Академический Документы
Профессиональный Документы
Культура Документы
org/terms
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 Müller, 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 Optics–Design, Fabrication, and Test, Donald C. O’Shea, 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 für 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 Steinbrüchel 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-Óptica 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
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
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.
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.
2 Optical Components 25
2.1 Components Based on Reflection 25
2.1.1 Plane mirror 25
2.1.2 Retroreflector 26
2.1.3 Phase-conjugate mirror 26
2.1.4 Cat’s-eye retroreflector 27
2.1.5 Roof mirror 28
2.2 Components Based on Refraction 29
2.2.1 Plane-parallel plate 29
2.2.2 Prisms 31
4 Anamorphic Optics 59
4.1 Two Alternative Matrix Representations 59
4.2 Orthogonal and Nonorthogonal Anamorphic Descriptions 60
4.3 Cascading 60
4.4 Rotation of an Anamorphic Component with Respect to the
Optical Axis 61
4.4.1 Rotation of an “orthogonal” system 62
4.4.2 Rotation of a “nonorthogonal” system 65
4.5 Examples 65
4.5.1 Rotated anamorphic thin lens 65
4.5.2 Rotated thin cylindrical lens 66
4.5.3 Cascading two rotated thin cylindrical lenses 67
4.5.4 Cascading two rotated thin anamorphic lenses 68
4.5.5 “Quadrupole” lens 69
4.5.6 Telescope built by cylindrical lenses 72
4.5.7 Anamorphic collimation lens 72
4.6 Imaging Condition 73
4.7 Incorporating Sensitivities and Tolerances in the Analysis 75
5 Optical Systems 77
5.1 Single-Pass Optics 77
5.1.1 Triplet synthesis 77
5.1.2 Fourier transform objectives and 4f arrangements 79
5.1.3 Telecentric lenses 80
5.1.4 Concatenated matrices for systems of n lenses 81
5.1.5 Dyson optics 82
5.1.6 Variable single-pass optics 84
5.2 Double-Pass Optics 92
5.2.1 Autocollimator 92
5.3 Multiple-Pass Optics 95
5.4 Systems with a Divided Optical Path 99
5.4.1 Fizeau interferometer 99
5.4.2 Michelson interferometer 101
6 Outlook 111
Bibliography 113
Index 119
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
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 approxi-
mated description would also allow for cascading subsystems to describe a com-
pound 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
The height in the second reference plane depends on the distance traveled and on
the starting angle,
y (2) = y (1) + t tan β (1) . (1.4)
Figure 1.1 Propagation in a homogeneous medium. The two reference planes are at a
distance t.
Under the assumption that the paraxial approximation is valid, i.e., for small an-
gles β (1) , we can linearize the trigonometric function in Eq. (1.4) as
y (2) = y (1) + tβ (1) . (1.5)
Equations (1.3) and (1.5) can now be combined and written as a matrix relation,
(2)
(1)
y 1 t y
(2)
= . (1.6)
β 0 1 β (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
T = . (1.7)
0 1
Figure 1.2 Refraction at a spherical surface. The spherical surface separates two media
with refractive indices n1 and n2 .
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) , (1.9)
(1) (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 β (1) and β (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) = β (1) + α, (1.11)
i (2)
=β (2)
+ α. (1.12)
Substituting these equations into Eq. (1.10), we find
n1 n1 − n2
β (2) = β (1) + α. (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 α appearing in
Eq. (1.13) as
y (1)
tan α ∼ = . (1.14)
R
Linearizing the trigonometric function for small angles (tan α ∼ = α), Eq. (1.14) is
substituted into Eq. (1.13) and we have
n1 n1 − n2 (1)
β (2) = β (1) + y . (1.15)
n2 n2 R
This is the linearized input–output 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
(2)
= n1 −n2 n1 . (1.16)
β n2 R n2
β (1)
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
1 0
S= . (1.18)
− R2 −1
The second coordinate nβ can also be introduced as a modified ray slope (Sieg-
man, 1986) as
dr(z)
r (z) ∼
= n(z) . (1.20)
dz
The interpretation of this coordinate in terms of slope can be fruitful in some cir-
cumstances.
1 0 1 t 1 0
S = n2 −n3 n2 n1 −n2 n1 , (1.22)
n3 R2 n3
0 1 n2 R1 n2
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
1 − n−1 t
R1 n
t
n
S= . (1.23)
− n−1
R1
− 1−n
R2
+ n−1 1−n t
R1 R2 n
1 − 1−n t
R2 n
This might suggest the following abbreviations:
n−1
P1 = , (1.24)
R1
1−n
P2 = . (1.25)
R2
1 − P1 nt t
S= n
. (1.26)
−P1 − P2 + P1 P2 nt 1 − P2 nt
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
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
0 1 b a11 a12 y
= , (1.29)
β out 0 1 a21 a22 0
or
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:
a22
a=− . (1.35)
a21
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
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:
1 h2 a11 a12 1 h1 1 0
= , (1.36)
0 1 a21 a22 0 1 − f1 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 left-
hand side, we have
1 n − 1 1 − n (n − 1)(1 − n)t
= + − . (1.40)
f R1 R2 nR1 R2
Its principal planes are at
n−1 t
h1 = −f , (1.41)
R1 n
1−n t
h2 = −f . (1.42)
R2 n
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 , respec-
tively.
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.
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 β 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 input–output
relation is out
in
y s11 s12 y
out
= . (1.45)
β s21 s22 β in
To ensure that β out has a single value for a given ray height y in , it has to be inde-
pendent of β in . A look at the input–output 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 β 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 b 1 g
with B = 0 1 and G = 0 1 . Considering the input–output relation again, we
find that y out is independent of β in if
s12 = 0. (1.48)
We can apply this condition immediately to find the imaging relation for a thin
lens. The corresponding matrix chain is
1 b 1 0 1 g 1 − fb g + b − bg f
S= = . (1.49)
0 1 − f1 1 0 1 − f1 1 − fg
Using s12 = 0, we have the well-known imaging condition
1 1 1
+ = , (1.50)
g b f
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 par-
allel 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 input–output
relation for S, we see that setting
s21 = 0 (1.51)
ensures that β 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 approxima-
tion, 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
1 0 1 d 1 0 1 − fd1 d
S= = . (1.52)
− f12 1 0 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. (1.53)
f1 f2 f1 f2
The setting of d = f1 + f2 solves this equation and we have
f2
− f1 f1 + f2
S= f1
(1.54)
0 f2
for the system matrix of the telescopic arrangement. It represents a Newtonian tele-
scope if both focal lengths are positive. If the focal length of the first lens is nega-
tive, 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 − fd1 d
S= . (1.55)
− f11 − f12 + f1df2 1 − fd2
The term s21 is related to the focal length of the doublet (measured with respect to
its second principal plane).
1 1 1 d
= + − . (1.56)
f f1 f2 f1 f2
[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
f1 f 2
f =− . (1.58)
E
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
(Pérez, 1996)
s11 0 1 0 s11 0
S= = , (1.65)
0 s22 − f1 1 s11
s22
f
s11 0 s11 0 1 0
S= = s22 . (1.66)
0 s22 f
s22 − f1 1
Depending on the application, the matrices appearing in the product can then be
further decomposed by applying the same set of rules.
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)
βa βaout
out
yb yb
→ . (1.68)
βb βbout
ya ya
out
=A , (1.69)
βa βa
out
yb yb
out
=A . (1.70)
βb βb
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 βa 0 0 a11 ya
⎢y β 0 0 ⎥ ⎢ a12 ⎥ ⎢ ybout ⎥
⎥ ⎢ ⎥ ⎢
⎢ b b ⎥
⎢ ⎥⎢ ⎥=⎢ ⎥. (1.71)
⎣ 0 0 ya βa ⎦ ⎣ a21 ⎦ ⎣ βaout ⎦
0 0 yb βb a22 βbout
Because the matrix is partitioned, two sets of linear equations can be solved inde-
pendently. If the determinant
ya βa
D = det = 0, (1.72)
yb βb
D = 0 is equivalent to the condition that the input ray vectors are linearly in-
dependent. 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 input–output 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.
y 0
= P
(k)
. (1.77)
β (k) β (k)
This implies that
y (k)
β (k)
= (k)
. (1.78)
p12
The aperture stop is at the position z(k) for which β (k) takes the minimum value of
all the candidates. It has the height yas = y (k) if (k) is the label for that minimum.
