00 0.0420 0.2523 0.0073 0.1389
The standard deviation of the aberration given by Eq. (2-168) is reduced by a factor of 10
if we introduce r cos q and r3 cos q aberrations with W1 = 0.3 W5 and W3 = - 1.2 W5 .
Hence, the peak value of a PSF aberrated by a small value of W5 and W3 = - 1.2 W5
occurs at x m = - 0.6 W5 . According to Eq. (2-158), the corresponding centroid occurs at
< x > = - 0.4W5 . Therefore, the separation between the peak and the centroid is 0.2 W5 .
For large values of W 5 , the minimization of variance with respect to W3 and W1 does not
lead to a maximum of the PSF.
168 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
(
W (r, q) = W5r5 + W3r3 cos q ,) (2-172)
where
W3 = - 1.2W5 . (2-173)
According to Eq. (2-171), the point in the image plane with respect to which the
aberration variance is minimum is given by
x m = - 0.6W5 . (2-174)
The aberrated PSF along the x axis is given by Eq. (2-162), where
(
B = 2t 2W5 + 2t W3 - x t 1 2 ) . (2-176)
Figure 2-41 shows the aberrated PSF I(x) for several values of W5 with W3 given by Eq.
(2-173). The values of x m , x p , and < x > and the corresponding irradiances Im , I p , and
Ic are given in Table 2-16. Note that x m , x p , and < x > are all negative. Moreover, their
magnitude for the values of W5 considered is very large. Therefore, in Figure 2-41, the
horizontal coordinate is chosen to be x - x m .
2.9.3 Comments
The results given here are applicable to both imaging systems, e.g., those used for
optical surveillance, as well as to laser transmitters used for active illumination of a
target. In both cases, the LOS of the optical system is extremely important. A LOS error
of a surveillance system will produce an error in the location of the target. In the case of a
Table 2-16. Typical values of x m , x p , and < x > and corresponding irradiances Im ,
I p , and Ic for primary and secondary coma given by Eq. (2-172).
W5 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
1.0 – 0.60 – 0.59 – 0.40 0.9676 0.9682 0.8763 0.3721
2.0 – 1.20 – 1.18 – 0.80 0.8765 0.8784 0.5870 0.0030
3.0 – 1.80 – 1.77 – 1.20 0.7429 0.7459 0.2981 0.0014
4.0 – 2.40 – 2.37 – 1.60 0.5886 0.5914 0.1173 0.0465
2.10 Diffraction OTF for Primary Aberrations 169
laser transmitter, a large LOS error may cause the laser beam to miss the target
altogether. Whereas for static aberrations we may be able to calibrate the LOS, for
dynamic aberrations it is the analysis given here that will determine the tolerances on
aberrations of the type rn cos q and rn sin q .
Although we have defined the LOS of an optical system in terms of the centroid of
its PSF, it could have been defined in terms of the peak of the PSF (assuming that the
aberrations are small enough so that the PSF has a distinguishable peak). For an
aberration-free PSF, its peak value and its centroid both lie at its origin, regardless of the
amplitude variations across its pupil. The two are not coincident when aberrations are
present. The precise definition of the LOS will perhaps depend on the nature of the
application of the optical system. Moreover, in practice, only a finite central portion of
the PSF will be sampled to measure its centroid, and the precision of this measurement
will be limited by the noise characteristics of the photodetector array used for such a
measurement.
Now we discuss the full OTFs for primary aberrations. The MTFs, PTFs, and the real
and imaginary parts of the OTFs are considered for various values of a primary
aberration. The phenomenon of contrast reversal is pointed out and discussed.
r ÛÛ r
[ ]
t(vi ) = Sex1 ÙÙ exp iQ( p, q; vi ) dp dq
ıı
(2-177a)
È ÛÛ ÛÛ ˘
= Sex1 Í ÙÙ cos Q dp dq + i ÙÙ sin Q dp dq ˙ , (2-177b)
Î ıı ıı ˚
where
r
Q ( p, q; vi ) = F ( p + l Rvi 2, q ) - F ( p - l Rvi 2, q ) (2-178)
is the phase aberration difference function and the integration is carried over the overlap
region of two pupils centered at ( m l Rvi 2, 0) . The aberration function F ( p, q ) in the
( )
rotated coordinate system is obtained from the nominal function F x p , y p by replacing
x p with p cos f - q sin f and y p with p sin f + q cos f . For a rotationally symmetric
system, there is no loss of generality if we assume that the point object for which the
aberration function is under consideration lies along an axis parallel to the x p axis. The
aberration function in that case depends on terms with integral powers of x p and
x 2p + y 2p . Hence, the aberration function F ( p, q ) can be obtained from the function
( )
F x p , y p by replacing x 2p + y 2p by p 2 + q 2 and x p by p cos f - q sin f .
170 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
q
p
yp
xp
(0,0)
ni
lR
Figure 2-42. Geometry for evaluating the OTF. The centers of the two pupils are
( )
located at (0, 0) and l R ( x, h) in the x p , y p coordinate system and m (l R 2) (vi , 0)
( )
12
in the ( p, q ) coordinate system, where vi = x 2 + h 2 and f = tan 1 ( h x) . The
shaded area is the overlap area of the two pupils. When normalized by the pupil
radius a, the centers of the two pupils of unity radius lie at m v along the p axis.
(
Ïbd x p 2 + y p 2 ,
Ô
) Defocus
Ô
(
Ôas x p + y p
F( x p , y p ) = Ì
2 2 2
) (
+ bd x p 2 + y p 2 , Spherical + Defocus) (2-179)
(
Ôaa x p 2 + bd x p 2 + y p 2 ,
Ô
)
Astigmatism + Defocus
(
Ôac x p x p 2 + y p 2 .
Ó ) Coma
Letting
Ï Bd r2 , Defocus
Ô
Ô As r 4 + Bd r2 , Spherical + Defocus
F(r, q) = Ì (2-181)
Ô Aa r cos q + Bd r
2 2 2
, Astigmatism + Defocus
Ô
Ó Ac r cos q ,
3
Coma
where the two types of aberration coefficients are related to each other according to
Bd = a 2 bd , (2-182a)
2.10 Diffraction OTF for Primary Aberrations 171
As = a 4 as , (2-182b)
Aa = a 2 aa , (2-182c)
and
Ac = a 3 ac . (2-182d)
Ï4 Bd pv Defocus
Ô
r ÔÔ s ( )
8 A pv p 2 + q 2 + v 2 + 4 Bd pv Spherical + defocus
Q ( p, q; v ) = Ì
( )
(2-183)
Ô4 Aa v p cos f - q sin f cos f + 4 Bd pv Astigmatism + defocus
2
Ô
( )
ÔÓ Ac v 6 p 2 + 2 q 2 + 2 v 2 cos f - 4 Ac pqv sin f , Coma
where ( p, q ) are now normalized by the pupil radius so that p 2 + q 2 £ 1 and the spatial
frequency v is normalized by the cutoff frequency 1 l F so that v £ 1. In the normalized
coordinates, the OTF given by Eq. (2-177a) becomes
r 1 ÛÛ r
t( v ) = ÙÙ
p ıı
[
exp iQ( p, q; v ) dp dq] . (2-184)
We note from Figure 2-42 that in the ( p, q ) coordinate system, the two unit circles
are centered at mv along the p axis and the limits of p and q are m ÊË 1 - q 2 - vˆ¯ and
m 1 - v 2 , respectively. Their overlap region, which forms the region of integration in the
above integrals, is symmetric in p and q. Hence, if Q is an odd function of p and/or q, the
imaginary part of the integral vanishes. This is true for defocus, spherical aberration, and
astigmatism. It is also true for coma for a spatial frequency parallel to the y p axis, i.e., for
f = p 2 . In such cases the OTF is real and, depending on whether its value for a certain
spatial frequency is positive or negative, its phase, the PTF, for that frequency is zero or
p, respectively. A phase of p is sometimes associated with a negative value of the MTF.
It represents contrast reversal, i.e., for example, bright regions in the object appear as
dark regions in the image. Similarly, the dark regions of the object appear as bright
regions in the image.
That the OTF is real for a symmetric aberration can be seen even more easily by
virtue of the corresponding symmetry of the PSF, namely that
• •
[
t (x, h) = Ú Ú PSF( x, y) exp 2 pi(x x + hy) dx dy
• •
]
• •
[
= Ú Ú PSF( - x, - y) exp 2pi(x x + hy) dx dy
• •
]
• •
[ ]
= Ú Ú PSF( x ¢, y ¢) exp - 2pi(x x ¢ + hy ¢) dx ¢ dy ¢
• •
= t * (x, h) . (2-186)
2.10.2 Defocus
Figure 2-43 shows how the OTF of a defocused system varies with the spatial
frequency. We note that it is real and radially symmetric; i.e., its value depends on the
value of v but not on the value of f. For Bd <~ 0.64l , the OTF is positive for all spatial
frequencies. However, for larger values of Bd it becomes negative, corresponding to a
PTF of p, for certain bands of spatial frequencies. It becomes negative for smaller and
smaller spatial frequencies as the amount of defocus Bd increases.
To illustrate the significance of the OTF and, in particular, the contrast reversal, we
consider, as shown in Figure 2-44a, a 2D object that is sinusoidal along the vertical axis
with a spatial frequency that increases linearly in the horizontal direction. The maximum
frequency in the object is chosen to equal the cutoff frequency of the aberration-free
system. This frequency is normalized to unity. The aberration-free or the diffraction-
1.00
Defocus
Bd = 0 W(r) = Bdr2
0.75
0.50 0.25
t
0.5
0.25
2 1 0.64
3
0.00
– 0.25
0.00 0.25 0.50 0.75 1.00
n
Figure 2-43. OTFs of a defocused system. Bd represents the peak defocus
aberration in units of l.
2.10 Diffraction OTF for Primary Aberrations 173
(a)
(b)
(c)
1.0
0.8
0.6
Bd = 0
t
0.4 Bd = 2l
0.2
0.0
– 0.2
0.0 0.2 0.4 0.6 0.8 1.0
n
(d)
limited image of the object is shown in Figure 2-44b. The monotonic reduction in
contrast with increasing spatial frequency is quite evident from this figure. A defocused
image corresponding to Bd = 2l is shown in Figure 2-44c. It is clear that the contrast in
the image reduces with frequency rapidly to zero, reverses its sign back and forth as the
frequency increases, with practically zero values for frequencies v ≥ 0.3 . As a
convenience, the aberration-free and defocused OTFs are shown in Figure 2-44d to
illustrate the regions of zero and near-zero contrast as well as the regions of contrast
reversal.
The OTF is independent of the sign of Bd . Hence the OTFs corresponding to two
image planes symmetrically defocused about the Gaussian image plane are identical
when the Fresnel number of the system is large. However, when the Fresnel number is
small, the two symmetrical planes correspond to different magnitudes of the defocus
aberration and, therefore, yield different OTFs. Moreover, the cutoff frequency in that
case is D l z instead of D l R , as may be seen by replacing R with z in Eq. (1-73b).
2.10.4 Astigmatism
The OTF aberrated by astigmatism Aa r2 cos2 q is real and independent of the sign
of Aa when Bd = 0 or when the sign of Bd is also changed. Figure 2-46 shows it for a
defocused image plane Bd = - Aa 2, corresponding to minimum aberration variance
across the pupil, for various values of Aa . The OTF in this case is four-fold symmetric,
with one axis lying in the tangential plane. We note from the figure that it becomes
negative for smaller values of Aa when f = p 4, compared to when f = 0 . For example,
the OTF for f = 0 is positive for Aa £ 1.28l . However, it is negative for f = p 4, when
Aa = 1 l and v > 0.35 . Figure 2-47 shows the OTF for Aa = 1l in various image planes.
As discussed in Section 4.3.3 of Part I, Bd Aa = 0, –1/2, and –1 correspond to image
planes in which the tangential focal line, circle-of-least (astigmatic)-confusion, and
sagittal focal line, respectively, are observed. The OTF has a biaxial (or inversion)
symmetry, except for observations in the plane of least confusion, in which case it has a
2.10 Diffraction OTF for Primary Aberrations 175
1.00
Balanced spherical
W(p) = As(pL p2)
0.75
0.25
0.50
...
(a)
t 0.25
0.00
-0.25
0.00 0.25 0.50 0.75 1.00
__ v
1.00..------------,--------,.--------r--------,
~,
\', ~ Spherical and defocus
f~••
W(p) =Asp4+ BdP2
\
0.75 \ ':.•',.......
\ As= 1
\ ', ......
\ " ........ Bd =-1
\\,-1··~·············..........
0.50
(b)
t .......
... ..
- ..-.::.;..""'.-
0.00 - - - -" - - - - i:.- -~ ..,-~~...:."-'-:=-=-----~-::;;;.::~'-=--~-_1
,;;:g. . . . ""
- 0.25 L -_ _ _ _--L_ _ _ _ _...L.._ _ _ _----'_ _ _ _ _....J
Figure 2-45. OTFs for spherical aberration. (a) Various values of As and image
planes Bd = -As corresponding to minimum aberration variance. (b) Fixed As and
various image planes. As and Bd are in units of A..
four-fold symmetry, as stated earlier. The PTF is either 0 or Te depending on the spatial
frequency . For <jJ = 0, the OTF in the image plane Bd = - Aa is the same as the
aberration-free OTF, as may be seen from the third of Eqs. (2-183). Similarly, the OTF
for Bd = 0 is the same as the defocused OTF in Figure 2-43 for Bd = 1 A.. For <jJ = Te14 ,
the OTFs in the image planes corresponding to Bd I Aa = 0 and - 1 are equal. The OTF
for <jJ = Tel2 is independent of astigmatism, as may be seen from Eq. (2-183).
176 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1 . 00~---------.-----------,----------.----------.
I'~"
'~ ....
\ " .. .' .. Balanced astigmatism
W(p ,e) =Aa(p 2cos 2e- p2/2)
0.75 " \ ...... ....... ::
'\' \ ' ","........
\ $= n/4
'\
,
\
\
". " .............
0.50
\\ \ "'" 0. ~5···..... (b)
t 0.25 \'\ \2
~,
- 0.25 '------------------'---------------'---------------'----------------'
0.00 0.25 0.50 0.75 1.00
_v
Figure 2-46. OTFs for various amounts of astigmatism in defocused image planes
corresponding to minimum-aberration variance. As indicated in Figure 1-7, cjl
represents the angle the spatial frequency vector makes with the x axis. Aa
represents the peak value of astigmatism aberration in units of A. and a defocused
image plane is represented by Bd = -Aa/2. (a) cjl= 0 and (b) cjl= 1t/4.
2.10.5 Coma
The OTF aberrated by coma Acp3 cose is a complex function, except for <I> = n/2 ,
i.e., when the radiance of the sinusoidal object varies along a direction that is normal to
the axis about which the PSF is symmetric. A conjugate function is obtained when the
sign of Ac is changed. The OTF dependence on the spatial frequency is shown in terms of
its MTF and PTF in Figures 2-48 and 2-49, respectively. Figure 2-48 shows the MTF for
several values Ac when <I> = 0, n/4 , and n/2. The corresponding PTF is shown in Figure
2.10 Diffractiobn OTF for Primary Aberrations 177
1.00 ..------------,---------,---------,--------,
0.50
....
t 0.25
(a)
0.00
-0 . 25L-----~-----~------L-----~
0.00 0.25 0.50v
__ 0.75 1.00
1.00 , . . - - - - - - - - - - - r - - - - - - - , - - - - - - - - - - , - - - - - - - - ,
0.50
....
(b)
t 0.25
,,
,,
, ,,
,,
,
0.00 - - - - - - - - - ' .... . . =-_=-_- - - - - - - - - - - -
-0 . 25~----~-----~-----~-----~
0.00 0.25 0.50v
__ 0.75 1.00
2-49, except when <\> = n12, in which case the PTF is zero. The value of 'P in this figure
is in radians. Note that if a wavefront tilt is added to the coma aberration to minimize its
variance across the pupil, as discussed in Section 2.3.3, it only introduces a PTF varying
linearly with the spatial frequency. The real and imaginary parts I't( v) Icos 'P (v) and
178 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.00..,-----------r------,-----------r-------,
Coma
W(p) =Acp3 cosS
0.75 $= 0
0.50 (a)
t ,,
1,
... ...
0.25 ... ...
0.00
0.00 0.25 0.50 0.75 1.00
_ v
1.00
Coma
W(p) =Acp3 cosS
0.75 $= 7t/4
...
0.50 ,, - (b)
t ...
1 ... ...
...
0.25
0.00
0.00 0.25 0.50 0.75 1.00
_ v
1.00
Coma
W(p) =Acp3 cosS
0.75 $=7t/2
t 0.50 (c)
0.25
Figure 2-48. MTFs for coma. Ac represents the peak value of coma aberration in
units of A. (a) cp = 0, (b) cp = n/4, and (c) cp = n/2 for which the OTF is real.
2.10 Diffraction OTF for Primary Aberrations 179
I't( V) Isin 'P (v), respecti vel y, of the OTF are shown in Figures 2-50 and 2-51 for <1> = 0
and n14, respectively. As discussed in Section 1.6.4, the slope of the real part at the origin
is independent of the aberration and is equal to that of the MTF. Moreover, the slope of
the imaginary part at the origin, which yields the centroid of the PSF (see Section 1.8.1),
increases as the aberration increases. Finally, as discussed in Section 1.9.3, the Strehl
ratio is obtained by integrating the real part, and the integral of the imaginary part is zero.
40.----------,----------,-----------r----------,
Coma
W(p) =Acp3 cose
30 $=0
9-
" "
t 20
" ""
(a)
2 "
""
10
0.25 0.50v
__ 0.75 1.00
30.----------.-----------.-----------,----------~
Coma
W(p) =Acp3 case
$= 11:/4
20
(b)
10
O~--~~~~~~==~----~-----------L----------~
Figure 2-49. PTFs for coma. Ac represents the peak value of coma aberration in
units of A. (a) cI> = 0 and (b) cI> = 1r./4.
180 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.00 ..,.--------.----------r-------,---------,
Coma
W(p) = Acp3 cose
$= 0
p 0.50
Q)
a:
(a)
t 0.25
0.00
1.00
Coma
W(p) =Acp3 cose
0.75
$ =n/4
p 0.50
Q)
a: (b)
t 0.25
0.00
-0.25
0.00 0.25 0.50 0.75 1.00
_v
Figure 2-50. Real part Ret of the OTF of a system aberrated by coma. (a) cj) = 0 and
(b) cj) = 1r./4 .
2.10 Diffraction OTF for Primary Aberrations 181
1.00
Coma
W(p) = Acp 3 cose
0.75
$= 0
l-'
0.50
E '.
". (a)
t 0.25
0.00
-0.25
0.00 0.25 0.50 0.75 1.00
__ v
1.00
Coma
W(p) = Acp3 cose
0.75
$= 1t/4
l-'
0.50
E (b)
t 0.25
-O . 25~--------~~--------~----------~----------~
0.00 0.25 0.50 0.75 1.00
__ v
Figure 2-51. Imaginary part Imt of the OTF of a system aberrated by coma.
(a) cp = 0 and (b) cp = 1[/4.
182 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Bd <
~ ± l 20 v , (2-187)
Aa <
~ ± l 10 v in the plane Bd = - Aa 2 , (2-188)
and
As <
v ¯ d ( )
Ê 0.106 + 0.33ˆ in the plane B = - 1.33 - 2.2 v + 2.8v 2 A
~ ± lË s , (2-191)
2.11.2 Defocus
Equation (2-187) for defocus tolerance can be obtained as follows. Substituting the
aberration difference function Q( p, q; v) = 4 Bd pv for the defocus phase aberration
( )
Bd p 2 + q 2 into Eq. (2-184), and noting that the limits of integration for p and q are
m ÊË 1 - q 2 - vˆ¯ and m 1 - v 2 , respectively, we obtain
2.11 Hopkins Ratio 183
1 v2 1 q2 v
1 Û Û
t (v) = Ù dq Ù exp( 4iBd pv) dp
p ı ı
1 v2 Ê 2 ˆ
Á 1 q v˜
Ë ¯
1 v2
Û
sin È4 Bd v ÊË 1 - q 2 - vˆ¯ ˘ dq .
1
= (2-192a)
p Bd v Ù
ı ÍÎ ˙˚
0
It is evident that t (v) is independent of the sign of Bd , i.e., it is symmetric about the
Gaussian image plane.
The OTF may be written in a different form by first integrating over q and then over
p. The limits of integration for p and q in this case are different. They may be obtained
from Figure 2-42 by noting that the centers of the two unit circles lie at mv along the p
axis. Thus, the OTF may be written
1 v 1 ( p + v)2
1 Û Û
t (v) = dp exp ( 4iBd pv) dq
p Ù
Ù
ı ı
(1 v ) 1 ( p + v)2
1 v
4 Û
1 - ( p + v) cos ( 4 Bd pv) dp
2
=
p Ù
ı
0
1
4Û
= Ù 1 - s 2 cos [ 4 Bd v(s - v)] ds , (2-192b)
pı
v
where we have kept only the real part since the OTF is real and we have let p + v = s in
the last step.
Letting q = sin a , where 0 £ a £ b with cosb = v (see Figure 2-5), Eq. (2-192a)
yields the aberration-free result of Eq. (2-44b) if we let Bd Æ 0 . Expanding the sine
function in a power series, we may write
1 v2
4 Û ÈÊ 8
1 - q 2 - vˆ¯ - Bd2 v 2 1 - q 2 ( ) + ...˘˙ dq
32
t (v) = Ù ÍË
p ı Î 3 ˚
0
Ê b ˆ
1Á 8 2 2Û
= 2 - sin 2 - Bd v Ù cos d + ...˜
b b 4
a a
pÁ 3 ı ˜
Ë 0 ¯
184 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
=
1
p
(
2b - sin 2b - 4 Bd2 v 2b + ... )
2È
( ) - 2 Bd2 v 2 cos 1 v + ...˘ .
12
= cos 1 v - v 1 - v 2 (2-193)
p ÎÍ ˚˙
The first two terms in Eq. (2-193) correspond to the aberration-free OTF.
Noting that
p
cos 1 v =
2
- v + O v3 ( ) , (2-194)
we obtain
t (v) = 1 -
4v
p
( )
- 2 Bd2 v 2 + O v 3 . (2-195)
As expected, the slope of the defocused OTF at the origin is - 4 p , same as that of the
aberration-free OTF. The Hopkins ratio for small values of v may be written
t (v; Bd )
H (v) =
t (v; Bd = 0)
~ 1 - 2 Bd2 v 2 . (2-196)
For small values of v, the Hopkins ratio based on the geometrical OTF is also given by
Eq. (2-196) [see Eq. (2-228)]. Of course, since the aberration-free geometrical PSF is a
point (or a Dirac delta function), the corresponding OTF is a constant independent of the
spatial frequency. Hence, the geometrical Hopkins ratio for a certain spatial frequency is
simply the geometrical OTF for that frequency.
Bd = 0.1 v rad
or
Bd = l 20 v
= 1 20 Fvi . (2-197)
Dz = 8 lF 2 20 v
= 0.4 F v i , (2-198)
H (v) ~ 1 - 1 s Q2 . (2-199)
2
where
2
s Q2 = Q 2 - Q . (2-200)
We noted in Section 2.3 that a maximum possible value of the Strehl ratio of a
system for a particular aberration is obtained by a minimization of its variance across the
exit pupil, provided the Strehl ratio thus obtained is not too small. The approximate Eq.
(2-199) shows that the Hopkins ratio of a system at a particular spatial frequency is
maximum when the variance of the aberration difference function across the overlap
region of two displaced exit pupils is minimum, the displacement of the pupils being
dependent on the spatial frequency under consideration. An investigation31 of balancing
secondary spherical aberration with primary spherical aberration and defocus, and
r
secondary coma with primary coma, shows that unless H (v ) > ~ 0.r7 , a decrease in the
value of s Q may be associated with a decrease in the value of H (v ) for v £ 0.1. Thus,
the variance of the aberration or that of the aberration difference function is a useful
criterion of image quality, provided the corresponding Strehl or Hopkins ratio obtained is
relatively high. For spatial frequencies v > 0.1, it is more appropriate to use the Strehl
ratio as the criterion of image quality and aberration balancing.
We also noted in Section 2.3 that the Strehl ratio of an aberrated system depends
approximately on the variance of the aberration function across its pupil and not on the
type of the aberration. Similarly, we note from Eq. (1-214) that the Hopkins ratio at a
certain spatial frequency depends approximately on the variance s Q2 of the aberration
difference function across the overlap region of two displaced pupils (displacement
depending on the spatial frequency) and not on the type of the aberration. As in the case
186 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
of the Strehl ratio, a better approximation to the Hopkins ratio is obtained by using the
exponential relation
1 v2 1 q2 v
16 Bd2 v 2 Û Û
s 2d (v) = Ù dq Ù p 2 dp
2b - sin 2b ı ı
1 v2 1 q2 +v
1 v2
64 Bd2 v 2 Û Ê ˆ
3
= Ù Ë 1 - q - v¯ dq
2
3(2b - sin 2b) ı
0
( )
4v 2 ÈÍ 1 + 4v 2 2b - 13 + 2 v 2 sin 2b˘˙
Î
1
3 ˚
( )
= . (2-204)
2b - sin 2b
The variance of the aberration difference function for other aberrations can be
calculated in a similar manner. For the primary aberration function
it is given by32
{ ( )
s Q2 (v) = 4v 2 4 P1 Bd2 + 4 Aa2 P6 cos 2 f + P1 sin 2 f sin 2 f + 8 P1 Bd Aa sin 2 f
[ (
+ Ac2 4 P3 cos 2 f + 4 P4 - P52 sin 2 f )]
+ 16 P2 As2 + 16 P7 Bd As + 128 P4 Aa As } , (2-206)
where
am 2b - bm sin 2b
Pm = , (2-207)
2b - sin 2b
( )
a1 = 1 + 4v 2 4 , b1 = Ê13 + 2 ˆ 12 ,
Ë
1
8v ¯
(2-208)
2.12 Geometrical OTF 187
1 17 2 29 139 2 292 4 28 6
a2 = + v + 8v 4 + 8v 6 , b2 = + v + v + v , (2-209)
8 6 24 10 45 45
1 1 2 9 7 2 1 4
a3 = + v , b3 = + v - v , (2-210)
24 4 40 90 90
3 17 2 101 83 2 17 4
a4 = + v + 4v 4 , b4 = + v - v , (2-211)
8 4 40 15 30
(
a5 = 1 + 4v 2 , b5 = 11 + 4v 2 3 , ) (2-212)
5 1 2
a6 = 1 4 , b6 = - v , (2-213)
12 6
and
1 7 25 2 2 4
a7 = 1 + 2 v 2 + 2 v 4 , b7 = + v + v .
6 6 9 9 (2-214)
A comparison of the diffraction and geometrical OTFs was given in Section 1.6.7.
Since the aberration-free image of a point object in geometrical optics is a point, or a
Dirac delta function, the corresponding OTF is unity for all spatial frequencies. There is
no cutoff frequency as there is in the diffraction OTF. Of course, it would be unreal to
consider the geometrical OTF beyond the diffraction cutoff frequency. Whereas the
diffraction OTF depends strongly on the wavelength, the geometrical OTF is independent
of it, except for any variation due to chromatic aberration.
In this section, we consider the aberrated geometrical OTF. It was shown in Section
1.6.6 that the slope of the geometrical MTF at the origin is zero, unlike that of the
diffraction MTF. For low spatial frequencies, a simple but approximate expression is
given. It is shown that the MTF averaged over all angular orientations of a spatial
frequency vector is maximum when the variance of the ray aberration, or the root mean
square (rms) radius of the PSF with respect to its centroid, is minimum. The OTF
reduction for a radially symmetric aberration is directly related to the rms radius of the
PSF. The general formulas derived here are applied to obtain the OTFs for primary
aberrations. A closed-form exact expression is obtained in the case of astigmatism and/or
defocus. Generally, the geometrical OTF agrees with the diffraction OTF for very low
spatial frequencies when the aberration is larger than about a wavelength.24-26,33-36 This is
a region of practical interest since the MTF at high frequencies for such aberrations is
negligibly small. Gostick, for example, has shown that the geometrical MTF for very
small spatial frequencies obtained according to Eq. (1-150) in the case of secondary
spherical aberration or secondary coma gives an excellent approximation to the
diffraction MTF.35
188 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
(
t g (x, h) = 1 p a 2 ) Ú Ú exp [2pi (x x + hy)] dx p d yp , (2-215)
where ( x, y) are the ray aberrations (i.e., they are coordinates of a ray in the image plane
with respect to the Gaussian image point). Using polar coordinates for a pupil point; i.e.,
letting
Ê ∂W sinq ∂W ∂W cosq ∂W ˆ
( x, y) = 2 F Á cosq – , sinq + ˜ . (2-217)
Ë ∂r r ∂q ∂r r ∂q ¯
the MTF and the PTF for small spatial frequencies are approximately given by [see Eqs.
(1-144a) and (1-144b)]
and
respectively. Since the aberration-free MTF is unity for all spatial frequencies, the
aberrated MTF given by Eq. (2-219) represents the geometrical Hopkins ratio.
~ 1 - 2 p2vi
2
t g (vi , f) Ú Ú [( x - xc ) cos f + ( y - yc ) sin f]
2
dx p d y p . (2-222)
a
2p
1 Û
t g (vi ) = t g (vi ) df
2p Ù
ı
0
2.12 Geometrical OTF 189
~ 1 - pv2i
2
Ú Ú [( x - xc ) + ( y - yc )]
2
dx p d y p . (2-223)
a
Comparing Eqs. (2-222) and (2-223), we note that the average MTF is also equal to the
MTF for f = ± 45o . 33 Equation (2-223) shows that, for small spatial frequencies, the
average MTF is maximum when the standard deviation of the ray aberration, called the
spot sigma (or the rms radius of the ray spot diagram with respect to its centroid), is
minimum. Balanced aberrations in terms of linear combinations of Zernike polynomials
yielding minimum ray spot sigma are given in Section 4.4 of Part I.33,35 The centroid of
the PSF for an aberration balanced in this manner lies at the origin, i.e., xc = 0 = yc for
such an aberration.
∂W
( x, y) = 2 F (cos q, sin q) , (2-225)
∂r
and using polar coordinates for a spatial frequency and a pupil point, we obtain
1 2
t g (vi ) ~ 1 - 8(pFvi ) Ù ÊÁ ∂W ˆ˜ r dr
2Û
(2-226a)
ı Ë ∂r ¯
0
= 1 - (pv is s )
2
, (2-226b)
where s s is the spot sigma (see Chapter 4 in Part I). It is evident that the OTF for any
frequency is maximum when the spot sigma is minimum.
2.12.3 Defocus
W (r) = Bd r2 . (2-227)
W (r) = As r 4 + Bd r2 ; (2-229)
Ê 8 ˆ
s 2s = 8F 2 Á 2 As2 + Bd2 + AsBd ˜ . (2-230)
Ë 3 ¯
In this case, a closed-form expression for the exact OTF can be obtained. Substituting for
the ray aberrations
[
( x, y) = 4 Fr ( Aa + Bd ) cos q, Bd sin q ] , (2-233)
and using polar coordinates for a spatial frequency, Eq. (2-215) yields
1 2p
1Û Û
[
t g (vi , f) = Ù Ù exp 8pFGivi r cos(q - y ) r dr dq
pı ı
]
0 0
1
Û
= Ù 2 J 0 (8pFGivi r ) r dr
ı
0
2 J1 (8pFGvi )
= , (2-234a)
8pFGvi
where
[( A + B ) cos f + B sin f]
2 12
G = a d
2 2
d
2
(2-234b)
and
2.13 Incoherent Line and Edge Spread Functions 191
Bd
tan y = tan f . (2-234c)
Aa + Bd
x x3
J1 ( x ) = - + ... , (2-235)
2 16
we obtain
a result that can also be obtained by substituting Eq. (2-233) into Eq. (2-219) and noting
that xc = 0 = yc for the aberration under consideration. The smallest image spot is
obtained when Bd = - Aa 2, in which case it is circular with a uniform irradiance. In that
case, G reduces to Aa 2, y Æ - f , and t g becomes independent of f , as expected for a
uniform circular image. This is not true of the diffraction image, which is nonuniform and
results in an OTF that depends on f , as was seen in Figure 2-46.
2.12.6 Coma
Finally, we consider coma given by
and
( xc , yc ) = (2 FAc , 0) . (2-239)
and
The OTF given by Eqs. (2-240) can also be obtained by determining the ray aberration
along the p axis and substituting its mean value and variance into Eqs. (1-155) (see
Problem 8).
The balanced aberrations and their tolerance are given for a Struve ratio of 0.8. The
expressions for the LSF and ESF are also given in terms of the OTF. Plots of numerical
results for primary aberrations are provided.
2.13.1 Theory
2.13.1.1 LSF from PSF
The line-spread function (LSF) of an imaging system is, by definition, the image of a
line object. For an object lying along the yo axis, it is given by [see Eq. (1-219)]
( )
12
or, substituting ri = xi2 + yi2 , we obtain
•
Û PSF ( ri )
LSF( xi ) = 2Ù ri dri , (2-242)
(r )
12
ı i
2
- xi2
xi
i.e., the LSF is the Abel transform40 of the PSF. Substituting Eq. (2-15), the aberration-
free LSF is given by
•
8 Û J12 ( pri )
LSF( xi ) = 2Ù dri . (2-243)
p ı r r2 - x2 1 2
i i xi i ( )
2.13.1.2 LSF from Pupil Function37
The infinite limits over the PSF integration can be avoided by writing it in terms of
the pupil function, as was done in obtaining Eq. (1-221a). For a circular pupil of radius a,
the normalized LSF in terms of the pupil function may be written
2
a a 2 y 2p
Û
Ù dy p
ı
Û
Ù
ı
[ ( Ê 2 pi
exp iF x p , y p exp Á -
Ë l
)]
R
ˆ
xi x p ˜ dx p
¯
a
a 2 y 2p
LSF ( xi ) = ,
a
È a 2 y 2p ˘
Í ˙
Û Í Û ˙
Ù dy p Ù dx p
ı Í ı ˙
a Í a 2 y 2p ˙
Î ˚
2.13 Incoherent Line and Edge Spread Functions 193
or
2
1 1 y n2
3 Û Û
LSF ( x ) =
16 Ù
ı
dyn Ù
ı
[ ]
exp iF( x n , yn ) exp ( - pi xx n ) dx n , (2-244)
1 1 y n2
( )
where F is the phase aberration function, ( x n , yn ) = x p , y p a are the coordinates of a
pupil point normalized by the pupil radius, and x = xi l F is the distance of an image
point in units of l F . The LSF is normalized by its aberration-free central value.
For a line object along a yo¢ axis (or parallel to it) making an angle b with the yo
axis (see Figure 1-9), the LSF( xi¢) is obtained by writing the pupil function in the
( )
coordinate system x ¢p , y ¢p by replacing x p by x ¢p cos b - y ¢p sin b and y p by
x ¢p sin b + y ¢p cos b . For a rotationally symmetric imaging system, since the aberration
function for a point object along the x o axis consists of terms with integral powers of x p
and x 2p + y 2p , the aberration function in the rotated coordinate system is obtained by
replacing x p by x ¢p cos b - y ¢p sin b and x 2p + y 2p by x ¢p2 + y ¢p2 .
2
1 1 y n2
=
3Û
Ù dy n
16 ı
Û
Ù
ı
{[
exp i F( x n , y n ) - F( y n ) ]}d x n
1 1 y n2
3Û
1 1 y n2
Û Ô [
Ï1 + i F( x n , y n ) - F( y n ) ] ¸
Ô
= Ù dy n Ù Ì 1 ˝d x n
16 ı
1
ı
1 y n2 Ó 2
[
Ô - F( x n , y n ) - F( y n ) ] 2
+ ...Ô
˛
1
2
~ 3Û
4Ù
ı
(
dyn 1 - yn2 ) ÈÍÎ1 - 12 s ( y )˘˙˚
2
F n
1
~ 3Û
4Ù
ı
( )[
dyn 1 - yn2 1 - s 2F ( yn ) ]
1
194 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1
3Û
(
= 1 - Ù 1 - yn2 s 2F ( yn ) dyn
4ı
) , (2-245)
1
where
1
È 1 yn2 ˘ 1 y n2
Í Û ˙ Û
F ( yn ) = Í Ù dx n ˙ Ù F( x n , yn ) dx n
Í ı ˙ ı
ÍÎ 1 yn2 ˙˚ 1 y n2
1 y n2
1 Û
= Ù F( x n , yn ) dx n (2-246)
2 1- yn2 ı
1 y n2
is the mean value of the aberration function along the x n axis for a given value of yn and
s 2F ( yn ) = F 2 ( yn ) - F( yn )
2
(2-247)
is the corresponding aberration variance. We have already shown [see Eq. (1-221c)] that
this ratio is less than or equal to 1.
Now we outline the derivation of Struve ratio for defocus and astigmatism and give
tolerance for a Struve ratio of 0.8 for some aberrations.
2.13.1.3.1 Defocus
(
F( x n , yn ) = Bd x n2 + yn2 ) , where Bd is in radians.
1 y n2
F ( yn ) =
2 1-
Bd
yn2
Û
Ù
ı
(x 2
n )
+ yn2 dx n
1 y n2
=
Bd
3
(
1 + 2 yn2 ) .
1 y n2
Bd2
F 2 ( yn ) =
Û
(x )
2
Ù
2
n + yn2 dx n
2 1- yn2 ı
1 y n2
= Ê +
1 4 2 8 4ˆ 2
y + y B .
Ë 5 15 n 15 n ¯ d
4 Bd2
s 2F ( yn ) = ( )
2
1 - yn2 .
45
2.13 Incoherent Line and Edge Spread Functions 195
1
128 Bd2
Û
ı
( 2 2
)
Ù 1 - yn s F ( yn ) dyn = 1575 .
1
s 2F ( yn ) =
4
( ) (A + Bd )
2 2
1 - yn2 a
45
= 0 if Bd = - Aa .
Hence, a line object parallel to the yo axis is not affected by astigmatism if the image is
observed in a defocused plane Bd = - Aa .
(
F( x n¢ , yn¢ ) = Aa yn¢ 2 + Bd x n¢ 2 + yn¢ 2 ) .
s 2F ( yn¢ ) =
16
( )
2
1 - yn¢ 2 Bd2 .
45
Hence, a line object parallel to the x o axis is not affected by astigmatism if its image is
observed in the Gaussian image plane ( Bd = 0 ).
b = p 4:
F( x n¢ , yn¢ ) =
1
2
Aa ( x n¢ - yn¢ ) + Bd x n¢ 2 + yn¢ 2
2
( ) .
s 2F ( yn¢ ) =
1
( 4
) ( ) (B )
2
1 + 13 yn¢ 2 - 14 yn¢ 4 Aa2 + 1 - yn¢ 2 2
d + Bd Aa .
45 45
Aa2
s 2F ( yn¢ ) =
3
( )
1 - yn¢ 2 yn¢ 2 .
1
16 Aa2
Û
ı
( 2 2
)
Ù 1 - yn¢ s F ( yn¢ ) dyn¢ = 315 .
1
The balanced aberrations and tolerance for a Struve ratio of 0.8 are given in Table 2-
16. Comparing with the Strehl ratio tolerances given in Tables 2-5 and 2-6, we note that
the tolerance for defocus is slightly larger than that for a Strehl ratio of 0.8; the tolerance
for spherical aberration is about the same; the tolerance for coma is about the same when
b = 0 , but it is less severe for b = p 2; the tolerance for secondary coma is more severe
for b = 0 and less severe for b = p 2 . Note that when b = 0 , the line object and its
Gaussian image lie along the y axis, which is perpendicular to the symmetry axis of the
coma PSF.
The LSF along any other direction may be obtained by using the corresponding profile of
the transfer function. Thus, for example, the LSF for a line object along a direction yo¢ (or
parallel to it) making an angle b with the yo axis is given by [see Eq. (1-229)]
Table 2-17. Balanced aberrations and their tolerance for a Struve ratio of 0.8.
3 1 Tolerance for a
Aberration Balanced aberration
4 Ú
–1
(1 – yn2 ) s w2 ( yn ) dyn Struve ratio of
0.8 in units of l
Primary 4
As (r – 0.984 r ) 2 0.1835 As2 1.09 in the plane
spherical Bd = – 0.984As
As r4
Secondary Aa yn2 cos q for b = p / 2 0.0124 W62 4.02
spherical
6
W6 r
where x ¢ and x ¢ are parallel to each other and normal to y ¢ , and t (x¢, 0) is the OTF
along the direction x ¢ .
For the aberration-free case, the OTF is real given by Eq. (2-44). Hence, noting that
-1 £ x £ 1 , Eq. (2-248a) reduces to
1
Û
LSF( x ) = 2Ù t (x, 0) cos (2 p x x )dx , (2-249)
ı
0
2È
( )
1 2˘
t (x, 0) = cos 1
x - x 1 - x2 , -1 £ x £ 1 . (2-250)
p ÍÎ ˙˚
1
Û
LSF(0) = 2Ù t (x, 0) dx = 8 3p . (2-251)
ı
0
1
Û
LSF( x ) = (3p 4) Ù t (x, 0) cos (2 p x x )dx (2-252a)
ı
0
3
= H1 (2 p x ) , (2-252b)
8p x 2
where H1 ( x ) is the first-order Struve function.41 Its behavior for small values of x may be
obtained by expanding cos (2p x x ) in a power series and integrating term by term. Thus,
1
Û
[
LSF( x ) = (3p 4) Ù t (x, 0) 1 - (2 p x x ) + O x 4 dx
ı
2
( )] (2-253a)
0
1
3p
= 1-
4
(2 p x)2 Û
Ù t (x, 0) x dx + O x
ı
2 4
( ) . (2-253b)
0
1
Û
LSF(0) = (3p 4) Ù t (x, 0) dx . (2-254)
ı
0
and obtain
È t ¢ (0, 0) ˘
LSF( x ) ~ (3p 4) Í 2 ˙+O x
ÍÎ (2 p x ) ˙˚
( )
4
= 3 4p 2 x 2 , (2-256)
where we have used Eq. (2-50) in the last step. Since t ¢ (0, 0) = - 4 p is independent of
the aberration, so is (approximately) the asymptotic behavior of the LSF.
1 x0
Û Û
Lc ( x 0 ) = 4Ù dx t (x, 0) Ù cos(2px x ) dx
ı ı
0 0
1
Û sin(2 px x 0 )
= 2 Ù t (x, 0) dx . (2-258)
ı px
0
L c ( x 0 Æ •) = 1 . (2-259)
1
1 Û sin(2 px x )
ESF( x ) = + t( x , 0 ) dx (2-260)
2 Ù
ı px
0
=
1
2
[1 + Lc ( x)] . (2-261a)
ESF (± x ) = [1 ± Lc ( x ) ] 2 . (2-261b)
Moreover, since Lc ( 0) = 0 , it is evident from Eq. (2-261a) that, regardless of the value or
the type of a symmetrical aberration, the central value of the ESF is equal to 1/2, i.e.,
If the aberration is not radially symmetric as, for example, in the case of coma, then
the ESF will depend on the orientation of the edge. For an object along a direction yo¢
making an angle b with the yo axis (see Figure 1-9), Eq. (1-233) may be written
1
1 1 Û t( x ¢ , 0 )
ESF( x ¢) = - exp( - 2 p ix ¢ x ¢) d x ¢ . (2-262a)
2 2 pi Ù
ı x¢
1
is shown in Figure 2-55. It has a value of 0.5 at the origin. The value of the ESF aberrated
by primary coma at the Gaussian image of the edge as a function of the peak value Ac of
coma is shown in Figure 2-56. When b = p 2 , this value is equal to 0.5 regardless of the
value of the aberration. The ESF for primary coma is shown in Figure 2-57.
0.8
0.6
LSF(x)
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
x
(a)
100
10–1
LSF(x)
10–2
10–3
0 0.5 1 1.5 2 2.5 3 3.5 4
x
(b)
Figure 2-52. Aberration-free LSF normalized to unity at the center. (a) Linear scale.
(b) Log scale. x is in units of lF. The radius of the Airy disc is 1.22 in these units.
2.13 Incoherent Line- and Edge-Spread Functions 201
0 .8
Bd = 1/4
0 .6
3:
LL
"l (a)
!
0 .4
0 .2
00 0 .5 1.5 2.5
- x
0 .8
As = 1/4
0.6
x
li:'
(f)
...J (b)
!
0.4
0.2
0.8
0.6
As = 1
3:
LL
"l
! (c)
0.4
0.2
. .. ... . .. . .
Figure 2-53. Aberrated LSF. (a) Defocus. (b) Primary spherical aberration Asp4. (c)
Balanced spherical aberration As (p4 - p2).
Ed and As are in units of A and x is in
units of AF.
202 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Ac =1.0
0.8
0.6
x
iC
~
t (a)
0.4
0.2
0_
3 -2 - 1
-x
Ac =1
0.8 ,/2
,,
~
\
\
,/4
0.6
x
iC
~
(b)
0.4
I
I
0.2 I
0
-3 -2 - 1
- x
Ac= 1.5
0.8
0.6
R , \
U. I \
~ I (c)
Tt/2
t I
0.4
~=O
0.2
°_3 -2 - 1
- x
0.8
0.6
ESF(x)
0.4
0.2
0
4 3 2 1 0 1 2 3 4
x
Figure 2-55. Aberration-free ESF. The Gaussian image of the edge has a value of
zero for x < 0 and unity for x ≥ 0 . The vertical dashed line indicates the Gaussian
image of the edge object and x is in units of lF. The edge object lies along a
direction making an angle b with the yo axis and the variation with x shown here is
along a direction that is normal to it.
0.5
0.4
ESF(0)
0.3
0.2
p/4
0.1
b=0
0.0
0 0.5 1.0 1.5 2.0 2.5
Ac
Figure 2-56. ESF value at the Gaussian image of the edge as a function of the peak
value Ac of primary coma Ac r3 cos q in units of l .
204 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1.0 Ac = 0.5
0.8
p/4
0.6
ESF(x)
0
0.4 b = p/2
0.2
0.0
–8 –6 –4 –2 0 2 4 6 8
x
1.0
Ac = 1.0
0.8
p/4
ESF(x)
0.6
0
0.4 b = p/2
0.2
0.0
–8 –6 –4 –2 0 2 4 6 8
x
1.0
Ac = 1.5
0.8
p/4
0.6
ESF(x)
0
0.4 b = p/2
0.2
0.0
–8 –6 –4 –2 0 2 4 6 8
x
Figure 2-57. ESF for primary coma Ac r3 cos q for various values of Ac in units of l .
2.14 Miscellaneous Topics 205
In this section, we develop an expression for the polychromatic PSF (PPSF) such that
parametric numerical analysis can be carried out in terms of the mean wavelength and the
spectral bandwidth of the object radiation. 44,45 Although the object radiance will generally
vary with wavelength (e.g., as in the case of a tungsten lamp or solar illumination), we
assume it to be constant for simplicity of numerical calculations. The primary objective of
such calculations is to see how the diffraction rings loose their sharpness as the spectral
bandwidth is increased, and to determine the bandwidth for which they completely
disappear. In practice, of course, the spectral transmittance of the imaging system and
spectral sensitivity of the observation medium, e.g., a film or a solid-state detector array,
must also be taken into account to determine the measured PPSF.
p p dl È 2 J1 ( p ri l F ) ˘
2
I (ri ; dl ) = Í ˙ , (2-263)
4l2 F 2 Î p ri l F ˚
where p is the image power per unit spectral bandwidth ( p dl is the total image power),
and l is the mean wavelength. Assuming for simplicity that the polychromatic object
radiates uniformly as a function of wavelength, the PPSF is obtained by integrating Eq.
206 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
I (ri ; Dl ) = Ù Í ˙ dl . (2-264)
4 F 2 ı Î p ri l F ˚ l2
l1
Normalizing the PPSF with the central value p p dl 4l2m F 2 of the QPSF, we may write
l2
l2 Û È 2 J ( p r l F ) ˘ 1
2
In (ri ; Dl ) = m Ù Í 1 i ˙ dl , (2-265)
dl ı Î p ri l F ˚ l2
l1
Dl 1
= , (2-266)
dl 1 - ( Dl 2 l m ) 2
Ê Dl ˆ
l1 = l m Á1 - ˜ (2-267a)
Ë 2l m ¯
and
Ê Dl ˆ
l 2 = l m Á1 + ˜ , (2-267b)
Ë 2 lm ¯
Dl 10 4
In (0; D l ) = . (2-268)
lm Ê Dl ˆ
2
1- Á ˜
Ë 2l m ¯
Figure 2-58a shows the variation of the normalized central irradiance as a function of the
normalized spectral bandwidth.
107
106
105
n (0)
104
103
102
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Dl/l m
(a)
106
104
1 Dl / lm = 1.99
102
In (r; Dl)
0.25
0.1
100
2
10
2
10
4
10
4
10
6
10
0 1 2 3 4 5 6 7 8 9 10
r
(b)
Figure 2-58. (a) Aberration-free central irradiance of the PPSF normalized by the
central value of the QPSF as a function of the relative spectral bandwidth Dl l m .
(b) Aberration-free PPSF for various relative spectral bandwidths. The units of r
are l m F.
208 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Dl
1+
2l m
4 ¥ 10 4 Û
In ( r; D l ) = Ù J12 ( p r l n ) d l n , (2-269)
p2r 2 ı
Dl
1
2l m
which is suitable for parametric numerical analysis. Figure 2-58b shows the PPSF for
various values of the relative spectral bandwidth Dl l m . The diffraction rings begin to
lose their identity as Dl l m increases. However, the ring structure does not disappear
until Dl l m = 1, i.e., until the spectral bandwidth equals the mean wavelength.
l2
1 Û
t p (vi ) = t(vi ) d l
Dl Ù
ı
l1
l2
Or,
q
Û È
( xv) - xv(1 - x 2 v 2 ) ˘dx , 0 £ v £1 ,
2 12
t p (v) = Ù cos
1
(2-271)
p (q - 1) ı ÎÍ ˚˙
1
0.8
q=1
0.6
2
tp (n)
4
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
n
2 + Dl lm
q = . (2-272)
2 - Dl lm
(
Pen = p ho2 Sen zo2 B . ) (2-273)
The total power in the exit pupil and, therefore, in the image is given by Pex = hPen ,
where h is the transmission factor of the system, as in Eq. (2-3). This power is uniformly
distributed in the circular Gaussian image of radius hg = Mho , where M is the
magnification of the image. Hence, the irradiance of the Gaussian image is given by
r
( )
Ig rg = Pex p hg2
(
= h Sen zo2 M 2 B )
r
= Ig for rg £ hg
= 0, otherwise , (2-274)
r
where rg is the position vector of a point on the Gaussian image.
r
where ri is the position vector of a point in the image plane and
r r
s = ri - rg
[ ( )]
12
= ri2 + rg2 - 2 ri rg cos q i - q g . (2-276)
The irradiance distribution is radiallly symmetric since the integral does not depend on
q i ; indeed we may let q i = 0 in Eq. (2-276). Letting
r = ri l F, r g = rg l F, and bg = hg l F , (2-277)
where
( )
12
t = r 2 + r2g - 2 ri r cos q g . (2-279)
( )
bg
Û 2 J1 pr g
2
Ii ( 0 ) = I g Ù dr g
ı rg
0
[ ( )
= Ig 1 - J 02 p bg - J12 p bg ( )] . (2-280)
The variation of Ii (0) with bg is similar to the variation of the encircled power of the
Airy pattern [see Eq. (2-24) and Figure 2-2]. It approaches Ig asymptotically.
The diffraction image can also be obtained by working in the spatial frequency
domain, i.e., by inverse Fourier transforming its spectrum, which itself is equal to the
product of the spectrum of the Gaussian image and the OTF of the imaging system. Since
the Gaussian image is radially symmetric, its spectrum (equal to its Fourier transform) is
given by its zero-order Hankel transform:
hg
r Û
(
Ĩg (vi ) = 2 p Ig Ù J 0 2 pvi rg rg drg
ı
)
0
( ) (
= Ig hg vi J1 2pvi hg ) . (2-281)
r
( ) (
= Ig hg vi J1 2pvi hg t(vi ) , ) (2-282)
r
where t(vi ) is the aberrated OTF. The irradiance distribution of the image is obtained by
r
inverse Fourier transforming the spectrum, i.e., Ii (ri ) is given by
r r r r r
Ii (ri ) = Ú I˜i (vi ) exp( - 2pivi ◊ ri ) dvi
r r r r
( )
= Ig hg Ú (1 vi ) J1 2 pvi hg t(vi ) exp( - 2 pivi ◊ ri ) dvi . (2-283)
The OTF of a system with a radially symmetric aberration is also radially symmetric
r
[see Eq. (1-83)] and, therefore, we may replace t(vi ) by t(vi ) . Accordingly, the image
spectrum is also radially symmetric. Its inverse Fourier transform, which gives the
irradiance distribution of the diffraction image, is equal to its zero-order Hankel
transform, i.e.,
212 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
( )
= 2 p Ig hg Ú J1 2 pvi hg J 0 (2 pvi ri ) t(vi ) dvi . (2-284)
( )
Ii (r ) = 2 p I g bg Ú J1 2 pvbg J 0 (2 pvr ) t(v) dv . (2-285)
(
Ii (0) = 2 p Ig bg Ú J1 2 pvbg t(v) dv .) (2-286)
For an aberration-free system, the integration can be carried out in a closed form [see
Eqs. (2-50) and (2-24)] with the result
[ ( )
Ii (0) = Ig 1 - J 02 p bg - J12 p bg ( )] , (2-287)
For a system with a radially symmetric aberration, the power in a circle of radius rc
(in units of l F ) is given by
rc
Û
P (rc ) = 2 p (l F ) Ù Ii (r ) r dr .
2
(2-288)
ı
0
rc
P (rc )
Û
( )
= (2 pl F ) Ig bg Ú dv J1 2 pvbg t(v) Ù J 0 (2 pvr ) r dr
2
ı
0
( ) ( )
= 2 Pex rc bg Ú J1 2 pvbg J1 (2 pvrc )t(v) (1 v)dv . (2-289)
Letting rc = bg , we obtain the power in a circle whose radius is equal to that of the
Gaussian image:
( ) ( )
P bg = 2 Pex Ú J12 2 pvbg t(v) (1 v)dv . (2-290)
( )
It is easy to see from Eq. (2-289) that the power P rc ; bg in a circle of radius rc due to a
( )
source of radius bg and the power P bg ; rc in a circle of radius bg due to a source of
radius rc are related to each other according to
( ) (
bg2 P rc ; bg = rc2 P bg ; rc ) . (2-291)
For a very small disc, i.e., for very small values of bg , e.g., a pinhole,
2.14 Miscellaneous Topics 213
(
J1 2pvi hg ) ~ pvi hg , (2-292)
Thus, as expected, it reduces to the irradiance distribution of the image of a point object.
This result may also be obtained from Eq. (2-285). In that case, the PSF is the zero-order
Hankel transform of the OTF because of the radial symmetry.
It is much simpler to calculate the encircled power from Eq. (2-289) using the
transfer function, than, for example, by first calculating the irradiance distribution of the
extended image and then integrating it over a circle, as was done by Goldberg and
McCulloch49 for an aberration-free system. Substituting the value of t(v) in Eq. (2-285),
we can calculate the irradiance distribution of the diffraction images.
The encircled power of the aberration-free image of a disc is shown in Figure 2-61a
for several disc sizes. Again, the curve for bg = 1 4 closely resembles the encircled
power of the Airy pattern, confirming the size of a pinhole that can be treated as a point
source. The power in a circle whose radius is equal to that of the Gaussian image of a disc
is shown in Figure 2-61b as a function of the disc size. It increases monotonically as the
disc size increases. For small values of bg , this circle contains only a small fraction of the
total power since it is spread by diffraction. It contains a significant fraction of the total
power when its size is significantly larger than the Airy disc, since then the relative effect
of diffraction is small.
Figure 2-62a shows how the central irradiance varies as a function of bg for several
defocus values. It increases according to Eq. (2-286) and approaches unity (actually Ig )
asymptotically. The axial irradiance of a disc of various sizes is shown in Figure 2-62b.
For bg = 1 4, it is approximately the same as for a point source [see Eq. (2-84c)]. For
example, it reduces to a value of nearly zero for integral values of Bd (in units of
wavelength). As the disc becomes larger, the axial irradiance does not reduce to zero for
such values of Bd .
214 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
0.8
2
0.6
i(r)
0.4 1
1/2
0.2
bg = 0
1/4
0
0 0.5 1 1.5 2 2.5 3
r
Figure 2-63 shows the defocused images of a disc of various sizes. We note that the
image generally resembles the object. In particular, it is bright in the central region and
dim in the outer region. As defocus increases, the irradiance decreases in the central
region and increases in the outer region. As stated earlier, if the disc is small and the
defocus is large, the irradiance at the center may be smaller than that in the outer region;
see for example, Figure 2-63a for Bd = 1l . This behavior is similar to that for a point
source, and it disappears as the disc size increases.
In Section 2.15.5, we consider the image of a coherently illuminated disc and show
that, unless the disc is very small, the central irradiance of the image can be much lower
than that in the surrounding region, whether or not the image is defocused. Moreover,
defocus can increase the central irradiance. As shown in Figure 2-77, the irradiance at the
center is lower than that in the surrounding region, for example, when bg = 2 and
Bd = 0 . The central irradiance increases significantly as defocus is introduced; i.e., for
Bd π 0 .
2.14 Miscellaneous Topics 215
0.8 bg 1/4
1/2
1
2
0.6
P(rc)
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
rc
(a)
1.0
0.8
0.6
P(bg)
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
bg
(b)
Figure 2-61. (a) Encircled power of the aberration-free image of an incoherent disc
normalized by the total power Pex . rc is in units of lF. (b) Power in a circle of
radius equal to that of the Gaussian image of the disc as a function of its size.
216 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Bd 0
0.8
1/4
1/2
0.6
(0 bg)
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
bg
(a)
0.8
4
0.6
2
(0 Bd)
1
bg 1/4
0.4
0.2
0
0 0.5 1 1.5 2 3 4
Bd
(b)
Figure 2-62. (a) Central irradiance of the defocused image of an incoherent disc. Bd
is in units of l . (b) Axial irradiance normalized to unity at the center. The actual
values (normalized by I g ) in increasing order of bg are 0.14, 0.825, 0.91, and 0.94.
