Sei sulla pagina 1di 46

Technical note TN-2007/00025

Issued: 3/2007
Ray-Optics Analysis of Homogeneous
Uniaxially Anisotropic Media
M. Sluijter
Philips Research Eindhoven
Unclassied
c _Koninklijke Philips Electronics N.V. 2007
TN-2007/00025 Unclassied
Concerns: Interim Report of PhD thesis
Period of
Work:
01/2006-11/2006
Notebooks: None
Authors address M. Sluijter
Prof Holstlaan 4, HTC 34
5656 AE Eindhoven
WB 3011, MS 31
Tel: 040-2747569
Supervisor: Dr. D.K.G. de Boer
Professor: Prof. Dr. Ir. J.J.M. Braat
c _ KONINKLIJKE PHILIPS ELECTRONICS N.V. 2007
All rights reserved. Reproduction or dissemination in whole or in part is
prohibited without the prior written consent of the copyright holder.
ii c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Title: Ray-Optics Analysis of Homogeneous Uniaxially Anisotropic Media
Author(s): M. Sluijter
Reviewer(s): D.K.G. de Boer, T.C. Kraan
Technical
Note:
TN-2007/00025
Additional
Numbers:
Subcategory:
Project:
Customer:
Keywords: geometrical optics, optical anisotropy, uniaxially anisotropic media,
optical indicatrix, polarization ray tracing, polarization vector, Fresnel
coecient, Fresnel equations, phase shift, reectance, transmittance
Abstract: In this report, we introduce a model based on geometrical optics
which describes the propagation of light in homogeneous uniaxially
anisotropic media. We call this model the OpIn model. A ray-tracing
algorithm based on the OpIn model is also introduced. Results from a
test case obtained by the OpIn model are compared with results from
the commercial ray-tracing program ASAP. It appears that results ob-
tained by ASAP show a number of defects, while the OpIn model
produces plausible and solid results. This gives condence in the use
of the OpIn model and the corresponding ray-tracing algorithm.
Conclusions:
c _ Koninklijke Philips Electronics N.V. 2007 iii
TN-2007/00025 Unclassied
iv c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Contents
1 Introduction 1
1.1 Optical anisotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Geometrical optics of anisotropic media 5
2.1 Dielectric tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Harmonic electromagnetic plane waves . . . . . . . . . . . . . . . . . . . . 6
2.3 Optical indicatrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4 Uniaxially anisotropic media . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Geometrical analysis of the polarization vectors . . . . . . . . . . . . . . . 11
2.6 Polarization ray tracing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6.1 Anisotropic-anisotropic interface . . . . . . . . . . . . . . . . . . . 13
2.6.2 Wave vectors in the plane of incidence . . . . . . . . . . . . . . . . 14
2.6.3 Fresnel equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6.4 Total reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6.5 Transmittance and reectance . . . . . . . . . . . . . . . . . . . . . 17
2.7 OpIn model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Numerical ray tracing of anisotropic media 21
3.1 Ray-tracing algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Phase shift and reectance of internally reected waves . . . . . . . . . . 23
3.3 Phase shift and reectance of isotropic-anisotropic interfaces . . . . . . . . 25
3.4 ASAP as a bench mark . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4 Conclusions 31
A 37
A.1 Plane waves in (in)homogeneous media . . . . . . . . . . . . . . . . . . . . 37
A.2 Polarization eigen modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
A.3 Ordinary polarization vector . . . . . . . . . . . . . . . . . . . . . . . . . . 38
A.4 Extraordinary polarization vector . . . . . . . . . . . . . . . . . . . . . . . 39
B 40
B.1 Fresnel rhomb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
c _ Koninklijke Philips Electronics N.V. 2007 v
Unclassied TN-2007/00025
Section 1
Introduction
1.1 Optical anisotropy
Anisotropy [1][2][3] (the opposite of isotropy) is the property of being directionally de-
pendent. Anisotropy can be found in many elds of interest. In this report, we will focus
on optical anisotropy only. Materials which are optically anisotropic are also known as
birefringent materials. This is because these materials are known to cause double refrac-
tion. Double refraction means that a single incident beam of light is split into two beams
of light when refracted at the surface of a birefringent material. One beam is labelled
the ordinary ray and the other is labelled the extraordinary ray. The extraordinary ray
is named so since it does not exhibit the ordinary behavior imposed by Snells law. Un-
like an isotropic material, a birefringent material has two refractive indices instead of one.
Most birefringent materials are crystals. A well-known example of a birefringent
crystal is calcite. The basic molecular unit of calcite is CaCO
3
which assumes a pyramidal
structure as shown in Figure 1.1. The carbon and oxygen atoms form the base of the
pyramid, with carbon lying in the center of the triangle of oxygen atoms. The calcium
atom is positioned above the carbon atom, at the top of the pyramid. The gure shows
two directions of propagation through the crystal.
First, we consider light entering from below along the line through both the carbon
and calcium atoms. The molecule, and thus the crystal, is symmetric with respect to
this direction. This direction of symmetry is called the optical axis of the crystal. The
oscillations of the electromagnetic eld are represented by the two transverse vectors
E