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)
β β
Its start coordinates in the object plane are
in
y 0
= yas ,
β 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
S = QP , (1.80)
and the coordinates of the axial ray as its leaves the system are
out
s
y yas 12
out
= . (1.81)
β p12 s22
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)
β β
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
β=− y . (1.83)
p12
Therefore, the input coordinates of the principal ray are given by
in
y y field
= , (1.84)
β in − pp11
12
y field
and the output coordinates after passage through the whole system are
out
y 1
=y Sfield
. (1.85)
β out − pp11
12
It is interesting to note the following symmetry that exists between the axial ray and
the principal ray:
0 y
P = as for the axial ray, (1.86)
β β
field
y 0
= P −1 for the principal ray, (1.87)
β β
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 ex-
tension 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.)
(yaout βb − ybout βa )(ya βbout − yb βaout ) − (βaout βb − βbout βa )(ya ybout − yb yaout ) n1
= .
(ya βb − yb βa )2 n2
(1.90)
L = −n(ya βb − βa yb ). (1.93)
In both these planes, the height of the axial ray is equal to zero. Therefore, the
above equation reduces to
n−1 1−n
P1 = , P2 = .
R1 R2
We state the relation for the Petzval radius Rp of a system made up of N refractive
surfaces, without proof, as
1 ∼ Dk
N
= , (1.96)
Rp n
k=1 k
where Dk = nk /Rk is the refractive power of the kth surface, nk is the dif-
ference 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 corre-
sponds 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 quan-
tity is Dyson’s system (Dyson, 1959). This optical arrangement will be considered
in more detail in Section 5.1.
0 0 tβ0
→ + , (1.97)
β0 β0 0
y0 y0 tβ0
→ + , (1.98)
β0 β0 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.
Figure 1.8 Effect of the translation matrix in phase space. The spatial coordinate is repre-
sented along the y-axis and the angular coordinate is represented along the β-axis.
Figure 1.9 Effect of the refraction matrix in phase space. As in Fig. 1.8, the spatial coor-
dinate is represented along the y-axis and the angular coordinate is represented along the
β-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 conserva-
tion is a consequence of Eq. (1.19).
From the theoretical point of view, it is more convenient to use coordinates
(y, nβ). The corresponding “volume” (or “surface”) is then always conserved. The
phase-space approach is common in laser technology (Hodgson and Weber, 1997).
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:
where uk+1 and uk are angles. We can combine these equations to obtain
hk+1 hk
= − ek . (1.104)
uk uk
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:
hk nk nk
uk = 1 − + uk . (1.107)
Rk nk nk
hk 1 0 hk
= nk −nk . (1.108)
uk
nk
Rk nk nk uk
the system, and that is described by the product matrix, as a function of the in-
put 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 ] = 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 Herzberger’s article, we take a description of a lens using his symbols, but
arrange the linear transformation as a matrix as follows:
x2 1 − nd12 φ1 − nd12 x1
= 12 12
. (1.114)
ξ2 φ1 + φ 2 − n φ1 φ2 1 − n φ2
d12 d12
12
ξ01
12
where x is the distance between intersections of the optical axis and the ray with
the reference plane and ξ is the inclination angle of the ray. The subscripts indicate
the corresponding surfaces. φk 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:
φ , − d12 d12
− n12
x2 1 n12 x1
= . (1.115)
ξ2 φ , − d12 , φ − d12 , φ ξ01
1 n12 2 n12 2
For the general case, the recursion law reads as follows (Herzberger, 1943):
φ1 , − d12 , . . . , − dk−1,k − d12 , φ2 , . . . , − dk−1,k
This equation establishes a link between the input coordinates and the output coor-
dinates 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 ex-
plicitly 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 proce-
dure. 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.
1 0
A= . (2.1)
0 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.
2.1.2 Retroreflector
Unlike a plane mirror, a retroreflector redirects the beam back into the same direc-
tion 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
A= . (2.2)
0 −1
−1 a12
A= . (2.3)
0 −1
1 0
A= . (2.4)
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
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.
1 g 1 0 1 b 1 0
S=
0 1 − f1 1 0 1 0 1
1 b 1 0 1 g
× . (2.5)
0 1 − f1 1 0 1
−1 2(f − g)
S(g = 0, b = f ) = (2.7)
0 −1
and
−1 0
S(g = f, b = f ) = . (2.8)
0 −1
The form of the second matrix is equal to the matrix of a retroreflector stated in
Eq. (2.2).
1 2t
T = . (2.9)
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 0 1 2t −1 −2t
S= = . (2.11)
0 −1 0 1 0 −1
Therefore, a roof reflector acts like a retroreflector [Eq. (2.3)] in the plane that is
perpendicular to the roof edge.
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
S= . (2.12)
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
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
y 1 0 y
= 1 (2.16)
β (1)
0 n β in
for the rays entering the plate and
out
y 1 0 y (2)
out = (2.17)
β 0 n β (2)
for those leaving it. In between, they pass through a homogeneous medium of
refractive index n. Putting this together, we have
1 0 1 t 1 0 1 nt
S= = . (2.18)
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:
out,1
in
in
y 1 z1 1 t y 1 t + z1 y
= = . (2.19)
β out,1 0 1 0 1 β in 0 1 β in
At the intersection with the optical axis, we have y out,1 = 0. This implies
y in
z1 = − − t. (2.20)
β in
y 1 z2 1 nt y in 1 nt + z2 y in
= = . (2.21)
β out,2 0 1 0 1 β in 0 1 β in
2.2.2 Prisms
Prisms have a vast range of applications in several fields of optics. In optical spec-
troscopy, for example, an important branch of science, prisms serve both as disper-
sive means (Demtröder, 1999) and as components for beam shaping. In optical data
storage technology, prisms are also used for beam shaping purposes, namely, to cir-
cularize the asymmetric beam emitted by a semiconductor laser
(Marchant, 1990).
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 α and the
incidence angle measured against the normal of the first face of the prism are small,
Snell’s law can be linearized and the approximated expression for the deviation
angle δ is as follows (Heavens and Ditchburn, 1991):
δ = (n − 1)α. (2.24)
It is interesting to observe that the angle of incidence does not appear in this ap-
proximated expression.
In terms of the matrix method, the thin prism can be written in the following
way:
y 1 0 y 0
= + . (2.25)
β 0 1 β −(n − 1)α
We will apply this later to get an approximated expression for an axicon.
where φ1 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 expres-
sion of the Brewster condition in terms of the angle φ1 :
φ1 = arccot(n). (2.28)
φ1 + φ2 = α. (2.31)
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., φ3 = 0. Using the trigonometric identity
arcsin x = 90 deg − arccos x, Eq. (2.32) can be reformulated as follows:
φ3 = arcsin n sin(α − 90 deg + φ0 ) . (2.34)
For φ3 = 0, this equation implies that
α − 90 deg + φ0 = 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 φ0 . The angle φ1 , with respect to this normal inside a medium of refractive index
n is, determined by Snell’s law as
1
φ1 = arcsin sin φ0 . (2.36)
n
If φ0 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 φ0 = , (2.37)
h
w1
cos φ1 = . (2.38)
h
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 φ1
= . (2.39)
w0 cos φ0
Together with Snell’s law [Eq. (2.36)], this leads to the beamwidth transfer function
for a refracting surface,
w1 cos{arcsin[(1/n) sin φ0 ]}
= . (2.40)
w0 cos φ0
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 φ2 )]
= . (2.41)
w1 cos φ2
In a prism of apex angle α, the link between the known angle φ1 and the angle φ2
is given by α = φ1 + φ2 .