2. 14 Miscellaneous Topics 217
bg = 1
0.8
0.6
(a)
§
0.4
0.2
00 2.5 3.5
-· r
bg =2
0.8
0.6
f (b)
0.4
0.2
00 0.5 3.5
-r
bg =2.3
0.8
0.6
§
(c)
0.4
0.2
00 0.5 3.5
-r
Based on geometrical optics, the image of a distant point object in the absence of a
lens will be approximately the same size as the pinhole if the pinhole is large. Reducing
the pinhole size reduces the image size until diffraction by the pinhole spreads it. Petzval
obtained the relationship by minimizing the image spot radius representing the sum of the
geometrical and diffraction contributions to it. The image radius based on diffraction (see
Figure 2-64) is approximately equal to the distance from the center of the image at which
a disturbance from the center of the pinhole cancels that from its edge. Thus, the
disturbance at the point P in the image plane at a distance L from the plane of the pinhole
is zero when the difference in optical path lengths between the disturbances from the
points O and A is equal to l 2, i.e., when
OP - AP = OC = l 2 . (2-294)
rd = L q
= L(OC a) (2-295)
= Ll 2 a
Adding the geometrical and diffraction contributions, we obtain the image spot radius as
ri = a + L l 2 a . (2-296)
rd
A
a
q q
O C
B
L
Figure 2-64. Diffraction spot radius based on the cancellation of disturbances from
the center and the edge of a pinhole of radius a. The camera length L >> a .
2.14 Miscellaneous Topics 219
The smallest spot radius is obtained by differentiating with respect to a and equating the
result to zero. This yields the result that
a P = (L l 2)
12
. (2-297)
Substituting Eq. (2-297) into Eq. (2-296), we obtain the minimum spot radius
ri = 2 aP (2-298a)
= (2 Ll )
12
. (2-298b)
The difference between a pinhole camera and a regular camera is that the former
does not have a lens to form the image. The lens in a regular camera converts a diverging
spherical wave from a point object into a spherical wave converging to a point in the
image plane. The absence of a lens in a pinhole camera may be thought of as a lens of
zero power. It implies that, except for diffraction by the pinhole, the diverging spherical
wave continues as a diverging wave toward the image plane. The corresponding defocus
aberration is given by
1Ê 1 1ˆ 2
( )
W rp = Á - ˜ rp
2 Ë Li Lo ¯
, (2-299)
where rp is the radial distance of a point in the plane of the pinhole from its center and
Lo and Li are the object and image distances from it (where Lo is numerically negative
according to our sign convention). It represents the sum of the sags of two spherical
wavefronts passing through the center of the pinhole with their centers of curvature lying
at the object and image points. The image will be practically diffraction limited according
to the Rayleigh criterion, if the peak value of the aberration is less than or equal to l 4 .
For a distant object ( Lo Æ - •), letting rp = a , Eq. (2-299) yields the Petzval result of
Eq (2-297). The image spot for a point object is approximately the Airy disc with a radius
of 0.61l Li a , or 0.61 times the value estimated by Petzval.
Another approach to optimize the pinhole radius is to maximize the central irradiance
of the diffraction pattern of a circular aperture. The axial irradiance for a circular aperture
is given by Eq. (2-84), where P is the power entering the pinhole and Bd is the peak
defocus phase aberration. For a point source of intensity Io , the power entering the
pinhole is given by the solid angle pa 2 L2o that it subtends at the source multiplied by
Io . Hence, the axial irradiance is given by
4 Io Ê p a 2 Li - Lo ˆ
I (0) = sin 2 Á ˜ . (2-300)
( Li - Lo )2 Ë 2 l Li Lo ¯
For a given object distance Lo , the maximum central irradiance is obtained at an image
distance Li given by
220 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
∂I (0)
= 0 , (2-301)
∂Li
or
È pa2 Ê1 1 ˆ˘ pa2 Ê1 1 ˆ Lo
tan Í Á - ˜˙ = - Á - ˜ (2-302)
ÍÎ 2 l Ë Li Lo ¯ ˙˚ 2l Ë i
L Lo ¯ Li
or
p p
tan ÈÍ ( Ni + No )˘˙ = ( Ni + No ) i
N
, (2-303)
Î2 ˚ 2 No
1
-
1
~ 1 ÈÍ1 + 42 No + Ê1 - 82 ˆ No2 ˘˙ , (2-304)
Li Lo f Î p
•
Ë p ¯ ˚
or
Ni ~ 1 - Ê1 - 42 ˆ No + Ê1 - 82 ˆ No2 , (2-305)
Ë p ¯ Ë p ¯
where f• = a 2 l is the focal length of the pinhole (i.e., the image distance when the
object distance is infinity).
Unless the object distance is very small, we may write Eq. (2-304) approximately as
1 1 1
- = , (2-306)
Li Lo fe
where
1
~ 1 + 0.4 No (2-307)
fe f•
is the effective focal length of the pinhole. We note that fe > f• and the maximum
irradiance is obtained at a point closer to the pinhole as the object moves closer to it, or
the pinhole subtends more than one Fresnel zone. This is in contrast to the case of a thin
lens for which the real image of a real object moves in the same direction as the object.
Rayleigh 52 also investigated Eq. (2-300) and, based on experimental observations,
concluded that the best image was obtained when a R = 0.95 ( Li l ) , which is equivalent
12
to the pinhole intercepting only 90% area of the first Fresnel zone.
2.14 Miscellaneous Topics 221
Figure 2-65 shows how Ni varies with No according to Eq. (2-305). We note that
Ni decreases as the object moves closer to the pinhole (so that No increases), reaches a
minimum value, and then increases monotonically. The value of a as a function of Lo is
shown in Figure 2-66 for several values of Li . All the quantities are in units of l in this
figure. We note that, for a given value of Li , a does not change much as Lo increases.
2.5
2
Ni
1.5
0.5
0 1 2 3 4 5
No
350
L i = 100,000
300
70,000
250
a
200 40,000
150
10,000
100
50
105 106 107 108
ΩLoΩ
Figure 2-66. Optimum value of pinhole radius a as a function of object distance for
several values of camera length Li . All three quantities are in units of wavelength.
222 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Since the focal length of the pinhole camera, approximately equal to a 2 l , depends
on the wavelength, it suffers from chromatic aberration. Similarly, since the pinhole
appears to be elliptical from an off-axis point object, its focal length for an object in the
horizontal plane differs from that in a vertical plane. Hence, it suffers from astigmatism.
However, it is free of distortion.
For an isoplanatic coherent object, the spatial frequency spectrum of its amplitude
image is given by the product of the spectrum of its Gaussian amplitude image and the
coherent transfer function of the imaging system. We give expressions for the aberration-
free point-, line-, and edge-spread functions. The transfer function is shown to be simply
a scaled version of the pupil function. The effect of the aberrations on the line- and edge-
spread functions is also discussed. As an example of an extended object, the image of a
coherent disc is considered. It is shown that, unless the disc is very small, defocus does
not necessarily decrease the central irradiance of its image.
r 1 Û r Ê 2 pi r r ˆ r
CSF( ri ) = Ù
lR ı
( )
G rp exp Á -
Ë lR
rp ri ˜ d rp
¯
◊ , (2-308)
where we have omitted the factor of - i in the front of the right-hand side of the
equation. If a system with a circular exit pupil of radius a is aberration free, then its
relative pupil function is constant, say Go , that is inversely proportional to zo , and may
be written
(r )
G rp = Go
r
, for rp £ a ,
= 0 , otherwise . (2-309)
a
2p Û Ê 2 p rp ri ˆ
= Ù J0 Á ˜ rp drp
lR ı Ë lR ¯
0
p a 2 2 J1 ( p r )
= , (2-310)
lR pr
r r
( )
If F rp is the phase aberration function where rp £ a , and if the amplitude across the
pupil is uniform, then
r r
[
CTF( vi ) = exp iF(l Rvi ) ] . (2-312)
Thus, for an aberration-free system with a circular exit pupil of radius a, all frequencies
r
vi £ 1 2l F are transmitted by the system without any amplitude or phase distortion.
224 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
0.8
0.6
CSF (r)
0.4
0.2
– 0.2
0 1 2 3 4
r
r
The spatial frequency vic = 1 2l F is called the cutoff frequency of the system in the
sense that object frequencies corresponding to image frequencies greater than 1 2l F are
not transmitted by it. We note that the cutoff frequency for coherent imaging is half of
that for incoherent imaging.
LSF( x ¢) =
1Û
2Ù
ı
{[
exp i F(x ¢, 0) - px ¢ x ¢ dx ¢ , ]} (2-313)
1
[ ]
where F(x ¢, 0) is the aberration function and exp iF(x ¢, 0) is the CTF along the
direction x ¢ , x ¢ is in units of l F , x ¢ is in units of 1 2l F, and division by 2 on the
right-hand side normalizes the aberration-free LSF to unity at the origin. For an even
aberration such as defocus or astigmatism, the LSF may be written
1
Û
[ ]
LSF( x ¢) = Ù exp iF(x ¢, 0) cos( px ¢ x ¢) dx ¢ .
ı
(2-314a)
0
It is complex and symmetric about x ¢ = 0 . For an odd aberration such as coma, the LSF
is real and given by
2.15 Coherent Imaging 225
1
Û
[
LSF( x ¢) = Ù cos F(x ¢, 0) - px ¢ x ¢ dx ¢ .
ı
] (2-314b)
0
sin p x
LSF( x ) = , (2-315)
px
Figure 2-69 shows the irradiance distribution associated with a defocused LSF, i.e.,
when F(x) = Bd x 2 , where Bd is in radians. However, the value of Bd given in the figure
is in units of wavelength. The curve for Bd = 0 is the aberration-free case and represents
simply the square of the LSF shown in Figure 2-68. A defocus of Bd = l corresponds to
a PSF with a central value of zero. Figure 2-70 shows the irradiance distribution
associated with the LSF aberrated by one wave of spherical aberration balanced with
defocus, i.e., for W (r) = As r 4 + Bd r2 with As = l . The optimum balancing defocus for
imaging of a point object is Bd = - As , but it is given by Bd = - 0.86 As for a line object.
The irradiance distribution associated with the LSF aberrated by one wave of coma is
shown in Figure 2-71, i.e., for W (r, q) = Ac r3 cos q , where Ac = l , or
F(x ¢, 0) = Ac x ¢ 3 cos q , - 1 £ x ¢ £ 1 , and Ac is in radians. When b = p 2 , the CTF is
unity, since it is equivalent to letting q = p 2 in the aberration function. Hence, the LSF
associated with this orientation is the same as the aberration-free LSF.
1
0.8
0.6
0.4
LSF (x)
0.2
0.2
0.4
6 4 2 0 2 4 6
x
1
Bd = 0
0.8
0.6 1/4
Re LSF (x)
1/2
0.4
(a)
0.2 1
0
0.2
0.4
0 0.5 1 1.5 2 2.5 3 3.5 4
x
0.6
1/2
0.4
m LSF (x)
1/4
0.2
1 (b)
0
Bd = 0
0.2
0.4
0 0.5 1 1.5 2 2.5 3 3.5 4
x
Bd = 0
0.8
1/4
0.6
ΩLS F (x)Ω2
(c)
0.4
1/2
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
x
Figure 2-69. Defocused coherent LSF. (a) Real part. (b) Imaginary part. (c)
Irradiance distribution associated with the LSF. The curve Bd = 0 represents the
aberration-free LSF. x is in units of lF.
2.15 Coherent Imaging 227
0.8
As = 1
0.6 Bd = 0.86
Re LSF (x)
1
0.4
(a)
0.2
0.2
0 0.5 1 1.5 2 2.5 3 3.5 4
x
0.6
0.4 As = 1
1
m LSF (x)
0.2
Bd = 0.86
0
(b)
0.2
0.4
0.6
0.8
0 0.5 1 1.5 2 2.5 3 3.5 4
x
3.5
As = 1
3
Bd = 0.86
2.5 1
ΩLS F (x)Ω2
(c)
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
x
Figure 2-70. Coherent LSF aberrated by one wave of spherical aberration balanced
with defocus (a) Real part. (b) Imaginary part. (c) Irradiance distribution associated
with the LSF. The optimum defocus for the LSF is Bd = - 0.86 As as opposed to
Bd = - As for optimizing a PSF. x is in units of lF.
228 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
1
Ac = 1
p/2
p/4
0.5
b=0
Re LSF (x)
0.5
4 3 2 1 0 1 2 3 4
x
(a)
1
p/2 Ac = 1
p/4
0.8
0.6
LSF2 (x)
b=0
0.4
0.2
0
4 3 2 1 0 1 2 3 4
x
(b)
Figure 2-71. Coherent LSF aberrated by one wave of coma (a) LSF, which is real.
(b) Irradiance distribution associated with the LSF. The line object makes an angle
b with the yo axis and the variation with x shown here is along a direction that is
normal to it. The aberration-free case is obtained when b = p 2.
2.15 Coherent Imaging 229
ESF( x ¢) =
1
-
1 Û
2 2 pi Ù
ı
{[
exp i F(x ¢, 0) - px ¢ x ¢
1
x ¢
dx ¢ . ]} (2-316)
1
For an even aberration, Eq. (2-316) can be written in terms of its real and imaginary parts
as
where
1
1 1Û 1
Re ESF( x ¢) = + Ù cos F(x ¢, 0) sin( px ¢ x ¢) dx ¢ (2-318a)
2 pı x¢
0
and
1
1Û 1
Im ESF( x ¢) = Ù sin F(x ¢, 0) sin( px ¢ x ¢) dx ¢ . (2-318b)
pı x¢
0
We note that
and
Moreover, Re ESF(0) = 0.5 and Im ESF(0) = 0, regardless of the value of the aberration.
For an odd aberration function, the ESF is real given by
1
1 1Û
ESF( x ¢) =
2 pı
[
- Ù sin F(x ¢, 0) - px ¢ x ¢
1
x¢
dx ¢ . ] (2-320)
0
independent of the orientation of the edge object. The ESF is real and obeys Eq. (2-319a).
Equation (2-321) can also be obtained by integrating the LSF from - • to x, as may be
230 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
seen from Eq. (1-228). The aberration-free ESF is shown in Figure 2-72. Its maxima and
minima occur at integral values of x. Its value at x = 0 is 0.5, and its principle maximum
lies at x = 1 with a value of 1.09. It approaches unity as x Æ • and 0 as x Æ - •. The
irradiance distribution associated with the aberration-free ESF is shown in Figure 2-73. It
has a value of 0.25 at x = 0 and a value of 1.19 at the principle maximum at x = 1. It is
evident from the figure that the edge appears to be shifted toward the bright region. There
are strong fluctuations in the irradiance distribution on the bright side of the Gaussian
image of the edge. Such fluctuations are absent in the incoherent ESF shown in Figure 2-
55. Moreover, the spatial period of the fluctuations on the dark side (shadow region) is
half of that on the bright side. It results from the squaring operation of the positive and
negative fluctuations of the amplitude ESF on the dark side.
Figure 2-74 shows the real and imaginary parts and the irradiance distribution
associated with a defocused ESF. As expected for an even aberration, the real part has a
value of 0.5 at x = 0 and the imaginary part is zero, regardless of the value of defocus.
Accordingly, the value of irradiance at the Gaussian image of the edge is 0.25
independent of the defocus value. The irradiance distributions do not bear any symmetry
about x = 0 , unlike the incoherent ESF (compare with Figure 2-55). The ring pattern
shifts more and more to the right as defocus increases. The ESF aberrated by a half wave
of coma is shown in Figure 2-75a. It is real, as expected for an odd aberration. The
corresponding irradiance distribution is shown in Figure 2-75b. We note that the ring
pattern is strongly dependent on the orientation of the edge. When b = p 2, for example,
an aberration-free image is obtained.
0.8
0.6
ESF (x)
0.4
0.2
6 4 2 0 2 4 6
x
Figure 2-72. Aberration-free coherent ESF. x is in units of lF. The dotted curve
represents the Gaussian image of the edge object.
2.15 Coherent Imaging 231
1.2
0.8
ESF2 (x)
0.6
0.4
0.2
6 4 2 0 2 4 6
x
(a)
100
1
10
ESF2 (x)
2
10
3
10
4
10
6 4 2 0 2 4 6
x
(b)
1.2 ,----~-~-~-~-~-~-~-~-___,
0 .8
~O.6
en
w
(a)
'"
a:
! 0.4
0 .2
-1
0.3 ,----,-----,-----.------,------,------,------,-----.--___,
"-
en
w
E
(b)
-0.3
-0~3 - 2 - 1
- x
1.4 ,--,-----,-----.------,------,------,---~-~-___,
1.2
E O's
"-
en
(c)
w
To.6
0.4
0.2
Figure 2-74. Defocused coherent ESF. (a) Real part. (b) Imaginary part. (c)
Irradiance distribution associated with the ESF. The curve Bd = 0 represents the
aberration-free ESF. Bd is in units of A and x is in units of AF.
2.15 Coherent Imaging 233
1.2
p/2
1 p/4
b=0
0.8
Ac = 0.5
0.6
ESF (x)
0.4
0.2
0.2
0.4
3 2 1 0 1 2 3 4 5 6
x
(a)
1.4
Ac = 0.5
1.2 p/2
1
p/2
b=0
ESF2 (x)
0.8
0.6
0.4
0.2
0
3 2 1 0 1 2 3 4 5 6
x
(b)
Figure 2-75. Coherent ESF aberrated by coma Ac r3 cos q, where Ac = 0.5 l . (a) ESF,
which is real. (b) Irradiance distribution associated with the ESF. The edge object
lies along a direction making an angle b with the yo axis and the variation with x
shown here is along a direction that is normal to it. The aberration-free case is
obtained when b = p 2.
234 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
(
= Ag hg vi J1 2pvi hg ) ( ) . (2-322)
r
The spectrum Ũi (vi ) of the diffraction amplitude image of an extended coherent object is
given by Eq. (1-252). Accordingly, the amplitude distribution of the diffraction image is
given by
r r r r r
Ui (ri ) = Ú U˜ i (vi ) exp( - 2pivi ◊ ri ) dvi
r r r r r
= Ú U˜ g (vi ) CTF(vi ) exp( - 2pivi ◊ ri ) dvi . (2-323)
r
Substituting for Ũ g (vi ) from Eq. (2-322) and assuming a radially symmetric aberration
( )
F rp and, therefore, a radially symmetric CTF, Eq. (2-323) reduces to
1 lF 2p
Û Û
Ui (r ) = Ù Ú U˜ g (vi ) CTF(vi )vi dvi Ù exp( - 2 pivi ri cos q) d q
ı ı
0 0
1
Û
( )
= p Ag bg Ù J1 pvbg J 0 ( pvr ) exp[iF(v)] dv ,
ı
(2-324)
0
Ii ( r ) = Ui ( r )
2
. (2-325)
2.15 Coherent Imaging 235
For very small values of bg , e.g., a pinhole, J1 pvbg ( ) ~ pvbg 2 and Eq. (2-324) reduces
to
1
(
Ui ( r ) = p 2
2
) Û
Ag bg2 Ù
ı
J 0 ( pvr ) exp[iF(v)] v dv . (2-327)
0
The integral on the right-hand side is proportional to the zero-order Hankel transform of
the CTF(v) [see, for example, Eq. (2-49b)]. Hence, Eq. (2-327) may be written
If we take the modulus square of both sides, we obtain the incoherent PSF, as expected,
since the distinction between a coherent and an incoherent disc disappears as it gets
smaller. Thus, a pinhole, whether coherently or incoherently illuminated, is a source of
coherent illumination.
[
Figure 2-76a shows the central irradiance of the defocused image F(v) = Bd v 2 of a ]
coherent disc as a function of its radius. The irradiance is normalized by Ag2 . We note
that for small discs bg £ 1.3 , the aberration-free central irradiance is higher than a
corresponding defocused value. However, for larger discs, the aberration-free value is not
necessarily higher than a defocused value. Their relative values depend on the disc radius.
Figure 2-76b shows the axial irradiance of a disc of various sizes. When the disc is very
small, e.g., bg = 1 4, it behaves like a point source. Note that the central value in this case
is normalized to be 4 to accentuate the secondary maxima. For a small disc, the irradiance
is maximum at the center of its image. For a large disc, the maximum occurs at a
defocused point, as evidenced by the curves for bg = 2 and 4.
Figure 2-77a shows the irradiance distribution of the defocused images for bg = 1.
The aberration-free image looks more like the Gaussian image compared to the defocused
images. Figure 2-77b shows the image distributions for bg = 2 . The aberration-free
distribution has a dip in the middle of the image and does not resemble the Gaussian
image. It is also clear that the central value for a defocus of one wave is much larger than
the corresponding aberration-free value, contrary to the case of an incoherent disc
[Compare with Figure 2-63b]. It is a coincidence that, for the disc size under
consideration, the image distribution for a half wave of defocus is similar to that for one
wave of defocus. They are quite different, for example, for bg = 2.3 , as may be seen from
236 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
3.5 1
2.5 1/2
i (0, bg)
2
Bd 0
1.5
1/4
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
bg
(a)
4.5
bg 1/4
3.5
3 2
Bd)
2.5
i(0
1
1.5
4
1
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Bd
(b)
Figure 2-76. (a) Central irradiance of the defocused image of a coherent disc as a
function of its radius. The radius bg of the Gaussian image is in units of lF, and
the defocus coefficient Bd is in units of l . (b) Axial irradiance of a coherent disc of
various sizes. The central value in the case bg = 1 4 is normalized to 4 to accentuate
the secondary maxima. The actual central values in increasing order of bg are 0.02,
1.70, 0.61, and 0.71.
2. 15 Coherent Imaging 237
1 .8 r--~---'--------'-----'----'---~--~------'
(a)
2.5 3.5
, I
1\\.
II.
"\
1
II
(b)
0.5
0
0 3.5
-' r
3.5
bg = 2.3
I
1 I
2.5 I
I
:s 112 '. I
I
(c)
1.5
Figure 2-77. Irradiance distribution of the defocused images of a coherent disc. (a)
The disc radius is bg = 1 in units of 'AF. (b) bg = 2. (c) bg = 2.3. The Gaussian
image is also indicated in each figure.
238 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Figure 2-77c. It is evident that, whereas the image of an incoherent disc resembles the
object, the image of a coherent disc can be quite different from it.
This may all be explained qualitatively as follows. If we divide the disc into radial
zones (similar to the Fresnel zones), the parts of a given zone contribute to the central
amplitude in phase. However, different zones do not contribute in phase (unless the phase
difference between two zones happens to be 2p ). The complex amplitudes contributed
by different zones can partially cancel each other. This cancellation may be reduced by
the defocus aberration, resulting in a larger defocused central irradiance. The relative
phase of a zone does not matter in the case of an incoherent disc because it is the
irradiance contributions from different zones that are added to obtain the net irradiance.
Hence, the central irradiance increases as the radius of an incoherent disc increases.
r -i È Ê r2 ˆ ˘ Û r È ikr 2 Ê 1 1 ˆ ˘ Ê 2 pi r r ˆ r
U (r ; z) = exp Íik Á z + ˜ ˙ Ù Uo ( rl ) exp Í l Á - ˜ ˙ exp Á - r ◊ rl ˜ drl .
lz ÍÎ Ë 2 z ¯ ˙˚ ı Î 2 Ë z f ¯ ˚ Ë lz ¯
(2-329)
If we let z = f , we obtain
È Ê rf2 ˆ ˘ Û
(r )
U rf ; f =
-i
lf
exp Íik Á f +
ÍÎ Ë
˜
2 f ¯˙ ı
r Ê 2 pi r r ˆ r
˙ Ù Uo ( rl ) exp Á -
Ë lf
rf ◊ rl ˜ drl
¯
, (2-330)
˚
2.15 Coherent Imaging 239
r
where rf is the position vector of a point in the focal plane of the lens. We see that except
for a phase factor outside the integral, the amplitude in the focal plane of the lens is the
Fourier transform of the object.
r -i È Ê r2 ˆ ˘ Û r È ikr 2 1 1 ˘ Ê 2 pi r r ˆ r
U (r ; z) = exp Íik Á z + ˜ ˙ Ù Uo ( ro ) exp Í o Ê - ˆ ˙ exp Á - r ◊ ro ˜ dro .
lz ÍÎ Ë 2 z ¯ ˙˚ ı Î 2 Ë z d ¯ ˚ Ë lz ¯
(2-331)
If we let z = d , we obtain
r -i È Ê r2 ˆ ˘ Û r Ê 2 pi r r ˆ r
U ( rd ; d ) = exp Íik Á d + d ˜ ˙ Ù Uo ( ro ) exp Á - rd ◊ ro ˜ dro , (2-332)
ld ÍÎ Ë 2 d ¯ ˙˚ ı Ë ld ¯
r
where rd is the position vector of a point in the observation plane at a distance d from the
object plane. Again, except for a phase factor outside the integral, the amplitude in the
plane of convergence of the spherical wave is given by the Fourier transform of the
object.
We can also obtain the Fourier-transform relationship from the imaging equation by
determining the amplitude in a defocused plane. As illustrated in Figure 2-78, consider an
object illuminated by a plane wave and imaged by a thin lens. The Fourier transform of
the object is obtained in the focal plane of the lens, while its image is obtained in the
image plane determined by Gaussian optics. If the object is illuminated by a spherical
wave, the Fourier transform is obtained in a plane where the focus of the spherical wave
is imaged by the lens. We also illustrate how the image can be altered by spatial filtering
certain frequency components in the Fourier-transform plane.
From Eq. (1-236), the complex amplitude in the focal plane of a lens of focal length f
is given by
r È Ê rf2 ˆ ˘ Û r r Ê r2 ˆ
(
U f rf ; f = - )i
lf
Í
exp ik Á f - zg +
ÍÎ Ë
˜ ˙ Ù d ro Uo ( ro ) exp Á - ik o ˜
2 f ¯˙ ı Ë 2 zo ¯
˚
r r
Û r r Ê ik 2 ˆ È 2 pi r Ê rf M ro ˆ ˘ r
Ù
( )
¥ Ù G rp ; ro exp Á - rp ˜ exp Í-
Ë 2 zo ¯ ÍÎ l
rp Á -
Ë f
◊ ˙ d rp
zg ˜¯ ˙
, (2-333)
ı ˚
where
1 1 1
- = (2-334)
z g zo f
240 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
(–)zo zi
zg
M = . (2-335)
zo
Note that according to our sign convention (see Section 1.3.2 of Part I), zo is numerically
negative. If the system is aberration free and the lens is quite large, we may replace
r r
( )
G rp ; ro by a constant that is inversely proportional to the distance of the object zo and
extend the region of integration in Eq. (2-333) to infinity. Therefore, the integral in this
equation becomes the (inverse) Fourier transform of an imaginary Gaussian function
r r
( )
evaluated at the spatial frequency rf f 1 - ro zo 1 l . Hence, within a constant, Eq. (2-
333) reduces to
r È ik Ê zo ˆ r f ˘ Û
2
r Ê ik r r ˆ r
(
U f rf ; f ) =
1
f
exp Í Á + ˜ ˙ Ù Uo ( ro ) exp Á - rf ◊ ro ˜ d ro
ÍÎ 2 Ë
1
f ¯ f ˙˚ ı Ë f ¯
, (2-336)
where we have dropped the unessential phase factors. Thus, the complex amplitude in the
back focal plane of the lens is proportional to a Fourier transform of the object. The phase
factor varying as rf2 vanishes when the object is placed in the front focal plane of the lens
(so that zo = - f ). The image is formed at infinity in this case, and a second lens must be
used to observe the image in its focal plane. The finite size of the pupil causes some
vignetting, which can be avoided by placing the object directly against the lens. (Based
on the angular spectrum concept, vignetting is equivalent to the lens missing certain high
spatial frequencies in the object.) One must contend with the quadratic phase factor in
that case, as may be seen by letting zo Æ 0 in Eq. (2-336) (unless the transform is
received on a spherical surface). A measurement of the irradiance distribution yields the
power spectrum of the object in either case. It should be noted that in order to obtain
high-quality Fourier transform and image of an object, the lens must be aberration free
for forming the image of the source illuminating the object (e.g., aberration-free focusing
2.15 Coherent Imaging 241
of the plane wave) and for forming the image of the object; i.e., it must be aberration free
for two conjugates.
To demonstrate how the image of an object can be altered by spatial filtering its
Fourier transform, we consider a wire mesh as an object illuminated by a plane wave, as
illustrated in Figure 2-78, and first explained by Porter59 nearly a century ago. A
diffraction pattern or a Fourier transform of the wire mesh is seen in the focal plane of the
lens, as illustrated in Figure 2-79a. Each spot in the focal plane is actually surrounded by
the diffraction rings characteristic of the mesh aperture. As expected, the image of the
wire mesh in the image plane consists of both vertical and horizontal lines. Figures 2-79b
through 2-79e show how the image changes if a mask is used in the focal plane to
spatially filter out certain parts of the Fourier transform. If the mask transmits only the
central horizontal spots (which are generated by the vertical wires of the mesh), as in
Figure 2-79b, then the image consists of only vertical lines. Similarly, if only the central
vertical spots are transmitted, as in Figure 2-79c, then the image consists of only the
horizontal lines. If the central spot is also blocked, then the image undergoes a contrast
reversal; the wire images appear as bright horizontal or vertical lines with wide dark
spaces between them. If the central spot and only the second spot on each side of it are
transmitted, then the spacing of the image lines is reduced by a factor of two. If only the
spots on a central diagonal are transmitted, as in Figure 2-79d, then the image consists of
diagonal lines that are absent in the object. Similarly, if the orthogonal central diagonal
spots are transmitted, as in Figure 2-79e, the image consists of orthogonal diagonal lines,
which are also absent in the object. If both sets of diagonal spots are transmitted, as in
Figure 2-79f, the image appears as if the object has been rotated by 45˚.
Lenses used to form and observe Fourier transforms of objects are called Fourier
lenses, and the process of selectively masking portions of a Fourier transform is called
spatial filtering. Indeed, this field has come to be known as Fourier optics.
Now we briefly compare coherent and incoherent imaging. We have already seen
how different the images of a line, an edge, and a disc are under coherent and incoherent
illuminations (compare Figure 2-52 with 2-69; 2-55 with 2-73; and 2-63 with 2-77). Here
we compare the frequency spectra and two-point resolution of the coherent and
incoherent images of an object. We illustrate that we can not a priori assert that one type
of image is always better than the other.
(a)
(b)
(c)
(d)
(e)
(f)
Figure 2-79. Spatial filtering of the spectrum of an object consisting of a wire mesh
and the corresponding changes in its image. A mask selectively transmits a certain
portion of the object spectrum and filters the rest. (a) No mask. (b) Central
horizontal spots transmitted, giving an image consisting of vertical lines. (c) Central
vertical spots transmitted, giving an image consisting of horizontal lines. (d)
Diagonal spots transmitted, giving an image consisting of diagonal lines absent in
the object. (e) The orthogonal diagonal spots, again giving an image with orthogonal
lines that are absent in the object. (f) Both central diagonal spots transmitted, giving
an image as if the object has been rotated by 45o .
2.15 Coherent Imaging 243
variations across the pupil. Thus, the bandwidth of frequencies transmitted by a system is
twice as wide for an incoherently illuminated object as for the one illuminated coherently.
However, an examination of the spectra of the observed irradiance images of coherently
and incoherently illuminated objects shows that it is not a simple matter to decide which
image is better.
r r r
Ii ( ri ) = U g ( ri ) ƒ CSF( ri )
2
. (2-338)
r r
= U g ( ri ) ƒ CSF( ri )
2 2
. (2-340)
Comparing Eqs. (2-339) and (2-341), it is evident that the spectra of the coherent and
incoherent irradiance images are different from each other. However, we can not
conclude that one image is better than the other.
Object
Incoherent Image
Coherent Image
Figure 2-80. Incoherent and coherent images of an extended object. The object is the
Taj Mahal built by the Mughal Emperor Shah Jahan. For clarity, a magnified
image of its boxed portion only is shown. The edge ringing associated with coherent
imaging is quite evident in the coherent image.
2.15 Coherent Imaging 245
I ( x) = Ì 1 ˝ +Ì ˝ , (2-342)
ÓÔ p( x - 0.61) ˛Ô ÔÓ p( x + 0.61) ˛Ô
For two coherently illuminated point objects separated by the Rayleigh resolution,
the corresponding irradiance distribution is given by
I( x) = + exp(id ) , (2-343)
p( x - 0.61) p( x + 0.61)
where d is the phase difference between the two object points. This distribution is also
symmetrical about x = 0 . It is shown in Figure 2-82 for three different values of d . We
note that, since J1 (1.22 p) = 0 , I ( ± 0.61) = 1, regardless of the value of d . There is no dip
in the irradiance when the point objects are in phase; hence, they can not be resolved.
However, when their phases are opposite of each other, then the irradiance at the center is
zero, making it easier to resolve them. When d = p 2 , then the irradiance distribution is
the same as for two incoherent point objects. Thus, the ability to resolve points of a
coherently illuminated object is strongly dependent on the phase relation among them.
Now we consider the effect of defocus on the images and show that it is possible to
obtain an asymmetric image of the two coherent point objects.62,63 From Eq. (2-308), the
complex amplitude distribution of the defocused image of a point object can be written
1
Û
[ ]
A(r ) = 2 Ù exp iF(r) J 0 ( p r r) r dr ,
ı
(2-344)
0
where F(r) = Bd r2 is the defocus phase aberration, Bd being its peak value in radians.
Hence, the irradiance distribution of the defocused image of two incoherent point objects
246 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
0.9
0.8
0.7
0.6
(x)
0.5
0.4
0.3
0.2
0.1
0
4 3 2 1 0 1 2 3 4
x
(a)
(b)
Figure 2-81. (a) Irradiance distribution along the x axis of the image of two
incoherent point objects separated by the Rayleigh resolution of 1.22ll F . The
central value is 0.73. (b) 2D image.
2.15 Coherent Imaging 247
1.5
d=0
p/2
(x)
0.5
0
4 3 2 1 0 1 2 3 4
x
Figure 2-82. Irradiance distribution of the image of two coherent point objects
separated by the Rayleigh criterion of resolution 1.22ll F with a relative phase of d.
When d = p 2 , the distribution is the same as for two incoherent point objects. The
central values are 1.47, 0.74 and 0. The principal peaks have a value of 1.11 at x =
0.78 when d = p .
È 1 2
ÍÛ
[ ]
I ( x ) = 4 Í Ù exp iF(r) J 0 [ p( x - 0.61)r] r dr
Í ı0
Î
2˘
1
Û ˙
[ ]
+ Ù exp iF(r) J 0 [ p( x + 0.61)r] r dr ˙ .
ı
(2-345)
˙
0 ˚
The distribution is symmetric about x = 0 , and it is shown in Figure 2-83 for defocus
wave aberrations of Bd = 0 , l 4 , and l 2. The peak value is given by I ( ± 0.61) = 1,
0.84, and 0.48 for these values of Bd . The irradiance in the central region decreases and
increases in the surrounding region. Figure 2-83b shows that when normalized to the
same peak value, the central dip lowers slightly with increasing value of Bd .
1.0
Bd = 0
0.8
1/4
0.6
(x)
0.4
1/2
0.2
0
4 3 2 1 0 1 2 3 4
x
(a)
0.8
0
1/4
1/2
0.6
(x)
0.4
0.2
Bd = 1/2
1/4
0
0
4 3 2 1 0 1 2 3 4
x
(b)
Figure 2-83. (a) Irradiance distribution of the defocused image of two incoherent
point objects separated by the Rayleigh criterion of resolution 1.22ll F . The central
values in increasing order of defocus are 0.73, 0.61, and 0.33. The principal peaks
have a value of 1, 0.84, and 0.48 at x = ± 0.61. Bd is in units of l . (b) The
distributions are normalized to unity at the principal peaks.
2.15 Coherent Imaging 249
È 1
Û
I ( x ) = 4 ÍÍ Ù exp iF(r) J 0 [ p( x - 0.61)r] r dr
[ ]
ı
ÍÎ 0
2˘
1
Û ˙
[ ]
+ exp(id ) Ù exp iF(r) J 0 [ p( x + 0.61)r] r dr ˙ .
ı
(2-346)
˙
0 ˚
This distribution is shown in Figure 2-84. It is symmetric for d = 0 and p . The central
value decreases with defocus in Figure 2-84a, and the side peaks decrease in value in
Figure 2-84b (the central value is zero in this case). However, the defocused distribution
is quite asymmetric when d = p 2 . Although the presence of two point objects can be
inferred from this figure, one would incorrectly infer that they are of unequal intensity. If
the defocus aberration is negative, as when z > R , then the object appearing dimmer in
Figure 2-84c will lie on the right-hand side of the origin.
The 2D defocused images of two incoherent and coherent point objects separated by
the Rayleigh resolution are shown in Figures 2-85 and 2-86. An example of an incoherent
two-point object is a double star observed by a telescope. In principle, one should be able
to resolve the image of two points that are closer to each other than the Rayleigh
resolution. However, in practice, the aberrations of the imaging system and the noise in
the image recording device (e.g., a film or an array of photodetectors) will limit the
resolution of a system whether the objects are coherent or not. 64
1.5
d=0
Bd = 0
1/4
1
(x)
(a)
1/2
0.5
0
4 3 2 1 0 1 2 3 4
x
1.4
d=p
1.2
Bd = 0
1
0.8 1/4
(x)
(b)
0.6
1/2
0.4
0.2
0
4 3 2 1 0 1 2 3 4
x
1.4
d = p/2
1.2
0.8 Bd = 0
(x)
(c)
0.6
1/4
0.4
0.2
1/2
0
4 3 2 1 0 1 2 3 4
x
Bd = 0 Bd = 0
Bd = 1 4 Bd = 1 4
Bd = 1 2 Bd = 1 2
(a) Incoherent points (b) Coherent points in phase
Figure 2-85. Aberration-free and defocused images of two object points separated by
the Rayleigh resolution. (a) Incoherent points. (b) Coherent points in phase ( d = 0 ).
252 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
Bd = 0 Bd = 0
Bd = 1 4 Bd = 1 4
Bd = 1 2 Bd = 1 2
(a) (b)
Figure 2-86. Aberration-free and defocused images of two coherent point objects
with a phase difference of (a) d = p . (b) d = p 2.
2.16 Summary 253
2.16 SUMMARY
Some of the key results of this chapter are summarized below:
This chapter starts with a description of the Airy pattern, i.e., the aberration-free
image of a point object formed by a system with a circular pupil, showing that 83.8% of
the total light in the image is contained in the Airy disc of radius 1.22 in units of l F .
Aberrations of the system reduce the central irradiance and, for small aberrations, its
value relative to the aberration-free value, called the Strehl ratio, is given approximately
( )
by exp - s F2 [see Eq. (2-53c)], where s F2 is the variance of the phase aberration across
the system pupil. This expression estimates the Strehl ratio for primary aberrations with
an error £ 10% provided it is ≥ 0.3. The Strehl ratio determines the aberration tolerance
for high-quality imaging systems. For example, a Strehl ratio of 0.8 is obtained for a
sigma value of l 14 , regardless of the nature of the aberration. It is different from the
Rayleigh’s quarter wave rule, which gives different values of the Strehl ratio for a quarter
wave of different aberrations.
Since an aberration with a smaller variance yields a higher Strehl ratio, we balance
an aberration of a certain order with one or more aberrations of lower orders to reduce its
variance. As shown in Section 2.4, the balanced classical aberrations are represented by
the Zernike circle polynomials. Thus, these polynomials are not only orthogonal across a
circular pupil, but they also represent balanced classical aberrations. Of course, for a large
classical aberration, a balanced aberration may not yield the highest possible Strehl ratio.
For spherical aberration Asr 4 , the Strehl ratio is maximum at the point of minimum
variance, i.e., corresponding to a defocus aberration of -Asr 2 for As < ~ 2.3 l . Similarly,
in the case of coma Ac r cos q , maximum Strehl ratio is obtained at the point of
3
minimum variance (4 3) Ac F for Ac < ~ 0.7l . Astigmatism Asr cos q yields maximum
2 2
Although aberrations always reduce the central irradiance, the peak value of the axial
irradiance of a focused beam for small Fresnel numbers does not lie at the geometrical
focus of the beam, but at a point closer to the focusing aperture. This is due to an increase
in the central irradiance due to the inverse square-law effect being larger than the
decrease due to the defocus aberration. For example, as discussed in Section 2.5.2, the
axial irradiance of a beam with a Fresnel number of unity focused at a distance R peaks at
0.6R. The inverse-square law increases the irradiance at this point by a factor of 2.78, but
the corresponding defocus reduces it by a factor of 0.68, with the net result that peak
irradiance is 1.9 times the focal-point irradiance. However, as illustrated in Figure 2-13,
this does not mean that optimum focusing of a beam on a target at a certain distance is
254 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
anything other than focusing on the target. Since the target distance is fixed, focusing the
beam at any other point along its axis will only decrease the central value due to the
defocus aberration. The depth of focus for a certain Strehl ratio is quite large for small
Fresnel numbers (as in the case of a laser transmitter) but quite small for large Fresnel
numbers (as for a photographic camera).
For systems with large Fresnel numbers, as in most imaging applications, the near-
and far-field regions can be defined quantitatively. As discussed in Section 2.5.3, a
collimated beam compared to a beam focused at a distance of D 2 l represents a
negligible defocus aberration of only l 8 . Accordingly, the near-field or the Fresnel
region is represented by distances shorter than D 2 l , and the far-field or the Fraunhofer
region corresponds to distances equal to or larger than D 2 l . It should be evident,
though, that Fresnel diffraction represents the defocused Fraunhofer diffraction.
For large Fresnel numbers, the PSF of an aberration-free system is symmetric about
the Gaussian image plane, as discussed in Section 2.7. The PSF in the Gaussian image
plane aberrated by a radially symmetric aberration is independent of its sign. Hence, the
sign of the aberration cannot be determined from it. While the PSF is not symmetric
about the Gaussian image plane, the axial irradiance is symmetric when the aberration is
represented by a Zernike polynomial Z n0 (r) . Thus, for example, the axial irradiance for
Seidel spherical aberration is symmetric about the defocused image plane that yields
minimum aberration variance (see Figure 2-25). For astigmatism r 2 cos 2 q , the axial
irradiance is also symmetric about the defocused image plane that yields minimum
variance. Moreover, the sign of the aberration does not affect the PSF in this defocused
plane. The PSFs in two image planes located symmetrically about the Gaussian image
plane are the same except for a rotation by p 2 if the sign of the aberration for one is
opposite to that for the other. The PSF for coma r 3 cos q is symmetric about the
Gaussian image plane. Moreover, it rotates by p in any image plane when the sign of the
aberration is changed.
2.16 Summary 255
(
The line of sight of an aberrated system defined in terms of the centroid x i , y i )
of its PSF is given by Eq. (2-151). Aberrations varying as cosq contribute to x i and
those varying as sinq contribute to y i . The primary and secondary coma varying as
r 3 cos q and r 5 cos q , respectively, give different PSFs but the same line of sight for the
same value of their coefficients. For a given value of the standard deviation, an aberration
of a higher order gives a larger LOS error, and two Zernike aberrations with different
values of n having the same standard deviation give different LOS errors, even though
they give (approximately) the same Strehl ratio. Hence, in tolerancing an optical system,
one should be careful in allocating equal standard deviation to two aberration terms that
also contribute to the LOS error.
viii. OTF
The aberration-free OTF, representing the fractional area of overlap of two circular
pupils separated by a distance that is proportional to the spatial frequency, decreases
monotonically from a value of unity at the origin to zero at the cutoff frequency 1 l F .
The slope of the OTF at the origin is - 4 p , regardless of any aberration. OTF for a
symmetric aberration is real, e.g., for defocus, astigmatism, and spherical aberration. Its
real part may be positive or negative in certain frequency bands. A negative value
represents contrast reversal, i.e., the bright regions of an object at those frequencies
appear dark in the image, and dark regions appear as bright. Figure 2-44 illustrates this
for defocus. The OTF is positive for a defocus aberration £ 0.64l and for spherical
aberration £ 2.2l . Its value for astigmatism depends on the orientation of the spatial
frequency. The OTF for coma is complex (except for spatial frequencies along the y
axis), and is expressed in terms of its real and imaginary parts (as in Figures 2-48 and 2-
49), or its MTF and PTF (as in Figures 2-50 and 2-51). The Hopkins ratio for a certain
spatial frequency, representing the ratio of its MTFs with and without aberration, is
( )
approximately given by exp - s Q2 2 , where s Q2 is the variance of the aberration
difference function for that frequency [see Eq. (2-203)]. The tolerances for primary
aberrations for a Hopkins ratio of 0.8 and spatial frequencies less than a tenth of the
cutoff frequency are given in Section 2.11.1. The geometrical MTF for a radially
symmetric aberration depends on the spot sigma, and is maximum when the spot sigma is
minimum.
The LSF (line-spread function) of an imaging system is the Abel transform of its PSF
[see Eq. (2-243)], and is related to the first-order Struve function, as in Eq. (2-252b). Its
central value normalized to unity in the absence of aberrations is called the Struve ratio.
The balanced aberrations and their tolerance for a Struve ratio of 0.8 are given Table 2-
17. Except for an additive constant, the ESF (edge-spread function) represents the
cumulative LSF [see Eq. (2-261b)].
256 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
As illustrated in Figure 2-63, the defocused image of a disc is generally bright in the
central region and dim in the outer region. As defocus increases, the irradiance decreases
in the central region and increases in the outer region. In the case of a coherently
illuminated disc, the central irradiance of the image can be much lower than that in the
surrounding region, whether or not the image is defocused. Moreover, defocus can
increase the central irradiance, as in Figure 2-77.
A pinhole camera consists of a box with a pinhole on one side and film on the other.
A relationship between the pinhole radius a and the box length L can be obtained in
several different ways. Petzval minimized the image spot radius resulting from the
geometrical and diffraction contributions, and obtained the relationship a P = (L l 2) .
12
The same result is obtained by letting the defocus aberration due to lack of focusing by a
lens to be a quarter wave. By maximizing the central irradiance of the diffraction pattern
of the pinhole in the image plane, or by letting the Hopkins ratio due to defocus
aberration be 0.8, the relationship obtained is a = (L l) , which is in close agreement
12
According to the Rayleigh criterion, two point objects of equal intensity are just
resolved when the principal maximum of the Airy pattern of one falls on the first
minimum of the other, i.e., if their angular separation is 1.22 l D, or the separation
between their Gaussian images is 1.22l F . For two incoherent point objects, the image
irradiance distribution has a dip at the center with a value of 0.73 compared to a peak on
its either side with a value of unity. The distribution for two coherent point objects and
the ability to resolve them depends on their relative phase. It is the same as for incoherent
point objects when their phase difference is p 2. When the phase difference is p, the
central irradiance is zero and the two points are resolved easily. There is no dip at the
center of the image when they are in phase and, therefore, they cannot be resolved.
A defocus aberration reduces the central irradiance of the image of two incoherent
point objects and increases it in the surrounding region. Figure 2-83b shows that, when
normalized to the same peak value, the central dip lowers slightly with increasing value
of defocus. The distribution for two coherent point objects out of phase by p 2 is quite
asymmetric, as in Figures 2-84b and 2-86b, yielding a wrong inference that they are of
unequal intensity.
258 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
REFERENCES
1. For properties of Bessel functions, the reader may refer to G. N. Watson, Treatise on
the Theory of Bessel Functions (Cambridge U. P., New York, 1944).
3. Lord Rayleigh, Phil. Mag. (5)11, 214 (1881); also his Scientific Papers (Dover, New
York, 1964), Vol. 1, p. 513.
4. V. N. Mahajan, “Included power for obscured circular pupils,” Appl. Opt. 17, 964–
968 (1978).
5. V. N. Mahajan, “Strehl ratio for primary aberrations: some analytical results for
circular and annular pupils,” J. Opt. Soc. Am. 72, 1258–1266 (1982), Errata, 10, 2092
(1993).
7. Lord Rayleigh, Phil. Mag. (5) 8, 403 (1879); also in his Scientific Papers (Dover,
New York, 1964) Vol. 1, p. 432.
8. W. B. King, “Dependence of the Strehl ratio on the magnitude of the variance of the
wave aberration,” J. Opt. Soc. Am. 58, 655–661 (1968).
10. V. N. Mahajan, “Zernike circle polynomials and optical aberrations of systems with
circular pupils,” Appl. Opt. 33, 8121–8124 (1994).
11. V. N. Mahajan, “Zernike polynomials and optical aberrations,” Appl. Opt. 34, 8060–
8062 (1995).
12. A. B. Bhatia and E. Wolf, “On the circle polynomials of Zernike and related
orthogonal sets,” Proc. Cambridge Philos. Soc. 50, 40–48 (1954).
13. V. N. Mahajan, “Axial irradiance and optimum focusing of laser beams,” Appl. Opt.
22, 3042–3053 (1983).
14. Y. Li and E. Wolf, “Focal shifts in diffracted converging spherical waves,” Optics
Commun. 39, 211–215 (1981).
15. D. S. Burch, “Fresnel diffraction by a circular aperture,” Am. J. Phys. 53, 255–260
(1985).
21. V. N. Mahajan, “Line of sight of an aberrated optical system,” J. Opt. Soc. Am. A 2,
833–846 (1985).
22. H. H. Hopkins, “The frequency response of a defocused optical system,” Proc. Roy.
Soc. A 231, 91–103 (1955).
23. W. H. Steel, “The defocused image of sinusoidal gratings,” Optica Acta 3, 65–74
(1956).
24. M. De, “The influence of astigmatism on the response function of an optical system,”
Proc. Roy. Soc. A 233, 91–104 (1955).
27. R. Barakat, “Computation of the transfer function of an optical system from the
design data for rotationally symmetric aberrations. Part I. Theory,” J. Opt. Soc. Am.
52, 985–991 (1962).
28. R. Barakat, “Part II. Programming and numerical methods,” J. Opt. Soc. Am. 52,
992–997 (1962).
29. R. Barakat and A. Houston, “Transfer function of an optical system in the presence
of off-axis aberrations,” J. Opt. Soc. Am. 55, 1142–1148 (1965).
30. H. H. Hopkins, “The aberration permissible in optical systems,” Proc. Phys. Soc.
(London) B 52, 449–470 (1957).
31. W. B. King, “Correlation between the relative modulation function and the
magnitude of the wave aberration difference function,” J. Opt. Soc. Am. 59, 692–697
(1969).
32. S. Szapiel, “Hopkins variance formula extended to low relative modulations,” Optica
Acta 33, 981–999 (1986).
260 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
34. W. Lukosz, “Der Einfluss der Aberrationen auf die optische Uebertragungsfunktion
bei kleinen Orts-Frequenzen,” Optica Acta 10, 1–19 (1963).
35. R. W. Gostick, “OTF-based optimization criteria for automatic optical design,” Opt.
Quant. Elect. 8, 31–37 (1976).
36. J. Braat, “Polynomial expansion of severely aberrated wave fronts,” J. Opt. Soc. Am.
A 4, 643–650 (1987).
37. H. H. Hopkins and B. Zalar, “Aberration tolerances based on the line spread
function,” J. Mod. Opt. 34, 371–406 (1987).
38 R. Barakat and A. Houston, “Line spread function and cumulative line spread
function for systems with rotationally symmetry," J. Opt. Soc. Am. 54, 768–773
(1964).
39. R. Barakat and A. Houston, “Line spread and edge spread functions in presence of
off-axis aberrations,” J. Opt. Soc. Am. 55, 1132–1135 (1965).
40. R. Bracewell, The Fourier Transform and Its Applications (McGraw-Hill, New
York, 1965), p. 262.
42. H. F. Willis, “A formula for expanding an integral as series,” Philos. Mag. 39, 455–
459 (1948). There is a minus sign missing on the right-hand side of the formula
corresponding to Eq. (2-255).
43. G. A. Korn and T. M. Korn, Mathematical Handbook for Scientists and Engineers
(McGraw-Hill, New York, 1968), p. 878.
44. W. S. Kovach, “Energy distribution in the PSF for an arbitrary passband,” Appl. Opt.
13, 1769–1771 (1974).
45. H. S. Dhadwal and J. Hantgan, “Generalized point spread function for a diffraction-
limited aberration-free imaging system under polychromatic illumination,” Opt. Eng.
28, 1237–1240 (1989).
47. W. Weinstein, “Images of incoherently illuminated bright and opaque disks,” J. Opt.
Soc. Am. 45, 1006–1008 (1955).
48. R. Barakat and A. Houston, “Image of an incoherently illuminated disk,” J. Opt. Soc.
Am. 55, 881–883 (1965).
References 261
51. J. Petzval, “Bericht über dioptrische Untersuchungen,” Wien Ber. XXVI, 33–90
(1857).
52. Lord Rayleigh, “On pin-hole photography,” Phil. Mag. XXXI, 87-99 (1891); also in
his Scientific Papers Vol I, 429–440 (Dover, New York, 1965).
54. X. Jiang, Q. Lin, and S. Wang, “Optimum image plane of the pinhole camera,”
Optik, 97, 41–42 (1994).
55. K. Sayanagi, “Pinhole imagery,” J. Opt. Soc. Am. 57, 1091–1099 (1967).
56. G. Reynolds and J. H. Ward, “Coherence theory solution to the pinhole camera,” J.
Soc. Photo. Instr. Eng. 5, 3–8 (1966).
57. R. E. Swing and D. P. Rooney, “General transfer function for the pinhole camera,” J.
Opt. Soc. Am. 58, 629–635 (1968).
59. A. B. Porter, “On the diffraction theory of microscopic vision ,” Phil. Mag. 11, 154–
166 (1906).
60. Lord Rayleigh, “Investigations in optics, with special reference to the microscope:
Resolving, or separating, power of optical instruments,” Phil. Mag. VIII, 261-264
(1879); also in his Scientific Papers Vol I, 415–418 (Dover, New York, 1965).
61. Lord Rayleigh, “On the resolving-power of telescopes I,” Phil. Mag. X, 116-119
(1880); also in his Scientific Papers Vol I, 488–490 (Dover, New York, 1965).
62. T. S. McKechnie, “The effect of defocus on the resolution of two points,” Optica
Acta 20, 253–262 (1973).
63. D. K. Cook and G. D. Mountain, “The effect of phase angle on the resolution of two
coherently illuminated points,” Optical and Quan. Elec. 10, 179–180 (1978).
64. A. J. den Dekker and A. van den Bos, “Resolution: a survey,” J. Opt. Soc. Am. 14,
547–557 (1997).
262 OPTICAL SYSTEMS WITH CIRCULAR PUPILS
PROBLEMS
3. (a) Calculate the mean and mean square values of spherical, coma, and astigmatism
aberrations. Show that their standard deviations are given as in Table 2-4. (b)
Similarly, show that the corresponding balanced aberrations are given as in Table
2-5. (c) Consider balancing defocus aberration with spherical aberration. Determine
the relative amount of the balancing spherical aberration and the standard deviation
of the balanced aberration.
5. (a) Determine the depth of focus of a photographic camera with a lens of diameter 2
cm and an f-number of 5 forming an image at a wavelength of 0.5 mm with a Strehl
ratio of at least 0.8. (b) Now consider a laser transmitter focusing a beam of
diameter 25 cm and a wavelength of 10.6 mm on a receiver at a distance of 1.47
km. Determine the depth of focus for a Strehl ratio of 0.8. Also, determine the point
and the value of maximum irradiance for a transmitted power of 1 kW. Is the
receiver in the near or the far field of the transmitter?
6. (a) Compare the axial irradiance of a beam focused at a distance of D2 l with that
of a corresponding collimated beam. Determine the location and the value of the
principal maximum in the case of the focused beam. (b) Show that the PSF
aberrated by a certain amount of coma rotates by p when the sign of its aberration
coefficient is changed.
8. (a) Compare the Strehl ratios, peak irradiances and their locations, and centroid
irradiances and locations of the images of a point object formed by a system with a
half wave of primary coma with those for a half wave of secondary coma. (b)
Estimate the Strehl ratio if both aberrations are present simultaneously. Calculate
the position of the centroid also.
9. Show that a power series expansion of the OTF aberrated by astigmatism in the
presence of defocus is given by
t (v, f) = 1 -
4v
p [ ( ) ] ( )
- 2 Bd2 + Aa2 + 2 Aa Bd cos 2 f v 2 + O v 3 .
Determine the corresponding Hopkins ratio for small values of v in the Gaussian,
sagittal, and tangential image planes. It should be evident that the slope of the OTF
at the origin is independent of the defocus aberration coefficient Bd .
10. Using Eqs. (1-155), show that the coma-aberrated geometrical OTF for small
spatial frequencies is given by Eqs. (2-240).
12. Consider two point objects of equal intensity, phase angle d , and degree of
coherence g located symmetrically about the axis of an imaging system. Obtain
the conditions under which the irradiance distribution of a defocused image along
the line joining their Gaussian image points is symmetric about the optical axis
(see Reference 62).
CHAPTER 3
265
Chapter 3
Optical Systems with Annular Pupils
3.1 INTRODUCTION
In this chapter, we discuss the imaging properties of a system with an annular pupil
in a manner similar to those for a system with a circular pupil. The two-mirror
astronomical telescopes discussed in Chapter 6 of Part I are a typical example of an
imaging system with an annular pupil. The linear obscuration ratios of some of the well-
known telescopes are 0.36 for the 200-inch telescope at Mount Palomar, 0.37 for the 84-
inch telescope at the Kitt-Peak observatory, 0.5 for the telescope at the McDonald
Observatory, and 0.33 for the Hubble Space Telescope.