. These oscillations of the electromagnetic eld both interact with the electrons in
the same way when propagating through the calcite along the optical axis.
Subsequently, we consider light entering from the left. From this direction, the two
vibration directions E
||
and E

have dierent interactions with the electrons in the base


of the pyramid. The speed of the E

-component wave v

is reduced more than the


speed of the E
||
-component wave v
||
. Since the refractive index n =
c
v
, with c the speed
of light, n

> n
||
. The measured values for calcite are n

= 1.658 and n
||
= 1.486 for
light with a wavelength = 589.3 nm. The ordinary refractive index is denoted n

and
the extraordinary refractive index is denoted n
||
. For calcite, n

> n
||
. Then, the crystal
is called uniaxial negative. If n

< n
||
, a crystal is called uniaxial positive.
Cubic crystals such as diamond (C) or salt (NaCl) are optically isotropic and have
c _ Koninklijke Philips Electronics N.V. 2007 1
TN-2007/00025 Unclassied
Figure 1.1: Pyramidal structure of a calcite molecule (CaCO
3
). The optical axis is
parallel to the direction of symmetry, which is the line joining the C and Ca atoms.
Since the crystal has one direction of symmetry, it is uniaxially anisotropic.
Figure 1.2: Example of a monoclinic crystalline system. This type of crystal has two
optical axes and therefore, it is biaxially anisotropic.
2 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
only one index of refraction. Crystals which possess two directions of symmetry are bi-
axially anisotropic and have three refractive indices. This type of crystal also generates
two beams of light when light is refracted at a birefringent surface. One example is mica,
which crystallizes in monoclinic forms. A monoclinic structure is depicted in Figure 1.2.
The optical properties of uniaxially anisotropic media are essential for many appli-
cations such as liquid crystal displays [4], switchable lenticulars for auto-stereoscopic 3D
displays [5] or GRIN lenses for active beam manipulation [6]. Therefore, it is most de-
sired to understand and predict the propagation of light in optical systems which contain
optically anisotropic media.
1.2 Outline
Optical anisotropy has already been studied intensively the last few decades. In this
report, however, our goal is to describe optical anisotropy in a unique and universal way.
In Chapter 2, we introduce a model which describes the optical anisotropy of uni-
axially anisotropic media within the geometrical optics approach. For the moment, we
restrict ourselves to homogeneous anisotropic media. With homogeneity we mean to say
that the direction of the optical axis is position independent inside the medium. In the
future, our goal is to model the optical anisotropy of inhomogeneous anisotropic media.
In Chapter 3, we present a ray-tracing algorithm based on the model derived in
Chapter 2. We will use this algorithm to test the model and use the commercial ray-
tracing program ASAP as a bench mark.
c _ Koninklijke Philips Electronics N.V. 2007 3
TN-2007/00025 Unclassied
4 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Section 2
Geometrical optics of anisotropic
media
The electromagnetic eld associated with the propagation of light is characterized by
the smallness of the wavelength which is in the order of 10
5
cm. In geometrical optics,
it is assumed that
0
0 with
0
the vacuum wavelength [1][2][3]. Geometrical optics
works well as long as the wavelength is small compared to the characteristic dimension
of the problem at hand. Optical phenomena may be deduced from geometrical consid-
erations, by determining the paths of light rays and calculating the associated intensity
and polarization. In this chapter, we introduce a model based on geometrical optics
which describes the propagation of light in homogeneous uniaxially anisotropic media.
In particular, we focus on the refraction and reection of light at plane interfaces between
anisotropic media.
2.1 Dielectric tensor
The most fundamental equations in electrodynamics are the macroscopic Maxwell equa-
tions. These read in SI units [7]:
E+
B
t
= 0, H
D
t
= J, D = , B = 0. (2.1)
In these equations, E and H are the electric eld vector and the magnetic eld vector,
respectively. The quantities D and B are the electric displacement and the magnetic
induction, respectively. The quantities and J are the electric charge density and the
current density, respectively, and may be considered as the sources of the elds E and
H.
The macroscopic Maxwell equations are completed with the macroscopic material
equations
D =
0
E, (2.2)
B =
0
H. (2.3)
Here, and represent the dielectric tensor and the permeability tensor, respectively,
and
0
is the vacuum permittivity and
0
is the vacuum permeability. If the medium is
optically isotropic, both and reduce to scalars.
c _ Koninklijke Philips Electronics N.V. 2007 5
TN-2007/00025 Unclassied
For the remainder of this report we assume that the medium is homogeneous, non-
conducting (J=0), free of electrical charge (=0) and non-magnetic ( is the unit tensor),
but we allow electrical anisotropy. According to Equation 2.2, this means that the vector
D is not in general in the direction of the vector E. In matrix form, the linear relation
between D and E is given by
D
i
=
0