In this way, one finds
w3 cos[arcsin(n · sin{α − arcsin[(1/n) sin φ0 ]})]
= . (2.42)
w1 cos{α − arcsin[(1/n) sin φ0 ]}
Substituting for w1 from Eq. (2.40), the beamwidth transfer function for a prism
with the apex angle α is obtained as
w3 cos[arcsin(n · sin{α − arcsin[(1/n) sin φ0 ]})] cos{arcsin[(1/n) sin φ0 ]}
= .
w0 cos{α − arcsin[(1/n) sin φ0 ]} cos φ0
(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.
A prism for expansion along one axis 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 φ0
= n. (2.44)
cos φ0
To avoid reshaping at the other surface of the prism, i.e., compression in the direc-
tion of the axis that has been expanded before, it is convenient to have the beam
exiting the prism perpendicular to its rear surface (φ3 = 0). Combining this condi-
tion [Eq. (2.35)] with the Brewster condition leads to the equation,
This implies that the following relation between the apex angle of the prism and its
refractive index should hold:
α = 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 α − arcsin(sin φ0 /n) first. Using Snell’s law and the trigonometric
identity arcsin x = 90◦ − arccos x, one has
sin φ0
α − arcsin = α − 90 deg + φ0 . (2.48)
n
Combining this equation with Eq. (2.46) and the Brewster condition, it follows that
sin φ0
α − arcsin = 0. (2.49)
n
This implies that the first factor in Eq. (2.43) is equal to 1. Using Snell’s law and
the trigonometric identity stated before, the second factor can be expressed as
If this is combined with the Brewster condition (tan φ0 = n), one finds that the
second factor is equal to the refractive index n.
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
φ3 = 0 ⇒ = 1. (2.51)
w1
In this case, it is sufficient to consider
w1 cos φ1
= . (2.52)
w0 cos φ0
Using the Brewster condition (1/ cos φ0 = n/ sin φ0 ), it is concluded that
w1 cos φ1
= n. (2.53)
w0 sin φ0
The sine term can now be expressed by Snell’s law (sin φ0 = n sin φ1 ) as
w1 cos φ1
= = cot φ1 . (2.54)
w0 sin φ1
At this point, it is convenient to make use of the alternative expression of the Brew-
ster condition (cos φ1 = 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 φ0
M= . (2.55)
n 1 − sin2 φ0
Its equivalence with Eq. (2.52) can directly be seen using Snell’s law (sin φ0 =
n sin φ1 ).
A prism for compression along one axis An analogous relation holds for a prism
that reduces the beamwidth along an axis instead of expanding it, namely,
w3 1
= . (2.56)
w0 n
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 φ2 = sin φ3 ).
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 dis-
adjustment 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 opti-
mum 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
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 α 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).
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.
Relation of incident and exit angle in the case of a two-prism arrangement Hav-
ing a look at the result for a single prism, the following equations can be written
immediately:
sin φ4
φ7 = arcsin n sin α − arcsin
II II
, (2.60)
nII
sin φ0
φ3 = arcsin n sin α − arcsin
I I
, (2.61)
nI
where the angles are designated as depicted in Fig. 2.22 and α II is the apex angle
of the second prism and nII is its index of refraction, while the variables α I and nI
describe the first prism. In the equation for φ4 , the relative orientation of the two
The transfer function for the beamwidth altered by two prisms Exploiting the fact
stated earlier that the beamwidth transfer function can be composed as the prod-
uct of the transfer function of the refracting surfaces that intervene, the transfer
function for the two prisms can now be written directly as
w7 cos(arcsin{nII · sin[α II − arcsin(sin φ4 /nII )]}) cos[arcsin(sin φ4 /nII )]
=
w0 cos[α II − arcsin( sinnIIφ4 )] cos φ4
cos(arcsin{nI · sin[α I − arcsin(sin φ0 /nI )]}) cos[arcsin(sin φ0 /nI )]
× .
cos[α I − arcsin(sin φ0 /n)] cos φ0
(2.64)
The link between the two prisms is again established via the following equation:
sin φ0
φ4 = arcsin n sin α − arcsin
I I
− β. (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, φ0 = arctan nI ,
α I = arccot nI , (2.66)
φ4 = −β,
Second prism, −β = arctan nII ,
α II = arccot nII . (2.67)
Under these conditions, the beam leaves the last prism perpendicular to its second
surface, i.e., φ7 = 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.
Of course, it is also possible to arrange two prisms in such a way that the in-
cident beam is perpendicular to each of the first surfaces of the prisms to realize
compressive beam shaping. If α I = arccot(nI ) and α II = arccot(nII ) hold, one
finds
w7 1
= I II . (2.69)
w0 nn
Tolerancing The formulas presented here [Eqs. (2.63)–(2.65)] allow for the in-
vestigation 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.
Figure 2.25 Tunnel diagram of the prism with two internal reflections.
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 α 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 1 0 y 0
= + , with δ = (n − 1)α. (2.82)
β (1) 0 1 β in −δ
This ray then propagates through air,
(2)
(1)
in
y 1 t y 1 t y −tδ
= = + . (2.83)
β (2) 0 1 β (1) 0 1 β in −δ
The condition that the incident light is collimated (β in = 0) implies that
y (2) = y in − tδ. (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
L= = . (2.85)
δ (n − 1)α
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 (Ri-
oux et al., 1978; Ichie, 1994). Here, an arrangement will be considered that de-
creases 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:
tan(β − α)
H =d , with β = arcsin(n sin α) and α = 90 deg − γ .
1 − [tan(β − α)/ tan γ ]
(2.86)
Figure 2.27 Adaptive axicon. The adaptive axicon is characterized by the angle γ . The two
components that form the adaptive axicon are separated at a distance d.
H ∼
= d tan(β − α). (2.87)
Now, we use the fact that α = Θ, where Θ is the angle of the axicon depicted in
Fig. 2.27, and linearize the equation as
H (d; n, Θ) ∼
= d(nΘ − Θ) = d(n − 1)Θ. (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 Θ = 90 deg − γ .
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 perpen-
dicular to the entrance surface for the first time. From the corresponding triangle,
we see that this propagation can be described using the matrix
1 t (y, β) d −y
B= , where t (y, β) = . (2.91)
0 1 tan β
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
C= . (2.92)
0 1
Proceeding further, either case 1 or case 2 can occur, and the corresponding matri-
ces are
−+ −1 0
D = for case 1, (2.93)
0 1
+− 1 0
D = for case 2. (2.94)
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
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 = . (2.97)
0 n
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, β) div s where s = . (2.98)
tan ϕ
The overall length of the rod is
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, β) − trest (y, β)]
S = F EDCBA = (−1)M n
, (2.100)
0 −1
with trest (y, β) = T − t (y, β) − Ms, t (y, β) = (d − y)/ tan β, and M and s as in
Eq. (2.98).
This system matrix is quite different from the matrices encountered earlier in-
sofar 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.
y 1 0 y 0
= +γ . (3.1)
β (2) 0 nn12 β (1) 1 − nn12
Figure 3.1 Tilted surface. The surface that separates the two media with refractive indices
n1 and n2 is tilted by an angle γ .