Expressions for the PSF, OTF, and encircled, ensquared, and excluded powers are
given. The Strehl ratio of an aberrated system is considered and tolerances for primary
aberrations are discussed. Symmetry properties of aberrated PSFs are discussed, and
pictures of the PSFs for primary aberrations are given as examples. The line of sight of an
aberrated system is discussed in terms of the centroid of its PSF. Numerical results are
given and compared with the corresponding results for systems with circular pupils
wherever possible and appropriate.
r Ê 2 pi r r ˆ r
2
[(
Ii ( ri; ) = Ii (0; ) Sex
2
)]
( ) Û
ı Ë R
◊
Ù exp Á- l rp ri ˜ d rp
¯
, (3-1)
where
( )
Sex ( ) = p 1 - 2 a 2 (3-2)
267
268 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Defocused
image plane
ExP
Gaussian
a image plane
a
Figure 3-1. Imaging by a system with an annular exit pupil of inner and outer radii
a and a, respectively.
is the clear area of the obscured exit pupil. The quantity 2 in Eq. (3-2) is sometimes
referred to as the area obscuration ratio. The optical wavefront at the exit pupil is
spherical with a radius of curvature R and center of curvature at the Gaussian image
point. The central irradiance is given by
(
= pPex 1 - 2 ) 4l2 F 2 . (3-3b)
where Pex is the total power in the exit pupil, and, therefore, in the image. For an object
of intensity Bo radiating at a wavelength l at a distance z0 from the entrance pupil of
area Sen ( ) , the total power is given by
[ ]
Pex = h Sen ( ) z02 Bo , (3-4)
where h is the transmission factor of the system for light propagation from its entrance to
its exit pupil. The quantity F in Eq. (3-3b) is given by
F = R D , (3-5)
where D = 2 a is the outer diameter of the exit pupil. It represents the focal ratio
(f-number) of the image-forming light cone exiting from the exit pupil. The integration in
r
Eq. (3-1) is carried over the clear area of the exit pupil such that the position vector rp of
r
a point in its plane satisfies a £ rp £ a .
As in Section 2.2.1, we express the position vectors of points in the pupil and image
planes in polar coordinates according to
r
( )
rp = rp cos q p , sin q p , a £ rp £ a , 0 £ q p < 2 p , (3-6)
3.2 Aberration Free System 269
and
r
ri = ri (cos q i , sin q i ) , 0 £ q i < 2 p . (3-7)
2
a 2p
È 2ip ˘
[
Ii (ri , q i ; ) = Ii (0; ) Sex
2
( ) Û Û
ı ı
]
Ù Ù exp Í-
Î l R
(
rp ri cos q p - q i ˙ rp drp dq p
˚
) . (3-8)
a 0
Comparing Eqs. (2-7) and (3-8), we note that the significant difference between the
two lies in the lower limit of the integration over rp ; in Eq. (2-7), the lower limit is 0,
indicating an unobscured pupil; in Eq. (3-8) it is , indicating an obscured pupil. The
values of Sex are different in the two equations by a factor of 1 - 2 . The values of Pex
would also be different by this factor if the pupil irradiance were the same in both cases.
(
= r cos q p , sin q p ) , (3-9b)
r r
r = ri l F (3-10a)
and
r r
[
I ( r ; ) = Ii ( ri ; ) Pex Sex (0) l2 R 2 ] . (3-11)
Note that in Eq. (3-11), we have normalized the irradiance by the central irradiance for a
system with a circular pupil. Using normalized quantities, Eq. (3-8) may be written
2
1 2p
[ ( )] Û Û
[ ( )]
1
I (r, q i ; ) = p 2 1 - 2 Ù Ù exp - pir r cos q p - q i r dr dq p . (3-12)
ı ı
0
È1 ˘2
Û
[ (
I ( r; ) = 4 1 - 2 )] Í
ÍÙ
˙
J ( prr) r dr ˙
ı 0
. (3-13)
ÍÎ ˙˚
Carrying out the radial integration by using Eq. (2-14), we finally obtain
270 OPTICAL SYSTEMS WITH ANNULAR PUPILS
È 2 J1 ( pr ) 2 J ( pr ) ˘
2
1
I ( r ; ) = - 2 1 . (3-14)
(1 - ) 2 Í pr
Î pr ˙˚
We note that the irradiance distribution is radially symmetric about the Gaussian image
point r = 0 , as may be expected for a radially symmetric (annular) pupil function. It is
not normalized to unity at the center. Its central value is given by 1 - 2 . Except for a
normalization factor, Eq. (3-14) also gives the PSF of the system. It follows from Eq. (1-
61) that
r r
PSF ( ri ; ) = I i ( ri ; ) Pex (3-15)
We note that as Æ 0 , Eq. (3-14) for the annular pupil reduces to Eq. (2-15) for the
circular pupil. In order that the total power be the same for the two pupils, the irradiance
across the annular pupil must be higher than that for a circular pupil by a factor of
(1 - )
1
2
. For a given total power Pex in the exit pupil, the central irradiance I (0;) is
smaller by a factor of 1 - 2 compared to that for a circular pupil. However, if the pupil
irradiance is the same in both cases, as in astronomical observations, then Pex () is also
smaller than Pex (0) by a factor of 1 - 2 . Hence, I (0;) will be smaller than I (0; 0) by a
(
factor of 1 - 2 . ) 2
The principal maximum of the image irradiance distribution lies at the Gaussian
image point r = 0 , since all the Huygens’ spherical wavelets originating at the spherical
wavefront in the exit pupil arrive in phase at this point and, accordingly, interfere
constructively. From Eq. (3-14), we note that the image irradiance is zero at those values
of r for which
J1 ( p r ) = J1 ( pr ) , r π 0 . (3-16)
These values of r locate the minima of the irradiance distribution. Noting Eq. (2-19), we
find that the secondary maxima lie at those values of r that satisfy
J2 ( p r ) = 2 J2 ( pr ) , r π 0 . (3-17)
The irradiance distribution for a system with a very thin annulus pupil ( Æ 1) may
be obtained from Eq. (3-13) by noting that r ~ 1, and the variation of J 0 ( prr) is
negligibly small over the variation of r that it can be replaced by J 0 ( pr ) . Hence, Eq. (3-
13) reduces to
I (r; Æ 1) = J 02 ( p r ) (3-18)
Substituting Eqs. (3-10a) and (3-11) into Eq. (3-19) and defining a normalized or
fractional encircled power
we obtain
rc l F
Û
(
P(rc ; ) = p 2 Ù I (r; ) rdr .
ı
2
) (3-21)
0
If we let rc be in units of lF and substitute Eq. (3-14) into Eq. (3-21), we obtain
È 1 ˘
1 Í Û du ˙
P(rc ; ) = P(rc ) + P ( rc ) - 4 Ù J1 ( p rc u) J1 ( prc u)
2
, (3-22)
1 - 2 Í ı u˙
ÍÎ 0 ˙
˚
where
ÛÛ
Pi (rs ; ) = ÙÙ Ii (ri ; ) ri dri dq i , (3-24)
ıı
rs
where the integration is carried over the square region. Following the same procedure as
in the case of circular pupils (see Section 2.2.3), Eq. (3-24) reduces to
Ps (rs ; ) = Pc ( 2 rs ; - ) ( Û
Ù
p 1 - 2 ı
8
) [ J1 (p rsu) - J1 (prsu)]
2
cos 1
(1 u)
du
u
, (3-25)
1
where
272 OPTICAL SYSTEMS WITH ANNULAR PUPILS
and
are the fractional ensquared and encircled powers, respectively, and rs and rc are in units
of l F . The first term on the right-hand side of Eq. (3-25) represents the image power
contained in a circle of radius 2rs . The second term gives the image power contained in
a region lying between a circle of radius 2rs and a square of full-width 2rs . Both of
these terms require numerical integration.
[sin p (r - 1 4) -
8
]
2
I ( r ; ) = sin p (r - 1 4) . (3-28)
p r (1 - )
4 3 2
Noting that the average of a sine square is half and the average of the product of two sines
with different arguments is zero, the average irradiance (indicated by a bar) for large
values of r may be written
4
I (r; ) ~ . (3-29)
p r (1 - )
4 3
Xc (rc ; ) ~ ( Û
p 2 Ù
2
ı
) I (r; ) r dr
rc
2
= , (3-30)
p 2 rc (1 - )
Xs (rs ; ) ~ Û Û
Ù dx Ù dyI (r; )
ı ı
|x| > rs | y| > r s
4 2
= , (3-31)
p 3rs (1 - )
3.2 Aberration Free System 273
( )
12
where r = x 2 + y 2 and the subscripts c and s on X indicate a large circular or a square
region of exclusion centered on the Gaussian image point. From Eqs. (3-30) and (3-31),
we note that
The factor of 0.9 between Xs and Xc is independent of . We also note that excluded
power for an annular pupil is larger by a factor of (1 - ) compared to that for a
1
The approximate result of Eq. (3-29) and those that follow from it, although obtained
for the aberration-free case, are valid even when aberrations are present in the system.
This may be seen by substituting Eq. (3-42) given later into Eq. (1-154) and considering
the normalizations used in Eq. (3-29). It should be noted, however, that the value of r for
which Eq. (3-29) is valid increases as aberrations are introduced into the system.
Figure 3-3 shows the irradiance distribution given by Eq. (3-14) for = 0 , 0.25, 0.5,
and 0.75 normalized to unity at the center. This figure and Table 3-2 also show how
Pc , Ps and Ps - Pc vary with rc in graphical and tabulated forms. The value of Pc in a
given dark ring decreases or increases with in a manner similar to how its radius varies,
although a peak or a valley in its variation is not achieved for the same value of . For
small values of , the first maximum of Ps - Pc is the highest. However, for large values
of one of the secondary maxima is the highest. As increases, the secondary maxima
of irradiance become increasingly more significant, and therefore, Eqs. (3-30) and (3-31)
give accurate results for increasingly larger values of rc and rs , respectively. For
example, if the difference between actual results (given in Table 3-2) and those obtained
from Figure 3-2 is to be less than 2.6% of the total power, rc must be larger than 1.7, 1.8,
and 3.8 when is equal to 0.25, 0.50, and 0.75, respectively. A general rule of thumb is
274 OPTICAL SYSTEMS WITH ANNULAR PUPILS
1 1
10 – 1 10 – 1
= 0 = 0.5
(r; R)
(r; R)
10 – 2 10 – 2
10 – 3 10 – 3
10 – 4 10 – 4
0 1 2 3 4 5 0 1 2 3 4 5
r r
Figure 3-2a. Irradiance and encircled power distributions for an annular pupil. is
the obscuration ratio of the pupil. The example of a circular pupil is shown for
comparison. The dashed curves are for a Gaussian beam discussed in Chapter 4.
1.0
0.9 I
P
=0
0.8
0.7 0.25
I (r) P(rc)
0.6
0.5
0.50
0.4
0.3 0.75
0.2
0.1
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
r; rc
Figure 3-2b. The irradiance and encircled power distributions for various values of
.
that for accurate results (less than 1% error), rc must be large enough so that X is less
than 10%. It is interesting to note that although a square region of a certain width has
27% more area than a circular region of the same diameter, the difference between the
ensquared and encircled powers is less than 9% of the total power, regardless of the value
of . (A maximum difference of 8.97% occurs when = 0.54 .)
3.2 Aberration Free System 275
Table 3-1. Positions r of PSF maxima and minima for an annular pupil in units of
lF, and the corresponding irradiance and the encircled power.
Max/
Min r, rc I(r) P(rc) r, rc I(r) P(rc) r, rc I(r) P(rc) r, rc I(r) P(rc) r, rc I(r) P(rc)
Max 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0
Min 1.22 0 0.838 1.21 0 0.818 1.17 0 0.764 1.11 0 06.82 1.06 0 0.584
Max 1.63 0.0175 0.867 1.63 0.0206 0.853 1.63 0.0304 0.818 1.61 0.0475 0.766 1.58 0.0707 0.702
Min 2.23 0. 0.910 2.27 0 0.906 2.36 0 0.900 2.42 0 0.899 2.39 0 0.885
Max 2.68 0.0042 0.922 2.68 0.0031 0.914 2.69 0.0015 0.904 2.73 0.0011 0.902 2.77 0.0033 0.893
Min 3.24 0 0.938 3.18 0 0.925 3.09 0 0.908 3.10 0 0.904 3.30 0 0.903
Max 3.70 0.0016 0.944 3.70 0.0024 0.936 3.68 0.0037 0.926 3.64 0.0028 0.916 3.66 0.0007 0.905
Min 4.24 0 0.952 4.32 0 0.949 4.37 0 0.947 4.22 0 0.929 4.04 0 0.907
Max 4.71 0.0008 0.957 4.71 0.0004 0.951 4.74 0.0004 0.949 4.75 0.0016 0.938 4.66 0.0028 0.922
Min 5.24 0 0.961 5.15 0 0.953 5.16 0 0.951 5.42 0 0.949 5.31 0 0.939
Max 5.72 0.0004 0.964 5.71 0.0008 0.959 5.69 0.0006 0.955 5.73 0.0001 0.950 5.79 0.0008 0.944
Min 6.24 0 0.968 6.35 0 0.965 6.23 0 0.959 6.07 0 0.950 6.43 0 0.950
Max 6.72 0.0003 0.970 6.73 0.0001 0.966 6.74 0.0004 0.962 6.67 0.0006 0.955 6.72 0.0001 0.950
Min 7.25 0 0.972 7.14 0 0.967 7.35 0 0.966 7.27 0 0.961 7.03 0 0.950
Max 7.73 0.0002 0.974 7.72 0.0003 0.970 7.72 0.0001 0.967 7.77 0.0003 0.963 7.65 0.0004 0.954
Min 8.25 0 0.975 8.34 0 0.974 8.11 0 0.967 8.38 0 0.966 8.22 0 0.958
Max 8.73 0.0001 0.977 8.74 0.0001 0.975 8.72 0.0003 0.971 8.72 0.0000 0.966 8.77 0.0004 0.962
Min 9.25 0 0.978 9.16 0 0.975 9.38 0 0.974 9.06 0 0.967 9.46 0 0.966
Max 9.73 0.0001 0.979 9.72 0.0001 0.977 9.75 0.0000 0.975 9.70 0.0002 0.970 9.78 0.0000 0.966
Min 10.25 0 0.980 10.30 0 0.979 10.16 0 0.975 10.32 0 0.973 10.13 0 0.966
Max/
Min r, rc I(r) P(rc) r, rc I(r) P(rc) r, rc I(r) P(rc) r, rc I(r) P(rc) r, rc I(r) P(rc)
Max 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0
Min 1.000 0 0.479 0.95 0 0.372 0.90 0 0.269 0.85 0 0.172 0.81 0 0.082
Max 1.54 0.0963 0.618 1.48 0.1203 0.512 1.41 0.1395 0.389 1.35 0.1527 0.256 1.28 0.1600 0.124
Min 2.29 0 0.829 2.17 0 0.717 2.06 0 0.560 1.95 0 0.376 1.85 0 0.184
Max 2.76 0.0124 0.859 2.69 0.0306 0.784 2.58 0.0533 0.649 2.47 0.0734 0.456 2.35 0.0861 0.229
Min 3.49 0 0.901 3.39 0 0.873 3.22 0 0.761 3.05 0 0.554 2.90 0 0.284
Max 3.78 0.0004 0.902 3.84 0.0045 0.886 3.74 0.0192 0.808 3.57 0.0401 0.619 3.40 0.0566 0.328
Min 4.12 0 0.903 4.52 0 0.902 4.38 0 0.865 4.16 0 0.695 3.95 0 0.379
Max 4.50 0.0009 4.80 0.0001 0.903 4.86 0.0050 0.880 4.68 0.0218 0.741 4.46 0.0404 0.421
Min 5.05 0 0.910 5.11 0 0.903 5.52 0 0.899 5.27 0 0.795 5.00 0 0.468
Max 5.66 0.0022 0.923 5.58 0.0004 0.905 5.91 0.0005 0.901 5.78 0.0110 0.824 5.51 0.0299 0.507
Min 6.30 0 0.938 6.00 0 0.906 6.47 0 0.903 6.37 0 0.857 6.05 0 0.549
Max 6.81 0.0008 0.943 6.61 0.0016 0.916 6.72 0.000 0.903 6.87 0.0048 0.872 6.56 0.0224 0.584
Min 7.50 0 0.950 7.19 0 0.925 6.97 0 0.903 7.47 0 0.889 7.10 0 0.622
Max 7.79 0.0000 0.950 87.75 0.0013 0.943 7.53 0.0004 0.905 7.95 0.0016 0.894 6.61 0.0169 0.652
Min 8.12 0 0.950 8.40 0 0.944 7.98 0 0.906 8.57 0 0.901 8.16 0 0.685
Max 8.62 0.0001 0.951 8.87 0.0004 0.947 8.58 0.0010 0.913 8.98 0.0003 0.902 8.67 0.0127 0.711
Min 9.05 0 0.952 9.53 0 0.950 9.13 0 0.919 9.58 0 0.903 9.21 0 0.739
Max 9.68 0.0004 0.957 9.80 0.0000 0.950 9.69 0.0011 0.927 9.83 0.0000 0.903 9.72 0.0094 0.761
Min 10.31 0 0.962 10.11 0 0.950 10.28 0 0.935 10.10 0 0.903 10.26 0 0.784
276 OPTICAL SYSTEMS WITH ANNULAR PUPILS
1.0 1.0
In In Ps
Ps
0.8 0.8 Pc
Pc
(r) Pc(rc) Ps(rs)
0.2 0.2
0.0 0.0
0 1 2 3 4 0 1 2 3 4
r, rc, rs r, rc, rs
1.0
In Ps
0.8 Pc
(r) Pc(rc) Ps(rs)
= 0.75
0.6
10(Ps – Pc)
0.4
0.2
0.0
0 2 4 6 8
r, rc, rs
Figure 3-3. Encircled and ensquared power distributions for an annular pupil. The
irradiance distribution and the difference between ensquared and encircled power
distributions are also shown.
n = 2 (1 - ) , (3-33)
3.2 Aberration Free System 277
Table 3-2. Encircled and ensquared powers for a centrally obscured circular pupil
with a linear obscuration ratio of .
rc , rs Pc Ps Pc Ps Pc Ps Pc Ps
which is equal to the ratio of the outer diameter and the width of the annulus, provided
that n is an integer. The group minima are the lowest ring maxima and correspond to ring
numbers that are multiples of n, e.g., 10, 20, 30, etc., for = 0.8 . The radius of a ring
group is also a multiple of n (in units of l F ) since the spacing between two successive
maxima or minima is approximately unity. The central bright spot or the first dark ring of
radius 1.22 contains 83.8% of the total power in the image when = 0 . For = 0.8 , as
may be seen from Table 3-1, the first dark ring has a radius of 0.85 and contains only
17.2% of the total power. However, the central ring group in this case has a radius of
10.10 and contains 90.3% of the total power. When n is not an integer, the distribution
becomes complex. For example, for = 0.7 , n = 6.67 , and the distribution has a double
periodicity with the number of maxima in the two periods equal to 6 and 7 (two integers
closest to n).
r r r r r
Û
( ) ( )
t ( vi ; ) = Pex1 Ù A rp A rp - l R vi d rp
ı
, (3-34)
(r ) r
A rp = [ Pex Sex ( )] , a £ rp £ a
12
= 0, otherwise . (3-35)
The OTF is radially symmetric and represents the fractional area of overlap of two annuli
whose centers are separated by a distance l Rvi . Figures 3-5a through 3-5c show the
overlap region of the two annuli for various separations for >£ 1 3. We note from the
figure that to obtain the OTF, we need expressions for the overlap area of two circles of
equal radii whose centers are separated by a distance l Rvi and two circles of different
radii also separated by l Rvi . An expression for the former is given by Eq. (2-41). For the
latter, we refer to Figure 3-6 showing two circles of radii a1 and a2 separated by a
distance d, and let
(r )
A1 rp = [ Pex Sex ( )] ,
12 r
rp £ a1
= 0, otherwise (3-36)
and
(r )
A2 rp = [ Pex Sex ( )] ,
12 r
rp £ a2 . (3-37)
3.2 Aberration-Free System 279
10-2 10-2
10-3 -
i 10-4
10-5
10-6
10-7
o
10-1 E =0.8
10-2
10-5
10-6
10-7
10 20 30 40 50 o 10 20 30 40 50
--+ r --+r
Figure 3-4a. Irradiance distribution for systems with circular (E = 0) and annular
(E *- 0) pupils. The case E~ 1 represents the limiting case of a totally obscured
pupil. In practice, it approximates the PSF for a system with a very thin annular or
a ring pupil.
280 OPTICAL SYSTEMS WITH ANNULAR PUPILS
=0 = 0.5
= 0.8 Æ1
Figure 3-4b. 2D PSFs for systems with circular ( = 0) and annular ( π 0) pupils.
3.2 Aberration Free System 281
0 < n< n=
n n
< n
n
a1 a2
q1
q2
O A O¢
Figure 3-6. Overlap area of two circles of radii a1 and a2 separated by a distance d.
r r r r
Û
( ) ( ) [( ◊ ) (
Ù A1 rp A2 rp - d d rp = [ Pex Sex ( )] q1a1 - OA AB + q 2 a2 - AO¢ AB
ı
2 2
◊ )]
(
= [ Pex Sex ( )] q1a12 + q 2 a22 - da1 sin q1 ) , (3-38)
where
d 2 + a12 - a22
cos q1 = (3-39a)
2 da1
and
d 2 + a22 - a12
cos q 2 = . (3-39b)
2 da2
t (v; ) =
1
1 - 2
[ ]
t (v) + 2 t (v ) - t12 (v; ) , 0 £ v £ 1 , (3-40)
where t (v) is given by Eq. (2-44) and represents the OTF of the system if there were no
obscuration, v is a normalized radial spatial frequency defined by Eq. (2-43), and
284 OPTICAL SYSTEMS WITH ANNULAR PUPILS
( )
= (2 p) q1 + 2 q 2 - 2 v sin q1 , (1 - ) 2 £ v £ (1 + ) 2 (3-41b)
= 0, otherwise . (3-41c)
4v 2 + 1 - 2
cos q1 = (3-41d)
4v
and
4v 2 - 1 + 2
cos q 2 = , (3-41e)
4 v
1+
< v < 1 , t (v; ) > t (v)
2
( )
1
by a factor of 1 - 2 . It may be seen from Figure 3-5 that the overlap area in this
frequency range is independent of , but the fractional area is larger owing to the smaller
area of the obscured exit pupil. For a thin annular pupil, as Æ 1, a sharp peak near the
cutoff frequency is obtained. The peak frequency represents the spatial frequency of
fringes obtained in a 2D analog of a Young’s double-slit aperture. 3
How t (v;) varies with v is shown in Figure 3-7 for various values of , including
zero. We note that an annular pupil gives a higher OTF at high frequencies but a lower
OTF at low frequencies, compared to the OTF for a corresponding circular ( = 0) pupil.
This is the frequency domain analog of smaller radius of the central bright ring and
higher secondary maxima of the PSF for an annular pupil compared to those for a circular
pupil. Table 3-3 gives numerical values of t (v;) for = 0 (0.05) 0.95 and v = 0 (0.05) 1.
As pointed out in Section 2.2.5, the slope of t (v) at the origin is equal to - 4 p .
From Eq. (3-40) we find that the slope of the OTF for an annular pupil at the origin is
given by
t ¢(0; ) = - 4 p (1 - ) . (3-42)
As pointed out in Section 1.6.4, this slope does not change when aberrations are
introduced into the system. Equation (3-42) may also be obtained from Eq. (1-167) by
noting that t (v;) is radially symmetric, v is in units of 1 l F , L = 2 p (1 + ) a , and
( )
Sex = p 1 - 2 a 2 . We also note that
3.2 Aberration Free System 285
1.0
0.8
= 0
0.6
t (n; )
0.25
0.4 0.50
0.75
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
n
Since, according to Eqs. (1-83) and (1-84), the PSF and the OTF of a rotationally
symmetric system are related to each other by a zero-order Hankel transform, Eq. (3-40)
may also be obtained by taking such a transform of Eq. (3-14). This was done by
O’Neill 3 who obtained an expression for the OTF similar to Eq. (3-40) except that he
used an angle y instead of the angles q1 and q 2 of Eq. (3-41b), where, as may be seen
from Figure 3-6, y is the angle OBO¢ so that
y = p - (q1 + q 2 ) (3-44a)
and
1 + 2 - 4v 2
cos y = . (3-44b)
2
To show that his expression for the OTF is equivalent to the one given by Eq. (3-40), we
note from Figure 3-6 that
AB = a1 sin q1 = a2 sin q 2 .
Therefore,
= a2 a1 = sin q1 sin q 2
and
286 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Table 3-3. Aberration-free OTF t (v; ) of optical systems with annular pupils.
v\ 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
0 1 1 1 1 1 1 1 1 1 1
0.05 0.936 0.934 0.930 0.925 0.921 0.915 0.909 0.909 0.894 0.884
0.10 0.873 0.870 0.869 0.852 0.842 0.831 0.819 0.805 0.789 0.769
0.15 0.810 0.807 0.798 0.782 0.766 0.749 0.731 0.709 0.685 0.656
0.20 0.747 0.744 0.734 0.718 0.695 0.670 0.645 0.616 0.583 0.544
0.25 0.685 0.682 0.672 0.655 0.630 0.597 0.563 0.526 0.484 0.435
0.30 0.624 0.620 0.610 0.592 0.566 0.532 0.488 0.441 0.389 0.345
0.35 0.564 0.560 0.549 0.531 0.504 0.468 0.422 0.374 0.334 0.299
0.40 0.505 0.501 0.490 0.470 0.442 0.413 0.381 0.347 0.307 0.272
0.45 0.447 0.443 0.431 0.417 0.400 0.380 0.357 0.330 0.298 0.260
0.50 0.391 0.390 0.385 0.378 0.367 0.354 0.337 0.316 0.291 0.261
0.55 0.337 0.338 0.340 0.340 0.336 0.328 0.317 0.303 0.284 0.261
0.60 0.285 0.285 0.288 0.291 0.297 0.297 0.294 0.286 0.274 0.257
0.65 0.235 0.235 0.237 0.240 0.245 0.251 0.258 0.260 0.257 0.247
0.70 0.188 0.188 0.190 0.192 0.196 0.201 0.207 0.214 0.224 0.227
0.75 0.144 0.144 0.146 0.148 0.150 0.154 0.159 0.164 0.172 0.181
0.80 0.104 0.104 0.105 0.106 0.108 0.111 0.114 0.119 0.124 0.131
0.85 0.068 0.068 0.069 0.070 0.072 0.073 0.075 0.078 0.081 0.085
0.90 0.037 0.037 0.038 0.038 0.039 0.040 0.041 0.043 0.045 0.047
0.95 0.013 0.013 0.013 0.014 0.014 0.014 0.015 0.015 0.016 0.017
1 1 0 0 0 0 0 0 0 0 0
v\ 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95
0 1 1 1 1 1 1 1 1 1 1
0.05 0.873 0.859 0.841 0.818 0.788 0.745 0.682 0.576 0.364 0.163
0.10 0.746 0.718 0.683 0.637 0.577 0.492 0.365 0.254 0.164 0.080
0.15 0.621 0.579 0.526 0.457 0.367 0.289 0.224 0.165 0.108 0.054
0.20 0.498 0.441 0.371 0.310 0.258 0.211 0.167 0.124 0.082 0.041
0.25 0.377 0.327 0.283 0.243 0.206 0.169 0.134 0.100 0.066 0.033
0.30 0.306 0.270 0.236 0.204 0.173 0.143 0.113 0.084 0.056 0.028
0.35 0.266 0.236 0.207 0.179 0.151 0.125 0.099 0.074 0.049 0.024
0.40 0.241 0.213 0.187 0.161 0.136 0.112 0.089 0.066 0.044 0.022
0.45 0.227 0.199 0.173 0.149 0.126 0.104 0.082 0.061 0.040 0.020
0.50 0.224 0.192 0.166 0.141 0.119 0.097 0.077 0.057 0.037 0.019
0.55 0.231 0.194 0.163 0.138 0.115 0.093 0.073 0.054 0.036 0.018
0.60 0.235 0.207 0.169 0.138 0.114 0.091 0.071 0.052 0.034 0.017
0.65 0.233 0.213 0.186 0.148 0.117 0.092 0.071 0.052 0.034 0.016
0.70 0.222 0.211 0.194 0.167 0.128 0.097 0.073 0.952 0.034 0.016
0.75 0.192 0.195 0.190 0.175 0.151 0.111 0.078 0.055 0.035 0.017
0.80 0.139 0.149 0.163 0.166 0.157 0.135 0.194 0.061 0.037 0.017
0.85 0.091 0.098 0.106 0.118 0.134 0.136 0.120 0.077 0.042 0.019
0.90 0.050 0.054 0.058 0.065 0.073 0.085 0.104 0.102 0.060 0.023
0.95 0.018 0.019 0.021 0.023 0.026 0.030 0.037 0.048 0.070 0.041
1 0 0 0 0 0 0 0 0 0 0
3.3 Strehl Ratio and Aberration Tolerance 287
q1 + q 2 ˆ q - q2 ˆ
= - tan Ê cot Ê 1 .
Ë 2 ¯ Ë 2 ¯
Or,
p q - q2 ˆ 1+ p q + q2 ˆ
tan Ê + 1 = tan Ê - 1
Ë2 2 ¯ 1- Ë 2 2 ¯
giving
1 Ê1 + yˆ
(p + q1 - q 2 ) = tan 1 Á 1 - tan 2 ˜ .
2 Ë ¯
q1 + 2 q 2 = p 2 -
1
2
( )1
( )
1 + 2 y + 1 - 2 ( p + q1 - q 2 )
2
Ê1 + yˆ
= p 2 -
1
2
( ) ( )
1 + 2 y + 1 - 2 tan 1 Á
Ë1 -
tan ˜
2¯
. (3-45a)
d AB = (OA + O¢A) AB ,
or
È Ê1 + yˆ˘
1
( ) (
t12 (v; ) = (2 p) Íp 2 - 1 + 2 y - sin y + 1 - 2 tan 1 Á
2
)
Ë1 -
tan ˜ ˙ .
2 ¯˚
(3-46)
Î
In Eq. (3-40), if we replace Eq. (3-41b) by Eq. (3-46), we obtain O’Neill’s expression for
the OTF, an expression that was obtained by Steel4 earlier using the convolution
approach. However, it seems that Eq. (3-41b) is simpler than Eq. (3-46).
[( )] Û Û
[ ( )] [ ( )]
2
I (r, q i ; ) = p 1 - 2 Ù Ù exp iF r, q p ; exp - pi r cos q p - q i r dr dq p , (3-47)
ı ı
0
( )
where F r, q p ; is the aberration of the system. By definition, the Strehl ratio of the
image or the system is given by the ratio of the irradiances at the center r = 0 with and
without aberrations. The aberration-free central irradiance is unity [as may be seen from
Eq. (3-47)] in units of Pex Sex () l2 R 2 [see Eq. (3-3a)]. Hence, the central irradiance
according to Eq. (3-47) gives the Strehl ratio, i.e.,
2
1 2p
[( )] Û Û
[ ]
2
S = p 1 - 2 Ù Ù exp iF(r, q; ) r dr dq , (3-48)
ı ı
0
where we have dropped the subscript p on the angle q p for simplicity. Approximate
expressions for the Strehl ratio when the aberration is small are given by Eqs. (1-204)
through (1-206), i.e.,
~ (1 - s 2F 2)
2
S1 , (3-49a)
S2 ~ 1 - s 2F , (3-49b)
and
S3 ~ exp (- s 2F ) , (3-49c)
where
is the variance of the aberration across the uniformly illuminated annular pupil. The mean
and the mean square values of the aberration are obtained from the expression
1 2p
[( )] ÛÛ
1
Ù Ù F (r, q; ) r dr dq ,
n
< F > = p 1- 2 n
(3-50b)
ıı
0
F(r; ) = As r 4 + Bd r2 . (3-51)
We determine the amount of defocus Bd such that the variance sF2 is minimized; i.e., we
calculate sF2 and let
∂s 2F
= 0 (3-52)
∂Bd
Table 3-4. Primary aberrations and their standard deviations for a system with a
uniformly illuminated annular pupil.
Aberration F( r,, q) sF
Spherical As r 4 (4 - )
12
2
- 6 4 - 6 + 4 8 As 3 5
Coma Ac r3 cos q (1 + )
12
2
+ 4 + 6 Ac 2 2
Astigmatism Aa r2 cos 2 q (1 + ) 2 12
Aa 4
Astigmatism and coma aberrations can be treated similarly. Table 3-5 lists the form
of a balanced primary aberration and its standard deviation. Also listed in the table is the
location of the diffraction focus, i.e., the point with respect to which the aberration
variance is minimum so that the Strehl ratio is maximum at it. We note that in the case of
coma, the balancing aberration is a wavefront tilt whose amount depends on . Thus,
maximum Strehl ratio is obtained at a point that is displaced from the Gaussian image
point but lies in the Gaussian image plane. In the case of astigmatism, the amount of
balancing defocus is independent of .
Figure 3-8 shows how the standard deviation of an aberration, for a given value of
the aberration coefficient Ai , varies with the obscuration ratio of the pupil. In Figures
3-8a and 3-8b, the amounts of defocus and tilt required to minimize the variance of
spherical aberration and coma, respectively, are also shown. We observe from these
figures that the standard deviation of spherical and balanced spherical aberrations and
defocus decreases as increases. Correspondingly, the tolerance in terms of their
aberration coefficients As and Bd , for a given Strehl ratio, increases. Thus, for example,
the depth of focus for a certain value of the Strehl ratio increases as increases. The
standard deviation of coma, astigmatism, balanced astigmatism, and tilt increases as
increases. The standard deviation of balanced coma first slightly increases, achieves its
maximum value at = 0.29 and then decreases rapidly as increases. The factor by
which the standard deviation of an aberration is reduced by balancing it with another
aberration is reduced in the case of spherical aberration, but increases in the case of coma
and astigmatism, as increases.
Table 3-5. Balanced primary aberrations, their standard deviation, and diffraction
focus.
Balanced
spherical [ (
As r 4 - 1 + 2 r 2 ) ] 1
6 5
1 - 2( )
2
As [0,0,8(1 + )F A ]
2 2
s
Balanced
coma
Ê
Ac Á r3 -
2 1 + 2 + 4 ˆ
r˜ cos q (1 - ) (1 + 4 + )
2 2 4 12
Ac Í
(
È 4 1 + 2 + 4 ) ˘
FAc , 0, 0 ˙
3 1 + 2
Ë ¯
6 2 (1 + ) 2 12
(
Í 3 1+
Î
2
) ˙
˚
Balanced
(
A r 2 cos 2 q - 1 2
astigmatism a
) 1
(1 + 2
+ 4 )
12
Aa (0, 0, 4 F A )
2
a
2 6
3.3 Strehl Ratio and Aberration To lerance 29 1
~~
.£
U
Q) '"
0;
.0
u
Q)
'-'
c
",-& 0.1 5 Q)
0.06 '-'
U
co 0.6 0.06 ~
c E
t e
t
0
'"
0; ~
~u 0.4
0.10 0.04
.0-
0.05 0.02 t 0.2
0.40 0.30
0.25
0.35
0.20
'"
~ ~
",<> 0.30 ",<> 0. 15
t t 0. 10
0.25
0.05
0.20 0.00
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
--+E --+ E
(c) (d)
0.75
0.70
<i: 0.65
~
t 0.60
0.55
0.50
0.0 0.2 0.4 0.6 0.8 1.0
--+E
(e)
Table 3-6. Exact expressions for Strehl ratio for primary aberrations.*
Aberration F( r, q ; ) S
Spherical As r 4
[ p / 2 As (1 – 2 )2 ]
{
¥ [C( 2 As / p ) – C( 2 As / p 2 )]2 + [ S( 2 As / p ) – S( 2 As / p 2 )]2 }
Balanced spherical As [r4 – (1 + 2 )r2 ] {
[2 p / As (1 – 2 )2 ] C 2 [( As / 2 p)1/ 2 (1 – 2 )] + S 2 [( As / 2 p)1/ 2 (1 – 2 )]}
2
È 1 ˘
Coma Ac r3 cos q (1 – 2 ) 2 Í
Î Ú
2
J0 ( Ac x 3 / 2 ) dx ˙
˚
È 2 (1 + 2 + 4 ) ˘ 2
Ac Ír3 – r˙ cos q Ï È Ê 2 1 + 2 + 4 1/ 2 ˆ ˘ ¸Ô
Ô 1
Balanced coma ÍÎ 3 (1 + 2 ) ˙˚ (1 – 2 ) 2 Ì
ÔÓ
Ú
2
J0 Í Ac Á x 3 / 2 –
ÍÎ Ë 3 1 + 2
x ˜ ˙ dx ˝
¯ ˙˚ Ô
˛
Astigmatism Ad r2 cos2 q (1 – 2 ) 2 [ H 2 ( Aa / 2) + 2 H 2 ( 2 Aa / 2) – 2 2 H ( Aa / 2) H ( 2 Aa / 2)
¥ cos[(1 – 2 )( Aa / 2) – a( Aa / 2) + a( 2 Aa / 2)]}
2
ÏÔ • ¸Ô
Balanced astigmatism Aa r2 (cos2 q – 1 / 2)
ÔÓ k =0
Â
Ì[ 4 / Aa (1 – )] J2 k +1 ( Aa / 2) – J2 k +1 ( Aa / 2)˝
2 2
Ô˛
2
È sin[ Ad (1 – 2 ) / 2] ˘
Curvature of field (defocus) Ad r2 Í ˙
ÍÎ Ad (1 – ) / 2 ˙˚
2
2
È 2J (A ) 2 J ( At ) ˘
Distortion (tilt) At r cos q (1 – 2 ) 2 Í 1 t – 2 1 ˙
ÍÎ At At ˙˚
* A i is the coefficient of aberration in radians, C (b ) = Ú0b cos(p x 2 / 2) dx, S (b ) = Ú0b sin(p x 2 / 2) dx,
H (b ) = [ J 02 (b ) + J12 (b )], a (b ) = tan –1[ J1 (b ) / J 0 (b )].
b
Û
C(b) = Ù cos p x 2 2 dx
ı
( ) (3-53a)
0
and
b
Û
(
S(b) = Ù sin p x 2 2 dx ,
ı
) (3-53b)
0
◊
respectively. The quantities H ( ) and a( ) are given by ◊
[ ]
12
H (b) = J 02 (b) + J12 (b) (3-54a)
and
respectively. We note that in the case of coma the integration must be carried out
numerically. In the case of balanced astigmatism, only the first few terms of the infinite
series need to be considered for adequate precision.
3.3 Strehl Ratio and Aberration Tolerance 293
Figures 3-9a and 3-9b show how the Strehl ratio of a primary aberration varies with
its standard deviation for = 0.5 and 0.75, respectively. Approximate as well as exact
results are shown in these figures. The curves for a given aberration and the
corresponding balanced aberration can be distinguished from each other by their behavior
for large s w values (near 0.25l). For example, coma is shown by the evenly dashed
curves; the higher dashed curve is for coma and the lower is for balanced coma. The same
holds true for astigmatism. The following observations may be made from Figures 3-9a
and 3-9b.
i. For small values of s w , the Strehl ratio is independent of the type of aberration. It
depends only on its variance.
ii. The expressions for S1 and S2 underestimate the true Strehl ratio.
iii. The expression for S3 overestimates the true Strehl ratio for > ~ 0.5 . It gives the
Strehl ratio with an error of less than 10% for S >
~ 0 . 4. For smaller obscurations, the error
is less than 10% for S ~ 0.3. The percent error is defined to be 100(1 - S3 S ) .
>
iv. S3 gives a better approximation for the true Strehl ratio than S1 and S2 .
The reason is that, for small values of s F , it is larger than S1 by approximately sF4 4 .
Of course, S1 is larger than S2 by s F
4
4.
v. The Strehl ratio depends strongly on the standard deviation of an aberration but
weakly on its detailed distribution over a wide range of Strehl ratio values and not just for
large values of it.
Using S1 to estimate the Strehl ratio, Figure 3-10 shows how the aberration
coefficient Ai (in units of l ) of a primary aberration for 10% error varies with the
obscuration ratio. It is evident that this coefficient increases with obscuration in the case
of spherical, balanced spherical, and balanced coma, but decreases in the case of
astigmatism, balanced astigmatism, and coma. Thus, S1 estimates the true Strehl ratio
with a small error for a larger aberration as increases in the case of spherical, balanced
spherical, and balanced coma, but a smaller aberration in the case of astigmatism,
balanced astigmatism, and coma.
1.0
= 0.5
0.8
0.6
S
0.4 S3
0.2
S2
S1
0.0
0.00 0.05 0.10 0.15 0.20 0.25
sw
Figure 3-9a. Strehl ratio for annular pupils with = 0.5 as a function of the
standard deviation sw of an aberration in units of l. The Strehl ratio for a given
value of the standard deviation for classical coma is practically the same as that for
balanced coma. For large values of sw , the Strehl ratio for classical astigmatism is
larger than that for balanced astigmatism. Spherical...., Coma----, Astigmatism–.–.
3.3 Strehl Ratio and Aberration Tolerance 295
1.0
= 0.75
0.8
0.6
S
S3
0.4
0.2
S2
S1
0.0
0.00 0.05 0.10 0.15 0.20 0.25
sw
Figure 3-9b. Strehl ratio for annular pupils with = 0.75 as a function of the
standard deviation sw in units of l . For large values of sw , the Strehl ratio for
balanced coma is higher than that for coma. The opposite is true for astigmatism.
Note that the curves for coma and astigmatism are practically identical.
Spherical...., Coma----, Astigmatism–.–.
296 OPTICAL SYSTEMS WITH ANNULAR PUPILS
1.0 5.0
Balanced
astigmatism
0.6 3.0
Ai
Astigmatism
0.4 2.0
Coma
Balanced
coma
0.2 1.0
Spherical
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
1.0 0.75
BS
BC
0.9 0.65
BA
0.8 0.55
S
S
0.7 0.45
C
0.6 0.35
0.5 0.25
0.0 0.2 0.4 0.6 0.8 1.0
• n
F(r, q; ) = Â Â c nm Z nm (r, q; ) , £r £1 , 0 £ q £ 2p , (3-55a)
n =0 m =0
where cnm are the orthonormal expansion coefficients that depend on the field angle of
the object, n and m are positive integers, n - m ≥ 0 and even, and
[
Z nm (r, q; ) = 2( n + 1) (1 + d m 0 ) ]1/ 2Rnm (r; ) cos mq . (3-55b)
1 1 2p
m¢
Ú Ú Z n (r, q; ) Z n ¢ (r, q; ) r dr d q = d nn ¢ d mm ¢
m
. (3-55c)
(
p 1- 2
)
The annular polynomials Z nm (r, q; ) are similar to the circle polynomials Z nm (r, q)
discussed in Section 2.4, except that they are orthogonal over an annular pupil. Thus, they
are also unique in that they are the only polynomials in two variables r and q, which
(a) are orthogonal over an annulus, (b) are invariant in form with respect to rotation of the
coordinate axes about the origin, and (c) include a polynomial for each permissible pair
of n and m values. They can be obtained from the corresponding circle polynomials by
the Gram-Schmidt orthogonalization process.9
È ( n m) 2 ˘
Rnm (r; ) = N nm Í Rnm (r) - Â (n - 2i + 1) < Rnm (r) Rnm 2i (r; ) > Rnm 2i (r; )˙ , (3-56a)
Î i ≥1 ˚
where
2 1 m
< Rnm (r) Rnm¢ (r; ) > = Ú Rn (r) Rn ¢ (r; ) r dr
m
(3-56b)
1 - 2
and Nnm is a normalization constant such that the radial polynomials satisfy the
orthogonality relation
1 1 - 2
Ú Rn (r; ) Rn ¢ (r; ) r dr = d
m m
. (3-56c)
2(n+ 1) nn ¢
For m = 0 , the radial polynomials are equal to the Legendre polynomials Pn (◊)
according to
R20n
È 2 r2 - 2
(r; ) = Pn Í - 1
˘
˙ .
( ) (3-57)
Í 1 - 2 ˙
Î ˚
Thus, they can be obtained from the circle radial polynomials R20n (r) by replacing r by
[(r 2
- 2 ) (1 - )] 2 12
; i.e.,
ÈÊ r2 - 2 ˆ 1 2˘
R20n (r; ) = R20n ÍÁ 2 ˜
˙ . (3-58)
ÍÎË 1 - ¯ ˙˚
{( ) (1 - 2(n +1) )}
12
= r n 1 - 2 . (3-59b)
Moreover,
Rnn 2 (r; ) =
nrn - (n - 1) 1 - 2 n [( ) (1 - ( ) )] r
2 n 1 n 2
. (3-59c)
( ) - (n - 1)(1 - ) (1 - ( ) )˘˚˙¸˝˛
12
(
Ï 1 - 2
)
1È 2 2 ( n +1
) 2n 2 2 n 1
Ì n 1- 2
Ó ÎÍ
It is evident that the radial polynomial Rnn (r; ) differs from the corresponding circle
polynomial Rnn (r) only in its normalization. We also note that
Rnm (1; ) = 1, m = 0
π 1, m π 0 . (3-60)
1 1 2p
c nm = Ú Ú F(r, q; )Z n (r, q; ) r dr d q ,
m
(
p 1- 2
) 0
(3-61)
as may be seen by substituting Eq. (3-55a) and using the orthonormality of the
polynomials. The Zernike annular radial polynomials for n £ 8 are listed in Table 3-7.
The number of Zernike (or orthogonal) aberration terms in the expansion of an aberration
function through a certain order n is given by Eqs. (2-63a) and (2-63b), as in the case of a
circular pupil.
n (r, q; ) = c nm Z n (r, q; ) .
Fm m
(3-62)
3.4 Balanced Aberrations and Zernike Annular Polynomials 299
n m Rnm (r; )
0 0 1
1 1 (
r 1 + 2 )1 2
2 0 ( 2r 2
) (1 - )
- 1 - 2 2
2 2 r (1 + + )
2 2 4 12
3 (1 + ) r - 2 (1 + + ) r
2 3 2 4
(1 - ) [(1 + ) (1 + 4 + )]
3 1 12
2 2 2 4
3 3 r (1 + + + )
3 2 4 6 12
4 0 [6r - 6 (1 + ) r + 1 + 4 + ] (1 - )
4 2 2 2 4 2 2
4r - 3 [(1 - ) (1 - )] r
4 8 6 2
4 2
Ï 1 2¸
Ì(1 - ) Í16 (1 - ) - 15 (1 - ) (1 - )˙
È 2 1 ˘ 10 8 2 6
˝
Ó Î ˚ ˛
( )
12
4 4 r 4 1 + 2 + 4 + 6 + 8
( ) ( ) (
10 1 + 4 2 + 4 r5 - 12 1 + 4 2 + 4 4 + 6 r3 + 3 1 + 4 2 + 10 4 + 4 6 + 8 r )
(1 - ) [(1 + 4 + ) (1 + 9 + 9 )]
5 1 12
2 2 2 4 2 4
+ 6
5 r - 4 [(1 - ) (1 - )] r 5 10 8 3
5 3 12
Ì( ) ÈÎÍ25 (1 - ) - 24 (1 - ) (1 - )˘˚˙¸˝˛
Ï1- 2 1 12 10 2 8
Ó
( )
12
5 5 r5 1 + 2 + 4 + 6 + 8 + 10
6 0 [20 r 6
( ) (
- 30 1 + 2 r 4 + 12 1 + 32 + 4 r2 - 1 + 9 2 + 9 4 + 6 ) ( )] (1 - )
2 3
( )
15 1 + 4 2 + 10 4 + 4 6 + 8 r6 - 20 1 + 4 2 + 10 4 + 10 6 + 4 8 + 10 r 4 ( )
6 2
( + 6 1 + 4 + 10 + 20 + 10 + 4 +
2 4 6 8 10 12
)r 2
(1 + ) [(1 + 4 )( )]
12
2 2 2
+ 10 4 + 4 6 + 8 1 + 9 2 + 454 + 656 + 458 + 9 10 + 12
6r - 5 [(1 - ) (1 - )] r 6 12 10 4
6 4 12
Ì( ) ( ) ( ) (1 - )˘ ¸˝
Ï1- È36 1 - - 35 1 -
2 1 14 12 2 10
Ó Í
Î ˚˙ ˛
( )
12
6 6 r6 1 + 2 + 4 + 6 + 8 + 10 + 12
300 OPTICAL SYSTEMS WITH ANNULAR PUPILS
n m Rnm (r; )
7 5
7r7 - 6 1 - 14 [( ) (1 - )] r 12 5
12
(
Ï 1 - 2
) ( ) - 48 (1 - ) (1 - )˘˙˚¸˝˛
1È 14 2
Ì 49 1 - 16 12
Ó ÍÎ
( )
12
7 7 r7 1 + 2 + 4 + 6 + 8 + 10 + 12 + 14
8 0
( ) ( )
70 r8 - 140 1 + 2 r6 + 30 3 + 82 + 34 r 4 - 20 1 + 6 2 + 6 4 + 6 r2 + e80 ( )
(1 - ) 2 4
(
a17 = 35 1 + 9 2 + 9 4 + 6 ) A71
(
b71 = - 60 1 + 9 2 + 154 + 9 6 + 8 ) A71
(
c17 = 30 1 + 9 2 + 254 + 256 + 9 8 + 10 ) A71
(
d71 = - 4 1 + 9 2 + 454 + 656 + 458 + 9 10 + 12 ) A71
( ) (1 + 9 ) (1 + 16 )
3 12 12
A71 = 1 - 2 2
+ 9 4 + 6 2
+ 36 4 + 16 6 + 8
(
a73 = 21 1 + 4 2 + 10 4 + 20 6 + 10 8 + 4 10 + 12 ) A73
(
b73 = - 30 1 + 4 2 + 10 4 + 20 6 + 20 8 + 10 10 + 4 12 + 14 ) A73
(
c73 = 10 1 + 4 2 + 10 4 + 20 6 + 358 + 20 10 + 10 12 + 4 14 + 16 ) A73
( ) (1 + 4 + 10 + 20 + 10 + 4 + )
2 12 1 2
A73 = 1 - 2 2 4 6 8 10
e80 = 1 + 16 2 + 36 4 + 16 6 + 8
3.4 Balanced Aberrations and Zernike Annular Polynomials 301
1 2p 1 2p
Û Û Û Û
Fm
n (r, q) n (r, q) r dr d q
= Ù Ù Fm Ù Ù r dr d q
ı ı ı ı
0 0
= 0 , n π 0, m π 0 . (3-63a)
For m = 0 , this may be seen with the help of Eq. (3-55b) and the fact that R00 (r; ) = 1 is
a member of the polynomial set. The orthonormality Eq. (3-55c) yields the result that the
mean value of Rn0 (r; ) is zero. When m π 0 , the average value of cos mq is zero.
Similarly, the mean square value of the aberration is given by
1 2p 1 2p
Û Û Û Û
[Fm
n (r, q; )]
2
n (r, q; )
= Ù Ù Fm
ı ı
[ ] 2
r dr d q Ù Ù r dr d q
ı ı
0 0
= cnm
2
. (3-63b)
(F mn ) 2
2
s 2nm = - Fm
n
= cnm
2
, n π 0, m π 0 . (3-64)
Thus, each expansion coefficient, with the exception of c00 , represents the standard
deviation of the corresponding aberration term. The variance of the aberration function is
accordingly given by
• n
s 2F = F 2 (r, q; ) - F(r, q; )
2
= Â 2
 c nm . (3-65)
n =1 m = 0
The balanced primary aberrations discussed in Section 3.3.3 can be easily identified
with the corresponding Zernike annular polynomials. For example, for n = 4 and m = 0 ,
Eq. (3-62) becomes
= [ ( )
5 c40 6 r 4 - 6 1 + 2 r2 + 1 + 4 2 + 4 ( )] (1 - )
2 2
. (3-66a)
302 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Comparing this with the balanced spherical aberration given in Table 3-5, namely,
[ (
F bs (r) = As r 4 - 1 + 2 r2 ) ] , (3-66b)
we note the following. The aberration F04 contains a constant (independent of r and q)
term. This term does not change the standard deviation of the balanced aberration or the
Strehl ratio corresponding to it. In Eq. (3-66a), as in Eq. (3-66b), the spherical aberration
is balanced with an amount of defocus that is - 1 + 2 times the amount of the ( )
aberration. Comparing the coefficients of the r 4 term, we find immediately that the
standard deviation of the balanced spherical aberration is given by
s bs = c40
( )
2
= As 1 - 2 6 5 , (3-66c)
=
[( )
2 2 c 31 3 1 + 2 r3 - 2 1 + 2 + 2 r cos q ( )] . (3-67a)
( ) [( )( )]
12
1 - 2 1 + 2 1 + 4 2 + 4
Ê 2 1 + 2 + 4 ˆ
F bc (r, q; ) = Ac Á r 3 - r˜ cos q , (3-67b)
Ë 3 1 + 2 ¯
s bc = c31
( )( ) ( )
12 12
= Ac 1 - 2 1 + 4 2 + 4 6 2 1 + 2 . (3-67c)
( ) (1 + )
12
= 2 6 c 22 r2 cos 2 q - 1 2 2
+ 4 . (3-68a)
(
Fba (r, q; ) = Aa r 2 cos 2 q - 1 2 ) , (3-68b)
3.4 Balanced Aberrations and Zernike Annular Polynomials 303
sba = c22
( )
12
= Aa 1 + 2 + 4 2 6 . (3-68c)
= (
3 c20 2r2 - 1 - 2 ) (1 - )
2
, (3-69a)
F d (r; ) = Bd r2 (3-69b)
sd = c20
(
= Bd 1 - 2 ) 2 3 . (3-69c)
( )
12
= 2c11r cos q 1 + 2 , (3-70a)
s t = c11
( )
12
= Bt 1 + 2 2 . (3-70c)
F 00 (r, q; ) = c 00 , (3-71)
which represents a uniform (piston) aberration. Obviously, it has no effect on the standard
deviation or the Strehl ratio of a system with a single exit pupil. In a multiexit pupil
system, any relative piston errors among the subpupils will cause some destructive
interference.
304 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Thus, we see that Zernike polynomials can be identified with balanced aberrations;
that, in fact, is their advantage. Here we have discussed only the primary aberrations. In
general, the aberration function of an optical system may consist of higher-order
aberrations. Moreover, in a system without an axis of rotational symmetry, the aberration
function will consist of terms not only in cos mq but in sin mq as well.
1 2
[ ( )] Û
( )
2
I (r; z; ) = 2 R z 1 - 2 Ù exp i Bd r J 0 ( p r rR z ) r dr
2
, (3-72)
ı
Bd = N Ê - 1ˆ
R
(3-73)
Ëz ¯
is the peak defocus phase aberration (relative to a zero aberration at the center of the
annular pupil), and
N = a2 l R (3-74)
is the Fresnel number of the exit pupil as observed from the focal point if there were no
obscuration. The corresponding Fresnel number of the annular pupil is given by
(
N = N 1 - 2 ) . (3-75)
If we let z = R , Eq. (3-72) reduces to Eq. (3-13), except for the difference due to
different normalizations in the two equations; the irradiance is normalized to values of 1
and 1 - 2 at the Gaussian image point in these equations, respectively. As in the case of
systems with circular pupils, the irradiance distribution is asymmetric about the Gaussian
image plane for the same three reasons, unless N is very large so that z ~ R .
3.5 Defocused System 305
or
Pex Sex ( )
Ii (0; z; ) = S , (3-76b)
l2 z 2
where
{ [ ( ) ] [ B (1 - ) 2]}
2
S = sin Bd 1 - 2 2 d
2
. (3-77)
Equation (3-77) differs from the corresponding Eq. (2-84c) for systems with circular
( )
pupils in that the quantity Bd in the latter has been replaced by Bd 1 - 2 . It represents
the peak defocus phase aberration at the outer edge of the annular pupil relative to its
value at the inner edge. Also, the area Sex () given by Eq. (3-2) and as a result the focal-
point irradiance given by Eq. (3-3) are smaller for an annular beam by a factor of 1 - 2
compared to a circular beam.
( )
The axial irradiance is minimum and equal to zero when Bd 1 - 2 is an integral
multiple of 2p. The corresponding z values are given by
( )
R z = 1 + 2 n N 1 - 2 , n = ± 1 , ± 2 , K . (3-78)
When N is very large (>>10), the axial irradiance is zero when z is different from R by an
( )
integral multiple of ± 8l F 2 1 - 2 . Accordingly, the depth of focus of a system with an
( )
annular pupil is larger by a factor of 1 1 - 2 , compared to a corresponding system with
a circular pupil. The maxima of axial irradiance, obtained by equating the derivative of
Eq. (3-76b) with respect to z equal to zero, are given by the solutions of
[ ( ) ] (
tan Bd 1 - 2 2 = ( R z ) Bd 1 - 2 2 , z π R . ) (3-79)
Figure 3-12a shows how the axial irradiance of an annular beam with = 0.5 varies
for N = 1, 10, and 100. Comparing it with Figure 2-11, we note that the effect of the
obscuration is to reduce the irradiance at the principal maximum, but to increase it at the
secondary maxima. Also, the maxima and minima occur at smaller z values for an annular
aperture. As in the case of circular beams, the axial irradiance of annular beams also
becomes symmetric about the focal point z = R as N increases.
306 OPTICAL SYSTEMS WITH ANNULAR PUPILS
N=1 N = 10 N = 100
1.5 = 0.5 1.5 = 0.5 1.5 = 0.5
(0; z)
lu
1.0 1.0 1.0
lg
0.5 0.5 0.5
Figure 3-12a. Axial irradiance of an annular beam with = 0.5 focused at a distance
R. The minima of irradiance occur at z R = 3 11, 3/19, 3/27, when N = 1. The
irradiance is in units of the focal-point irradiance of a corresponding circular beam
with the same total power. Accordingly, the focal-point irradiance in this figure is
1 - 2 = 0.75 . The axial irradiance becomes symmetric about the focal point as N
increases. The dashed curves are for a Gaussian beam with g = 1, as discussed in
Chapter 4.
lg
0.6 0.6 0.6
z
Figure 3-12b. Central irradiance (in units of Pex Sex l 2 z 2 ) at a distance z from the
plane of the exit pupil when a beam is focused at various distances R. The quantity
N z = a 2 l z represents the Fresnel number of the exit pupil as observed from the
target. The dashed curves are for a Gaussian beam.
Although the principal maximum of axial irradiance does not lie at the geometrical
focus, the maximum central irradiance on a target at a fixed distance z is obtained when
the beam is focused on it, i.e., when R = z , as may be seen by differentiating Eq. (3-76b)
with respect to R and equating the result to zero. It is evident that the central irradiance on
a target when the beam is focused on it is Pex Sex ( ) l2 z 2 . Figure 3-12b illustrates how
the central irradiance on a target at a fixed distance z varies when the beam is focused at
various distances R along its axis. The irradiance in this figure is in units of
Pex Sex ( ) l2 z 2 . The quantity N z = a 2 l z in this figure represents the Fresnel number of
3.5 Defocused System 307
a circular exit pupil as observed from the target. We note that as N z increases, the
curves become symmetric about the point R = z .
