Matrix Methods For Optical Layout

Tutorial Texts Series
Matrix Methods for Optical Layout, Gerhard Kloos, Vol. TT77

Fundamentals of Infrared Detector Materials, Michael A. Kinch, Vol. TT76
Practical Applications of Infrared Thermal Sensing and Imaging Equipment, Third Edition, Herbert
Kaplan, Vol. TT75
Bioluminescence for Food and Environmental Microbiological Safety, Lubov Y. Brovko, Vol. TT74
Introduction to Image Stabilization, Scott W. Teare, Sergio R. Restaino, Vol. TT73
Logic-based Nonlinear Image Processing, Stephen Marshall, Vol. TT72
The Physics and Engineering of Solid State Lasers, Yehoshua Kalisky, Vol. TT71
Thermal Infrared Characterization of Ground Targets and Backgrounds, Second Edition, Pieter A. Jacobs,
Vol. TT70
Introduction to Confocal Fluorescence Microscopy, Michiel Mller, Vol. TT69
Artificial Neural Networks An Introduction, Kevin L. Priddy and Paul E. Keller, Vol. TT68
Basics of Code Division Multiple Access (CDMA), Raghuveer Rao and Sohail Dianat, Vol. TT67
Optical Imaging in Projection Microlithography, Alfred Kwok-Kit Wong, Vol. TT66
Metrics for High-Quality Specular Surfaces, Lionel R. Baker, Vol. TT65
Field Mathematics for Electromagnetics, Photonics, and Materials Science, Bernard Maxum, Vol. TT64
High-Fidelity Medical Imaging Displays, Aldo Badano, Michael J. Flynn, and Jerzy Kanicki, Vol. TT63
Diffractive OpticsDesign, Fabrication, and Test, Donald C. OShea, Thomas J. Suleski, Alan D.
Kathman, and Dennis W. Prather, Vol. TT62
Fourier-Transform Spectroscopy Instrumentation Engineering, Vidi Saptari, Vol. TT61
The Power- and Energy-Handling Capability of Optical Materials, Components, and Systems, Roger M.
Wood, Vol. TT60
Hands-on Morphological Image Processing, Edward R. Dougherty, Roberto A. Lotufo, Vol. TT59
Integrated Optomechanical Analysis, Keith B. Doyle, Victor L. Genberg, Gregory J. Michels, Vol. TT58
Thin-Film Design Modulated Thickness and Other Stopband Design Methods, Bruce Perilloux, Vol. TT57
Optische Grundlagen fr Infrarotsysteme, Max J. Riedl, Vol. TT56
An Engineering Introduction to Biotechnology, J. Patrick Fitch, Vol. TT55
Image Performance in CRT Displays, Kenneth Compton, Vol. TT54
Introduction to Laser Diode-Pumped Solid State Lasers, Richard Scheps, Vol. TT53
Modulation Transfer Function in Optical and Electro-Optical Systems, Glenn D. Boreman, Vol. TT52
Uncooled Thermal Imaging Arrays, Systems, and Applications, Paul W. Kruse, Vol. TT51
Fundamentals of Antennas, Christos G. Christodoulou and Parveen Wahid, Vol. TT50
Basics of Spectroscopy, David W. Ball, Vol. TT49
Optical Design Fundamentals for Infrared Systems, Second Edition, Max J. Riedl, Vol. TT48
Resolution Enhancement Techniques in Optical Lithography, Alfred Kwok-Kit Wong, Vol. TT47
Copper Interconnect Technology, Christoph Steinbrchel and Barry L. Chin, Vol. TT46
Optical Design for Visual Systems, Bruce H. Walker, Vol. TT45
Fundamentals of Contamination Control, Alan C. Tribble, Vol. TT44
Evolutionary Computation Principles and Practice for Signal Processing, David Fogel, Vol. TT43
Infrared Optics and Zoom Lenses, Allen Mann, Vol. TT42
Introduction to Adaptive Optics, Robert K. Tyson, Vol. TT41
Fractal and Wavelet Image Compression Techniques, Stephen Welstead, Vol. TT40
Analysis of Sampled Imaging Systems, R. H. Vollmerhausen and R. G. Driggers, Vol. TT39
Tissue Optics Light Scattering Methods and Instruments for Medical Diagnosis, Valery Tuchin, Vol. TT38
Fundamentos de Electro-ptica para Ingenieros, Glenn D. Boreman, translated by Javier Alda, Vol. TT37
Infrared Design Examples, William L. Wolfe, Vol. TT36
Sensor and Data Fusion Concepts and Applications, Second Edition, L. A. Klein, Vol. TT35
Tutorial Texts in Optical Engineering

Volume TT77
Bellingham, Washington USA
Library of Congress Cataloging-in-Publication Data
Kloos, Gerhard.
Matrix methods for optical layout / Gerhard Kloos.
p. cm. -- (Tutorial texts series ; TT 77)
ISBN 978-0-8194-6780-5
1. Optics--Mathematics. 2. Matrices. 3. Optical instruments--Design and construction. I. Title.
QC355.3.K56 2007
681'.4--dc22
2007025587
Published by
SPIE
P.O. Box 10
Bellingham, Washington 98227-0010 USA
Phone: +1 360 676 3290
Fax: +1 360 647 1445
Email: spie@spie.org
Web: spie.org
Copyright 2007 Society for Photo-optical Instrumentation Engineers

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.
Printed in the United States of America.
Introduction to the Series

Since its conception in 1989, the Tutorial Texts series has grown to more than 70
titles covering many diverse fields of science and engineering. When the series
was started, the goal of the series was to provide a way to make the material
presented in SPIE short courses available to those who could not attend, and to
provide a reference text for those who could. Many of the texts in this series are
generated from notes that were presented during these short courses. But as
stand-alone documents, short course notes do not generally serve the student or
reader well. Short course notes typically are developed on the assumption that
supporting material will be presented verbally to complement the notes, which
are generally written in summary form to highlight key technical topics and
therefore are not intended as stand-alone documents. Additionally, the figures,
tables, and other graphically formatted information accompanying the notes
require the further explanation given during the instructors lecture. Thus, by
adding the appropriate detail presented during the lecture, the course material can
be read and used independently in a tutorial fashion.
What separates the books in this series from other technical monographs and
textbooks is the way in which the material is presented. To keep in line with the
tutorial nature of the series, many of the topics presented in these texts are
followed by detailed examples that further explain the concepts presented. Many
pictures and illustrations are included with each text and, where appropriate,
tabular reference data are also included.
The topics within the series have grown from the initial areas of geometrical
optics, optical detectors, and image processing to include the emerging fields of
nanotechnology, biomedical optics, and micromachining. When a proposal for a
text is received, each proposal is evaluated to determine the relevance of the
proposed topic. This initial reviewing process has been very helpful to authors in
identifying, early in the writing process, the need for additional material or other
changes in approach that would serve to strengthen the text. Once a manuscript is
completed, it is peer reviewed to ensure that chapters communicate accurately the
essential ingredients of the processes and technologies under discussion.
It is my goal to maintain the style and quality of books in the series, and to
further expand the topic areas to include new emerging fields as they become of
interest to our reading audience.
Arthur R. Weeks, Jr.

University of Central Florida
Contents
Preface
xi
1 An Introduction to Tools and Concepts

1.1 Matrix Method
1.2 Basic Elements
1.2.1 Propagation in a homogeneous medium
1.2.2 Refraction at the boundary of two media
1.2.3 Reflection at a surface
1.3 Comparison of Matrix Representations Used in the Literature
1.4 Building up a Lens
1.5 Cardinal Elements
1.6 Using Matrices for Optical-Layout Purposes
1.7 Lens Doublet
1.8 Decomposition of Matrices and System Synthesis
1.9 Central Theorem of First-Order Ray Tracing
1.10 Aperture Stop and Field Stop
1.11 Lagrange Invariant
1.11.1 Derivation using the matrix method
1.11.2 Application to optical design
1.12 Petzval Radius
1.13 Delano Diagram
1.14 Phase Space
1.15 An Alternative Paraxial Calculation Method
1.16 Gaussian Brackets
1
1
2
2
3
5
5
6
7
10
12
13
14
16
18
18
18
19
19
20
21
22
2 Optical Components
2.1 Components Based on Reflection
2.1.1 Plane mirror
2.1.2 Retroreflector
2.1.3 Phase-conjugate mirror
2.1.4 Cats-eye retroreflector
2.1.5 Roof mirror
2.2 Components Based on Refraction
2.2.1 Plane-parallel plate
2.2.2 Prisms
25
25
25
26
26
27
28
29
29
31
viii
2.2.3 Axicon devices

2.3 Components Based on Reflection and Refraction
2.3.1 Integrating rod
2.3.2 Triple mirror
47
49
49
52
3 Sensitivities and Tolerances

3.1 Cascading Misaligned Systems
3.2 Axial Misalignment
3.3 Beam Pointing Error
53
55
56
57
4 Anamorphic Optics
4.1 Two Alternative Matrix Representations
4.2 Orthogonal and Nonorthogonal Anamorphic Descriptions
4.3 Cascading
4.4 Rotation of an Anamorphic Component with Respect to the
Optical Axis
4.4.1 Rotation of an orthogonal system
4.4.2 Rotation of a nonorthogonal system
4.5 Examples
4.5.1 Rotated anamorphic thin lens
4.5.2 Rotated thin cylindrical lens
4.5.3 Cascading two rotated thin cylindrical lenses
4.5.4 Cascading two rotated thin anamorphic lenses
4.5.5 Quadrupole lens
4.5.6 Telescope built by cylindrical lenses
4.5.7 Anamorphic collimation lens
4.6 Imaging Condition
4.7 Incorporating Sensitivities and Tolerances in the Analysis
59
59
60
60
5 Optical Systems
5.1 Single-Pass Optics
5.1.1 Triplet synthesis
5.1.2 Fourier transform objectives and 4f arrangements
5.1.3 Telecentric lenses
5.1.4 Concatenated matrices for systems of n lenses
5.1.5 Dyson optics
5.1.6 Variable single-pass optics
5.2 Double-Pass Optics
5.2.1 Autocollimator
5.3 Multiple-Pass Optics
5.4 Systems with a Divided Optical Path
5.4.1 Fizeau interferometer
5.4.2 Michelson interferometer
61
62
65
65
65
66
67
68
69
72
72
73
75
77
77
77
79
80
81
82
84
92
92
95
99
99
101
ix
5.4.3 Dyson interferometer

5.5 Nested Ray Tracing
103
107
6 Outlook
111
Bibliography
113
Index
119
Preface
This book is intended to familiarize the reader with the method of Gaussian matrices and some related tools of optical design. The matrix method provides a means
to study an optical system in the paraxial approximation.
In optical design, the method is used to find a solution to a given optical task,
which can then be refined by optical-design software or analytical methods of aberration balancing. In some cases, the method can be helpful to demonstrate that
there is no solution possible under the given boundary conditions. Quite often it is
of practical importance and theoretical interest to get an overview on the solution
space of a problem. The paraxial approach might then serve as a guideline during
optimization in a similar way as a map does in an unknown landscape.
Once a solution has been found, it can be analyzed under different points of
view using the matrix method. This approach gives insight on how degrees of
freedom couple in an optical device. The analysis of sensitivities and tolerances is
common practice in optical engineering, because it serves to make optical devices
or instruments more robust. The matrix method allows one to do this analysis in a
first order of approximation. With these results, it is then possible to plan and to
interpret refined numerical simulations.
In many cases, the matrix description gives useful classification schemes of
optical phenomena or instruments. This can provide insight and might in addition
be considered as a mnemonic aid.
An aspect that should not be underestimated is that the matrix description represents a useful means of communicating among people designing optical instruments, because it gives a kind of shorthand description of main features of an optical instrument.
The book contains an introductory first chapter and four more specialized chapters that are based on this first chapter. Sections 1.11.14 are intended to provide a
self-contained introduction into the method of Gaussian transfer matrices in paraxial optics. The remaining sections of the chapter contain additional material on how
this approach compares to other paraxial methods.
The emphasis of Chapters 3 and 4 is on refining and expanding the method of
analysis to additional degrees of freedom and to optical systems of lower symmetry.
The last part of Chapter 4 can be skipped at first reading.
To my knowledge, the text contains new results such as theorems on the design
of variable optics, on integrating rods, on the optical layout of prism devices, etc.
xii
Preface
I tried to derive the results in a step-by-step way so that the reader might apply
the methods presented here to her/his design problems with ease. I also tried to
organize the book in a way that might facilitate looking up results and the ways of
how to obtain them.
It would be a pleasure for me if the reader might find some of the material
presented in this text useful for her/his own engineering work.
Gerhard Kloos
June 2007
Chapter 1
An Introduction to Tools and

Concepts
1.1 Matrix Method
Ray-transfer matrices is one of the possibile methods to describe optical systems
in the paraxial approximation. It is widely used for first-order layout and for the
purpose of analyzing optical systems (Gerrard and Burch, 1975). The reason why
the paraxial approximation is often used in the first phase of a design or of an optical
analysis becomes obvious if we have a look at the law of refraction in vectorial form
as follows:
n1 a N = n2 b N ,
(1.1)
This
where a is the vector of the ray incident on the interface with the normal N.
interface separates two homogeneous media with indices of refraction n1 and n2 .
For optical-layout purposes, we need
The refracted ray is described by the vector b.
an explicit expression of this ray in terms of the other quantities because we have
to trace the ray through the optical system. Using vector algebra, Eq. (1.1) can be
rewritten in the following way:

2

n
n
n
1
1
1
1 (N a )2 N .
b = a
(1.2)
N a 1
n2
n2
n2
The form obtained like this is complicated and it is difficult to trace the ray without
making use of a computer. Therefore, a linearized form of this law would be helpful
for thinking about the optical system, and this is the motivation for starting with a
paraxial layout.
It would be a precious tool for analyzing optical instruments if the approximated description would also allow for cascading subsystems to describe a compound system. The method of ray-transfer matrices provides this advantage and
cascading of subsystems is performed by matrix multiplication.
Another aspect, which might be sometimes underestimated, is that paraxial
descriptions, and especially the matrix method, provide a convenient shorthand
notation to communicate and discuss ideas to other optical designers. In a way, this
Chapter 1
branch of optics is axiomatic like thermodynamics, for example. The framework

of the underlying theory can be reduced to a limited number of basic rules and
elements. But combining these rules and elements allows one to study a great
variety of optical systems.
1.2 Basic Elements

We will now look for linearized relations that describe three situations, namely,
propagation of a ray, and its refraction and reflection. The matrices obtained in this
way serve as building blocks of the matrix description.
1.2.1 Propagation in a homogeneous medium
Let us first consider the propagation of a ray in a homogeneous medium. We assume that the ray propagates in the yz plane and choose the z-axis as the optical
axis. In any plane perpendicular to the optical axis, the ray can now be described
by its distance from the optical axis, y, and by the angle , which it has with a
line parallel to the optical axis. As the ray propagates along the optical axis, these
coordinates may change and take different values in different planes perpendicular
to the optical axis. We now choose two reference planes separated by a distance t
inside a homogeneous medium (Fig. 1.1) and determine the inputoutput relationship. The ray starts with the coordinates [y (1) , (1) ]. Due to the propagation along
a rectilinear line, the angle remains unchanged,
(2) = (1) .
(1.3)
The height in the second reference plane depends on the distance traveled and on
the starting angle,
y (2) = y (1) + t tan (1) .
(1.4)
Figure 1.1 Propagation in a homogeneous medium. The two reference planes are at a
distance t.
An Introduction to Tools and Concepts
Under the assumption that the paraxial approximation is valid, i.e., for small angles (1) , we can linearize the trigonometric function in Eq. (1.4) as
y (2) = y (1) + t (1) .
(1.5)
Equations (1.3) and (1.5) can now be combined and written as a matrix relation,
(1)
(2)

1 t
y
y
=
.
(1.6)
(2)
0 1
(1)
The matrix depends on the distance of the two reference planes. We will later refer
to it as the translation matrix T , defined as

1 t
.
(1.7)
T =
0 1
1.2.2 Refraction at the boundary of two media
Now, we will try to obtain a linearized expression for the refraction of a ray at a
spherical surface described by the radius R. This surface separates two homoge-
Figure 1.2 Refraction at a spherical surface. The spherical surface separates two media
with refractive indices n1 and n2 .
Chapter 1
neous media of refractive indices n1 and n2 . Let us first draw a line representing
the ray as it hits the spherical surface in a reference plane (Fig. 1.2). We consider
how the input and output variables are changed in this single reference plane where
the refraction takes place. The distance from the optical axis remains unchanged
for the ray leaving the reference plane, i.e.,
y (2) = y (1) .
(1.8)
The change in angle is described by the law of refraction,

n1 sin i (1) = n2 sin i (2) ,
(1)
(1.9)
(2)
where the angles i and i refer to the normal vector that is perpendicular to the
surface.
Assuming that the paraxial approximation is valid, Eq. (1.9) can be linearized
as
n1 i (1) = n2 i (2) .
(1.10)
But we need expressions in terms of the angles (1) and (2) that are measured
with respect to a line parallel to the optical axis. To obtain relations between these
angles and the angles appearing in Eq. (1.9), we have a closer look at the triangles
in Fig. 1.2. Applying the exterior angle theorem for triangles twice, we have
i (1) = (1) + ,
i
(2)
(2)
+ .
(1.11)
(1.12)
Substituting these equations into Eq. (1.10), we find

n1
n1 n2
(2) = (1) +
.
(1.13)
n2
n2
Neglecting the small distance between the intersection of the spherical surface with
the optical axis and the reference plane, we approximate the angle appearing in
Eq. (1.13) as
y (1)
tan
.
(1.14)
=
R
Linearizing the trigonometric function for small angles (tan
= ), Eq. (1.14) is
substituted into Eq. (1.13) and we have
n1
n1 n2 (1)
(2) = (1) +
(1.15)
y .
n2
n2 R
This is the linearized inputoutput relation we were looking for. In combination
with Eq. (1.8), we can write it in matrix form as
(2)

(1)
y
1
0
y
=
.
(1.16)
n1
n1 n2
(2)
(1)
n2 R
n2
The corresponding matrix will be used later as the refraction matrix R, defined as

1
0
R = n1 n2 n1 .
(1.17)
n2 R
n2
Figure 1.3 The unfolding of a spherical mirror.
1.2.3 Reflection at a surface

A geometrical consideration quite similar to the one that led to Eq. (1.17) can also
be used to find the matrix for a spherical concave mirror. In this case, the output
ray remains on the same side of the reference plane.
It is interesting to note that we can formally obtain the matrix of an unfolded
spherical concave mirror by setting n1 = 1 and n2 = 1 in Eq. (1.17), i.e.,

1
0
.
(1.18)
S=
R2 1
Unfolding refers to the symmetry operation (or coordinate break) depicted in
Fig. 1.3. This can be helpful in finding the matrix chain of a compound optical
system. Please note that some signs might change in the system matrix with respect to the starting system because reference is made to an optical axis with a
different direction after the coordinate break.
1.3 Comparison of Matrix Representations Used in the

Literature
In the literature, different notations used to write the ray-transfer matrices can be
found. Many authors use coordinates that have n as the second coordinate, where
n is the index of refraction (Guillemin and Sternberg, 1984). An advantage of
this notation is that the determinant value of the ray-transfer matrices is always 1.
This provides a useful check during calculations and can also simplify theoretical
arguments based on the determinant. In the description used here, the determinant
of the ray-transfer matrix A has the value
det A =
n1
,
n2
(1.19)
with n1 as the refractive index of the medium at the entrance reference plane and
n2 as the refractive index of the medium at the exit reference plane.
Chapter 1
The second coordinate n can also be introduced as a modified ray slope (Siegman, 1986) as
dr(z)
r (z)
.
(1.20)
= n(z)
dz
The interpretation of this coordinate in terms of slope can be fruitful in some circumstances.
1.4 Building up a Lens

With the prerequisite of Eqs. (1.7) and (1.17), we can determine the matrix of a
spherical lens. The refraction at the first surface is expressed by the matrix R (a) .
The ray is then propagated through the lens using the translation matrix T and
finally refracted at the second surface of the lens. To describe this refraction, the
matrix R (b) is used. The combined effect is calculated as the product of these
matrices,
S = R (b) T R (a) .
(1.21)
More explicitly, this equation reads as

1
1
0
S = n2 n3 n2
0
n3 R2
n3
n1 n2
n2 R1
n1
n2
(1.22)
where t is the thickness of the lens and R1 and R2 are the radii of the first and
the second surfaces of the lens, respectively. Because the lens is in air, we can
specialize the set of refractive indices as n1 = 1, n2 = n, and n3 = 1. Therefore,
we have

t
t
1 n1
R1 n
n
S=
.
(1.23)
1n t
t
n1
1n
+ n1
1 1n
R1
R2
R1 R2 n
R2 n
This might suggest the following abbreviations:
n1
,
R1
1n
.
P2 =
R2
P1 =
With these abbreviations, Eq. (1.23) then takes the form

t
1 P1 nt
n
.
S=
P1 P2 + P1 P2 nt 1 P2 nt
(1.24)
(1.25)
(1.26)
The so-called thin lens is obtained by letting the lens thickness t tend to zero in
Eq. (1.26),

1
0
S=
.
(1.27)
P1 P2 1
1.5 Cardinal Elements

To identify the lower-left entry in the matrix of the thin lens, we first look at a lens
described by a more general matrix of the form

a11 a12
A=
.
(1.28)
a21 a22
Its focal plane can be found by letting a ray parallel to the optical axis pass through
the lens and determine the distance b from the exit reference plane to the plane
where it intersects the optical axis. Expressing this in matrix notation, we have
in

a11 a12
y
0
1 b
=
,
(1.29)
0 1
a21 a22
0
out
or
0
out
a11 + ba21
a12 + ba22
a21
a22
y in
0
(1.30)
This implies that

0 = (a11 + ba21 )y in .
(1.31)
This equation should hold for all values of y in . Therefore, it follows that
a11 + ba21 = 0.
(1.32)
The position of the second focal plane of the lens described by the matrix A is
therefore determined by
a11
b= ,
(1.33)
a21
and we can identify b as the focal length f of the lens.
Applying this result to the thin lens of Eq. (1.27), for which a11 = 1 holds, we
see that the lower-left entry represents the negative inverse of its focal length, i.e.,
the matrix of the thin lens is

1 0
F =
.
(1.34)
f1 1
The second focal plane is one of the cardinal elements of a lens. The position of the
first focal plane is calculated on the same footing, but by letting a parallel ray enter
from the other side into the system or by finding the distance for which the light
from a point source in front of the lens is collimated. In both ways, the following
result is obtained for the position of the first focal plane:
a=
a22
.
a21
(1.35)
A straightforward way to obtain other cardinal elements is by direct comparison

with the thin lens. We are interested in finding the positions of the planes with
Chapter 1
respect to which a lens given by the matrix A could be described similar to a thin
lens. To this end, we take the following approach:

a11 a12
1 0
1 h2
1 h1
=
,
(1.36)
f1 1
0 1
a21 a22
0 1
where h1 and h2 are the distances that have to be determined. The corresponding
planes are called principal planes and, together with the focal planes, they are the
cardinal elements of a lens. After performing the matrix multiplication on the lefthand side, we have

a11 + a21 h1 a12 + a11 h1 + a22 h2 + a21 h1 h2

1 0
=
.
(1.37)
f1 1
a21
a22 + a12 h2
The position of the first principal plane is therefore given by
h1 =
1 a11
,
a21
(1.38)
measured with respect to the first reference plane of the lens. The position of the
second principal plane is at
1 a22
h2 =
,
(1.39)
a21
measured with respect to the second reference plane of the lens.
A beautiful illustration of the principal planes concept is given by Lipson et
al. (1997). We can trace typical rays through the lens and draw this on a piece
of paper. If we now fold this paper along the lines that represent the principal
planes, we can hold it in such a way that the part between the principal planes
is perpendicular to the other parts. These other parts are combined to represent a
simplified arrangement (Fig. 1.4), which corresponds to a thin lens.
This is in complete analogy to Eq. (1.36). The results on the cardinal elements
are collected in Fig. 1.5.
With these prerequisites, we can state the cardinal elements of the thick lens
given by Eq. (1.23). The equation for the focal length f of the lens is
1
n 1 1 n (n 1)(1 n)t
+
.
=
f
R1
R2
nR1 R2
(1.40)
Its principal planes are at

h1 = f
n1 t
,
R1 n
(1.41)
h2 = f
1n t
.
R2 n
(1.42)
Figure 1.4 Principal planes visualized by folding. The optical system is described by the
matrix A. It has the focal points F1 and F2 and its principal planes are at h1 and h2 , respectively.
Figure 1.5 Cardinal elements. The focal points F1 and F2 and the positions h1 and h2 of
the principal planes serve to characterize the optical system given by the matrix A.
Chapter 1
10
1.6 Using Matrices for Optical-Layout Purposes