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 trans-
formation to the coordinate system of the misaligned element is presented in the
literature on laser technology (Siegman, 1986). Figure 3.2 shows an optical ele-
ment that is decentered and tilted. Decentering and tilt are described by the error
vector
y1
1 = (3.2)
β1
with respect to the optical axis. In the coordinate system of the misaligned ele-
ment, 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)
β (2) a21 a22 β (1) β
is expressed as
⎡ ⎤ ⎡ ⎤ ⎡ (1) ⎤
y (2) a11 a12 y y
⎣ β (2) ⎦ = ⎣ a21 a22 β ⎦ ⎣ β (1) ⎦ . (3.4)
1 0 0 1 1
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.
r A b r
= T , (3.5)
1 0 1 1
We observe that the error vector b of a system with N subsystems is formed ac-
cording to the rule
This implies that, for the general case of N 3 × 3 matrices, the product matrix can
be expressed by the following formula (Kloos, 2007):
⎡ ⎤
1
N−1 k+1
A b ⎢ A (k)
( (l)
A )b (k) +b (N ) ⎥
0 T =⎣ ⎦, (3.8)
1 k=N k=1 l=N
0 T 1
with 1k=N A(k) = A(N ) · · · · · A
(1)
.
Please note that the symbol is used here to designate a matrix multiplication
(and not a multiplication of scalar quantities).
1 z
. (3.9)
0 1
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
1 z 0 a11
A =A+ z. (3.10)
0 1 0 a21
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)
0 a 11 z
S = A(N ) · · · · · A(k+1) (k)
A(k−1) · · · · · A(1) . (3.11)
0 a21 z
mirror is well positioned at a distance f with the respect to the lens, the system
matrix is
−1 2f
S= (3.12)
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
f2
− f1
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 dis-
tance f in front of the lens. In this case, the results for the system matrix and the
error matrix of the cat’s-eye retroreflector appear as follows:
−1 0 0 0
S= , S = 2z 1 . (3.14)
0 −1 f2
0
y
,
β in
where β 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
= β in
. (3.15)
β out s22
One would like to have a collimated beam in the y–z plane and a high divergence
in the x–z 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.
The matrix given by Eq. (4.3), in which the block matrices R and S vanish, de-
scribes what might be called an orthogonal anamorphic system. In the more general
case of a nonorthogonal anamorphic system, there is coupling between the orthog-
onal 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 nonorthog-
onal anamorphic description. Another example for a component with nonvanish-
ing 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
M II R II MI RI M II M I + R II S I M II R I + R II N I
= . (4.4)
S II N II SI NI S II M I + N II S I S II R I + N II N I
⎡ ⎤ ⎡ ⎤
x out II I
x in
⎢ α out ⎥ M II R I ⎢ in ⎥
⎢ ⎥= M M ⎢ α ⎥. (4.5)
⎣ y out ⎦ N II S I II I
N N ⎣ y in ⎦
out
β β in
For the case where an orthogonal anamorphic description is appropriate for subsys-
tem I, it follows that
⎡ out ⎤ ⎡ in ⎤
x
x
⎢ α out ⎥ II I
M M R N II I ⎢ in ⎥
⎢ ⎥ ⎢ α ⎥
⎣ y out ⎦ = S II M I N II N I ⎣ y in ⎦ . (4.6)
β out β 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 II I
x in
⎢ α out ⎥ 0 ⎢ α ⎥
⎢ ⎥= M M ⎢ ⎥ (4.7)
⎣ y ⎦out
0 II
N N I ⎣ y in ⎦ .
β out β in
The matrix of a compound system described by representation B also follows by
multiplication of block matrices,
II
I
II I
A B II A BI A A + B II C I AII B I + B II D I
= . (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 representa-
tion B, the symmetry expressed by Eq. (4.7) is less apparent in this matrix repre-
sentation. But this representation has other advantages.
cos ϕ 0
Cϕ = , (4.11)
0 cos ϕ
sin ϕ 0
Sϕ = . (4.12)
0 sin ϕ
β out β in
where X and Y are quadratic submatrices. They form the system matrix E as
X 0
E= . (4.15)
0 Y
This matrix is transformed according to the following transformation law into the
new, angle-dependent system matrix E (ϕ):
where TϕT is the transposed matrix of Tϕ . The transposed matrix can be formed
easily in the block-matrix notation and we find
Cϕ −Sϕ X 0 Cϕ Sϕ
E (ϕ) = . (4.18)
Sϕ Cϕ 0 Y −Sϕ Cϕ
Cϕ XCϕ + Sϕ Y Sϕ Cϕ XSϕ − Sϕ Y Cϕ
E (ϕ) = . (4.19)
Sϕ XCϕ − Cϕ Y Sϕ Sϕ XSϕ + Cϕ Y Cϕ
By definition, Cϕ and Sϕ are diagonal matrices. The above equation can therefore
be simplied as
Cϕ2 X + Sϕ2 Y Sϕ Cϕ (X − Y )
E (ϕ) = . (4.20)
Sϕ Cϕ (X − Y ) Sϕ2 X + Cϕ2 Y
The upper-right and lower-left 2 × 2 block matrices are equal. The matrix differ-
ence 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:
Representation B
After abbreviating the rotation matrix as
cos ϕ sin ϕ
R= , (4.24)
− sin ϕ cos ϕ
R 0
T = . (4.26)
0 R
A B
E= . (4.29)
C D
The transformation law can directly be applied to this general matrix as
T
R 0 A B R 0
E (ϕ) = . (4.30)
0 RT C D 0 R
It follows that
R T AR R T BR
E (ϕ) = . (4.31)
R T CR R T DR
The calculation of the matrix entries in the special case of an orthogonal anamor-
phic system is straightforward, i.e.,
X U
E= , (4.35)
V Y
wherein U and V are nonvanishing block matrices. As before, the transformation
formula is applied, in which ϕ appears as the parameter for the rotation angle,
After multiplying the matrices and using the diagonality of the block matrices Cϕ
and Sϕ , we find
Cϕ2 X − Cϕ Sϕ (U + V ) + Sϕ2 Y Cϕ Sϕ (X − Y ) + Cϕ2 U − Sϕ2 V
E (ϕ) = .
Cϕ Sϕ (X − Y ) − Sϕ2 U + Cϕ2 V Sϕ2 X + Cϕ Sϕ (U + V ) + Cϕ2 Y
(4.37)
This can be decomposed into two matrices, F (ϕ) and G (ϕ),
with
Cϕ2 X + Sϕ2 Y Sϕ Cϕ (X − Y )
F (ϕ) = , (4.39)
Sϕ Cϕ (X − Y ) Sϕ2 X + Cϕ2 Y
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 that—apart
from the angle-dependent matrices—the matrix F (ϕ) only contains the matrices X
and Y , whereas the matrix G (ϕ) 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 (ϕ) of an anamorphic lens that has been rotated
by an angle ϕ 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.
Using matrix representation B, the system matrix of the unrotated thin lens can
then be stated as follows:
⎡ ⎤
1 0 0 0
⎢
A B ⎢ 0 1 0 0 ⎥ ⎥
E= =⎢ −1 0 1 0 ⎥ . (4.41)
C D ⎣ fx ⎦
0 − f1y 0 1
R T AR R T BR I O
E (ϕ) = = , (4.42)
R T CR R T DR T
R CR I
where
1 0 0 0
I= and O= .
0 1 0 0
Therefore, the system matrix of the anamorphic thin lens rotated by an angle ϕ
about the optical axis is
⎡ ⎤
1 0 0 0
⎢ 0 1 0 0 ⎥
⎢ ⎥
E (ϕ) = ⎢ cos2 ϕ
− fx − sinfy ϕ
2
sin ϕ cos ϕ − f1x + 1
1 0 ⎥ . (4.43)
⎣ fy ⎦
2 cos2 ϕ
sin ϕ cos ϕ − f1x + 1
fy
− sinfx ϕ − fy
0 1
I O
F1 = , (4.46)
P1 (f1 , α) I
where
1 0 0 0
I= , O= , and
0 1 0 0
2
− cosf1 α − sin αfcos α
P1 (f1 , α) = 1
2 .