2 1 1
È 2R ˘ Û Û
I ( r; z; ) = Í
( -
˙ Ù Ù cos Bd r 2 - s 2
) [ ( )] J 0 (prr) J 0 (prs) r s dr ds . (3-80)
˚˙ ı ı
2
ÍÎ z 1
If we let r = 0 and note that J 0 (0) = 1 , we obtain a different form of the expression for
axial irradiance, namely,
1 1
[ ( )] [ ( )]
2Û Û
I (0; z; ) = 2 R z 1 - 2
Ù Ù cos Bd r - s r s dr ds .
2 2
(3-81)
ı ı
where Q (r, s; rc ) is given by Eqs. (2-108a) and (2-108b). The integrals in Eqs. (3-80)
through (3-82) may be evaluated by the Gauss quadrature method, according to which,
for a function ¶(r, s)
1 1
M M i 1
Û Û
[
Ù Ù f (r, s) d r ds = (1 - ) 2
ı ı
]2 Â w i2 f (ri , si ) + 2 Â Â ( )
w i w j f ri , s j , (3-83)
i =1 i=2 j =1
where, as in Eq. (2-118), M is the number of points of a 1D quadrature, w i are the weight
factors, and
ri = si
[
= 1 + + (1 - ) xi ] 2 , (3-84)
xi being the i-th zero of the M th-order Legendre polynomial. Note that by letting Bd = 0
in Eqs. (3-80) through (3-82), we can calculate the focal-plane distributions as well.
Equations (3-80) through (3-83) are generalizations of Eqs. (2-104) through (2-107) for
circular beams to annular beams with F (r) = Bd r2 . Note that with slight modification,
Eqs. (3-80) through (3-82) can be applied to diffraction calculations involving any
radially symmetric amplitude and phase distributions at the exit pupil. For example, if
spherical aberration As r 4 were present, the cosine factor in these equations would
[ ]
become cos F (r) - F (s) , where F (r) = Bd r2 + As r 4 .
308 OPTICAL SYSTEMS WITH ANNULAR PUPILS
0.6 1.2
0.514 = 0.5
0.4 0.8
P(rc)
(r)
0.3 0.6
Pg
0.2 0.4
Pu
0.1 0.2
lg
lu
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5
r; rc
Bd = p a 2 l z . (3-85)
It represents the peak phase aberration of a plane wavefront with respect to a reference
sphere centered at a distance z from the exit pupil. Equation (3-76b) for the axial
irradiance reduces to
[ ( )
I (0; z; ) = 4 I0 sin 2 p a 2 1 - 2 2 lz ] , (3-86)
where
is the irradiance at the exit pupil. The axial irradiance is maximum and equal to 4 I0 at z
values given by
(
z = a 2 1 - 2 ) l(2 n + 1) , n = 0, 1, 2, K . (3-88)
(
z = a 2 1 - 2 ) 2 l(n + 1) , n = 0, 1, 2, K (3-89)
These z values for the location of maxima and minima correspond to those axial positions
at which the annular exit pupil subtends an odd or even number of Fresnel’s half-wave
zones, respectively.
( )
For z > a 2 1 - 2 l, it decreases monotonically to zero. For
(
z ≥ D2 1 - 2 l , ) (3-90)
( )
corresponding to a defocus aberration of £ l 8 1 - 2 , it decreases approximately as
z 2 . For z satisfying Eq. (3-90), a collimated beam gives an axial irradiance at a distance
z that is ≥ 0.95 times the irradiance at this point if the beam were focused at it, i.e.,
S ≥ 0.95. This is illustrated in Figure 3-14, where the axial irradiance of a collimated
beam is plotted as a function of z. The distance z in this figure is in units of
( )
D2 1 - 2 l , which may be called the far-field distance of the annular exit pupil. The
axial irradiance is normalized by the exit pupil irradiance Pex Sex (0) for a circular beam
of the same power. The pupil irradiance and, therefore, the maxima of axial irradiance are
( )
1
higher for an annular beam compared to a circular beam by a factor of 1 - 2 . It is
evident that a collimated beam yields practically the same irradiance on a target lying in
the far field of the annular exit pupil as a beam focused on it.
310 OPTICAL SYSTEMS WITH ANNULAR PUPILS
N=0
5
= 0.5
lg
4
(0;z)
lu
2
0
0.0 0.1 0.2 0.3 0.4 0.5
z
Figure 3-14. Axial irradiance of a collimated annular beam normalized by the exit
pupil irradiance Pex Sex (0) for a circular beam of the same power. The pupil
irradiance and, therefore, the maxima of axial irradiance are higher for an annular
( )
1
beam compared to a circular beam by a factor of 1 - 2 . The axial distance z is
(
in units of the far-field distance D 2 1 - 2 l . The dashed curve is for a )
corresponding Gaussian annular beam, which is discussed in Chapter 4 .
The irradiance distribution in a plane at a distance z can be obtained from Eq. (3-72)
by letting R Æ • and noting that the units of irradiance in this equation are
Pex Sex ( ) l2 R 2 and those of r are l F = l R D . Thus, for a collimated annular beam, we
may write
1 2
I (r; z; ) =
4 Û
( )
Ù exp i Bd r J 0 ( p rr) r dr
2
, (3-91)
(1 - )2 2 ı
where the units of irradiance are Pex Sex ( ) l2 z 2 and those of r are lz D. For
( )
z ≥ D2 1 - 2 l , since the effect of the defocus aberration is negligibly small, Eq. (3-
91) reduces to the aberration-free result, Eq. (3-14):
È 2 J1 ( p r )
2
1 2 J1 ( pr ) ˘
I (r; z; ) ~ Í - 2 ˙ . (3-92)
(1 - )2 2
Î pr pr ˚
Equation (3-90) is the far-field condition for an annular exit pupil and Eq. (3-92)
represents its far-field or Fraunhofer diffraction pattern. Except for the units of irradiance
and z, Eq. (3-92) is the same as Eq. (3-14). Hence, the discussion of Section 3.2 on
encircled power, is applicable to this equation.
3.6 Symmetry Properties of an Aberrated PSF 311
where Anm is the aberration coefficient. We will refer to the plane in which the center of
the reference sphere of radius of curvature R lies as the Gaussian image plane. If the
image is observed in another plane that lies at a distance z from the exit pupil, then the
aberration becomes
p a2 Ê 1 1 ˆ
Bd = - (3-94a)
l Ë z R¯
= p N Ê - 1ˆ .
R
(3-94b)
Ëz ¯
1 2p 2
I (r , q i ; z ) =
(R z) 2
Û Û Ï È R ˘¸
Ù Ù exp Ìi ÍF (r, q) - p z rr cos (q - q i ) ˙ ˝ r dr dq . (3-95)
[( )]
2
p 1- 2 ı ı Ó Î ˚˛
0
i. When N is small, the irradiance distribution is asymmetric about the Gaussian image
plane whether or not the system is aberrated. The distribution in any plane normal to the z
axis is m-fold symmetric. Moreover, the tangential plane and all planes containing the z
axis and making angles of p j m with the tangential plane, where j = 1, 2, ..., m, are
planes of symmetry. When m = 0 , the irradiance distribution in any observation plane is
radially symmetric. The aberration function given by Eq. (3-93) also possesses these
symmetry properties.
ii. For large values of N, we may let z ~ R in which case Eq. (3-95) reduces to
2
1 2p
I ( r, q i ; z ) =
1 Û Û
{[
Ù Ù exp i F(r, q) - p rr cos(q - q i ) ] } r dr d q , (3-96)
[( )]
2
p 1- 2 ı ı
0
312 OPTICAL SYSTEMS WITH ANNULAR PUPILS
where
pa2
Bd = ( R - z) . (3-97)
l R2
We note from Eq. (3-97) that the magnitude of Bd for z = R + D is the same as for
z = R - D , but its sign does change. Thus, for large Fresnel numbers, two planes located
symmetrically about the Gaussian image plane correspond to Bd values that are equal in
magnitude but opposite in sign.
The irradiance distribution is not symmetric about the Gaussian image plane when
m = 0 , i.e., for spherical aberration. However, the irradiance distributions in two planes
located symmetrically about the Gaussian image plane are identical if they are for
spherical aberrations of equal magnitude but opposite signs. The two aberration functions
in this case are different from each other only in their signs.
When m is odd, as for coma, the irradiance distribution is symmetric about the
Gaussian image plane, even though the corresponding aberration functions are not equal
to each other. When m is even, as for astigmatism, the distribution in the Gaussian image
plane is 2m-fold symmetric although the aberration function is only m-fold symmetric.
The irradiance distributions in two planes located symmetrically about the Gaussian
image plane differ from each other by a rotation of p m about the z axis. The
corresponding aberration functions have equal magnitude but opposite signs.
iii. The axial irradiance for large Fresnel numbers [obtained by letting r = 0 in Eq. (3-
96)] is symmetric about the Gaussian image plane. This is evident from the symmetry
properties of the distribution when m is not equal to zero. When m is equal to zero,
although the distribution is not symmetric about the Gaussian image plane, the axial
irradiance is. To see this, we let r = 0 in Eq. (3-96), which for spherical aberration may
be written
1 2
I (0; z ) =
4 Û
[ ] ( )
Ù exp i An 0 Rn (r; ) exp iBd r r dr
0 2
. (3-98)
(1 - )2 2 ı
Now, since
R20n r;
È 2 r2 - 2
( ) = Pn Í
( ˘
- 1˙ ,
) (3-99)
Í 1- 2
˙
Î ˚
◊
where Pn ( ) is a Legendre polynomial of degree n, we let
[(
x = 2 r2 - 2 ) (1 - )] - 1
2
(3-100)
1 2
I (0; z ) =
1 Û
4 Ù
ı
[ ( ) ] [
exp i Bd x 1 - 2 2 exp i An 0 Pn /2 ( x ) dx ] . (3-101)
1
F(r) = As r 4 + Bd r2 . (3-102)
( ) {[
S = Èp 2 As 1 - 2 ˘ C(b+ ) + C(b )] [
+ S(b+ ) + S(b )] }
2 2 2
, (3-103)
ÎÍ ˙˚
where
b± = [(1 - ) A ± d] (2p A )
2
s s
12
, (3-104)
and
(
d = Bd + 1 + 2 As ) . (3-105)
It is evident that the Strehl ratio as a function Bd represents the axial irradiance. Hence,
the axial irradiance of a system aberrated by spherical aberration As is independent of the
sign of d. Now d = 0 corresponds to a defocus value Bd = - 1 + 2 As , which, in turn, ( )
represents its optimum value giving minimum variance of the aberration given by Eq. (3-
102). Indeed, for this defocus value, if we add an appropriate value of piston aberration,
the aberration of Eq. (3-102) takes the form of Zernike annular polynomial R40 (r;) .
Hence, the axial irradiance for primary spherical aberration is symmetric about the axial
point with respect to which the aberration variance is minimum, or it is symmetric about
the Gaussian image plane when the aberration is represented by the Zernike annular
polynomial R40 (r;) . It should be noted that the point of axial symmetry in the case of
astigmatism Aa r2 cos 2 q corresponds to Bd = - Aa 2, which is independent of the value
of . Similarly, in the case of coma Ac r3 cos q , the axial irradiance is symmetric about
the Gaussian image plane.
iv. A change in the sign of the aberration coefficient Anm when m is even has no effect
on the irradiance distribution in the Gaussian image plane. Thus, the sign of such an
aberration cannot be determined from this PSF. However, the irradiance distribution in a
defocused image plane does change when the sign of the aberration coefficient is
314 OPTICAL SYSTEMS WITH ANNULAR PUPILS
The axial irradiance for astigmatism is shown in Figure 3-17. It is symmetric about
the point Bd = - Aa 2 , as indicated by an arrow on each curve. Figure 3-18 shows the
axial irradiance for one wave of coma. The irradiance for coma without any tilt is along
the z axis. For coma optimally balanced with tilt, the axial irradiance is along an axis that
is parallel to the z axis but passing through the diffraction focus (1.4, 0). The PSFs for
coma are discussed in Section 3.9.
1.0 0.20
Aberration free W(r) = Asr4
0.8 Bd = – 1(1+2) As = 1 0.16
= 0.5
0.6 0.12
I(r; Bd)
– 0.5(1+2)
0.4 0.08
– 1.5(1+2)
0 (5¥)
0.2 – 2(1+2) (5¥)) 0.04
0.0 0.00
0 1 2 3 4
r
Figure 3-15. PSFs for = 0.5 and one wave of spherical aberration in various
defocused image planes. The right-hand side vertical scale is for PSFs for Bd = 0
( )
and - 2 1 + 2 l since they have been multiplied by 5.
1.0
1 W(r) = Asr4
= 0.5
0.8 2
As = 0
0.6
I(0; Bd)
0.4
0.2
0.0
–6 –5 –4 –3 –2 –1 0 1 2 3
Bd
Figure 3-16. Axial irradiance for = 0.5 and various values of spherical aberration.
It is symmetric about the point Bd = - As 1 + 2 . ( )
316 OPTICAL SYSTEMS WITH ANNULAR PUPILS
1.0
W(r,q) = Aar2cos2q Aa = 0
0.8 = 0.5
0.6
I(0; Bd)
1
0.4
0.2
2
3
0.0
–6 –5 –4 –3 –2 –1 0 1 2 3
Bd
Figure 3-17. Axial irradiance for = 0.5 and various values of astigmatism. It is
symmetric about a point where an arrow is indicated on a curve and corresponds to
Bd = - Aa 2 .
0.6
W(r,q) = Acr3cosq
0.5 Ac = 1
= 0.5
0.4
Through
(1.4, 0)
I(Bd)
0.3
0.2
On axis
0.1
(0, 0)
0.0
–4 –3 –2 –1 0 1 2 3 4
Bd
Figure 3-18. Axial irradiance for = 0.5 with one wave of coma. It is symmetric
about the Gaussian image plane Bd = 0.
3.8 2D PSFs 317
3.8 2D PSFs
The PSFs for spherical aberration shown in Figure 3-15 or for coma shown in
Section 3.9 are useful for quantitative assessment. However, as stated in Section 2.8.5
regarding the PSFs for circular pupils, they do not lend themselves easily to what they
may look like when observed in practice. This is especially true when the PSF is not
radially symmetric. Accordingly, we give computer-generated 2D pictures of the PSFs in
this section for various values of a primary aberration. The emphasis of these pictures is
on the structure of a PSF, i.e., on the distribution of its bright and dark regions, and not
on its irradiance distribution. Some of the symmetry properties of the aberrated PSFs
discussed above are evident from these pictures.
Figures 3-19 and 3-20 illustrate the defocused PSFs for = 0.5 and = 0.8. From
( )
Eq. (3-77) the axial irradiance is zero when Bd 1 - 2 is equal to an integral number of
wavelengths. Thus, when = 0.5, the central irradiance of the PSFs observed in image
planes corresponding to Bd = ± 4 3 , ± 8 3, etc., is zero. Similarly, when = 0.8, the
central irradiance is zero in planes corresponding to Bd = ± 2.78 , ± 5.56 , etc. Hence,
there is a dark spot at the center of these PSFs. The PSFs aberrated by spherical
( )
aberration are shown in Figures 3-21 and 3-22. Both the classical As r 4 and balanced
[ ( ) ]
As r 4 - 1 + 2 r2 aberrations are considered. There is no dark spot at the center of the
PSFs for these aberrations.
(
Figures 3-23 and 3-24 show the PSFs aberrated by classical Aa r2 cos 2 q and )
( )
balanced Aa r2 cos 2 q - r2 2 astigmatism, respectively for = 0.5. The balancing
defocus is independent of the value of . For a given value of Aa , the PSFs in planes
Bd = 0 and Bd = - Aa are the sagittal and tangential images. They are identical to each
other except for a rotation of one with respect to the other by p 2. This is a special case
of the general result that the PSFs in two planes located symmetrically about the plane
Bd = - Aa 2 for balanced astigmatism are identical except for a rotation by p 2. For
example, when Aa = 3 l , the PSFs in defocused planes Bd = 1.5 l and Bd = - 4.5 l
differ from each other only by the rotation. The PSFs have bilateral symmetry. The PSF
in a defocused plane Bd = -1.5 l corresponds to balanced astigmatism and has a 4-fold
symmetry. Similar results are obtained when = 0.8. The sagittal and tangential images
for a large value of deviate considerably from the line images of geometrical optics
even when the aberration is large. See, for example, the sagittal image for Aa = 3 l , i.e.,
in the plane Bd = 0 . The central bright line appears broken near the center when
compared with the corresponding image for = 0 in Figure 2-35.
The irradiance distribution for coma is symmetric about the tangential plane. Thus, it
has a line symmetry in any observation plane, the line lying in the tangential plane. The
distribution in two observation planes located symmetrically about the Gaussian image
plane are identical. A change in the sign of the aberration coefficient A c produces a
rotation of the distributions by p about the axis. The PSFs aberrated by coma are shown
in Figures 3-27 and 3-28. The conical shape of the PSFs for large aberration when = 0
is lost for large values of .
318 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Bd = 0 Bd = 4 3
Bd = 0.5 Bd = 8 3
Bd = 1 Bd = 16 3
Figure 3-19. Defocused PSFs for = 0.5 . The central irradiance of a PSF is zero
( )
when Bd 1 - 2 = ± nl , where n is an integer.
3.8 2D PSFs 319
Bd = 1 Bd = 2.78
Bd = 2 Bd = 5.56
Bd = 3 Bd = 8.33
Figure 3-20. Defocused PSFs for = 0.8 . The central irradiance of a PSF is zero
( )
when Bd 1 - 2 = ± nl , where n is an integer.
320 OPTICAL SYSTEMS WITH ANNULAR PUPILS
As = 0.5, Bd = 0 As = 1, Bd = - 1.25
As = 1, Bd = 0 As = 2, Bd = - 2.5
As = 2, Bd = 0 As = 3, Bd = - 3.75
As = 0.5, Bd = 0 As = 1, Bd = - 1.64
As = 1, Bd = 0 As = 2, Bd = - 3.28
As = 2, Bd = 0 As = 3, Bd = - 4.92
Gaussian image plane (
Defocused plane Bd = - 1 + 2 As )
Figure 3-22. PSFs for = 0.8 and spherical aberration.
322 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Aa = 1, Bd = 0 Aa = 1, Bd = 0.5
Aa = 3, Bd = 1.5 Aa = 3, Bd = 0
Figure 3-23. PSFs for = 0.5 and astigmatism observed in various image planes.
3.8 2D PSFs 323
Aa = 1 4 Aa = 1 2
Aa = 1 Aa = 2
Aa = 3
( )
Figure 3-24. PSFs for = 0.5 and balanced astigmatism Aa r2 cos 2 q - r2 2 . Thus,
Bd = - Aa 2 and the PSFs are 4-fold symmetric.
324 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Aa = 1, Bd = 0 Aa = 1, Bd = 0.5
Aa = 1, Bd = - 1.5
Aa = 3, Bd = 0 . Aa = 3, Bd = 1.5
Figure 3-25. PSFs for = 0.8 and astigmatism observed in various image planes.
3.8 2D PSFs 325
Aa = 1 4 Aa = 1 2
Aa = 1
Aa = 2 Aa = 3
( )
Figure 3-26. PSFs for = 0.8 and balanced astigmatism Aa r2 cos 2 q - r2 2 . Thus,
Bd = - Aa 2 and the PSFs are 4-fold symmetric.
326 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Ac = 1 4 Ac = 1 2
Ac = 1 Ac = 2
Ac = 3
Figure 3-27. PSFs for = 0.5 and coma Ac r3 cos q. They are symmetric about the
horizontal axis.
3.8 2D PSFs 327
Ac = 1 4 Ac = 1 2
Ac = 1 Ac = 2
Ac = 3
Figure 3-28. PSFs for = 0.8 and coma Ac r3 cos q. They are symmetric about the
horizontal axis.
328 OPTICAL SYSTEMS WITH ANNULAR PUPILS
where cnm and snm are the orthonormal Zernike aberration coefficients. For a system
with a uniformly illuminated annular pupil of inner and outer radii a and a ,
respectively, the centroid of its aberrated PSF, following Eqs. (1-181a) and (1-181b), is
given by
2p
Û
< xi , yi > = [ R Sex ( )] Ù
ı
[W (a, q; ) - W (a, q; )] a (cos q,sin q) d q , (3-107)
0
where we have made use of Eqs. (2-150a) and (2-150b). Substituting Eq. (3-106) into Eq.
(3-107), we obtain
[ ( )] [ ]
•
< xi , yi > = 2 F 1 - 2 Â ¢ 2(n + 1) Rn (1; ) - Rn ( ; ) (cn1 , sn1 ) ,
1 1
(3-108)
n =1
where a prime on the summation sign indicates a summation over odd integral values of
n. Thus, the only aberrations that contribute to the LOS error are those with m = 1.
Aberrations of the type Rn1 (r;) cos q contribute to < xi > and those of the type
Rn1 (r;) sin q contribute to < yi > . This follows from the symmetry of the aberrations.
We note that, whereas two aberration terms with m = 1 but different values of n give
(approximately) the same Strehl ratio if their coefficients are equal in magnitude, their
contribution to the LOS error is different for a different aberration.
Since a radial polynomial Rn1 (r;) consists of terms in rn , rn 2 , ..., and r, with their
coefficients varying with , there is no loss of generality if we consider aberrations of the
type rn cos q , where n is an odd integer, to determine their contribution to the LOS error.
Thus, we consider, for simplicity, an aberration
where Wn is its peak value (at the outer edge of the annular pupil relative to a value of
3.9 Line of Sight of an Aberrated System 329
zero at its center). Substituting Eq. (3-109) into Eq. (3-107), we obtain
( n 1) 2
< x > = 2Wn  2i , (3-110)
i=0
where Wn is in units of l , < x > is in units of l F , and it is understood that < y > = 0 .
As discussed in Section 2.9, when = 0 , the LOS error depends only on Wn , but not on
the power n of r in Eq. (3-109). However, when π 0 , the LOS error does depend on n.
The reason for this is that, even if the aberration value along the outer perimeter of an
annular pupil is the same for different values of n, its value along the inner perimeter is
different. Hence, for a given value of Wn an annular pupil gives aberrated PSFs with
different centroids for different orders of the aberration. Of course, these observations
also hold for aberrations represented by Zernike annular polynomials, as may be seen by
noting that whereas Rn1 (1; 0) = 1 , regardless of the value of n, Rn1 (1;) and Rn1 (; ) do
depend on n.
Following the procedure used in Section 2.9.1, it can be shown that for an
aberration
•
W (r, q) = Â ¢ Wnr cos q
n
, (3-111)
n =1
the irradiance distribution of the image of a point object formed by an aberrated optical
system with an annular exit pupil may be written
2
È1 ˘
[ (
I ( r , q i ; ) = 4 1 - 2 )] ÍÛ J ( p B) r dr ˙
ÍÙı
0 ˙
, (3-112)
ÍÎ ˙˚
where B is given by Eq. (2-155) and Wn is in units of l. The centroid of the distribution
is given by
• ( n 1) 2
<x> = 2  ¢Wn  2i . (3-113)
n =1 i=0
The PSF simply shifts such that its peak and centroid locations move from (0, 0) to
where
B = (2t W3 - x ) t 1 2 . (3-118)
Figure 3-29 shows how I(x) varies with x for several values of W3 (in units of l) varying
from 0 to 2 and 2 = 0.5 . The centroid of the PSFs is given by
(
< x > = 2W3 1 + 2 ) . (3-119)
Thus, for a given value of W3 , the centroid for an annular pupil shifts by an amount that
( )
is larger by a factor of 1 + 2 than that for a circular pupil.
1.0
W3 = 0
2 = 0.5 0.5
0.8
0.6
l
1.5
0.4
0.2 2
0.0
3 2 1 0 1 2 3
x
Figure 3-29. PSF for several typical values of primary coma W3 in units of l and
2 = 0.5 .
3.9 Line of Sight of an Aberrated System 331
For small values of W3 , the peak value of the aberrated PSF occurs at a point such
that, if the aberration is measured with respect to a reference sphere centered at this point,
the variance of the aberration across the annular pupil is minimum. From the properties of
the Zernike annular polynomials, we find that the polynomial R31 (r;) cos q gives the
optimum combination of r3 cos q and r cos q terms that lead to a minimum variance.
Since
R31 (r; ) =
( ) (
3 1 + 2 r3 - 2 1 + 2 + 4 r ) , (3-120)
(1 - ) [(1 + ) (1 + 4 + )]
12
2 2 2 4
we note that, for small values of W3 , the peak value of the aberrated PSF occurs at
(
x m = 4 W3 1 + 2 + 4 3 1 + 2 ) ( ) , (3-121)
where the subscript m refers to the point corresponding to minimum aberration variance.
From the form of the aberration, it is understood that ym = 0 . Thus an amount W3 of
primary coma shifts the centroid and peak of the PSF by different amounts, the
( ) ( )
2
movement of the peak being 2 1 + 2 + 4 3 1 + 2 of the movement of the centroid.
Figure 3-30 shows how the irradiance Im at x m , the peak irradiance I p , and the
irradiance Ic at < x > vary with W3 . Figure 3-31 shows how x m , x p (the point at which
the peak irradiance occurs) and < x > vary with W3 . The observations made above about
the PSFs aberrated by primary coma are evident from these figures. Several typical values
of x m , x p , and < x > and the corresponding irradiances Im , I p , and Ic are noted in
Table 3-9 for 2 = 0.5 . We note that the peak value lies approximately at x m for small
values of W3 . For large values of W3 , it occurs at a point that is closer to the origin than
x m . The distance of the peak from the origin does not increase monotonically but
fluctuates as W3 increases. Since according to Eq. (3-110), the distance of the centroid
increases linearly with W3 , it is clear that the separation between the locations of the
centroid and the peak increases as W3 increases.
Table 3-9. Typical values of x m , x p , and < x > and the corresponding irradiances
Im , I p , and Ic for annular pupils aberrated by primary coma* when 2 = 0.5 .
W5 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
*The aberrated central irradiance I(0), i.e., the Strehl ratio, is also given here.
332 OPTICAL SYSTEMS WITH ANNULAR PUPILS
1.0
2 = 0.5
0.8
0.6
l
0.4
lc
0.2
lp
lm
0.0
0 2 4 6 8 10
W3
30
2 = 0.5
20
x
<x>
xm
10
xp
0
0 2 4 6 8 10
W3
The aberrated PSF along the x-axis is given by Eq. (3-117), where
(
B = 2t 2 W5 - x t 1 2 ) . (3-123)
Figure 3-32 shows how I ( x;) varies with x for several values of W5 and 2 = 0.5 .
Following Eq. (3-110), the centroid of the PSF is given by
(
< x > = 2 W5 1 + 2 + 4 ) . (3-124)
(
x m = W5 1 + 2 + 4 + 6 ) (1 + ) 2
. (3-125)
The values of x m , x p , and < x > and the corresponding irradiances Im , I p , and Ic are
noted in Table 3-10 for the values of W5 considered in Figure 3-32 and 2 = 0.5 .
1.0
W5 = 0
2 = 0.5
0.8
0.5
0.6
l
1
0.4
0.2 1.5
0.0
3 2 1 0 1 2 3
x
Figure 3-32. PSF for several typical values of secondary coma W5 in units of l and
2 = 0.5 .
334 OPTICAL SYSTEMS WITH ANNULAR PUPILS
Table 3-10. Typical values of x m , x p , and < x > and the corresponding irradiances
Im , I p , and Ic for annular pupils aberrated by secondary coma* when 2 = 0.5 .
W5 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
*The aberrated central irradiance I(0), i.e., the Strehl ratio, is also given here.
R15 (r; ) =
( ) ( ) (
10 1 + 4 2 + 4 r 5 - 12 1 + 4 2 + 4 4 + 6 + 3 1 + 4 2 + 10 4 + 4 6 + 8 r ) . (3-126)
(1 - ) ÈÍÎ(1 + 4 )( )
2 12
2 2
+ 4 1 + 9 2 + 9 4 + 6 ˘
˙˚
(
W (r, q) = W 5r 5 + W 3r 3 cos q , ) (3-127)
where
(
W3 = - 1.2 W5 1 + 4 2 + 4 4 + 6 ) (1 + 4 2
+ 4 ) . (3-128)
According to Eq. (3-126), the point in the image plane with respect to which the
aberration variance is minimized is given by
(
x m = - 0.6 W5 1 + 4 2 + 10 4 + 4 6 + 8 ) (1 + 4 2
+ 4 ) . (3-129)
Substituting Eqs. (3-127) and (3-128) into Eq. (3-113), we obtain the centroid
(
< x > = - W5 0.4 + 2 2 + 7.2 4 + 2 6 + 0.48 ) (1 + 4 2
+ 4 ) . (3-130)
Substituting Eq. (3-127) into Eq. (3-112), we find that the aberrated PSF along the x-axis
is given by Eq. (3-117), where
(
B = 2t 2 W5 + 2tW3 - x t 1 2 ) . (3-131)
3.9 Line of Sight of an Aberrated System 335
Figure 3-33 shows the aberrated PSF I(x) for several values of W5 with W3 given by Eq.
(3-128) and 2 = 0.5 . The values of x m , x p , and < x > and the corresponding irradiances
Im , I p , and Ic are given in Table 3-11. Note that x m , x p and < x > are all negative.
Moreover, their magnitude for the values of W5 considered is very large. Therefore, in
Figure 3-33, the horizontal coordinate is chosen to be x - x m .
1.0
W5 = 0
2 = 0.5
W3 ~ – 1.52 W5 5
0.8 xm ~ – 1.52 W5
0.6
10
0.4
15
0.2
20
0.0
–3 –2 –1 0 1 2 3
x – xm
Figure 3-33. PSF for aberration given by Eq. (3-127) and 2 = 0.5 . Note that the
horizontal coordinate is x - x m .
Table 3-11. Typical values of x m , x p , and < x > and the corresponding irradiances
Im , I p , and Ic for annular pupils aberrated by a combination of primary and
secondary coma given by Eq. (3-127).
W5 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
3.10 SUMMARY
Astronomical telescopes generally have annular pupils. Many of the results of
imaging with such pupils are also applicable to propagation of annular beams. From an
analysis standpoint, the principal effect of the central obscuration of the pupil compared
to imaging with a circular pupil is to change the lower limit of a radial integration from 0
to . The main results of this chapter are summarized below.
i. PSF
The obscuration of an annular pupil not only blocks the light incident on it, but it also
reduces the size of the central bright spot and brightens the secondary maxima of the PSF.
Because of the smaller area of the annular pupil and less light transmitted by it, the
central irradiance of the PSF is smaller than its corresponding value for a circular pupil
( )2
by a factor of 1 - 2 . As illustrated in Figure 3-4, the PSF consists of not only the
bright and dark rings but also a periodic ring group structure. The number of rings in a
group is given by 2 (1 - ). It may be seen, for example, from Table 3-1 that the first dark
ring for = 0.8 has a radius of 0.85 (in units of l F ) and contains only 17.2% of the total
image light compared to the corresponding numbers of 1.22 and 83.8% for a circular
pupil. However, the central ring group consists of 10 rings, has a radius of 10.10, and
contains 90.3% of the light. The limiting PSF as Æ 1 is given by J 02 (p r ) . The 2D
aberrated PSFs are illustrated in Section 3.8.
ii. OTF
Since the aberration-free OTF represents the fractional overlap area of two annular
pupils separated by a distance that is proportional to the spatial frequency, the cutoff
frequency is determined by the outer diameter of the pupil. Consequently, its value of
1 l F is independent of the obscuration ratio. The obscuration reduces the OTF at low
spatial frequencies, but increases it slightly at the high frequencies. As may be seen from
Figure 3-5, the overlap area in the frequency range (1 + ) 2 < v < 1 is independent of ,
but the fractional area is larger owing to the smaller area of the annular pupil, thus
yielding a higher value of the OTF compared to its corresponding value for a circular
pupil. The slope of the OTF at the origin for an annular pupil is - 4 p (1 - ) , regardless
of the aberration. It is thus larger by a factor of (1 - ) compared to that for a circular
1
pupil.
The aberrations of a system with an annular pupil are balanced to reduce their
variance in the same manner as for a circular pupil. The balanced primary aberrations are
listed in Table 3-5. The amount of balancing defocus in the case of astigmatism is
independent of . As for a circular pupil, the Strehl ratio for a small aberration is
( )
approximately given by exp - s F2 , where s F is the standard deviation of the phase
aberration across the annular pupil. For primary aberrations, it overestimates the true
3.10 Summary 337
As shown in Figure 3-8, the standard deviation of spherical, balanced spherical, and
defocus aberrations decreases as increases. Correspondingly, their tolerance for a given
Strehl ratio increases. The standard deviation of coma, astigmatism, balanced
astigmatism, and tilt increases as increases. The standard deviation of balanced coma
first slightly increases, achieves its maximum value at = 0.29 and then decreases
rapidly as increases. The factor by which the standard deviation of an aberration is
reduced by balancing it with another aberration is reduced in the case of spherical
aberration, but increases in the case of coma and astigmatism, as increases.
The depth of focus for a certain value of the Strehl ratio increases by a factor of
(1 - 2 ) 1 as increases. As in the case of a circular pupil, it is small or large depending
on whether the Fresnel number of the focusing beam is large or small. Similarly, a beam
focused with a small Fresnel number yields maximum central irradiance on a target when
the beam is focused on it even though a larger on-axis irradiance is obtained at a point
closer to the focusing pupil.
( )
The far-field distance of an annular pupil is given by D 2 1 - 2 l , where a
collimated beam compared to a focused beam represents a defocus aberration of
( )
l 8 1 - 2 and yields a Strehl ratio of 0.95. Thus a collimated beam yields practically
the same irradiance on a target lying in the far field of the annular exit pupil as a beam
focused on it. As in the case of a circular pupil, the near-field or Fresnel diffraction is
equivalent to a defocused far-field or Fraunhofer diffraction.
The LOS error is defined as the displacement of the centroid of an aberrated PSF
from its aberration-free location. As shown in Section 3.9, only the coma aberrations
contribute to the LOS error. Aberrations varying as cos q contribute a LOS error in the
tangential plane and those varying as sinq contribute in the sagittal plane. Unlike a
circular pupil, the primary and secondary comas with the same coefficient yield different
LOS errors. A primary coma W 3r 3 cos q , with W 3 in units of l , yields a LOS error of
( )
W 3 1 + 2 in units of l F . Similarly, a secondary coma W 5r 5 cos q yields an error of
( )
W 5 1 + 2 + 4 . For small aberrations, the peak value of an aberrated PSF lies at the
point with respect to which the aberration variance of coma is minimum. The centroid
does not coincide with this point.
References 339
REFERENCES
1. V. N. Mahajan, “Included power for obscured circular pupils,” Appl. Opt. 17,
964–968 (1978).
3. E. L. O’Neill, “Transfer function for an annular aperture,” J. Opt. Soc. Am. 46,
285–288 (1956). Note that a term of - 2 h2 is missing in the second of O’Neill’s
Eq. (26).
5. V. N. Mahajan, “Strehl ratio for primary aberrations: some analytical results for
circular and annular pupils,” J. Opt. Soc. Am. 72, 1258–1266 (1982); errata, J.
Opt. Soc. Am. A 10, 2092 (1993).
10. V. N. Mahajan, “Axial irradiance and optimum focusing of laser beams,” Appl.
Opt. 22, 3042–3053 (1983).
12. V. N. Mahajan, “Line of sight of an aberrated optical system,” J. Opt. Soc. Am. A
2, 833–846 (1985).
340 OPTICAL SYSTEMS WITH ANNULAR PUPILS
PROBLEMS
1. Consider an optical imaging system with an annular exit pupil of obscuration ratio
= 0.5 forming the image of a point object with a focal ratio of 8. Let the pupil of
outer radius 2 cm be uniformly illuminated with an irradiance of 0.1 W/cm2 at a
wavelength of 0.5 mm. (a) Determine the radius of the central bright spot of the
image, the amount of power contained in it, and the value of the irradiance at its
center. (b) Give the radius of the first bright ring of the image and the maximum
value of its irradiance distribution. How much power is contained in the bright
ring? (c) Give the diameter of the third dark ring and the power lying outside it. (d)
How much power would a square detector centered on the image collect if its half
width is equal to the radius of the third dark ring? (d) What is the spatial frequency
for which the MTF is equal to 0.5? What is the cutoff frequency of the system?
2. (a) Show that the standard deviation of a primary aberration is given by the
expressions listed in Table 3-4. (b) Also show, using Eq. (3-52), that a balanced
primary aberration is given by the expressions listed in Table 3-5.
4. If the beam in Problem 3 has 0.5 l of primary spherical aberration, (a) determine
the range of distance across which a Strehl ratio of 0.8 is obtained. (b) Determine
the axial point about which the axial irradiance is symmetric. (c) Give approximate
maximum and minimum values of the power contained in a circular detector whose
radius is equal to that of the central bright spot in the focal plane as it is moved
along the axis in the range calculated in (a).
5. Determine the location of the point about which the axial irradiance of a beam
aberrated by 2 l of secondary spherical aberration W6r6 is symmetric.
6. Determine the location of and irradiance value at the centroid and peak of the focal
plane distribution if the beam in Problem 3 is aberrated by l 4 of primary coma.
CHAPTER 4
341
342 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
If the Gaussian function is very narrow (i.e., its standard deviation is very small)
compared to the radius of the pupil, it is said to be weakly truncated. In essence, the pupil
343
344 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
can be assumed to be also infinitely wide with the result that the diffraction rings
disappear and a Gaussian beam remains Gaussian as it propagates. A focused beam has
the smallest radius, called its waist, in a plane that is closer to the pupil than the focal
plane. However, maximum central irradiance on a target at a fixed distance is still
obtained when the beam is focused on it. Considering the waist of a beam incident on a
lens as an object, an imaging equation is developed in which the waist of the transmitted
beam acts as the image. When the waist of the incident beam lies in the front focal plane,
the waist of the transmitted beam lies in its back focal plane, and not at infinity as in
conventional imaging. Similarly, whereas in conventional imaging of a real object by a
positive lens forming a real image, there is a minimum separation between an object and
its image, there is no minimum separation between a real object waist and a real image
waist. A beam quality factor M 2 is defined that accounts for the difference in the
divergence of a practical beam from that of an ideal Gaussian beam.
(r ) ( ) [ (r )]
P rp = A rp exp i F rp , (4-1)
where
2
( ) (
A rp = A0 exp È- rp w ˘ ,
ÎÍ ˙˚) (4-2)
r
with A0 as a constant describes the Gaussian amplitude at the exit pupil and rp = rp .
The quantity w, called the Gaussian radius, represents the radial distance from the center
of the pupil at which the amplitude drops to e 1 of the amplitude at the center. The
r r
aberration function F( rp ) represents the phase aberration at a point rp in the plane of the
exit pupil. The amplitude at its center is determined from the total power according to
4.2 General Theory 345
r r
Ú A (rp ) d rp
2
Pex = . (4-3)
For an aberration-free system, the wavefront at its exit pupil is spherical passing
through its center and with a center of curvature at a distance R from it. This distance
represents the distance between the planes of the exit pupil and the Gaussian image in the
case of an imaging system. In the case of a laser transmitter, it represents the distance of
the beam focus from the plane of the exit pupil. The following discussion and the results
are equally applicable to the propagation of a laser beam, as in a laser transmitter, and to
an imaging system with a Gaussian pupil.
The irradiance distribution of the diffracted beam normalized by its total power in a
plane at a distance z is given by Eq. (1-62):
r 1 Û r Ê 2 pi r r ˆ r
2
Ii ( ri ; z ) =
l2 R 2 Ù
ı
( )
P rp ; z exp Á -
Ë lR ¯
◊
rp ri ˜ d rp , (4-4)
r
where ri is the position vector of a point in the focal plane with respect to the focal point,
(r ) (r )
P rp ; z = A rp exp iF rp ; z [ (r )] (4-5)
and
(r )
F rp ; z = F rp + (r ) p Ê1 1ˆ 2
-
l Ë z R¯
rp . (4-6)
It is assumed here that z ~ R so that their difference enters only through the defocus
aberration. In particular, the variation of the irradiance due to the inverse-square law
dependence is negligible.
The optical transfer function (OTF) of a system is given by the autocorrelation of its
pupil function [see Eq. (1-73b)]:
r r r r r r 2 r
t (v i ) = Û
ı
( ) (
Ù P rp P * rp - l R v i d rp ) Ú P (r )
p d rp . (4-7)
t (v i , f) = Pex1Ú Ú P( p + l Rv i 2 , q) P * ( p - l Rv i 2, q) dp dq , (4-8)
where ( p, q ) is a coordinate system whose axes are rotated by an angle f with respect to
( )
the x p , y p system and whose origin lies at a distance lRv i from that of the x p , y p ( )
system along the p axis, as illustrated in Figure 2-42. Thus, in the ( p, q ) coordinate
system, the centers of the two pupils are located at m (lR / 2) (v i , 0 ) . The pupil function
(
P( p, q ) is obtained from the pupil function P x p , y p by replacing x p with )
p cos f - q sin f and y p with p sin f + q cos f . Now we apply the above equations to
systems with circular and annular pupils.
346 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Gaussian beams that are very narrow so that their truncation by a lens is negligible,
are shown to remain Gaussian as they propagate. Such a beam also yields maximum
irradiance on a target when it is focused on the target, although its waist and therefore the
principal maximum of axial irradiance lie in a plane that is closer to the lens. It is shown
that an aberration-free beam may be considered weakly truncated when the beam radius
is half of that of the pupil. However, the radius of an aberrated beam must be smaller than
one third of the pupil radius in order that it may be treated as a weakly truncated beam.
Considering the waist of a beam incident on a lens as the object, an imaging equation is
developed in which the waist of the transmitted beam acts as the image. The aberration
tolerance for a weakly truncated beam is shown to be significantly larger due to its
narrowness. Unlike a uniform focused beam, the axial irradiance of a Gaussian focused
beam aberrated by spherical aberration is not symmetrical about any point. However,
their symmetry properties for astigmatism and coma are similar to those for a uniform
beam.
and
2
g = (a w ) , (4-10)
Substituting Eq. (4-11) into Eq. (4-3), we obtain the power transmitted by the pupil:
4.3 Systems with Circular Pupils 347
1
Pex = 2 A02 Sex Ú (
exp - 2gr 2 r dr ) (4-12a)
0
[
= A02 (Sex 2 g ) 1 - exp(- 2 g ) ] , (4-12b)
where Sex = p a 2 is the area of the exit pupil. Squaring both sides of Eq. (4-11) and
substituting for A02 from Eq. (4-12b), we may write the pupil irradiance distribution in
units of Pex Sex :
(
I (r) = 2 g exp - 2 g r2 ) [1 - exp (- 2 g )] . (4-13)
The parameter g characterizes the truncation of a Gaussian beam by the exit pupil. Its
square root represents the ratio a w of the pupil and beam radii. Large values of g
represent narrow Gaussian beams, and small values imply wide beams. As g Æ 0 ,
I (r) Æ 1; i.e., a Gaussian beam reduces to a uniform beam. The pupil irradiance
[
decreases from a maximum value of 2 g 1 - exp( - 2 g ) at its center to a minimum ]
[ ]
value of 2 g exp(2 g ) - 1 at its edge. Figure 4-1 shows the pupil amplitude (normalized
to unity at the center) and irradiance distributions for g = 1, 2, and 3. The amplitude
A(r) has a value of 1 e (or approximately 0.368) at r = 1, 1/2, and 1/3 for g = 1, 2,
and 3, respectively, compared to a value of unity at the center. The corresponding value
of irradiance at the center of the pupil is 2.31, 8.00, and 18.00, compared to a value of
unity for a uniform pupil. The corresponding irradiance at the edge of the pupil relative to
its value at the center is given by 1 e 2 = 0.1353 , 1 e 4 = 0.0183 , and 1 e 6 = 2.479 ¥ 10 3 .
If we let the upper limit of the integral in Eq. (4-12a) approach infinity, we obtain the
total power Pinc incident on the pupil plane:
•
Pinc = 2 A02 Sex Ú (
exp - 2gr 2 r dr )
0
A02 Sex
= . (4-14)
2g
2 Pinc
A02 = . (4-15)
pw 2
Hence, from Eqs. (4-12b) and (4-14), the fraction of the power transmitted by the pupil is
given by
1.0
g =1
0.8
0.6
A (r)
2
0.4 3
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
r
(a)
18
16
14
g =3
12
I (r)
10
8
2
6
4
1
2 0
0
0.0 0.2 0.4 0.6 0.8 1.0
r
(b)
Figure 4-1. Amplitude and irradiance distributions at the exit pupil for
g = a w = 1 , 2, and 3. (a) Amplitude. (b) Irradiance. The amplitude is normalized
to unity at the center, but the irradiance is in units of Pex Sex .
in Figure 4-2 labeled as beam. Its value is 86.47% when g = 1, 98.89% when g = 1.5,
99.97% when g = 2, and practically 100% when g = 3. Thus, the fractional power
exp( - 2g ) lying outside the pupil is negligible when g ≥ 2.
The power transmitted by the pupil is again given by Eq. (4-12b). Hence, the fractional
transmitted power is given by
4.3 Systems with Circular Pupils 349
0.8
;:: 0.6
Cii
c
~
0:
t 0.4
0.2
l-exp(- 2y)
~rQns = pex / Pine = 2y (4-18)
4.3.2.1 PSF
For simplicity of equations and numerical analysis, we use normalized quantities
and
(4-20)
where F = Rj2a is the focal ratio of the beam. Note that in Eq. (4-20), the irradiance is
normalized by the focal-point irradiance of a uniform beam of the same total power [see
Eq. (2-2a)].
350 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Substituting Eqs. (4-9), (4-19), and (4-20) into Eq. (4-4), the irradiance distribution
for an aberration-free system may be written
2
1 2p
I ( r; q i ; g ) = p 2
Ú Ú
0 0
[ ( )]
I (r) exp -pirr cos q p - q i r dr dqp . (4-21)
The amount of power in the focal plane contained in a circle of radius rc centered at the
focal point is given by
rc
Pi (rc ; g ) = 2 p Ú Ii (ri ; g ) ri dri . (4-23)
0
Substituting Eqs. (4-19), (4-20), and (4-22) into Eq. (4-23) and defining a normalized
power
we obtain
rc
(
P(rc ; g ) = p 2 2 )Ú I (r; g ) rdr , (4-25)
0
where rc is in units of l F.
Figure 4-3 shows the image-plane irradiance and encircled-power distributions for
g = 0 , 1, 2, and 3. For clarity, the irradiance distributions are also plotted on a
logarithmic scale. It is evident that the Gaussian illumination broadens the central disc
but reduces the power in the secondary rings. As g increases, the secondary maxima
reduce, which is the objective of apodization. For large values of g , the diffracted beam
is also Gaussian (see Section 4.3.7 for more details). At and near the focal point, a
uniform beam gives a higher irradiance than a Gaussian beam. Similarly, the encircled
power is greater for a uniform beam for small values of rc , but smaller for large values.
Of course, P(rc ) Æ 1 as rc Æ • , regardless of the value of g .
The positions of maxima and minima and the corresponding irradiance and
encircled-power values are given in Table 4-1 for g = 1. The corresponding results for a
uniform beam are given in parentheses. It is evident that the corresponding maxima and
minima for a Gaussian beam are located at higher values of r than those for a uniform
beam. Thus, the diffraction rings for a Gaussian pupil are larger. For example, the radius
4.3 Systems with Circu lar Pupil s 35 1
/
/
"
,.,.
0.6
.0'
i:L
2
-
t 0.4
,.
3
0 .2
0
10
-'- - ' ,
10 - 1
10 - 2
10 - 3
10 - 4 I
-
II
II
,' ,
10 - 5
t •I "II .,
10 - 6
, ,
II
i
I,'..
I '
.....
II ".
I I '.
10 - 7
10 - 8
10- 9
10 - 10
0 0.5 1.5 2 2.5 3 3.5 4 4 .5 5
-- r
Figure 4-3. PSF and encircled power for a Gaussian pupil with = 0, 1, 2 and 3. Ii
The irradiance is in units of P.x Sex / 'A 2 R2, and the encircled power is in units of p.x.
T and Tc are in units of 'AF. The irradiance distribution normalized to unity at the
center is also shown on a log scale to highlight the differences between the secondary
maxima of uniform and Gaussian beams.
352 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Max 0 0.924 0
(0) (1) (0)
Min 1.43 0 0.955
(1.22) (0) (0.838)
Max 1.79 0.0044 0.962
(1.64) (0.0175) (0.867)
Min 2.33 0 0.973
(2.23) (0) (0.910)
Max 2.76 0.0012 0.976
(2.68) (0.0042) (0.922)
Min 3.30 0 0.981
(3.24) (0) (0.938)
Max 3.76 0.0005 0.983
(3.70) (0.0016) (0.944)
Min 4.29 0 0.985
(4.24) (0) (0.952)
Max 4.75 0.0002 0.986
(4.71) (0.0008) (0.957)
of the central bright spot is 1.43 for the Gaussian beam compared to 1.22 for a uniform
beam. Moreover, whereas the principal maximum for a Gaussian beam is only slightly
lower (0.924 compared with 1), the secondary maxima are lower by a factor > 3
compared with the corresponding maxima for a uniform beam. Lower secondary maxima
is indeed the objective of apodization. At and near the focal point, a uniform beam gives a
higher irradiance than a Gaussian beam. Thus, Iu > Ig for r < 0.42 . For larger values of
r, Ig > Iu , except in the secondary rings, where again Iu > Ig . The encircled power
Pu >< Pg for rc <> 0.63 Of course, as rc Æ • , Pu Æ Pg Æ 1.
A detailed examination of how the Airy pattern evolves into a Gaussian distribution
shows, for example, that its first two dark rings merge into one at the position of the
second dark ring as g Æ 2.63 , or its first two secondary maxima merge into one at the
position of the second maximum as g Æ 5.60 . 4 Moreover, the radius of the central bright
disc increases discontinuously with g with steps at the aforementioned g values and
g = 8.64 and 11.71, its value being greater than 2g for g >> 1.
4.3 Systems with Circular Pupils 353
Figure 4-4 shows how the focal-point irradiance 1(0; y) varies with y. It has a
maximum value of unity for y = 0, i.e. , for uniform illumination, and decreases
monotonically as y increases. It demonstrates Eq. (1-192) in that, for a fixed total power
Pex in the pupil, any amplitude variations across it (but without any aberrations) reduce
the focal-point irradiance. Now, the increase in y can be due to an increase in a or a
decrease in co. If a is kept fixed and cois decreased, the value of the focal-point irradiance
for fi = 1,2, and 3is 0.924, 0.482, and 0.222, respectively, compared to its value of
unity for a uniform pupil. However, if co is kept fixed and a is increased, then the focal-
point irradiance for a Gaussian pupil increases as tanh( y /2), as may be seen by
multiplying the right-hand side of Eq . (4-26) by the normalizing factor ?"xsexi'J.} R2 .
This is also illustrated in Figure 4-4 showing that the increase is < 4% for ~ 2. The fi
normalizing factor in this case is 2n ~x co !,).} R2. For a uniform pupil, however, the
2
0.8
0.6
;::
g Fixed a
0.4
Fixed OJ
0.2
For a given total beam power Pinc incident on a pupil of fixed radius a, the
transmitted power Pex increases as 0) decreases, but the corresponding focal-point
irradiance decreases. Hence, for a given total power P;nc, there is an optimum value of 0)
that yields the maximum focal-point irradiance. 5 To determine this value, we write the
focal-point irradiance given by Eq. (4-26) in units of P;IlC ,,? R2 : Sex!
1(0; y) = [1- exp(- 2y)] tanh(y/ 2)/(Y/ 2)
dl(O ; y)
o , (4-29)
dy
we find that 1(0; y) is maximum and equal to 0.8145 when y = l.255 or 0) = 0.893a .
The corresponding irradiance at the edge of the pupil is 8.1 % and the transmitted power
P;rans is 91.87 %.
4.3.2.3 OTF
From Eq. (4-8) and Figure 2-42, the OTF for an aberration-free Gaussian pupil III
0.6
g'"
0.4
2 2.5 3
(
t (v ; g ) = a 2 Pex ) Ú Ú A( p + v , q) A( p - v , q) dp dq , 0£ v£1
1 v2 1 q2 v
(
8g exp -2 gv 2 ) Û Û
=
[
p 1 - exp( -2 g ) ] Ù dq
ı
Ù
ı
[ ( )]
exp -2 g p 2 + q 2 dp , (4-30)
0 0
For large values of g (e.g., g ≥ 4 ), the contribution to the integral in Eq. (4-30) is
negligible unless v = 0 , in which case it represents the Gaussian-weighted area of a
quadrant of the pupil, and the equation reduces to
(
t (v ; g ) = exp -2gv 2 ) , 0£v £1 . (4-31)
As shown in Section 4.3.7.6, this result may also be obtained by zero-order Hankel
transforming the PSF for z = R . The corresponding defocused OTF is also considered in
this section.
Figure 4-6 shows how the OTF varies with v for several values of g . We note that
compared to a uniform pupil (i.e., for g = 0 ), the OTF of a Gaussian pupil is higher for
low spatial frequencies, and lower for the high.6 Moreover, as g increases, the bandwidth
of low frequencies for which the OTF is higher decreases and the OTF at high
frequencies becomes increasingly smaller. This is due to the fact that the Gaussian
weighting across the overlap region of two pupils whose centers are separated by small
values of v is higher than that for large values of v. If we consider an apodization such
that the amplitude increases from the center toward the edge of the pupil, then the OTF is
lower for low frequencies and higher for the high. Thus unlike aberrations, which reduce
the MTF of a system at all frequencies within its passband [see Eq. (1-107)], the
amplitude variations can increase or decrease the MTF at any of those frequencies.
~ 1 - s F2 (4-32b)
~ exp ( - s F2 ) , (4-32c)
where
356
OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
\
\
0.8
\
\
\
\
\
\
0.6
\
'">
-;::
\ 0
t \
\
0.4
fY =~
\
2\
\
\
\
0.2 \
-+y
Figure 4-6. The OTF of a Gaussian pupil. A uniform pupil corresponds to 'Y = 0,
and a large value of 'Y represents a weakly truncated pupil.
(4-33)
is the variance of the phase aberration across the Gaussian-amplitude weighted pupil. The
mean and the mean square values of the aberration are obtained from the expression
fo 2J A(p)[<l>(p, 8)]"
0
PdPd8/ f YA(p)
0 0
pdpd8
I 21t
11:
[
1 - exp -
Y(
y
)]f f exp(-yp2)[<l>(p,8)f pdpd8 (4-34)
o 0
Table 4-2 lists the primary aberrations and their standard deviations for increasing
values of y. It is evident that the standard deviation of an aberration decreases as y
increases. This is due to the fact that while an aberration increases as p increases, the
amplitude decreases more and more rapidly as y increases, thus reducing its effect more
and more compared to that for a uniform pupil. Accordingly, for a given small amount of
aberration Ai' the Strehl ratio for a Gaussian pupil is higher than that for a uniform pupil.
Similarly, the aberration tolerance for a given Strehl ratio is higher for a Gaussian pupil.
Its approximate value can be obtained from Eq. (4-32c).
4.3 Systems with Circular Pupils 357
Table 4-2. Primary aberrations and their standard deviations for optical systems
with Gaussian pupils. For comparison, the results for a uniform pupil ( g = 0 ) are
also given.
Seidel Aberration sF ( g = 0) sF ( g = 1) sF ( g =2 ) sF ( g ≥3 )
Spherical, As r 4 2 As As As As 2 5 As
=
3 5 3 .35 3.67 6.20 g2
Coma, Ac r3 cos q Ac Ac Ac Ac 3 Ac
=
2 2 2.83 3.33 6.08 g3 2
Astigmatism, Aa r2 cos 2 q Aa Aa Aa Aa
4 4.40 6.59 2g
Defocus, Bd r2 Bd Bd Bd Bd Bd
=
2 3 3.46 3.55 4.79 g
Tilt, Bt r cos q Bt Bt Bt Bt
2 2.19 2.94 2g
Balanced F( r, q ; g = 0) F( r, q ; g = 1) (
F r, q ; g = 2 ) (
F r, q ; g ≥ 3 )
Aberration
Ê 4 2ˆ
Spherical (
As r 4 r2 ) (
As r 4 0.933r 2 ) (
As r 4 0.728 r 2 ) As Á r 4
Ë
r ˜
g ¯
Ê 2 ˆ Ê 3 2 ˆ
Coma Ac Á r 3
Ë
r˜ cos q
3 ¯
(
Ac r 3 )
0.608 r cos q A c r 3 ( )
0.419 r cos q A c Á r
Ë
r˜ cos q
g ¯
Astigmatism
(
A a r 2 cos 2 q 12 ) (
A a r 2 cos 2 q 12 ) (
A a r 2 cos 2 q 12 ) (
A a r 2 cos 2 q 12 )
Balanced sF ( g = 0) sF ( g = 1) sF ( g =2 ) sF ( g ≥3 )
Aberration
Spherical As As As As 2 As
=
6 5 13.42 13.71 18.29 g2
Coma Ac Ac Ac Ac Ac
=
6 2 8.49 8.80 12.21 g3 2
Astigmatism Aa Aa Aa Aa Aa
=
2 6 4.90 5.61 9.08 2g
Table 4-5. Factor by which the standard deviation of a Seidel aberration across an
aperture is reduced when it is optimally balanced with other aberrations.