j=x,y,z

ij
E
j
, (2.4)
where i represents the coordinates x, y and z. The elements
ij
are constants of the
medium and constitute the dielectric tensor. This tensor is real and symmetric, which
implies
ij
=
ji
, with the complex conjugate of [1][8].
It is always possible to nd a coordinate system in which the o-diagonal elements
of the dielectric tensor are zero. Hence, the dielectric tensor can be written as
=

x
0 0
0
y
0
0 0
z

, (2.5)
where
x
,
y
, and
z
are the relative principal dielectric constants. The x, y and z axes
are the principal dielectric axes of the crystal structure of the (homogeneous) anisotropic
medium. These axes form the principal coordinate system. The principal relative di-
electric constants are related to the principal indices of refraction n
x
, n
y
and n
z
by

i
= n
2
i
, i = x, y, z, (2.6)
where the constant is the relative magnetic permeability.
2.2 Harmonic electromagnetic plane waves
In order to describe the propagation of light in homogeneous (an)isotropic media, we
assume a monochromatic time harmonic plane wave given by (see Appendix A.1)
E = E
0
e
i(+)
, (2.7)
where denotes the constant part of the phase factor and denotes the variable part of
the phase factor given by
= t k r, (2.8)
with r = (x, y, z) a position vector, k the wave vector and the angular frequency. The
wave vector k is dened as
k =

c
n()

k, (2.9)
where

k is a unit vector in the direction of propagation, n the eective index of refraction
and c the speed of light. In addition, we dene the complex amplitude of the time
harmonic plane wave as

E = E
0
e
i
. (2.10)
In terms of

E, the electric eld vector reads
E =

Ee
i(krt)
. (2.11)
6 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Figure 2.1: Directions of the wave vector, eld vectors and Poynting vector.
In the same way, the magnetic eld vector reads
H =

He
i(krt)
. (2.12)
By substituting the electric eld vector E and the magnetic eld vector H into the
Maxwell equations we obtain
k E =
0
H (2.13)
and
k H = D. (2.14)
The energy ow of a light ray is described by the time average of the Poynting
vector [1]. The Poynting vector is dened as the ow of energy per unit area of an
electromagnetic wave in terms of electric and magnetic quantities. The time-averaged
Poynting vector S) is given by
S) =
1
2
Re(EH