In the derivation of the position of the second focal plane, we considered the optical
arrangement formed by a lens, which was given by the matrix A, and a translation
matrix T , i.e.,
S = T A.
(1.43)
On this combined optical arrangement, the condition s12 = 0 was then imposed to
ensure that the ray height in the output plane was independent of the ray angle in
the input plane. We used this condition because in the paraxial approximation, it
characterizes a point in the second focal plane. This way of reasoning can also be
applied to other situations.
Its application to the first focal plane is convenient; to this end, we consider the
combined arrangement given by the matrix product,
S = AT .
(1.44)
We then impose a condition on the combined matrix S that expresses (in the linear
approximation) that a bundle of rays at a given ray height y in but with different
angles in in the entrance plane of S will be transformed into a parallel beam, i.e.,
a bundle of rays with the same angle, at the exit plane. The general inputoutput
relation is
out

in
y
s11 s12
y
=
.
(1.45)
out
s21 s22
in
To ensure that out has a single value for a given ray height y in , it has to be independent of in . A look at the inputoutput relation suggests that this condition is
met if we choose
s22 = 0.
(1.46)
This choice determines the distance contained in the translation matrix and thereby
the position of the first focal plane, which corresponds to the matrix A.
At this point, we have conditions for the first focal plane (s22 = 0) and for the
second focal plane (s11 = 0), and we might ask: what is the characteristic feature
of a ray-transfer matrix S that describes imaging? The rays leaving at a point at y in
in the object plane with different angles in intersect in a point at y out in the image
plane. If the matrix A describes a lens, we have to add two spacings on both sides
to model imaging, so we have
S = BAG,
(1.47)
1 g
1 b
with B = 0 1 and G = 0 1 . Considering the inputoutput relation again, we
find that y out is independent of in if
s12 = 0.
This is the characteristic feature of a matrix S that represents imaging.
(1.48)
11
We can apply this condition immediately to find the imaging relation for a thin
lens. The corresponding matrix chain is

1 fb g + b bg
1 0
1 b
1 g
f
S=
.
(1.49)
=
0 1
0 1
f1 1
f1
1 fg
Using s12 = 0, we have the well-known imaging condition
1 1
1
+ = ,
g b
f
(1.50)
which expresses that b varies in a hyperbolic way as a function of g and vice versa.
The signs of the distances are positive in Eq. (1.50) because the direction of the
distances is chosen as the direction of the optical axis.
To find a relation for the first focal plane, we asked under which conditions parallel rays leaving the system might be independent of the input angle. Alternatively,
we can consider the situation where the rays leaving the system are independent of
the ray height in the entrance plane. This is the case if a collimated input beam is
transformed into a collimated output beam. Making reference to the inputoutput
relation for S, we see that setting
s21 = 0
(1.51)
ensures that out does not depend on the ray height y in in the input reference plane.
Because collimated rays are considered, no additional translation matrices have to
be introduced here and therefore S = A. Earlier, we related the matrix entry a12
to the negative inverse of the focal length of an optical system via Eq. (1.33). This
matrix entry takes the value of zero now, which corresponds to the case of an afocal
system.
Typical examples for such systems are telescopes. In the paraxial approximation, we might model a telescopic arrangement using thin lenses. We choose two
lenses with focal lengths f1 and f2 , separated by a distance d. Concatenation of the
corresponding matrices gives us the system matrix

d
1 fd1
1
0
1
0
1 d
S=
. (1.52)
=
0 1
f12 1
f11 1
f11 f12 + f1df2 1 fd2
Now, we impose the condition that s21 = 0 should hold. This implies that
1
1
d
+
= 0.
f1 f2 f1 f2
The setting of d = f1 + f2 solves this equation and we have
f2
f1 f1 + f2
S=
f1
0
f2
(1.53)
(1.54)
Chapter 1
12
Figure 1.6 The significance of zero-matrix entries.
for the system matrix of the telescopic arrangement. It represents a Newtonian telescope if both focal lengths are positive. If the focal length of the first lens is negative, the matrix describes a Galilean telescope, which is composed of a concave and
a convex lens. Optical arrangements of this type also serve as transmissive beam
expanders (Das, 1991) and intracavity telescopes (Siegman, 1986). The results on
the significance of special matrix entries are summarized in Fig. 1.6.
1.7 Lens Doublet

We encountered telescopic arrangements as the first examples of a lens doublet and
we now have a closer look at optical systems composed of two lenses. The matrix
that describes two lenses separated by a distance d forms the starting point of our
discussion:

1 fd1
d
S=
.
(1.55)
f11 f12 + f1df2 1 fd2
13
The term s21 is related to the focal length of the doublet (measured with respect to
its second principal plane).
1
1
d
1
+
.
=
f
f1 f2 f1 f2
(1.56)
[As shown before, this principal plane is at a distance z = (1 s11 )/(s21 ) from the
second reference plane of the system.] To facilitate the discussion, it is convenient
to reference the intermediate distance to the second focal plane of the first lens and
to the first focal plane of the second lens by setting
d = f1 + E + f2 .
(1.57)
With this setting, the equation for the focal length of the doublet reduces to
f =
f1 f 2
.
E
(1.58)
It can now be discussed in terms of the signs of the three parameters that intervene.
Depending on whether f1 < 0 or f1 > 0, f2 < 0 or f2 > 0, or E < 0, E = 0, or
E > 0, twelve cases can be distinguished. The case where f1 > 0 and f2 > 0 and
E = 0, for example, represents the Galilean telescope.
At this point, it is near at hand to make a distinction between divergent (f < 0)
and convergent (f > 0) doublets in terms of their three parameters. A compound
microscope represented as a doublet is characterized by f1 > 0 and f2 > 0 and
E > 0, and it is interesting to note that it is an example of a divergent system
(Prez, 1996)
1.8 Decomposition of Matrices and System Synthesis

In the layout of a new optical system, it is advantageous to know how the raytransfer matrix of a given optical system can be factorized. Let us consider the
design of an optical device with given properties and that some of these features
can be expressed in terms of a system matrix. To realize the device, it is now of
interest to systematically explore in which ways a device with the given features
can be realized. To this end, it is useful to divide the device into subsystems,
the combination of which would create the desired functionality. In the matrix
description, this is equivalent to considering matrix products of the target matrix,
and this is where factorizing the system matrix comes into play. The problem of
a synthesis of optical systems using this approach has been studied in depth by
Casperson (1981).
In what follows, we will consider optical systems that have both their object
and image planes in air. Therefore, n1 = 1 and n2 = 1 and the determinant of the
system matrix S can be written as
det S =
n1
= 1.
n2
(1.59)
Chapter 1
14
Therefore, the condition

s11 s22 s12 s21 = 1
(1.60)
is contained implicitly in Eqs. (1.61) and (1.62). A generalization is possible and

can, for example, be found in the work of Casperson (1981).
The appropriate factorization depends on the matrix entries. If we consider a
nonimaging problem, we can assume s12 = 0 for the system matrix. Such a matrix
can be factorized as

1
0
1 s12
1
0
s11 s12
= s22 1
.
(1.61)
S=
s11 1
s21 s22
1
0 1
1
s12
s12
If the lower-left entry of the system matrix can be assumed to be nonzero (s21 = 0),
i.e., if we do not look for an afocal system, the following matrix decomposition
is appropriate:

s11 s12
1 s11s211
1 0
1 s22s211
S=
=
.
(1.62)
s21 s22
0
1
s21 1
0
1
What is left are the cases in which both s12 = 0 and s21 = 0. These cases correspond to optical systems that are imaging and afocal devices. In the above-cited
work, four possibilities for a decomposition of this diagonal matrix are given. The
system matrix is either decomposed in a product of matrices A and B with a21 = 0
and b21 = 0 as

s11 ts22
s11 0
1 t
=
,
(1.63)
S=
0 1
0
s
0 s22

22
s11 0
s11 ts11
1 t
S=
,
(1.64)
=
0 1
0
s22
0 s22
or a product of matrices with a12 = 0 and b12 = 0 as

1 0
s11 0
s11 0
=
,
S=
s11
0 s22
f1 1
s22
f

s11 0
s11 0
1 0
S=
= s22
.
0 s22
f1 1
s22
f
(1.65)
(1.66)
Depending on the application, the matrices appearing in the product can then be
further decomposed by applying the same set of rules.
1.9 Central Theorem of First-Order Ray Tracing

We will now turn to a theorem that is of prime importance to ray tracing using
the matrix method. It can be applied to different sets of rays. Its main content is
that the number of rays necessary to characterize an optical system in the linear
approximation is rather small.
15
Let us consider two rays labeled a and b that are traced through the optical
system described by the matrix A. Each ray vector entering the system is mapped
onto an output ray vector as follows:
out
ya
ya

,
(1.67)
a
aout

out
yb
yb

.
(1.68)
b
bout
The mapping is given by the system matrix A. Therefore, we have

out
ya
ya
=A
,
out
a
a
out
yb
yb
=A
.
out
b
b
(1.69)
(1.70)
We assume that we can completely determine the four ray coordinates and that
we want to use this information to determine the system matrix A. Its entries are
therefore the unknown variables of the problem, and we can state it by rewriting
the above equations as the following system of linear equations:
out
ya a 0 0
a11
ya
y
0 0 a12 ybout
b
b
(1.71)
=
.
0 0 ya a a21 aout
0
yb
bout
a22
Because the matrix is partitioned, two sets of linear equations can be solved independently. If the determinant

ya a
= 0,
(1.72)
D = det
yb b
the problem has a unique solution, namely,
a11 =
a12 =
a21 =
a12 =
det
a
ybout b
yaout
D
det
yaout
yb ybout
(1.73)
(1.74)
(1.75)
(1.76)
ya
a
bout b
det
aout
det
D
ya aout
yb bout
Chapter 1
16
D = 0 is equivalent to the condition that the input ray vectors are linearly independent. We can therefore conclude that the ray-transfer matrix is completely
determined if we know a set of two linearly independent input ray vectors and the
corresponding output ray vectors. In the linear approximation, the passage of any
other third ray through the system is then known because we can trace it through
the system using the matrix A. Putting it in different words, the theorem states that
in the approximation used, the inputoutput relation is completely characterized
once the input and output data of two linearly independent rays are known.
This gives the theoretical basis of why an optical system can be characterized
to such an extent by just tracing the principal ray and the axial ray.
1.10 Aperture Stop and Field Stop

The aperture stop is defined as the opening of an optical system that limits the input
angle at zero height in the object plane. A ray with these coordinates can be transported through the system. If the input angle of a ray is slightly greater than this
critical angle, the ray is blocked. We might have several candidates in the system to
cause this blockage, and which of them forms the aperture stop can be determined
in the following way using the matrix method. We label the free diameters of the

candidates as y (k) . To every candidate now corresponds a matrix P (k) that maps
the start ray into the reference plane at z(k) ,

(k)
0
y
(k)
.
(1.77)
= P
(k)
(k)
This implies that
(k)
y (k)
(k)
p12
(1.78)
The aperture stop is at the position z(k) for which (k) takes the minimum value of

= y (k) if (k) is the label for that minimum.
all the candidates. It has the height yas
The axial ray is the ray that starts at zero height in the object plane and that
passes through the aperture stop at the maximum possible height. If we suppose
that the matrix P describes the mapping of the ray from the object plane to the
aperture plane, we can trace this ray to that plane using

0
y
P
= as .
(1.79)
Its start coordinates in the object plane are

in

y
0
= yas ,
in
p12
and this ray can now be traced through the complete optical system. We describe
the second part of the system, i.e., the part between the aperture plane and the image
17
plane, by Q. Therefore, the system matrix is

S = QP ,
and the coordinates of the axial ray as its leaves the system are
out
s
y
yas
12
=
.
out
p12 s22
(1.80)
(1.81)
While the axial ray starts at zero height in the object plane and passes through the
aperture stop at its margin, the principal ray starts at the marginal height of the
object (if this corresponds to the field stop) and passes through the aperture stop
at zero height. In the matrix description, we can express this relation by using the
matrix P , which describes the mapping from the object plane to the plane of the
aperture stop, as
field

y
0
P
.
(1.82)
=

To be able to trace the principal ray through a complete system, we need its input
angle, which we can calculate from the following equation:
=
p11 field
y .
p12
Therefore, the input coordinates of the principal ray are given by
in

y field
y
=
,
in
pp11
y field
12
and the output coordinates after passage through the whole system are
out

y
1
field
=y S
.
out
pp11
12
(1.83)
(1.84)
(1.85)
It is interesting to note the following symmetry that exists between the axial ray and
the principal ray:

y
0
= as
P
for the axial ray,
(1.86)
field
y
0
= P 1
for the principal ray,
(1.87)
where the corresponding inverse matrix has been used. Das (1991) expressed this
symmetry relation by writing . . . the field stop is nothing but the new aperture
stop, when the object is placed at the center of the actual aperture stop.
(Please note that in writing the symmetry relation it was assumed that the extension of the object can be identified with the extension of the field stop. This is
quite often the case, but more intricate situations are possible.)
Chapter 1
18
1.11 Lagrange Invariant

1.11.1 Derivation using the matrix method
We know that
det A =
n1
n2
(1.88)
holds for a ray-transfer matrix. During the derivation of the central theorem of firstorder ray tracing, the following result was obtained for two linearly independent
rays that pass through the system described by this matrix:

a11 a12
1 yaout b ybout a ya ybout yb yaout
A=
.
(1.89)
=
D aout b bout a ya bout yb aout
a21 a22
If we use this result in Eq. (1.88), we have
(yaout b ybout a )(ya bout yb aout ) (aout b bout a )(ya ybout yb yaout )
n1
=
.
(ya b yb a )2
n2
(1.90)
Performing the multiplications, we find
yaout bout aout ybout
n1
= .
ya b a yb
n2
(1.91)
This equation can now be rearranged slightly, to separate input and output quantities, as

n1 (ya b a yb ) = n2 yaout bout aout ybout .
(1.92)
We can therefore conclude that the following quantity is conserved during the passage through the system:
L = n(ya b a yb ).
(1.93)
This is the Lagrange invariant.

1.11.2 Application to optical design
We now turn to an application of the Lagrange invariant that is useful when designing imaging systems. Let us consider the invariance condition given by Eq. (1.92)
and specialize it for the case where ray a is the axial ray and ray b is the principal
ray. From the discussion earlier, we know that these are two linearly independent
rays and are therefore suitable to characterize the system. These rays pass through
an imaging system. As reference planes, the object plane and the image plane are
natural choices, so we have

n(yar pr ar ypr ) = n (yar
pr
ar
ypr
).
(1.94)
19
In both these planes, the height of the axial ray is equal to zero. Therefore, the
above equation reduces to

ypr
).
n(ar ypr ) = n (ar
(1.95)

can be identified with the aperture of the system in the
The quantities ar and ar

are the heights of
object plane and the image plane, respectively, and yar and yar
the object and the image. In this way, the Lagrange invariant allows one to establish
a direct relation between these quantities. This relation is, for example, of use to
study the magnification and angular magnification of an optical instrument.
1.12 Petzval Radius

The Petzval radius is the reciprocal of the curvature of the image field. This is a
concept of aberration theory, and it is surprising that we can determine its value
from paraxial quantities. The quantities to be considered are quite similar to those
that we encountered building up a lens, namely,
P1 =
n1
,
R1
P2 =
1n
.
R2
We state the relation for the Petzval radius Rp of a system made up of N refractive
surfaces, without proof, as
N
1 Dk
,
(1.96)
=
Rp
n
k=1 k
where Dk = nk /Rk is the refractive power of the kth surface, nk is the difference in refractive indices, and Rk is the radius of curvature of the kth surface.
A proof can be found in textbooks on optics (Born, 1933; Born and Wolf, 1980)
The Petzval radius should tend to infinity to have a flat image field. This corresponds to finding a combination for the right-hand side of the equation that brings
its value close to zero. An example for an optical device that minimizes this quantity is Dysons system (Dyson, 1959). This optical arrangement will be considered
in more detail in Section 5.1.
1.13 Delano Diagram

The Delano diagram or yy diagram is a visual tool of paraxial analysis (Delano,
1963; Shack, 1973; Besenmatter, 1980). It is created by tracing a principal ray (y)
and an axial ray (y) through an optical system and by drawing the corresponding
ray heights (y,
y) in a single diagram. The position on the optical axis does not
appear explicitly in this diagram and this representation is therefore somewhat abstract compared to a usual ray trace. But in many cases it can give an overview that
is like a kind of shorthand notation for the system. Figure 1.7 shows four Delano
diagrams that represent the different cases of matrices with zero entries.
Chapter 1
20
Figure 1.7 Delano diagram, four cases.
1.14 Phase Space

Looking at the set of coordinates used in matrix optics, it is near at hand to try a
representation in a plane using the ray height as one coordinate, and the ray angle as
the other coordinate. A set of rays in a given reference plane can then be represented
as a surface in this abstract plane.
To familiarize ourselves with this concept, let us choose a set of rays whose
coordinates are represented by a rectangle in this so-called phase space. We would
like to see what might be the effect of the translation matrix T on this set of rays.

0
0
t0

+
,
(1.97)
0
0
0

y0
y0
t0

+
,
(1.98)
0
0
0

0
0

,
(1.99)
0
0

y0
y0

.
(1.100)
0
0
Graphically, this can be expressed as in Fig. 1.8. The corresponding mapping for
the refraction matrix is shown in Fig. 1.9.
21
Figure 1.8 Effect of the translation matrix in phase space. The spatial coordinate is represented along the y-axis and the angular coordinate is represented along the -axis.
Figure 1.9 Effect of the refraction matrix in phase space. As in Fig. 1.8, the spatial coordinate is represented along the y-axis and the angular coordinate is represented along the
-axis.
It is an important feature of this phase space that the volume (or surface in
the two-dimensional case considered here) is conserved if we consider mappings
between input and output planes where the refractive indices are 1. This conservation is a consequence of Eq. (1.19).
From the theoretical point of view, it is more convenient to use coordinates
(y, n). The corresponding volume (or surface) is then always conserved. The
phase-space approach is common in laser technology (Hodgson and Weber, 1997).
1.15 An Alternative Paraxial Calculation Method

An alternative paraxial method (Berek, 1930) uses distances sk measured along the
optical axis and ray heights hk measured perpendicular to it as a set of coordinates
for the description of a ray. For readers who use this method, it might be interesting
Chapter 1
22
to see how both methods are connected. They will be familiar with the so-called
transition equations. These equations state how a ray is transformed that leaves
a lens surface labeled with k and and that reaches another surface (labeled with
k + 1) situated at a distance ek as follows:
sk+1 = sk ek ,
hk+1
,
sk+1 =
uk+1
h
sk = k ,
uk
(1.101)
(1.102)
(1.103)
where uk+1 and uk are angles. We can combine these equations to obtain
hk
hk+1
=
ek .
uk
uk
(1.104)
Setting uk+1 = uk , we can write the linear relation,

hk+1
uk+1
ek
hk
uk
which is quite similar to the relation for the translation matrix T .

The other method also makes use of the relation

1
1
1
1
nk
= nk

Rk
sk
Rk
sk
(1.105)
(1.106)
for the two sides of a refracting surface with radius Rk . Using uk = hk /sk for the
paraxial angle, we can write this equation as follows:

nk
nk
hk
1 + uk .
(1.107)
uk =
Rk
nk
nk
Assuming hk = hk in addition, we have an augmented linear relation of the form

hk
uk
nk nk
Rk nk
0
nk
nk
hk
uk
(1.108)
This relation corresponds to the refraction matrix R of the matrix description.
1.16 Gaussian Brackets

While cascading matrices in order to determine a system matrix, the recursive character of the problem became clear. But at this point, we were unable to state the
underlying recursion law that would allow us to express the ray that finally leaves
23
the system, and that is described by the product matrix, as a function of the input ray without performing the matrix multiplication. The use of the mathematical
concept of Gaussian brackets makes it possible to state this recursion formula.
An introduction to the algebra of Gaussian brackets and the recursion law for
cascaded linearized optical systems was given by Herzberger (1943). In this text,
the Gaussian brackets are defined by the following recursion formula:
[a1 , . . . , ak ] = a1 [a2 , . . . , ak ] + [a3 , . . . , ak ],
(1.109)
with [ ] = 1. This implies

[a1 ] = a1 ,
(1.110)
[a1 , a2 ] = a1 a2 + 1,
(1.111)
[a1 , a2 , a3 ] = a1 a2 a3 + a1 + a3 ,
(1.112)
[a1 , a2 , a3 , a4 ] = a1 a2 a3 a4 + a1 a2 + a1 a4 + a3 a4 + 1.
(1.113)
From Herzbergers article, we take a description of a lens using his symbols, but
arrange the linear transformation as a matrix as follows:

1
nd12
1 nd12
x2
x1
12
12
=
.
(1.114)

d
d
12
12
2
01
1 + 2 n 1 2 1 n 2
12
12
where x is the distance between intersections of the optical axis and the ray with
the reference plane and is the inclination angle of the ray. The subscripts indicate
the corresponding surfaces. k is the refractive power of the kth surface, dk,k+1 is
the distance between the kth and the (k + 1)st surface, and nk,k+1 is the refractive
index between the kth and the (k + 1)st surface. Using Eqs. (1.110)(1.112), this
can be recast in the following way:
d12

, d12
n12
x2
1
x1
n12
=
.
(1.115)

2
01
, d12 ,
d12 ,
1
n12
n12
For the general case, the recursion law reads as follows (Herzberger, 1943):

1 , d12 , . . . , dk1,k d12 , 2 , . . . , dk1,k
x1
n12
nk1,k
12
nk1,k
x
=
. (1.116)

d12

dk1,k

01
,...,
,
,...,
1
nk1,k
n12
This equation establishes a link between the input coordinates and the output coordinates of an optical system. The optical properties of the system are described by
the matrix entries, which are Gaussian brackets. In order to state these entries explicitly in terms of refractive powers, distances, and refractive indices, the recursive
relations have to be evaluated using Eq. (1.109) in a kind of backtracking procedure. The approach of the matrix method is advantageous, because it allows us to
obtain the system matrix by concatenating matrices, i.e., by matrix multiplication.
24
Chapter 1
In cases, where it is useful to state inputoutput relations in a recursive way, the

method of Gaussian brackets provides an alternative approach.
In the framework of the matrix method presented in this text, I found no way to
state the recursion law with a similar conciseness. Therefore, I would like to draw
the readers attention to the Gaussian-brackets method that allows such a formulation. It is beyond the scope of this book to derive the method here, and interested
readers are referred to Herzbergers work. In optical design, Gaussian brackets
were applied to the layout of zoom systems (Pegis and Peck, 1962; Tanaka, 1979,
1982).
Chapter 2
Optical Components
2.1 Components Based on Reflection
Mirrors are a key component of many optical devices. The matrix for the reflection
at a spherical mirror was considered in the introductory chapter. We will now
discuss reflectors more generally.
2.1.1 Plane mirror
The matrix describing the reflection at a plane mirror can be obtained by taking the
matrix for reflection at a spherical reflector and letting the radius of the spherical
mirror tend to infinity. In this way, the unity matrix is obtained as

A=
1 0
0 1
(2.1)
The signs that appear in this matrix are surprising at first, and it is instructive to
derive the matrix also in an alternative way. In Fig. 2.1, the reflection of a ray at a
plane mirror, which is perpendicular to the optical axis, is depicted. In the matrix
representation used here, we unfold the ray using the reference plane of the mirror
as the plane of the coordinate break. Figure 2.2 shows the result of this unfolding.
It is this coordinate break that causes the positive signs in the matrix of the plane
reflector.
Figure 2.1 Plane mirror.
Chapter 2
26
Figure 2.2 Unfolded plane mirror.
2.1.2 Retroreflector
Unlike a plane mirror, a retroreflector redirects the beam back into the same direction from where it came. In the matrix description, this is expressed by a negative
sign of the matrix entry a22 . Reflection at a retroreflector is combined with a change
in the height of the beam. The height of the incident beam is changed from y in
to y out = y in . This corresponds to a parallel shift and is expressed by a negative
sign of the matrix entry a11 . The complete matrix of the retroreflector reads as

1 0
.
(2.2)
A=
0 1
More generally, a retroreflector has the following matrix:

1 a12
A=
.
0 1
(2.3)
2.1.3 Phase-conjugate mirror

An element that redirects an incident beam into itself without any shift in ray height
would be described by the following matrix (Lam and Brown, 1980):

1 0
.
(2.4)
A=
0 1
There are devices that exploit nonlinear optical effects and that are able to operate in
such a way on an incident laser beam. To describe these so-called phase-conjugate
Optical Components
27
mirrors in the paraxial approximation, the matrix stated above can be used. The
nonlinear effect itself is far beyond the scope of this approach. Retroreflectors, on
the other hand, can be advantageously modeled using the matrix method.
2.1.4 Cats-eye retroreflector
A prominent example of this type of optics is the cats-eye retroreflector. It basically consists of a lens and a plane mirror (Fig. 2.3). The distance between the
lens and the mirror is chosen as the focal length of the lens (with respect to the
backward principal plane of the lens). The lens will be approximated as a thin lens.
To describe the plane mirror, we can use the matrix given earlier [Eq. (2.1)]. The
distances involved are first designated by g and b. Following the ray through the
unfolded arrangement, we find the following matrix chain:

S=

1 0
1 g
1
f1 1
0 1
0

1 0
1 b
f1 1
0 1
b
1
1
0
1 0
0 1
g
.
1
After performing the matrix multiplications, we have

2
1 + 2 fb fg 1 2 fg 2g + 2b 4 bg
+ 2 gf fb 1
f
S=
.