Reduction Factor
Balanced Uniform Gaussian Gaussian Weakly Truncated
Aberration ( g = 0) ( g = 1) ( g =2 ) Gaussian g ≥ 3 ( )
Spherical 4 3.74 2.95 5 = 2.24
Diffraction Focus
Balanced Uniform Gaussian Gaussian Weakly Truncated
Aberration ( g = 0) ( g = 1) ( g =2 ) (
Gaussian g ≥ 3 )
Ê 32 2 ˆ
Spherical (0, 0, 8F A ) (0, 0, 7.46 F A ) (0, 0, 5.82 F A )
2
s
2
s
2
s Á 0, 0, F As ˜
Ë g ¯
Astigmatism (0 , 0 , 4 F A ) (0 , 0 , 4 F A )
2
a
2
a (0 , 0 , 4 F A )
2
a (0 , 0 , 4 F A )
2
a
2 2
1 2p È1 2p
˘
S = Ú Ú A(r) exp[i F(r, q)] r dr dq Í Ú Ú
ÍÎ 0
A(r) r dr dq ˙
˙˚
0 0 0
2 1 2p 2
ÏÔ g ¸Ô
= Ì
p 1[- exp(- g ) ] Ú0 Ú0
˝ exp - g r ( 2
) exp[i F(r, q)] r dr dq , (4-35)
ÓÔ ˛Ô
2 2
1
È g ˘
S = Í ˙ Ú exp( - g x ) f ( x ) dx , (4-36)
Î1 - exp ( - g ) ˚ 0
where
ÔÔ s[(
Ï exp i A x 2 + B x
d )]
Spherical + defocus
f ( x) (
= Ì J 0 A c x 3 2 + B t x1 2 )
Coma + tilt (4-37)
Ô
[ ]
ÔÓ exp i (0.5Aa + Bd ) x J 0 (0.5Aa x ) Astigmatism + defocus .
The coefficient of a balancing aberration for different values of g is given in Table 4-3.
For example, the defocus aberration that balances spherical aberration is given by
Bd As = - 1, - 0.933 , and - 4 g when g = 0 , 1, and ≥ 3, respectively. Similarly, the
tilt aberration that balances coma for these values of g is given by Bt Ac = - (2 3) ,
360 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Figure 4-7 shows how the Strehl ratio varies with spherical aberration As with and
without balancing with defocus. We note from Figures 4-7a and 4-7b that the balanced
aberration yields a much higher Strehl ratio for small values of g but As < ~ 3 l . Thus,
4
when spherical aberration As r is present in a system, the central irradiance in a
defocused observation plane yielding minimum aberration variance is higher than that at
the Gaussian image point or the beam focus. For a weakly truncated Gaussian beam,
however, the improvement in Strehl ratio due to balancing is not as significant, as
illustrated in Figure 4-7c for g = 3. Moreover, in this case, the balancing of spherical
aberration reduces the Strehl ratio when As >
~ 8.3l . Of course, the value of Bd defining
the defocused plane for minimum aberration variance depends on the value of g .
Although for moderate values of As , the Strehl ratio for balanced spherical
aberration is greater than its corresponding value without balancing with defocus, a larger
value is obtained in a plane that is different from the one that yields minimum variance
when the aberration is large. Figure 4-8a shows how the Strehl ratio of a uniform beam
for various values of balanced spherical aberration varies as a function of the defocus
aberration d. It should be evident that d represents the deviation of defocus aberration
from its value required for minimum variance of spherical aberration. We note that the
Strehl ratio is symmetrical about d = 0 , as expected.10 However, when As ≥ 2.3 l ,
minimum variance does not yield maximum Strehl ratio. For example, when As = 3 l,
the maximum central irradiance is obtained in a plane corresponding to d = ± 1 l , or
Bd = - 2 l or - 4 l . In fact, a minimum of irradiance is obtained in the plane d = 0 , or
Bd = - 3 l , corresponding to minimum variance. A similar result is obtained for a
Gaussian beam with g = 1, as illustrated in Figure 4-8b, except that the Strehl ratio is not
symmetric about d = 0 . For a weakly truncated Gaussian beam with g = 3, maximum
Strehl ratio is obtained for d > 0 even for As = 2 l , as shown in Figure 4-8c. The peak
shifts to larger values of d as As increases.
For a uniform pupil, balanced coma yields a higher Strehl ratio than the Seidel coma
only for Ac £ 2.3 l , as illustrated in Figure 4-9. For larger aberrations, Seidel coma
yields a higher Strehl ratio than balanced coma, i.e., the irradiance at the Gaussian image
point is higher than at the point with respect to which the aberration variance is minimum.
Similar results are obtained for a Gaussian pupil, although the cross-over point occurs for
larger values of Ac as g increases. For g = 3, the cross-over occurs at Ac = 7 l . For
larger aberrations, the difference between the Strehl ratios for balanced coma and Seidel
coma increases as g increases.
Figure 4-10 shows the Strehl ratio for astigmatism. The increase in Strehl ratio by
balancing astigmatism with defocus is not as significant as in the case of balanced
spherical aberration or balanced coma. As in the case of coma, balanced astigmatism
yields a higher Strehl ratio than Seidel astigmatism for Aa < 1 l for small values of g .
For a weakly truncated Gaussian beam with g = 3, the cross-over point is Aa ~ 4 l .
4.3 Systems with Circular Pupils 361
(/)
(a)
Y= 1
,,
,
',Bd =- 0 .933 As
(/) \
t ,
,,
,
,, (b)
0.2 "
0.8
0 .6
(/)
t
0.4
(c)
0.2
Y= 0
0.8
"
,'1\ \
0.6
en
t
0.4
(a)
0.2
y =l
0.8
0.6
en
t
0.4
(b)
0.2
0
- 4 - 3
0 .8
0.6
en
t
0.4
(C)
0 .2
0
- 4 - 3 -2 - 1 0
- 8
Figure 4-8. Strehl ratio for balanced spherical aberration as a function of defocus
aberration a, which represents the deviation of defocus aberration from its value for
optimum balancing of spherical aberration for minimum variance.
4. 3 Systems with Circular Pupil s 363
,,
,, .fi=3
0.8 "- ,
,,
,,
, B, = - (2/9) Ao
0.6 ,,
<f)
,
t Bt = a
0.4
(C)
0.2
0
0 10
-Ac
Figure 4-9. Strehl ratio of a beam aberrated by coma Ac in units of wavelength A..
(a) Uniform beam (1=0). (b) Gaussian beam with 1=1. (c) Weakly truncated
Gaussian beam with .fi
= 3. BI represents the coefficient of the balancing tilt
aberration.
364 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Y= a
0.8
0.6
(J)
t
0.4
(a)
0.2
,
,,
a
a
Y= 1
(J)
(b)
0.2
\ Sd= a
a
a
-Aa
,, .fi= 3
0.8
" ,,
,,
,
' , Bd= - Aa/2
0.6 ,,
(J) Bd=O ", ,
t
0.4
(c)
0.2
a
a
-Aa
Since for small aberrations, the Strehl ratio depends on the aberration variance
regardless of the type of the aberration, it is useful to determine the range of validity of
the approximate Eq. (4-32c). It has been shown in Section 2.3.4 that for a uniform pupil
( )
the approximate expression exp - s 2F for the Strehl ratio yields a value that is within
10% of its true value as long it is ≥ 0.3 . 11 For small values of g , the approximate
expression estimates the true value of Strehl ratio quite well for s w < ~ l 4 , as may be
seen from Figures 4-11a through 4-11d. However, Figures 4-12a and 4-12b show that
unless the Strehl ratio is ≥ 0.9 or s w < ( 2
)
~ l 20 , exp - s F significantly underestimates
the true value of the Strehl ratio. Moreover, the Strehl ratios for a given value of s w of
the various aberrations are quite different from each other. In particular, the Strehl ratio
for balanced spherical aberration or balanced astigmatism is significantly smaller than its
corresponding value without balancing. For a uniform pupil, however, the Strehl ratio for
a given value of s w is the same for both spherical and balanced spherical aberrations. Of
course, the aberration coefficient As for a given value of s w is quite different for the
aberration with and without balancing.
From Tables 4-2 and 4-4 we note that, for a weakly truncated beam, the balancing of
a primary aberration reduces its standard deviation by a factor of 5 , 3 , and 2 in
the case of spherical aberration, coma, and astigmatism, respectively. These reduction
factors are listed in Table 4-5 for the uniform ( g = 0) , e 2 truncated ( g = 1) and the
weakly truncated (large g) Gaussian beams. The reduction factors decrease in the case of
spherical aberration and coma, but increase in the case of astigmatism. Table 4-7 gives
the reduction factors that relate the peak value Ai of a primary aberration at the edge of a
circular aperture and the standard deviation of its corresponding balanced aberration for
various values of g .
Table 4-7. Reduction factor for primary aberrations by which the peak aberration
coefficient Ai must be divided to obtain the standard deviation for a Gaussian
circular beam with various values of g.
0.6
({)
t
0.4
(a)
0.2 "~.. ~~- . - . - -_._. _.- .-
.....
~--'-""---
-..,..._..c;-~ ________ _
°0~--L-~
0 .-1-~--0~.2
~-~_~OL.3~~~~-OL
.4_~~~0. 5
y = 0, Balanced aberration
- - spherical
- - - coma
0.8 , - - . astigmatism
, ..... . exp (- ,, ~ )
0.6
({)
t
0.4
(b)
0.2 .~
·X:.
,'.
....-::.:--:.-'-::.-.::-'::: ...-:-
,::",-,... .- . - - - '- '
O ~--L-~--~-~--~-~~~~~-L-~-~
o 0.1 0.2 0.3 0.4 0.5
Figure 4-11. Strehl ratio S of a primary aberration with and without balancing as a
function of its standard deviation Ow in units of wavelength A. (a) Uniform beam
('1 = 0) with a primary aberration. (b) Uniform beam with a balanced primary
aberration. (Cont.)
It is evident that the reduction factor increases as y increases. Accordingly, while the
sigma value of an aberration for a certain Strehl ratio may be small, its peak value may
not be. This is why the expression exp (- O'~) does not give a good estimate of the true
4.3 Systems with Circul ar Pupils 367
0.6
(fJ
t
0.4
(c)
0.2
0.6
UJ
t
0.4
(d)
0.2
' ... ~
°0L---~---0~.-1---L--~0~.2----L---~
0 .-
3 ~~·~-~·~---0~.4----L---~
0.5
Figure 4-11. (Cont.) Strehl ratio S of a primary aberration with and without
balancing as a function of its standard deviation O'w in units of wavelength A. (c)
Gaussian beam with 'Y = 1 and a primary aberration. (d) Gaussian beam with 'Y = 1
and a balanced primary aberration.
Strehl ratio in Figures 4-12a and 4-12b, except for very small aberrations. Similarly, since
the reduction factor is larger for spherical aberration than that for coma, which itself is
larger than that for astigmatism, the Strehl ratio estimate is also poorer on the same order.
368 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
0.8
0.6
(J)
t
0.4
~, ................ _c. .
2 )"
exp (- a W ".
... ~ ,.....
, . ...
--- _
'-'-A'-·"",,-._-._- ._.:~_
••••• 0><5' " .... .... . ... 8A - .- . (a)
0.2 " ,, .... ....,..,... .
"" -'-
......82
".
o L-~--~--~--~--~---L~~~~--~~
o 0.1 0.2 0.3 0.4 0.5
0.8
0.6
(J)
t
0.4
....
... ·-A· ..... ·
-.-.
.... - - - -
0.2
exp (-
2 ".
a W ) ".
. . ..
-'-.-'-.....--'- (b)
'.
where cnm are the expansion coefficients, n and m are positive integers including zero, n –
m ≥ 0 and even, and
[
Z nm (r, q; g ) = 2( n + 1) (1 + d m 0 ) ]1/ 2Rnm (r; g ) cos mq . (4-39)
1 2p 1
m¢
Ú Ú Z n (r,q; g )Z n ¢ (r,q; g )A(r ) r dr d q 2p Ú A(r) r dr = d nn ¢ d mm ¢
m
, (4-40)
0 0 0
È (n m) 2 ˘
Rnm (r; g ) = M nm Í Rnm (r) - Â (n - 2i + 1) < Rnm (r) Rnm 2i (r; g ) > Rn 2i (r; g )˙˙ . (4-41)
ÍÎ i ≥1 ˚
The angular brackets indicate an average over the Gaussian pupil; i.e.,
1 1
< Rnm (r) Rn 2 i (r; g ) > = Ú Rnm (r) Rn 2 i (r; g ) A(r) r dr Ú A(r) r dr . (4-42)
0 0
The normalization constant Mnm is chosen such that the radial polynomials obey the
orthogonality relation
1 1
1
Ú Rnm (r; g ) Rnm¢ (r; g ) A(r) r dr Ú A(r) r dr = d . (4-43)
n + 1 nn ¢
0 0
Note that except for the normalization constant, the radial polynomial Rnn (r; g ) is
identical to the corresponding polynomial for a uniformly illuminated circular pupil
Rnn (r) ; i.e.,
where the coefficients anm , etc., depend on g. The number of Zernike (or orthogonal)
aberration terms in the expansion of an aberration function through a certain order n is
given by Eqs. (2-63a) and (2-63b), as in the case of circle polynomials. The Zernike-
Gauss expansion coefficients are given by
1 2p 1
c nm = Ú Ú F(r, q; g )Z nm (r, q; g ) A(r) r dr d q 2 p Ú A(r) r dr , (4-46)
0 0 0
as may be seen by substituting Eq. (4-38) and utilizing the othonormality of the
polynomials.
1 2p 1
n ( r, q; g ) = Ú Ú F n ( r, q; g ) A ( r ) r dr d q 2 p Ú A ( r ) r dr
Fm m
0 0 0
= 0 , n π 0, m π 0 . (4-48)
For m = 0 , this may be seen with the help of Eq. (4-43) and the fact that R00 (r; g ) = 1 is a
member of the polynomial set. The orthogonality Eq. (4-43) yields the result that the
mean value of Rn0 (r; g ) is zero. When m π 0 , the average value of cos mq is zero.
Similarly, the mean square value of the aberration is given by
2 1 2p 2 1
[F m
n (r, q; g )] = Ú Ú
0 0
[F m
n (r, q; g )] A(r) r dr d q 2 p Ú A(r) r dr
0
2
= cnm . (4-49)
m 2 2
s 2nm = (F )n - F nm
2
= c nm , n π 0, m π 0 . (4-50)
Thus, each expansion coefficient, with the exception of c00 , represents the standard
deviation of the corresponding aberration term.
4.3 Systems with Circular Pupils 371
From Eqs. (4-47) and (4-50), we note that the standard deviation of an aberration can
be obtained immediately by comparing its form with the corresponding orthonormal
aberration represented by Eq. (4-47), without having to calculate the integrals in Eqs. (4-
45) and (4-48). The variance of an aberration consisting of two or more terms of the form
of Eq. (4-47) is given by the sum of the variances of each of the aberration terms. The
variance of the aberration function is accordingly given by
2 • n
s F2 = F 2 (r, q; g ) - F(r, q; g ) = Â 2
 c nm . (4-51)
n =1 m = 0
Piston R00 1 1 1 1
2
Field curvature R20 a20r2 + b20 2
2.04989r – 0.85690 2r – 1 ( gr2 – 1) / 3
(defocus)
Astigmatism R22 a22r2 1.14541r2 r2 ( g / 6 )r2
3
Coma R31 a31r3 + b31r 3.11213r 3 – 1.89152r 3r – 2 r Êg ˆ
g / 2 Á r3 – r˜
Ë2 ¯
Spherical aberration R40 a40r4 + b40r2 + c40 6.12902r4 – 5.71948r2 + 0.83368 6 r4 – 6 r2 + 1 ( g 2r4 – 4 gr2 + 2) / 2 5
1
*a11 = (2 p2 ) –1/2 , a 20 = [3( p4 – p22 )] –1/2, b20 = – p2 a 20 , a 22 = (3 p4 ) –1/2 , a 13 = ( p – p42 / p2 ), b31 = – ( p4 / p2 ) a 13 ,
2 6
–1/2
{
a 40 = 5 [ p8 – 2 K 1 p6 + (K 12 + 2 K 2 ) p4 – 2 K 1 K 2 p2 + K 22 ] } , b40 = – K 1 a 40 , c40 = K 2 a 40 ,
p 0 = 1, K1 = ( p6 – p 2 p 4 ) / ( p 4 – p 22 ), K 2 = ( p 2 p6 – p 42 ) / ( p 4 – p 22 ) .
372 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
[
P(r) = A0 exp - ( g - iBd ) r2 ] , (4-52)
where
R
Bd ( z ) = p N Ê - 1ˆ (4-53)
Ëz ¯
is the coefficient of defocus phase aberration representing its peak value. Here
N = a2 lR (4-54)
is the Fresnel number representing the number of Fresnel zones in the pupil as observed
from the focus. The defocus aberration is generally not symmetrical about the Gaussian
image plane. For small values of N (e.g., £ 10), z can be much different from R for Bd to
achieve a significant value. Hence, the defocus aberration for two image planes located
symmetrically about the Gaussian image plane has different magnitude as well as
different sign. However, when the Fresnel number is very large (>> 10), Bd becomes
large even for very small differences between z and R. Accordingly, the small defocus
aberration tolerance dictates that z be approximately equal to R. Hence, Eq. (4-53) may be
written
z
Bd ( z ) = p N Ê1 - ˆ , (4-55)
Ë R¯
p ND
Bd ( R + D ) = - . (4-56)
R
Therefore, the defocus aberrations for two image planes located symmetrically about the
Gaussian image plane have the same magnitude but opposite signs. The longitudinal
defocus may be written in terms of the focal ratio F = R 2 a of the image-forming light
cone as
D = - 8 Bd F 2 , (4-57)
where Bd is now the coefficient of wave aberration [obtained by multiplying the phase
aberration coefficient Bd in Eq. (4-55) by l 2 p ]. An observation made in a plane beyond
the focal plane so that z > R introduces a negative defocus aberration, and one closer to
the pupil so that z < R introduces a positive aberration, as may be seen from Eq. (4-55).
4.3 Systems with Circular Pupils 373
4.3.4.2 PSF
Substituting Eq. (4-48) into Eq. (4-4) and replacing R by z, we obtain the PSF in a
defocused image plane. The irradiance distribution in a defocused image plane at a
distance z from the plane of the exit pupil may be written
1 2
2
Ê 2R ˆ
I (r; z; g ) = Á
Ë z ¯
˜ Ú ( )
I (r) exp iBd r J 0 (prr ) r dr
2
, (4-58)
0
where the irradiance is in units of Pex Sex l2 R 2 , as in Eq. (4-21), but r is in units of
lz D. Equation (4-58) represents the modified form of the defocus-free Eq. (4-22) due
to an observation in a defocused plane. The modification consists of the inverse-square
law dependence on the distance z and the defocus aberration Bd r2 . The encircled power
is obtained by substituting Eq. (4-58) into Eq. (4-25).
From Eq. (4-58), we note that the irradiance distribution is independent of the sign
of Bd . However, it is asymmetric about the Gaussian image plane; i.e., the irradiance
distributions in two observation planes located at z = R ± D are not identical. There are
three reasons for this asymmetry.13,14 First, the inverse square law dependence on z
increases I (r ) for z < R and decreases it for z > R . Second, Bd is asymmetric. Third, the
◊
argument of the Bessel function J 0 ( ) depends on z. For small Fresnel numbers, all three
factors contribute to the asymmetry. The defocus aberration reduces the central
irradiance, but the inverse-square dependence on z increases it for z < R . Since the
smaller the Fresnel number, the larger the depth of focus, the irradiance on and near the
axis can be significantly higher for z < R than the focal-point irradiance. This is
illustrated by considering the axial irradiance of a beam with different Fresnel numbers.
However, when the Fresnel number is very large, the depth of focus is very small and the
2
effect of inverse-square dependence on z is negligible. Thus, the ( R z ) factor in Eq. (4-
58) can be replaced by unity. Moreover, since Bd only changes sign when going from
one symmetric plane to the other, the irradiance distribution (which is independent of the
sign of Bd ) is symmetric about the Gaussian image plane.
If we let r = 0 and note that J 0 (0) = 1 , we obtain a different form of the expression for
axial irradiance, namely,
1 1
I (0; z; g ) = (2 R z )
2
Ú Ú [ I (r) I (s)]
0 0
12
[ ( )]
cos Bd r2 - s 2 rs dr ds . (4-60)
1 1
P(rc ; z; g ) = 2 p 2
Ú Ú [ I (r) I (s)]
0 0
12
[ ( )]
cos Bd r2 - s 2 Q(r, s; rc ) rs dr ds , (4-61)
where Q(r, s; rc ) is given by Eqs. (2-108a) and (2-108b). The integrals in Eqs. (4-59)
through (4-61) can be evaluated by the Gauss quadrature method, as discussed in
Section 2.6.
For large Fresnel numbers, Figure 4-13 shows the defocused irradiance and encircled
power distributions for a defocus of 0.5, 1, 1.5, and 2 waves. The corresponding
longitudinal defocus is obtained by multiplying the wave aberration by 8 F 2 . For clarity,
the irradiance values have been multiplied by five in Figures 4-13b through 4-13d. For
small values of defocus, the central irradiance is higher for the uniform beam, but the
encircled power is higher for the Gaussian (except for very small circles). However, its
value in the case of a uniform beam is zero for an integral number of waves of defocus.
The cancellation of contributions of the Fresnel zones is complete. This is not true for a
Gaussian pupil, since the amplitudes of the various zones are different. For one wave of
defocus, the PSFs for g = 2 and 3 are practically the same for r ≥ 0.8 . It is shown in
Section 4.3.4.4 that the corresponding OTFs are practically equal to each other for any
spatial frequency. Defocused PSFs in the form of isophotes (i.e., contours of constant
irradiance) have been given for g = 0, 1, 2, 3, and • . 15 Since a given value of z
corresponds to a certain value of Bd according to Eq. (4-53), the PSFs given in Figure 4-
1 can be applied to systems with small Fresnel numbers provided they are multiplied by
( R z)2 . Of course, the units of r also change from lR D to lz D. Results for such
systems have been observed experimentally for N ~ 0.5 and 1, g ~ 0.5 and 1, and various
values of z,16 and for g = 0 and N ~ 0.5 , 1, and 5.17
0.8
I
. .
." I
,
FY=?!<:-</
/ .:" I 0
(a)
0 .8
1 .. '
~ 0 .6
iL
0.4
(b)
0 .2 '.,
'"''
o ~~~~~~--~----~--~----~==~=-~~
a 1.5 2 .5 3.5
0.8
_ 0.6
~u
iL
1 ..···
0 .4 -
(c)
0.2
0.8
/
/
/
"3' 0.6 /
/
il:"
,,/ ./ /
,,/ //
/
0.4 /
/
/
....
1.. . · ··
........ / ","" .. ...
(d)
/r .... .../'
0.2 ._ . _._ . - . _ . _ . _._/ 0(. . _ ......_/ ..........
..'
/ /
/ ./
/ /
" .,; ... ... :: ~ . ,::.
o ~~~~~~~~==~======~~~
o 0.5 1.5 2 2.5 3 3.5 4
0.500
0.5 1.0
Pg .,"" , -- ---
~.-'-
0.4 .- 0.8
.- .-
""
-
i
0.3 0.6 t
~u
0.2 0.4
0.1 0.2
It goes through a series of maxima and minima as a function of z because of the cos Bd
term. By equating its derivative with respect to z to zero, we obtain the positions of its
maxima and minima as the solutions of
Ê lz B ˆ
2Á - 2 d 2 ˜ (cosh g - cos Bd ) = - sin Bd . (4-63)
Ë Sex Bd + g ¯
They occur at approximately those z values at which the pupil subtends an odd or an even
number of Fresnel zones, respectively. Since cosh g > 1 and cos Bd £ 1, the minima are
not equal to zero [unless g = 0 , as shown by Eq. (2-84) for a uniform beam].
Figure 4-15 shows how the axial irradiance of a focused Gaussian beam with g = 1
differs from that of a corresponding uniform beam when the Fresnel number N = 1, 10,
and 100.14 The numbers 0.924, 0.482, and 0.222 on the curves in Figure 4-15a for N = 1
represent the irradiance value at the focal point relative to a value of unity for a uniform
pupil. We note that the principal maximum is higher for the uniform beam compared with
that for the Gaussian beam. It lies closer and closer to the pupil as g increases. However,
the secondary maxima are higher for the Gaussian beam. Moreover, the axial minima for
the uniform beam have a value of zero at z values given by Bd = 2p n or
1
z R = (1 + 2 n N ) , where n is a nonzero positive or a negative integer. For example,
when N = 1, the minima are located at z R = 1 3, 1 5, 1 7, K and there is no minimum
for z < R . Qualitatively, as stated in Section 4.3.4.2, the even number of Fresnel zones
that the pupil subtends at these locations cancel out each other to yield a zero irradiance.
Since these zones have different weights in the case of a Gaussian pupil, the cancellation
is not complete and the irradiance is a minimum but not zero. We note from Figures 4-
15b and 4-15c that the curves become symmetric about the focal point z = R as N
increases, regardless of the value of g , as expected from the discussion in Section
4.3.4.2.
2.5
N=l
2 .046
,-...
I '
\
1.824 \
\
\
\
1.5 \
N
g ".
0 .222
0
0 0.5 1.5
- z/ R
1.4
N = 10
1.2
N O.8
t 0.6
0.4 (b)
0.2
0 ~~~~L-~~~
--- __L-~~~~~~~~
0.5 0.6 0.7 1.3 1.4 1.5
- z /R
N = 100
0.8
0.6
N
g
0.4
(c)
I _ _ \
0.2 i ... 3 "' .... '\
0
0.9
7 ~ =~~-~:~~~;<.J \'~>-~--==~ -_~--
-
Figure 4-15. Axial irradiance of a beam focused at a fixed distance R with a Fresnel
number N = a 2 /AR = 1, 10, 100. The irradiance is in units of the focal-point
irradiance PexSex /A2
R2 for a uniform beam.
4.3 Systems with Circular Pupils 379
2
È g ˘ 1
= Í
1 - exp ( )˚
- g
˙ 2
g + Bd2
[
1 - exp( - 2 g ) - 2 exp( - g ) cos Bd ] . (4-64)
Î
1
S = 2
. (4-65)
1 + (Bd g )
Figure 4-17 shows how the Strehl ratio varies with the defocus aberration Bd in units
of wavelength l for various values of g . It is evident that for a given value of S, the
defocus tolerance increases as g increases. Moreover for a Gaussian pupil, the Strehl
ratio when Bd is an integral number of wavelengths is not zero, as it is for a uniform
pupil (see Section 2.5.2 also). For a certain Strehl ratio, the value of Bd can be obtained
from Figure 4-17, and the corresponding depth of focus can be obtained from Eq. (4-57).
For example, a Strehl ratio of 0.8 is obtained for Bd = l 4 when g < ~ 1. Hence, the depth
2 2
of focus is approximately ± 2 l F ; however, its value is ± 5l F when g = 3.
10
-, , N=1
/
/
,,
\
\
\
\
,
\ ..fi=3
N
6 ,
\
g
\
,,
4
,,
, ,,
, (a)
,, '- 2.- -- '- '-
,.r/ ' -
0
0 0.5 1.5
-ziR
1.4
N = 10
1.2
I
/
/
/
/
/
N 0.8 /
g fi=3'
/
/
/
0.6 /
/
/
/
I
0.4
I / (b)
/
I
I
/
I
0.2
I
0
0.5 0.6 0.7 0.8 0.9 1.1 1.2 13 1.4 15
- ziR
';("\' , N = 100
,/ i I
."
I
I
,, ;
I I I
I I
0.8
I I
I I I
I
0.6 , I
I
I
I
N , I
I
I
g .f'Y=2, I
0.4 , \
,, 2;
; ,,
(C)
/
I
,,
0.2
/ /
\
,,
I / ,
\~,--
I
/
Figure 4-16. Axial irradiance of a focused beam normalized to unity at the focal
point.
4.3 Systems with Circular Pupils 381
0.8
0.6
(/) . 3
t
0.4 .......
'.
,
', 2 ' .. ........
" .
\ '. . ...... .
0.2 \ 1.5 ' .
\
0
0 0.5 2.5
Figure 4-17. Strehl ratio of a Gaussian beam as a function of the defocus wave
aberration coefficient Bd in units of wavelength A, showing how it increases as '1
increases. For a certain value of the Strehl ratio, the value of Bd is obtained from
this figure which, in turn, is used to obtain the corresponding value of the depth of
focus from Eq. (4-57).
the derivative of the axial irradiance given by Eq. (4-59) with respect to Rand
determining the value of R that yields the maximum irradiance. When doing so, the
normalization factor ?"xSex j').? R2 should be substituted in this equation with the
consequence that the R2 factor in front of its right-hand side disappears. We find that R
must equal z. Since the observation distance z is fixed, the inverse-square law dependence
of irradiance on the distance is also fixed. Hence, focusing the beam at any distance other
than z introduces defocus aberration and reduces the central irradiance. Figure 4-18
illustrates this point, where the central irradiance Iz(O; R) on a target at a fixed distance z
is shown as a function of the focal distance R . The irradiance is in units of ?"xSexlA? z2 ,
and Nz = a 2 /AZ represents the Fresnel number of the pupil as observed from the target.
The peak irradiance for the Gaussian beam with y = 1 is 0.924 compared to 1 for a
uniform beam. Moreover, as in Figure 4-15 or 4-16, the curves become symmetric about
the axial pointR = z as N z increases. Thus , regardless of the value of the Fresnel
number, the focal point is indeed the diffraction focus , even though a higher central
irradiance is obtained for small Fresnel numbers at axial points closer to the pupil. This is
different from, for example, the case of a beam aberrated by a small amount of spherical
aberration, in which case an appropriate amount of defocus improves the central
irradiance (see Section 4.3.5). In other words, defocusing the beam in such a case yields a
hjgher central irradiance, or the corresponding defocused image is better. Hence, the
defocused point (whose location depends on the amount of spherical aberration and the
type of pupil illumination, e.g., uniform or Gaussian) is the diffraction focus. Although
we can not improve the central irradiance on a target by defocusing the beam, the fact
that the principal maximum of axial irradiance lies at a point other than the focal point
382 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
implies a large depth of focus. For example, for a uniform beam with N = 1, as in Figure
4-15a, the axial irradiance is larger than the focal-point irradiance for 0.435 < z R < 1 . As
discussed earlier (see Figure 4-17), defocus tolerance or depth of focus for a Gaussian
beam is even higher.
For a focused Gaussian beam, the location of the plane of minimum field spread
(i.e., the spot radius at which the irradiance is 1 e 2 of its value at the center) or minimum
root mean square spot radius have also been investigated.21-23 For small Fresnel numbers,
their location is also closer to the pupil and not at the focal plane. It is also interesting to
note that in the case of a uniform beam with N > 12, the principal maximum of encircled
power for any radius rc lies at the focal plane, i.e., the Airy pattern yields the highest
encircled power. For N > 1.31 , the Airy disc contains the highest encircled power, i.e.,
the power in a circle of radius 1.22 (in units of l F ) as a function of z has the highest
value in the focal plane. Similarly, for N > 2.68 , the principal maximum of encircled
power in the second ring, i.e., for rc = 2.23 , also lies in the focal plane. 24
2
1
I ( r; z ) = 4 Bd2 Ú (
I (r) exp i Bd r 2
) J (p rr) r dr
0 , (4-66)
0
where
Bd = Sex l z (4-67)
represents the phase aberration of a plane wavefront with respect to a reference sphere
centered at a distance z and passing through the center of the pupil. In Eq. (4-66), the
irradiance in both the pupil and the observation planes is in units of the pupil irradiance
Pex Sex for a uniform circular beam. As in Eq. (4-58), r is in units of lz D.
2g
I (0; z ) = [cosh g - cos(p 4 z)] . (4-68)
[1 + (4 g z p) ] sinh g
2
The results for a uniform beam can be obtained by letting g Æ 0 . Thus, Eq. (4-68)
reduces to
z = 1 4 (2 n + 1) , n = 0 , 1 , 2 , K . (4-71a)
z = 1 8 (n + 1) , n = 0 , 1 , 2 , K . (4-71b)
These z values for the location of maxima and minima correspond to those axial positions
at which the circular exit pupil subtends an odd or an even number of Fresnel’s half-wave
384 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Figure 4-19 illustrates how the axial irradiance of collimated uniform and Gaussian
beams vary with distance z from the pupil. Compared to Figure 4-15, Figure 4-19
corresponds to N = O. Unlike the principal maximum of a focused beam in Figure 4-15a,
the maximum farthest from the pupil has a lower value than those closer to it in the case
of a Gaussian beam. Moreover, whereas the maxima of axial irradiance of a collimated
uniform beam have the same value of 4 and minima have a value of zero, the maxima in
the case of a Gaussian beam are higher and the minima are nonzero. Because of their
different amplitudes, the Fresnel zones do not cancel completely at the location of the
minima.
Now we consider a beam focused at the far-field distance and compare its axial
irradiance with that of a corresponding collimated beam. In particular, we show that beam
focusing yields a negligible improvement over a collimated beam, which further
decreases as y increases. From Eg. (4-62), the axial irradiance of a beam focused at the
far-field distance D 2 /A may be written
1(0; z; y) = (-
1t )2 ( 22 Y 2 ) - .-(coshy
1 - cosBd ) , (4-72)
4z Bd + Y smhy
20
18
\
\
\
16
\
\
\
14
12
N
g 10 \
\
\
8 l;j"-'J'\ \
'-'''''',
2 "
\
\
6
.. ~ "- .
, .... .....
'
4 , .......... ~~~-
2
-- -- -- --- -.-.-.
0
0 0.1 0.2 0.3 0.4 0.5
-z
Figure 4-19. Axial irradiance of a collimated Gaussian beam in units of the pupil
irradiance Pex / Sex at a uniform pupil. Tbe distance z is units of tbe far-field distance
D
2
/'A..
4.3 Systems with Circular Pupils 385
where
p Ê1 ˆ
Bd = -1 (4-73)
4 Ëz ¯
and the irradiance is in units of Pex Sex representing the illumination of a uniform pupil.
Figure 4-20 shows a comparison of the axial irradiance of a focused beam with that
for a collimated beam. We note that the axial irradiance of a focused beam differs from
that of a collimated beam in the near field. However, a collimated beam yields practically
the same irradiance on a target lying in the far field as a beam focused on it; in other
words, beam focusing does not significantly increase the power concentration on the
target. In the case of a uniform beam, a collimated beam gives a value of 0.5858 at the
far-field distance D2 l compared to a value of 0.6168 for a focused beam. Thus a
collimated beam gives an axial irradiance at a distance z in the far field that is ≥ 0.95
times the irradiance at this point when the beam is focused at it. Similarly, a collimated
beam gives an irradiance of 0.5428, 0.2895, and 0.1360 compared to a value of 0.5700,
0.2973, and 0.1370 for a focused beam when g = 1, 2, and 3, respectively. Thus, as g
increases, the axial irradiance of a collimated beam at the far-field distance D2 l
approaches the corresponding value for a beam focused at this point. The results for g = 0
and 1 are quite similar; the only significant difference being that the axial minima are
zero when g = 0 and nonzero when g = 1.
4.3.4.4 OTF
Substituting Eqs. (4-50) and (4-12b) into Eq. (4-8), the defocused OTF for a
Gaussian pupil can be written
1 v2 1 q2 v
t (v; g ; B d ) =
(
4 g exp -2g v 2 )¥ Û
Ù (
dq exp -2gq 2 ) Û
Ù ( )
exp -2gp 2 + 4iB d vp dp .
[
p 1 - exp( -2g ) ] ı
0 Ê
ı
2 vˆ
Á 1 q ˜
Ë ¯
(4-74)
Letting exp(i x ) = cos x + i sin x , the imaginary part of the integral vanishes upon
integration being an odd function of p with symmetric limits of integration, and Eq. (4-
74) reduces to
1 v2 1 q2 v
(
8g exp -2 g v 2 ) Û Û
t (v ; g ; Bd ) =
[
p 1 - exp( -2 g ) ] (
Ù dq exp -2 g q
ı
2
) Ù
ı
( )
exp -2 gp 2 cos( 4 Bd vp)dp ,
0 0
(4-75)
yielding a real OTF, as expected for a symmetric pupil function. The defocused OTF
given by Eq. (4-75) differs from the in-focus OTF given by Eq. (4-30) only by the
cos( 4Bd vp) factor in the p-integral. Letting g = 0 , the corresponding OTF for a uniform
pupil is given by [see Eq. (2-192a)]
386 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
y=0
N
g
(a)
0 .6 0.8 1.2
N
g
(b)
Collimated (c)
(d)
o L---~----~====~============~
o 0.2 0.4 0.6 0.8 1.2
f
o
(4-76)
Both Eqs. (4-75) and (4-76) show that the defocused OTF is independent of the sign
of Bd . Hence it is asymmetric about the Gaussian image plane for small Fresnel numbers
and symmetric for large, just like the PSF. For small Fresnel numbers , the cutoff
frequency is Vc = D/A Z instead of D/AR, as may be seen from Eqs. (4-4) and (4-8) by
replacing R with z. Figure 4-21 shows the defocused OTF corresponding to the PSFs
considered in Figure 4-l3. As the amount of defocus increases, the OTF becomes
negative in the low-frequency region, representing contrast reversal, and nearly zero in
the high-frequency region. As 'Y increases, the contrast reversal disappears. The OTF is
positive for all frequencies when .fi
~ 2. It is a coincidence that for one wave of
defocus, the OTFs for .fi = 2 and 3 are practically identical. From the approximate
expression given later [see Eq. (4-140)], the OTFs are given by exp(-27.74v 2 ) and
exp( -26.77v 2 ) , respectively.
0.4
'.,
1....\
'.\
"\
";" (a)
0.2 \
\
"\ . ~:.
'.
. \
,
0.8
.. \\
",\.fi
\
0.6
=3
\
'"
~ 0.4
.
'.
". 2 ,
\
\
t . \
..,
0.2
, "
\ (b)
\,.
o .. ...... "'+
............ '. ~. '.~"-~-...-.-.. .-~ - --_._-
-0 . 2 ~-~-~--~-~-~--~-~-~--~-~
1 . 0 ~-.---.--r--.---.--.--.---.--.--,
", '\
Bd= 1.5 A
0.8 \., \:,\,
: '.,
"\,
\.,
, -.f(= 3
0.6
'.,
'.,',,
'" 0.4
~
"
t
\ "
'\
" '
2\ \
0.2 '.'- , \ (c)
\ \
-, '
o "~'~~~~:~~~~;::~~.~-~.-~-==~-~-=-=.-~=-:=-~---------~~
- 0.2 0~-~--:-
0.~
2 --~----:0:-'.4:----'----::-
0 '::c
.6---'--0:"-.:
8 ---~-~
1.0 ..---.,..---,----.-----,-----.------.-----,-----,--,....--.
.\
'.'\\
\'
O.S
\\
'.' ,
"
"
- 0.2 O
L---'------:-
0.-'-
2 - - - - - ' - -0=-'.4' - - - ' - - -0-'-
.6---'----=-'S _-,-_......J
0.-=-
Table 4-9 lists the standard deviation of defocus aberration with and without
balancing with spherical aberration or astigmatism, showing that it decreases as y
increases. Spherical aberration reduces the standard deviation by a factor of 4, 3.74, and
fY
.J5 when = 0 , 1, and 3, respectively. The corresponding reduction factors for
astigmatism are .J3i2,
1.27, and ..fi. The reduction factor is the same whether spherical
aberration is balanced with defocus aberration or defocus aberration is balanced with
spherical aberration, but the balanced aberrations have different forms in the two cases.
For example, for a uniform beam, spherical aberration Asp4 balanced with defocus
aberration is A,(p4 - p2) , but defocus aberration BdP2 balanced with spherical
aberration is Bd[p2 - (15/16) p4]. Similarly, astigmatism Aap2 cos 2 e balanced with
defocus aberration is Aa(p2 cos 2 e - p2), but defocu s aberration balanced with
astigmatism is Bd[p 2 -(2/3)p 2 cos 2 e]. The difference comes from the fact that optimal
balancing is done with respect to defocus aberration in one case and spherical aberration
or astigmatism in the other.
If the defocus aberration is balanced with an aberration <l>(p, e) , then Eq. (4-58) is
modified and yields the axial irradiance according to
2
T(O' Z'Y) =
, ,
2y
1- exp(-2y)
(!i)2 fl 2f exp(- yp2)exp{i[<l>(P, e)+BdP2]} pdp de
1tZ 0
lt
0
(4-77)
I 2
TO' z· Y =
( , , )
2Y -
1 - exp (- 2 y) Z
(R)2 f exp( - 2
y x) exp[i( A sx + BdX)] dx (4-78)
o
390 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Standard Deviation
Aberration, Bd r2 + F( r, q) g=0 g =1 g ≥3
Bd Bd Bd Bd
Defocus, Bd r2 =
2 3 3.46 3.55 g
2 2
1
2g Ê Rˆ
I ( z; g ) = Á ˜
1 - exp ( - 2 g ) Ë z ¯ Ú [ ]
exp( - g x ) exp i(0.5 Aa + Bd ) x J 0 (0.5 Aa x ) dx ,
0
(4-79)
Ú ( 2 2
)
exp i Aa r cos q d q = exp 0.5i Aa r ( 2
)Ú (
exp 0.5iAa r2 cos 2q d q )
0 0
( ) (
= 2 p exp 0.5i Aa r 2 J 0 0.5Aa r 2 ) . (4-80)
12
g = 0, N = 1
10
As = – (15/16)Bd
I(z)
6
4
(a)
Bd As = – (5/16)l
2
As = 0
0
0 0.5 1 1.5
z/R
10
g = 1, N = 1
6 As = – Bd
I(z)
Bd
(b)
As = – l/3
2
As = 0
0
0 0.5 1 1.5
z/R
10
÷g = 3, N = 1
Bd
8 As = – 9.23l
6
I(z)
(c)
As = – 7.39l
2
As = 0
As = – (9/5)Bd
0
0 0.2 0.4 0.6 0.8 1.0
z/R
z value where the principal maximum lies in the absence of spherical aberration.
Compared to when As = 0 , the irradiance is smaller in the vicinity of the focal point and
larger in the vicinity of the principal maximum. The peak moves closer to the pupil, the
secondary maxima are higher, and the minima are no longer zero. If the amount of
spherical aberration varies as Bd varies with z R so that As = - (15 16) Bd , then the
aberration variance is minimum and the axial irradiance is higher for any value of z
(although the difference in the vicinity of the focal point is negligible), as illustrated by
the dashed curve. The peak value now lies even closer to the pupil, which is more than
ten times the focal-point irradiance. Thus, the peak value with spherical aberration is
more than five times the peak value without it. Figure 4-22b shows similar results for a
Gaussian beam with g = 1. Minimum variance of the defocus aberration when balanced
with spherical aberration is obtained in this case when As = - Bd . (A more precise
relationship is As = - 0.995 Bd .) The maxima for a Gaussian beam are lower compared to
those for a uniform beam.
Figure 4-23 shows how the axial irradiance is modified when defocus aberration is
balanced with astigmatism. The increase is not as dramatic as in the case of spherical
aberration when g = 0 or 1. This is consistent with the fact that aberration balancing also
in this case does not reduce the standard deviation of the defocus aberration significantly.
However, the axial irradiance of a weakly truncated beam with g = 3 increases
significantly and monotonically as z decreases.
Figure 4-24 shows the Strehl ratio for a given value of defocus aberration Bd as a
function of spherical aberration As . Both Bd and As are in units of wavelength l . It is
evident that, as the amount of defocus aberration increases, the maximum Strehl ratio is
obtained for an As value that is different from the one yielding minimum variance. For
example, when Bd = 3 l , maximum Strehl ratio is obtained when As = - 2.2 l instead of
- 2.8 l when g = 0 , and As = - 2 l instead of 3 l when g = 1. When g = 3, the
Strehl ratio peaks to the right of the expected value of As = - (9 5) Bd .
4.5
y = O, N= 1
3.5
IB, I
N
2.5
A, = - (2/3)B,
2
1.5
(a)
0.5
,
-,
0
0 0.5 1.5
~ z/R
4.5
y = 1, N = 1
3.5
\IB, I
2.5
~ A, = - O.77B,
1.5
.... Aa = 0
(b)
",
0.5
J\/
,
-,
0
0 0.5 1.5
- z /R
14 .
{1= 3, N = 1
12
10 ,
4
(c)
,
2 . .. . . .
-,
Aa = 0 ~.: : ..
0
0 0.2 0.4 0.6 0.8 1.0
- z /R
Figure 4-23. Axial irradiance of a focused beam aberrated by astigmatism Aa. The
defocus aberration IBdl in units of wavelength is also shown. (a) Uniform beam
( 'Y = 0). (b) Gaussian beam with 'Y = 1. (c) Weakly truncated Gaussian beam with
fY =3.
394 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
y=o
0.8
0.6
CI)
t
0.4
(a)
,
\
0.2 \
y= 1
0,8
0.6
CI)
t
0.4
(b)
/
0.2
I
/
3 " ~
"> - -.. . . . -
... - - -.. -..- .- ...
.....
I / - ·-· , .i , ._ ,;'"
I .' . .
I I /4
........ " ".
0
0
- -As
, ..... . - .--.,;,:: - .-
/
/
0.6 / .,..,"" :.,:. . . .
/2
CI)
/
t
0.4
(c)
0.2
0
0 12 16 20
- - As
Figure 4-24. Strehl ratio of a beam for a given value of defocus aberration Bd as a
function of spherical aberration As' Both Bd and As are in units of wavelength A..
(a) Uniform beam (y = 0). The Strehl ratio in this case is zero when Bd is an
integral number of wavelengths, as discussed in Section 2.5.2. (b) Gaussian beam
with y = 1. (c) Weakly truncated Gaussian beam with fY = 3.
4.3 Systems with Circular Pupils 395
2
2 1 2p
2 g (Bd p)
I (0; z; g ) =
1 - exp (- 2 g ) Ú Ú ( ) {[
exp - gr 2 exp i F(r, q) + Bd r 2 ]} r dr dq , (4-81)
0 0
where
Bd = ( p 4 z ) . (4-82)
Now the irradiance is in units of the pupil irradiance P Sex for a uniform pupil and z is in
units of the far-field distance D2 l , where D = 2 a is the pupil diameter.
1 2
2 gBd2
I (0; z; g ) =
1 - exp (- 2 g ) Ú [( 2
)]
exp(- g x ) exp i As x + Bd x dx (4-83)
0
and
1 2
2 gBd2
I (0; z; g ) =
1 - exp (- 2 g ) Ú [ ]
exp(- g x ) exp i (0.5Aa + Bd ) x J 0 (0.5Aa x ) dx , (4-84)
0
Figure 4-25 illustrates the effect of spherical aberration on the axial irradiance of a
collimated beam. The aberration-free axial irradiance is included in this figure for
comparison. The solid curves show that the axial irradiance increases dramatically near
the pupil, especially for small values of g , as a varying amount of spherical aberration
minimizing the variance of the defocus aberration is introduced. Similar results are
obtained when astigmatism is introduced, as illustrated in Figure 4-26, though the
increase at axial points near the pupil is not as dramatic. Of course, the irradiance at a
certain distance in the absence of an aberration is higher if the beam is focused at that
distance. For example, the irradiance of a uniform beam focused at a distance z is
(p 4 z)2 . Its value at z = 0.1 , corresponding to Bd = 1.25 l , is 62 compared to a value of
2 or 44 for a collimated beam with As = 0 or As = - (15 16) Bd , respectively.
120 6
g 0, N 0
100 5
80 4
As 0
I(z)
60 3
40 2 (a)
20 As – (15/16) Bd 1
Bd
0 0
0 0.1 0.2 0.3 0.4 0.5
z
120 6
g 1, N 0
100 5
80 4
As 0
I(z)
60 3
40 2 (b)
20 As – Bd 1
Bd
0 0
0 0.1 0.2 0.3 0.4 0.5
z
40
÷g 3, N 0
30
I(z)
20
(c)
10 As – (9/5) Bd
As 0
Bd
0
0 0.1 0.2 0.3 0.4 0.8
z
Figure 4-25. Axial irradiance of a collimated beam, i.e., one with a Fresnel number
N = 0 , aberrated by spherical aberration As . The axial irradiance when As = 0 is
shown for comparison. The defocus aberration Bd in units of wavelength is also
shown. (a) Uniform beam ( g = 0 ). (b) Gaussian beam with g = 1. (c) Weakly
truncated Gaussian beam with g = 3 . In figures (a) and (b), the right-hand side
scale is for As = 0 and Bd .
4.3 Systems with Circular Pupils 397
g 0, N 0
6
Aa – (2/3) Bd
5
(z) 3
Aa 0
(a)
2
1
Bd
0
0 0.2 0.4 0.6 0.8
z
14
g 1, N 0
12
10
8
(z)
6 Aa – 0.77 Bd
(b)
4
Aa 0
2
Bd
0
0 0.2 0.4 0.6 0.8
z
50
÷g 3, N 0
40
30
(z)
Aa – Bd
20
(c)
10
Aa 0
Bd
0
0 0.2 0.4 0.6 0.8
z
Figure 4-26. Axial irradiance of a collimated beam, i.e., one with a Fresnel number
N = 0 , aberrated by astigmatism Aa . The axial irradiance when Aa = 0 is shown for
comparison. The defocus aberration Bd in units of wavelength is also shown. (a)
Uniform beam ( g = 0 ). (b) Gaussian beam with g = 1. (c) Weakly truncated
Gaussian beam with g = 3 .
398 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
2
1
I ( r; g ) = 4 Ú
0
[( 4
I (r) exp i As r + Bd r 2
)] J (p rr) r dr
0 . (4-85)
Figure 4-27 shows the aberrated PSFs for As = 1, 2, 3, and 4 waves with and without
balancing by defocus. The PSFs are normalized to unity at the center. Spherical
aberration broadens the PSF, makes the diffraction minima nonzero, and increases the
value of the secondary maxima. For balanced aberrations, however, the radius of the
central bright spot and positions of the minima and maxima do not change, but power
flows from the central spot to the diffraction rings. Thus, whereas the Gaussian amplitude
at the pupil increases the size of the central bright spot but decreases the secondary
maxima, the aberrations do not change the size of the central bright spot but increase the
secondary maxima. Hence, the advantage of the Gaussian apodization in reducing the
secondary maxima is lost if spherical aberration is present. The amount of spherical
aberration must be £ l 4 to take advantage of the positive attribute of apodization. For a
weakly truncated pupil, the effect of spherical aberration on the spot size is similar, but
there are no diffraction rings.
2
1 2p
Û Û Ï R ¸
¥ Ù Ù I (r) exp Ìi ÈÍ F(r, q; g ) - p r r cos (q - q i )˘˙ ˝ r dr dq , (4-86)
ı ı ÓÎ z ˚˛
0 0
with Bd given by Eq. (4-53). The reasoning used to determine the symmetry properties
and the results thus obtained are the same, except that, as illustrated in Figures 4-8a and
4-8b, the axial irradiance for spherical aberration (expressed by a Zernike-Gauss
polynomial) is not symmetric about the Gaussian image plane even for large Fresnel
numbers. Letting z ~ R for a system with a large Fresnel number, the axial irradiance for
spherical aberration may be written
4.3 Systems with Circular Pupils 399
g 0, Spherical
0.8
0.6
(r)
0.4
(a)
3
1 2
0.2
As 0
0
0 1 2 3 4 5
r
g 0, Balanced spherical
0.8
0.6
(r)
As 3
0.4
(b)
0.2
2
1
0
0 1 2 3 4 5
r
g 1, Spherical
0.8
0.6
(r)
0.4
(c)
0.2 3
1 2
As 0
0
0 1 2 3 4 5
r
Figure 4-27. PSFs aberrated by spherical and balanced spherical aberrations. (a)
and (b) Uniform pupil ( g = 0 ). (c) Gaussian pupil with g = 1. Aberration-free PSFs
are also included for comparison. (Cont.)
400 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
g 1, Balanced spherical
0.8
0.6
(r)
As 3
0.4
(d)
0.2
2
1
0
0 1 2 3 4 5
r
÷g 3, Spherical
0.8
0.6
(r)
0.4 2
3
(e)
0.2
As 0
0
0 1 2 3 4 5
r
÷g 3, Balanced spherical
0.8
0.6
(r)
0.4
(f)
0.2
0
0 1 2 3 4 5
r
2
1
Û
I (0; z; g ) = 4 Ù
ı
( ) [
I (r) exp iBd r2 exp i A n 0 Rn0 (r; g ) r dr ] . (4-88)
0
Unlike the radial polynomials Rn0 (r) (dsicussed in Section 2.4 for a uniform pupil), the
radial polynomials Rn0 (r; g ) are not given by the Legendre polynomials and do not
possess their symmetry property, Moreover, the arguments made following Eq. (2-145)
do not hold. In the absence of any aberration, Eq. (4-86) reduces to Eq. (4-58), which
shows that the aberration-free PSF is symmetric about the Gaussian image plane.
Accordingly, the axial irradiance is also symmetric. From the form of the Zernike-Gauss
circle polynomial for astigmatism, we note that the point of axial symmetry for
astigmatism Aa r2 cos 2 q is the same, namely, Bd = - Aa 2, for uniform and Gaussian
pupils.
A(r) = (
2 g exp - gr2 ) . (4-89)
Similarly, Eq. (4-10) for the pupil irradiance distribution may be written
(
I (r) = 2 g exp - 2gr2 ) , (4-90a)
or
2
( ) ( )
I rp = 2 Pex p w 2 exp È- 2 rp w ˘ .
ÍÎ ˙˚ ( ) (4-90b)
In Eq. (4-90a), I(r) is in units of Pex Sex , as in Eq. (4-10). In Eq. (4-90b), no such
normalization is used.
402 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
4.3.7.2 PSF
Substituting Eq. (4-90a) into Eq. (4-58) and replacing the upper limit of integration
by infinity, we obtain
2
2 •
R
I ( r ; z; g ) = 8 g Ê ˆ Ú [ 2
]
exp - ( g - iBd ) r J 0 ( prr ) r dr . (4-91)
Ë z¯
0
Ú (- ar ) J0 (br) r dr ( )
2
= (1 2a ) exp - b 2 4a , Re a > 0 , (4-92)
0
I ( r; z ) = ( R z )
2
[2 g ( B 2
d )] [
+ g 2 exp - g p 2 r 2 2 Bd2 + g 2 ( )] , (4-93a)
( ) (
= 2 Pex p w 2z exp - 2 r 2 w 2z ) , (4-93b)
where
2 2
w 2z = (l z p w ) + w 2 (1 - z R) (4-93c)
is the beam radius at a distance z. In Eq. (4-93a), the irradiance is in units of Pex Sex l2 R 2
and r is in units of lz D, as was the case in Eq. (4-58). In Eq. (4-93b) these quantities
are not normalized to illustrate that the diffracted irradiance is independent of the pupil
radius a. Comparing Eqs. (4-90) and (4-93), we note that, when the truncation of the
beam by the pupil is negligible, the diffraction rings disappear and a Gaussian beam
remains Gaussian as it propagates.27,28
Ú ( )
exp - br 2 J 0 (p r r) r dr , (4-94)
0
J n (p rr) 1
lim =
rn 2 n n! (4-95)
rÆ 0
4.3 Systems with Circular Pupils 403
and
d È J n ( p rr ) ˘ J n +1( p rr )
Í n ˙ = pr , (4-96)
dr ÍÎ r ˙˚ rn
n
1 È Ê p2 r 2 ˆ •
Ê - pr ˆ J pr ˘ .
ÍexpÁ - - exp( - b ) Â n( )˙ (4-97)
2 b ÍÎ Ë 4b ˜¯ n = 0 Ë 2b ¯ ˙˚
Thus, the diffracted amplitude consists of the sum of a Gaussian term and other terms,
which vanish in the limit of large g . Hence, for a weakly truncated beam, Eq. (4-58)
reduces to Eq. (4-93a), as expected.
[
P(rc ; z ) = 1 - exp - g p 2 rc2 2 Bd2 + g 2 ( )] , (4-98a)
{
= Pex 1 - exp - 2 rc2 w 2z[( )] } , (4-98b)
I (0; z ) = ( R z )
2
[2 g ( B 2
d + g2 )] (4-100a)
= 2 Pex p w 2z . (4-100b)
Of course, for large values of g, Eq. (4-62) also reduces to Eq. (4-93a), as expected. If we
let z = R in Eqs. (4-93a) through (4-93c), we obtain the focal-plane irradiance
distribution
(
I (r; R) = (2 g ) exp - p 2 r 2 2 g ) (4-101a)
( ) (
= 2 Pex p w 2R exp - 2 r 2 w 2R ) , (4-101b)
where
wR = lR pw (4-101c)
is the beam radius in the focal plane. The focal-point irradiance is given by
404 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
I (0; R) = 2 g (4-102a)
= 2 p Pex w 2 l2 R 2 , (4-102b)
If we equate to zero the derivative of Eq. (4-93a) with respect to z, we obtain the z
value z p at which the axial irradiance is maximum:
1
[
z p R = 1 + (g p N )
2
] (4-103a)
1
[
= 1 + (w R w )
2
] . (4-103b)
It is evident that z p < R , i.e., the peak value lies at a point that is closer to the pupil
compared to the focal point. Substituting Eq. (4-103a) into Eq. (4-100a) and Eq. (4-103b)
into Eq. (4-100b), we obtain the peak value of the axial irradiance
( ) (
I 0; z p = (2 g ) + 2 g p 2 N 2 ) (4-104a)
= 2 Pex p w 2z p , (4-104b)
where
1
[
w 2z p = w 2 1 + ( w w R )
2
] . (4-104c)
showing that w z p < w R . Thus, the radius of the beam waist is smaller than the beam
radius in the focal plane.