− sin αfcos
1
α
− sinf1 α
The spacing between both cylindrical lenses is taken into account by using the
translation matrix D,
I dI
D= . (4.47)
O I
The second cylindrical lens has the block matrix F2 ,
I O
F2 = , (4.48)
P2 (f2 , β) I
with 2
− cosf2 β − sin βfcos β
P2 (f2 , β) = 2
2 .
− sin βfcos
2
β
− sinf2 β
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 multi-
plications instead of operating on the matrix entries of P1 or P2 , i.e.,
I O I dI I O
S= . (4.50)
P2 (f2 , β) I O I P1 (f1 , α) I
I + dP1 (f1 , α) dI
S= .
P1 (f1 , α) + P2 (f2 , β) + dP1 (f1 , α)P2 (f2 , β) I + dP2 (f2 , β)
(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 equa-
tions, Eq. (4.51) allows us to get a convenient overview of the five-dimensional
parameter space of this anamorphic optical setup.
The submatrix for the second anamorphic lens has the analogous form
⎡ 2 2 ⎤
x y − cosf x β − sinf y β sin β cos β − f1x + f1y
P2 f2 , f2 , β = ⎣ 2
2
2
2
2
2 ⎦ . (4.53)
sin β cos β − f1x + f1y − sinf x β − cosf y β
2 2 2 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 . (4.54)
In block-matrix notation, this equation reads as
I O I dI I O
S= y y . (4.55)
P2 (f2x , f2 , β) I O I P1 (f1x , f1 , α) I
Finally, we have for the system matrix of the two rotated anamorphic lenses,
I + dP1 (f1i , α) dI
S= ,
P1 (f1i , α) + P2 (f2i , β) + dP1 (f1i , α)P2 (f2i , β) I + dP2 (f2i , β)
(4.56)
wherein i takes the values 1 and 2.
I O
S= . (4.57)
P1 (f1 , α) + P2 (f2 , β) 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 1 12 12
P1 (−f, −45 deg)+P2 (+f, +45 deg) = 2 − . (4.59)
f − 12 1
2 f 1
2
1
2
β out 1
f
0 0 1 β in
we see that a direct link has been established between y in and α out and between x in
and β 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 instru-
ment 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 represen-
tation 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 one-
dimensional 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 posi-
tioned 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 follow-
ing matrix chain to characterize the experimental situation (Hodgson and Weber,
1997):
S = Tb−f0 Qf Tf0 Ff0 Ta , (4.62)
wherein three translation matrices appear and Qf stands for the quadrupole lens
we obtained before,
⎡ ⎤
1 0 0 0
⎢ 0 ⎥
⎢ 1 0 0 ⎥ I O
Qf = ⎢ 0 − 1 1 0 ⎥= , (4.63)
⎣ f ⎦ Q I
1
f
0 0 1
and Ff0 represents a standard lens,
⎡ ⎤
1 0 0 0
⎢ 0
⎢ 1 0 0 ⎥
⎥ I O
Ff0 = ⎢ − 1 0 ⎥
1 0 ⎦ = . (4.64)
⎣ f0 P I
0 − f10 0 1
Rewriting the matrix chain in block-matrix form, we have
I (b − f0 )I I O I f0 I I O I aI
S= .
O I Q I O I P I O I
(4.65)
After multiplying the submatrices, we obtain
(4.66)
0 −1/f −1/f0 0 1
QP = = I. (4.67)
−1/f 0 0 −1/f0 ff0
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 subma-
trices appearing in the system matrix. The upper-left submatrix can be evaluated as
follows:
− fb0 0 0 − b−f 0
0 b−f0
I+ + f
+ b−f0 f
0 − fb0 − b−f
f
0
0 f
0
1 − fb0 0
= . (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
f0 (b−f0 )
(a + b)I + abP + (b − f0 )(a + f0 )Q + (b − f0 )f0 aQP = a + b − ab
f0 I− f I.
(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:
f0 (b − f0 )
(a + b)I + abP + (b − f0 )(a + f0 )Q + (b − f0 )f0 aQP = − I. (4.70)
f
x out b 1 0 x in f0 (b − f0 ) 0 1 α in
= 1− − . (4.71)
y out f0 0 1 y in f 1 0 β in
x out 1− b
f0
x in
= . (4.72)
y out − f0 (b−f0 ) in
β
f
This reflects that the instrument indeed operates as a phase-space analyzer: the
two phase-space coordinates (x in , α in ) are imaged on the analyzing screen (with
different scaling factors).
⎢ 0 1 0 ⎥
S=⎢ ⎥. (4.74)
⎣ − f1 − f1 + ff1 +f 2 f1 +f2
0 1 − f2 0 ⎦
1 2 1 f2
0 0 0 1
The system matrix for an anamorphic lens at the given distance then follows di-
rectly by concatenating both matrices,
⎡ ⎤
1 d 0 0
⎢ −1 1 − fdx 0 0 ⎥
⎢ fx ⎥
E = FG = ⎢ ⎥. (4.77)
⎣ 0 0 1 d ⎦
f
0 0 − f1y 1 − dy
At this point, we can introduce the condition for collimation in the y–z plane by
setting d = fy in Eq. (4.75),
⎡ ⎤
1 fy 0 0
⎢ −1 1 − ffxy 0 0 ⎥
⎢ ⎥
E = ⎢ fx ⎥. (4.78)
⎣ 0 0 1 fy ⎦
0 0 − f1y 0
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.
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 x in α in
=A +B . (4.80)
y out y in β in
Rays leaving at (x in , y in ) should all arrive at (x out , y out ) in the paraxial approxima-
tion. This implies that (x out , y out ) should not depend on B ϕin , where (x in , y in ). We
might state this as
in
α 0
B = . (4.81)
β in 0
I bI ⎢ − 1 ⎥ I aI
S = Tb F Ta = ⎣ fx
0 ⎦ . (4.83)
O I I O I
0 − f1y
I + bP (a + b)I + abP
S= , (4.84)
P I + aP
with
− f1x 0
P = .
0 − f1y
At this point, we can identify the upper-right submatrix with the matrix B,
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.
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 ⎤
r A B r r
⎣ ϕ out ⎦ = ⎣ C D ϕ ⎦ ⎣ ϕ in ⎦ . (4.89)
1 0 T 0 T 1 1
1 0 1 d2 1 0 1 d1 1 0
S= . (5.1)
− f13 1 0 1 − f12 1 0 1 − f11 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 1 d1 1 1 1 d1
= + − and = + − .
f12 f1 f2 f1 f2 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,
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.
s11 s12
0 s22
= T4 F3 T3 F2 T2 F1
s11 0
s21 s22
= T4 F3 T3 F2 T2 F1 T1
s11 s12
= F3 T3 F2 T2 F1 T1
s21 0
f22 = E2 E3 . (5.8)
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).
1 f 1 0 1 f
S= . (5.10)
0 1 − f1 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 so-
called 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 above-
mentioned type (Fouckhardt, 1994),
f2
0 f2 0 f1 − f1 0
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 out-
put 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
y 0 f2 m11 m12 0 f1 y
=
β out − f12 0 m21 m22 − f11 0 β in
f
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.
S = QP . (5.13)
We can now state the condition of telecentricity in image space by the mapping
out
y 0
=Q . (5.14)
0 β
The condition that 0 = q22 β 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 for-
mulated. Starting from the mapping from the object plane into the plane of the
aperture stop,
0 y in
=P , (5.15)
β 0
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.,
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
1 0 1 fk−1 + Ek + fk
Fk = and Tk = . (5.17)
− f1k 1 0 1
The special form of the translation matrix is chosen to simplify the product matri-
ces.
0 −f12
1
n = 1: F1 T1 = − , (5.18)
1 E1
f1
n = 4:
F4 T4 F3 T3 F2 T2 F1 T1
1
=
f1 f2 f3 f4
f22 f42 − E2 E3 f42 +E1 f22 f42 + E3 f12 f42 − E1 E2 E3 f42
× .