2 b
1
2 fg fb 1 + 1 2 fb
f f
(2.5)
(2.6)
We know that letting b = f makes the arrangement work as a cats-eye retroreflector. It is instructive to see what happens if we choose the entrance and exit planes
of the system either at the position of the lens or at a distance f in front of it. The
first alternative corresponds to setting g = 0 and the second one to letting g = f .
In this way, we find

1 2(f g)
(2.7)
S(g = 0, b = f ) =
0
1
Figure 2.3 Cats-eye arrangement.
Chapter 2
28
and
S(g = f, b = f ) =
1 0
0 1
(2.8)
The form of the second matrix is equal to the matrix of a retroreflector stated in
Eq. (2.2).
2.1.5 Roof mirror
A roof mirror is formed by two plane mirrors meeting at a right angle. This optical
arrangement is also being designated as a double mirror (DeWeerd and Hill, 2004).
We will consider the plane that is perpendicular to the roof edge. In this plane,
the mirror can be unfolded as shown in Fig. 2.4 using an xy plane as the plane
of symmetry. The coordinates of the rays that leave the mirror after reflection can
be found by tracing a rectilinear line through the unfolded arrangement. If we first
ignore the coordinate break caused by the unfolding operation, the figure shows the
passage of a ray through a distance 2t. This propagation can be described by the
following matrix:

1 2t
.
(2.9)
T =
0 1
Additionally, it has to be taken into account that the orientation of the reference
axis changes due to the unfolding operation. The corresponding sign change of the
coordinates in the new coordinate system can be seen in Fig. 2.4 and expressed by
the following matrix:

1 0
.
(2.10)
0 1
Combining, this gives the component matrix,

1
1 2t
1 0
=
S=
0
0 1
0 1
2t
1
(2.11)
Therefore, a roof reflector acts like a retroreflector [Eq. (2.3)] in the plane that is
perpendicular to the roof edge.
Figure 2.4 Unfolding the roof mirror.
Optical Components
29
Figure 2.5 Arrangement with roof mirror and plane mirror.
It is interesting to see how a combination of this roof mirror and a plane mirror
as depicted in Fig. 2.5 would act on a ray. To this end, it is advantageous to unfold
the optical arrangement with respect to the reference plane of the plane mirror. This
leads to the following matrix chain:

1 2t
1 b
1 0
1 b
1 2t
.
(2.12)
S=
0
1
0 1
0 1
0 1
0
1
The evaluation of this matrix product gives the system matrix of the arrangement
of Fig. 2.5 as follows:

1 4t + 2b
S=
.
(2.13)
0
1
This equation has some similarity to the equation of a plane mirror, but is of the
following more general form:

1 a12
A=
.
(2.14)
0 1
2.2 Components Based on Refraction

Lenses are of course very prominent examples of optical components based on
refraction. Because their description using the matrix method are treated in other
chapters, other components based on refraction are considered here.
2.2.1 Plane-parallel plate
The plane-parallel plate is a component that is often encountered in optical setups.
It can be a simple glass plate used for path-length compensation or a component
that is used to influence the polarization and has the form of a plate. Its system
Chapter 2
30
matrix can be concluded as a special case of the system matrix of a thick lens
[Eq. (1.23)] in air by letting both radii tend to infinity. In this way, we find

1 nt
S=
,
(2.15)
0 1
where t is the thickness of the plane-parallel plate and n is its refractive index.
To recall this equation during design work, we may also apply the linearized
law of refraction twice. In matrix notation, this reads as
(1)

in
1 0
y
y
=
(2.16)
1
(1)
in
0 n
for the rays entering the plate and
out

1
y
=
out
0
0
n
y (2)
(2)
(2.17)
for those leaving it. In between, they pass through a homogeneous medium of
refractive index n. Putting this together, we have

1 nt
1 0
1 0
1 t
.
(2.18)
S=
=
0 n
0 1
0 n1
0 1
This matrix can be used to determine the shift z depicted in Fig. 2.6, which occurs
if a plane-parallel plate is introduced into a convergent beam. For comparison, we
first describe the situation without a plane-parallel plate in the following way:
in

in
out,1

1 t
y
y
1 z1
1 t + z1
y
=
=
. (2.19)
out,1
0 1
0 1
in
0
1
in
At the intersection with the optical axis, we have y out,1 = 0. This implies
z1 =
y in
t.
in
Figure 2.6 Shift caused by a plane-parallel plate.
(2.20)
Optical Components
31
The situation with a plane-parallel plate can be described by

out,2

1 nt
1 nt + z2
y in
y in
y
1 z2
=
=
. (2.21)
out,2
0 1
in
in
0 1
0
1
Using y out,2 = 0, we find
z2 =
y in
t
.
in
(2.22)
The difference of Eqs. (2.22) and (2.20) gives the shift z we are looking for in the
paraxial approximation,

1
z = z2 z1 = t
1 .
(2.23)
n
Because the term in brackets is generally negative, the shift z takes a positive
value as is expected from Fig. 2.6.
2.2.2 Prisms
Prisms have a vast range of applications in several fields of optics. In optical spectroscopy, for example, an important branch of science, prisms serve both as dispersive means (Demtrder, 1999) and as components for beam shaping. In optical data
storage technology, prisms are also used for beam shaping purposes, namely, to circularize the asymmetric beam emitted by a semiconductor laser
(Marchant, 1990).
2.2.2.1 Two types of prisms
It is useful to divide prisms into two distinct groups that have different properties
with respect to dispersion. A convenient graphical tool to make this distinction
is the so-called tunnel diagram (Yoder, 1985). To draw this diagram, the prism
has to be optically unfolded as if it has a surface from which the beam of light
is reflected back into the prism. Figures 2.72.10 show two examples that are
representative for the two groups. The tunnel diagram in Fig. 2.8 shows that the
Dove prism (Fig. 2.7) is equivalent to a plane-parallel plate. There is a variety of
prisms that can be understood in terms of this basic component. The second prism
(Fig. 2.9), on the other hand, can be reduced to a prism with an apex angle using
the tunnel diagram (Fig. 2.10).
Figure 2.7 Dove prism.
Chapter 2
32
Figure 2.8 Tunnel diagram of the Dove prism.
Figure 2.9 Prism with one internal reflection.
Figure 2.10 Tunnel diagram of the prism with one internal reflection.
We can conclude from these examples that there are the following:
1. Prisms that are reducible to a plane-parallel plate.
2. Prisms that are not reducible to a plane-parallel plate.
Another way of stating this important difference is talking of nondispersive (Wolfe,
1995) and dispersive prisms (Zissis, 1995). The first case can be treated with the
matrix of a plane-parallel plate, which has already been derived. We will therefore
have a closer look at the other type of prisms.
Thin-prism approximation for dispersive prisms If both the prism angle and the
incidence angle measured against the normal of the first face of the prism are small,
Optical Components
33
Snells law can be linearized and the approximated expression for the deviation
angle is as follows (Heavens and Ditchburn, 1991):
= (n 1).
(2.24)
It is interesting to observe that the angle of incidence does not appear in this approximated expression.
In terms of the matrix method, the thin prism can be written in the following
way:

0
y
y
1 0
.
(2.25)
+
=
(n 1)
0 1

We will apply this later to get an approximated expression for an axicon.
In many cases, the linear approximation is not sufficient and it is necessary to perform the trigonometric calculations
to trace a ray through a prism.
Using beam shaping, especially laser beam circularization, is a representative
example of how prism arrangements can be analyzed with trigonometric transfer
functions. Semiconductor lasers emit with a high beam divergence perpendicular
to the junction and with a low beam divergence parallel to it. In several applications, it is of importance to transform this elliptical Gaussian beam into a circular
Gaussian beam or, more generally speaking, to adapt the elliptical Gaussian beam
to a given optical system by expanding or compressing it along an axis. There are
also applications in which the intensity profile that results in a plane perpendicular
to the optical axis differs from a circular one after the Gaussian beam is shaped.
Only beam shaping of collimated light is considered here, i.e., prisms that can be
used to perform the expansion or compression after the beam emitted by the laser
diode has passed a collimating lens are described.
Of course, prism arrangements are not the only way to realize beam shaping
in an optical system. The reader who would like to know more about alternative
techniques is referred to Dickey and Holswade (2000).
Trigonometric description of dispersive prisms
2.2.2.2 Brewster condition
To minimize losses, the so-called Brewster condition important. Reflective losses

at a surface are minimal if a polarized beam is parallel to the plane of incidence,
i.e., a p-polarized beam, is incident at Brewsters angle (Young, 1997; Marchant,
1990). This angle depends on the refractive index in the following way:
0 = arctan(n).
(2.26)
The Brewster condition is also often expressed by saying that the ray reflected by
the surface and the ray refracted into the medium are perpendicular with respect to
each other, i.e.,
0 = 90 deg 1 ,
(2.27)
Chapter 2
34
where 1 is the angle that the refracted ray forms with the surface normal pointing
into the medium. This implies that the following equation is an alternative expression of the Brewster condition in terms of the angle 1 :
1 = arccot(n).
(2.28)
2.2.2.3 Refracting prism
To derive a relation between the angle of incidence 0 and the exit angle 3 of the refracting prism
characterized by its apex angle and the refractive index n, the law of refraction
has to be applied twice. This leads to
Relation of incident and exit angle in the case of a single prism
3 = arcsin(n sin 2 ),

1
1 = arcsin
sin 0 ,
n
(2.29)
(2.30)
where 1 and 2 are the corresponding angles inside of the prism as depicted in
Fig. 2.11. In a prism, these angles are related by
1 + 2 = .
If these equations are put together, the following relation is obtained:

1
.
3 = arcsin n sin arcsin
sin 0
n
(2.31)
(2.32)
In an analogous way, the corresponding equation for the inverse relation can be
derived as

1
.
(2.33)
sin 3
n
Figure 2.11 Refracting prism.
Optical Components
35
From Eq. (2.32), a relation can be derived that links the incidence angle and
the apex angle for the case where the beam exiting the prism is perpendicular to
the second surface of the prism, i.e., 3 = 0. Using the trigonometric identity
arcsin x = 90 deg arccos x, Eq. (2.32) can be reformulated as follows:

3 = arcsin n sin( 90 deg + 0 ) .
(2.34)
For 3 = 0, this equation implies that
90 deg + 0 = 0.
(2.35)
This relation is used later.

The transfer function for the beamwidth altered by a single prism To study the
beam-shaping effect of a prism on a collimated beam, it is appropriate to consider
first how a single refracting surface changes an incident beam of width w0 . The
angle of incidence of this beam, with respect to the surface normal is, designated
by 0 . The angle 1 , with respect to this normal inside a medium of refractive index
n is, determined by Snells law as

1
1 = arcsin
(2.36)
sin 0 .
n
If 0 is not zero, the beamwidth is increased after transition from a medium of lower
refractive index to a medium of higher refractive index. Figure 2.12 shows a cut
through a surface that forms the boundary between air and the prism material. The
beamwidth after refraction is called w1 . From the figure, the following relations for
the cosines of the two angles are obvious:
w0
cos 0 =
,
(2.37)
h
w1
.
(2.38)
cos 1 =
h
Figure 2.12 Change of beamwidth at a refracting surface.
Chapter 2
36
If they are combined, a relation follows that is an expression for the beamwidths as
a function of incident and exiting angle, i.e.,
w1
cos 1
=
.
w0
cos 0
(2.39)
Together with Snells law [Eq. (2.36)], this leads to the beamwidth transfer function
for a refracting surface,
w1
cos{arcsin[(1/n) sin 0 ]}
=
.
w0
cos 0
(2.40)
The corresponding equation for the other refracting surface of the prism, i.e., the
transition from a medium with refractive index n to air, is
w3
cos[arcsin(n sin 2 )]
=
.
w1
cos 2
(2.41)
In a prism of apex angle , the link between the known angle 1 and the angle 2
is given by = 1 + 2 .
In this way, one finds
cos[arcsin(n sin{ arcsin[(1/n) sin 0 ]})]
w3
=
.
w1
cos{ arcsin[(1/n) sin 0 ]}
(2.42)
Substituting for w1 from Eq. (2.40), the beamwidth transfer function for a prism
with the apex angle is obtained as
cos[arcsin(n sin{ arcsin[(1/n) sin 0 ]})] cos{arcsin[(1/n) sin 0 ]}
w3
=
.
w0
cos{ arcsin[(1/n) sin 0 ]}
cos 0
(2.43)
This equation shows that the transfer functions of the refracting surfaces can be
multiplied to obtain the transfer function of the prism. This is a general property of
the beamwidth transfer functions.
It is of interest to consider the situation
where the light beam is incident on the prism in accordance with the Brewster
condition [Eq. (2.26)], i.e.,
sin 0
= n.
(2.44)
cos 0
A prism for expansion along one axis
To avoid reshaping at the other surface of the prism, i.e., compression in the direction of the axis that has been expanded before, it is convenient to have the beam
exiting the prism perpendicular to its rear surface (3 = 0). Combining this condition [Eq. (2.35)] with the Brewster condition leads to the equation,
= 90 deg + arctan(n).
(2.45)
Optical Components
37
Figure 2.13 Expanding prism.
This implies that the following relation between the apex angle of the prism and its
refractive index should hold:
= arccot(n).
(2.46)
In this situation of practical importance (Fig. 2.13), the relation for the beamwidths
is especially simple, i.e.,
w3
= n.
(2.47)
w0
This formula can be concluded from Eq. (2.33) using the Brewster condition and
the additional condition expressed in Eq. (2.46). To this end, it is helpful to consider
the argument arcsin(sin 0 /n) first. Using Snells law and the trigonometric
identity arcsin x = 90 arccos x, one has

sin 0
= 90 deg + 0 .
arcsin
(2.48)
n
Combining this equation with Eq. (2.46) and the Brewster condition, it follows that

sin 0
= 0.
(2.49)
arcsin
n
This implies that the first factor in Eq. (2.43) is equal to 1. Using Snells law and
the trigonometric identity stated before, the second factor can be expressed as
cos[arcsin(sin 0 /n)]
sin 0
=
.
cos 0
cos 0
(2.50)
If this is combined with the Brewster condition (tan 0 = n), one finds that the
second factor is equal to the refractive index n.
Chapter 2
38
It seems worthwhile to present a shortcut to derive Eq. (2.47) for the case where
it can already be assumed that the second surface of the prism leaves the beamwidth
unaltered, i.e.,
w3
3 = 0
= 1.
(2.51)
w1
In this case, it is sufficient to consider
cos 1
w1
=
.
w0
cos 0
(2.52)
Using the Brewster condition (1/ cos 0 = n/ sin 0 ), it is concluded that

w1
cos 1
=
n.
w0
sin 0
(2.53)
The sine term can now be expressed by Snells law (sin 0 = n sin 1 ) as
w1
cos 1
=
= cot 1 .
w0
sin 1
(2.54)
At this point, it is convenient to make use of the alternative expression of the Brewster condition (cos 1 = n) in order to obtain Eq. (2.47) and we have
w3
w3 w1
=
= n.
w0
w1 w0
The type of prism considered here has numerous applications in optical devices.
In the literature, the abbreviation M = w1 /w0 for magnification can sometimes be
encountered in conjunction with the following equation (Hanna et al., 1975):

1 n2 sin2 0
.
(2.55)
M=
n 1 sin2 0
Its equivalence with Eq. (2.52) can directly be seen using Snells law (sin 0 =
n sin 1 ).
An analogous relation holds for a prism
that reduces the beamwidth along an axis instead of expanding it, namely,
A prism for compression along one axis
w3
1
= .
w0
n
(2.56)
Such a prism can be realized by letting a beam pass undeviated through the first
surface, so that the beam shaping occurs at the exit surface (Fig. 2.14). In contrast
to the beam-shaping effect described earlier, in this case a transition from a medium
with a higher refractive index to a medium with a lower index of refraction is of
importance (n sin 2 = sin 3 ).
Optical Components
39
Figure 2.14 Compressive prism.
Equation (2.56) can be derived in a way very similar to the derivation of

Eq. (2.47) if it is assumed that w2 /w0 = 1 holds, i.e., that perpendicular incidence
on the first surface is realized. Additionally, it has to be assumed that
tan 3 = n
cot 2 = n.
(2.57)
Combined with Snells law, the two equivalent equations lead to

w3
cos 3
1
1
=
=
= ,
w2
cos 2
n
cot 2
(2.58)
and combined with the assumption of perpendicular incidence finally gives

w2 w3
1
w3
=
= .
w0
w0 w2
n
(2.59)
For the purpose of tolerancing, it is advantageous to have explicit

formulas for the quantities of interest as functions of physical quantities that are
controlled by adjustment (the incidence angle 0 ) or by manufacturing (the refractive index n, the apex angle ). Equations (2.32) and (2.43) are appropriate starting
points for such a sensitivity analysis.
An important point is the question of how the exiting angle changes if the input
angle 0 is not well adjusted. In Fig. 2.15, the output angle is plotted as a function
opt
opt
of the deviation = 0 0 from the optimum angle 0 = arctan(n) for
an expanding prism (Fig. 2.15) made of the standard glass BK 7. The wavelength
considered is = 405 nm. This determines the refractive index used in Eq. (2.32),
which is n = 1.53024. In Fig. 2.16, the change of beamwidth calculated from
Eq. (2.43) is shown as a function of the deviation for the same prism.
Tolerancing
40
Chapter 2
Figure 2.15 Exit angle versus adjustment error for the prism of Fig. 2.13.
Figure 2.16 Beamwidth versus adjustment error for the prism of Fig. 2.13.
Figure 2.17 gives a representation of the change of the exiting angle with disadjustment for a compressive prism (Fig. 2.14). To allow for comparison with
Fig. 2.9, the same material and wavelength as before are chosen in this example.
Figure 2.18 shows the dependence of the beamwidth on a deviation from the optimum input angle.
Being a function of wavelength and temperature, the refractive index that has to
be considered in the analysis might change in practice, depending on the conditions
Optical Components
41
Figure 2.17 Exit angle versus adjustment error for the prism of Fig. 2.14.
Figure 2.18 Beamwidth versus adjustment error for the prism of Fig. 2.14.
under which the laser is operated. Figure 2.19 shows how the exiting angle changes
in the case of an expanding prism (Fig. 2.13) if there is a deviation n = n nopt
from the optimum refractive index. The apex angle is considered to be a fixed
value. The same holds for the angle of incidence. Again, Eq. (2.32) can be used
for a simple analysis. Figure 2.20 shows the same dependence for a compressive
prism (Fig. 2.14).
Chapter 2
42
Figure 2.19 Exit angle versus a change in refractive index for the prism of Fig. 2.13.
Figure 2.20 Exit angle versus a change in refractive index for the prism of Fig. 2.14.
2.2.2.4 Two-prism arrangements
Figures 2.21 and 2.22 show two ways of arranging two prisms. The ways that the
dispersion of light is influenced by these arrangements is quite different. While the
dispersion of the prisms adds if the two refracting edges are on the same side, it
subtracts if the prisms are arranged in a way that the apex angles are on different
sides (Barr, 1984). The first case corresponds to typical prism arrangements applied
Optical Components
43
Figure 2.21 Arrangement of two prisms to increase dispersion.
Figure 2.22 Arrangement of two prisms to decrease dispersion.
in optical spectroscopy to increase dispersion, while the second case is of importance here. Arrangements of two prisms in which the second case is realized are
common in devices for optical recording (Okuda et al., 1995) and also have been
used in dye laser systems (Niefer and Atkinson, 1988).
Having a look at the result for a single prism, the following equations can be written
immediately:

sin 4
II
II
,
(2.60)
nII

sin 0
I
I
,
(2.61)
nI
Relation of incident and exit angle in the case of a two-prism arrangement
where the angles are designated as depicted in Fig. 2.22 and II is the apex angle
of the second prism and nII is its index of refraction, while the variables I and nI
describe the first prism. In the equation for 4 , the relative orientation of the two
Chapter 2
44
prisms intervenes. It is expressed by the variable = 3 4 as

sin 0
I
I
.
nI
(2.62)
If this expression is substituted into the equation for 7 , the output angle of the
two-prism arrangement is found as a function of the input angle, i.e.,

1
II
II
I
7 = arcsin n sin arcsin II sin arcsin n sin I
n

sin 0
arcsin
.
(2.63)
nI
Exploiting the fact
stated earlier that the beamwidth transfer function can be composed as the product of the transfer function of the refracting surfaces that intervene, the transfer
function for the two prisms can now be written directly as
The transfer function for the beamwidth altered by two prisms
w7
cos(arcsin{nII sin[ II arcsin(sin 4 /nII )]}) cos[arcsin(sin 4 /nII )]
=
w0
cos 4
cos[ II arcsin( sinnII4 )]
cos(arcsin{nI sin[ I arcsin(sin 0 /nI )]}) cos[arcsin(sin 0 /nI )]