Comparing Eqs. (4-102a) and (4-104a), we note that the peak axial irradiance is
higher than the corresponding focal-point irradiance by 2 g p 2 N 2 (in units of
Pex Sex l2 R 2 ) or by 2 Pex p w 2. As illustrated in Figure 4-15a for g = 3 and N = 1, the
value at z = 0 is given by 2 g p 2 or 1.824. The peak is located at z = 0.1086 R with a
value of 2.046, which is 9.2 times the corresponding focal-point irradiance of 0.222 and
1.12 times the central pupil irradiance of 1.824. Equation (4-100a) shows that the defocus
aberration of Bd = 81 p or 4.10 l for this point reduces the irradiance by a factor of
0.1086, but the inverse-square law increases it by a factor of 84.79. Equations (4-103a)
and (4-104c) can also be written in the form
2˘ 1
z p R = È1 + p N g
ÎÍ
( ) ˚˙
(4-105a)
4.3 Systems with Circular Pupils 405
and
2 1
ÍÎ (
w 2z p = w 2 È1 + p N g ˘
˙˚ ) , (4-105b)
Even though the peak of axial irradiance does not lie at the focal point, maximum
central irradiance on a target at a fixed distance z is obtained when the beam is focused on
it, as discussed in Section 4.3.4.3.1. Similarly, although the beam waist does not lie in the
focal plane, the radius of the beam in the plane of the target is minimum when the beam
is focused on it. This may be seen by considering R as a variable and letting ∂w z ∂R = 0
from Eq. (4-93c). We find that the beam radius w z on the target is minimum with a value
of l z pw when R = z . Correspondingly, the axial irradiance I (0; z ) on it is maximum,
as may be seen from Eq. (4-93b). If we choose R such that the beam waist lies on the
target, its value is given by Eq. (4-103b) with z p = z :
R
z [ 2
= 1 + (w R w ) > 1 , ] (4-106)
where w R = l R p w is the beam radius in the focal plane. As expected, the beam is
focused beyond the target. The beam radius on the target is the radius of the beam waist,
which according to Eq. (4-104d), is given by
1
[
w 2z = w2R 1 + ( w R w )
2
]
2
Ê lz ˆ
= Á
Ë pw¯
˜ [1 + ( w R w)
2
], (4-107)
showing, as expected, that it is larger than the beam radius l z pw when the beam is
focused on the target. Accordingly, the axial irradiance on the target is smaller when the
waist of the beam lies on it compared to when the beam is focused on it. Again, as
explained in Section 4.3.4.3.1, the reason for this is that, while the effect of the inverse-
square law dependence on the target distance is fixed, the defocus aberration reduces the
axial irradiance when the beam is focused anywhere except on the target.
406 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Figure 4-29 shows how the axial irradiance of a focused Gaussian beam varies when
g = 2 and N = 1, 10, 100. Once again, the solid curves in this figure have been
obtained by using Eq. (4-58); the dashed curves represent their corresponding
approximations given by Eq. (4-100a). It is evident that Eq. (4-100a) represents the true
axial irradiance quite well. The only significant difference occurs when N = 1, in that the
true results show secondary maxima and minima, but the approximate result shows only
0.5 1.0
g =2
0.4 0.8
Pg
0.3 0.6
Pg(r;R)
lg(r;R)
0.2 0.4
0.1 lg 0.2
0.0 0.0
0 1 2 3
r; rc
0.6 0.6
1.0
0.4 0.4
0.5
0.2 0.2
the principal maximum. For larger values of g ; e.g., g = 2.5 , the secondary maxima
and minima disappear and the true and approximate results overlap each other at the scale
of Figure 4-29. Hence, we conclude that the truncation of an aberration-free Gaussian
beam by a pupil has a negligible effect on the irradiance distribution as the beam
propagates when g ≥ 2.
r R È Ê 1 r2 ˆ ˘
•
U ( ri ; z ) = - 2i 2 g Ê ˆ exp Íik Á z + i ˜ ˙ [ ]
exp - ( g - iBd ) r2 J 0 ( pr ri D l z ) r dr
Ú
Ë z¯ ÍÎ Ë 2 z ¯ ˙˚ 0
R i È Ê 1 r2 ˆ ˘
= - 2g Ê ˆ
Ë z ¯ ( g - iBd ) [ 2
exp Íik Á z + i ˜ ˙ exp - ( pri D l z ) 4( g - iBd ) .
ÍÎ Ë 2 z ¯ ˙˚ ]
(4-108)
[ (
z Rz = Sex Bd l z Bd2 + g 2 - 1 )]
408 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
1-z R
= 2 - 1 . (4-109)
(1 - z R)2 (
+ lz p w2 )
At the waist position z p , Rz p = • , implying a plane wave. Moreover, at the focal plane,
Rz = - R . A negative value of Rz indicates a diverging spherical wave. For z > R ,
z Rz < - 1 and the beam continues to expand as it propagates. Substituting for g and
Bd , and noting that the units of diffracted irradiance are Pex Sex l2 R 2 , the complex
amplitude given by Eq. (4-108) may also be written
w0 Ê r2 ˆ Ê 1 ri2 ˆ
U (ri ; z ) = - iA0
wz
{ }
exp i [kz + f( z )] expÁ ip i ˜ exp Á -
Ë lRz ¯
2˜
Ë 2 wz ¯
, (4-110)
where
pw 20
tan f( z ) = (4-111)
lz
is a phase factor. It is seen that the diffracted amplitude is independent of the pupil radius
a, as expected for a weakly truncated pupil, equivalent to free-space propagation.
{ [
I ( r; z ) = 2 g 1 + ( 4 g z p )
2
] } exp {- 8g z r [1 + (4g z p) ] }
2 2 2
, (4-112a)
( ) (
= 2 Pex p w 2z exp - 2 r 2 w 2z ) , (4-112b)
where
2
ÍÎ (
w 2z = w 2 È1 + l z p w 2 ˘ .
˙˚ ) (4-112c)
2
ÍÎ (
Rz = - z È1 + p w 2 l z ˘ .
˙˚ ) (4-112d)
ÎÍ
(
I z = I0 È1 + z / z R ˘)
˚˙
(4-113)
and
2
ÍÎ (
w 2z = w 20 È1 + z / z R ˘ )
˙˚
(4-114)
where
z R = pw 20 / l (4-115)
is a distance, called the Rayleigh range, from the plane of the beam waist to a plane in
which the central irradiance decreases by a factor of 2 and the beam radius increases by a
factor of 2 . The beam radius w z increases monotonically as z increases, and at large
distances it is approximately given by
w z = w 0 ( z z R ) = lz pw 0 . (4-116)
As illustrated in Figure 4-30, the beam expands as a hyperbola with asymptotes, indicated
by the dashed lines, inclined to its axis at an angle
q = w z z = l pw 0 , (4-117)
called the far-field divergence or diffraction angle of the beam. The backward
propagation of the beam is also shown in the figure. For large values of z, the beam radius
increases linearly with it. Equation (4-112c) for the radius of curvature of the beam
wavefront at a distance z may also be written
2
Rz = - z È1 + z R
ÎÍ
( / z) ˘ .
˚˙
(4-118)
w0 q ÷2w 0
z=0
z = – zR z = zR
Figure 4-30. Beam radius and divergence angle. The radius of the beam is w 0 and
its wavefront is planar at z = 0. Its radius increases by a factor of 2 and its
wavefront becomes spherical with a radius of curvature mz R at z = ± z R .
410 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
2
ÍÎ (
w l2 = w 20 È1 + z0 / z R ˘ )
˙˚
(4-119)
and
2
Rl = z 0 ÈÍ1 + z R / z 0 ˘˙ ,
( ) (4-120)
Î ˚
where
z R = pw 20 / l (4-121)
is the Rayleigh range of the incident or object beam. We assume that the lens is large
enough that its radius is ≥ 2w l so that the weak-truncation approximation is satisfied.
The radius of curvature Rl+ of the wavefront exiting from the lens is given by
1 1 1
+ = + . (4-122)
Rl Rl f
The beam transmitted by the lens first converges and then expands as it propagates to the
right of the lens. Its radius w ¢z at a distance z ¢ from the lens is given by Eq. (4-93c) with
w0 wl
w0 wf
( ) z0 z0
f
Figure 4-31. Imaging of a Gaussian beam by a lens of focal length f . The incident
or the object beam has a waist of radius w 0 at a (numerically negative) distance z 0
from the lens, and the transmitted or the image beam has a waist of radius w¢0 at a
distance z 0¢ .
4.3 Systems with Circular Pupils 411
appropriate substitutions:
2 2
Ê lz ¢ ˆ 2Ê z¢ ˆ
w ¢z2 = Á ˜ + w l Á1 - + ˜
Ë pw l ¯ Ë Rl ¯
2
=
(w 0 z ¢ ) 2 + (w 0 z ¢ )
2 z02 + z R2 Ê 1 1 z0 ˆ (4-123)
Á z¢ - f - 2 ˜ .
z02 + z R2 2
zR Ë z0 + z R2 ¯
lf f
w ¢f = = , (4-124)
pw 0 zR
showing that the beam radius in the back focal plane of the lens is independent of the
location z0 of the incident waist.
The radius of the waist of the transmitted beam according to Eq. (4-104c) is given by
2˘ 1
È Ê ˆ
Í ˙
Á w ˜
w ¢02 = w l2 Í1 + Á l ˜
˙
Í Áw + ˜ ˙
Í Ë Rl ¯ ˙
Î ˚
1
È Ê pw 2 ˆ 2 ˘
2Í l
= wl 1 + Á ˜ ˙
Í Ë l Rl+ ¯ ˙
Î ˚
w 20 f 2 .
= (4-125)
(z 0 + f )2 + z R2
It lies at a distance z0¢ from the lens which, according to Eq. (4-103b), is given by
1
z0¢ È Ê lR + ˆ 2 ˘
= Í1 + Á l2 ˜ ˙
Rl+ ÍÎ Ë pw l ¯ ˙˚
or
z0¢ =
[
f z0 ( z0 + f ) + z R2 ] . (4-126)
2
( z0 + f ) + z R2
Considering the incident and transmitted waists as an object and its corresponding
image formed by the lens, Eq. (4-126) relating their distances from the lens may be
written
412 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
1 1 1
- = , (4-127)
z0¢ z0 + z R2 ( z0 + f )
/ f
3
zR
= 0 1/4
f
2
1/2
1
1 2
z0 /f
2 (a)
4
4 3 2 1 0 1 2 3 4
z0 /f
3
zR
0
f
M
1/4
2
1/2
1
(b)
1
0
–4 –3 –2 –1 0 1 2 3 4
zo/ f
Figure 4-32. Incident and transmitted waist relationships. (a) Object and
corresponding image distances. (b) Transverse magnification.
4.3 Systems with Circular Pupils 413
transmitted beam lies in its back focal plane, and not at infinity as in conventional
imaging. Accordingly, the transmitted waist lies in the back focal plane only if the
incident waist lies in the front focal plane. Otherwise, the transmitted waist lies beyond
the back focal plane if z0 f < - 1 , or inside the back focal plane if z0 f > - 1 . It lies in
the focal plane, i.e., z0¢ = 0 if Rl = 0 or z0 + f = - z R2 z0 .
The radius and position of the corresponding transmitted waist are given by Eqs. (4-118)
and (4-126) according to
w0
w ¢0 = 1/ 2 (4-129)
[1 + ( zR f )2 ]
and
f
z0¢ = 2 . (4-130)
1 + ( f zR )
M = w ¢0 / w 0
1
= 1/ 2 , (4-131)
2
Ï
Ó
[ (
Ì 1 + z0 / f )] ( + zR / f
2¸
) ˝˛
which reduces to conventional image magnification as z R f Æ 0 , provided the negative
sign associated with an inverted image is ignored. A magnification of unity is obtained
when
[
z0 f = - 1 ± 1 - ( z R f )
2
] , (4-132)
and correspondingly
z0¢ f = - z0 f . (4-133)
The radii of curvature of the wavefronts incident on and exiting from the lens are given
by Rl = - 2 f = - Rl+ . As illustrated in Figure 4-32b, the maximum value of
414 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
magnification is f / z R , which occurs when the object waist lies in the front focal plane,
i.e., when z0 = - f . Hence, unity magnification in this case occurs only if f = z R .
Moreover, M £ 1 for z R f ≥ 1.
z R¢ = pw ¢02 l = M 2 z R . (4--134)
4.3.7.6 OTF
If we write the p-integral in Eq. (4-74) in the form
1 q2 v
Û 2
[
exp - (2 g ) Bd2 v 2 ] Ù
ı
(
exp - 2 g p + i 2 g Bd v ) dp ,
Ê 2 vˆ
Á 1 q ˜
Ë ¯
we see that, for large values of g , the contribution to the integral is negligible unless
v = 0 . Therefore, in such cases, Eq. (4-74) may be written
1 q2
t (v ; g ; Bd ) =
{[
8g exp - 2 g + (2 g ) Bd2 v 2 Û
2 ] } 1
( ) Û
( )
2
Ù dq exp -2 gq Ù exp -2 gp dp .
[
p 1 - exp( -2 g ) ı ] 0
ı
0
(4-136)
The 2D integral represents the Gaussian-weighted area of a quadrant of the pupil, which
may also be written
1
pÛ p
2Ù
ı
(
exp -2 gr2 r dr =
8 g
)
1 - exp( -2 g ) [ ] . (4-137)
0
{[
t (v ; g ; Bd ) = exp - 2 g + (2 g ) Bd2 v 2 ] } , 0 £v £1 , (4-138)
Substituting for the PSF from Eq. (4-86b) into Eq. (4-132), we obtain
(
t (v i ) = exp - p 2w 2z v i2 2 ) , 0 £ vi £ D lz , (4-140)
which is the same as Eq. (4-138). If we let Bd = 0 in Eq. (4-138), we obtain the defocus-
free OTF,
(
t (v ) = exp -2 gv 2 ) , 0 £v £1 , (4-141)
which is a zero-order Hankle transform of the defocus-free PSF given by Eq. (4-101a), as
expected. The defocus-free and defocused OTFs are illustrated for g = 3 in Figures 4-6
and 4-21, respectively.
and
• 2p
g n
n
<F > =
p Ú Ú ( )[
exp - g r2 F(r, q) r dr dq , ] (4-143)
0 0
respectively.
2 • 2
È g ˘
S = Í ˙ Ú [
exp - ( g - iBd ) x dx ]
Î1 - exp ( - g ) ˚ 0
1
= 2 , (4-144)
1 + ( Bd g )
416 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
a result that is already obtained in Eq. (4-65). Closed-form expressions for spherical
aberration and astigmatism can also be obtained. They are listed in Table 4-10 along with
the integral expressions for coma and balanced spherical aberration. For comparison, the
corresponding expressions for a uniform beam are also listed in the table.
The standard deviation of a primary aberration for a large value of g can be obtained
by calculating its mean and mean square values according to Eq. (4-143). The results thus
obtained are given in Table 4-2. The corresponding balanced aberrations and their
standard deviations are given in Tables 4-3 and 4-4, respectively. The balancing of an
aberration reduces the standard deviation by a factor of 5 , 3 , and 2 in the case of
spherical aberration, coma, and astigmatism, respectively, as noted in Table 4-5. The
diffraction focus for these aberrations is listed in Table 4-6. The amount of balancing
aberration decreases as g increases in the case of spherical aberration and coma, but does
not change in the case of astigmatism. For example, in the case of spherical aberration,
the amount of balancing defocus for a weakly truncated Gaussian beam is ( 4 g ) times
Aberration S ( g ≥3 )
Spherical, ÏÔ È 1 2 2 †
p2 Ê 1 ˆ˘ È1 Ê 1 ˆ ˘ ¸Ô
ÌÍ - S Á ˜˙ + Í - C Á ˜˙ ˝
As r 4 b¢ ÔÓ Î 2 Ë b¢ ¯ ˚ Î2 Ë b¢ ¯ ˚ Ô
˛
Balanced spherical, Eq. (4-36) with Bd = - (4 g ) As
g 2s s
As r 4 + Bd r2 and As =
2 5
Coma, g 3 2sc
Eq. (4-36) with Bt = 0 and Ac =
Ac r3 cos q 3
Balanced coma, Eq. (4-36) with Bt = - ( 2 g ) Ac
( )
Ac r3 + Bt r cos q
and Ac = g 3 2 s bc
Astigmatism, 12
Aa r2 cos 2 q [1 + ( A a g)
2
] [
= 1 + 2s 2a ]
12
Balanced astigmatism, 1 1
(
Aa r2 cos 2 q - 1 2 ) [1 + ( A a 2g )
2
] =
1 + s 2ba
1 1
Defocus, Bd r2 [1 + ( B d g)
2
] =
1 + s 2d
b b
† b = 3 5s s p , b ¢ = ps s 5 , C( b ) = Ú (
cos p x 2 2 dx , S(b) = ) Ú sin(p x 2) dx
2
0 0
4.3 Systems with Circular Pupils 417
the corresponding amount for a uniform beam. Similarly, in the case of coma, the
balancing tilt for a weakly truncated Gaussian beam is (3 g ) times the corresponding
amount for a uniform beam. The location of the diffraction focus is independent of the
value of g in the case of astigmatism, since the balancing defocus is the same regardless
of the value of g . Compared to the peak value of an aberration, its standard deviation is
smaller by a factor of g 2 2 , g 3 2 , and 2g in the case of spherical aberration, coma, and
astigmatism, respectively.
When a Gaussian beam is weakly truncated, i.e., when g is large, the quantity ps in
Table 4-8 reduces to1-3
ps = < rs > = (s 2 g ) ps 2 = (s 2) ! g s2
. (4-145)
As a result, we obtain simple expressions for the radial polynomials, which are listed in
Table 4-8. They are similar to Laguerre polynomials.12 If we normalize rp by w (instead
of by a), then g disappears from these expressions. The standard deviation of an
aberration can be obtained by comparing its form with the corresponding orthonormal
aberration in Eq. (4-39). Since the power in a weakly truncated Gaussian beam is
concentrated in a small region near the center of the pupil, the effect of the aberration in
its outer region is negligible. Accordingly, the aberration tolerances in terms of the peak
value of the aberration at the edge of the pupil (r = 1) may not be very meaningful. They
may instead be defined in terms of their value at the Gaussian radius.2,32
Comparing the standard deviation reduction factors given in Table 4-7 with those for
a weakly truncated Gaussian beam given in Table 4-4, we find that they agree with each
other with negligible difference for g ≥ 3. This provides a convenient definition of a
weakly truncated Gaussian beam, namely, that a ≥ 3w . Some authors have assumed that
g ≥ 2 provides a sufficient condition for the validity of the aberration analysis of a
weakly truncated Gaussian beam given here. When g = 2 , the standard deviation of
balanced spherical aberration according to the weakly truncated beam assumption is
given by As 8, whereas the true value is given by As 18.29 , which is significantly
different. When g = 3, the corresponding standard deviations are given by As 40.50
and As 43.52 , which are nearly equal to each other. The difference between the true and
approximate results is even less for g > 3. Similarly, the approximate and exact values
of the standard deviation of balanced coma are Ac 8 and Ac 12.21 for g = 2 , and
Ac 27 and Ac 27 .57 for g = 3. In the case of balanced astigmatism, however, even
when g = 2 , the approximate value of Aa 8 is nearly equal to the true value of
Aa 9.08 . Of course, for g = 3, the approximate and exact values of Aa 18 and
Aa 18 .06 , respectively, are practically equal to each other. When g = 2 , the standard
deviation of defocus aberration Bd r2 is Bd 4.80 compared to a true value of Bd 4 .
When g = 3, the true and approximate values are Bd 9.05 and Bd 9, respectively,
which are practically equal to each other.
418 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
When g = 2 , even though the true focal-plane distribution obtained from Eq. (4-
22) agrees quite well with the corresponding approximate result obtained from Eq. (4-
101a) (see Figure 4-28), the true and the approximate standard deviations of primary
aberrations are significantly different, as pointed out above. The reason for the
discrepancy in the case of an aberrated beam is simple. Even though the irradiance in the
region of the pupil w a £ r £ 1 is quite small compared with that at or near its center, the
amplitude in this region is not as small. Moreover, the aberration in this region can be
quite large and thus have a significant effect on the standard deviation. In the case of
spherical aberration, it increases as r 4 . In the case of coma and astigmatism, it increases
as r3 and r2 , respectively. Hence, we require a larger value of g, namely g ≥ 3, for the
aberrated-beam analysis of this section to be valid. This is also true of defocus, which
varies as r2 .
r¢ = g r (4-146)
is a radial variable in the pupil plane normalized by the beam radius w, and the aberration
coefficients Ai¢ represent the peak values of the Seidel aberrations at r¢ = 1. These
aberration coefficients are related to the coefficients Ai according to
The reason for defining the primed aberration coefficients in this manner is as follows.
Since the power in a weakly truncated Gaussian beam is concentrated in a small region
near the center of the pupil, the effect of the aberration in its outer region is negligible.
Table 4-11. Primary aberrations and their standard deviations for optical systems
( )
with weakly g ≥ 3 truncated Gaussian circular pupils.
Accordingly, the aberration tolerances in terms of the peak value of the aberration at the
edge of the pupil (r = 1) are not very meaningful. We note from Table 4-8 for the
Zernike-Gauss circle polynomials for a weakly truncated pupil (or Table 4-11) that the
point with respect to which the variance of an aberration is minimized is given by
Bd = - ( 4 g ) As (4-148a)
= - 4g As¢ , (4-148b)
Bt = - (2 g ) Ac (4-149a)
= - 2 g Ac¢ , (4-149b)
and
Bd = - (1 2) Aa (4-150a)
= - ( g 2) Aa¢ , (4-150b)
in the case of spherical aberration, coma, and astigmatism, respectively. From Table 4-11
we note that the balancing of a primary aberration reduces its standard deviation by a
factor of 5 , 3 , and 2 in the case of spherical aberration, coma, and astigmatism,
respectively. These reduction factors are listed in Table 4-5 for the uniform ( g = 0) , e 2 -
truncated ( g = 1) , and weakly truncated (large g) Gaussian beams. It is evident that as g
increases, the reduction factors decrease for spherical aberration and increase for
astigmatism. The amount of balancing aberration decreases as g increases in the case of
spherical aberration and coma, but it does not change in the case of astigmatism. For
example, in the case of spherical aberration, the amount of balancing defocus for a
weakly truncated Gaussian beam is ( 4 g ) times the corresponding amount for a uniform
beam. Similarly, in the case of coma, the balancing tilt for a weakly truncated Gaussian
beam is (3 g ) times the corresponding amount for a uniform beam. For large g , the
aberration tolerance in terms of the coefficients Ai¢ is given in Table 4-11 for a Strehl
ratio of 0.8.
Table 4-7 gives the reduction factors that relate the peak value Ai of a primary
aberration at the edge of a circular aperture and the standard deviation of its
corresponding balanced aberration for various values of l. In the case of balanced
aberrations, these numbers are given by 5a40 , 2 2 a31 , and 2 6 a22 for spherical
aberration, coma, and astigmatism, respectively. For example, for spherical aberration
As r 4 , the standard deviation of the corresponding balanced spherical aberration when
g = 2 is equal to As 18.29 .
parameters, the size and location of its waist. The beam radius in any plane can be
determined from Eq. (4-16) by placing in the plane a circular aperture concentric with the
beam and measuring the fractional power passing through it. The difficulty of positioning
an aperture in two dimensions can be avoided by using a slit or a wire and measuring the
transmitted flux.33 The size of the waist of a beam can be determined by focusing it with
a lens of known focal length, measuring the spot size in the focal plane, and utilizing Eq.
(4-124). Its position can be calculated from Eq. (4-117) by measuring its divergence
angle. Since it involves large distances, it can be determined more conveniently by an
interferometric method.34
In practice, laser beams are not diffraction-limited Gaussian beams, i.e., they are not
the fundamental TEM00 mode, due to limitations of the optical cavity or nonuniformity of
the lasing material. They are multimode beams which expand more than that predicted by
Eq. (4-93c) for a Gaussian beam, although a spot radius can be defined for each mode35
and its propagation can be investigated.36 The amplitude of a mode of order p in the plane
of the beam waist at z = 0 is given by37,38
Ê r2 ˆ Ê 1 ri2 ˆ
U (ri ; 0) = Ap L p Á 2 i 2 ˜ exp Á - 2˜ , (4-151)
Ë w0 ¯ Ë 2 w0 ¯
w 0 Ê ri2 ˆ Ê r2 ˆ Ê 1 ri2 ˆ
p
U (ri ; z ) = - i( - 1) Ap
wz Ë wz ¯
{ }
L p Á 2 2 ˜ exp i [kz + f( z )] expÁ ip i ˜ exp Á -
Ë lRz ¯
2˜
Ë 2 wz ¯
,
(4-152)
where
f p ( z ) = (2 p + 1) tan 1
(pw 2
0 / lz ) (4-153)
and w z and Rz are given by the same Eqs. (4-93c) and (4-109), as for the fundamental
Gaussian mode.
normalizing Gaussian beam) of the same width as the multimode beam, the divergence is
larger and Rayleigh range is smaller by a factor of M 2 according to
l l
Q = Mq = M = M2 (4-154)
pw 0 pW0
and
pw 20 1 pW02 W
zR = = 2 = 0 . (4-155)
l M l Q
Of course, the beam radius at a distance z from the beam waist, following Eq. (4-154), is
given by
2 È Ê lz 2˘
È zˆ ˘
Ê 2ˆ
Wz2 = W02 Í1 + Á 2
˜ ˙ = W0 Í1 + Á 2 M ˜
˙ . (4-156)
ÍÎ Ë zR ¯ ˙
˚ ÍÎ Ë pW0 ¯ ˙
˚
Similarly, from Eqs. (4-113) and (4-118), its central irradiance in terms of the central
irradiance I0 at the beam waist and the radius of curvature of its wavefront may be
written
I0
Iz = (4-157)
(
1 + l z M 2 p W 02 )2
and
2
Rz = - z ÈÍ1 + pW 02 l zM 2 ˘˙ .
( ) (4-158)
Î ˚
Thus, M 2 is a measure of the quality of a beam, and the higher its value the poorer its
quality. A value lying between 1 and 1.2 is considered very good. For a Gaussian
distribution with a standard deviation of s , its 1 e 2 radius is equal to 2s. Hence, if the
standard deviation s of a beam in a certain plane is calculated from its measured data,
then its beam radius is simply 2s. If s x and s y are the standard deviations along the x
and y axes, then the corresponding 1 e 2 beam widths are 2s x and 2s y . 40
properties of aberrated annular Gaussian beams and their line of sight in terms of their
centroids are discussed.
(
I (r; ) = 2 g exp - 2 g r2 ) [exp (-2 g ) 2
- exp ( - 2 g ) ] . (4-159)
1
As in Eq. (4-13), the irradiance is in units of Pex Sex . As g Æ 0 , I (r; ) Æ 1 - 2 , as ( )
expected for a uniformly illuminated pupil. Of course, Eq. (4-159) reduces to Eq. (4-13)
as Æ 0 . Figure 4-33a shows how the irradiance across a pupil with g = 1 and = 0.5
varies compared to a corresponding uniform annular pupil. It varies from a maximum
value of 2.57 at the inner edge of the pupil to a minimum value of 0.57 at its outer edge,
compared to a value of 1.33 for the corresponding uniform pupil. The ratio of the peak
values of the pupil irradiance for Gaussian and uniform annular beams is given by
(
Ig ( ) Iu ( ) = 2 g 1 - 2 ) {1 - exp [- 2 g (1 - ) ] } 2
. (4-160)
The variation of this ratio with is shown in Figure 4-33b for g = 0.5, 1, 2, and 3. It is
( )
evident that as g 1 - 2 increases, the ratio approaches a value of 2 g 1 - 2 . ( )
3 20
2.57 = 0.5
g =3
15
2 lg
u
/
g
10
lu 2
1
5
0.57 1
0.5
0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
r
(a) (b)
Figure 4-33. (a) Pupil irradiance distribution for uniform and Gaussian ( g = 1)
beams of a given total power Pex across an annular pupil with = 0.5 . (b) Ratio of
peak values of pupil irradiance as a function of for several values of g . The units
of pupil irradiance are Pex Sex , where Sex = p a 2 is the area of a circular pupil.
4.4 Systems with Annular Pupils 423
and
rc
(
P (rc ; g ; ) = p 2 2 )Ú I (r; g ; ) r dr . (4-162)
0
They differ from the corresponding Eqs. (4-22) and (4-25) for a circular pupil, in that
I (r; ) is different from I (r) and that the lower limit of integration is instead of zero.
Figure 4-34 shows the irradiance and encircled-power distributions for an annular
pupil with = 0.5 . Both Gaussian ( g = 1) and uniform annular pupils are considered in
this figure. Since Gaussian illumination broadens the central disc and reduces the power
in the secondary rings, and obscuration reduces the size of and power in the central disc
thereby increasing the power in secondary rings, we note that the differences between the
focal-plane distributions for Gaussian and uniform annular pupils are less compared with
those for corresponding circular pupils. However, note that the values of secondary
maxima decrease monotonically for a Gaussian beam; not so for a uniform beam. We
also note that the difference Pu - Pg between the encircled powers changes its sign from
positive to negative to positive, as rc increases.
The focal-point irradiance for an annular Gaussian beam may be obtained from
Eq. (4-154) by letting r = 0 . It is given by
The corresponding value for a uniform beam is equal to 1 - 2 . The ratio of the two is
given by
h = I (0; g ; ) I (0; 0; )
[(
= tanh g 1 - 2 2 ) ] [g (1 - ) 2] 2
. (4-164)
Figure 4-35a shows how h varies with g for several values of . It is evident that h
decreases as g increases, regardless of the value of . However, as shown in Figure 4-35b,
for a given value of g, h increases as increases. Note that for large values of g,
(
h Æ 2 g 1 - 2 ) . (4-165)
424 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
1.0
0.8 0.750
0.717
l(r; R); P(rc ; R)
0.6 Pg
0.4
Pu
= 0.5
0.2
lg
lu
0.0
0 1 2 3
r; rc
10 – 1
= 0.5
l(r; R)
10 – 2
10 – 3
10 – 4
0 1 2 3 4 5
r
1.0 1.0 0
1
0.8 0.8
0.6 0.6
h
h
0.75
2
0.4 0.5 0.4
=0
0.2 0.2 g =3
0.0 0.0
0 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0
g
(a) (b)
Figure 4-35. Focal-plane irradiance ratio for Gaussian and uniform beams as a
function of g and .
1 2p
g n
<Fn > = Ú Ú A(r) [F(r; q)] r dr d q , (4-166)
[ (
p exp - g 2
) - exp( -g ) ] 0
and
1 2p 2
g
[ ]
A(r) exp i F(r, q) r dr dq
]ÚÚ
S = , (4-167)
[ ( )
p exp - g 2 - exp( -g ) 0
The phase aberration function of a system with an annular exit pupil can be expanded
in terms of a complete set of Zernike-Gauss annular polynomials Z nm (r, q; g; ) that are
orthonormal over a unit annulus weighted by the Gaussian amplitude in the form
• n
F(r, q; g; ) = Â Â c nm Z nm (r, q; g; ) , £r £1 , 0 £ q £ 2p , (4-168)
n =0 m =0
where cnm are the expansion coefficients, n and m are positive integers including zero, n –
m ≥ 0, and
[
Z nm (r, q; g; ) = 2( n + 1) (1 + d m 0 ) ]1/ 2Rnm (r; g; ) cos mq . (4-169)
1 2p 1
Ú Ú Z n (r, q; g; ) Z n ¢ (r, q; g; ) A(r) r dr d q 2 p Ú A(r) r dr = d nn ¢ d mm ¢
m m¢
. (4-170)
0
È (n m) 2 ˘
Rnm (r; g ; ) = Mnm Í Rnm (r; ) - (n - 2i + 1) < Rnm (r; ) Rnm 2i (r; g ; ) > Rn
 2 i (r; g ; )˙ .
ÍÎ i ≥1 ˙˚
(4-171)
The angular brackets indicate an average over the annular Gaussian pupil; i.e.,
1 1
< Rnm (r; ) Rn 2 i (r; g; ) > = Ú Rnm (r; ) Rn 2 i (r; g; ) A(r) r dr Ú A(r) r dr . (4-172)
The normalization constant Mnm is chosen such that the radial polynomials obey the
orthogonality relation
1 1
1
Ú Rnm (r; g ; ) Rnm¢ (r; g ; ) A(r) r dr Ú A(r) r dr = d . (4-173)
n + 1 nn ¢
It should be noted that the radial polynomial Rnn (r; g ; ) is identical to the corresponding
polynomial for a uniformly illuminated annular pupil Rnn (r) , except for the normalization
constant, i.e.,
where the coefficients anm , etc., depend on g and . The number of Zernike (or
orthogonal) aberration terms in the expansion of an aberration function through a certain
order n is given by Eqs. (2-63a) and (2-63b), as in the case of circle or annular
polynomials for uniformly illuminated pupils. The Zernike-Gauss expansion coefficients
are given by
1 2p 1
c nm = Ú Ú F(r, q; g; ) Z nm (r, q; g; ) A(r) r dr d q 2 p Ú A(r) r dr , (4-176)
0
as may be seen by substituting Eq. (4-168) and utilizing the orthonormality of the
polynomials.
n (r, q; g; ) = c nm Z n (r, q; g; ) .
Fm m
(4-177)
= 0 , n π 0, m π 0 . (4-178)
For m = 0 , this may be seen with the help of Eq. (4-177) and the fact that Z 00 (r; g; ) = 1
is a member of the polynomial set. The orthogonality Eq. (4-173) yields the result that the
mean value of Rn0 (r; g ; ) is zero. When m π 0 , the average value of cos mq is zero.
Similarly, the mean square value of the aberration is given by
2 1 2p 2 1
[F m
n (r, q; g; )] = Ú Ú
0
[F m
n (r, q; g; )] A(r) r dr d q 2 p Ú A(r) r dr
2
= cnm . (4-179)
2
s 2nm = (F mn ) 2 - Fm
n
2
= cnm , n π 0, m π 0 . (4-180)
Thus, each expansion coefficient, with the exception of c00 , represents the standard
deviation of the corresponding aberration term.
From Eqs. (4-177) and (4-180), we note that the standard deviation of an aberration
can be obtained immediately by comparing its form with the corresponding orthonormal
aberration represented by Eq. (4-177), without having to calculate the integrals in Eqs. (4-
178) and (4-179). The variance of an aberration consisting of two or more terms of the
form of Eq. (4-177) is given by the sum of the variances of each of the aberration terms.
The variance of the aberration function is accordingly given by
428 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
2
2
sF = F 2 (r, q ) - F(r, q )
• n
2
= Â Â cnm . (4-181)
n =1 m = 0
ps = < rs >
Ë {
= Ê s exp g 1 - 2 [( )] - 1} {exp [g (1 - )] - 1}ˆ¯ + (s 2 g ) p
2
s 2 . (4-182)
Using these expressions, numerical results for the coefficients of the terms of a radial
polynomial for any values of g and can be obtained.
The coefficients for g = 1 and = 0, 0.25, 0.50, 0.75, and 0.90 are given in Table
4-12. For comparison, the coefficients for a uniformly illuminated pupil, i.e., for g = 0 ,
are given in parentheses in this table. An increase (decrease) in the value of a coefficient
anm of an orthogonal aberration Rnm (r; g ; ) cos mq implies a decrease (increase) in the
value of s F for a given amount of the corresponding classical aberration. This, in turn,
implies that for small aberrations, the system performance as measured by the Strehl ratio
is less (more) sensitive to that classical aberration when balanced with other classical
0.00 1.09367 2.04989 – 0.85690 1.14541 3.11213 – 1.89152 6.12902 – 5.71948 0.83368
0.25 1.04364 2.18012 – 1.00080 1.08940 3.01573 – 1.84513 6.95563 – 6.98197 1.25153
0.50 0.92963 2.70412 – 1.56449 0.93620 3.14319 – 2.06618 10.79549 – 13.08900 3.46706
0.75 0.80827 4.59329 – 3.51548 0.74439 4.55179 – 3.57767 31.47560 – 48.77879 18.39840
0.90 0.74453 10.53581 – 9.50324 0.63890 9.60573 – 8.69629 166.33359 – 300.66342 135.36926
4.4.5.1 PSF
The irradiance distribution in a defocused image plane given by Eq. (4-58) is now
replaced by
1 2
2
Ê 2R ˆ
I ( r; z; g; ) = Á
Ë z ¯
˜ Ú (
I (r; ) exp i Bd r 2
) J 0 (p rr) r dr . (4-183)
2
I (0; z; g ; ) = ( R z ) 2 g [ (B 2
d + g2 )] {coth [g (1 - )] 2
[ (
- cos Bd 1 - 2 )] [(
sinh g 1 - 2 )] } . (4-184)
430 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
0.6 1.2
0.514 = 0.5
0.4 0.8
P(rc)
l(r)
0.3 0.6
Pg
0.2 0.4
Pu
0.1 0.2
lg
lu
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5
r; rc
The positions of its maxima and minima are obtained by equating to zero its derivative
with respect to z. Thus, they are given by the solutions of
Ê lz ˆ
2Á
B
{ [( )]
- 2 d 2 ˜ cosh g 1 - 2 - cos Bd 1 - 2
Ë Sex Bd + g ¯
[ ( ) ]}
( ) (
= - 1 - 2 sin ÈB d 1 - 2 ˘ .
ÎÍ ˚˙ ) (4-185)
Figure 4-37 shows how the axial irradiance of a focused Gaussian beam differs from
that of a focused uniform beam when g = 1, = 0.5 , and Fresnel number N = 1, 10, 100.
We note that, as in the case of a circular beam, the principal maximum is higher for the
uniform beam compared with that for the Gaussian beam. However, the secondary
maxima are higher for the Gaussian beam. Moreover, whereas the axial minima for the
uniform beam have a value of zero, the minima for the Gaussian beam have nonzero
values. For a given value of , the locations of maxima and minima, except the principal
maximum, are nearly the same for the two beams. The effect of the obscuration is to
reduce the irradiance at the principal maximum but to increase it at the secondary
maxima. Also, the maxima and minima occur at smaller z values for an annular beam.
These z values correspond approximately to those axial points at which the annular pupil
subtends an odd or an even number of Fresnel’s half-wave zones, respectively. We note
that the curves become symmetric about the focal point z = R as N increases.
4.4 Systems with Annular Pupils 431
N=1 N = 10 N = 100
1.5 = 0.5 1.5 = 0.5 1.5 = 0.5
l (0; z)
lu
1.0 1.0 1.0
lg
0.5 0.5 0.5
Figure 4-37. Axial irradiance of an annular beam focused at a fixed distance R with
a Fresnel number N z = a 2 l R = 1 , 10, 100. The irradiance is in units of the focal-
point irradiance for a uniform annular beam. The Gaussian beam results shown in
this figure by the dashed curves are for g = 1.
Once again, even though the principal maximum of axial irradiance does not lie at
the focus, maximum central irradiance on a target at a given distance from the pupil is
obtained when the beam is focused on it. Figure 4-38 illustrates how the central
irradiance on a target at a fixed distance z varies when the beam is focused at various
distances R along its axis. The irradiance in this figure is in units of Pex Sex l2 z 2 . The
quantity N z = a 2 l z represents the Fresnel number of a circular pupil as observed from
the target. As in Figure 4-34, the maximum irradiance values for uniform and Gaussian
( g = 1) beams are 0.750 and 0.717, respectively, when = 0.5 . We note that as N z
increases, the curves become symmetric about R = z .
lg
0.6 0.6 0.6
z
Figure 4-38. Central irradiance on a target at a fixed distance z from the plane of the
annular exit pupil when a beam is focused at various distances R. The quantity
N z = a 2 l z represents the Fresnel number of the pupil as observed from the target.
The irradiance is in units of Pex Sex l 2 z 2 . The Gaussian beam results shown in this
figure by dashed curves are for g = 1.
432 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
1 2
I (r; z; g ; ) = 4 Bd2 Ú ( 2
)
I (r; ) exp i Bd r J 0 ( p r r) r dr , (4-186)
Bd = Sex l z (4-187)
and represents the phase aberration of a plane wavefront with respect to a reference
sphere centered at a distance z and passing through the center of the pupil. In Eq. (4-79),
the irradiance in both the pupil and the observation planes is in units of the pupil
irradiance Pex Sex for a uniform circular beam.
In the far field, i.e., for z ≥ D2 l , the irradiance and encircled-power distributions
are similar to the focal-plane distribution. The only difference is in scaling of the
diffraction pattern. Similarly, in the near field, i.e., for z < D2 l , these distributions
correspond to defocused distributions. The only significant difference is in the definition
of Bd .
{ [1 + (4g z p) ] }
I (0; z; g ; ) = 2 g
2
= - (1 - ) sin [ p (1 - ) 4 z ]} .
2 2
(4-189)
Figure 4-39 illustrates how the axial irradiance of collimated uniform and Gaussian
annular beams varies with distance z from the pupil for = 0.5 . It is similar to Figure
4-19 for a circular beam. Note, however, that the maxima are higher and minima are
lower for the annular beam.
4.4 Systems with Annular Pupils 433
N=0
5
= 0.5
lg
4
(0;z)
lu
2
0
0.0 0.1 0.2 0.3 0.4 0.5
z
Figure 4-39. Axial irradiance of a collimated annular Gaussian beam with = 0.5
and g = 1 compared with that for a corresponding uniform beam.
2
1 2p
Û Û Ï R ¸
¥ Ù Ù I (r; ) exp Ìi ÈÍ F(r, q; g ; ) - p r r cos (q - q i )˘˙ ˝ r dr dq , (4-190)
ı ı Ó Î z ˚ ˛
0
where
and
1 2
Û
I (0; z; g ; ) = 4 Ù
ı
( ) [ ]
I (r; ) exp iBd r2 exp i An 0 Rn0 (r; g ; ) r dr , (4-192)
respectively. As in Eq. (4-88), the Fresnel number in Eq. (4-192) is assumed to be large
so that z ~ R . Once again, the axial irradiance for spherical aberration is not symmetrical
434 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
about any point. This is illustrated in Figure 4-40 for primary spherical aberration As r 4
for several values of As , = 0.5 , and g = 1. Moreover, from the form of the Zernike-Gauss
polynomial for astigmatism, the point of axial symmetry for astigmatism Aa r2 cos 2 q is
the same as for circular pupils, namely, Bd = - Aa / 2 .
1 2p
2F Ê ∂W sin q ∂W ˆ
xi =
p Ú Ú I (r; ) ÁË cos q ∂r -
r
˜ r dr d q
∂q ¯
(4-193)
0
and
1.0
1
W(r) = Asr4 As = 0
I(r) = I0exp(– 2r2)
0.8 2
= 0.5
(0; Bd)
0.6
3
0.4
0.2
0.0
–6 –5 –4 –3 –2 –1 0 1 2 3
Bd
Figure 4-40. Axial irradiance of an annular Gaussian beam with = 0.5 and g = 1
aberrated by spherical aberration. The quantity I0 represents the coefficient of
( )
exp - 2r2 on the right-hand side of Eq. (4-152) for g = 1, and the aberration-free
central value of 0.717 has been normalized to unity.
4.5 Line of Sight of an Aberrated System 435
1 2p
2F Ê ∂W cos q ∂W ˆ
yi =
p Ú Ú I (r; ) ÁË sin q ∂r -
r
˜ r dr d q ,
∂q ¯
(4-194)
0
where W (r, q; g ; ) is the aberration function. Expanding the aberration function in terms
of Zernike-Gauss polynomials, we may write
• n 12
W (r, q; g ; ) = Â Â [2(n + 1) (1 + d m 0 )] Rnm (r; g ; )
n=0 m=0
where cnm and snm are the aberration coefficients. Substituting Eq. (4-195) into
Eqs. (4-193) and (4-194), we obtain
1
2 FSex • ∂
xi , yi = Â ¢ 2(n + 1) (c n1, sn1)
Pex n =1 Ú I (r; ) ∂r [r R1n (r; g; )] r dr . (4-196)
We note that, as in the case of systems with uniformly illuminated circular and annular
pupils, the only aberrations that contribute to the centroid are those with m = 1.
Aberrations that vary with q as cos q contribute to < xi > , and those that vary as sin q
contribute to < yi > . This may also be seen from the symmetry of these aberrations in a
manner similar to the discussion following Eq. (2-151).
We now consider aberrations given by Eq. (2-152). Substituting this equation into
Eq. (4-4) and following the procedure of Section 2.9.1, it can be shown that the irradiance
distribution of the aberrated image of a point object may be written
2
È1 ˘
I (r; q i ; g ; ) = 4 Í Ú I (r; ) J 0 ( p B) r dr ˙ , (4-197)
Î ˚
where B = (2tW3 - x ) t 1 2 , as in Eq. (2-164). Substituting Eq. (4-159) and the aberration
W = W3 r3 cos q into Eq. (4-193), we obtain
xi
Ï
Ô 1
= 4 W 3F Ì +
[ (
2 exp 2 g 1 - 2 - 1 ¸Ô
˝ .
)] (4-199)
ÔÓ 2 g [ (
exp 2 g 1 - 2 - 1 Ô
˛ )]
From the form of the radial polynomial R31 (r; g ; ) given in Table 4-3, the point x m in the
image plane with respect to which the aberration variance is minimum is given by
2W3 F p4 p2 , where ps is given by Eq. (4-182). Thus,
Ï
Ô2
xm = 2W 3F Ì +
g 4 exp g 1 - 2 [( )] - 1 ¸Ô
˝ . (4-200)
ÔÓ g (
1 + g 2 exp g 1 - 2 ) [( )] - (1 + g ) Ô˛
For small values of W3 , the peak value of the aberrated PSF occurs at this point.
Figure 4-41 shows how I ( x ) varies with x for several values of W3 when g = 1 and
= 0 or 2 = 0.5 . The values of x m , x p , and < x > and the corresponding irradiances
Im , I p ,and Ic for these values of W3 are given in Table 4-13. The values of I (0) are also
included in this table. The irradiances given here are normalized by the aberration-free
central irradiance for the annular Gaussian pupil.
1.0 1.0
W3 = 0 W3 = 0
= 0 2 =
0.5 0.5
0.8 g= 1 0.8 g= 1
0.5
1
0.6 0.6
l
1
1.5
0.4 0.4
2
1.5
2
0.2 0.2
0.0 0.0
3 2 1 0 1 2 3 3 2 1 0 1 2 3
x x
Figure 4-41. PSF for a Gaussian pupil with g = 1 aberrated by primary coma W3 (in
units of l).
4.5 Line of Sight of an Aberrated System 437
Table 4-13. Typical values of x m , x p , and < x > in units of lF and corresponding
irradiances Im , I p , and Ic in units of the aberration-free central irradiance for
Gaussian pupils with g = 1 aberrated by primary coma.*
W3 xm xp <x> Im Ip Ic I(0)
0 0 0 0 1 1 1 1
(0) (0) (0) (1) (1) (1) (1)
xi = 6W F
( 4
+ 2 g 1
) [ ( )] (
+ 1 2 g 2 exp 2 g 1 - 2 - 1 + g 1
+ 1 2g2 ). (4-201)
exp [2 g (1 - )] - 1
5 2
Figure 4-42 shows how I ( x ) varies with x for several values of W5 when g = 1 and = 0
or 2 = 0.5 . The values of x p , < x > , I p , Ic , and I (0) are given in Table 4-14. It is
evident from the data given in Tables 4-13 and 4-14 that the centroids of two PSFs
1.0 1.0
W5 = 0 W5 = 0
= 0 2 = 0.5
0.8 g= 1 0.8 g= 1
0.5 0.5
0.6 0.6
l
1 1.5
1
0.4 0.4
2
1.5
0.2 0.2 2
0.0 0.0
3 2 1 0 1 2 3 3 2 1 0 1 2 3
x x
Figure 4-42. PSF for a Gaussian pupil with g = 1 aberrated by secondary coma W5
(in units of l).
438 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
Table 4-14. Typical values of x p and < x > in units of lF and corresponding
irradiances I p and Ic in units of the aberration-free central irradiance of PSFs for
Gaussian pupils with g = 1 aberrated by secondary coma.*
W5 xp <x> Ip Ic I(0)
0 0 0 1 1 1
(0) (0) (1) (1) (1)
4.6 SUMMARY
In this chapter, we have compared the effects of diffraction, obscuration, and
aberrations on the propagation of uniform and Gaussian beams of a fixed total power.
The following general conclusions can be drawn from the discussion given here.
i. PSF
For a given total beam power Pinc incident on a pupil of fixed radius a, the fractional
transmitted power Pex Pinc increases as the Gaussian beam radius w decreases (see
Figure 4-2), but the corresponding focal-point irradiance decreases (see Figure 4-4).
Hence, for a given total power Pinc , there is an optimum value of w that yields the
maximum focal-point irradiance. This optimum value is obtained for g = 1.255 or
w = 0.893a (see Figure 4-5).
The effect of the Gaussian apodization of the pupil is to increase the size of the
central bright disc of the PSF and to decrease the power in the rings of the diffraction
pattern (see Figure 4-3). The focal-point irradiance for a focused Gaussian beam is
smaller than the corresponding value for a uniform beam of the same total power. It
4.6 Summary 439
decreases as g increases (see Figure 4-3). The encircled power for small circles is higher
for a uniform beam, but for large circles it is higher for a Gaussian beam (see Figure 4-3).
The minima of axial irradiance for a uniform beam have a value of zero, and those
for a Gaussian beam have nonzero values (see Figure 4-15). The Huygens’ secondary
wavelets do not cancel each other completely due to their different amplitudes. Whereas
the principal maximum of axial irradiance for a Gaussian beam has a smaller value than
that for a uniform beam, the secondary maxima for a Gaussian beam have higher values.
The principal maximum for a beam with a small Fresnel number lies closer and closer to
the pupil as g increases (see Figure 4-19). As for a uniform beam in such cases, even
though the principal maximum does not occur at the focus, maximum central irradiance
and encircled power are obtained on a target at a given distance from the pupil when the
beam is focused on it (see Figure 4-18).
The depth of focus increases as g increases (see Figure 4-17). For a < ~ w , the
Gaussian beams are somewhat less sensitive to aberrations than uniform beams.
Accordingly, the aberration tolerance is somewhat higher for the Gaussian beams.
However, this tolerance increases rapidly as a becomes increasingly larger than w. This is
understandable since, for a > > w , the power in the pupil is concentrated in a small region
near its center; and therefore the aberration in its outer region has little effect on the
irradiance distribution. The diffraction focus in the case of spherical aberration lies closer
and closer to the pupil compared to that for a uniform pupil as g increases. Similarly, it
lies closer to the Gaussian image point in the Gaussian image plane as g increases (see
Table 4-6). Whereas equal amounts of primary and secondary coma yield the same
centroid in the case of a uniform circular beam, they yield different centroids in the case
of a Gaussian beam. The objective of the Gaussian apodization is to decrease the
secondary maxima of the aberration-free PSF. However, this advantage is lost when, for
example, spherical aberration is present, unless its amount is less than about a quarter
wave (see Figure 4-27d).
spherical aberration and least sensitive to astigmatism, rather than being most sensitive to
coma and least sensitive to spherical aberration.
The imaging of a Gaussian beam by a lens, where its beam waist is the object, is
different from that of conventional imaging. For example, when the incident beam waist
lies in the front focal plane of the lens, the image waist lies in the back focal plane instead
of infinity, as in conventional imaging.
M 2 is a measure of the quality of a multimode beam, and the higher its value the
poorer its quality. M represents the factor by which its divergence is larger than that for a
corresponding Gaussian (single mode) beam. A beam with a value of M 2 lying between
1 and 1.2 is considered to be a high-quality beam.
v. OTF
The OTF of a Gaussian pupil is higher for low spatial frequencies and lower for the
high compared to that for a uniform pupil. Moreover, as g increases, the bandwidth of
low frequencies for which the OTF is higher decreases, and the OTF at high frequencies
becomes increasingly smaller (see Figure 4-6). Since the PSF for a weakly truncated
beam is Gaussian, its OTF is also Gaussian [see Eq. (4-138)].
The effect of a central obscuration in the pupil is opposite to that of the Gaussian
apodization. It reduces the size of the central disc and increases the power in the rings of
the diffraction pattern. Accordingly, as the obscuration increases, the difference between
the diffraction effects of uniform and Gaussian beams decreases (see Figure 4-34).
References 441
REFERENCES
9. V. N. Mahajan, “Strehl ratio of a Gaussian beam,” J. Opt. Soc. Am. A 22, 1824–
1833 (2005).
11. V. N. Mahajan, “Strehl ratio for primary aberrations in terms of their aberration
variance,” J. Opt. Soc. Am. 73, 860–861 (1983).
12. G. A. Korn and T. M. Korn, Mathematical Handbook for Scientists and Engineers
(McGraw-Hill, New York, 1968), p. 454.
13. V. N. Mahajan, “Axial irradiance and optimum focusing of laser beams,” Appl.
Opt. 22, 3042–3053 (1983).
14. V. N. Mahajan “Axial irradiance of a focused beam,” J. Opt. Soc. Am. A 22,
1814–1823 (2005).
442 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
18. Y. Li, “Variations of the axial intensity pattern formed by a focused truncated
Gaussian beam,” Opt. Commun. 68, 324–328 (1988).
19. Y. Li, “Encircled energy of diffracted converging spherical waves,” J. Opt. Soc.
Am. 73, 1101–1104 (1983).
20. K. Tanaka and O. Kanzaki, “Focus of a diffracted Gaussian beam through a finite
aperture lens: experimental and numerical investigations,” Appl. Opt. 26, 390–395
(1987).
24. Y. Li, “Establishment of the maximum encircled energy in the geometrical focal
plane,” Optica Acta 31, 1107–1118 (1984).
25. A. Yoshida and T. Asakura, “Propagation and focusing of Gaussian laser beams
beyond the conventional diffraction limit,” Opt. Commun. 123, 694–704 (1996).
26. D. Y. Jiang and J. J. Stamnes, “Focusing at low Fresnel numbers in the presence
of cylindrical or spherical aberration,” Pure Appl. Opt. 6, 85–96 (1997).
27. A. E. Siegman, Lasers (University Science Books, Mill Valley, CA, 1971).
28. J. Gaskill, Linear System, Fourier Transforms, and Optics (McGraw Hill, New
York, 1978).
29. P. Kuttner, “Image quality of optical systems for truncated Gaussian laser beams,”
Opt. Eng. 25, 180–183 (1986).
31. S. A. Self, “Focusing of spherical Gaussian beams,” Appl. Opt. 22, 658–661
(1983).
32. R. Herloski, “Strehl ratio for untruncated Gaussian beams,” J. Opt. Soc. Am. A 2,
1027–1030 (1985).
33. E. Stijns, “Measuring the spot size of a Gaussian beam with an oscillating wire,”
IEEE J. QE. 16, 1298–1299, (1980).
35. S. Nemoto and T. Makimoto “Generalized spot size for a higher-order beam
mode,” J. Opt. Soc. Am. 60, 578–580 (1979).
37. H. Kogelnik and T. Li, “Laser beams and resonators,” Appl. Opt. 24, 1346–1354
(1996).
38. A. E. Siegman, An Introduction to Lasers and Masers (McGraw Hill, New York,
1986).
41. V. N. Mahajan, “Line of sight of an aberrated optical system,” J. Opt. Soc. Am. A
2, 833–846 (1985).
444 OPTICAL SYSTEMS WITH GAUSSIAN PUPILS
PROBLEMS
2. (a) Show that the variance of a primary aberration in a system with a circular pupil
and Gaussian illumination with g = 1 is minimized when balanced as follows:
(
F bs (r) = As r 4 - 0.933r2 ) ,
( )
F bc (r, q) = Ac r3 - 0.608r cos q ,
and
(
F ba (r, q) = Aa r2 cos 2 q - 0.5r2 ) .
(b) Repeat problem (a) for g = 4 . (c) Repeat problem (a) for an annular pupil with
= 0.5 . Use the results given in Table 4-8. (d) Compare the results obtained in (a)
and (c) with those for a uniformly illuminated circular pupil.
3. It is shown in Section 4.3.2.2 that the optimum ratio of the beam and circular pupil
radii for maximum focal-point irradiance is 0.893. Determine the optimal ratio for
an annular pupil and plot the focal-point irradiance as a function of the obscuration
ratio [see D. A. Holmes, P. V. Avizonis, and K. H. Wrolstad, “On-axis irradiance
of a focused, apertured Gaussian beam,” Appl. Opt. 9, 2179-2180 (1970)].
4. Theorem 11 in Chapter 1 states that the slope of the real part of the OTF at the
origin is independent of the aberration. Demonstrate this by considering the OTF of
a defocused Gaussian pupil given by Eq. (4-75). Show that the slope at the origin is
given by 8g p [1 - exp( 2g ) ] , independent of the defocus coefficient. Show further
that, for large values of g , this result reduces to the slope obtained from the
corresponding OTF given by Eq. (4-31). Moreover, show that, as g Æ 0, the slope
reduces to the value obtained from the defocused OTF for a uniform pupil given by
Eq. (2-195), or from the astigmatism OTF given in Problem 9 of Chapter 2.
CHAPTER 5
RANDOM ABERRATIONS
445
446 RANDOM ABERRATIONS
We discuss the effects of two types of random aberrations: random wavefront tilt
causing random image motion, and random aberrations introduced by atmospheric
turbulence. The time-averaged Strehl ratio, point-spread function (PSF), optical transfer
function (OTF), and encircled power are discussed for the two types of aberrations. A
coherence length of atmospheric turbulence is defined, which limits the resolution of an
imaging system, regardless of how large its aperture is. Both long- and short-exposure
images are discussed, and expressions for the aberration variance are given in both cases.
These expressions can be used to define the requirements of a steering mirror for
corrections of wavefront tilt and a deformable mirror for corrections of wavefront
deformation or the aberrations, as in adaptive optics. The fluctuations of the angle of
arrival of the wave are considered both in terms of the wavefront tilt as well as the
centroid of the aberrated PSF. It is shown that in severe turbulence, a short-exposure
image breaks up into speckles of roughly the same size, which is determined by the
turbulence-free resolution of the system. Although much of our discussion is on systems
with circular pupils, systems with annular pupils are also considered, and differences
between the two types are outlined.
447
448 RANDOM ABERRATIONS
OTF, and encircled power for an imaging system with a circular exit pupil undergoing
Gaussian random motion. A simple approximate model based on a Gaussian
approximation of its motion-free PSF is also developed, and numerical results provided
by it are compared with the exact results. It is shown that the Strehl ratio is more sensitive
to obscuration in the case of transverse motion, but less sensitive in the case of
longitudinal motion.
where angular brackets indicate a time average. Thus, the time-averaged irradiance
distribution of the image formed by an imaging system undergoing random motion is
given by the convolution of its motion-free distribution and the probability density
function describing its motion. The corresponding PSF is obtained by dividing both sides
of Eq. (5-1) by Pex , where Pex is the total power in the exit pupil of the system and,
therefore, in the image. Following Eq. (1-56c), we note that the time-averaged PSF can
also be interpreted as the irradiance distribution of the image of an incoherent object
r
whose Gaussian-image irradiance distribution is given by p ( ri ) . Fourier transforming Eq.