−E2 f32 − E4 f22 + E2 E3 E4 −E1 E2 f3 − E1 E4 f22 − E3 E4 f12 + f12 f32 + E1 E2 E3 E4
2
(5.21)
Figure 5.2 Dyson optics. The two spherical surfaces are separated by a distance z.
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 0 1 R+ 1 0 1
A = 1−n = 1−n n
. (5.23)
R+
n 0 1 0 n1 R
1
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 1−n
− = . (5.24)
f R+
This suggests the choice
R+
z=− (5.25)
1−n
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+
R+
1 − 1−n 1 R+ 0 1 − 1−n
1
1−n 1 = 1−n n
1 . (5.26)
0 1 R+ n R n
+
1 0 1 0
= . (5.27)
− R2g 1 − 2(n−1)
nR+
1
In analogy to the first step, the matrix that describes the third step of the propagation
reads as
1 R+ R+
1 − 1−n
R+ 1
1 − 1
n n
= n n 1−n . (5.28)
− n−1
R+
1 0 1 − n−1
R+
0
h 1 0 0 h −1 4 hn
n = , (5.30)
− h1 0 2
nh
1 − h1 n1 0 −1
and after using Eq. (5.28), the system matrix takes the form
4R+
−1 − n(1−n)
S= . (5.31)
0 −1
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.
with
1 1 1 d1 1 1 1 d1
= + − and = + − .
f12 f1 f2 f1 f2 f23 f2 f3 f2 f3
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:
d1 + d2 + d3 d2 + d3 d3 d 1 d 2 + d3 d3
1− − − + +
f1 f2 f3 f1 f2 f3
1 1 d1 d2 d3
+ d2 d3 + − = 0. (5.33)
f1 f2 f1 f2 f3
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.
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.
with
d10 + d20 + d30 d20 + d30 d30 d10 d20 + d30 d30
a0 = 1 − − − + +
f1 f2 f3 f1 f2 f3
1 1 0
(d )(d )d0 0
+ d20 d30 + − 1 2 3 = 0, (5.38)
f1 f2 f1 f2 f3
1 d 0 + d30 d0 1 1 (−d10 + d20 )d30
a1 = + 2 + 3 − d30 + − , (5.39)
f2 f1 f2 f1 f3 f1 f2 f1 f2 f3
0
1 d3
a2 = −1 . (5.40)
f1 f2 f3
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 effec-
tively 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
In this polynomial, a0 also vanishes because we have assumed that the start
configuration fulfills the focusing condition,
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
Figure 5.3 Variable optics: exact compensation at three points. The deviation from com-
pensation 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 ap-
proach 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 de-
vices. 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.
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. 5.4–5.6.
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)
(a)
Figure 5.6 Application to four lenses. (a) Theorem on optically compensated systems—the
case of a vanishing upper-left matrix entry is represented.
In the case of no disturbing aberrations (faberr → ∞), the focal length of the com-
pensating device would take the following value:
d3
fcomp = . (5.45)
1 − [(d2 + d3 − d2 d3 )/fstat ]
(b)
(c)
give a condition for two entries in the matrix C of the subsystem used for compen-
sation:
c11 (t11 a11 + t11 a21 ) + c12 (t21 a11 + t21 a21 ) = 0. (5.47)
(d)
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 beam-
splitter 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-
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 auto-
collimator.
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 γ
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 g 1 0 1 h+d
A=
0 1 − f1 1 0 1
1 − fg h + d + g − fg (h + d)
= , (5.48)
− f1 − f1 (h + d) + 1
where f is the focal length of the thin lens and h, d, and g are the distances appear-
ing 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:
(2)
(1)
y 1 0 y 0
= + . (5.49)
β (2) 0 1 β (1) 2γ
The matrix chain for the passage from the tilted target to the detector reads as
follows:
1 d +e 1 0 1 g
B=
0 1 − f1 1 0 1
g
1 − d+e g + (d + e) 1 −
= f f
, (5.50)
− f1 1 − fg
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
y −1 0 y 2f .
= +γ (5.53)
β out 2 fg2 − f1 −1 β in 2 1 − fg
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.,
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 input–output relation of the collimator suggests that
g = f might be a convenient choice for the corresponding distance.
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 λ of a 2 × 2 matrix are calculated from its characteristic equa-
tion,
a11 − λ a12
det = 0, (5.55)
a21 a22 − λ
where det A stands for the determinant of the matrix A. This is equivalent to
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,
2
a11 + a22 a11 + a22
λ1/2 = ± − 1. (5.58)
2 2
It is the discriminant 2
a11 + a22
dis = −1 (5.59)
2
that determines the character of the eigenvalues. This discriminant provides a dis-
tinction 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
λ= . (5.60)
2
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
λ1 λ2 = 1. (5.61)
AN = F D N F −1 (5.62)
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 ma-
trix 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 calcu-
lated from the eigenvector spaces of the matrix A that correspond to the eigenvalues
λ1 and λ2 . The eigenvector space eig(A, λ1/2 ) follows as the solution space of the
equation, i.e.,
Knowing that both eigenvalue are complex and that λ1 λ2 = 1 holds, we can set
λ1/2 = e±iϕ with i as the imaginary unit, for which i 2 = −1 holds. Solving
Eq. (5.31) for these eigenvalues, we find
±iϕ
±iϕ
e − a22
eig A, e = . (5.65)
a21
λ1 0 e 0
D= = . (5.68)
0 λ2 0 e−iϕ
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
e − a22
eig(A, e±α ) = . (5.73)
a21
With this result, we can now build the rows of the transformation matrix. In this
case, it reads +α
Also here, the upper-right entry of the matrix needed special attention and a11 +
a22 = trace(A) = 2 cosh α 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
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 (Her-
riott et al., 1964) and optical delay lines (Herriott and Schulte, 1965). The Her-
riott 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 Fu-
jimoto (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 Sylvester’s theorem to the case where det A = 1 and/or
incorporating error vectors in homogeneous coordinates are given by Tovar and
Casperson (1995).
Figure 5.9 Fizeau interferometer. The beamsplitting element is represented by the horizon-
tal 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.,
y y b1
=A + , (5.77)
β β b2
wherein
1 − 2d − 2d 1 − fd 2d 1 − Rd
A= f
R
2 , (5.78)
− f2 1 − fd − R2 1 − fd 1 − 2d
f
− 2d
R
1 − fd
and
b1 d
= 2γ . (5.79)
b2 1− d
f
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)
β (1) 0 1 β in
During the second step, we make reference to the result of Sec. 2.1, i.e.,
(2)
(1)
y y b1
=A + . (5.81)
β (2) β (1) b2
The ray finally returns to the reference plane chosen directly behind the beamsplit-
ter and we have out
(2)
y 1 e y
= . (5.82)
β out 0 1 β (2)
When these three steps are combined, they result in
out
in
y 1 e 1 e y 1 e b1
= A + , (5.83)
β out 0 1 0 1 β in 0 1 b2
or
y out 1 e 1 e y in d + e(1 − fd )
= A + 2γ . (5.84)
β out 0 1 0 1 β in 1 − fd
Having a look at the sensitivity, the γ -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)
∂γ f
out
∂β d
=1− . (5.86)
∂γ f
1 t 1 h
1 a 1 a+ h
+t
n = n . (5.88)
0 1 0 1 0 1 0 1
y (1) 1 a+ h
+t y in
= n
. (5.89)
β (1) 0 1 β in
y 1 0 y 0
= + . (5.90)
β (2) 0 1 β (1) 2γ
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
out
(2)
y 1 t + hn + e x
out = . (5.92)
β 0 1 β (2)
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
y 1 a + 2 hn + e + t x in t + hn + e
= + 2γ . (5.93)
β out 0 1 β in 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 (β in = 0), the change in
position as it impinges on the detector is as follows:
h
y = y − y = 2γ t + + e .