.
cos[ I arcsin(sin 0 /n)]
cos 0
(2.64)
The link between the two prisms is again established via the following equation:

sin 0
I
I
.
(2.65)
nI
Expansion or compression If two expanding prisms of the type shown in Fig. 2.13
are combined, the following set of equations is valid in the ideal case, i.e., when no
adjustment errors and no manufacturing errors are present:

First prism,
0 = arctan nI ,

I = arccot nI ,
(2.66)
4 = ,
Second prism,

= arctan nII ,

II = arccot nII .
(2.67)
Under these conditions, the beam leaves the last prism perpendicular to its second
surface, i.e., 7 = 0.
The beamwidth ratio after passing the two prisms adjusted in this arrangement
is
w7
= nI nII .
(2.68)
w0
In practice, two prisms of equal refractive index are often used.
Optical Components
45
Figure 2.23 Exit angle versus adjustment error.
Of course, it is also possible to arrange two prisms in such a way that the incident beam is perpendicular to each of the first surfaces of the prisms to realize
compressive beam shaping. If I = arccot(nI ) and II = arccot(nII ) hold, one
finds
1
w7
= I II .
(2.69)
w0
nn
The formulas presented here [Eqs. (2.63)(2.65)] allow for the investigation of a variety of dependencies. As an example, the sensitivity of the exit
angle on the input angle for a pair of expanding prisms of BK 7 at a wavelength of
405 nm is shown in Fig. 2.23.
Tolerancing
2.2.2.5 Prism with one internal reflection
A modified prism is sometimes used in data-storage applications in order to realize

a 90-deg deflection. A description of the component can be found in a U.S. patent
filed by Marchant (1988). The front surface of the prism can also serve as a beamsplitter to divert a laser-monitoring signal from the main optical path to a power
detector (Latta et al., 1992). Figure 2.9 shows the working principle of this prism.
The tunnel diagram (Fig. 2.10) shows how this component corresponds to a prism
with the apex angle . Using = 90 deg , the equations derived for a single
prism can directly be applied as

1
3 = arcsin n cos arcsin
,
(2.70)
sin 0
n

1
0 = arcsin n cos arcsin
,
(2.71)
sin 3
n
Chapter 2
46
w3
cos[arcsin(n cos{ + arcsin[(1/n) sin 0 ]})] cos{arcsin[(1/n) sin 0 ]}
=
.
w0
sin{ + arcsin[(1/n) sin 0 ]}
cos 0
(2.72)
The condition of perpendicularity of the exiting beam with respect to the last surface gives
3 = 0,
+ 0 = 0.
(2.73)
The Brewster condition [0 = arctan(n)] corresponds to
= arctan(n).
(2.74)
Therefore, this type of prism can be reduced to the type of prism discussed earlier,
making use of the tunnel diagram.
2.2.2.6 Prism with two internal reflections
At first glance, it seems difficult if not impossible to realize the parallelism of incident and exiting beams with a single refracting prism oriented in such a way that the
incident beam is perpendicular to its first surface. Fantone (1986, 1991) proposed
a type of prism that fulfills these requirements. His design (Fig. 2.24) is based on
two reflections inside of the prism and the following equation:
cos(3) = n cos ,
(2.75)
where n is the refractive index of the prism material and is the angle depicted in
Fig. 2.24. It is related to the apex angle in the following way: = 90 deg .
The design equation stated by Fantone (1986, 1991) can be derived by taking
a look at Fig. 2.24 and referencing the corresponding tunnel diagram. From the
drawing, it can be seen that must complement the angle B to 90 deg in order to
realize the parallelism condition, i.e.,
B = 90 deg .
(2.76)
The tunnel diagram that represents the single prism and takes the two reflections
into account is given in Fig. 2.25. There are two equations that follow directly from
this diagram, namely,
t = 90 deg t ,
(2.77)
= 3.
(2.78)
Figure 2.24 Prism with two internal reflections.
Optical Components
47
Figure 2.25 Tunnel diagram of the prism with two internal reflections.
Using t = 1t + 2t , it can be concluded that

t = 2t .
(2.79)
The combination of these three equations gives

2t = 90 deg 3.
(2.80)
Using A = 2t , Eqs. (2.76) and (2.80) are combined with Snells law as
n sin A = sin B ,
(2.81)
and the design equation (2.75) follows directly. These considerations on prisms
were restricted to beams of collimated light. If uncollimated light is used, astigmatism becomes an important feature (Naumann and Schrder, 1992).
In the technical literature, a multitude of prism-based arrangements are described that are beyond the scope of this introductory text. A sophisticated device
for optical data storage described by Saito et al. (2000) is based on a birefringent
crystal prism, for example.
Aspects of achromatic prism designs are treated for optical recording applications (Kay and Gage, 1997) and for applications in dye laser systems (Duarte and
Piper, 1982, 1983; Duarte, 1985a, 1985b). Additional resources on this subject can
also be found in the patent literature (Duarte and Piper, 1983; Luecke, 1997; Grove
and Shuman, 1992).
2.2.3 Axicon devices
Axicons (McLeod, 1954, 1960) are components featuring rotational symmetry with
respect to the optical axis. Unlike a lens whose surface can be thought of as being
created by letting a part of a circle rotate around the optical axes, axicons often
resemble a cone, i.e., their surface can be thought of as being created by letting
a straight line rotate around the optical axis. Figure 2.26 shows an axicon that is
used for linear focusing (Rioux et al., 1978), i.e., being illuminated by collimated
Chapter 2
48
Figure 2.26 Axicon used for linear focusing.
light, it creates a focal line extended from the intersection of its exit surface with
the optical axis to the point designated by L. Making use of the matrix method, it
is easy to derive an expression for the length of the focal line in terms of the axicon
parameters. We assume that the angle is small. Under this condition, we can
use the equation derived for a thin prism [Eq. (2.25)] to model the passage of a ray
through the axicon as
(1)

in

y
y
1 0
0
, with = (n 1).
(2.82)
=
+
0 1
(1)
in
This ray then propagates through air,
(2)

(1)

in

y
y
y
1 t
1 t
t
.
=
=
+
0 1
0 1
(2)
(1)
in
(2.83)
The condition that the incident light is collimated ( in = 0) implies that

y (2) = y in t.
(2.84)
The length of the focal line is now determined as the t value for which y (2) = 0
holds and y is at its maximum height, i.e.,
y in,max
y in,max
=
.
(2.85)
(n 1)
Annular and radial focusing are other applications of axicons to laser machining
(Rioux et al., 1978).
Of special interest are adaptive axicon devices as annular beam expanders (Rioux et al., 1978; Ichie, 1994). Here, an arrangement will be considered that decreases the diameter of an incoming collimated beam (Fig. 2.27).
The difference in height H , which is obtainable with the axicon arrangement,
can be found using the following formula:
L=
H =d
tan( )
,
1 [tan( )/ tan ]
with = arcsin(n sin ) and = 90 deg .

(2.86)
Optical Components
49
Figure 2.27 Adaptive axicon. The adaptive axicon is characterized by the angle . The two
components that form the adaptive axicon are separated at a distance d.
It can be derived from a consideration of triangles that gives tan( ) =

H /(a + d) and tan = H /a. To get a better understanding of this expression,
we approximate it using tan( ) tan as follows:
H
= d tan( ).
Further approximated and using the law of refraction, this reads as

H
= d arcsin(n sin ) .
(2.87)
(2.88)
Now, we use the fact that = , where is the angle of the axicon depicted in
Fig. 2.27, and linearize the equation as
H (d; n, )
= d(n ) = d(n 1).
(2.89)
This is the approximated expression that links the difference in beam diameter to
the state of the compound axicon and to its parameters n and = 90 deg .
2.3 Components Based on Reflection and Refraction

2.3.1 Integrating rod
The integrating rod is treated as an example of a light guide. This optical component is often used to homogenize light distributions (Chang and Shieh, 2000;
Magarill, 2002).
Chapter 2
50
Figure 2.28 Integrating rod.
The component seems to be quite simple at first glance, but it provides some
interesting features of unexpected complexity. Figure 2.28 shows a ray propagating
through a light guide. As before, we are interested in having an explicit formula
that gives the ray leaving the integrating rod as a function of the input ray and the
parameters of the rod, namely, its length T , the height d describing the entrance
surface, and its refractive index n. The entrance surface is assumed to be quadratic.
The ray entering the light guide undergoes refraction and the matrix describing
this reads as

1 0
A=
.
(2.90)
0 n1
Depending on its entrance height and entrance angle, the ray propagates through
the light guide along a straight line before it hits one of the surfaces that are perpendicular to the entrance surface for the first time. From the corresponding triangle,
we see that this propagation can be described using the matrix

d y
1 t (y, )
B=
, where t (y, ) =
.
(2.91)
0
1
tan
The ray is then reflected back in the direction of the optical axis. At this point,
we distinguish two cases as shown in Fig. 2.29. Both cases have in common that
they contain a part of the passage of the ray through the rod, during which it is
reflected in such a way that its ray coordinates are the same after the reflections. It
is therefore possible to conceptually separate N unit cells with this property from
the rod and to describe them by
N

1 0
.
(2.92)
C=
0 1
Proceeding further, either case 1 or case 2 can occur, and the corresponding matrices are

1 0
+
for case 1,
(2.93)
D =
0 1

1 0
+
for case 2.
(2.94)
D =
0 1
To obtain an explicit formula later, we combine them into one matrix. To this
end, the integer exponent M is introduced, which is a function of the start ray
coordinates and the rod parameters,

(1)M
0
D=
.
(2.95)
0
(1)M+1
Optical Components
51
Figure 2.29 Integrating rod, two cases.
From Fig. 2.29, we see that this matrix is then followed by a matrix describing a
propagation of the ray along a rectilinear line,

1 trest
E=
.
(2.96)
0 1
And finally, the ray is refracted at the exit surface of the light guide,

1 0
.
F =
0 n
(2.97)
The exponent M is determined using the div operation, which gives the integer part
of a division. For example, 7.8 div 2 evaluates to 3. M corresponds to the number
of reflections minus 1. From a geometrical consideration, this number is calculated
as

2d
M = T t (y, ) div s where s =
.
(2.98)
tan
The overall length of the rod is
T = t (y, ) + Ms + trest .
(2.99)
From this relation, we can determine the distance appearing in matrix E. This can
now be put together to obtain the system matrix of the integrating rod as

1
1
[t
(y,
)
t
(y,
)]
rest
n
S = F EDCBA = (1)M
,
(2.100)
0
1
52
Chapter 2
with trest (y, ) = T t (y, ) Ms, t (y, ) = (d y)/ tan , and M and s as in
Eq. (2.98).
This system matrix is quite different from the matrices encountered earlier insofar as it contains the coordinates of the ray. It provides an explicit formula to
calculate the ray leaving the light guide as a function of the parameters of the light
guide and the input ray.
The matrix S reflects two important properties of a mixing rod. The intricate
dependence of the entry s21 on the input coordinates indicates the mixing property
with respect to the y coordinates. The entries s21 = 0 and s22 = (1)M , on the
other hand, indicate that the angle is conserved with the exception of a change in
sign.
2.3.2 Triple mirror
Triple mirrors are used as retroreflectors in interferometers and spectrometers, as
well as in measurement devices for civil engineering. Unfolding the triple mirror
is different from the geometrical operations that we used until now. Drawing the
tunnel diagram for a prism, we used a reflection with respect to a plane that allowed
us to unfold the device and to make it more accessible to an analytical treatment.
Also, in the case of the light guides studied before, we used a symmetry operation
that consisted of reflection with respect to a plane.
To unfold a triple mirror, it is advantageous to use a reflection with respect to a
point. This operation is called an inversion.
Chapter 3
Sensitivities and Tolerances

The analysis of the sensitivity of an optical system with respect to misadjustment
is an important part of the optical design. The matrix method provides useful
tools to study tilt, decentering, defocusing, and beam pointing error in a systematic way.
We will start with a simple example and consider the surface in Fig. 3.1 that
separates two media with different indices of refraction. The surface is slightly
tilted at an angle . From Eq. (1.17), we know how the situation without tilt is
described by a matrix. The question is now how this relation has to be modified to
incorporate the effect of tilt.
Starting with the linearized form of the law of refraction, we find from Fig. 3.1
that

n1 (1) = n2 (2) .
Solving this equation for (2) and taking into account that the distance of the ray is
not changed in the reference plane, this can be expressed as follows:
(2)

(1)
1 0
0
y
y
=
+
.
(3.1)
0 nn12
1 nn12
(2)
(1)
Figure 3.1 Tilted surface. The surface that separates the two media with refractive indices
n1 and n2 is tilted by an angle .
Chapter 3
54
Figure 3.2 Tilted element. The optical element is tilted with respect to the optical axis and
also decentered by y1 .
An additional additive term appears in this equation. The matrix in Eq. (3.1) is
the same that is used to describe the untilted surface. It is a typical feature of
considerations with respect to tilt and decentering that they can be described by a
so-called error vector that represents the effect of misalignment.
We will now have a look at how an error vector is transformed if the misaligned
element cannot be described by a single reference plane, but if two reference planes
separated by a distance L are needed. An approach involving a coordinate transformation to the coordinate system of the misaligned element is presented in the
literature on laser technology (Siegman, 1986). Figure 3.2 shows an optical element that is decentered and tilted. Decentering and tilt are described by the error
vector

y1

1 =
(3.2)
1
with respect to the optical axis. In the coordinate system of the misaligned element, the vector R1 is transformed by the element matrix A to the vector R2 . The
transformation law for the error vector is different and is described by a matrix
that propagates the error vector along a distance L. Both transformations are then
combined for the transformation of the main coordinate system. Figure 3.3 gives
an overview of the transformations.
To analyze tilt and/or decentering, we have a choice between two alternative
methods. On the one hand, we can use 2 2 matrices and two-dimensional vectors
to study the influence of error vectors on rays propagating through the system.
On the other hand, we can use 3 3 matrices to describe the same situation in
homogeneous coordinates. In these coordinates, an equation of the form
(2)

(1)

y
a11 a12
y
y
(3.3)
=
+

(2)
a21 a22
(1)
is expressed as

a11
y (2)
(2) = a21
0
1
a12
a22
0
(1)
y
y
(1) .
1
1
(3.4)
55
Figure 3.3 Transformation used to describe a misaligned element. This figure gives a
scheme for how the coordinates in the reference system and those in the coordinate system
of the misaligned element are interrelated.
The advantage of the use of homogeneous coordinates stems from the fact that
every transition can now be described by cascading matrices and no additional
vectors appear.
3.1 Cascading Misaligned Systems

To perform the multiplication of the 3 3 matrices, it is convenient to organize
them as partitioned matrices. Equation (3.4) then takes the form
(2)

(1)
A b
r
r
= T
,
(3.5)
1
1
0
1
where 0 T is the transposed vector of the two-dimensional zero vector.
The product of two matrices of this kind can now be calculated as follows:
(2)
(1)
(2) (1)
A
A
A A
A(2) b (1) + b (2)
b (2)
b (1)
=
.
(3.6)
1
1
1
0 T
0 T
0 T
We observe that the error vector b of a system with N subsystems is formed according to the rule
b = A(N ) A(2) b1 + A(N ) A(3) b (2) + + A(N ) b (N1) + b (N ) . (3.7)
Chapter 3
56
This implies that, for the general case of N 3 3 matrices, the product matrix can
be expressed by the following formula (Kloos, 2007):

A
0 T
b
1
1

(k)
k=N
N1
k+1

(l)
(k)
A )b
+b
k=1 l=N
0 T
(N )
(3.8)

(1)
with 1k=N A(k) = A(N ) A
.
Please note that the symbol is used here to designate a matrix multiplication
(and not a multiplication of scalar quantities).
3.2 Axial Misalignment

Up to now, we have been looking at decentering and tilt. Now, we will also take an
axial misalignment as defocusing into consideration.
To analyze the tolerance of a given subsystem described by the matrix A, we
can proceed in this way: first, the deviation z from the undisturbed position is
expressed by a matrix as follows:

1
0
z
1
(3.9)
Depending on the application, there are two possibilities to multiply the matrices
and it might even be more appropriate to use two axial-misalignment matrices to
model the subsystem. We take a tolerance of the entrance reference plane of the
subsystem as an example. This leads to

A
1
0
z
1

=A+
0 a11
0 a21
z.
(3.10)
As before, we are able to separate the matrix of the subsystem and a term describing
the effect of the misalignment. In the case of axial misalignment, the corresponding
term is an error matrix and not an error vector.
This description provides a means by which to study the effect of the deviation
z on an optical device with the system matrix S = A(N ) A(1) . Assuming that
the matrix A corresponds to the kth subsystem of the optical device, we find for the
error matrix,

(k)
z
0
a
11
A(k1) A(1) .
S = A(N ) A(k+1)
(3.11)
(k)
0 a21
z
As an illustrative example, let us consider the cats-eye retroreflector formed by a
lens and a plane mirror. If we model the lens as a thin lens and assume that the
57
mirror is well positioned at a distance f with the respect to the lens, the system
matrix is

1 2f
(3.12)
S=
0 1
if the entrance and exit reference plane is coincident with the thin lens. If we now
assume that the mirror is moved with respect to the lens by a distance z, we find
1

f
1
S = 2z
.
(3.13)
1
f1
f2
The factor 2 in the error matrix stems from the fact that the ray passes two times
through the spacing that is changed by the deviation.
Alternatively, we might refer to an entrance and exit reference plane at a distance f in front of the lens. In this case, the results for the system matrix and the
error matrix of the cats-eye retroreflector appear as follows:
0 0
1 0
,
S = 2z
S=
.
(3.14)
1
0
0 1
f2
3.3 Beam Pointing Error

The matrix method is also a convenient tool with which to study the beam pointing error. For example, we might consider a laser beam that should enter a given
instrument parallel to the optical axis. We are interested in seeing how a deviation
from this alignment will affect the performance of an optical instrument. In this
case, we can start with
in
y
,
in
where in describes the deviation from the target alignment.
Tracing this ray through a system with the system matrix S leads to an error
vector of the form

y out
s12
in
=
.
(3.15)
out
s22
Chapter 4
Anamorphic Optics
Most of the examples treated until now featured symmetry with respect to the optical axis and it was sufficient to consider a projection onto the yz plane to find
the system matrix. But there are many optical systems that are not symmetric with
respect to the optical axis. In those cases, it is necessary to use an extended form
of the matrix method.
4.1 Two Alternative Matrix Representations

The generalization is straightforward and leads to a 4 4 matrix description. There
are two ways to state the corresponding matrices, namely,
in
out
m11 m12 r11 r12
x
x
out m21 m22 r21 r22 in

(4.1)
y out = s11 s12 n11 n12 y in
out
s21 s22 n21 n22
in
or

a11
x out
y out a21

out = c11
out
c21
a12
a22
c12
c22
b11
b21
d11
d21
in
x
b12
y in
b22
d12 in
d22
in
(4.2)
The second form has the advantage that algebraic commutation relations are fulfilled between the matrices A, B, C, and D and that it allows for a generalization of
the determinant relation known from 2 2 ray matrices with determinant 1 (Siegman, 1986, p. 618).
The first form reduces to the following partitioned matrix if there is no coupling
between the two orthogonal planes:
out
in
x
m11 m12 0
x
0
out m21 m22 0
0 in
=
.
(4.3)
out
y 0
0
n11 n12 y in
out
0
0
n21 n22
in
In the case of rotation symmetry, M and N would be equal. In the case of some
illumination devices such as headlamps, for example, they can be rather different.
Chapter 4
60
One would like to have a collimated beam in the yz plane and a high divergence
in the xz plane to have a broad illumination in the plane of the road. In such a
case, n22 = 0 in the paraxial approximation and m22 takes a value different from
zero in accordance with the divergence angle that has to be reached. We will refer
to the first form as expressed by Eq. (4.1) as representation A and the second form
given by Eq. (4.2) will be designated as representation B.
4.2 Orthogonal and Nonorthogonal Anamorphic Descriptions

The matrix given by Eq. (4.3), in which the block matrices R and S vanish, describes what might be called an orthogonal anamorphic system. In the more general
case of a nonorthogonal anamorphic system, there is coupling between the orthogonal planes, and the block matrices R and S have nonzero entries that reflect this
coupling.
Examples of an orthogonal anamorphic system are encountered describing an
anamorphic lens or a prism in a coordinate system oriented in accordance with
the main axes of these components. In these cases, the anamorphic symmetry is
inherent to the optical component under study. The rotational symmetry can also
be broken if a component with rotational symmetry is tilted with respect to the
common optical axis of a system.
Components with a three-fold symmetry can be analyzed using a nonorthogonal anamorphic description. Another example for a component with nonvanishing block matrices R and S is an anamorphic lens with its main transversal axes
oriented in such a way that they are rotated with respect to the two axes of the
coordinate system used.
4.3 Cascading
The matrices of representation A or B can be cascaded by multiplying the corresponding block matrices. We consider the passage through two optical systems I
and II. In representation A, the overall matrix then takes the form

M II
S II
R II
N II
MI
SI
RI
NI
M II M I + R II S I
S II M I + N II S I
M II R I + R II N I
S II R I + N II N I
(4.4)
If subsystem II can be described as orthogonal anamorphic arrangement, we have
x out
II I
out
= M M
y out
N II S I
out
in
x
in
M II R I
.
II I
in
N N
y
in
(4.5)
Anamorphic Optics
61
For the case where an orthogonal anamorphic description is appropriate for subsystem I, it follows that
out
in
x
x

II I
II I
in
out
M M R N
(4.6)
y out = S II M I N II N I y in .
out
in
In the case where both subsystems can be described as orthogonal anamorphic in
the same coordinate system, the product matrix for the compound system takes the
simpler form of an orthogonal anamorphic arrangement,
out
in
x
x in
II I
out

0
= M M
(4.7)
out
II
I
y
y in .
0
N N
out
in
The matrix of a compound system described by representation B also follows by
multiplication of block matrices,
II
II I
A
A BI
A A + B II C I AII B I + B II D I
B II
=
.
(4.8)
C II D II
C I DI
C II AI + D II C I C II B I + D II D I
Due to the different arrangement of spatial and angular coordinates in representation B, the symmetry expressed by Eq. (4.7) is less apparent in this matrix representation. But this representation has other advantages.
4.4 Rotation of an Anamorphic Component with Respect to the

Optical Axis
Although cascading subsystems by concatenating the corresponding matrices is
relatively simple, describing the rotation of an anamorphic component or system
with respect to the optical axis is somewhat more involved. Making use of matrix
representation B, we can state the coordinate transformation as follows (Hodgson
and Weber, 1997):

x
x
cos
sin
0
0
y sin cos
0
0
=
y ,
(4.9)

0
0
cos
sin
0
0
sin cos
where is the angle of rotation. The rotation matrices appear as block matrices in
Eq. (4.9). This is not the case if we turn to matrix representation A, i.e.,

x
cos
0
sin
0
x

0
cos
0
sin
.