(5-1), we obtain
r r r
t ( vi ) = t ( vi ) P ( vi ) , (5-2)
r r
where t ( vi ) and t ( vi ) are the motion-free and time-averaged OTFs, respectively,
r
corresponding to image spatial frequency v i , and
r r r r r
P ( vi ) = Ú p ( ri ) exp (2 p i vi ◊ ri ) d ri (5-3)
( ) (
p (ri ) = 1 / 2 p s i2 exp - ri2 / 2s i2 ) , (5-5)
r r
where ri = ri . Substituting Eq. (5-5) into Eq. (5-3) and letting vi = vi , Eq. (5-2) for the
time-averaged OTF may be written
r r
(
t ( vi ; s i ) = t ( vi ) exp - 2 p 2 s i2 vi2 ) . (5-6)
5.2.2.2.1 Theory
Consider systems with circular pupils and radially symmetric motion-free PSFs and,
therefore, radially symmetric motion-free OTFs. Substituting Eq. (5-6) into Eq. (5-4) and
using Eq. (2-12), we obtain the time-averaged irradiance
where
(
t (vi ; s i ) = t (vi ) exp - 2 p 2 s i2 vi2 ) (5-8)
where
(
t (v; s ) = t (v) exp - 2 p 2 s 2 v 2 ) (5-10)
is the time-averaged OTF. The motion-free OTF t (v) is given by Eq. (2-44).
Defining the Strehl ratio of the image as the ratio of the central irradiances with and
without image motion, the time-averaged Strehl ratio can be written
S (s ) = I (0; s ) I (0)
1
= 8 Ú t (v; s ) v dv .
0
(5-11)
450 RANDOM ABERRATIONS
Note that the motion-free central irradiance I(0) is unity in units of Pex Sex l2 R 2 . ( )
The encircled power in terms of the OTF is given by Eq. (1-89). Accordingly, the
time-averaged fractional encircled power is given by
1
P (rc ; s ) = 2 p rc Ú t (v; s ) J1 (2 p rc v) dv . (5-12)
0
This result may also be obtained by substituting Eq. (5-9) into Eq. (2-22b) and using Eq.
(1-88). Numerical results for an aberration-free system undergoing Gaussian random
motion may be obtained by substituting Eq. (2-44) into Eq. (5-10) and substituting the
expression obtained into Eqs. (5-9), (5-11), and (5-12). However, the integration in Eq.
(5-11) thus obtained can be carried out analytically, yielding the results
( ){
S (s ) = 2 p 2 s 2 1 - exp - p 2 s 2 ( ) [ I (p s ) + I (p s )] }
0
2 2
1
2 2
, (5-13)
where I0 (∑) and I1 (∑) are the hyperbolic Bessel functions of zero order and first order,
respectively. In the following, we consider an aberration-free system undergoing random
Gaussian motion as described above.
[
Ig (r ) = exp - ( p r 2)
2
] . (5-14)
The corresponding approximation of the OTF is given by its (slightly modified) zero-
order Hankel transform according to Eq. (2-49a), i.e.,
(
t g (v) = exp - 4v 2 ) . (5-15)
t g (v; s ) [
= exp - 2 p 2 s 2 + 2 p 2 v 2 ( ) ] . (5-16)
The time-averaged irradiance distribution is also Gaussian with a variance equal to the
sum of the variances of the motion-free distribution and the image motion; i.e.,
1
I g ( r; s ) = [ ( p 2 ) (s
2 2
+ 2 p2 )] [ (
exp - r 2 2 s 2 + 2 p 2 )] . (5-17)
Sg (s ) = Ig (0; s )
1
(
= 1 + p2s2 2 ) . (5-18)
Sg (s ) Æ 2 p 2 s 2 . (5-19)
This result may also be obtained from Eq. (5-13), since, for large arguments, the
hyperbolic Bessel functions approach
12
I m ( z ) Æ e z (2 p z ) . (5-20)
[
= 1 - exp - rc2 2 s 2 + 2 p 2 ( )] . (5-21)
Figure 5-1 shows how the exact and approximate time-averaged Strehl ratios given
by Eqs. (5-13) and (5-18), respectively, vary with s . The exact results are shown by the
solid curves and the corresponding approximate results are shown by the dashed curves.
The values of the time-averaged Strehl ratio are given in Table 5-1 for s = 0 (0.05) 1. We
note that the Gaussian approximation given by Eq. (5-18) overestimates the Strehl ratio.
( )
However, the maximum fractional difference S - Sg S is less than 12% and occurs at
s = 1.1. The time-averaged irradiance distributions obtained according to Eqs. (5-9) and
(5-17) for typical values of s are shown in Figure 5-2. The corresponding encircled
powers obtained by use of Eqs. (5-12) and (5-21) are shown in Figure 5-3. Once again,
we note that the Gaussian approximation overestimates the irradiance and encircled
power (at least within the Airy disc). However, the differences between the exact and
approximate results are not large. In the case of encircled power, the difference increases
monotonically as the radius of the circle increases.
452 RANDOM ABERRATIONS
1.0 1.0
0.8 0.8
0.6 0.6
S
S
0.4 0.4
<S> <Sg>
0.2 0.2
<S>
<Sg>
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0 1 2 3 4 5
s s
Figure 5-1. Time-averaged Strehl ratio for transverse Gaussian random image
motion s in units of lF. The dashed curves represent the Gaussian approximation
of the exact results represented by the solid curves.
Table 5-1. Time-averaged Strehl ratio for various values of the transverse Gaussian
image motion characterized by standard deviation s in units of lF along each of
the two axes of the image plane.
s S Sg
0 1 1
0.05 0.988 0.988
0.10 0.953 0.953
0.15 0.898 0.900
0.20 0.831 0.835
0.25 0.756 0.764
0.30 0.680 0.693
0.35 0.606 0.623
0.40 0.538 0.559
0.45 0.476 0.500
0.50 0.422 0.448
0.55 0.375 0.401
0.60 0.334 0.360
0.65 0.298 0.324
0.70 0.268 0.293
0.75 0.241 0.265
0.80 0.218 0.241
0.85 0.198 0.219
0.90 0.181 0.200
0.95 0.165 0.183
1 0.152 0.169
5.2 Random Image Motion 453
1.0 1.0
s= 0 s = 0.25
0.8 0.8
0.6 0.6
0.4 0.4
<I>
0.2 0.2
<I>
<Ig>
<Ig>
0.0 0.0
0 1 2 3 0 1 2 3
r r
1.0 1.0
s = 0.5 s = 0.75
0.8 0.8
0.6 0.6
<Ig>
0.4 0.4
<I>
<I>
0.0 0.0
0 1 2 3 0 1 2 3
r r
1.0
s= 1
0.8
0.6
<I>
0.4
0.2
<Ig>
0.0
0 1 2 3
r
1.0 1.0
<Pg> <Pg>
0.8 <P> 0.8 <P>
0.6 0.6
P(rc)
P(rc)
0.4 0.4
0.2 0.2
s= 0 s = 0.25
0.0 0.0
0 1 2 3 0 1 2 3
rc rc
1.0 1.0
<Pg>
<Pg>
0.8 0.8
<P> <P>
0.6 0.6
P(rc)
P(rc)
0.4 0.4
0.2 0.2
s = 0.5 s = 0.75
0.0 0.0
0 1 2 3 0 1 2 3
rc rc
1.0
0.8 <Pg>
<P>
0.6
P(rc)
0.4
0.2
s= 1
0.0
0 1 2 3
rc
Figure 5-3. Time-averaged encircled power for transverse Gaussian random image
motion s. The solid and dashed curves represent the exact and approximate results,
respectively.
5.2 Random Image Motion 455
5.2.2.3.1 Theory
I i ( ri ; ; s i ) = 2p Pex Ú t ( v i ; ; s i ) J 0 ( 2p ri v i ) v i dv i , (5-22)
where
(
t ( v i ; ; s i ) = t ( v i ; ) exp - 2p 2 s i2 v i2 ) (5-23)
is the time-averaged OTF of the system. The corresponding PSF is obtained by dividing
both sides of Eq. (5-22) by Pex . The motion-free OTF t (vi ; ) is discussed in Section
3.2.6. Using normalized quantities defined by Eqs. (3-10a), (3-11), and (2-47), and letting
s = s i l F , Eq. (5-22) becomes
1
[ (
I (r; ; s ) = 8 1 - 2 )] Ú
0
t (v; ; s ) J 0 (2 p rv) v dv , (5-24)
where
(
t (v; ; s ) = t (v; ) exp - 2 p 2 s 2 v 2 ) . (5-25)
The motion-free OTF t (v; ) is given by Eq. (3-40). The Strehl ratio of the image, i.e.,
the ratio of the central irradiances with and without image motion, is given by
Note that the motion-free central irradiance I (0;) is unity in units of Pex Sex ( ) l2 R 2 .
The encircled power in terms of the OTF is given by Eq. (1-89). Following this
equation and Eq. (3-20), the time-averaged fractional encircled power may be written
If we let = 0 , Eqs. (5-24) through (5-27) for the time-averaged irradiance, OTF, Strehl
ratio, and encircled power reduce to the corresponding Eqs. (5-9) through (5-12) for a
system with a circular exit pupil.
Figure 5-4 shows how the Strehl ratio varies with s for = 0 (0.25) 0.75 . It
decreases monotonically as s increases. Numerical values of the Strehl ratio for
s = 0 (0.05) 1 and = 0 (0.05) 0.95 are given in Table 5-2. We note that as increases,
the drop in Strehl ratio due to image motion for a given value of s increases. This occurs
because the motion-free PSF (normalized to unity at the origin) for a larger value of is
smaller for small values of r for r £ 1. Figure 5-5 shows how the irradiance distribution is
affected by the image motion. We note that as s increases, the central irradiance
decreases, the minima of the distribution are filled in, and the distribution becomes
smoother. The ring structure of the motion-free image disappears for s ≥ 0.5 , which is
approximately the spacing between adjacent maxima and minima.
Figure 5-5 also shows how the encircled power changes due to the image motion. We
note that the power contained in small circles decreases as s increases, because power
1.0
0.8
0.6 = 0.00
< S(; s)>
0.25
0.4 0.50
0.75
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
s
Figure 5-4. Time-averaged Strehl ratio as a function of s for several typical values
of .
5.2 Random Image Motion 457
Table 5-2. Time-averaged Strehl ratio for transverse Gaussian random image motion
characterized by s in units of lF.
s \ 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
0 1 1 1 1 1 1 1 1 1 1
0.05 0.988 0.988 0.988 0.987 0.987 0.987 0.987 0.986 0.986 0.985
0.10 0.953 0.952 0.952 0.952 0.951 0.950 0.948 0.947 0.945 0.943
0.15 0.898 0.898 0.898 0.896 0.895 0.892 0.890 0.887 0.883 0.879
0.20 0.831 0.831 0.830 0.828 0.825 0.822 0.817 0.813 0.807 0.801
0.25 0.756 0.756 0.754 0.752 0.748 0.744 0.738 0.732 0.724 0.716
0.30 0.680 0.679 0.678 0.674 0.670 0.665 0.658 0.650 0.642 0.632
0.35 0.606 0.606 0.603 0.600 0.595 0.589 0.582 0.573 0.564 0.554
0.40 0.538 0.537 0.535 0.531 0.526 0.520 0.512 0.503 0.494 0.483
0.45 0.476 0.476 0.473 0.469 0.464 0.458 0.450 0.442 0.433 0.423
0.50 0.422 0.421 0.419 0.415 0.410 0.404 0.397 0.389 0.380 0.371
0.55 0.375 0.374 0.372 0.368 0.364 0.358 0.351 0.244 0.336 0.328
0.60 0.334 0.333 0.331 0.328 0.323 0.318 0.312 0.306 0.299 0.293
0.65 0.298 0.298 0.296 0.293 0.289 0.285 0.279 0.274 0.268 0.263
0.70 0.268 0.267 0.265 0.263 0.260 0.256 0.251 0.247 0.242 0.237
0.75 0.241 0.241 0.239 0.237 0.234 0.231 0.227 0.223 0.220 0.216
0.80 0.218 0.218 0.216 0.215 0.212 0.209 0.206 0.203 0.201 0.198
0.85 0.198 0.198 0.197 0.195 0.193 0.191 0.188 0.186 0.184 0.182
0.90 0.181 0.180 0.179 0.178 0.176 0.175 0.173 0.171 0.169 0.168
0.95 0.165 0.165 0.164 0.163 0.162 0.160 0.159 0.158 0.157 0.156
1.00 0.152 0.152 0.151 0.150 0.149 0.148 0.147 0.146 0.145 0.145
s \ 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95
0 1 1 1 1 1 1 1 1 1 1
0.05 0.985 0.984 0.983 0.983 0.982 0.981 0.980 0.979 0.979 0.981
0.10 0.941 0.939 0.936 0.933 0.930 0.927 0.924 0.920 0.917 0.916
0.15 0.875 0.870 0.865 0.859 0.853 0.847 0.841 0.834 0.827 0.823
0.20 0.794 0.787 0.779 0.771 0.762 0.752 0.743 0.733 0.723 0.715
0.25 0.708 0.698 0.688 0.677 0.666 0.655 0.643 0.630 0.618 0.608
0.30 0.622 0.611 0.599 0.587 0.575 0.562 0.549 0.536 0.524 0.513
0.35 0.543 0.531 0.519 0.507 0.494 0.481 0.468 0.456 0.443 0.433
0.40 0.472 0.461 0.449 0.437 0.425 0.413 0.402 0.390 0.379 0.370
0.45 0.412 0.402 0.391 0.380 0.369 0.359 0.348 0.338 0.328 0.321
0.50 0.362 0.352 0.343 0.333 0.324 0.315 0.306 0.297 0.289 0.283
0.55 0.320 0.312 0.304 0.296 0.288 0.280 0.273 0.265 0.258 0.253
0.60 0.286 0.279 0.272 0.265 0.259 0.252 0.246 0.240 0.234 0.229
0.65 0.257 0.251 0.246 0.240 0.235 0.229 0.224 0.218 0.212 0.203
0.70 0.233 0.228 0.224 0.219 0.215 0.210 0.206 0.201 0.196 0.187
0.75 0.212 0.209 0.205 0.202 0.198 0.194 0.190 0.186 0.181 0.174
0.80 0.195 0.192 0.189 0.187 0.184 0.181 0.177 0.174 0.169 0.163
0.85 0.180 0.178 0.176 0.173 0.171 0.169 0.166 0.163 0.159 0.153
0.90 0.166 0.165 0.164 0.162 0.160 0.158 0.156 0.153 0.150 0.144
0.95 0.155 0.154 0.153 0.152 0.150 0.149 0.147 0.144 0.141 0.136
1.00 0.144 0.144 0.143 0.143 0.142 0.140 0.139 0.137 0.134 0.131
458 RANDOM ABERRATIONS
1.0
<I>
<P>
0.8
0.6
= 0.25
0.4
s=1
0.75
0.2 0.5
0.25
0
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
r; rc
1.0
<I>
<P>
0.8
s=1
< (r)> <P(rc)>
0.75
0.6
0.5 = 0.5
0.25
0
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
r; rc
1.0
s=1
= 0.75
0.8
<I>
0.75
< (r)> <P(rc)>
0.6
<P>
0.5
0.4
0.25
0
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
r; rc
Figure 5-5. Time-averaged irradiance and encircled power distributions for annular
pupils for several typical values of transverse Gaussian random image motion s.
5.2 Random Image Motion 459
flows out of the central bright disc of the irradiance distribution. However, for large
circles, the power difference becomes alternately positive and negative as more power
flows inward or outward, respectively, from a bright ring of the irradiance distribution.
The range of s values for which this effect occurs increases as increases. Moreover,
whereas the central irradiance drops rapidly as s increases, it is found that the encircled
power for large circles (rc ≥ 3) does not change significantly for the S values considered
here. The time-averaged PSFs may again be approximated by a Gaussian, as in the case
of circular pupils.
When the Fresnel number is small, the depth of focus is large. Hence, the effect of a
small amount of defocus on the image is negligible. However, when the Fresnel number
is large, the depth of focus is small and the image is degraded, when z differs even
slightly from R. In this case z is approximately equal to R, and the defocus phase
aberration may be written
pD
Bd = - , (5-28)
4 lF 2
1 2
2
Ê 2 ˆ Û
I (r; D; ) = Á ˜
Ë 1 - 2 ¯ ı
2
( )
Ù exp - 2 pi Dr J 0 ( p r r) r dr , (5-29)
where we have let D be in units of 8lF 2 . Thus, a value of D = 1 in these units represents
a defocus phase aberration of 2p or a wave aberration of one wave. The corresponding
Strehl ratio is given by
2
S( ) = Ì
[ (
Ï sin p D 1 - 2
Ô )] ¸Ô
˝ . (5-30)
(
ÔÓ p D 1 -
2
) Ô˛
1
p (D) =
2p s
(
exp - D2 2 s 2 ) , (5-31)
where s is its standard deviation (in the same units as D ). The time-averaged PSF may
be written
•
1 Û
I (r; s; ) = Ù I (r; D; ) exp - D 2 s dD .
2p s ı
2 2
( ) (5-32)
•
S(s; ) =
1 ÙÔ
Ì
[ (
Û Ï sin p D 1 - 2
)] ¸Ô 2 2
(
˝ exp - D 2 s dD .) (5-33)
2p s Ù
Ù Ô pD 1 -
ıÓ
2
( ) Ô˛
•
The time-averaged encircled power, i.e., the image power in a circle of radius rc in units
of l F , is given by
rc
p2 Û
P( rc ; s; ) =
2
( )
1 - 2 Ù I ( r; s; ) r dr ,
ı
(5-34)
0
where on the right-hand side we have multiplied by the aberration-free central irradiance
( )
Pp 1 - 2 4 l2 F 2 , divided by the total power P, and taken into account the fact that r
and rc are in units of lF .
We note from Eq. (5-30) that the static Strehl ratio is zero for integral values of
( )
D 1 - 2 . Thus, for a circular pupil, for example, it is zero when the defocus wave
aberration is one wave or the longitudinal defocus D (in units of 8lF 2 ) is unity.
However, the time-averaged value of the dynamic Strehl ratio for s = 1 is 0.3483.
Similarly, for an annular pupil with = 0.5, it is zero when D = 4 3, but the time-
averaged dynamic Strehl ratio for s = 4 3 is approximately 0.35.
5.2 Random Image Motion 461
s
s
Figure 5-6. Strehl ratio S(s; ) as a function of the standard deviation s of the
longitudinal Gaussian random image motion. is the obscuration ratio of the
annular pupil. s is in units of 8l F 2 and its numerical value represents the peak
defocus wave aberration in units of wavelength.
Figure 5-7 shows the time-averaged PSF I (r; s; ) and encircled power
P(rc ; s; ) for various values of s and . The PSFs have been normalized to unity at
the center to illustrate that for small values of , they are simply scaled by the Strehl ratio
in the region r <~ 0.5. Hence, the encircled power for rc <~ 0.5 is also scaled by the Strehl
ratio. The larger the value of s , the higher the relative irradiance in the region of the
diffraction rings. It is evident that the normalized PSFs overlap for larger and larger
values of r as increases and, consequently, the encircled power for larger and larger
values of rc can be obtained from the aberration-free value by multiplying it with the
corresponding Strehl ratio. For small values of , the diffraction rings tend to disappear
as s increases. However, for ≥ 0.5, the diffraction rings remain intact even for
relatively large values of s . This is seen clearly by comparing Figure 5-8, showing the
unnormalized PSFs for = 0.9 and values of s as large as 5, with Figure 5-7e. Even the
zero-value secondary minima, which can only gain in value because of the image motion,
remain practically zero in this case.
462 RANDOM ABERRATIONS
Table 5-3. Time-averaged Strehl ratio S(s; ) for annular pupils with an
obscuration ratio for various values of the standard deviation s of the
longitudinal random image motion in units of 8l F 2 . (Cont.)
0 1 1 1 1 1 1 1 1 1 1
0.1 0.9683 0.9684 0.9689 0.9697 0.9707 0.972 0.9736 0.9754 0.9774 0.9796
0.2 0.8865 0.887 0.8884 0.8909 0.8942 0.8985 0.9037 0.9097 0.9165 0.9239
0.3 0.783 0.7838 0.7862 0.7901 0.7957 0.8028 0.8115 0.8218 0.8336 0.8469
0.4 0.6822 0.6831 0.686 0.6907 0.6975 0.7063 0.7172 0.7303 0.7456 0.7632
0.5 0.5953 0.5963 0.5992 0.6043 0.6114 0.6207 0.6324 0.6466 0.6635 0.6834
0.6 0.5241 0.5251 0.528 0.5329 0.5399 0.5491 0.5607 0.575 0.5922 0.6126
0.7 0.4665 0.4674 0.4701 0.4748 0.4814 0.4902 0.5014 0.5152 0.5319 0.552
0.8 0.4195 0.4203 0.4229 0.4273 0.4335 0.4418 0.4524 0.4654 0.4814 0.5008
0.9 0.3807 0.3815 0.3839 0.388 0.3938 0.4016 0.4115 0.4239 0.439 0.4574
1 0.3483 0.349 0.3513 0.3551 0.3606 0.3679 0.3772 0.3888 0.4031 0.4206
1.2 0.2972 0.2979 0.2999 0.3033 0.3081 0.3146 0.3228 0.3331 0.3459 0.3615
1.4 0.2591 0.2597 0.2614 0.2645 0.2688 0.2745 0.2819 0.2912 0.3026 0.3167
1.6 0.2295 0.2301 0.2317 0.2344 0.2382 0.2434 0.2501 0.2584 0.2688 0.2815
1.8 0.206 0.2065 0.2079 0.2104 0.2139 0.2186 0.2247 0.2323 0.2417 0.2533
2 0.1868 0.1872 0.1886 0.1908 0.194 0.1983 0.2039 0.2109 0.2195 0.2302
2.2 0.1709 0.1713 0.1725 0.1745 0.1775 0.1815 0.1866 0.1931 0.201 0.2109
2.4 0.1574 0.1578 0.1589 0.1608 0.1636 0.1673 0.172 0.178 0.1854 0.1946
2.6 0.1459 0.1463 0.1473 0.1491 0.1517 0.1551 0.1596 0.1651 0.172 0.1806
2.8 0.136 0.1363 0.1373 0.139 0.1414 0.1446 0.1488 0.154 0.1605 0.1685
3 0.1273 0.1277 0.1286 0.1301 0.1324 0.1354 0.1393 0.1442 0.1503 0.1579
4 0.0966 0.0968 0.0975 0.0987 0.1005 0.1028 0.1058 0.1095 0.1142 0.1201
5 0.0778 0.078 0.0785 0.0795 0.0809 0.0828 0.0852 0.0883 0.0921 0.0969
5.2 Random Image Motion 463
Table 5-3. (Cont.) Time-averaged Strehl ratio S(s; ) for annular pupils with an
obscuration ratio for various values of the standard deviation s of the
longitudinal random image motion in units of 8l F 2 .
s\ 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95
0 1 1 1 1 1 1 1 1 1 1
0.1 0.9819 0.9843 0.9867 0.9891 0.9915 0.9937 0.9957 0.9975 0.9988 0.9997
0.2 0.932 0.9405 0.9493 0.9583 0.9671 0.9755 0.9833 0.99 0.9953 0.9987
0.3 0.8615 0.8774 0.8943 0.9118 0.9295 0.9469 0.9633 0.9778 0.9894 0.9972
0.4 0.7831 0.8052 0.8294 0.8554 0.8826 0.9102 0.937 0.9613 0.9814 0.995
0.5 0.7063 0.7326 0.7621 0.7949 0.8305 0.8679 0.9056 0.9411 0.9713 0.9923
0.6 0.6367 0.6649 0.6975 0.7349 0.7768 0.8226 0.8707 0.9179 0.9593 0.9889
0.7 0.5761 0.6047 0.6385 0.6782 0.7242 0.7765 0.8336 0.8921 0.9456 0.9849
0.8 0.5241 0.5522 0.586 0.6264 0.6747 0.7313 0.7957 0.8646 0.9304 0.9805
0.9 0.4798 0.5069 0.5399 0.5801 0.629 0.6882 0.758 0.836 0.9138 0.9754
1 0.4418 0.4678 0.4996 0.5389 0.5875 0.6478 0.7212 0.8068 0.8962 0.9699
1.2 0.3807 0.4043 0.4335 0.4702 0.5166 0.5761 0.6527 0.7487 0.8585 0.9573
1.4 0.334 0.3554 0.3821 0.4159 0.4593 0.5163 0.5922 0.6932 0.819 0.9429
1.6 0.2973 0.3168 0.3413 0.3724 0.4128 0.4665 0.5399 0.642 0.7789 0.927
1.8 0.2677 0.2856 0.3081 0.3369 0.3744 0.4249 0.495 0.5957 0.7395 0.9098
2 0.2434 0.2599 0.2807 0.3074 0.3424 0.3897 0.4563 0.5544 0.7015 0.8914
2.2 0.2232 0.2385 0.2578 0.2826 0.3153 0.3598 0.4229 0.5175 0.6654 0.8722
2.4 0.206 0.2202 0.2382 0.2615 0.2921 0.334 0.3939 0.4848 0.6316 0.8524
2.6 0.1913 0.2046 0.2214 0.2432 0.272 0.3116 0.3684 0.4556 0.6001 0.8321
2.8 0.1785 0.191 0.2068 0.2273 0.2545 0.2919 0.3459 0.4295 0.5709 0.8116
2.9 0.1727 0.1848 0.2002 0.2201 0.2466 0.283 0.3356 0.4175 0.5572 0.8013
3 0.1673 0.1791 0.194 0.2134 0.2391 0.2745 0.3259 0.4061 0.544 0.791
3.2 0.1574 0.1686 0.1827 0.201 0.2254 0.2591 0.3081 0.385 0.5191 0.7705
3.4 0.1487 0.1592 0.1726 0.19 0.2132 0.2453 0.2921 0.3659 0.4962 0.7503
3.6 0.1408 0.1508 0.1636 0.1802 0.2023 0.2329 0.2777 0.3486 0.475 0.7303
3.8 0.1337 0.1433 0.1555 0.1713 0.1924 0.2216 0.2645 0.3328 0.4554 0.7108
4 0.1273 0.1365 0.1481 0.1632 0.1834 0.2114 0.2526 0.3183 0.4372 0.6918
4.2 0.1215 0.1303 0.1414 0.1559 0.1752 0.2021 0.2417 0.305 0.4204 0.6733
4.4 0.1162 0.1246 0.1353 0.1492 0.1677 0.1936 0.2317 0.2927 0.4047 0.6554
4.6 0.1114 0.1194 0.1297 0.143 0.1608 0.1857 0.2224 0.2814 0.3901 0.6381
4.8 0.1069 0.1146 0.1245 0.1373 0.1545 0.1785 0.2139 0.2709 0.3765 0.6214
5 0.1028 0.1102 0.1197 0.1321 0.1487 0.1718 0.206 0.2612 0.3638 0.6053
464 RANDOM ABERRATIONS
s
< s
> < ; s)>
(a)
< s;
> < s;
>
(b)
< s;
> < ; s; )>
(c)
< s;
> < ; s; )>
(d)
< s;
> < ; s; )>
(e)
< s;
> < ; s; )>
Figure 5-8. Same as Figure 5-7e for = 0.9, except that the PSFs are not normalized
to unity at the center.
5.3 Imaging through Atmospheric Turbulence 467
The fluctuating image can be stabilized if the wavefront tilt is corrected in real time
with a steering mirror. A tilt-corrected image is equivalent to a time-averaged short-
exposure image. The characteristics of such an image are considered and compared with
those of a long-exposure image. The angular resolution of the system for a long-exposure
image is about l r0 compared to the diffraction-limited resolution l D . For a short-
exposure image the resolution is significantly better than that for a long-exposure image.
The exposure time may be 0.1 sec or less. The location of such an image is random, but it
is not relevant in astronomy. When r0 << D , the image breaks up into small spots, called
speckles, of roughly the same size, which depends on D, while the overall size of the
image is determined by r0 . An example of a speckled short-exposure image of a point
object is illustrated and compared with the corresponding aberration-free image. The
image wander, sometimes referred to as the (time-varying) angle of arrival, is considered
in terms of the wavefront tilt (called the Z-tilt) or the centroid of the image (called the G-
tilt), and expressions for their standard deviations are given.
The purpose of a large ground-based telescope has generally not been better
resolution but to collect more light so that dim objects may be observed. Of course, with
the advent of adaptive optics,7-9 the resolution can be improved by correcting the phase
aberrations with a deformable mirror. The amplitude variations are negligible in near-
field imaging, i.e., when the far-field distance D 2 l >> L or D >> l L , where l is the
wavelength of the starlight and L is the propagation path length through the turbulence. In
principle, a diffraction-limited image can be obtained if the aberrations are corrected
completely in real time by the deformable mirror. However, in far-field imaging, i.e.,
when D 2 l << L or D << l L , significant amplitude variations are introduced in
addition to the phase variations. Since the amplitude variations cannot be corrected by
adaptive optics, even a complete correction of the phase errors does not yield a
diffraction-limited image. The characteristics of a short-exposure far-field image are also
discussed.
r r r
Whereas N ( r ) ~ 1, n( r ) is only on the order of 10 6 . It should be evident that n( r )
r rr
has a mean value n ( r ) = 0 . The structure function n ( r1 , r2 ) of the refractive index
fluctuations represents the mean square value of the difference of refractive index at two
r r
points r1 and r2 , i.e.,
rr r r
n (r1, r2 ) = [n (r1) - n (r2 )] 2
5.3 Imaging through Atmospheric Turbulence 469
r r r r
= n 2 ( r1 ) + n 2 ( r2 ) - 2 n ( r1 ) n( r2 ) . (5-36)
The turbulent atmosphere consists of packets of air called eddies, each with a
characteristic value of its refractive index. However, it is reasonable to assume that the
turbulence is locally statistically homogeneous so that the mean square value of the
refractive index is the same at every point, i.e.,
r r
n 2 ( r1 ) = n 2 ( r2 ) . (5-37)
The homogeneity of turbulence also implies that the correlation function of refractive
index fluctuations,
r r r r
Rn ( r1 , r2 ) = n ( r1 ) n ( r2 ) , (5-38)
r r
depends on the difference r2 - r1 of the position vectors of the two points. Thus, letting
r r r
r = r2 - r1 , (5-39)
r
The power spectral density Qn (k ) of the refractive index fluctuations is given by the 3D
Fourier transform of their correlation function, i.e.,
r r r r
Qn (k ) = Ú Rn ( r ) exp (2 p i k r ) d 3r , ◊ (5-41)
r
where k is a 3D vector representing a spatial frequency in units of cycles/m. It gives a
r
measure of the relative abundance of eddies with dimensions k 1 , where k = k . For a
spatial period L of the refractive index fluctuations, k = 1 L . Similarly, we may write
r r r r 3
Rn ( r ) = Ú Qn (k ) exp (- 2 p i k ◊ r ) d k . (5-42)
Rn (r ) = n (0) n (r ) . (5-43)
For statistically homogeneous and isotropic turbulence, the structure function of the
refractive index fluctuations can be written in terms of their corresponding correlation
function in the form
n (r ) = 2 [ Rn (0) - Rn (r )] . (5-44)
n (r ) = Cn2 r 2 3 , l 0 r L0 , (5-45)
where Cn2 (in units of m 2 3 ) is called the refractive index structure parameter. The
quantities l0 and L0 are called the inner and outer scales of turbulence, representing the
smallest and the largest eddies, respectively. Typical values of Cn2 vary from 10 13 m 2 3
for strong turbulence to 10 17 m 2 3 for weak turbulence. Values of l0 are on the order
of a few millimeters, and those of L0 vary from 1 m to 100 m. Substituting Eq. (5-42)
into Eq. (5-44), letting l0 Æ 0 and L0 Æ • (and noting that d 3k = k 2 dk sin q dq df,
0 £ q £ p , 0 £ f < 2 p , and carrying out the angular integrations), we obtain
•
Û Ê sin 2 p k r ˆ 2
n (r ) = 8p Ù Qn (k ) Á1 - ˜ k dk . (5-46)
ı Ë 2p k r ¯
0
Noting that
•
Û aÊ sin bx ˆ G ( a) sin ( p a 2)
Ù x Ë1 - bx ¯ dx = - , - 3 < a < -1 , (5-47)
ı b a +1
0
it can be shown that the right-hand side of Eq. (5-46) reduces to the right-hand side of Eq.
(5-45) if
Qn (k ) = 9.69 ¥ 10 3 Cn2 k 11 3
. (5-48)
If we divide r by 2p in Eq. (5-41) [and, therefore, in Eqs. (5-42) and (5-46)], then the
numerical coefficient in Eq. (5-48) is replaced by 9.69 ¥ 10 3 (2 p)2 3 or 0.033. Now
k = 2 p L is a wavenumber, similar to the optical wavenumber k, which for an optical
wavelength l , is k = 2p l . Similarly, if we divide k by 2p , as was done by Fried,4 then
the coefficient becomes 9.69 ¥ 10 3 (2 p)11 3 or 8.18.
(r )
PR rp [ (r ) (r )]
= exp l rp + i F R rp . (5-49)
(r r ) (r ) (r )
M rp , rp¢ = PR rp PR* rp¢
5.3 Imaging through Atmospheric Turbulence 471
∫ P1P2*
[
= exp (l 1 + l 2 ) + i (F1 - F 2 ) ] , (5-50)
where the angular brackets indicate an ensemble or time average, and we have used the
r r
subscripts 1 and 2 to represent the point locations rp and rp¢ on the wave, respectively.
If the refractive index fluctuations are statistically stationary, i.e., they are statistically
homogeneous and isotropic, then so are the fluctuations in l and F which they generate.
Therefore,
= 0 , (5-51b)
where the two averaged quantities in the first term on the right-hand side of Eq. (5-51a)
cancel each other due to homogeneity of turbulence, and those in the second term cancel
due to its isotropy. Thus, l1 + l 2 and F1 - F 2 are uncorrelated random variables. Hence,
Eq. (5-50) may be written
(r r )
M rp , rp¢ = exp (l 1 + l 2 ) [
exp i (F1 - F 2 ) ] . (5-52)
For a Gaussian random variable x with a mean value of x and a standard deviation of
s , it is easy to show that
•
1
Ú exp (bx ) exp[- ( x - x ) ]
2
exp(bx ) = 2s 2 dx
2 ps
•
1
= exp ÈÍ b 2 ( x - x ) + b x ˘˙ ,
2
(5-53)
Î 2 ˚
where b is an arbitrary constant. As the wave propagates through the random atmosphere,
it accumulates randomly both phase and amplitude variations. Each of these two random
variables can be expressed as an integral along the propagation path. Since the
atmospheric path length is much longer than the correlation length of these variable, the
accumulation consists of a sum of a large number of terms that are statistically
independent. By the central limit theorem, we can infer that they obey Gaussian
statistics.10 Letting both l and FR be Gaussian random variables, where FR has a mean
value of zero, we obtain,
It is shown later from conservation of energy [see Eq. (5-90)] that the left-hand side of
Eq. (5-54) is unity. Hence, the argument of the exponent on its right-hand side must be
zero, or
472 RANDOM ABERRATIONS
2
l2 = l - l . (5-55)
l1 = l 2 (5-56)
and
l12 = l 22 . (5-57)
Hence,
[
= exp l 2 + l1l 2 - 2 l 2
+2 l ]
= exp ( l l 1 2 - l2 ) , (5-58)
where in the last step we have used Eq. (5-55). Now we define a log-amplitude structure
function l according to
= (l 1 - l 2 )2
= 2 l 2 - l 1l 2 ,
(5-59)
r r
and not on rp and rp¢ separately. In view of Eq. (5-59), Eq. (5-58) may be written
È 1 ˘
exp (l 1 + l 2 ) = exp Í - l ( r )˙ . (5-61)
Î 2 ˚
Since F1 and F 2 are Gaussian random variables with zero mean values, F1 - F 2 is
also a Gaussian random variable. Hence, following Eq. (5-53), we obtain
1
[
exp i (F1 - F 2 ) ] = exp ÈÍ- ( F1 - F 2 ) 2 ˘
˙˚
Î 2
È 1 ˘
= exp Í - F ( r )˙ , (5-62)
Î 2 ˚
5.3 Imaging through Atmospheric Turbulence 473
where
[F (rr ) - F (rr ¢ )]
2
F ( r ) = p p (5-63)
is the phase structure function of turbulence. Substituting Eqs. (5-61) and (5-62) into Eq.
(5-52), we may write the MCF as
È 1 ˘
M ( r ) = exp Í - w( r ) ˙ , (5-64)
Î 2 ˚
where
w (r ) = l ( r ) + F ( r ) (5-65)
The wave structure function for a spherical wave propagating through Kolmogorov
turbulence is given by 4
L
53
w ( r ) = 2.914 k 2 r 5 3 Ú C n2 ( z )( z L ) dz . (5-66)
0
where k = 2p l and z varies along the atmospheric path of total length L from a value of
zero at the source (or the object plane) to a value of L at the receiver (or the image plane).
where
L 35
ÔÏ 53 Ô¸
Ì Ú C n ( z ) (z L) dz ˝
65 2
r0 = 0.1847l (5-70)
ÔÓ 0 Ô˛
C n2 ( h) = 8.2 ¥ 10 26
W 2 h10e h
+ 2.7 ¥ 10 16
e h 1.5
+ Ae 10 h
, (5-71)
where h is the height from the ground in km, A = 1.7 ¥ 10 14 , and W = 21 m/s is the wind
speed. Its variation with h is shown in Figure 5-9. It decreases rapidly for the first few
kilometers, dips at about 5.9 km, rises slightly due to turbulence in the jet stream peaking
at about 9.8 km, and decreases monotonically to negligible values beyond about 20 km.
35
ÈL 53 ˘
6 5 ÍÛ 2 Ê zˆ ˙
= 0.185l ÍÙ C n ( z ) Á 1 - ˜ dz ˙ . (5-72)
ı Ë L ¯
ÍÎ 0 ˙˚
As shown in Figure 5-10 by the dashed curve, the value of r0 in the plane of the observer
on ground does not change very much with the height L of the point source.
For a plane wave propagating up or down through the atmosphere, the factor z/L in
Eq. (5-66) and in turn, in Eq. (5-70), is replaced by unity. It is easy to see this for starlight
observed from the ground. A plane wave can be thought of as a spherical wave
originating at an infinite distance and traveling through a uniform medium for which
Cn2 = 0 , except for the propagation path through the atmosphere. The value of z L in the
region for which Cn2 π 0 is infinitesimally different from unity. Hence, the starlight
5.3 Imaging through Atmospheric Turbulence 475
Figure 5-9. Variation of the refractive index structure parameter C n h with the
2
Figure 5-10. Variation of r0 in the plane of the observer with the path length L from
the source to the observer. The propagation of light is along the zenith at a
wavelength λ = 0.5 μm. The solid curve is for a point source on ground such that the
spherical wave propagates upwards to a space observer looking down directly at the
source. The dashed curve is for a point source in space so that a spherical wave
propagates down and an observer on the ground looks at it directly above. The
dotted curve is for plane wave propagation with an observer on the ground (or in
space) looking up (or down) at a space (or ground) object.
476 RANDOM ABERRATIONS
L 35
6 5Ê ˆ
r0 = 0.185l ÁÚ C n2 ( z ) dz ˜ . (5-73)
Ë0 ¯
The variation of r0 with the separation L of the plane-wave source and the observer
is shown in Figure 5-10 by the dotted curve. We note that the variation is small and
occurs for small values of L. For large values of L, approaching infinity as in ground-
based astronomy, the value of r0 is close to 5 cm. Moreover, the value of r0 for a ground
observer looking up through the atmosphere at an incoming plane wave is smaller than its
value for a corresponding spherical wave, as expected from the smaller factor of 1 - z L
compared to z L .
We also note that r0 is smaller when a satellite is observed from the ground
compared to when an object on the ground is observed from a satellite. As an observer
moves to a higher altitude above the atmosphere, the value of r0 increases linearly with
the altitude L. Consequently, the image degradation is much smaller when an object on
the ground is observed from space than when a space object is observed from the ground.
In the first case, the object is near the region of turbulence, and it is observed from far
away. In the second case, the object is away from the region of turbulence, but it is
observed from nearby. The reason for the superior image quality in the first case is
similar to the observability of an object behind a diffuse shower glass. One can see some
detail in the object when it is in contact with the shower glass. However, as soon as the
object is moved slightly away from the shower glass, it appears only as a halo, illustrating
complete loss of image resolution. This does not, however, mean that reciprocity of wave
propagation does not hold.13,14 For example, if the wavefront errors of a wave from a
point source in space propagating downwards are measured on the ground, and a
conjugate correction is introduced in a beam transmitted upwards with a deformable
mirror, the beam focus in space will be diffraction limited (neglecting any measurement
or correction error), illustrating that the atmosphere introduces the same wavefront errors
whether a beam is propagating up or down through it.
The coherence length for horizontal propagation and, therefore, nearly constant
C n2 ( h) can be obtained from Eqs. (5-72) and (5-73). Thus,
Ô
r0 = Ì
(
Ï1.68 C 2 L k 2
n ) 35 Plane Wave (5 – 74)
Ó (
Ô 3.02 C n2 L k 2 ) 35 Spherical Wave . (5 – 75)
5.3 Imaging through Atmospheric Turbulence 477
r r r r 2
F (r ) = [F (r1 ) - F (r1 + r )]
= 2 [ RF (0) - RF (r )] , (5-76)
In the near field , the amplitude variations are negligible and, therefore,
F ( r ) = w ( r ) , (L << D 2 l) .
(5-78)
F ( r ) = (1 2) w ( r ) , (L >> D 2 l) . (5-79)
For in-between ranges, the multiplying factor varies smoothly between 1 and 1/2.5 The
correlation function RF (r ) of phase fluctuations and the corresponding power spectral
density QF (k ) are related to each other by a 2D Fourier transform. Thus,
r r
RF (r ) = Ú ◊
QF (k ) exp ( - 2 p i k r ) d 2 k (5-80a)
= 2 p Ú QF (k ) J 0 (2 p k r ) k d k , (5-80b)
where we have used Eq. (1-81) to carry out the angular integration. Similarly, we may
write
r r
QF ( k) = Ú ◊
RF ( r ) exp (2p i k r ) d 2 r (5-81a)
= 2p Ú RF r J 0 (2p k r ) r d r . (5-81b)
•
[
F (r ) = 4 p Ú QF (k ) 1 - J 0 (2 p kr ) k d k . ] (5-82)
0
Noting that
•
Û p ba 1
Ùx
a
[1 - J 0 (bx)] dx = , 1<a <3 , (5-83)
ı [
2 G (a + 1) 2 sin p (a - 1) 2
a 2
] [ ]
0
478 RANDOM ABERRATIONS
it can be shown that the right-hand side of Eq. (5-82) reduces to the right-hand side of Eq.
(5-69) if
If the system is aberration free and has uniform transmittance, a spherical wavefront of
uniform amplitude centered at the Gaussian image point emerges from its exit pupil.
However, if the system is aberrated and has nonuniform transmittance, then a distorted
r r
wavefront of nonuniform amplitude emerges from its exit pupil. Let FL rp and AL rp
r ( ) ( )
be the phase and amplitude transmittance corresponding to a point r p on the exit pupil. A
dimensionless pupil function of the system may be written
(r ) (r )
PL rp = AL rp exp i F L rp [ (r )] . (5-86)
The total power in the exit pupil and, therefore, in the image is given by
r 2 r
Pex = Ien Ú P (r ) L p d rp
(r )
= Ien Ú AL2 rp d rp
r
. (5-87)
When the medium between the object and the imaging system is inhomogeneous as
in ground-based astronomy, it introduces random phase and amplitude variations across
r r
the wavefront propagating through it. Let FR ( ren ) and l ( ren ) be the random phase and
r
(dimensionless) log amplitude introduced by atmospheric turbulence at a point ren on the
entrance pupil. The instantaneous object power at the entrance pupil is given by
r r
Pen = Ien Ú exp [2l ( ren )] d ren . (5-88)
Hence, the time-averaged total power at the entrance pupil may be written
r r
Pen = Ien Ú [
exp 2l ( ren ) ] d ren . (5-89)
r
[
exp 2 l ( ren ) ] = 1 . (5-90)
As the wave propagates through the imaging system, it undergoes additional phase and
r
amplitude variations. The total phase aberration at a point r p on the exit pupil may be
written
(r )
F rp = F L rp + F R rp (r ) (r ) , (5-91)
r r
where the position vector r p is related to the position vector ren by the pupil
r r
( )
magnification m p i.e., rp = m p ren . The pupil function of the overall system (i.e.,
including the effects of atmospheric turbulence) representing the wavefront at the exit
pupil may be written
(r )
P rp = PL rp PR rp (r ) (r ) , (5-92)
where
(r ) [ (r )
PR rp = exp l rp + i F R rp (r )] (5-93)
(r )
is truncated by the pupil function PL rp and represents the complex amplitude variation
introduced by turbulence. The instantaneous total power at the exit pupil is given by
r 2 r
Pex = Ien Ú P(r ) p d rp
r r r
Ú AL (rp ) exp [2l (rp )] d rp
2
= Ien . (5-94)
r
Pex = Ien Ú A (r )
2
L p exp 2 l rp[ (r )] r
d rp
= Ien Ú (r )
AL2 rp d rp
r
, (5-95)
where we have made use of Eq. (5-90). Comparing Eqs. (5-87) and (5-95), we note that,
as expected, the average total power at the exit pupil is the same as the corresponding
total power in the absence of atmospheric turbulence. If the transmittance of the optical
r
system is uniform; e.g., if the transmittance is 100% so that AL rp = 1 m 2p , then ( )
Pex (
= Ien m 2p Sex ) (5-96a)
where Sex is the area of the exit pupil. It should be evident that in this case, the irradiance
at the exit pupil is uniform, with a value of Iex = Ien m 2p .
480 RANDOM ABERRATIONS
From Eq. (1-62), where the pupil function is not dimensionless, the instantaneous
irradiance distribution of the star image formed by the overall system is given by
r Û r Ê 2 pi r rˆ r
2
Pex
Ii ( ri ) =
Sex l2 R 2 ı
( )
Ù P rp exp ÁË - l R rp ◊ ri ˜ d rp
¯
. (5-97)
Substituting Eq. (5-92) into Eq. (5-97), the time-averaged distribution may be written
r r r
(r ) (r )
Pex
Ii ( ri ) = ÚÚ PL (rp ) PL (rp¢)
*
2 2 PR rp PR* rp¢
Sex l R
È 2p i r r
) ◊ rr ˙˚ d rr d rr ¢
˘
¥ exp Í -
Î lR
rp - rp¢( i p p . (5-98)
Dividing both sides of Eq. (5-98) by Pex yields the PSF. The quantity in the angular
brackets is the MCF. Substituting for it from Eq. (5-64), we may write
r ÛÛ P r * r
Ii ( ri ) =
Pex
2
È 1
( ) ( )
ÙÙ L rp PL rp¢ exp Í- w
Sex l R ıı 2
Î 2
( rr p
r
- rp¢ ˘˙
˚
)
È 2p i r r
) ◊ rr ˙˚ d rr d rr ¢
˘
¥ exp Í -
Î lR
rp - rp¢( i p p . (5-99)
Because of a Fourier transform relationship between the PSF and the OTF, we identify
r r r
( )
rp - rp¢ l R with a spatial frequency vi . Thus, we let
r r r
rp - rp¢ = l R v i . (5-100)
r
Substituting Eq. (5-100) into Eq. (5-99) and carrying out the integration over r p , we may
write
r Û r È 1 ˘ r r r
I i (ri ) = Pex Ù t L (v i ) exp Í - w (l R v i )˙ exp (- 2pi v i
ı Î 2 ˚
◊ ri ) d v i , (5-101)
where
r
(r ) (r
t L ( vi ) = Sex1 Ú PL rp PL* rp - l R vi d rp
r
) r
(5-102)
r
is the OTF of the (turbulence-free) optical system and vi = vi .
and
r r
vi = v l F , (5-103c)
where
Pex Sex
I (0) = (5-104)
l2 R 2
where
r r
[
t ( v ) = t L ( v ) exp - (1 2) w ( vD) ] (5-106a)
r
= t L ( v ) M ( vD) (5-106b)
is the time-averaged OTF of the overall system. This OTF being equal to the product of
the OTF of the optical system and the MCF M ( vD) associated with atmospheric
turbulence is a consequence of the fact that the pupil function of the overall system is
equal to the product of the pupil function of the optical system and the complex
amplitude variation introduced by turbulence. It should be evident that, since vD = v il R ,
the modification of the system OTF due to turbulence is independent of the diameter D of
the exit pupil of the system. The MCF acts as a reduction factor for the OTF of the
overall system.
Substituting Eq. (5-69) for the wave structure function in terms of the coherence
length r0 into Eq. (5-106a), we may write the time-averaged OTF of the overall system
r r
t (v ; D r0 ) = t L ( v ) M (v; D r0 ) , (5-107)
where
[
M (v; D r0 ) = exp - 3.44 (vD r0 )
53
] (5-108)
is the long-exposure atmospheric MCF, representing the MTF reduction factor or the
Hopkins ratio (defined in Section 1.10 and discussed further in Section 2.11). As stated
earlier, M (v; D r0 ) is independent of the pupil diameter D, as may be seen by replacing
the normalized spatial frequency v by vi l F and noting that F = R D . The exponent in
482 RANDOM ABERRATIONS
Eq. (5-108) varies with wavelength as l 1 3 . The vectorial nature or the orientation of the
r
spatial frequency v enters into Eq. (5-107) through the aberrations of the imaging
system. For numerical analysis, we will consider only aberration-free systems so as to
focus on the effects of aberrations introduced by turbulence. Accordingly, the spatial
frequency will enter our analysis only as a scalar quantity.
Since exp ( - 3.44) ~ 0.03 , atmospheric turbulence reduces the overall system MTF
corresponding to a spatial frequency v = r0 D by a factor of 0.03. From Eqs. (5-64) and
(5-69), we find that r0 represents a correlation length such that the correlation of complex
amplitudes at two points on a wave separated by a distance r0 is 0.03. Moreover, by
definition, the MCF represents the degree of spatial coherence of the wave at the
receiver, and thus the visibility of the fringes formed in a two-pinhole experiment and
observed in the vicinity of a point that is equidistant from the two pinholes. Note that
because of the random nature of atmospheric turbulence, the time-averaged irradiances at
the two pinholes are equal to each other. Hence, r0 represents a partial coherence length
of the wave so that its degree of coherence corresponding to two points on it separated by
r0 is only 0.03, or that the visibility of the fringes formed by the secondary waves from
these points is 0.03. The value of r0 on a mountain site may vary from 5 to 10 cm in the
visible region of the spectrum and increases with wavelength as l6 5 . 15
S ( D r0 ) = I (0; D r0 )
1
= 8 Ú t (v; D r0 ) v d v . (5-110)
0
It gives the average central irradiance relative to its aberration-free value. Similarly, from
Eqs. (1-89) and (2-21b), it follows that the average fractional encircled power is given by
1
P (rc ; D r0 ) = 2 p rc Ú t (v; D ro ) J1 (2 p rc v) d v , (5-111)
0
where rc is in units of l F . This result may also be obtained by substituting Eq. (5-109)
into Eq. (2-22b) and using Eq. (1-88). Equations (5-109) through (5-111) are similar to
Eqs. (5-9) through (5-12), respectively. Their differences lie in the difference of the
average OTFs given by Eqs. (5-10) and (5-107).
5.3 Imaging through Atmospheric Turbulence 483
Pex Sex
Ii (0; D r0 ) = S ( D r0 ) (5-112a)
l2 R 2
Pex Sa
= h ( D r0 ) , (5-112b)
l2 R 2
Since Sex = p D2 4, the quantity h is proportional to the central irradiance for a fixed
total power Pex . For example, we may have a ground-based laser transmitter whose
total power is fixed, but whose beam diameter can be selected to change the central
irradiance on a target. From Eqs. (5-107) and (5-110), we may write
21
Ê Dˆ
h ( D r0 ) = 8 Á ˜
Ë r0 ¯ Út
0
L [ 53
(v) exp - 3.44(vD r0 ) v d v , ] (5-114)
where t L (v) is the aberration-free OTF of the imaging system given by Eq. (2-44).
= 1 , (5-115)
53
where x = 3.44 (vD r0 ) and the integral over x is the gamma function G (6 5) . Hence,
the central irradiance given by Eq. (5-112b) approaches a limiting value of
Pex Sa l2 R 2 , which is equal to the aberration-free value for a system with an exit pupil
of diameter r0 . Since Sa ~ ro2 ~ l12 5 , the limiting value varies with wavelength as l2 5 .
Figure 5-11a shows how turbulence degrades the system MTF for several values of
D r0 . The aberration (or turbulence)-free MTF (which corresponds to D r0 = 0 ) is also
484 RANDOM ABERRATIONS
shown for comparison. For example, when D r0 = 2 , not only is the MTF at any
frequency reduced, but the cutoff frequency is practically reduced from a value of 1 to
0.5. Similarly, when D r0 = 5 , the cutoff frequency is practically reduced to 0.2. The
OTF at any frequency reduces as D r0 increases.
Figure 5-12 shows the average irradiance distribution normalized to unity at the
center for several values of D r0 . The diffraction rings disappear even for small values of
D r0 such as 1, the PSFs become smooth and may be approximated by Gaussian
functions. The corresponding encircled power is also shown in this figure. As D r0
increases, a given fraction of the total power is contained in an increasingly larger circle.
As an example, whereas 84% of the total power is contained in a circle of radius
rc = 1.22 when there is no turbulence, it is contained in a circle of radius 1.9 when
D r0 = 1 .
5.3 Imaging through Atmospheric Turbulence 485
0.8
D/r0 = 0
0.6
<t>
1
2
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
v
(a)
1.0
<S>
0.1
0.01
0 2 4 6 8 10
D/r0
(b)
10
(D/r0)2
1
h
(D/r0)2 <S>
0.1
0.01
0.1 1 10
D/r0
(c)
Figure 5-11. Effect of atmospheric turbulence on time-averaged system
performance. (a) MTF. The diffraction-limited case is represented by D r0 = 0 . (b)
Strehl ratio. (c) Central irradiance for a fixed total power represented by h . Its
2
aberration-free value increases as ( D r0 ) , but its aberrated value approaches unity
as D r0 Æ • .
486 RANDOM ABERRATIONS
1.0
D/r0 = 0
1
0.8 2
3
< (r)> <P(rc)>
4
0.6
5
0.4
0.2
0.0
0 1 2 3 4 5 6
r; rc
1
8
S (; D r0 ) = Ú t (v; ; D r0 ) v dv , (5-117)
1 - 2 0
and
1
P (rc ; ; D r0 ) = 2p rc Ú t (v; ; D r0 ) J1(2p rc v ) d v , (5-118)
0
where
[
t (v; ; D r0 ) = t L ( v; ) exp - 3.44 (vD r0 )
53
] (5-119)
Figure 5-13 shows the time-averaged OTF for several values of D1ro. The OTF gain at
high frequencies disappears even for weak turbulence, as is evident from Figure 5-13b for
D1ro = 1. The turbulence dominates the OTF for large values of D1ro, and the effect of
obscuration becomes small, as illustrated in Figure 5-13c. Not only is the MTF at any
frequency reduced, but the effective cutoff frequency is also reduced, for example, from a
value of 1 to about 0.1 when D1ro= 10.
0.8
/'..
,-....
....0
Q- 0.6
\Ii (a)
;; 0.4
'--'
...
...........
0.2
0.8
/'..
,-....
....0
Q-
.
0.6
\Ii
~
(b)
;> 0.4
'--'
...
...........
0.2
0.8
/'..
,-....
....0
Q-
. ~
0.6
\Ii (c)
;>
'--' 0.4
...
...........
0.2
Figure 5-13. Time-averaged OTF for various values of E and D1ro. The solid curves
represent the aberration-free OTF, and the dotted and dashed curves represent the
corresponding long- and short-exposure OTFs. (a) E = 0, (b) E = 0.25, and (c) E = 0.5.
488 RANDOM ABERRATIONS
(
h (; D r0 ) = 1 - 2 (D r0 ) ) 2
S (; D r0 ) (5-120)
( )
varies with D r0 . Since Sex () = p 1 - 2 D2 4 , this quantity is proportional to the
aberrated central irradiance and thus shows its variation with D for a given value of and
a fixed total power Pex . Its aberration-free or diffraction-limited value varying as
( ) 2
1 - 2 ( D r0 ) is illustrated by the straight lines for several values of . For small values
of Dr0 , the atmospheric MTF reduction factor is approximately equal to unity.
Accordingly, S ( ; D r0 ) is also approximately equal to unity, and the aberrated value of
h increases with D r0 as its aberration-free counterpart. However, for larger values of
D r0 , it increases slowly with a negligible increase beyond a certain value of D r0 ,
10.0
0.25
=0
0.50
0.75
=0
1.0
h(D/r0; )
0.25
0.50
0.75
0.1
0.01
0.1 1.0 10.0 100.0
D/r0
h ( ; D r0 ) Æ 1 as D r0 Æ • , (5-121)
as in the case of circular pupils. The two asymptotes of h (; D r0 ) for a given value of
12
(
intersect at the point D r0 = 1 - 2 ) . The aberrated central irradiance is given by the
product of its aberration-free value Pex Sex ( ) l2 R 2 and the Strehl ratio S ( ; D r0 ) or
( )
Pex Sa l2 R 2 h ( ; D r0 ) . Hence, regardless of how large D is, the central irradiance is
less than or equal to the aberration-free central irradiance for a system with an exit pupil
of diameter r0 , equality approaching as D r0 Æ •. The limiting value of the central
irradiance is independent of the value of .
( )
In astronomical observations, Pex = p 1 - 2 D2 I0 , where I0 is the time-
averaged irradiance across the exit pupil, and Pex increases as D increases. However, if
the observation is made against a uniform background, then the background irradiance in
the image also increases as D2 . Hence, as in the case of a circular pupil, the detectability
of a point object is limited by turbulence to a value corresponding to an exit pupil of
diameter r0 , no matter how large the diameter D of the exit pupil is. In the case of a laser
transmitter with a fixed value of laser power Pex , the central irradiance on a target will
again be limited by its aberration-free value for an exit pupil of diameter r0 . Similarly,
the ratio of the signal and noise powers in an optical heterodyne detection of a
turbulence-degraded signal is limited to the aberration-free value corresponding to an exit
pupil of diameter r0 .