), (2.15)
with H

the complex conjugate of H. According to Equation 2.15, the time-averaged


Poynting vector is always perpendicular to the electric eld vector E and the magnetic
eld vector H. In general, D, H and k on the one hand, and E, H and S on the other
hand, form orthogonal triads. In addition, the vectors k, D, E and S are coplanar and
all perpendicular to B and H. These vectors are depicted in Figure 2.1.
2.3 Optical indicatrix
Eliminating H from equations 2.13 and 2.14 and using Equation 2.2 yields
k (k E) +

2
c
2
E = 0. (2.16)
c _ Koninklijke Philips Electronics N.V. 2007 7
TN-2007/00025 Unclassied
Figure 2.2: Octant of the optical indicatrix in k space with k
x
, k
y
and k
z
in units of

c
.
If we solve this equation for the eigenvectors E and the corresponding eigenvalues n, we
obtain the electromagnetic eigen modes: the ordinary and extraordinary wave. Using
Equation 2.5, we can write Equation 2.16 as
A(k, )E = 0, (2.17)
where
A(k, ) =

2
c
2

x
k
2
y
k
2
z
k
x
k
y
k
x
k
z
k
y
k
x

2
c
2

y
k
2
x
k
2
z
k
y
k
z
k
z
k
x
k
z
k
y

2
c
2

z
k
2
x
k
2
y

. (2.18)
Equation 2.17 has only nontrivial solutions for the eigenvector E if the determinant of
the matrix A vanishes. This demand leads to a relation between and k. For a given
frequency , the relation between and k represents a three-dimensional surface in
k space. This surface is known as the normal surface [4][10], Fresnel surface of wave
normals [8] or the optical indicatrix [11]. In general, the optical indicatrix consists of
two shells [1]. These two shells have four points in common. The two lines that go
through the origin and these points are known as the optical axes. Figure 2.2 shows
one octant of the optical indicatrix in the principal coordinate system, where the wave
vector k is represented in units of

c
.
Since the optical indicatrix consists of two shells, there are two values of k for each
direction of propagation

k. These two values of k correspond to two dierent phase
velocities (/k) of the waves propagating along the chosen direction. For propagation
along the optical axis, there is only one value of k and therefore only one phase velocity.
8 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
The two dierent phase velocities correspond to two eigen modes of the polarization.
The directions of the electric eld vectors associated with these eigen modes are obtained
from equations 2.16 and 2.9 and yield

E
x
E
y
E
z

k
x
n
2
n
2
x

k
y
n
2
n
2
y

k
z
n
2
n
2
z

, (2.19)
given that the denominators do not vanish (see Appendix A.2). The vector that denes
the direction of a eld vector is called the polarization vector. Note that in a non-
absorbing medium, the polarization vectors represent linear polarization states since all
the components in Equation 2.19 are real
1
. Given Equation 2.17 and Equation 2.19, it
can be shown that Equation 2.19 is only valid if

k
2
x
n
2
n
2
x
+

k
2
y
n
2
n
2
y
+

k
2
z
n
2
n
2
z
=
1
n
2
. (2.20)
Equation 2.20 is known as Fresnels equation [1] and can be solved for the eigenvalues
of index of refraction. Note that for any direction of propagation, there are in general
two solutions for n with two corresponding polarization eigen modes.
2.4 Uniaxially anisotropic media
Consider the dielectric tensor given by Equation 2.5 dened in the principal coordinate
system. Depending on the values of
i
and thus on the values of n
i
, we can distinguish
between three types of media.
If n
x
,= n
y
,= n
z
, the medium is called biaxially anisotropic [1][12]. In this case, the
medium contains two optical axes and the optical indicatrix is as depicted in Figure 2.2.
A medium is called uniaxially anisotropic (birefringent) if two of the principal indices
of refraction are equal. The index of refraction that corresponds to these two indices is
the ordinary index of refraction n
o
. The remaining principal index of refraction is the
extraordinary index of refraction n
e
. In the principal coordinate system, the principal
indices of refraction are dened as n
x
= n
y
= n
o
and n
z
= n
e
.
2
As a result, the optical
indicatrix is dened as
det(A(k, )) = (
k
2
x
+ k
2
y
n
2
e
+
k
2
z
n
2
o
1)(
k k
n
2
o
1) = 0, (2.21)
with k
x
, k
y
and k
z
in units of