(4.10)
y = sin
y
0
cos
0
0
sin
0
cos
Chapter 4
62
In order to be able to continue to calculate with block matrices, we might introduce

the following matrices:

cos
0
,
(4.11)
C =
0
cos

sin
0
.
(4.12)
S =
0
sin
With these abbreviations, Eq. (4.10) reads as

x
x
S
C
y = S C y
(4.13)
4.4.1 Rotation of an orthogonal system

Representation A
We are interested in the matrix of the anamorphic component after rotation. We
will start with the simpler case of a component to be rotated that has an orthogonal
anamorphic description,
out
in
x
x in

out
= X 0 ,
(4.14)
out
in
y
0 Y y
out
in
where X and Y are quadratic submatrices. They form the system matrix E as

X 0
.
(4.15)
E=
0 Y
This matrix is transformed according to the following transformation law into the
new, angle-dependent system matrix E ():
E () = T1 E T ,
(4.16)
where T1 is the inverse matrix of T , or

E () = TT E T ,
(4.17)
where TT is the transposed matrix of T . The transposed matrix can be formed

easily in the block-matrix notation and we find

C S
X 0
C S

.
(4.18)
E () =
0 Y
S C
S C
Anamorphic Optics
After multiplication, we have

C XC + S Y S

E () =
S XC C Y S
63
C XS S Y C
S XS + C Y C
(4.19)
By definition, C and S are diagonal matrices. The above equation can therefore
be simplied as

S C (X Y )
C2 X + S2 Y

.
(4.20)
E () =
S C (X Y ) S2 X + C2 Y
The upper-right and lower-left 2 2 block matrices are equal. The matrix difference that appears in these block matrices expresses the transversal anisotropy of
the anamorphic component or system. (The case X = Y would correspond to a
situation of transversal isotropy and could be treated with the simpler description
of the former chapters.)
To use this result in systems design or analysis, we have to make reference to its
matrix elements. The block matrices appearing in the above equation are therefore
given in expanded form to facilitate their use as follows:

cos2 x11 + sin2 y11 cos2 x12 + sin2 y12
2
2
C X + S Y =
, (4.21)
cos2 x21 + sin2 y21 cos2 x22 + sin2 y22

sin2 x11 + cos2 y11 sin2 x12 + cos2 y12
2
2
, (4.22)
S X + C Y =
sin2 x21 + cos2 y21 sin2 x22 + cos2 y22

x11 y11 x12 y12

S C (X Y ) = sin cos
.
(4.23)
x21 y21 x22 y22
Representation B
After abbreviating the rotation matrix as

cos
R=
sin
sin
cos
we might reformulate Eq. (4.9) in the following way:

x
x

y
y
R
0
= 0 R
(4.24)
(4.25)
The transformation matrix T is therefore formed by two identical block matrices,

R 0
.
(4.26)
T =
0 R
Chapter 4
64
Described by representation B, an orthogonal anamorphic system has a system matrix E of the following form:
axx 0 bxx 0
0 ayy
0 byy
.
E=
(4.27)
cxx
0 dxx 0
0 dyy
0 cyy
Unlike in the case for representation A, there are no zero block matrices, but an
orthogonal anamorphic arrangement is characterized by diagonal block matrices.
At this point, we can apply the transformation law (Hodgson and Weber, 1997)
to find the system matrix E () of the component or system rotated by an angle
as
E () = T 1 ET = T T ET .
(4.28)
Before specializing E to the case described by Eq. (4.27), it is convenient to have
a look at the more general case because it will bring out an advantage of the form
that the transformation matrix takes in representation B. Let us therefore consider
the more general anamorphic system with

A B
.
(4.29)
E=
C D
The transformation law can directly be applied to this general matrix as

T
R 0
A B
0
R

.
E () =
0 R
C D
0 RT
It follows that
E () =
R T AR
R T CR
R T BR
R T DR
(4.30)
(4.31)
The calculation of the matrix entries in the special case of an orthogonal anamorphic system is straightforward, i.e.,

axx 0
cos
sin
cos sin
R T AR =
.
(4.32)
sin cos
0 ayy
sin
cos
Multiplication of the matrices leads to

cos2 axx + sin2 ayy sin cos (axx ayy )
T
R AR =
.
sin cos (axx ayy ) sin2 axx + cos2 ayy
The corresponding result for the upper-right entry is

cos2 bxx + sin2 byy sin cos (bxx byy )
T
,
R BR =
sin cos (bxx byy ) sin2 bxx + cos2 byy
(4.33)
(4.34)
and the other two entries follow on the same footing.

Equation (4.31) is useful in studying anamorphic instruments or setups and will
be used in illustrative examples.
Anamorphic Optics
65
4.4.2 Rotation of a nonorthogonal system

Now, we will turn to the more general case of a nonorthogonal anamorphic component or system. Before rotation, its optical properties might be described by the
system matrix E,

X U
,
(4.35)
E=
V Y
wherein U and V are nonvanishing block matrices. As before, the transformation
formula is applied, in which appears as the parameter for the rotation angle,
E () = T1 E T .
(4.36)
After multiplying the matrices and using the diagonality of the block matrices C
and S , we find

E () =
C2 X C S (U + V ) + S2 Y
C S (X Y ) + C2 U S2 V
C S (X Y ) S2 U + C2 V
S2 X + C S (U + V ) + C2 Y

.
(4.37)
This can be decomposed into two matrices, F () and G (),

E () = F () + G (),
(4.38)
with
F () =
G () =
S C (X Y )
C2 X + S2 Y
S C (X Y ) S2 X + C2 Y
C S (U + V )
S2 U + C2 V
C2 U S2 V
C S (U + V )
(4.39)
(4.40)
The first matrix appearing in Eq. (4.38) corresponds to the orthogonal anamorphic
case and the second matrix describes the deviation from it. Please note thatapart
from the angle-dependent matricesthe matrix F () only contains the matrices X
and Y , whereas the matrix G () only contains the matrices U and V . The second
matrix has the interesting property that its trace is equal to zero.
4.5 Examples
4.5.1 Rotated anamorphic thin lens
A common problem in building up an instrument with anamorphic components is
to determine the system matrix E () of an anamorphic lens that has been rotated
by an angle about the optical axis of the device. This can be of interest as part of
a sensitivity analysis or because the component is intentionally positioned at this
angle. We will treat the lens in the thin-lens approximation.
Chapter 4
66
Using matrix representation B, the system matrix of the unrotated thin lens can
then be stated as follows:
1
0
0 0

1
0 0
A B
0
= 1
E=
.
(4.41)
0
1 0
C D
fx
0
f1y 0 1
Applying the transformation law [Eq. (4.31)], leads us to
E () =
where
R T AR
R T CR

I=
1 0
0 1
R T BR
R T DR
I
T
R CR

and
O=
0 0
0 0
O
I
(4.42)
Therefore, the system matrix of the anamorphic thin lens rotated by an angle
about the optical axis is
E () =
1
0
2
cos2
fx sinfy

sin cos f1x +

1
0
1

sin cos f1x +
fy
sinfx
cos2
fy
0 0
0 0

.
1
1 0
fy
0 1
(4.43)
4.5.2 Rotated thin cylindrical lens

In workshop practice and industry, cylindrical lenses are often applied. The system
matrix of a thin cylindrical lens rotated by an angle about the optical axis follows
from the corresponding result for the thin anamorphic lens by letting either 1/fy =
0 or 1/fx = 0. The system matrices that are obtained in this way for the cylindrical
lenses with different initial orientations are
E (fy , ) =
E (fx , ) =
0
0 0
1
0 0
1
,
cos2
fx
1 0
sin cos fx

2
sinfx
0 1
sin cos f1x
1
0
0 0
0
1
0 0
1
sin2
.
1 0
sin cos fy
fy
1
2
sin cos fy
cosfy
0 1
1
0
(4.44)
(4.45)
Anamorphic Optics
67
4.5.3 Cascading two rotated thin cylindrical lenses

A typical arrangement encountered to analyze anamorphic optical devices is a setup
of two cylindrical lenses. These lenses might have different focal lengths, f1 and
f2 , and in the more general situation they are rotated by different angles and
with respect to the optical axis. The two lenses are positioned with a distance d
between them.
Assuming a paraxial situation and assuming the thin-lens approximation to be
a sufficient description for the given purpose, we can analyze this arrangement
making use of Eq. (4.44). The paraxial optical properties of the two-lens arrangement in terms of the five parameters that intervene will then be expressed by the
system matrix that is formed by matrices for the two anamorphic lenses and a translation matrix.
We might note the first cylindrical lens with a block matrix F1 as

I
O
F1 =
,
(4.46)
P1 (f1 , ) I
where

I=
1 0
0 1
P1 (f1 , ) =
O=
2
cosf1
sin fcos
1
0 0
0 0
and
sin fcos
1

.
sinf1
The spacing between both cylindrical lenses is taken into account by using the
translation matrix D,

I dI
.
(4.47)
D=
O I
The second cylindrical lens has the block matrix F2 ,

I
O
,
F2 =
P2 (f2 , ) I
with

P2 (f2 , ) =
cosf2
sin fcos
2
sin fcos
2
sinf2
(4.48)

.
These three matrices are now concatenated to find the system matrix S of the
anamorphic two-lens arrangement as
S = F2 DF1 .
(4.49)
It is advantageous to use the block-matrix description to perform the matrix multiplications instead of operating on the matrix entries of P1 or P2 , i.e.,

I
O
I dI
I
O
.
(4.50)
S=
P1 (f1 , ) I
O I
P2 (f2 , ) I
Chapter 4
68
After multiplying the block matrices, we have

I + dP1 (f1 , )
dI
.
S=
P1 (f1 , ) + P2 (f2 , ) + dP1 (f1 , )P2 (f2 , ) I + dP2 (f2 , )
(4.51)
To trace rays through the anamorphic arrangement, it is then necessary to make
reference to the explicit form of the two matrices P1 and P2 in terms of focal
lengths and rotation angles as stated above. In conjunction with these two equations, Eq. (4.51) allows us to get a convenient overview of the five-dimensional
parameter space of this anamorphic optical setup.
4.5.4 Cascading two rotated thin anamorphic lenses
The generalization of the considerations made for cylindrical lenses to the analysis
of anamorphic lenses of a more general form is straightforward. But the notation
becomes more involved because we have to deal with a seven-parameter problem
instead of a five-parameter problem now. In place of making use of one of the matrices of a rotated cylindrical lens, we will make reference to the matrix given for
a rotated thin anamorphic lens [Eq. (4.43)]. We will describe the first anamorphic
y
lens by its two focal lengths f1x and f1 and the angle of rotation . The corresponding submatrix then reads as

2
2
sin cos f1x + f1y
cosf x sinf y
x y
1
1
1
1
. (4.52)

P1 f1 , f1 , =
2
2
sinf x cosf y
sin cos f1x + f1y
1
The submatrix for the second anamorphic lens has the analogous form

2
2
cosf x sinf y
sin cos f1x + f1y
x y
2
2
2
2
. (4.53)
P2 f2 , f2 , =

2
2
sin cos f1x + f1y
sinf x cosf y
2
These two lenses are placed at a distance d from each other. The system matrix
is determined as before by concatenating the two matrices for the lenses and a
matrix D that describes the transfer through the spacing between these lenses, i.e.,
S = F2 DF1 .
In block-matrix notation, this equation reads as

I dI
I
I
O
S=
y
y
O I
P1 (f1x , f1 , )
P2 (f2x , f2 , ) I
(4.54)
O
I
(4.55)
Finally, we have for the system matrix of the two rotated anamorphic lenses,

dI
I + dP1 (f1i , )
,
S=
P1 (f1i , ) + P2 (f2i , ) + dP1 (f1i , )P2 (f2i , ) I + dP2 (f2i , )
(4.56)
wherein i takes the values 1 and 2.
Anamorphic Optics
69
Please note that the concatenated block matrices of representation B take a

familiar form that is reminiscent of the nonanamorphic matrix description when
the matrix entries are scalars and no submatrices.
4.5.5 Quadrupole lens
With the new prerequisites of this chapter, we can study interesting optical devices as the so-called quadrupole lens. It is described in Hodgson et al. (1992), and
Hodgson and Weber (1997), as a convenient tool for an experimental phase-space
analysis. A look at the anamorphic matrix will give us insight into why this is so.
The quadrupole lens mainly can be modeled by two thin cylindrical lenses of
opposite focal length. These lenses are rotated at 90 deg with respect to each other
and the two-lens arrangement is rotated with respect to the optical axis. Using symbols as in Eq. (4.51), these angles of rotation are = 45 deg and = +45 deg.
We can assume a vanishing distance d = 0 between the two lenses. In fact, the
effect of the two crossed lenses can be incorporated in a single anamorphic lens
to form the quadrupole lens (Hodgson et al., 1992; Hodgson and Weber, 1997).
The system matrix for two rotated thin cylindrical lenses as given by Eq. (4.53)
then reduces to

I
O
S=
.
(4.57)
P1 (f1 , ) + P2 (f2 , ) I
Incorporating the special features of the quadrupole lens, we have

I
O
S=
.
(4.58)
P1 (f, 45 deg) + P2 (f, +45 deg) I
Using sin(45 deg) = 1/ 2 and cos(45 deg) = 1/ 2, we have for the sum of
the two submatrices
1

1 12 12
1
12
2
. (4.59)
P1 (f, 45 deg)+P2 (+f, +45 deg) =
1
1
1
f 12
f
2
2
2
Equation (4.58) can therefore be written in a surprising mathematical form, i.e.,
1
0 0 0
0
1 0 0
S = 0 1 1 0 .
(4.60)
f
1
f
0 1
Looking at the corresponding inputoutput relation for the quadrupole lens,
out
1
0 0 0
x in
x
in
y out
0
1 0 0
=
y ,
(4.61)
1
out
0 f 1 0 in
1
out
in
0 0 1
f
Chapter 4
70
we see that a direct link has been established between y in and out and between x in
and out .
To understand why this optical property of the quadrupole lens is advantageous
in a so-called phase-space beam analyzer, we will analyze such a device using the
description by anamorphic matrices in representation B. The purpose of this instrument is to create a phase-space representation of the light distribution in a given
object plane. On a screen in the image plane of the instrument, a typical representation is created with one axis of the screen representing the space coordinate of a
one-dimensional distribution in the object plane and the other axis representing the
angular coordinate of a one-dimensional distribution in the object plane. A onedimensional distribution is selected by using one slit close to the object plane to
restrict the spatial extension of the light distribution and another slit close to the
imaging lens to restrict the angular distribution accordingly. A complete analysis
can be performed by rotating the analyzing device, i.e., the two slits, the imaging
lens, and the quadrupole lens, about the optical axis. The quadrupole lens is positioned at the focal distance f0 of the imaging lens behind that lens. The quadrupole
lens has the length f . Naming the distance from the object plane to the imaging lens
a and the distance from this lens to the image plane b, one can write the following matrix chain to characterize the experimental situation (Hodgson and Weber,
1997):
S = Tbf0 Qf Tf0 Ff0 Ta ,
(4.62)
wherein three translation matrices appear
we obtained before,
1
0 0
0
1 0
Qf = 0 1 1
f
1
f
and Ff0 represents a standard lens,
1
0
0
1
Ff0 = 1
0
f0
0
f10
and Qf stands for the quadrupole lens

0
0
0

I O
,
=
Q I
(4.63)
0 0

0 0
I
1 0
P
0 1
Rewriting the matrix chain in block-matrix form, we have

I f0 I
I O
I
I (b f0 )I
S=
Q I
P
O
I
O
I
O
I
O
I
(4.64)
I
O
aI
I
.
(4.65)
After multiplying the submatrices, we obtain

S=
I + bP + (b f0 )Q + (b f0 )f0 QP
P + Q + f0 QP
(a + b)I + bP + (b f0 )(a + f0 )Q + (b f0 )f0 abQP

I + aP + (a + f0 )Q + f0 aQP
(4.66)
Anamorphic Optics
71
The product of the two submatrices takes diagonal form

QP =
1/f
0
0
1/f
1/f0
0
0
1/f0
1
I.
ff0
(4.67)
For the given purpose, we are interested in the functions that determine (x out , y out ).
It is therefore sufficient to have a look at the upper-left and the upper-right submatrices appearing in the system matrix. The upper-left submatrix can be evaluated as
follows:

bf0
0
fb0
0
bf
0
0
f
f
I+
+
+ bf0
0
0
fb0
bf
0
0
f
f

0
1 fb0
.
(4.68)
=
0
1 fb0
We observe the cancellation that takes place and causes that this submatrix has an
advantageous diagonal form. For the upper-right submatrix, we find
(a + b)I + abP + (b f0 )(a + f0 )Q + (b f0 )f0 aQP = a + b
ab
f0
f0 (bf0 )
I.
f
(4.69)
At this point, we can use the imaging condition 1/f0 = (1/a) + (1/b) to our
advantage. It implies that the following equation holds:
(a + b)I + abP + (b f0 )(a + f0 )Q + (b f0 )f0 aQP =
f0 (b f0 )
I. (4.70)
f
Putting these results together, we have the following inputoutput relation:

x out
y out
b
= 1
f0
1 0
0 1
x in
y in
f0 (b f0 )
0 1
1 0
in
in
. (4.71)
If we assume the input light distribution to be restricted to one dimension and

choose an orientation of the slits along the x-axis, for example, we may assume
that y in = 0 and in = 0. It follows that

x out
y out

b
f0
x in
0 ) in
f0 (bf

.
(4.72)
This reflects that the instrument indeed operates as a phase-space analyzer: the
two phase-space coordinates (x in , in ) are imaged on the analyzing screen (with
different scaling factors).
72
Chapter 4
4.5.6 Telescope built by cylindrical lenses

In industrial inspection devices, anamorphic arrangements are used that have a telescoping effect in one principal plane. One way to realize them is by mounting
cylindrical lenses. To describe such a setup, we can use the system matrix of two
cylindrical lenses. They are mounted to have a refracting effect in the same principal plane. Therefore, we can set = 0 deg and = 0 deg. In order to have
the wanted telescoping effect in that plane, the condition d = f1 + f2 has to hold,
where d is the distance between the two lenses. Depending on whether a Galilean
telescope (f1 < 0, f2 > 0) or a Keplerian telescope (f1 > 0, f2 > 0) is considered,
the signs of the focal lengths might differ.
For both cases, the submatrices in Eq. (4.53) take the following form:

f11 0
f12 0
P1 (f1 , 0 deg) =
and P2 (f2 , 0 deg) =
. (4.73)
0
0
0
0
With these submatrices and the telescoping condition, the system matrix of the
compound anamorphic telescope reads, using representation B, as
2
1 f1f+f
0 f1 + f2
0
1
0
1
0
f1 + f2
S=
(4.74)
.
f1 +f2
2
f1 f1 + ff1 +f
0 1 f2
0
1
2
1 f2
0
0
0
1
4.5.7 Anamorphic collimation lens
Representation A also has its advantages and can be convenient if a decoupling
of both orthogonal planes can be exploited during the calculation. This might be
illustrated using an example from LED technology where anamorphic lenses are
used to project the light from the LED chip onto the road or on the faade of a
building. Let us consider a lens that has a collimating effect in the yz plane and
that diverges the emitted light in the xz plane.
First, we note the matrix of an anamorphic lens with two different focal lengths
in representation A, i.e.,
1
0
0
0
1 1
0
0
fx
F =
(4.75)
.
0
1
0
0
0
0 f1y 1
The lens is positioned at a distance d from the LED. We can express this using the
matrix G,
1 d 0 0
0 1 0 0
(4.76)
G=
0 0 1 d .
0 0 0 1
Anamorphic Optics
73
The system matrix for an anamorphic lens at the given distance then follows directly by concatenating both matrices,
1
1
fx
E = FG =
0
0
d
1 fdx
0
0
0
0
0
0
1
f1y
d
f
1 dy
(4.77)
At this point, we can introduce the condition for collimation in the yz plane by
setting d = fy in Eq. (4.75),
1
1
E = fx
0
0
fy
1 ffxy
0
0
0
0
1
f1y
0
0
.
fy
0
(4.78)
The last entry of the matrix reflects the collimation condition. The entry in the
second row and the second column shows how the diverging effect in the orthogonal
plane scales with the two focal lengths.
4.6 Imaging Condition

In Section 4.5, examples of an anamorphic telescoping arrangement and a matrix
description of a lens that has a collimating effect in one plane were given. Here, we
will turn to the question how the imaging condition is reflected in this formalism.
We consider a system matrix expressed in representation B,

x out
a11
y out a21

out = c11
out
c21
a12
a22
c12
c22
b11
b21
d11
d21
b12
b22
d12
d22
x in
y in
in .
in
(4.79)
For imaging, the relation between the transversal spatial coordinates in the object
plane and in the image plane is of importance, i.e.,

x out
y out

=A
x in
y in

+B
in
in
(4.80)
Rays leaving at (x in , y in ) should all arrive at (x out , y out ) in the paraxial approximation. This implies that (x out , y out ) should not depend on B in , where (x in , y in ). We
might state this as
in

.
(4.81)
=
B
0
in
Chapter 4
74
In order to proceed, we might interpret this equation as a characteristic equation for

the matrix B with an eigenvalue 0 (van den Eerenbeemd, 2001). Because the determinant of a matrix is the product of its eigenvalues, it follows that the determinant
is zero in this case, i.e.,
det B = 0.
(4.82)
Equation (4.82) is the analytical expression of the imaging condition in the anamorphic case.
Example
To illustrate its application and its physical content, we will have a look at a simple
example. Let us consider a thin anamorphic lens between two spacings a and b.
This situation can be described by the following matrix product in representation B:
I
O
I bI 1
I aI
0
S = Tb F Ta =
.
(4.83)
fx
O I
O I
I
0
f1y
Multiplication of the block matrices gives

S=
I + bP
P
with

P =
(a + b)I + abP
I + aP
f1x
f1y
(4.84)

.
At this point, we can identify the upper-right submatrix with the matrix B,
B = (a + b)I + abP .
(4.85)
Therefore, the following equation corresponds to the anamorphic imaging condition:

a + b ab
0
fx
= 0.
(4.86)
det
0
a + b ab
fy
For a given a, this is a quadratic equation in b, i.e.,

ab
ab
a+b
a+b
= 0.
fx
fy
(4.87)
The imaging equation has two solutions in this anamorphic case. They determine
image positions that correspond to the two orthogonal planes. The positions of the
focal lines of anamorphic systems can be determined with a similar approach.
Anamorphic Optics
75
4.7 Incorporating Sensitivities and Tolerances in the Analysis

In this chapter, we put some effort into understanding the effect of a rotation about
the optical axis in an anamorphic system. In Chapter 3, we considered tilt and
decentering using homogeneous coordinates. A question that is near at hand at
this stage is whether both mathematical descriptions might be combined in order
to study tolerance anamorphic systems in more detail. This can in fact be done by
extending the transfer matrices of representation B by homogeneous coordinates,
out
in
x
a11 a12 b11 b12 x
x
y out a21 a22 b21 b22 y y in
out
= c11 c12 d11 d12 in .

(4.88)
out c21 c22 d21 d22 in

1
1
0
0
0
0
1
The misadjustment vectors for decentering and tilt appear in the fifth column of
this 5 5 matrix. For concatenating such matrices it is convenient to write them in
block-matrix form,
out
in
A B r
r
r
out = C D in .
(4.89)
0 T 0 T
1
1
1
The concatenation of two anamorphic subsystems I and II can then be stated as
II
I
A
A B I r I
B II r II
C II D II II C I D I I
1
1
0 T 0 T
0 T 0 T
II I
A A + B II C I AII B I + B II D I AII r I + B II I + r II
= C II AI + D II C I C II B I + D II D I C II r I + D II I + II . (4.90)
1
0 T
0 T
Chapter 5
Optical Systems
5.1 Single-Pass Optics
5.1.1 Triplet synthesis
A triplet is a system of three lenses. In a straightforward way, one might derive the
matrix of a three-lens arrangement by concatenating the matrices of thin lenses and
air spacings as

S=
1
f13
0
1
1 d2
0 1
1
f12
0
1
1
0
d1
1
1
f11
0
1
(5.1)
The matrix obtained in this way has the disadvantage of being relatively lengthy,
i.e.,

1 fd11 fd122
d1 + d2 df1 d2 2
S=
,
(5.2)
f13 f112 + fd1 f1 3 + f12d2f3
1 fd23 fd231
with
1
1
d1
1
=
+
f12
f1 f2 f1 f2
and
1
1
1
d1
=
+
.
f23
f2 f3 f2 f3
Again, it is advantageous for the discussion to reference the spacings that intervene
to the focal planes of the lenses. We saw already that this facilitated the discussion
of the doublet,
1
S=
f1 f2 f3
E2 f32
f22 + E2 E3
f12 f32
E3 f12
(5.3)
With this prerequisite, we can now state the design approaches for a focusing triplet,
an imaging triplet, an afocal triplet, or a collimating triplet. Making reference to
the consideration on the matrix entries, we can state the points of departure for the
triplet synthesis in these four cases.
Chapter 5
78
For a focusing arrangement:

s11 s12
0 s22
= T4 F3 T3 F2 T2 F1

1
E2 f32 E4 f22 + E2 E3 E4
=
f22 + E2 E3
f1 f2 f3
E1 E2 f32 E3 E4 f12 + f12 f32

E3 f12
.
(5.4)
For an imaging arrangement:

s11 0
s21 s22
= T4 F3 T3 F2 T2 F1 T1

1
E2 f32 E4 f22 + E2 E3 E4
=
f22 + E2 E3
f1 f2 f3
h12
E1 f22 E3 f12 + E1 E2 E3
with h12 = E1 E2 f32 E1 E4 f22 E3 E4 f12 + f12 f32 + E1 E2 E3 E4 .