out in
(5.94)
n
5.4.3 Dyson interferometer
The Dyson interferometer (Dyson, 1968, 1979; Steel, 1983; Heavens and Ditch-
burn, 1991, p. 197) is an example of a measurement head that is relatively insensi-
tive 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 princi-
ple. 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
βout
1 + 8 (1 − a )(1 − b )(−(a + b) + ab ) − f4 (1 − fb )(a + b − ab
f )
2
= −f4 (1 −fa )[1 − f1 (a + b − ab )] f − 4 (1 −
f )(1 − f )(a + b − f ) + (1 − f (a + b) +
a b ab 2 2ab 2
f f f f f f2
)
yin 4 a + b − ab 1 − 1
(a + b) + ab
× + 2γ f f f
2
. (5.95)
βin 2 1 − fa − f2 (a + b) + 2 fab2 + 1
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. (5.96)
For this arrangement, the transfer mapping reads as
yout 1 0 yin
= . (5.97)
βout 0 1 βin
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 cat’s
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
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 achiev-
able, 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:
1 f 1 0 1 f 0 f
G= = . (5.98)
0 1 − f1 1 0 1 − f1 0
For the central part of the unfolded optical path, the square of this matrix will be
convenient, that is,
−1 0
G =
2
. (5.99)
0 −1
This establishes a link to the retroreflector discussed in Sec. 2.1.
Figure 5.15 Dyson interferometer, and the unfolded path of the outer beam.
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 cγ or dy, respectively, the addition of the tilt-
error vector, and, again, a propagation by a distance of cγ or dy, respectively. Put
together, this gives
1 cγ 1 cγ y 0 y 2cγ 2
+ =C +
0 1 0 1 β 2γ β 2γ
1 2cγ
with C = (5.100)
0 1
and
y −2dγ 2 1 −2dγ
D + with D = . (5.101)
β 2γ 0 1
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
y 0 f y
= . (5.102)
β (1) − 1
f
0 β in
The following ray vector is one changed by the tilted mirror:
(2)
(1)
y y 2cγ 2
=C + . (5.103)
β (2) β (1) 2γ
Using Eq. (5.102), this gives
(2)
2cγ
y − f f y in 2cγ 2
= + . (5.104)
β (2) − f1 0 β in 2γ
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.,
(3)
(2)
y −1 0 y
= . (5.105)
β (3)
0 −1 β (2)
To this result, the transformation involving the matrix D will then be applied as
(4)
(3)
y 1 −2dγ y −2dγ 2
= + . (5.106)
β (4) 0 1 β (3) 2γ
After performing the multiplications, it follows that
(4)
2γ
y (c − d) −f y in −2dγ 2
= f
− . (5.108)
β (4) 1
f
0 β in 0
The term that is quadratic in γ 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 0 f y (4)
= . (5.109)
β out − f1 0 β (4)
y 1 0 y in
= . (5.110)
β out − f2γ2 (c − d) 1 β in
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 tilt-
dependent entry in the system matrix is brought to zero.
1 f2 + c 1 0 1 f2 1 f1 + b
S=
0 1 − f12 1 0 1 0 1
1 0 1 f1 + a
× . (5.111)
− f11 1 0 1
The distances appearing in this equation are related by the imaging conditions im-
posed 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
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
− fb1 0
A= , (5.114)
− f11 − fb1
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 f2 + c 1 0 1 f2 1 f1 + b
B=
0 1 − f12 1 0 1 0 1
− f2 f2 − f2 (f1 + b)
c c
= . (5.116)
− f12 − f1f+b
2
We can now draw conclusions from these considerations that characterize the illu-
mination 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
1 f2 + c 1 0 1 f2 − fc2 f2
C= = . (5.119)
0 1 − f12 1 0 1 − f12 0
Equation (5.117) might also be used at this point, and we have the mapping
out
y − f1f+b
2
f2 y ap
= , (5.120)
β out − f1 0 β ap
2
which implies that the angle β 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
y − f1f+b
2
0 y field-stop
= ,
β out − f1 − f1f+b β field-stop
2 2
which shows that adapting the diameter y field-stop of the field stop directly changes
the field on the target plane.
Altmann, J., Baumgart, R., and Weitkamp, C., “Two-mirror multipass absorption cell,”
Appl. Opt., 20, pp. 995–999 (1981).
Antonsen, E.L., Burton, R.L., Spanjers, G.G., and Engelman, S.F., “Herriott cell interfer-
ometry for millimeter-scale plasma measurements,” Rev. Sci. Instrum., 74, pp. 88–93
(2003a).
Antonsen, E.L., Burton, R.L., Spanjers, G.G., and Engelman, S.F., “Herriott cell augmenta-
tion of a quadrature heterodyne interferometer,” Rev. Sci. Instrum., 74, pp. 1609–1612
(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. 684–691 (1954).
Back, F.G., and Lowen, H., “Generalized theory of Zoomar systems,” J. Opt. Soc. Am.,
48(3), pp. 149–153 (1958).
Baker, D., “Field-widened interferometers for Fourier spectroscopy,” Spectrometric Tech-
niques, G.A. Vanasse (Ed.), pp. 71–106, Academic Press, New York (1977).
Barr, J.R.M., “Achromatic prism beam expanders,” Opt. Commun., 51(1), pp. 129–134
(1984).
Berek, M., Grundlagen der praktischen Optik—Analyse 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 appli-
cation of electrowetting,” Eur. Phys. J. E, 3, pp. 159–163 (2000).
Besenmatter, W., “Designing zoom lenses aided by the Delano diagram,” Proc. SPIE, 237,
pp. 242–255 (1980).
Born, M., Optic—Ein Lehrbuch der Electro-magnetischen Lichttheorie, p. 102, Springer,
Berlin (1933) (reprint 1985).
Born, M., and Wolf, E., Principles of Optics—Electromagnetic Theory of Propagation,
Interference and Diffraction of Light, pp. 225–226, Pergamon Press, Oxford (1980).
Breckinridge, J.B., and Schindler, R.A., “First-order optical design for Fourier spectrom-
eters,” 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 pro-
jectors with single digital micromirror devices,” Appl. Opt., 39, pp. 3202–3208 (2000).
Clark, A.D., Zoom Lenses, Adam Hilger, London (1973).
Das, P., Lasers and Optical Engineering, Springer-Verlag, Berlin, p. 302 (1991).
Delano, E., “First-order design and the y, ȳ diagram,” Appl. Opt., 2, pp. 1251–1256 (1963).
Demtröder, W., Laserspektroskopie—Grundlagen und Techniken, Chap. 4, Springer-
Verlag, 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 Shaping—Theory 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. 303–307 (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.
259–262 (1985a).
Duarte, F.J., Erratum, Opt. Commun., 54, p. 388 (1985b).
Duparré, J., Schreiber, P., Matthes, A., Pshenay-Severin, E., Bräuer, A., and Tünner-
mann, A., “Microoptical telescope compound eye,” Opt. Express, 13(3), pp. 889–903
(2005).
Dyson, J., “Unit magnification optical system without Seidel aberrations,” J. Opt. Soc. Am.,
49, pp. 713–716 (1959).
Dyson, J., “Optics in a hostile environment,” Appl. Opt., 7, pp. 569–580 (1968).
Dyson, J., Interferometry as a Measuring Tool, pp. 96–98, 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. 5008–5009
(1991).
Fouckhardt, H., Photonik—Eine Einführung in die integrierte Optoelektronik und technis-
che Optik, pp. 94–95, 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, Tech-
niques, and Design, pp. 1.1–1.109, McGraw-Hill, New York (1995).
Grenville, A., Hsieh, R.L., von Bünau, R., Lee, Y.-H., Markle, D.A., Owen, G., and
Pease, R.F.W., “Markle-Dyson optics for 0.25 μm lithography and beyond,” J. Vac.