Figure 5-15 shows how the difference in Strehl ratios for circular and annular pupils
[
D S (; D r0 ) = 10 S (0 ; D r0 ) - S (; D r0 ) ] (5-122)
varies with D r0 . It is evident that the Strehl ratio decreases to zero monotonically as
D r0 increases, irrespective of the value of . However, we note that DS > 0 for
D r0 <
~ 3 , i.e., S decreases faster for annular pupils than for circular pupils. The opposite
is true for D r0 >~ 3 . Some typical values of the Strehl ratio are given in Table 5-4 for
several values of D r0 and .
Figure 5-16 shows how the time-averaged irradiance distribution or the PSF and
encircled power change as D r0 increases for several values of . The PSFs are
normalized to unity at the center, the actual central value being the long-exposure (LE)
Strehl ratio given in Table 5-4. As increases, power flows from the central bright spot
into the diffraction rings. As D r0 increases, the diffraction rings disappear and the PSFs
become smooth, and a given fraction of total power is contained in a circle of larger and
larger radius.
490 RANDOM ABERRATIONS
1.2
1.0
DS (0.75)
0.8
<S (D/r0 ; 0)>
<S (D/r0 ; 0)> DS (e)
0.6
DS (0.50)
0.4
DS (0.25)
0.2
0.0
–0.2
0 2 4 6 8 10
D/r0
[
Figure 5-15. Variation of D S ( ; D r0 ) = 10 S (0; D r0 ) - S ( D r0 ) . The Strehl ]
ratio for = 0 shown in Figure 5-11b on a log scale is shown here on a linear scale.
Table 5-4. Time-averaged Strehl ratio for various values of and D r0 , where LE
and SE represent long- and short-exposure images, respectively.
\ D/ r0 1 2 5 10
LE SE LE SE LE SE LE SE
't\\\""'~~,
E= O
0.8
.........
,..-.,
,-0
0.6 \,\1 '~ \:? "'-.! 0
C3-- (a)
IjJ
'- 0.4
'-"
-......-
0.2
o 2 4 r 6 8 \0
(b)
2 4 r 6 10
~\~~""""
E= 0.5
0.8
0.6 \I \2 '<:~ ",10
(c)
0.4
0.2 -..........
o )~~
Df .' 1
L-~~_B~~=~~~~~
'. > __ ..-. ..., "'......... .
"""~"" ..........
______~·=~
· =
··
= ..
··= .. ·=..d··
o 2 4 r 6 \0
(a)
0.8
...........
~
.....
0
C:s- 0.6
(b)
w
.....u 0.4
'-"
0....
...........
0.2
0
0 2 4 6 8 10
0.8
...........
~
.....0
C:s- 0.6
w (c)
.....u
'-" 0.4
0....
...........
0.2
0
0 2 4 6 8 10
where r = r a is a normalized radial variable and a j are the expansion coefficients. The
polynomials may be written
where Rnm (r) are the radial polynomials given by Eq. (2-61c). As stated in Section 2.4, n
and m are positive integers (including zero) and n - m ≥ 0 and even. The index n
represents the radial degree or the order of the polynomial since it represents the highest
power of r in the polynomial, and m may be called the azimuthal frequency. The index
j is a polynomial-ordering number and is a function of n and m . The first eleven
orthonormal polynomials and the relationship among the indices j, n, and m are given in
Table 5-5. They are ordered such that an even j corresponds to a symmetric polynomial
varying as cosmq, while an odd j corresponds to an antisymmetric polynomial varying as
sinmq. For a given value of n, a polynomial with a lower value of m is ordered first. The
orthonormality of Zernike polynomials implies that
1 2p 1 2p
Ú Ú Z j (r, q) Z j ¢ (r, q) r dr dq Ú Ú r dr dq = d jj ¢ . (5-125)
0 0 0 0
494 RANDOM ABERRATIONS
Table 5-5. Orthonormal Zernike circle polynomials Z j ( r,, q) . The indices j, n, and m
are called the polynomial number, radial degree, and azimuthal frequency,
respectively. The polynomials Z j are ordered such that an even j corresponds to a
symmetric polynomial varying as cos mqq , while an odd j corresponds to an
antisymmetric polynomial varying as sin mqq. A polynomial with a lower value of n
is ordered first, and for a given value of n, a polynomial with a lower value of m is
ordered first.
j n m Z j ( r,, q) Aberration*
1 0 0 1 Piston
2 1 1 2 r cos q x tilt
3 1 1 2 r sin q y tilt
4 2 0 (
3 2r 2 - 1 ) Defocus
2
5 2 2 6 r sin 2q 45∞ Primary astigmatism
6 2 2 6 r2 cos 2 q 0∞ Primary astigmatism
7 3 1 (
8 3r3 - 2r sin q ) Primary y coma
8 3 1 8 (3r 3
- 2r) cos q Primary x coma
9 3 3 8 r3 sin 3 q
10 3 3 8 r3 cos 3 q
11 4 0 (
5 6r 4 - 6r2 + 1 ) Primary spherical
*The words “orthonormal Zernike circle” are to be associated with these names, e.g.,
orthonormal Zernike circle 0∞ primary astigmatism.
as may be seen by substituting Eq. (5-123) into Eq. (5-126) and using Eq. (5-125).
r r r r r r
( )
= 1 p 2 Ú d r Ú d r ¢ Z *j ( r ) Z j ¢ ( r ¢) RF ( r - r¢ ) , (5-127)
so that
r r r r r
Z j (r) = Ú S j (k ) exp ( - 2 p i k r ) d k .◊ (5-129)
J n +1 (2 pk )
Seven j (k, f) = 2(n + 1) i m ( -1)( n m) 2
cos mf , m π 0 , (5-130a)
k
J n +1 (2 pk )
Sodd j (k, f) = 2(n + 1) i m ( -1)( n m) 2
sin mf , m π 0 , (5-130b)
k
and
J n +1 (2 pk )
S j (k , f) = n + 1( -1)n 2 , m=0 , (5-130c)
k
r
where (k, f) are the polar coordinates of k . Substituting Eq. (5-129) into Eq. (5-127), we
obtain
r r r r r r r r
a*j a j ¢ = 1 p2 ( ) Ú d k S (k ) Ú d k ¢ S (k ¢) Ú d r ¢ exp [2p i (k - k ¢) ◊ r ¢]
*
j j¢
r r r r r r
¥ Ú d r RF ( r - r¢ ) exp [2p i k ◊ (r - r ¢) ] . (5-131)
r
The integration over r can be carried out using Eqs. (5-81a) and (5-84). Thus, since
r r
r = r a,
r r r r r r
Ú d r RF ( ) [
r - r ¢ exp 2 p i k ◊ (r - r ¢)]
) Ú d (rr - rr ¢) RF ( r r r r
r
= 1 a2 ( )
r - r ¢ exp 2p i (k a) [ ◊ (r - r ¢)]
ÏÔ 0.023 (a r )5 3 k 11 3 Near Field, L << D 2 l , (5 - 132a)
0
= Ì 53
ÔÓ(0.023 2)(a r0 ) k 11 3 Far Field, L >> D 2 l . (5 - 132b)
r r r
The integration over r ¢ yields d(k - k ¢) . Hence, Eq. (5-131) for L << D 2 l reduces to
r r r
a*j a j ¢ (
= 0.023 p 2 ) (a r )
0
53
Ú dkk 11 3 *
Sj (k ) S j ¢ (k ) . (5-133)
496 RANDOM ABERRATIONS
r r r
Substituting for S j (k ) and S j ¢ (k ) from Eqs. (5-130) into the integral over k , we obtain
r r r
Ú dkk 11 3 *
Sj (k ) S j ¢ (k )
2 12
= [
(n + 1) (n ¢ + 1) i ] m+ m¢
(-1)( n + n ¢ m m ¢) 2
Inn ¢ Imm ¢ , (5-134)
1 + d m0
where18
•
I nn ¢ ∫ Ú dk k 14 3
J n +1(2p k) J n ¢ +1(2 p k) (5-135)
0
=
[
p11 3 G (14 3) G (n + n ¢ - 5 3) 2 ] , n π 0 π n ¢ (5-136)
[ ] [
2 G (n - n ¢ + 17 3) 2 G (n ¢ - n + 17 3) 2 G (n + n ¢ + 23 3) 2 ] [ ]
and
◊
In Eq. (5-136), G ( ) is the gamma function. Thus, Eq. (5-133) may be written
( n + n ¢ 2m) 2 53
a *j a j ¢ = 0.1534 ( -1) [(n + 1)(n ¢ + 1) ]1 2 (Dr0 ) d mm ¢
G (14 3) G [(n + n ¢ - 5 3) 2] (5-138)
¥ .
G [(n - n ¢ + 17 3) 2] G [(n ¢ - n + 17 3) 2] G [(n + n ¢ + 23 3) 2]
The case when both n and n ¢ are equal to zero, i.e., when j and j ¢ are both equal to unity,
is discussed separately in the next section on aberration variance.
Numerical values of variance and covariance are given in Tables 5-6 and 5-7.19 We
note that the variance of an aberration coefficient depends on the value of n, but is
5.3 Imaging through Atmospheric Turbulence 497
F2 = RF (0)
•
53
= 0.046 p ( a r0 ) Úk
83
dk . (5-141)
0
D1 = F 2 - a12 = Â a 2j
j =2
Ú {1 - [2 J (2p k ) 2p k] }k
53 2
= 0.046 p ( a r0 ) 1
83
dk
0
= 1.0299 ( D r0 )
53
(rad ) 2
(5-142)
is finite. Of course, the piston aberration Z1 (r, q) , being a constant, has no effect on the
quality of an image. Since r0 μ l6 5 , the variance in terms of the optical path-length
errors is independent of l , as expected in the absence of atmospheric dispersion. In
obtaining Eq. (5-126), we have used the relation
•
2
Û ÔÏ È 2 J1 ( x ) ˘ Ô¸ p pG ( p + 2)
Ù Ì1 - Í ˙ ˝x dx = .
ı
0
ÔÓ Î x ˚ Ô˛
p 2
[ ] [ ]
2 G ( p + 3) 2 G ( p + 5) 2 G (1 + p 2) sin p( p - 1) 2 [ ]
(5-143)
498 RANDOM ABERRATIONS
ajaj¢ n n¢ ajaj¢
1
a 22 , a 32 1 1 4.49 ¥ 10
2
a 42 , a 52 , a 62 2 2 2.32 ¥ 10
2
a2 a8 , a3 a7 1 3 - 1.42 ¥ 10
3
a 72 , a 82 , a 92 , a 10
2
3 3 6.19 ¥ 10
3
a 4 a 11 , a 5 a 13 , a 6 a 12 2 4 - 3.88 ¥ 10
2 2 2 2 2 3
a 11 , a 12 , a 13 , a 14 , a 15 4 4 2.45 ¥ 10
3
a 7 a 17 , a 8 a 16 , a 9 a 19 , a 10 a 18 3 5 - 1.56 ¥ 10
2 2 2 2 2 2 3
a 16 , a 17 , a 18 , a 19 , a 20 , a 21 5 5 1.19 ¥ 10
4
a 11 a 22 , a 12 a 24 , a 13 a 23 , a 14 a 26 , a 15 a 25 4 6 - 7.60 ¥ 10
1
a 22 , a 32 1 1 0 4.49 ¥ 10
2
a2 a8 , a3 a7 1 3 2 - 1.42 ¥ 10
4
a 2 a 16 , a 3 a 17 1 5 4 7.54 ¥ 10
6
a 2 a 30 , a 3 a 29 1 7 6 - 9.52 ¥ 10
7
a 2 a 46 , a 3 a 47 1 9 8 8.61 ¥ 10
7
a 2 a 68 , a 3 a 67 1 11 10 - 1.41 ¥ 10
8
a 2 a 92 , a 3 a 93 1 13 12 3.24 ¥ 10
9
a 2 a 122 , a 3 a 121 1 15 14 - 9.29 ¥ 10
9
a 2 a 154 , a 3 a 155 1 17 16 3.14 ¥ 10
5.3 Imaging through Atmospheric Turbulence 499
The approximate expressions of Eq. (1-204) through (1-206) are not suitable for
calculating the time-averaged Strehl ratio for a random aberration, where D1 ∫ s 2F ,
especially when its value is small. For example, the approximate Strehl ratio given by
S1 ( D r0 ) [
~ exp -1.03( D r0 )5 3 ] (5-144)
is illustrated in Figure 5-17 by the dashed curve, which is quite steep owing to the
( D r0 )5 3 dependence of s 2F . Even for a small value of D r0 = 1 , it gives a Strehl ratio of
0.357, compared to a true value of 0.445. For larger values of D r0 , it underestimates the
Strehl ratio by larger factors. A much better approximation is given by
65
S2 (D r0 ) [
~ 1 + (D r0 )5 3 ] , (5-145)
as may be seen from Figure 5-17, where it is illustrated by the dotted curve. It slightly
overestimates the true value S .
100
<S>
10 1
<S>
10 2
2 4 6 8 10
D/r0
Figure 5-17. Variation of time-averaged Strehl ratio with D r0 . The solid curve
represents the exact value S given by Eq. (5-110), which is also shown in Figure 5-
6b. The dashed curve represents the approximate value S1 given by Eq. (5-144) in
terms of the aberration variance. Similarly, the dotted curve represents the
approximate value S2 given by Eq. (5-145).
500 RANDOM ABERRATIONS
•
= Â a j Z j (r, q) . (5-147)
j = J +1
DJ = [F J (r, q)]2
1 2p
2
Û Û È J ˘
ÍF(r, q) - Â a j Z j (r, q)˙
1
= p Ù Ù r dr dq
ı ı Î j =1 ˚
0
J
= D1 - Â a 2j , (5-148)
j =2
where D1 ∫ s F2 .
Numerical values of D J are given in Table 5-8. We note that when D = r0 , the phase
aberration with a standard deviation of 1 radian is obtained without any correction
(except for the piston mode). However, if the x and y tilts are corrected, we find from the
expression for D 3 , that D can be as large as 3.34 r0 for 1 radian of phase aberration. It is
evident that a large portion of the phase aberration introduced by turbulence is a random
wavefront tilt caused by large eddies and represented by the coefficients a 2 and a 3 .
The residual phase structure function after correction of the first J modes is given by
r r r r
J (r, r¢ ) = [F J (r) - F J (r¢)]2 (5-149)
where
5.3 Imaging through Atmospheric Turbulence 501
Table 5-8. Variance of residual phase errors for near-field propagation in units of
( D r0 )5 3 when J Zernike modes are corrected.17
D 9 = 0.0463 D 20 = 0.0220
r r r r
F (r, r¢ ) = [F(r) - F(r¢)]2 (5-151)
r r
is the long-exposure phase structure function. From Eq. (5-63), F (r, r¢ ) depends on
r r r r
r - r¢ , or from Eqs. (5-100) and (5-103c), on v = r - r¢ 2 . For Kolmogorov turbulence
in the near field, since F = W [see Eq. (5-79)], we may write from Eq. (5-69)
53
F ( v ) = 6.88(vD r0 ) . (5-152)
The MCF for the residual phase aberration after correction of the first J modes is
given by
r v
M J ( v ) = expÏÌi F J (r) - F J (r¢ )
[ ]2 ¸ = exp - (1 2) ( v )
˝ [ J ] . (5-153)
Ó ˛
The MTF at a certain frequency or the Strehl ratio for a given value of D r0 improves as
more and more modes are corrected. Of course, J = • represents complete correction,
i.e., the aberration-free case.
502 RANDOM ABERRATIONS
From Eq. (5-146), the wavefront tilt aberration introduced by turbulence is given by
[
= 2 a 2 R11(r) cos q + a 3 R11(r) sin q ]
= 2(a 2r cos q + a 3r sin q) . (5-155)
F t ( x , y ) = 2(a 2 x + a 3 y ) . (5-156)
l Ê ∂F t ∂F t ˆ
(x t , y t ) = 2F Á ,
2 p Ë ∂ x ∂y ¯
˜
5.3 Imaging through Atmospheric Turbulence 503
2l F
=
p
(a 2 , a 3 ) . (5-157)
1
(a , b) = (x , y )
R t t
2l
= (a , a ) .
pD 2 3
(5-158)
a 22 = a 32
5/3
Ê Dˆ
= 0.499 Á ˜ . (5-160)
Ë r0 ¯
Accordingly, the standard deviation of the angular fluctuations based on the Zernike tilts
are given by
2 l 12
(s a )Z = sb( ) Z
= a2
p D 2 [ ] . (5-161)
53
Since a 22 ~ (D r0 ) , Eq. (5-161) shows that the mean square fluctuation of the image
displacement decreases with the pupil diameter D as D 1 3 . Moreover, since r0 ~ l6 5 ,
we find that the fluctuation is independent of the wavelength, as expected in the absence
of any atmospheric dispersion. Substituting Eq. (5-160), we obtain the Z-tilt
2 l 12
(s a )Z = sb( ) Z
=
pD 2
a2 [ ]
56
l Ê Dˆ
= 0.4265 Á ˜ . (5-162)
D Ë r0 ¯
However, the latter approach gives insight into why the G-tilt is different from the Z-tilt,
why the two tilts are approximately equal to each other, and why the G-tilt is smaller.
From Eq. (1-180), the angular coordinates of the centroid (neglecting amplitude
variations introduced by turbulence) can be written
1 ÛÛ (
È ∂F x , y ) (
∂F x p , y p ) ˘˙ dx
(a , b) =
kSex ÙÙ
(
ÙÙ A x p , y p ) Í
Í
p
∂x p
p
,
∂y p ˙
p dy p , (5-163)
ıı Î ˚
( )
where A x p , y p defines the region of the pupil with a value of unity inside it and zero
outside. Because of the random nature of the aberration function F x p , y p introduced ( )
by atmospheric turbulence, their mean value a = b = 0. The mean square fluctuation
of a is given by (with a similar equation for b )
1 ÛÛ ÛÛ ∂F( x , y ) ∂F( x ¢ , y ¢ )
(sa2 )G =
k 2 Sex
2 ÙÙ A( x , y ) A( x ¢ , y ¢ )ÙÙ
ıı ∂x
,
∂x ¢
dx dy dx ¢ dy ¢ , (5-164)
ıı
where we have dropped the pupil subscript p for simplicity. The quantity in the angular
brackets can be written
∂F( x , y ) F( x ¢ , y ¢ ) ∂2
, = F( x , y )F( x ¢ , y ¢ )
∂x ∂x ¢ ∂x∂x ¢
∂2
∫ RF ( x - x ¢ , y - y ¢ )
∂x∂x ¢
∂2 È 2 1 ˘
=
∂x∂x ¢ ÍÎ F - 2 F ( x - x ¢ , y - y ¢ )˙˚
1 ∂2
= - ( x - x ¢, y - y ¢)
2 ∂x∂x ¢ F
1 ∂2
= ( x - x ¢, y - y ¢) , (5-165)
2 ∂x 2 F
where we have used Eqs. (5-76) and (5-77) for the phase structure and correlation
functions. Substituting Eq. (5-165) into Eq. (5-164), we may write
1 1 ÛÛ ∂2
(sa2 )G = 2 ÙÙ
2 k 2 Sex ıı
G ( x , y )
∂x 2
F ( x , y ) dx dy , (5-166)
where
G ( x, y) = ÚÚ A( x ¢ - x, y ¢ - y) A( x ¢, y ¢) dx ¢ dy ¢ (5-167)
represents the overlap area of two pupils with their centers separated by ( x , y ) . Because
of the isotropy of turbulence, the phase structure function and its derivatives can be
written
5.3 Imaging through Atmospheric Turbulence 505
F ( x , y ) ∫ F ÊË x 2 + y 2 ˆ¯ = F ( r ) (5-168a)
∂ ∂r x
( r ) = F¢ ( r ) = F¢ ( r ) (5-168b)
∂x F ∂x r
and
∂2 x2 y2
F ( r ) = ¢¢
F ( r ) + ¢
F ( r ) (5-168c)
∂x 2 r2 r3
where r is the separation distance between two circular pupils, each of diameter D.
Substituting Eq. (5-168d) into Eq. (5-166), we obtain
D 2p
1 1 Û Û È 1 ˘
( )
s a2
G
= 2 2 Ù Ù
2 k Sex ı ı
G ( r ) Í F¢¢( r ) cos 2 q + F¢ ( r ) sin 2 q ˙ rdrdq
Î r ˚
0 0
D
8 Û È 1 ˘
= Ù G ( r ) Í F¢¢( r ) + r F¢ ( r ) ˙ rdr . (5-169)
pk 2 D 4 ı Î ˚
0
1 6.88 25
F¢¢( r ) + ¢ ( r ) = 5 3 1 3 . (5-170)
r F r0 r 9
The fractional overlap area of two circular pupils each of diameter D separated by a
distance r is given by Eq. (2-44a), where v = r D . Hence,
D2 È 2 ˘
G (r) =
2 ÎÍ
cos 1
(r D) - (r D) 1 - (r D) ˙ .
˚
(5-171)
2 53
Ê l ˆ Ê Dˆ
= 0.1698 Á ˜ Á ˜ , (5-172)
Ë D ¯ Ë r0 ¯
or
506 RANDOM ABERRATIONS
56
l Ê Dˆ
(s a )G = 0.412 Á ˜
D Ë r0 ¯
, (5-173)
Now we determine the G-tilt from the Zernike expansion of the aberration function.
Neglecting the amplitude variations introduced by turbulence, the centroid ( x c , y c ) of an
aberrated PSF is given by the mean value of the gradient of the aberration function23
R ÛÛ È ∂W ( x , y ) ∂W ( x , y ) ˘
(x c , y c ) =
Sp ÙÙ Í , ˙ dx dy , (5-174)
ıı Î ∂x ∂y ˚
• •
W (r, q) = Â
n =0 m =0
[
 2 (n + 1) (1 + d m 0 ) ]1 2 Rnm (r) (c nm cos m q + snm sin m q) , (5-175)
where c nm and snm are the Zernike coefficients, it is shown in Section 2.9 that the
centroid of the corresponding PSF is given by 23
•
(x c , y c ) = 2F Â ¢ 2(n + 1) (c n1, sn1) , (5-176)
n =1
where a prime on the summation sign indicates a summation over only the odd integral
values of n. Thus, the only aberrations that contribute to the centroid are those with
m = 1. Aberrations of the type R1n (r) cos q contribute to x c , and those of the type
R1n (r) sin q contribute to y c , as may also be seen from symmetry considerations of the
aberrations. Keeping the tilt and only the primary coma terms (and thus neglecting the
higher orders of coma owing to their small magnitudes) in Eq. (5-176) and dividing by R,
we obtain the angular coordinate of the centroid,
2
a =
D
(
2 c11 + 8c 31 )
l
=
pD
(
2 a2 + 8a8 ) . (5-177)
l 12
=
pD [
4 a 22 + 8 a 82 + 8 2 a 2 a 8 ] . (5-178)
56
l Ê Dˆ
(s a )G = 0.4132 Á ˜
D Ë r0 ¯
. (5-179)
The G-tilt is smaller than the Z-tilt by approximately 3%. It is not surprising that the Z-
and G-tilts are approximately equal to each other. Since the Zernike-tilt aberrations
account for 87% of the aberration variance, the variance of the coma aberration(s) is
relatively small. The G-tilt is smaller than the Z-tilt because of the negative correlations
of the tilt coefficients with the coma coefficients.
It is evident from Table 5-6 or 5-7 that 87% of the phase aberration introduced by
turbulence is a random wavefront tilt caused by large eddies and represented by the
coefficients a 2 and a 3 . Thus, if the tilt is corrected, the variance of the phase aberration
is reduced by a factor of 1.03 0.134 ~ 7.7 . Hence, a significant improvement in the
quality of an image is obtained by correcting the wavefront tilt in (near) real time with a
steering mirror. A short-exposure image is equivalent to a tilt-free image, and a tilt-
corrected image yields a time-averaged short-exposure image. The value of s F without
any corrections is 1 radian when D = r0 . However, if the x and y tilts are corrected, we
find from the expression for D 3 that D can be as large a 3.34 r0 for the same value of s F .
Of course, if D r0 is very large, the residual aberration may still be too large for any
significant improvement of image quality.
This result simply represents the fact that a displacement of the image produces a linear
phase factor in the OTF, i.e., a linearly varying PTF. Taking the time-average of both
sides, we obtain
r r r r
◊
t (v ) = exp ( - 2 p i v rt ) t t (v ) . (5-182)
r r
Assuming that the random variations of exp (2p i v rt ) are uncorrelated with those of ◊
t t (v) , we may write Eq. (5-182)
r r r r
◊
t ( v ) = exp (-2p i v rt ) t t ( v )
r r ˘ t ( vr )
= exp ÈÍ- 2 p 2 (v ◊ rt )2 ˙˚ t
Î
r r r
[
= exp - p 2 v 2 rt rt ◊ ] t t (v ) , (5-183)
r
where in the middle step we have used Eq. (5-53) and the fact that rt is a Gaussian
random variable with a mean value of zero.
53
Letting x t and y t be in units of l F and noting that a 22 = a 32 = 0.448(D r0 ) ,
we may write from Eq. (5-157)
r r
◊
rt rt = xt2 + yt2
(
= 8 p 2 a 22) (5-184a)
2 53
= 3.584(v p) (D r0 ) . (5-184b)
Substituting Eqs. (5-107) and (5-184b) into Eq. (5-183), the tilt-corrected MTF of the
overall system may be written in the form
r r
[
t t ( v ) = t ( v ) exp 3.584 v 2 (D r0 )
53
]
r
= t L ( v ) M t ( v) , (5-185)
where
[
M t ( v ) = M ( v ) exp 3.584 v 2 (D r0 )
53
] (5-186a)
is the corresponding MCF representing the MTF reduction factor or the Hopkins ratio
associated with the tilt-corrected atmospheric turbulence. The MCF may be written in the
form
[
M t ( v ) = exp - (1 2)t ( v ) ] , (5-187)
5.3 Imaging through Atmospheric Turbulence 509
where
t ( v ) = F ( v ) - 16v 2 a 22
= 6.88 (vD r0 )
53
(1 - 1.042v1 3 ) (5-188)
The Strehl ratio and the h-function of the short-exposure image can be obtained
from the OTF. Thus,
1
St (D r0 ) = 8 Ú t t (v; D r0 ) v d v (5-189)
0
and
2
Ê Dˆ
ht (D r0 ) = Á ˜ St (D r0 ) . (5-190)
Ë r0 ¯
Figure 5-18a shows how the short-exposure MCF M t ( v ) varies with v for a few
values of D r0 . The figure illustrates what is evident from Eq. (5-186), that M t ( v ) > 1
for 0.884 < v < 1, regardless of the value of D r0 , i.e., the MTF reduction factor is
greater than unity in the high-frequency range. This has the consequence that the tilt-
r r
corrected MTF t t ( v ) is larger than the turbulence-free MTF t L ( v ) . This, of course,
is unphysical and hence incorrect, since the aberrations always reduce the MTF at every
spatial frequency [see Eq. (1-107)]. Equation (5-186b) and, in turn, Eq. (5-183) were
obtained from Eq. (5-182) under the assumption that the random variations of the
wavefront tilt and the tilt-corrected OTF were uncorrelated. This, in turn, implies that the
random wavefront tilt and the tilt-corrected aberrations introduced by turbulence are
uncorrelated. And that, of course, is not true, as we see from Tables 5-6 and 5-7. The tilt
aberration coefficients a2 and a3 are correlated with coma terms; or the correlations
a2 a8 , a3 a7 , a2 a16 , a3 a17 , etc. are not zero. Of course, the MTF reduction factor
given by Eq. (5-186) gives an incorrect value at all spatial frequencies within the
passband of the system, and not just the high-frequency region of 0.884 < v < 1. Fried5
circumvented this problem by replacing the number 1.042 in Eq. (5-186) by unity.
(Actually, his number was 1.026 owing to the use of a slightly different value of the
variance of the wavefront tilt aberration.) The MTF reduction factor thus obtained is £ 1,
as illustrated in Figure 5-18b. Although its value is physically satisfactory, it is
nevertheless incorrect.
510 RANDOM ABERRATIONS
1.2
Forbidden Region
1.0
0.8
~
>
'--'
~- 0.6
t
0.4 \
\2
5
0.2 \
\
\ 5
0
0 0.2 0.4 0.6 0.8
---+ v
(a)
0.8
Dlro = 2
~
0.6
\
> \
'--'
~-
t
0.4
\
\2
\
0.2 5
,
ooL----L----o~
.2-2~~----
OL.4----L---~0~.6~--~----OL
.8----L---~
---+ v
(b)
Figure 5-18. (a) Near-field short-exposure atmospheric MTF reduction factor
M t (v) given by Eq. (5-186b) and shown by the solid curves, illustrating the error
resulting from ignoring the correlation of the tilt aberration coefficient with the
coma coefficients. M t (v) > I is the forbidden region, since it makes the overall
system MTF greater than the corresponding diffraction-limited MTF. (b) MTF
reduction factor obtained when 1.042 in Eq. (5-186b) is replaced by unity. The
dashed curves represent the long-exposure or the uncorrected MTFs shown earlier
in Figure 5-11a.
5.3 Imaging through Atmospheric Turbulence 511
Figure 5-18c shows how the Strehl ratio St varies with D r0 . The uncorrected
Strehl ratio is also shown to illustrate the improvement made by the tilt correction. The
( ) 53
Strehl ratio estimate given by exp - s 2F , where s 2F = 0.134( D r0 ) is the tilt-corrected
time-averaged phase variance, and shown by the dashed curve, approximates the true
value reasonably well for D r0 £ 6 .
Figure 5-18d shows how ht varies with D r0 . For small values of D r0 , it increases
approximately as for a diffraction-limited system, since the aberration is small. However,
for D r0 > 11.4 , its value is wrongly greater than the corresponding diffraction-limited
2
value of ( D r0 ) . If we define htc as the h-function obtained by replacing 1.042 by
unity, its value, as shown in Figure 5-18d, is approximately equal to that of ht for small
values of D r0 , reaches a maximum of 3.5 for D r0 = 3.74 , and then decreases slowly
but monotonically to unity. Since the image displacement due to wavefront tilt decreases
as D 1 3 , the effect of tilt correction becomes negligible for large values of D r0 due to
the large residual phase errors.
Considering the correlation of the tilt coefficients with only the primary coma
coefficients (and thus neglecting their correlation with the coefficients of higher-order
coma), the short-exposure phase structure function 3 ( v ) can be written6
( )
3 ( v ) = F ( v ) - 16v 2 a 22 - 32 2 v 2 6v 2 - 6v + 1 a 2 a 8
( )
= 6.8839v 5 3 1 - 0.9503v1 3 - 0.5585v 4 3 + 0.5585v 7 3 (D r0 )
53
. (5-191)
53
It has a maximum value of 0.344(D r0 ) at v = 1, which is approximately 1/20 of the
corresponding value F ( v ) of the long-exposure image (i.e., without any correction or
for J = 1), illustrating a significant benefit of the tilt correction.
with
[
M 3 (v; D r0 ) = exp -(1 2)3 (v; D r0 ) ] . (5-193)
512 RANDOM ABERRATIONS
100
exp( s 2F )
<S>
10– 1
< St >
<S>
10– 2
2 4 6 8 10
D/r0
Figure 5-18. (c) Near-field tilt-corrected time-averaged Strehl ratio St . The result
obtained from Eq. (5-189) is shown by the solid curve, and its approximate value is
indicated by the dashed curve according to exp -s F 2
(
, where sF )
2 53
= 0.134( D r0 ) .
The corresponding uncorrected Strehl ratio S , shown earlier in Figure 5-10b, is
shown here for comparison.
103
102
ht
101
(D/r0)2
h
h tc
100
h
10–1
10–2 –1
10 100 101 102
D/r0
Figure 5-18. (d) Variation of near-field h with D r0 for long- and short-exposure
images. h : Long exposure or uncorrected; ht : Short exposure or tilt corrected, but
correlation of coefficients neglected; htc : Short-exposure or tilt-corrected with 1.042
2
replaced by unity; ( D r0 ) : Diffraction limited.
5.3 Imaging through Atmospheric Turbulence 513
It is evident that the correlation yields a slightly higher value for both, though the
effect is negligible for D/ ro :s 5. 24 The central irradiance 113 peaks at D/ ro = 4.3 with a
value of 5.3, but 11 tc peaks at 3.74 with a value of 3.5.
0.5
0.4 . . .... 2lltc
0.3 ,,
0.2
,,
\
(f)
t (b)
- O/ro
103
102
I=" 101
t 100 ~~======~ (c)
10- 1
10- 2
10- 1
- O/ro
Figure 5-19. (a) Comparison of 9lt {v) and 9ltc {v) with 9l3 {v) as a function of v, (b)
(SI) and (SIC) with S3 as a function of D/ro, and (c) TIL and Tl lc with Tl3 of Figure
5-18d.
514 RANDOM ABERRATIONS
a 42 = a 52 = a 62 , or a 72 = a 82 = a 92 = a10
2
. (5-195)
However, when π 0 , the variance depends on the values of both n and m, but it is
independent of whether it is a cosine or a sine mode. Thus, for example,
a 42 ( ) π a 52 ( ) = a 62 ( ) , or a 72 ( ) = a 82 ( ) π a 92 ( ) = a10
2
( ) . (5-196)
Similarly, the covariance for given values of n and n ¢ are independent of their m values
when = 0. For example,
However, when π 0 , equality is obtained only when the m values are also equal. For
example,
or
Figures 5-20 and 5-21 show the variance and covariance of Zernike coefficients as a
function of for some low-order terms. The variance of terms with m = n increases
monotonically with an increasing value of , e.g., tip and tilt variances a 22 and a 32 , or
astigmatism variances a 52 and a 62 . For other terms, e.g., coma a 72 and a 82 , or
5.3 Imaging through Atmospheric Turbulence 515
0.9
0.8
...............
M
~
.~
0.7 (a)
-----
0.6
0.5 (a:),(a:)
0.4
0 0.2 0.4 0.6 0.8
E
0.05
0.04
0.03
...............
M
~
.~
(b)
----- 0.02
0.01
0
0 0.2 0.4 0.6 0.8
E
0.007
0.014
.,.',-
0.006 ./
0.01 2
/ /
0.005 ,/
0.01 ./
-.. //~
./
/'
(an, (a,'o) ___ 0.004 ....// (a"4),(a~5 )
0.008
~--.
~
/
-.......- 0.003 .---
0.006 ----'--'0"
0.002
(a~ ),(an
0.004
(a~, ), (a~3 )
0.002 0.00 1 (a~2 )
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
E E
(c) (d)
-0.002
-0.004
-0.006
----- ' ",
(::l
---
'",
(::l -0.008
-0.01
-0.012
-0.014
(a)
0.001
-0.001
----- '",
(::l
(a 4all )
---
'",
(::l
-0.002
-0.003
-0.004 '--------'--------'---------'---------'---------'
o 0.2 0.4 0.6 0.8
(b)
defocus (a~) and spherical aberration (a)2)), they generally decrease monotonically and
approach zero as E~ 1. Similarly, the covariances of tilt with coma, e.g. , (a2(E)a8(E)) ,
and defocus with primary or secondary spherical aberration, i.e. , (a 4 (E) a)) (E)) or
(a 4(E)a22 (E)), approach zero as E~ 1.
Since the wavefront tilt produced by turbulence is independent of the shape of the
pupil , except for its correlation with the coma terms , (ai(E)) increases with E
approximately as 1 + E2. Accordingly, the angular fluctuation depends very weakly on the
obscuration ratio. For example, when E= 0.5, the constant 0.4265 on the right-hand side
ofEq. (5-162) is replaced by 0.4246.
Figure 5-22 shows how Ll)(E) == ()~(E) in units of (D/ro)5/3 varies with E. Its value
increases monotonically from 1.03 for a circular pupil (E= 0) to 1.84 for an
infinitesimally thin ring pupil (E ~ 1). The figure also shows the variation of Ll3 (E) (the
piston-removed and tilt-corrected variance) with E. It is nearly constant for E:S 0.5.
Figure 5-23 shows the relative tilt-corrected variance [(Ll) (E) - Ll3(E))/ Ll) (E)] as a function
of E. It increases from a value of approximately 87% for a circular pupil to nearly 91 % as
E~ 1.
2 0.2
1.8
0.175
1.6
0.15
1.4 Ll3
................................................., ....
0.125
1.2
0.1
o 0.2 0.4 0.6 0.8
Figure 5-22. Piston-removed aberration variance Ll) (E) == a! (E) and tilt-corrected
variance Ll3 (E) as a function of E.
518 RANDOM ABERRATIONS
0.91
0.9
<l
-...
,-.,
M 0.89
<lI
~
0.88
0.87
o 0.2 0.4 0.6 0.8
5-23. Relative tilt-corrected phase variance [([\1 (E) - [\3 (E))/ [\1 (E)] as a function of
the obscuration ratio E.
(5-200)
21
Ê Dˆ
St (D r0 ) = 8Á ˜ Ú t L (v) M t (v; D r0 ) v d v (5-201)
Ë r0 ¯ 0
and
2
Ê Dˆ
ht (D r0 ) = Á ˜ St (D r0 ) , (5-202)
Ë r0 ¯
respectively. The number 0.521 in Eq. (5-198) may be replaced by 0.5 for consistency
with the corresponding near-field images in which case 1.042 was replaced by 1 in Eq.
(5-186b).
When the phase errors are corrected completely, the MCF reduces to
[
M t ( v ) = exp - 1.72 (vD r0 )
53
] , (5-203)
with the corresponding equations for the Strehl ratio and the h-function. Figure 5-24a
shows the aberration-corrected MTF t c = t L M t as a function of the spatial frequency.
0.8
D/r0 = 0
0.6
< tc >
2
0.4
5
0.2
0
0 0.2 0.4 0.6 0.8 1
v
100
< Sc >
<S>
10– 1
<S>
< St >
10– 2
2 4 6 8 10
D/r0
hc
ht
100
(D/r0)2
h
h
10–1
10–2
10–1 100 101 102
D/r0
It is evident that even with full correction, its value at any frequency is less than the
corresponding diffraction-limited value, and decreases as D r0 increases. This is a
consequence of M t being less than unity due to scintillation or amplitude variations
produced by turbulence. Perhaps it is worth remembering that, in principle, the MTF of a
system with nonuniform amplitude across the pupil can be higher at some frequencies, as
in the case of a Gaussian pupil (see Figure 4-4). Figure 5-24b shows how the tilt-
corrected Strehl ratio St , aberration corrected Sc , and no correction S vary with
D r0 . It is seen that the improvement even with full aberration correction is not that
significant. Figure 5-24c shows the variation of ht with D r0 . It is evident that the
improvement in this case is also much smaller. The performance improves with full
aberration correction, but falls significantly short of its diffraction-limited value.
D /r0 = 10
s w = 0.4l
Aberration
(a)
(b) (c)
Figure 5-25. Aberration introduced by atmospheric turbulence corresponding to
D r0 = 10 . (a) Aberration shape. (b) Aberration interferogram. The standard
deviation of the tilt-free aberration introduced by turbulence is 0.4 l . (c)
Interferogram with 25 l of tilt.
522 RANDOM ABERRATIONS
In Figure 5-26b, the value of r0 is kept fixed while the value of D increases from r0
to 3 r0 and to 10 r0 , e.g., r0 = 10 cm and D = 10 cm , 30 cm, and 1 m, respectively. Now
the size of a speckle decreases as D increases, but the overall image size is approximately
constant. Thus, an increase in D does not significantly improve the resolution of the
system (as determined by the overall image size). For convenience, the pictures in Figure
5-26b are shown reduced by a factor of 1.5 compared to those in Figure 5-26a. Thus, for
example, the pictures corresponding to D r0 = 10 in these two figures are otherwise
similar.
The spot radius representing the point at which the irradiance drops to half of its
central value is given in Table 5-9 for both the short- and long-exposure images for
various values of D r0 , and compared with the corresponding diffraction-limited value.
We note that, whereas the LE radius is roughly D r0 times the diffraction-limited (DL)
radius, the SE radius for small values of D r0 , e.g., D r0 = 5, is roughly equal to the DL
radius, but for large values of D r0 , e.g., D r0 = 10 , it is about half the LE radius. Of
course, as increases and more and more light goes into the diffraction rings, these
factors change somewhat. For small values of D r0 , the diffraction ring structure of the
DL image, smoothed out in the LE image, returns in the SE image. This is more and more
evident for larger and larger values of .
If the image wander is corrected in real time and the image is averaged over time, the
speckles disappear and the image becomes smooth. Such PSFs and the corresponding
encircled power can be obtained from Eqs. (5-116) and (5-118), respectively, by
substituting the short-exposure OTF given by Eq. (5-192). The PSF and encircled power
thus obtained are shown by the dashed curves in Figure 5-16a and 5-16b, respectively, for
several values of . The short-exposure image when randomly displaced over time yields
a smooth long-exposure image, as illustrated by the dotted curves in these figures. Modal
correction of a Gaussian beam (including one with a central obscuration) propagating
through turbulence has also been discussed by Wang using the Zernike circle
polynomials,26 though such polynomials are not orthogonal over the Gaussian-weighted
(or an annular) pupil, as discussed in Chapters 3 and 4.
5.3 Imaging through Atmospheric Turbulence 523
D r0 = 1
D r0 = 3
D r0 = 10
D r0 = 1
D r0 = 3
D r0 = 10
Table 5-9. Angular spot radius (representing the point at which the irradiance is
50% of its central value) of short-exposure (SE), long-exposure (LE), and
diffraction-limited (DL) images in units of l D for various values of obscuration
ratio and D r0 .
\ D/ r0 1 2 5 10
DL LE SE LE SE LE SE LE SE
2
P ª 5.6 exp[ -0.1557(D r0 ) ] , D r0 ≥ 3.5 . (5-204)
Thus an observer can take a series of short-exposure images over a time period before the
turbulence characterized by r0 changes significantly. The good-quality images are
selected based on their full-width at half maximum or peak pixel value normalized by the
total illumination, centered on the peak pixel, and summed in a process called shift and
add. This approach has been referred to as lucky imaging.29-31
The image quality can also be improved by correcting the wavefront errors in (near)
real time with adaptive optics.7-9 In practice, a steering mirror with only three actuators is
used to correct the large tip and tilt of the wavefront. The residual aberration is corrected
by a deformable mirror, which is deformed by an array of actuators attached to it. The
signals for the actuators are determined either by sensing the wavefront errors with a
wavefront sensor in a closed loop to minimize the variance of the residual errors, or the
526 RANDOM ABERRATIONS
(a) (b)
(c) (d)
(e) (f)
(g)
Figure 5-27. Lucky imaging. (a) Poor-quality SE image, (b) corresponding good-
quality image, (c) LE image obtained by adding all of the 50,000 SE images taken at
a rate of 40/sec over a period of about 21 minutes, (d) LE image obtained by aligning
the centroids of the SE images, (e) 50% of the selected SE images aligned by peak
pixel value and added, (f) 10% of SE images added, and (g) 1% of SE images added.
5.4 Summary 527
actuators are actuated to produce Zernike modes (e.g., focus, two modes of astigmatism,
two modes of coma, etc.) iteratively until the sharpness of the image is maximized (see
Problem 1.7).32-34 The signals are independent of the optical wavelength provided
atmospheric dispersion is negligible. The two approaches are referred to as zonal and
modal approaches, respectively. The zonal approach has the advantage that the rate of
correction is limited only by the rate at which the wavefront errors can be sensed and the
actuators can be actuated. However, the amount of light that is used by the wavefront
sensor is lost from the image. In practice, however, the image beam is split into two parts.
The centroid of the image of one part is measured with a quad cell, and the tilt indicated
by it is corrected with a steering mirror. The resulting tilt-corrected image of the other
part with the residual aberration is corrected with a deformable mirror in a closed-loop
manner.35 In the modal approach, there is no loss of light, but the rate or the bandwidth of
correction 36-38 can be slow due to its iterative nature, especially when turbulence is severe
and a large number of modes must be corrected. Moreover, for imaging an extended
object, wavefront sensing requires a point source in its vicinity, but the modal approach is
applicable to the extended object itself. Adaptive optics has also been used in lucky
imaging to achieve diffraction-limited image quality.39
Of course, adaptive optics can improve the image quality only if the object lies
within an isoplanatic angle of turbulence. In the case of a ground-to-space laser
illuminating a satellite, the angular travel (point-ahead angle) of the satellite during the
round-trip time of the beam to the satellite must be less than the isoplanatic angle of
turbulence. An estimate of the isoplanatic angle can be obtained from 40
35
85 ÏL 2 53 ¸
J 0 = 0.058l6 5 ( cos J) Ì Ú C n ( z ) z dz ˝ . (5-205)
Ó0 ˛
Substituting for C N2 from Eq. (5-71) into Eq. (5-205), we obtain J 0 = 7 mrad for
l = 0.5 mm for a zenith observation. Thus, the H - V5 7 model for C n2 yields r0 = 5 cm
and J 0 = 7 mrad for visible light; hence the 5 7 subscript. If the object lies at an angle J
with the direction of the adaptive optics correction, the Strehl ratio of the image is
reduced by a factor of exp(- J J 0 ) .
5.4 SUMMARY
This chapter starts with a discussion of the effect of Gaussian random image motion.
Its focus, however, is on the aberrations introduced on a wave propagating through
atmospheric turbulence. A brief summary of the key results of this chapter is given
below.
The effect of image motion on the Strehl ratio, PSF, and OTF of a system is
discussed. It is shown that the drop in Strehl ratio for a certain value of the transverse
image motion is larger for a larger obscuration ratio of the system pupil, but it is smaller
in the case of the longitudinal image motion.
528 RANDOM ABERRATIONS
The effect of turbulence is discussed in terms of the MTF, Strehl ratio, PSF, and
encircled power for both long- and short-exposure images. The aberration function
introduced by turbulence is expanded in terms of the Zernike circle polynomials, and the
variance and correlation of the expansion coefficients are discussed. However, the effect
of the correlation of Zernike coefficients is negligible for D r0 <
~ 5 . The piston-removed
53
aberration variance in units of (D r0 ) increases monotonically from a value of 1.03 for
a circular pupil ( = 0) to a value of 1.84 for an infinitesimally thin ring pupil ( Æ 1)
(see Figure 5-22). The angle of arrival and the consequent image displacement varies
with the pupil diameter as D 1 3 . The image wander is approximately one-third of the
radius of the Airy disc [see Eq. (5-162), (5-173), or (5-179)]. It depends weakly on the
pupil obscuration. The time-averaged variance of some aberration coefficients increases
with and decreases for others (Figure 5-20). For example, the variance of tilt and
astigmatism increases, but that of defocus and spherical aberration decreases and
approaches zero as Æ 1. The covariance follows a similar trend (see Figure 5-21).
APPENDIX
r
where k is a 2D dimensionless spatial frequency variable as before. Using polar
r r
coordinates (k, f) and (r, q) for k and r , respectively, we may write
r
k = k (cos f, sin f) (5A-2a)
and
r
r = r (cos q, sin q) , (5A-2b)
so that
r r
◊
k r = k r cos (q - f) . (5A-3)
Let us consider an even value of j. Substituting Eq. (5-109a) into Eq. (5A-4), we obtain
1 2p
Seven j (k, f) = [ ]
2(n + 1) Ú Rnm (r) r dr Ú exp 2 p i k r cos (q - f) cos mq dq . (5A-5)
0 0
Letting
q-f = a , (5A-6)
2p
Û
= cos mf Ù exp(2 p i k r cos a ) cos ma da
ı
0
Appendix: Fourier Transform of Zernike Circle Polynomials 531
= 2 p i m J m (2 p k r) cos mf , (5A-7)
where we have used the fact that the integral over 0 to 2p of an odd function of a is zero
and that41
p
Ú exp(i cos x ) cos nxdx = p i Jn () .
n
(5A-8)
0
J n +1 (2p k)
2(n + 1) i m ( -1)(
n m) 2
= cos mf , m π 0 , (5A-9)
k
J n +1 (2 p k )
Sodd j (k, f) = 2(n + 1) i m ( -1)( n m) 2
sin mf , m π 0 , (5A-11)
k
and
J n +1 (2 p k )
S j (k , f) = n + 1( -1)n 2 , m=0 . (5A-12)
k
The Fourier transform of a Zernike annular polynomial can be obtained by writing its
radial polynomial in terms of the circle radial polynomials and then using the Fourier
transform of the circle polynomials. 6
532 RANDOM ABERRATIONS
REFERENCES
1. V. N. Mahajan, “Degradation of an image due to Gaussian image motion,” Appl.
Opt. 17, 3329–3334 (1978).
11. D. Fried, “Evaluation of r0 for propagation down through the atmosphere,” Appl.
Opt. 13, 2620–2622 (1974); errata 1, Appl. Opt. 14, 2567 (1975); errata 2, Appl.
Opt. 16, 549 (1977).
17. R. J. Noll, “Zernike polynomials and atmospheric turbulence,” J. Opt. Soc. Am.
66, 207–211 (1976).
21. V. N. Mahajan, Optical Imaging and Aberrations, Part I: Ray Geometrical Optics
(SPIE Press, Bellingham, Washington, Second Printing, 2001).
23. V. N. Mahajan, “Line of sight of an aberrated optical system,” J. Opt. Soc. Am. A
2, 833–846 (1985).
24. J. Y. Wang, “Optical resolution through a turbulent medium with adaptive phase
compensation,” J. Opt. Soc. Am. 67, 383–390 (1977).
25. G.-m. Dai, “Modal compensation of atmospheric turbulence with the use of
Zernike polynomials and Karhunen Loève functions,” J. Opt. Soc. Am. A 12,
2182–2193 (1995).
31. C. Mackay, J. Baldwin, N. Law, and P. Warner, “High resolution imaging in the
visible from ground without adaptive optics: New techniques and results,” Proc.
SPIE 5492, 128–135 (2004) [doi: 10.1117/12.550443].
37. C. B. Hogge and R. R. Butts, “Frequency spectra for the geometric representation
of wavefront distortions due to atmospheric turbulence,” IEEE Trans. Antennas
Propagation AP-24, 144–154 (1976).
39. C. Mackay, N. Law, and T. D. Stayley, “Diffraction limited imaging in the visible
from large ground-based telescopes: New methods for future instruments and
telescopes,” Proc. SPIE 7014, 7014C– 7014C-7 (2010) [doi: 10.1117/12.787439].
40. D. L. Fried, “Anisoplanatism in adaptive optics,” J. Opt. Soc. Am. 72, 52–61
(1982).
42. M. Born and E. Wolf, Principles of Optics, 7th ed. (Cambridge University Press,
New York, 1999), p. 910.
Problems 535
PROBLEMS
1. Consider a beam of light of wavelength l = 0.5 mm incident parallel to the optical
(or z) axis of a beam expander made up of two confocal (i.e., common focus)
paraboloidal mirrors. Let the diameters and the focal ratios of the mirrors be
D1 = 10 cm, D2 = 50 cm, and F1 = F2 = 2 . Let the exit pupil of the system be at its
secondary mirror.
(a) Determine the change in mirror spacing required to focus the beam at a distance
of 1 km.
(b) Determine the time-averaged Strehl ratio if M1 moves randomly along the x
and y axes with a standard deviation of 0.5 mm.
W (r, q) = (l 4) r3 cos q, 0 £ r £ 1 ,
then based on the centroid of the aberrated irradiance distribution in the focal plane,
determine the beam line of sight in units of l D2 .
(d) Determine the mirror figure fabrication tolerance for a Strehl ratio of 0.9 for the
beam expander.
(a) Determine D such that a long-exposure image is obtained with a Strehl ratio of
0.8 without the use of adaptive optics.
(c) Now assume that the value of D obtained in problem (b) is used to obtain a
long-exposure image of Strehl ratio 0.8 with the aid of a steering mirror. Specify
the angular steering requirements and compare them with the angular radius of the
Airy disc.
(d) Determine the total wavefront error in units of l if D = r0 . How much of the
error is a wavefront tilt (in units of l D) and how much is a wavefront
deformation (in units of l )? How do the total wavefront error and tilt angle scale
as D is increased?
(e) Now consider D = 1m. If a deformable mirror is used to correct the wavefront
deformation, specify the range of actuators needed.
Bibliography
M. Born and E. Wolf, Principles of Optics, 7th ed. (Cambridge University Press, New
York, 1999).
J. D. Gaskill, Linear Systems, Fourier Transforms, and Optics (John Wiley & Sons, New
York, 1978).
J. W. Goodman, Introduction to Fourier Optics, 3rd ed. (Roberts & Co., Greenwood
Village, CO, 2005).
V. N. Mahajan, Aberration Theory Made Simple, 2nd ed. (SPIE Press, Bellingham, WA,
2011) [doi: 10.1117/3.903924].
V. N. Mahajan, Optical Imaging and Aberrations, Part I: Ray Geometrical Optics (SPIE
Press, Bellingham, WA, 1998) [doi: 10.1173/3.265735].
D. J. Schroeder, Astronomical Optics, 2nd ed. (Academic Press, New York, 2000).
R. R. Shannon, The Art and Science of Optical Design (Cambridge University Press, New
York, 1997).
A. Walther, The Ray and Wave Theory of Lenses (Cambridge University Press, New
York, 1995).
537
References for Additional Reading
These references are the author’s collection as the editor of Milestone Series 74
entitled Effects of Aberrations in Optical Imaging, published by SPIE Press in 1994.
2. Lord Rayleigh, “On images formed without reflection or refraction,” Philos. Mag.
5, 214–218 (1881).
10. W. S. Kovach, “Energy distribution in the PSF for an arbitrary passband,” Appl.
Opt. 13, 1769–1771 (1974).
12. E. Wolf, “Light distribution near focus in an error-free diffraction image,” Proc.
Royal Soc. A 204, 533–548 (1951).
539
540 REFERENCES FOR ADDITIONAL READING
13. J. C. Dainty, “The image of a point for an aberration-free lens with a circular
pupil,” Opt. Comm. 1, 176–178 (1969).
14. R. E. Stephens and L. E. Sutton, “Diffraction image of a point in the focal plane
and several out-of-focus planes,” J. Opt. Soc. Am. 58, 1001–1002 (1968).
15. P. A. Stokseth, “Properties of a defocused optical system,” J. Opt. Soc. Am. 59,
1314–1321 (1969).
16. D. S. Burch, “Fresnel diffraction by a circular aperture,” Am. J. Phys. 53, 255–260
(1985).
18. T. S. McKechnie, “The effect of defocus on the resolution of two points,” Optica
Acta 20, 253–262 (1973).
20. D. K. Cook and G. D. Mountain, “The effect of phase angle on the resolution of
two coherently illuminated points,” Optical and Quan. Elec. 10, 179–180 (1978).
22. Y. Li and E. Wolf, “Focal shifts in diffracted converging spherical waves,” Opt.
Comm. 39, 211–215 (1981).
23. Y. Li, “Dependence of the focal shift on Fresnel number and f number,” J. Opt.
Soc. Am. 72, 770–774 (1982).
24. V. N. Mahajan, “Axial irradiance and optimum focusing of laser beams,” Appl.
Opt. 22, 3042–3053 (1983).
26. W. H. Steel, “The defocused image of sinusoidal gratings,” Optica Acta 3, 65–74
(1956).
32. W. H. Steel, “The problem of optical tolerances for systems with absorption,”
Appl. Opt. 8, 2297–2299 (1969).
33. V. N. Mahajan, “Strehl ratio for primary aberrations: some analytical results for
circular and annular pupils,” J. Opt. Soc. Am. 72, 1258–1266 (1982).
34. V. N. Mahajan, “Strehl ratio for primary aberrations in terms of their aberration
variance,” J. Opt. Soc. Am. 73, 860–861 (1983).
35. G. Martial, “Strehl ratio and aberration balancing,” J. Opt. Soc. Am. A 8, 164–170
(1991).
36. W. B. King, “Dependence of the Strehl ratio on the magnitude of the variance of
the wave aberration,” J. Opt. Soc. Am. 58, 655–661 (1968).
37. J. J. H. Wang, “Tolerance conditions for aberrations,” J. Opt. Soc. Am. 62, 598–
599 (1972).
39. H. H. Hopkins, “The aberration permissible in optical systems,” Proc. Phys. Soc.
B 70, 449–470 (1957).
40. W. B. King, “Correlation between the relative modulation function and the
magnitude of the variance of the wave-aberration difference function,” J. Opt.
Soc. Am. 59, 285–290 (1969).
46. V. N. Mahajan, “Zernike annular polynomials for imaging systems with annular
pupils,” J. Opt. Soc. Am. 71, 75–85; 1408 (1981).
47. V. N. Mahajan, “Zernike annular polynomials for imaging systems with annular
pupils,” J. Opt. Soc. Am. A 1, 685 (1984).
51. S. N. Bezdid'ko, “The use of Zernike polynomials in optics,” Soviet J. Opt. Tech.
41, 425–429 (1974).
56. G. Conforti, “Zernike aberration coefficients from Seidel and higher-order power-
series coefficients,” Opt. Lett. 8, 407–408 (1983).
58. R. J. Noll, “Zernike polynomials and atmospheric turbulence,” J. Opt. Soc. Am.
66, 207–2111 (1976).
61. R. Barakat and A. Houston, “Diffraction effects of coma,” J. Opt. Soc. Am. 54,
1084–1088 (1964).
64. V. N. Mahajan, “Line of sight of an aberrated optical system,” J. Opt. Soc. Am. A
2, 833–846 (1985).
65. H. H. Hopkins, “Image shift, phase distortion and the optical transfer function,”
Optica Acta 31, 345–368 (1984).
66. G. B. Airy, “On the diffraction of an annular aperture,” Philos. Mag. 18, 1–10,
132–133 (1841).
70. V. N. Mahajan, “Included power for obscured circular pupils,” Appl. Opt. 17,
964–968 (1978).
71. J. J. Stamnes, H. Heier, and S. Ljunggren, “Encircled energy for systems with
centrally obscured circular pupils,” Appl. Opt. 21, 1628–1633 (1982).
74. E. L. O'Neill, “Transfer function for an annular aperture,” J. Opt. Soc. Am. 46,
285–288 (1956).
81. Y. Li and E. Wolf, “Focal shift in focused truncated Gaussian beams,” Opt.
Comm. 42, 151–156 (1982).
85. R. Herloski, “Strehl ratio for untruncated aberrated Gaussian beams,” J. Opt. Soc.
Am. A 2, 1027–1030 (1985).
90. J. Y. Wang, “Optical resolution through a turbulent medium with adaptive phase
compensations,” J. Opt. Soc. Am. 67, 383–390 (1977).
547
548 Index