(5.5)
For an afocal arrangement:

1
s11 s12
E2 f32
= F3 T3 F2 T2 F1 =
0 s22
f1 f2 f3 f22 + E2 E3
For a collimating arrangement:

s11 s12
= F3 T3 F2 T2 F1 T1
s21 0

1
E2 f32
=
f1 f2 f3 f22 + E2 E3
f12 f32
E3 f12
E1 E2 f32 + f12 f32

E1 f22 E3 f12 + E1 E2 E3
(5.6)
.
(5.7)
As an example of how this synthesis approach can be applied, let us consider

the afocal arrangement of three lenses. To fulfill the condition given by Eq. (5.6),
the following equation should hold:
f22 = E2 E3 .
(5.8)
This is the imaging equation in Newtonian form. It expresses that an imaging

condition has to be realized for the central lens in order to have the given three-lens
arrangement operate as an afocal system. Such an arrangement is well known as
the terrestrial telescope.
In the literature, triplet synthesis is quite often considered for the first case, i.e.,
for three lenses arranged in such a way that an object at infinity is imaged at a finite
distance (Berek, 1930, Malacara and Malacara, 1994).
Optical Systems
79
The synthesis approach with concatenated matrices is not restricted to the four
types of matrices considered as examples. Depending on the application, other
target matrices can be chosen for the system matrix. Several examples of system
matrices are given by Goodman (1995).
5.1.2 Fourier transform objectives and 4f arrangements
Here, we will consider an example that features unique properties, namely, the
matrix

1
0
f
S=
.
(5.9)
f 0
A way to obtain it with a single lens is by choosing both the input reference plane
and the output reference plane at the distance of the focal length,

1 0
1 f
1 f
.
(5.10)
S=
f1 1
0 1
0 1
This matrix has the property of mapping incident ray heights on ray angles and
incident ray angles as ray heights and is referred to as the Fourier transform
arrangement (Goodman, 1995). It is central to Fourier transform objectives and socalled 4f arrangements, which play a prominent role in spatial filtering and optical
information processing (Goodman, 1968; Yu and Khoo, 1990). In the paraxial
approximation, we can model them by concatenating two subunits of the abovementioned type (Fouckhardt, 1994),

f2

0
f1
0
f1
0
f2
S = S2 S1 =
=
.
(5.11)
f12 0
f11 0
0
ff12
After the two transformations, ray heights in the input reference plane are now
imaged on ray heights in the output reference plane, and input ray angles on output ray angles. Figure 5.1 shows typical rays in an arrangement with equal focal
lengths. This 4f scheme is applied in order to influence the output ray parameters
in a controlled way, for example, by placing a mask structure in the intermediate
plane,
out

in

0
f2
0
f1
y
m11 m12
y
=
f12 0
f11 0
out
m21 m22
in

f
f21 m22 f1 f2 m21

y in
.
(5.12)
=
m12
in
ff1 m11
f f
1 2
While the wave-optical analysis of these systems is performed in terms of

Fourier transforms (Goodman, 1968, Yu and Khoo, 1990), the matrix method provides a very convenient way to analyze processing devices of this kind in the first
order of approximation.
Chapter 5
80
Figure 5.1 4f arrangement. The figure shows schematically how typical rays of light are
transformed by the 4f arrangement. The vertical lines represent the positions of the two
lenses.
5.1.3 Telecentric lenses

In many optical applications, telecentricity of a lens arrangement is an important
feature to make the device relatively insensitive against axial changes. We use the
definition of telecentricity given by Goodman (1995): A lens is telecentric if the
chief rays are parallel to one another. Most commonly, they are also parallel to the
lens axis and perpendicular to the object and/or image planes that are perpendicular
to the axis. . .
Therefore, the following three types of telecentric systems can be distinguished:
1. Telecentric in object space.
2. Telecentric in image space.
3. Doubly telecentric.
We considered the principal ray as a chief ray descriptive of the system in Chapter 1.
Making reference to the decomposition used there, we can describe the system by a
matrix mapping from the object plane to the plane of the aperture stop and another
matrix that maps from the aperture stop to the image plane,
S = QP .
(5.13)
We can now state the condition of telecentricity in image space by the mapping
out

y
0
.
(5.14)
=Q
0
The condition that 0 = q22 can only be fulfilled if q22 = 0 holds. In this way, the
telecentricity condition in image space imposes a condition on the matrix Q.
In an analogous way, the condition of telecentricity in object space can be formulated. Starting from the mapping from the object plane into the plane of the
Optical Systems
81
aperture stop,

=P
y in
0
(5.15)
we see that 0 = p11 y in implies p11 = 0. This is the condition imposed on the
matrix P .
Finally, the question is near at hand: What would a doubly telecentric lens
arrangement look like? In the framework of the matrix description, the answer is
found by concatenating the submatrices that express the telecentricity conditions in
object space and in image space to determine a system that fulfills both conditions,
i.e.,

S = QP =
q11
q21
q12
0
0
p21
p12
p22
q12 p21
0
q11 p12 + q12 p22

q12 p12
. (5.16)
The system matrix of a doubly telecentric system is therefore characterized by the

entry s21 = 0. This implies that such a telecentricity condition can only be realized
if an afocal arrangement is chosen as the basis of the optical design.
5.1.4 Concatenated matrices for systems ofnlenses

At the beginning of the chapter, the triplet was considered. For the purpose of
reference, four cases are listed here that correspond to concatenating matrices of
the form Fk Tk with

Fk =
1
f1k
0
1

and
Tk =
1
0
fk1 + Ek + fk
1
(5.17)
The special form of the translation matrix is chosen to simplify the product matrices.
f12
,
(5.18)
E1

1
f22
E1 f22
,
(5.19)
n = 2: F2 T2 F1 T1 =
E2 f12 + E1 E2
f1 f2
n = 3:

1
E2 f32
E1 E2 f32 + f12 f32
F3 T3 F2 T2 F1 T1 =
,
f1 f2 f3 f22 + E2 E3 E1 f22 E3 f12 + E1 E2 E3
(5.20)
n = 1: F1 T1 =
1
f1
0
1
Chapter 5
82
n = 4:
F4 T4 F3 T3 F2 T2 F1 T1
1
=
f1 f2 f3 f4

f22 f42 E2 E3 f42

E2 f32 E4 f22 + E2 E3 E4
+E1 f22 f42 + E3 f12 f42 E1 E2 E3 f42

2
E1 E2 f3 E1 E4 f22 E3 E4 f12 + f12 f32 + E1 E2 E3 E4
(5.21)
5.1.5 Dyson optics
Dyson (1959) presented a system with a Petzval radius that is advantageous for
flattening the image plane. Figure 5.2 shows the concentric optical layout. The
formula that establishes the relationship between the radii and the refractive index
of the lens is
Rg
n
=
.
(5.22)
R+
n1
Using this in the equation for the inverse Petzval radius [Eq. (1.96)] gives a value
of zero.
The Dyson optics influenced optical systems used for lithography (Grenville
et al., 1991, 1993; Hsieh et al., 1992; Owen et al., 1993). New et al. (1992)
presented a modified version of the optics in which the lens and mirror are arranged
nonconcentrically.
It will now be shown how the systems matrix of the Dyson optics can be obtained. Starting from Fig. 5.2, we can unfold the optics along the apex of the mirror
and consider three separate steps of the ray propagation: (1) propagation to the
reference plane at the apex of the mirror, (2) reflection by the mirror, and (3) propagation from the reference plane at the apex of the mirror to the image plane.
A reference plane directly in front of the plane part of the lens is chosen to start.
The ray will undergo a refraction at this plane and then pass through the lens. To
determine the refraction at the curved side of the lens, it is important to apply the
Figure 5.2 Dyson optics. The two spherical surfaces are separated by a distance z.
Optical Systems
83
sign convention consistently. The convention used here is that the radius is negative
if the surface is concave.
To have a quantity for the distance that is counted positive as the ray propagates
along the optical axis, R+ = R is introduced, where R is the usual radius in the
refraction matrix. The matrix for the ray transfer through the lens can then be
written as follows:

R+
1
1
0
1 0
1 R+
n
A = 1n
= 1n
.
(5.23)
1
n
0 1
0 n1
R+
R
n
+
The next step is a propagation by a distance z. Having a look at Fig. 5.2, we see that
this lens should focus a parallel beam on the apex of the mirror. We can identify
the component a21 as the negative reciprocal of the focal length of the plano-convex
lens,
1
1n
=
.
(5.24)
f
R+
This suggests the choice
R+
z=
(5.25)
1n
in the propagation matrix. [At this point, it is interesting to note that in combination
with the concentricity condition (R+ + z = Rg ), Eq. (5.22) follows directly.] The
ray transfer matrix to the reference plane of the mirror reads as

R+
1
R+

1 1n
0
1 R+
1 1n
n
.
(5.26)
= 1n
1
1n
1
0
1
R+
n
R
n
+
Making use of Eq. (5.22), the reflection matrix is

1
0
1
=
R2g 1
2(n1)
nR+
0
1
(5.27)
In analogy to the first step, the matrix that describes the third step of the propagation
reads as

R+
1
R+
R+
1
1
1 1n
1
n
n
n
n
1n
.
(5.28)
=
n1
0
n1
1
0
1
R+
R+
At this point, we use the following abbreviation:
h :=
R+
.
1n
(5.29)
If we now collect the matrices for the three steps, we get the following product for
the system matrix:

1 0
0 h
1 4 hn
h
n
,
(5.30)
=
2
0 1
1
h1 n1
h1 0
nh
Chapter 5
84
and after using Eq. (5.28), the system matrix takes the form

S=
4R+
1 n(1n)
0
1
(5.31)
In some way, this optics is similar to optics that will be treated as double-pass optics
in Chapter 5.2.
With the matrix method, we can check another useful property of the Dyson
system (Malacara and Malacara, 1994): It is both front and back telecentric.
5.1.6 Variable single-pass optics
The usage of the word zoom lens may vary from author to author. The definition that will be applied here was given by Clark (1973): A zoom lens is an
image-forming optical system with such properties that an axial movement of certain components will produce a change in the equivalent focal length of the system,
while keeping the resultant image fixed with respect to the desired image plane.
These systems will be examined first. Later, we will enlarge our view and also
consider varifocal systems that form no image in the proper sense.
Designing a zoom device, it is important to distinguish between two possible approaches, namely, mechanical compensation and optical compensation (Back
and Lowen, 1958; Johnson and Feng, 1992). In a mechanical zoom lens arrangement, the movement of one lens (or lens group) is compensated by the movement
of another lens (or lens group) in order to keep the position of the image fixed.
Generally, the movement of the second lens (or lens group) has to be nonlinear in
order to obtain compensation. In an optically compensated zoom lens device, on
the other hand, only one lens (or lens group) is moved and the optical layout is
devised in such a way that (approximate) constancy of the position of the image
plane is obtained.
Both cases can be studied using the triplet as an example. To consider a triplet
objective that creates an image at a finite distance and that has its object at infinity,
we can make use of the systems matrix stated before in Eq. (5.2),

1 fd11 fd122
d1 + d2 df1 d2 2
A=
,
(5.32)
f13 f112 + fd1 f1 3 + f12d2f3
1 fd23 fd231
with
1
1
1
d1
=
+
f12
f1 f2 f1 f2
and
1
1
d1
1
=
+
.
f2 f3 f2 f3
f23
On this matrix, the focusing condition is imposed, namely, a11 = 0. This leads
to the following equation, which establishes the relation that has to be fulfilled
between the focal lengths of the individual lenses and the spacings between them
in order to have a lens arrangement with the desired property:
Optical Systems
85

d1 + d2 + d3 d2 + d3
d3
d 1 d 2 + d3
d3
1
+
+
f1
f2
f3 f1
f2
f3

1
d1 d2 d3
1
+
= 0.
+ d2 d3
f1 f2
f1 f2 f3
(5.33)
Now, the focal length of the lens arrangement has to be changed while keeping this
condition. As mentioned above, there are two ways to achieve this aim, namely,
mechanical or optical compensation.
5.1.6.1 Mechanical compensation of the triplet
We assume that a starting configuration has been found for which Eq. (5.33) holds,
and we consider that a lens is moved linearly. To realize that Eq. (5.33) also holds
for these new positions, an additional lens can be moved. The movement that is
necessary for proper compensation can be calculated introducing shifts l1 and
l2 of the first and second lens as
d1 = d10 l1 + l2 ,
(5.34)
d2 =
(5.35)
d20
l2 .
By combining the three equations, we see that the movement of the first lens must
be nonlinear if the second lens is moved in a linear way.
5.1.6.2 Optical compensation of the triplet
For optical compensation of the triplet, we can either move only a single lens or
link two lenses together and move them as a lens group. Let us first consider
the case of a single moving lens. As an example, the central lens of the triplet
is chosen and moved by a distance d along the optical axis. The unzoomed
start configuration is designated by d10 , d20 , and d30 . This corresponds to setting
d1 = d10 + d, d2 = d20 d and d3 = d30 in Eq. (5.32), i.e.,

d30 d10 + d d20 + d30 d
d30
d10 + d20 + d30 d20 + d30 d
+
+
1
f1
f2
f3
f2
f3
f1

0
0
0

(d + d)(d2 + d)d3
1
1
+ d20 d d30
1
+
= 0.
(5.36)
f1 f2
f1 f2 f3
This will now be interpreted as a polynomial in d,
a0 + a1 (d) + a2 (d)2 = 0,
with

d10 + d20 + d30 d20 + d30 d30 d10 d20 + d30 d30
a0 = 1
+
+
f1
f2
f3
f1
f2
f3

0
0
0

(d )(d )d
1
1
+ d20 d30
1 2 3 = 0,
+
f1 f2
f1 f2 f3
(5.37)
(5.38)
Chapter 5
86

(d10 + d20 )d30
d 0 + d30
d0
1
1
1
+ 2
+ 3 d30
+
,
f1 f2 f3
f2
f1 f2
f1 f3
f1 f2
0

d3
1
1 .
a2 =
f1 f2 f3
a1 =
(5.39)
(5.40)
The term a0 is equal to zero if the triplet complies with the focusing condition in its
starting configuration (characterized by d = 0). This is what we assume in this
derivation. For nonvanishing a2 , the polynomial is of second degree. This implies
that the focusing condition can be exactly fulfilled for two zoom positions only.
If we decide to move the first lens, the degree of the polynomial would be even
lower, namely, one. This is because only the spacing after this lens is then effectively changed according to d1 = d10 + d, while the spacings that follow remain
unchanged. The spacing before the lens does not contribute because the object is
at infinity and this spacing is therefore related to the passage of a collimated beam.
Moving the first lens would therefore not be a good choice to fulfill Eq. (5.32) using
optical compensation.
Now, the case of two coupled moving lenses is considered. A common choice
is to move the outer lenses (Back and Lowen, 1954, 1958). They are moved by
the same distance d along the optical axis. The variable spacings between the
lenses then change according to d1 = d10 d, d2 = d20 + d and d3 = d30 d.
Equation (5.32) then takes the form
d10 + d20 + d30 d
d 0 + d30 d30 d
2
f1
f2
f3

0

0
0
0

1
d d
d + d d2 + d3
1
+ d20 + d d30 + d
+ 1
+ 3
+
f1
f2
f3
f1 f2
0
0
0
(d + d)(d2 + d)(d3 + d)
1
= 0.
(5.41)
f1 f2 f3
This can be rearranged as

a0 + a1 (d) + a2 (d)2 + a3 (d)3 = 0.
(5.42)
In this polynomial, a0 also vanishes because we have assumed that the start
configuration fulfills the focusing condition,
a1 (d) + a2 (d)2 + a3 (d)3 = 0.
(5.43)
From the term on the right-hand side, we see that this is a polynomial of third
degree in d. This implies that exact compensation can be realized for three zoom
positions (Fig. 5.3).
Alternatively, we might consider coupling the two lenses on the right-hand side
and moving them a distance d along the optical axis. This leads to d1 = d10 +
d, d2 = d20 d + d = d20 , and d3 = d30 + d. The interesting point here is
Optical Systems
87
Figure 5.3 Variable optics: exact compensation at three points. The deviation from compensation is shown as a function of the displacement d.
that the movements of the two lenses that enclose the spacing d2 leave its effective
value unchanged. Therefore, if we substitute into Eq. (5.33) and take the same
steps as before, we are led to a polynomial of degree two. This suggests that letting
the positions of the outer lenses vary is a better choice for a triplet when optical
compensation has to be achieved.
It is observed that the number of positions for which exact compensation can be
achieved depends on the variable spacings. They contribute to the number if they
are not in a parallel beam and if the two lenses that they separate are not moved
in the same way. In designing a zoom lens, it is therefore important to make this
choice carefully in order to obtain the best compensation possible.
In the examples discussed until now, the object was at infinity. The same approach can also be applied to zoom objectives that perform point-to-point imaging.
In this case, we make reference to the imaging condition [Eq. (1.48)] and apply it
to the system matrices.
On the same footing, other instruments can be treated that are not zoom lenses
in the proper sense of the above-mentioned definition, but that are also varifocal devices. They have applications as variable beam expanders or adaptive collimators,
to name just a couple.
In the design of all these devices, the following theorem, which generalizes the
observations made in this chapter, can be advantageously applied.
5.1.6.3 Theorem on optically compensated adaptive lens arrangements
The number of positions for which rigorous compensation can be obtained equals
the number of variable spacings that are not in a collimated beam and that are not
between two lenses moving in the same way.
88
Chapter 5
Figure 5.4 Theorem on optically compensated systems: application to two lenses. The
horizontal bars indicate the components that are moved. They might be moved together,
indicated by a bar that links two lenses, or separately.
The application of this theorem is illustrated in table form for different lens
numbers and different arrangements in Figs. 5.45.6.
5.1.6.4 Adaptive optics
Adaptive devices (Tyson, 1998) are constantly gaining ground in scientific and industrial applications. A basic problem of adaptive optics might be stated as follows:
the output of an optical instrument is deteriorated by an error source that introduces
aberrations into the system. As a countermeasure to compensate these aberrations
and to keep the output of the instrument stable, an adaptive subsystem is added.
We might designate an unwanted change in optical power introduced to the optical
system by 1/faberr , the focal length of the undisturbed system by fstat , and the focal
length of the subsystem used for compensation by fcomp .
Optical Systems
89
Figure 5.5 Theorem on optically compensated systems: application to three lenses. As in

Fig. 5.4, the horizontal bars indicate the components that are moved.
Considering this example, we can use Eq. (5.33) again but read it differently. It
now gives the focal length fcomp , which is necessary to keep the image position as
a function of an unwanted change in optical power due to aberrations. Substituting
f1 = faberr , f2 = fstat , and f3 = fcomp in Eq. (5.33), one gets
fcomp =
d3 + (d1 d3 /faberr )[(d2 /fstat ) 1]
.
1 [(d1 + d2 + d3 )/faberr ] [(d2 + d3 )/fstat ] + {(d1 /faberr )[(d2 + d3 )/fstat ]} + d2 d3 (1/faberr + 1/fstat )
(5.44)
Chapter 5
90
(a)
Figure 5.6 Application to four lenses. (a) Theorem on optically compensated systemsthe
case of a vanishing upper-left matrix entry is represented.
In the case of no disturbing aberrations (faberr ), the focal length of the compensating device would take the following value:
fcomp =
d3
.
1 [(d2 + d3 d2 d3 )/fstat ]
(5.45)
In this example, fcomp can be realized by a zoom device as treated before, or by a

liquid lens (Berge and Peseux, 2000). In the case of a zoom device, the next step in
the design would be to model the zoom optics in the thick-lens approximation. The
thin-lens approximation used here gives a guideline on what zoom range has to be
realized.
Optical Systems
91
(b)
Figure 5.6 (Continued.) (b) Theorem on optically compensated systemsthe case of a
vanishing upper-right matrix entry is represented.
To generalize this approach to adaptive optics using the matrix method, we

might consider the adaptive feedback loop illustrated in Fig. 5.7. The matrix of the
adaptive system is
S = CT A,
(5.46)
wherein A describes the aberrations introduced to the system, T represents the
static part of the system, and C describes the subunit of the system used for compensation.
Depending on the application, we might impose the condition s11 = 0 as the
target state of the adaptive system. In this example, the following equation would
Chapter 5
92
(c)
Figure 5.6 (Continued.) (c) Theorem on optically compensated systemsthe case of a
vanishing lower-left matrix entry is represented.
give a condition for two entries in the matrix C of the subsystem used for compensation:
c11 (t11 a11 + t11 a21 ) + c12 (t21 a11 + t21 a21 ) = 0.
(5.47)
5.2 Double-Pass Optics

5.2.1 Autocollimator
Autocollimators are optical measurement instruments used in tooling. A typical
measurement task is to quantify a slight tilt of a reflecting plate. This can be done
with an autocollimator setting as depicted in Fig. 5.8. The divergent beam from
the light source is collimated by the lens and directed to the target. Because of
Optical Systems
93
(d)
Figure 5.6 (Continued.) (d) Theorem on optically compensated systemsthe case of a
vanishing lower-right matrix entry is represented.
the slight tilt of the target, the reflected beam shows a slight deviation as it hits
the detector. Assuming that the autocollimator is well aligned, no tilt of the plate
corresponds to a spot at the center of the detector, i.e., at the position were the
detector has its intersection with the optical axis. The deviation from this target
position is a measure of tilt.
For the purpose of the matrix description, we unfold the optical arrangement of
the double-pass optics. We make a coordinate break at the position of the beamsplitter and reference the matrices to the corresponding optical axes before and
after the coordinate break. The passage through the beamsplitter is then described
by the unity matrix, but we have to keep in mind that we changed the coordi-
94
Chapter 5
Figure 5.7 Adaptive feedback loop. The unit A describes the aberrations introduced in
the system, T represents the static part of the system, and C describes the subunit of the
system used for compensation.
Figure 5.8 Autocollimator. The rays emitted from the light source are reflected from the
slightly tilted mirror and pass through the beamsplitter to the detector element of the autocollimator.
nate system. A similar coordinate break is made to model the reflecting plane.
Here, we do not assume that it is well aligned, but we want to take its tilt angle
into account. This could be done making reference to homogeneous coordinates.
At this point, the autocollimator is described in terms of 2 2 matrices (Kloos,
2007).
Using the unity matrix of the unfolded beamsplitter implicitly, the matrix chain
from the light source to the target is

1 0
1 g
1 h+d
A=
f1 1
0 1
0
1

1 fg h + d + g fg (h + d)
=
,
(5.48)
f1
f1 (h + d) + 1
Optical Systems
95
where f is the focal length of the thin lens and h, d, and g are the distances appearing in Fig. 5.8. The ray is then affected by the tilted target and we use the following
relation for a tilted plane mirror to describe it:
(1)