Technol. B, 9(6), pp. 3108–3112 (1991).
Grenville, A., Owen, G., and Pease, R.F.W., “Image monitor for Markle-Dyson optics,”
J. Vac. Sci. Technol. B, 11(6), pp. 2700–2704 (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., Kärkkäinen, P.A., and Wyatt, R., “A simple beam expander for frequency
narrowing of dye lasers,” Opt. Quant. Electron., 7, pp. 115–119 (1975).
Heavens, O.S., and Ditchburn, R.W., Insight into Optics, Wiley, Chichester (1991).
Herriott, D., Kogelnik, H., and Kompfner, R., “Off-axis paths in spherical mirror interfer-
ometers,” Appl. Opt., 3, pp. 523–526 (1964).
Herriott, D.R., and Schulte, H.J., “Folded optical delay lines,” Appl. Opt., 4, pp. 883–889
(1965).
Herzberger, M., “Gaussian optics and Gaussian brackets,” J. Opt. Soc. Am., 33(12), pp.
651–655 (1943).
Hodgson, N., Haase, T., Kostka, R., and Weber, H., “Determination of laser beam parame-
ters with the phase space beam analyzer,” Opt. Quant. Electron., 24, pp. S927–S949
(1992).
Hodgson, N., and Weber, H., Optical Resonators—Fundamentals, 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. 829–839 (2001).
Hsieh, R.L., Lee, Y.Y.-H., Maluf, N.I., Browning, R., Jerabek, P., and Pease, R.F., “Reflec-
tive masks for 1X deep ultraviolet lithography,” Proc. SPIE, 1604, pp. 67–77 (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 mov-
ing element,” Appl. Opt., 31, pp. 2274–2278 (1992).
Kay, D.B., and Gage, E.C., “Heads and lasers,” T.W. McDaniel, and R.H. Victora, Hand-
book of Magneto-Optical Data Recording—Materials, 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. 61–63 (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. 57–64
(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 elec-
tromagnetic beams,” U.S. Patent No. 4,759,616 (1988).
Marchant, A.B., Optical Recording—A 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. 592–597 (1954).
McLeod, J.H., “Axicons and their uses,” J. Opt. Soc. Am., 50(2), pp. 166–169 (1960).
Naumann, H., and Schröder, G., Bauelemente der Optik—Taschenbuch der technischen
Optik, Carl Hanser Verlag, München, p. 187 (1992).
New, R.M.H., Owen, G., and Pease, R.F.W., “Analytic optimization of Dyson optics,” Appl.
Opt. 31, pp. 1444–1449 (1992).
Niefer, R.J., and Atkinson, J.B., “The design of achromatic prism beam expanders for
pulsed dye lasers,” Optics Commun. 67(2), pp. 139–143 (1988).
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 Bünau, R., Jeong, H., Markle, D.A., and Pease, R.F., “The ef-
fect of gaps in Markle-Dyson optics for sub-quarter-micron lithography,” Jpn. J. Appl.
Phys. (Pt. 1), 32, pp. 5840–5844 (1993).
Pegis, R.J., and Peck, W.G., “First-order design theory for linearly compensated zoom
systems,” J. Opt. Soc. Am., 52(8), pp. 905–911 (1962).
Pérez, J.-P., Optique—Fondements et applications, Masson, Paris (1996).
Reynolds, G.O., DeVelis, J.B., Parrent, G.B., and Thompson, B.J., The New Physical Op-
tics Notebook: Tutorials in Fourier Optics, p. 154, SPIE, Bellingham, WA, American
Institute of Physics, New York (1989).
Rioux, M., Tremblay, R., and Bélanger, P.A., “Linear, annular, and radial focusing with
axicons and applications to laser machining,” Appl. Opt. 17, pp. 1532–1536 (1978).
Roberts, R.B., “Absolute dilatometry using a polarization interferometer,” J. Phys. E 8,
pp. 600–602 (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. 1315–1322 (2000).
Schreiber, P., Kudaev, S., Dannberg, P., and Zeitner, U.D., “Homogeneous LED-
illumination using microlens arrays,” Proc. SPIE, 5942, pp. 59420K-1–59420K-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. 1106–1113 (2003).
Shack, R.V., “Analytic system design with pencil and ruler—The advantages of the y–ȳ
diagram,” Proc. SPIE, 39, pp. 127–140 (1973).
Siegman, A.E., Lasers, University Science Books, Mill Valley, pp. 582–583 (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 Zoom-
Objektives,” Opt. Commun. 29(2), pp. 138–140 (1979).
Tanaka, K., “Paraxial analysis of mechanically compensated zoom lenses. 1: Four-
component type,” Appl. Opt. 21(12), pp. 2174–2183 (1982).
Torre, A., Linear Ray and Wave Optics in Phase Space—Bridging 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 applica-
tions in matrix optics,” J. Opt. Soc. Am. A, 12, pp. 578–590 (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 ’01—International Symposium on Optical Memory 2001 Technical Digest,
Washington, D.C.: Optical Society of America. pp. 196–197 (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. 4.1–4.29 McGraw-Hill, New York (1995).
Wong, A.K.-K., Resolution Enhancement Techniques in Optical Lithography, pp. 7–8,
SPIE, Bellingham, WA (2001).
Wooters, G., and Silvertooth, E.W., “Optically compensated zoom lens,” J. Opt. Soc. Am.,
55(4), pp. 347–351 (1965).
Yoder, P.R., “Non-image-forming optical components,” Proc. SPIE, 531, pp. 206–220
(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. 5.1–5.16, McGraw–Hill, New York (1995).
I phase space, 20
image field, 19 phase-conjugate mirror, 26
imaging, 10 phase-space analysis, 69
imaging condition, 11, 12, 20, 73 phase-space beam analyzer, 70
imaging equation, 74, 78 plane mirror, 25
imaging triplet, 77 plane-parallel plate, 29
integrating rod, 49 principal planes, 8
internal reflection, 45, 46 principal ray, 17
prism for compression along one
K axis, 38
Keplerian telescope, 72
prism for expansion along one axis, 36
Köhler’s illumination system, 107
prisms, 31
L
Lagrange invariant, 18 Q
laser beam circularization, 33 quadrupole lens, 69
laser resonators, 99
laser technology, 95 R
law of refraction, 1, 4 ray slope, 6
lens, 6 recursion formula, 23
lens doublet, 12 reflection, 5, 25
light guide, 49 refraction, 3, 29
liquid lens, 90 refraction matrix, 21
retroreflector, 26
M roof mirror, 28
magnification, 19 rotation of an anamorphic component, 61
matrix representations, 5
mechanical compensation, 84, 85 S
Michelson interferometer, 101 sensitivities, 53, 75
microscope, 13 single-pass optics, 77
microscopy, 107 spatial filtering, 79
misadjustment, 53 spherical mirror interferometers, 99
mixing rod, 52 spherical surface, 4
multiple-pass optics, 95 stable, 99
Sylvester’s theorem, 98
N symplectic systems, 111
nested ray tracing, 107 synthesis of optical systems, 13
nondispersive, 32
system synthesis, 13
nonorthogonal anamorphic system, 60
O T
optical compensation, 84, 85 telecentric lenses, 80
optical delay lines, 99 telecentricity, 80
optical resonators, 95 telescope, 13, 72
optics, 84 telescopes, 11
orthogonal anamorphic system, 60 terrestrial telescope, 78
thin-prism approximation, 32
P tilt, 53
partitioned matrix, 59 tolerances, 53, 75
Petzval radius, 19 tolerancing, 39
transfer function, 35 V
translation matrix, 3, 21 variable, 84
trigonometric description of dispersive varifocal systems, 84
prisms, 33
triple mirror, 52 W
triplet, 77, 84 Wigner distribution functions, 112
tunnel diagram, 31
U Z
unstable, 99 zoom lens, 84