(2)

y
1 0
0
y
.
(5.49)
=
+
0 1
2
(2)
(1)
The matrix chain for the passage from the tilted target to
follows:

1 0
1 d +e
1
B=
f1 1
0
1
0

g
1 d+e
g
+
(d
+
e)
1
f
f
=
1 fg
f1
the detector reads as

g
1

(5.50)
where e is the distance from the beamsplitter to the detector. Additionally, the
setting of the autocollimator has to be taken into account. The lens should direct
collimated light onto the target plane and it should focus the reflected light onto the
detector surface. This is expressed by the following two equations:
h + d = f,
(5.51)
d + e = f.
(5.52)
Combing these equations and the relations stated above, we find the relation for the
autocollimator as
out

in
1 0
2f
y
y

.
=
+
(5.53)
1
2 fg2 f1
2 1 fg
out
in
As read on the detector, the difference between the y value of the tilted plate and
the y value that would correspond to a perfectly aligned plate is of importance, i.e.,
y out = y out ( ) y out (0) = 2f .
(5.54)
It is this relation that reflects how the tilt of the target is translated to a difference
in position by the autocollimator.
Furthermore, a look at the inputoutput relation of the collimator suggests that
g = f might be a convenient choice for the corresponding distance.
5.3 Multiple-Pass Optics

Multiple-pass optics are encountered in spectroscopy as absorption cells and in
laser technology as optical resonators. To describe the multiple passage of a ray
through the same optical system, we could apply the same system matrix N times.
This corresponds to raising the system matrix to its Nth power and applying it to
Chapter 5
96
the vector of the input ray. Therefore, the Nth power of the ray transfer matrix is
an important tool for studying multiple-pass optics.
In this chapter, we will use some basic results from the eigenvalue theory of
linear algebra. To have explicit formulas for the matrix A raised to its Nth power,
it is useful to have a look at the eigenvalues of the matrix. They are helpful in order
to diagonalize the matrix and they are also closely related to the optical problem of
tracing a ray through a periodic system.
The eigenvalues of a 2 2 matrix are calculated from its characteristic equation,

a11
a12
det
= 0,
(5.55)
a21
a22
where det A stands for the determinant of the matrix A. This is equivalent to
2 (a11 + a22 ) + a11 a22 a12 a21 = 0.
(5.56)
This quadratic equation has the two solutions,

a11 + a22
1/2 =
a11 + a22
2
2
a11 a22 + a12 a21 .
(5.57)
For a system matrix, det A = n1 /n2 holds. Here, we will assume that both indices
of refraction are equal. Therefore, we have the simpler equation,
1/2
a11 + a22
=
It is the discriminant

dis =
a11 + a22
2
a11 + a22
2
2
1.
(5.58)
2
1
(5.59)
that determines the character of the eigenvalues. This discriminant provides a distinction that is of fundamental importance to the optics of multipass systems. If
dis < 0, the eigenvalues have an imaginary part, i.e., they have to be described by
complex numbers. If dis > 0, the eigenvalues are given by real numbers. In the
limiting case of dis = 0, there is only a single, real eigenvalue, namely,
=
a11 + a22
.
2
(5.60)
From Eq. (5.58), we see that the trace a11 + a22 of the matrix A determines the sign
of the discriminant. Because the assumption det A = 1 was made, all eigenvalues
satisfy the condition
1 2 = 1.
(5.61)
Optical Systems
97
We have to determine the corresponding matrices AN . To find them, we will make

a theorem from linear algebra. It states that AN can be calculated as
AN = F D N F 1
(5.62)
if A can be transformed to the diagonal matrix D by a transformation matrix F and

its inverse matrix F 1 by the transformation
A = F DF 1 .
(5.63)
To understand this theorem, we can concatenate the matrix A of Eq. (5.63) N times
and see how the inverse matrices cancel out in the product matrix.
Now, we will have a look at the case where dis < 0 and calculate the matrix AN that corresponds to this case. First, the transformation matrix F has to be
determined. From linear algebra, we know that the rows of this matrix are calculated from the eigenvector spaces of the matrix A that correspond to the eigenvalues
1 and 2 . The eigenvector space eig(A, 1/2 ) follows as the solution space of the
equation, i.e.,

x1
0
a11 1/2
a12
=
.
(5.64)
a21
a22 1/2
x2
0
Knowing that both eigenvalue are complex and that 1 2 = 1 holds, we can set
1/2 = ei with i as the imaginary unit, for which i 2 = 1 holds. Solving
Eq. (5.31) for these eigenvalues, we find
i

e
a22
i
eig A, e
=
.
(5.65)
a21
With this result, the transformation matrix can now be composed as
i
e a22 ei a22
.
F =
a21
a21
(5.66)
The inverse matrix reads

F
a12
a12
ei + a22
ei a22
Using these transformations, the matrix D takes the diagonal form
+i

e
1 0
0
=
.
D=
0 2
0
ei
(5.67)
(5.68)
We can now apply the theorem to find the matrix AN . Most entries follow directly
by performing the matrix multiplication. Calculating the upper-right entry of the
Chapter 5
98
matrix, the equations

x+y
xy
cos
,
2
2
= trace(A) = 2 cos , and
sin x sin y = 2 sin
(5.69)
a11 + a22
(5.70)
det A = 1
(5.71)
have to be used. Finally, we have

1
sin[(N + 1)] a22 sin(N)
a12 sin N
N
.
A =
a21 sin(N)
a22 sin(N) sin[(N 1)]
sin
(5.72)
In this case, the matrix AN is described by trigonometric functions. Equation (5.72)
is designated as Sylvesters theorem in the scientific literature (Gerrard and Burch,
1975).
To complete the picture, we will now have a look at the case where dis > 0. We
will proceed in an analogous way, but for a different pair of eigenvalues. Taking
into account that both eigenvalues are real and that 1 2 = 1 holds, we can use
1/2 = e as a starting point. Using this input, we then calculate

e a22
.
(5.73)
eig(A, e ) =
a21
With this result, we can now build the rows of the transformation matrix. In this
case, it reads
+
e a22 e a22
,
(5.74)
F =
a21
a21
and its inverse matrix is
F
1
=
+
a21 (e e )
a21
a21
e + a22
e a22
(5.75)
Using 2 sinh() = e e and applying the theorem An = F D n F 1 , also in this

case, we have

1
sinh[(n + 1)] a22 sinh(n)
a12 sinh(n)
n
.
A =
a21 sinh(n)
a22 sinh(n) sinh[(n 1)]
sinh
(5.76)
Also here, the upper-right entry of the matrix needed special attention and a11 +
a22 = trace(A) = 2 cosh and det A = 1 had to be used.
The interesting point now is to compare both results. In the first case, we have
a matrix made up of periodic functions that describes a periodic passage of the ray
through the multipass optics. This is a stable solution. In the second case, on the
other hand, exponential functions are encountered. The ray parameters diverge and
Optical Systems
99
this solution is therefore unstable. From this, we can conclude that the trace of
the system matrix is a decisive quantity to characterize the function of multipass
optics.
Stable passages of the rays are desired in spherical mirror interferometers (Herriott et al., 1964) and optical delay lines (Herriott and Schulte, 1965). The Herriott cell found applications in spectroscopy (Altmann et al., 1981, Hovde et al.,
2001) as an absorption cell and in interferometry (Antonsen et al., 2003a, 2003b).
It consists of two spherical mirrors separated by almost their radius of curvature.
Considerations on multipass cavities of this kind are given by Sennaroglu and Fujimoto (2003).
The distinction between stable and unstable periodic systems is fundamental in
the theory of laser resonators (Siegman, 1986, Hodgson and Weber, 1997, Svelto,
1998). The second case is also of importance in laser technology because so-called
unstable optical resonators find broad application in high-power laser systems. The
literature on optical resonators is abundant; there are many studies that make use of
the matrix method and apply it to the design and analysis of lasers.
Generalizations of Sylvesters theorem to the case where det A = 1 and/or
incorporating error vectors in homogeneous coordinates are given by Tovar and
Casperson (1995).
5.4 Systems with a Divided Optical Path

The method of ray-transfer matrices as presented here is not appropriate to describe
the phenomenon of interference. But it is a useful tool for analyzing problems that
arise while setting up and operating interferometers. It gives quantitative answers
on how to balance the optical beam paths of two-beam interferometers and on the
adjustment sensitivity of these instruments. The method can be used to derive
guidelines on how to make interferometric devices more robust.
The results obtained in this chapter can also be used in spectrometer design because the key optics of these instruments is often based on principles of interferometry. A typical example is a Fourier spectrometer based on a Michelson interferometer arrangement (Baker, 1977). An interesting account on designing spectrometers
that are robust with respect to tilt was given by Breckinridge and Schindler (1981).
5.4.1 Fizeau interferometer
Figure 5.9 shows the optical layout of a Fizeau interferometer used for lens testing. We will have a look at the measurement arm of the interferometer in order to
analyze its sensitivity with respect to the tilt of the spherical mirror.
The entrance plane of the measurement arm is chosen directly behind the beamsplitter. From this plane, the ray first propagates along a distance e before it is refracted by the lens under test. The optical unit formed by the lens and the concave
spherical mirror corresponds to an arrangement that was considered in Section 2.1,
Chapter 5
100
Figure 5.9 Fizeau interferometer. The beamsplitting element is represented by the horizontal bar. It separates the light from the laser into a reference beam and a beam that passes
through the measurement arm. Both beams are then directed by an additional beamsplitter
mounted at 45 deg to the detector or a camera.
i.e.,
wherein
A=
and
y

=A

2d
1 fd
1 2d
f
R

2
f2 1 fd R2 1 fd

b1
b2

= 2
b1
b2
(5.77)

2d 1 Rd

,
1 2d
2d
1 fd
f
R
(5.78)
d
1
d
f
(5.79)
We can therefore take its system matrix and error vector and apply it here to analyze
the Fizeau interferometer. Doing this, the measurement arm is implicitly unfolded
with respect to the apex of the spherical mirror.
The first propagation is described by a simple transfer matrix,
(1)

in
y
1 e
y
=
.
(5.80)
(1)
in
0 1
During the second step, we make reference to the result of Sec. 2.1, i.e.,
(2)
(1)

y
y
b1
=A
+
.
(2)
(1)
b2
(5.81)
Optical Systems
101
The ray finally returns to the reference plane chosen directly behind the beamsplitter and we have
(2)
out

y
y
1 e
=
.
(5.82)
0 1
out
(2)
When these three steps are combined, they result in
out

in

y
1 e
1 e
y
1 e
b1
=
A
+
,
out
in
b2
0 1
0 1
0 1
or

y out
out
1 e
0 1
1
0
e
1
y in
in
+ 2
d + e(1 fd )
1 fd
(5.83)

.
(5.84)
Having a look at the sensitivity, the -dependent vector on the right-hand side
is of interest. The sensitivities with respect to mirror tilt of the ray leaving the
measurement arm now follow immediately as
out

y
d
=d +e 1
,
(5.85)
f
out
d
=1 .
(5.86)
f
This sensitivity consideration would imply that d = f would be a convenient
choice to reduce the tilt error. In contrast to this finding, experimental arrangements corresponding to d = f + R are also often encountered. The reason for
this is that such an interferometric configuration allows for the measurement of odd
and even terms of the aberrations, while the first-mentioned arrangement allows for
the measurement of only the even terms. This is because the reflected beam passes
through the opposite half of the lens after reflection in the case of d = f and is not
reflected into itself.
For the arrangement that is suitable to measure odd aberrations, the sensitivity
analysis provides hints to minimize the tilt sensitivity of the interferometer:
out
R
= .
(5.87)
d=f +R
f
This implies that it is advantageous to choose a radius of the spherical concave
mirror that is small compared to the focal length under test.
5.4.2 Michelson interferometer
The Michelson interferometer is a generic type of interferometer and can be considered as a kind of workhorse of interferometry. We will consider a simple problem
that is encountered, for example, while adjusting such an interferometer for the
Chapter 5
102
Figure 5.10 Michelson interferometer. A beamsplitting cube with a thickness h serves to

separate and recombine reference and measurement beams of this Michelson interferometer.
Figure 5.11 Michelson interferometer, and measurement arm.
measurement of small vibrations of a sample provided with a mirror. The experimental situation is schematically shown in Fig. 5.10. The light beams originating
from both arms of the interferometer have to be superimposed carefully on the detector surface in order to observe and evaluate an interference pattern. It is then
of interest to know how the position of a laser beam that enters the beamsplitter
cube parallel to the optical axis would be changed by an unintentional tilt of the
sample. Using the ray-transfer matrix method, this and similar questions can be
answered in a rather straightforward way. To this end, the measurement arm of the
interferometer is first unfolded as shown in Fig. 5.11.
Using the matrix derived for a plane-parallel plate to describe the passage
through the beamsplitter cube, the ray transfer to the reference plane of the sample
under test is expressed by the following product matrix:

1 t
0 1
h
n
1
0
1 a
0 1
1
0
a+
h
n
+t
(5.88)
This implies that

y (1)
(1)
1
0
a+
h
n
+t
y in
in
(5.89)
The effect of an unintentional tilt of the sample is expressed by an error vector
Optical Systems
103
added to the term for an unfolded plane reflector, i.e.,
(1)

(2)

y
y
1 0
0
.
=
+
0 1
2
(2)
(1)
(5.90)
The propagation from the reference plane of the sample to the detector can be
written as a product of matrices in an analogous way as in Eq. (5.88), i.e.,

1 e
1 hn
1 t
1 t + hn + e
=
.
(5.91)
0 1
0 1
0 1
0
1
In this way, an equation for the output ray vector is obtained as
(2)
out

x
1 t + hn + e
y
=
.
out
0
1
(2)
(5.92)
By combining Eqs. (5.89), (5.90), and (5.92), the transformation that describes the
measurement arm of this Michelson interferometer is obtained as

out

1 a + 2 hn + e + t
t + hn + e
x in
y
. (5.93)
=
+ 2
out
in
0
1
1
The answer to the initial question can now be directly concluded from this equation.
For a laser beam that is first parallel to the optical axis ( in = 0), the change in
position as it impinges on the detector is as follows:

h
out
in
y = y y = 2 t + + e .
(5.94)
n
5.4.3 Dyson interferometer
The Dyson interferometer (Dyson, 1968, 1979; Steel, 1983; Heavens and Ditchburn, 1991, p. 197) is an example of a measurement head that is relatively insensitive with respect to tilt. This layout had some influence on later work in the field of
optical interferometry (Roberts, 1975).
Figure 5.12 shows a schematic layout of the interferometer. It is interesting
to analyze the device in order to understand the underlying compensation principle. To this end, the optical path is unfolded in order to determine an appropriate
representation in terms of transfer matrices. If the thicknesses of the quarter-wave
plate and of the reflecting element on the Wollaston prism are neglected, Fig. 5.13
represents the central beam.
The transfer matrix for this beam takes the following form:

yout
out

1 + 8 (1 a )(1 b )((a + b) + ab )
= f4 (1 fa )[1 f1 (a + b ab )] f 4 (1
f
f
f
f
f
2
f4 (1 fb )(a + b ab
f )
a
b
ab
2
f )(1 f )(a + b f ) + (1 f (a + b) +

1
ab
1
4 a + b ab
(a
+
b)
+
yin
2
f
f
f

.
+ 2
in
2 1 fa f2 (a + b) + 2 fab2 + 1
2ab 2
)
f2
(5.95)
Chapter 5
104
Figure 5.12 Dyson interferometer. A Wollaston prism serves as the beamsplitting element.
A quarter-wave plate near a small mirror on (or near) the Wollaston prism changes the
polarization in both paths of the interferometer accordingly, so that the light beams are
recombined when passing through the Wollaston prism again.
Figure 5.13 Dyson interferometer, and the unfolded path of the central beam.
For the purpose of tilt compensation, we are looking for a system with a vanishing
tilt-error vector. From Eq. (5.95), it is obvious that the second component of this
vector will vanish if a = f is chosen. The first component will then also be equal
to zero for any choice of b.
A look at the systems matrix A suggests that b = f would be a good choice to
let the matrix entry a12 vanish. Therefore, the following settings are made:
a = f,
b = f.
For this arrangement, the transfer mapping reads as

yin
1 0
yout
=
.
0 1
out
in
(5.96)
(5.97)
Using the concepts of Secs. 1.1 and 5.1, the stages of the unfolded optical path can
be interpreted as a Fourier lens arrangement, followed by a tilted mirror and a cats
eye that is again followed by a tilted mirror and a Fourier lens arrangement.
While the equation for the central beam was given without detailed derivation,
we will have a closer look at the outer beam because its compensation is more
Optical Systems
105
critical and its analysis will contain the central beam as a special case. Looking
at the outer beam, it is necessary to take into account the difference in beam path
that is illustrated in Fig. 5.14. This leads to an unfolded optical path as shown in
Fig. 5.15.
We directly consider the case for which we expect that compensation is achievable, namely, a = f , b = f . Keeping general values for a and b, as in Eq. (5.95),
would be useful when defocusing errors have also to be analyzed.
Separate steps of the ray transfer are considered separately first, and then
lumped together in order to facilitate the computation. The propagation from one
focal plane of the lens to the other focal plane is described by the following matrix:

G=
1 f
0 1
1
f1
0
1
1
0
f
1
0
f1
f
0
(5.98)
For the central part of the unfolded optical path, the square of this matrix will be
convenient, that is,

1 0
2
.
(5.99)
G =
0 1
This establishes a link to the retroreflector discussed in Sec. 2.1.
Figure 5.14 Dyson interferometer, outer beam.
Figure 5.15 Dyson interferometer, and the unfolded path of the outer beam.
Chapter 5
106
To incorporate the effect of the tilted mirror on the ray vector, the matrix C and
a corresponding error vector are defined for the upper side of the reflector, and the
matrix D and its corresponding error vector for the opposite side. They contain
the propagation by a distance of c or dy, respectively, the addition of the tilterror vector, and, again, a propagation by a distance of c or dy, respectively. Put
together, this gives
2c 2
y
0
y
1 c
1 c
+
=C
+
0 1
0 1
2

1 2c
(5.100)
with C =
0
1

and
D
2d 2
2

with D =
1 2d
0
1
(5.101)
Please note the different signs that follow from Fig. 5.14.
Now, a first step of the propagation can be written as follows:
(1)

in
0 f
y
y
=
.
1
(1)
in
f
(5.102)
The following ray vector is one changed by the tilted mirror:

(2)
(1)

y
y
2c 2
.
=C
+
(2)
(1)
2
Using Eq. (5.102), this gives
(2)
2c
f
y
=
(2)
f1
f
0
y in
in
2c 2
2
(5.103)
(5.104)
In order to describe the propagation through the lens, a reflection by a mirror, which
is assumed to be untilted, and an additional propagation back through the lens, the
use of the square of the matrix G is now convenient, i.e.,
(2)
(3)

y
y
1 0
=
.
(5.105)
(3)
0 1
(2)
To this result, the transformation involving the matrix D will then be applied as
(4)

(3)

y
1 2d
y
2d 2
.
(5.106)
=
+
(4)
(3)
2
0
1
After performing the multiplications, it follows that

(4)
2

(c d) f
y in
c 2
y
2c 2 2d 2
f
+
.
=
1
(4)
in
2
2
0
f
(5.107)
Optical Systems
107
This intermediate step is instructive because the compensation occurs here as

(4)
2
(c d) f
y in
2d 2
y
f
.
(5.108)
=
1
(4)
in
0
0
f
The term that is quadratic in will be neglected because it is beyond the scope
of the linear approximation used here. To transform to the output ray vector, the
matrix G is applied again as

y out
out
0
f1
f
0
y (4)
(4)
(5.109)
This can be written as follows in terms of the input vector:

out

1
0
y in
y
=
.
f22 (c d) 1
out
in
(5.110)
It follows from this equation that it is appropriate to adjust the measurement head
in a symmetric way so that the condition c = d is fulfilled. In this case, the tiltdependent entry in the system matrix is brought to zero.
5.5 Nested Ray Tracing

Khlers illumination system provides a typical example for nested ray tracing.
This illumination system is widely applied in optical microscopy. Other fields of
application are microcamera systems (Reynolds et al., 1989) and the illumination
units of systems used in optical lithography (Wong, 2001).
The optical arrangement is depicted in Fig. 5.16. It consists of a collector lens
and a condenser lens. A field stop is situated in the immediate vicinity of the
collector lens. The condenser images the field stop onto the target plane that has to
be illuminated. The collector lens, on the other hand, images the light source onto
the plane of the aperture stop.
Attributing the focal length f1 to the collector lens and the focal length f2 to
the condenser, the matrix chain can be written as follows:

S=
1
0

f2 + c
1
1
f11
0
1
1
f12
0
1
1 f1 + a
0
1
1 f2
0 1
.
1 f1 + b
0
1
(5.111)
The distances appearing in this equation are related by the imaging conditions imposed on the nested optical arrangement. To apply the imaging condition for the
collector, we have a look at the matrix A that describes the passage of a ray from
Chapter 5
108
Figure 5.16 Khlers illumination system.
the light source to the aperture stop, i.e.,

A=
1
0
f1 + b
1
1
f11
0
1
1
0
f1 + a
1
fb1
f11
f1
fa1
ab
f1

.
(5.112)
The imaging condition implies that a12 = 0. Therefore, we have

f12 = ab,
(5.113)
and can express one of the distances appearing in Eq. (5.112) in terms of the other
distance. In this way, the matrix A takes the form

0
fb1
A=
,
(5.114)
f11 fb1
and the system matrix changes accordingly to

f1 f2
1 bc
2
b
.
S = f1 f2 b
0
f1 f2
(5.115)
Now, the other condition has to be incorporated into the system matrix. To this end,
we consider the matrix B that is an expression for the transit of a ray from the field
stop to the target that has to be illuminated, i.e.,

1
0
1 f2 + c
1 f2
1 f1 + b
B=
f12 1
0
1
0
1
0 1

c
c
f2 f2 f2 (f1 + b)
=
.
(5.116)
f1f+b
f12
2
Specializing this matrix of a subsystem to the case where an imaging condition
holds between the plane of the field stop and the target plane is equivalent to setting
Optical Systems
109
b21 = 0. This implies

c=
f22
.
f1 + b
(5.117)
Using this in Eq. (5.114), the system matrix of a paraxial model for Khlers illumination system reads as

2
f1f+b
f1bf2
S=
.
(5.118)
b
0
f1 f2
We can now draw conclusions from these considerations that characterize the illumination scheme. First, we consider the impact of changes of the aperture stop. The
passage of a ray from the plane to the aperture stop to the target plane is represented
by the matrix

fc2 f2
1
0
1 f2 + c
1 f2
C=
.
(5.119)
=
f12 1
0
1
0 1
f12 0
Equation (5.117) might also be used at this point, and we have the mapping

out

2
f1f+b
f2
y ap
y
=
,
(5.120)
out
ap
f1
0
2
which implies that the angle out in the plane of the target is controlled by the
diameter y ap of the aperture stop.
The corresponding mapping for the field stop follows from combining
Eqs. (5.116) and (5.117), i.e.,

out

2
0
f1f+b
y field-stop
y
=
,
out
field-stop
f1f+b
f1
2
which shows that adapting the diameter y field-stop of the field stop directly changes
the field on the target plane.
Chapter 6
Outlook
The method of Gaussian matrices is useful for finding a paraxial solution to a given
optical-design problem. This solution can then be refined by analytical methods of
aberration theory or by optical-design software.
A generic way to proceed is as follows: a paraxial start layout is determined
using the matrix approach. Its properties and parameter dependencies can then be
analyzed in a first order of approximation with this method. The paraxial model
can be implemented as a start configuration in an optical-design program. The necessary paraxial element is included in most optical-design packages. Let us say
that we started with a lens in the thin-lens approximation to find and understand the
first layout. Then we can make the transition to a thick spherical lens with the same
focal length and extend the validity of the solution from the paraxial domain to the
complete aperture while trying to keep the main properties of the layout. The control features of the software can be used for this purpose. Depending on the degrees
of freedom available for the design task, we might improve the aberration control
in further steps of optimization by allowing for an aspheric lens. This method also
proved efficient in designing variable optical devices such as zoom lenses.
There are some approaches in the literature for extending the matrix theory
that were not treated in this book: To overcome the restriction to the paraxial region, trigonometric expressions as matrix entries are used and the concept of the
so-called exact matrices are introduced in Das (1991). The relation to Seidel aberrations is described by Guillemin and Sternberg (1984). And an extension of the
matrix method is also proposed in Brouwers book on instrument design (Brouwer,
1964).
In this book, the focus is on optical design, and a most natural step from the
practical point of view is to use the matrix description as groundwork (and, in a way,
as a framework) and then turn to aberration control for refinement of the design
obtained. But the theory of Gaussian matrices also lays a solid groundwork for
some in-depth theoretical studies of subjects such as Fourier optics and symplectic
systems (Guillemin and Sternberg, 1984).
The matrix description is considered a good starting point to dive into the theoretical description of light propagation through optical systems by integral transformations (Guillemin and Sternberg, 1984; Hodgson and Weber, 1997). The Collins
112
Outlook
integrals can be classified in a straightforward way by making reference to Gaussian

matrices (Hodgson and Weber, 1997).
A similar statement holds for Wigner distribution functions. Considerations on
ray matrices and on Wigner distribution functions on a high theoretical level can be
found in the book of Torre (2005).
In laser technology, extensive use is made of the theory of Gaussian matrices to
design optical resonators. In fact, many results in matrix optics stem from research
in that field.
A surprising application of tools similar to those presented in Chapter 3 for
sensitivity analysis and tolerancing is made by Duparr et al. (2005) and Schreiber
et al. (2005). They apply this mathematical description to analyze lens arrays.
There are many other interesting applications of the matrix method and related
paraxial methods to optical design and instrumentation. This text could only contain a part of them, and the reader will presumably discover new fascinating optical
results during her/his design work in the field of optics.
Bibliography
Altmann, J., Baumgart, R., and Weitkamp, C., Two-mirror multipass absorption cell,
Appl. Opt., 20, pp. 995999 (1981).
Antonsen, E.L., Burton, R.L., Spanjers, G.G., and Engelman, S.F., Herriott cell interferometry for millimeter-scale plasma measurements, Rev. Sci. Instrum., 74, pp. 8893
(2003a).
Antonsen, E.L., Burton, R.L., Spanjers, G.G., and Engelman, S.F., Herriott cell augmentation of a quadrature heterodyne interferometer, Rev. Sci. Instrum., 74, pp. 16091612
(2003b).
Back, F.G., and Lowen, H., The basic theory of varifocal lenses with linear movement and
optical compensation, J. Opt. Soc. Am., 44(9), pp. 684691 (1954).
Back, F.G., and Lowen, H., Generalized theory of Zoomar systems, J. Opt. Soc. Am.,
48(3), pp. 149153 (1958).
Baker, D., Field-widened interferometers for Fourier spectroscopy, Spectrometric Techniques, G.A. Vanasse (Ed.), pp. 71106, Academic Press, New York (1977).
Barr, J.R.M., Achromatic prism beam expanders, Opt. Commun., 51(1), pp. 129134
(1984).
Berek, M., Grundlagen der praktischen OptikAnalyse und Synthese optischer Systeme,
Walter de Gruyter, Berlin (1930) (reprint 1970).
Berge, B., and Peseux, J., Variable focal lens controlled by an external voltage: An application of electrowetting, Eur. Phys. J. E, 3, pp. 159163 (2000).
Besenmatter, W., Designing zoom lenses aided by the Delano diagram, Proc. SPIE, 237,
pp. 242255 (1980).
Born, M., OpticEin Lehrbuch der Electro-magnetischen Lichttheorie, p. 102, Springer,
Berlin (1933) (reprint 1985).
Born, M., and Wolf, E., Principles of OpticsElectromagnetic Theory of Propagation,
Interference and Diffraction of Light, pp. 225226, Pergamon Press, Oxford (1980).
Breckinridge, J.B., and Schindler, R.A., First-order optical design for Fourier spectrometers, Spectrometric Techniques, G.A. Vanasse (Ed.), Vol. II, Academic Press, New
York (1981).
Brouwer, W., Matrix Methods in Optical Instrument Design, Chap. 9, Benjamin, New York
(1964).
Casperson, L.W., Synthesis of Gaussian beam optical systems, Appl. Opt., 20, pp. 2243
2249 (1981).
Chang, C.-M., and Shieh, H.-P.D., Design of illumination and projection optics for projectors with single digital micromirror devices, Appl. Opt., 39, pp. 32023208 (2000).
Clark, A.D., Zoom Lenses, Adam Hilger, London (1973).
114
Bibliography
Das, P., Lasers and Optical Engineering, Springer-Verlag, Berlin, p. 302 (1991).
Delano, E., First-order design and the y, y diagram, Appl. Opt., 2, pp. 12511256 (1963).
Demtrder, W., LaserspektroskopieGrundlagen und Techniken, Chap. 4, SpringerVerlag, Berlin (1999).
DeWeerd, A.J., and Hill, S.E., Reflections on handedness, Phys. Teacher, 42, pp. 275
279 (2004).
Dickey, F.M., and Holswade, S.C., Laser Beam ShapingTheory and Techniques, Marcel
Dekker, New York (2000).
Duarte, F.J., and Piper, J.A., Dispersion theory of multiple-prism beam expanders for
pulsed dye laser, Opt. Commun., 43(5), pp. 303307 (1982).
Duarte, F.J., and Piper, J.A., Erratum, Opt. Commun., 45, p. 420 (1983).
Duarte, F.J., Note on achromatic multiple-beam expanders, Opt. Commun., 53(4), pp.
259262 (1985a).
Duarte, F.J., Erratum, Opt. Commun., 54, p. 388 (1985b).
Duparr, J., Schreiber, P., Matthes, A., Pshenay-Severin, E., Bruer, A., and Tnnermann, A., Microoptical telescope compound eye, Opt. Express, 13(3), pp. 889903
(2005).
Dyson, J., Unit magnification optical system without Seidel aberrations, J. Opt. Soc. Am.,
49, pp. 713716 (1959).
Dyson, J., Optics in a hostile environment, Appl. Opt., 7, pp. 569580 (1968).
Dyson, J., Interferometry as a Measuring Tool, pp. 9698, Machinery Publ., Brighton
(1979).
Fantone, S.D., Optical system with anamorphic prism, U.S., Patent No. 4,627,690 (1986).
Fantone, S.D., Anamorphic prism: A new type, Appl. Opt., 30(34), pp. 50085009
(1991).
Fouckhardt, H., PhotonikEine Einfhrung in die integrierte Optoelektronik und technische Optik, pp. 9495, Teubner, Stuttgart (1994).
Gerrard, A., and Burch, J.M., Introduction to Matrix Methods in Optics, Wiley, New York
(1975).
Goodman, J.W., Introduction to Fourier Optics, McGraw-Hill, New York (1968) (reissued
1988).
Goodman, D.S., General principles of geometric optics, Handbook of Optics, M. Bass,
E.W. Van Stryland, D.R. Williams, W.L. Wolfe (Eds.), Vol. I: Fundamentals, Techniques, and Design, pp. 1.11.109, McGraw-Hill, New York (1995).
Grenville, A., Hsieh, R.L., von Bnau, R., Lee, Y.-H., Markle, D.A., Owen, G., and
Pease, R.F.W., Markle-Dyson optics for 0.25 m lithography and beyond, J. Vac.
Technol. B, 9(6), pp. 31083112 (1991).
Grenville, A., Owen, G., and Pease, R.F.W., Image monitor for Markle-Dyson optics,
J. Vac. Sci. Technol. B, 11(6), pp. 27002704 (1993).
Grove, S.L., and Shuman, C.A., Anamorphic achromatic prism for optical disk heads,
U.S. Patent No. 5,155,633 (1992).
Guillemin, S., and Sternberg, S., Symplectic Techniques in Physics, Cambridge University
Press, Cambridge, England (1984).
Hanna, D.C., Krkkinen, P.A., and Wyatt, R., A simple beam expander for frequency
narrowing of dye lasers, Opt. Quant. Electron., 7, pp. 115119 (1975).
Heavens, O.S., and Ditchburn, R.W., Insight into Optics, Wiley, Chichester (1991).
Bibliography
115
Herriott, D., Kogelnik, H., and Kompfner, R., Off-axis paths in spherical mirror interferometers, Appl. Opt., 3, pp. 523526 (1964).
Herriott, D.R., and Schulte, H.J., Folded optical delay lines, Appl. Opt., 4, pp. 883889
(1965).
Herzberger, M., Gaussian optics and Gaussian brackets, J. Opt. Soc. Am., 33(12), pp.
651655 (1943).
Hodgson, N., Haase, T., Kostka, R., and Weber, H., Determination of laser beam parameters with the phase space beam analyzer, Opt. Quant. Electron., 24, pp. S927S949
(1992).
Hodgson, N., and Weber, H., Optical ResonatorsFundamentals, Advanced Concepts and
Applications, Springer-Verlag, Berlin (1997).
Hovde, D.C., Hodges, J.T., Scace, G.E., and Silver, J.A., Wavelength-modulation laser
hygrometer for ultrasensitive detection of water vapor in semiconductor gases, Appl.
Opt., 40, pp. 829839 (2001).
Hsieh, R.L., Lee, Y.Y.-H., Maluf, N.I., Browning, R., Jerabek, P., and Pease, R.F., Reflective masks for 1X deep ultraviolet lithography, Proc. SPIE, 1604, pp. 6777 (1992).
Ichie, K., Laser scanning optical system using axicon, Eur. Patent No. EP 0 627 643
(1994).
Johnson, R.B., and Feng, C., Mechanically compensated zoom lenses with a single moving element, Appl. Opt., 31, pp. 22742278 (1992).
Kay, D.B., and Gage, E.C., Heads and lasers, T.W. McDaniel, and R.H. Victora, Handbook of Magneto-Optical Data RecordingMaterials, Subsystems, Techniques, p. 62,
Noyes Publications, Westwood, NJ (1997).
Kloos, G., Entwurf und Auslegung optischer Reflektoren, Expert Verlag, Renningen (2007).
Lam, J.F., and Brown, W.F., Optical resonators with phase-conjugate mirrors, Opt. Lett.,
5, pp. 6163 (1980).
Latta, M.R., Strand, T.C., and Zavislan, J.M., Effect of track crossing on focus servo
signals: Feedthrough, Proc. SPIE, 1663, 157163 (1992).
Lipson, S.G., Lipson, H.S., and Tannhauser, D.S., Optik, Springer-Verlag, Berlin (1997).
Luecke, F.C., Achromatic anamorphic prism pair, U.S. Patent No. 5,596,456 (1997).
Magarill, S., First order property of illumination system, Proc. SPIE, 4768, pp. 5764
(2002).
Malacara, D., and Malacara, Z., Handbook of Lens Design, Marcel Dekker, New York
(1994).
Marchant, A.B., Method and apparatus for anamorphically shaping and deflecting electromagnetic beams, U.S. Patent No. 4,759,616 (1988).
Marchant, A.B., Optical RecordingA Technical Overview, Addison-Wesley, Reading,
MA (1990).
McLeod, J.H., The axicon: a new type of optical element, J. Opt. Soc. Am., 44(8),
pp. 592597 (1954).
McLeod, J.H., Axicons and their uses, J. Opt. Soc. Am., 50(2), pp. 166169 (1960).
Naumann, H., and Schrder, G., Bauelemente der OptikTaschenbuch der technischen
Optik, Carl Hanser Verlag, Mnchen, p. 187 (1992).
New, R.M.H., Owen, G., and Pease, R.F.W., Analytic optimization of Dyson optics, Appl.
Opt. 31, pp. 14441449 (1992).
Niefer, R.J., and Atkinson, J.B., The design of achromatic prism beam expanders for
pulsed dye lasers, Optics Commun. 67(2), pp. 139143 (1988).
116
Bibliography
Okuda, I., Takishima, S., Oono, M., Maruyama, K., and Noguchi, M., Optical system
for optical disc apparatus including anamorphic prisms, U.S. Patent No. 5,477,386
(1995).
Owen, G., Grenville, A., von Bnau, R., Jeong, H., Markle, D.A., and Pease, R.F., The effect of gaps in Markle-Dyson optics for sub-quarter-micron lithography, Jpn. J. Appl.
Phys. (Pt. 1), 32, pp. 58405844 (1993).
Pegis, R.J., and Peck, W.G., First-order design theory for linearly compensated zoom
systems, J. Opt. Soc. Am., 52(8), pp. 905911 (1962).
Prez, J.-P., OptiqueFondements et applications, Masson, Paris (1996).
Reynolds, G.O., DeVelis, J.B., Parrent, G.B., and Thompson, B.J., The New Physical Optics Notebook: Tutorials in Fourier Optics, p. 154, SPIE, Bellingham, WA, American
Institute of Physics, New York (1989).
Rioux, M., Tremblay, R., and Blanger, P.A., Linear, annular, and radial focusing with
axicons and applications to laser machining, Appl. Opt. 17, pp. 15321536 (1978).
Roberts, R.B., Absolute dilatometry using a polarization interferometer, J. Phys. E 8,
pp. 600602 (1975).
Saito, K., Sato, S., Shino, K., and Taniguchi, T., Device for magneto-optic signal detection
with a small crystal prism, Appl. Opt., 39(8), pp. 13151322 (2000).
Schreiber, P., Kudaev, S., Dannberg, P., and Zeitner, U.D., Homogeneous LEDillumination using microlens arrays, Proc. SPIE, 5942, pp. 59420K-159420K-9
(2005).
Sennaroglu, A., and Fujimoto, J.G., Design criteria for Herriott-type multi-pass cavities
for ultrashort pulse lasers, Opt. Express, 11(9), pp. 11061113 (2003).
Shack, R.V., Analytic system design with pencil and rulerThe advantages of the yy
diagram, Proc. SPIE, 39, pp. 127140 (1973).
Siegman, A.E., Lasers, University Science Books, Mill Valley, pp. 582583 (1986).
Steel, W.H., Interferometry, Cambridge University Press, Cambridge, England (1983).
Svelto, O., Principles of Lasers, Plenum Press, New York (1998).
Tanaka, K., Allgemeine Gausssche Theorie eines mechanisch kompensierten ZoomObjektives, Opt. Commun. 29(2), pp. 138140 (1979).
Tanaka, K., Paraxial analysis of mechanically compensated zoom lenses. 1: Fourcomponent type, Appl. Opt. 21(12), pp. 21742183 (1982).
Torre, A., Linear Ray and Wave Optics in Phase SpaceBridging Ray and Wave Optics in
the Wigner Phase-Space Picture, Elsevier, Amsterdam (2005).
Tovar, A.A., and Casperson, L.W., Generalized Sylvester theorems for periodic applications in matrix optics, J. Opt. Soc. Am. A, 12, pp. 578590 (1995).
Tyson, R.K., Principles of Adaptive Optics, Academic Press, New York (1998).
van den Eerenbeemd, J., Combining astigmatism of a plane parallel plate and a cylindrical
lens, ISOM 01International Symposium on Optical Memory 2001 Technical Digest,
Washington, D.C.: Optical Society of America. pp. 196197 (2001).
Wolfe, W.L., Nondispersive Prism, Handbook of Optics, M. Bass, E.W. Van Stryland,
D.R. Williams, W.L. Wolfe (Eds.), Vol. II: Devices, Measurements, and Properties,
pp. 4.14.29 McGraw-Hill, New York (1995).
Wong, A.K.-K., Resolution Enhancement Techniques in Optical Lithography, pp. 78,
SPIE, Bellingham, WA (2001).
Wooters, G., and Silvertooth, E.W., Optically compensated zoom lens, J. Opt. Soc. Am.,
55(4), pp. 347351 (1965).
Bibliography
117
Yoder, P.R., Non-image-forming optical components, Proc. SPIE, 531, pp. 206220
(1985).
Young, M., Optik, Laser, Wellenleiter, Springer-Verlag, Berlin (1997).
Yu, F.T.S., and Khoo, I.C., Principles of Optical Engineering, Wiley, New York (1990).
Zissis, G.J., Dispersive prisms and gratings, Handbook of Optics, M. Bass, E.W. Van
Stryland, D.R. Williams, and W.L. Wolfe (Eds.), Vol. II: Devices, Measurements, and
Properties, pp. 5.15.16, McGrawHill, New York (1995).
Index
4f arrangements, 79
A
absorption cell, 99
adaptive axicon, 48
adaptive feedback loop, 94
adaptive optics, 88
adjustment sensitivity, 99
afocal system, 11, 12, 20
afocal triplet, 77
anamorphic optics, 59
anamorphic thin lens, 65
angular magnification, 19
aperture stop, 16
autocollimator, 92, 94
axial misalignment, 56
axial ray, 16
axicon devices, 47
B
balance the optical beam paths, 99
beam pointing error, 53, 57
beam shaping, 33
Brewster condition, 33
C
cardinal elements, 7
cascading, 60, 67, 68
cats-eye retroreflector, 27, 56
central theorem of first-order ray
tracing, 14
chief rays, 80
collimating system, 12, 20
collimating triplet, 77
collimation lens, 72
Collins integrals, 112
compensated adaptive lens
arrangements, 87
compensation, 87, 103

cylindrical lens, 66
D
decentering, 53
decomposition of matrices, 13
defocusing, 53
Delano diagram, 19
determinant, 5
dispersive prisms, 32
divided optical path, 99
double-pass optics, 92
doubly telecentric lens, 81
Dyson interferometer, 103
Dyson optics, 82
E
eigenvalue theory, 96
error vector, 54
F
feedback loop, 91
field stop, 17
Fizeau interferometer, 99
focal length, 8
focusing condition, 12, 20
focusing triplet, 77
Fourier optics, 111
Fourier transform objectives, 79
G
Galilean telescope, 72
Gaussian brackets, 22
H
Herriott cell, 99
homogeneous coordinates, 54, 75
120
I
image field, 19
imaging, 10
imaging condition, 11, 12, 20, 73
imaging equation, 74, 78
imaging triplet, 77
integrating rod, 49
internal reflection, 45, 46
K
Keplerian telescope, 72
Khlers illumination system, 107
L
Lagrange invariant, 18
laser beam circularization, 33
laser resonators, 99
laser technology, 95
law of refraction, 1, 4
lens, 6
lens doublet, 12
light guide, 49
liquid lens, 90
M
magnification, 19
matrix representations, 5
mechanical compensation, 84, 85
Michelson interferometer, 101
microscope, 13
microscopy, 107
misadjustment, 53
mixing rod, 52
multiple-pass optics, 95
N
nested ray tracing, 107
nondispersive, 32
nonorthogonal anamorphic system, 60
O
optical compensation, 84, 85
optical delay lines, 99
optical resonators, 95
optics, 84
orthogonal anamorphic system, 60
P
partitioned matrix, 59
Petzval radius, 19
phase space, 20
phase-conjugate mirror, 26
phase-space analysis, 69
phase-space beam analyzer, 70
plane mirror, 25
plane-parallel plate, 29
principal planes, 8
principal ray, 17
prism for compression along one
axis, 38
prism for expansion along one axis, 36
prisms, 31
Q
quadrupole lens, 69
R
ray slope, 6
recursion formula, 23
reflection, 5, 25
refraction, 3, 29
refraction matrix, 21
retroreflector, 26
roof mirror, 28
rotation of an anamorphic component, 61
S
sensitivities, 53, 75
single-pass optics, 77
spatial filtering, 79
spherical mirror interferometers, 99
spherical surface, 4
stable, 99
Sylvesters theorem, 98
symplectic systems, 111
synthesis of optical systems, 13
system synthesis, 13
T
telecentric lenses, 80
telecentricity, 80
telescope, 13, 72
telescopes, 11
terrestrial telescope, 78
thin-prism approximation, 32
tilt, 53
tolerances, 53, 75
tolerancing, 39
121
transfer function, 35
translation matrix, 3, 21
trigonometric description of dispersive
prisms, 33
triple mirror, 52
triplet, 77, 84
tunnel diagram, 31
V
variable, 84
varifocal systems, 84
U
unstable, 99
Z
zoom lens, 84
W
Wigner distribution functions, 112
Dr. Gerhard Kloos is a physicist. He received his Ph.D. from Eindhoven University of Technology, where he worked on modulation laser interferometry and
electromechanics. He has published several scientific papers on optical technology,
materials science, and electromechanical effects, and a book on reflector design.
Dr. Kloos does research at an international enterprise that produces lighting and
sensor systems for the automotive industry. His design and development activities
currently focus on LED optics, projection systems, and adaptive optics. The main
areas of his research are polymer optics, actuators, and novel adaptive devices.

Matrix Methods For Optical Layout

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Matrix Methods For Optical Layout

Caricato da

Copyright:

Formati disponibili

Tutorial Texts Series

Matrix Methods for Optical Layout, Gerhard Kloos, Vol. TT77

Tutorial Texts in Optical Engineering

Bellingham, Washington USA

Library of Congress Cataloging-in-Publication Data

Copyright 2007 Society for Photo-optical Instrumentation Engineers

Introduction to the Series

Arthur R. Weeks, Jr.

1 An Introduction to Tools and Concepts

2.2.3 Axicon devices

3 Sensitivities and Tolerances

5.4.3 Dyson interferometer

An Introduction to Tools and

branch of optics is axiomatic like thermodynamics, for example. The framework

1.2 Basic Elements

An Introduction to Tools and Concepts

The change in angle is described by the law of refraction,

Substituting these equations into Eq. (1.10), we find

An Introduction to Tools and Concepts

Figure 1.3 The unfolding of a spherical mirror.

1.2.3 Reflection at a surface

1.3 Comparison of Matrix Representations Used in the

1.4 Building up a Lens

With these abbreviations, Eq. (1.23) then takes the form

An Introduction to Tools and Concepts

1.5 Cardinal Elements

This implies that

A straightforward way to obtain other cardinal elements is by direct comparison

a11 + a21 h1 a12 + a11 h1 + a22 h2 + a21 h1 h2

Its principal planes are at

An Introduction to Tools and Concepts

1.6 Using Matrices for Optical-Layout Purposes

An Introduction to Tools and Concepts

The setting of d = f1 + f2 solves this equation and we have

Figure 1.6 The significance of zero-matrix entries.

1.7 Lens Doublet

An Introduction to Tools and Concepts

1.8 Decomposition of Matrices and System Synthesis

Therefore, the condition

is contained implicitly in Eqs. (1.61) and (1.62). A generalization is possible and

1.9 Central Theorem of First-Order Ray Tracing

An Introduction to Tools and Concepts

1.10 Aperture Stop and Field Stop

Its start coordinates in the object plane are

An Introduction to Tools and Concepts

plane, by Q. Therefore, the system matrix is

Therefore, the input coordinates of the principal ray are given by

1.11 Lagrange Invariant

This is the Lagrange invariant.

An Introduction to Tools and Concepts

1.12 Petzval Radius

1.13 Delano Diagram

Figure 1.7 Delano diagram, four cases.

1.14 Phase Space

An Introduction to Tools and Concepts

1.15 An Alternative Paraxial Calculation Method

Setting uk+1 = uk , we can write the linear relation,

which is quite similar to the relation for the translation matrix T .

This relation corresponds to the refraction matrix R of the matrix description.

1.16 Gaussian Brackets

An Introduction to Tools and Concepts

with [ ] = 1. This implies

In cases, where it is useful to state inputoutput relations in a recursive way, the