c
. According to Equation 2.21, the optical indicatrix
consists of two concentric shells. One shell represents a sphere with radius n
o
and the
other one an oblate ellipsoid of revolution with semi-axes n
o
and n
e
. Figure 2.3 shows
one octant of the optical indicatrix. The medium has only one optical axis and in the
principal coordinate system, it is exactly aligned with the z-axis.
If all three principal indices of refraction are equal, the medium is optically isotropic.
Then, the optical indicatrix is a sphere with radius n.
1
Equation 2.19 represents reected waves by taking

k complex. This topic is further discussed in
Subsection 2.6.4.
2
For the moment, we assume positive birefringence, i.e. n
e
> n
o
.
c _ Koninklijke Philips Electronics N.V. 2007 9
TN-2007/00025 Unclassied
Figure 2.3: Octant of the optical indicatrix of an uniaxially anisotropic medium in the
principal coordinate system. The vector components k
x
, k
y
and k
z
are in units of

c
.
Consider Figure 2.3. The sphere with radius n
o
describes the relation between and
k for the ordinary (O) wave. The ellipsoid of revolution describes the relation between
and k for the extraordinary (E) wave. Figure 2.4 shows the complete optical indicatrix
in the principal coordinate system. The polarization vector of the E-wave can directly
be obtained from Equation 2.19 and reads
E
e
=

k
x
n
2
n
2
o

k
y
n
2
n
2
o

k
z
n
2
n
2
e

. (2.22)
The eective index of refraction n is determined by
1
n
2
=

k
2
x
+

k
2
y
n
2
e
+

k
2
z
n
2
o
. (2.23)
The eective index of refraction for the O-wave reads
n = n
o
. (2.24)
Since Equation 2.24 is singular if n
2
=
i
, we consider Equation 2.16 and derive an
expression for the polarization vector of the O-wave (see Appendix A.3). The result
reads
E
o
=

k
y

k
x
0

. (2.25)
10 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Figure 2.4: The optical indicatrix of an uniaxially anisotropic medium in the principal
coordinate system.
In the next section, we derive more elegant expressions for the polarization vectors by
using the geometrical properties of the optical indicatrix.
2.5 Geometrical analysis of the polarization vectors
From Equation 2.25 it is clear that E
o
is orthogonal to the direction of propagation

k.
As a result, E
o
is tangent to the optical indicatrix for the O-wave, which is a sphere
with radius n
o
. In addition, E
o
is perpendicular with respect to the z-axis, which is the
optical axis in the principal coordinate system. We can generalize this property
3
and
write the electric polarization vector of the O-wave as
E
o
= k
o
o, (2.26)
where k
o
is the ordinary wave vector and o is a unit vector in the direction of the optical
axis. Hence, the polarization vector E
o
is always orthogonal to the optical axis. Accord-
ing to Equation 2.13, the magnetic polarization vector H
o
is perpendicular to both k
o
and E
o
. As a result, H
o
is also tangent to the optical indicatrix.
According to Equation 2.21, the ellipsoid of revolution in the principal coordinate
system is given by
H(k
x
, k
y
, k
z
) =
k
2
x
+ k
2
y
n
2
e
+
k
2
z
n
2
o
1 = 0. (2.27)
The polarization vector E
e
is tangent to the optical indicatrix since it can be demon-
strated that (see Appendix A.4)
H E
e
= 0. (2.28)
3
The orientation of the individual polarization vectors with respect to the optical indicatrix should
be independent of the choice of coordinate system in which we dene these polarization vectors.
c _ Koninklijke Philips Electronics N.V. 2007 11
TN-2007/00025 Unclassied
Figure 2.5: Octant of the optical indicatrix in the principal coordinate system. The elec-
tric polarization vectors of the O-waves are indicated in blue. The electric polarization
vectors of the E-waves are indicated in red.
In addition, E
e
(k
e
c) = 0, with k
e
the extraordinary wave vector. Hence, E
e
is
always in the plane dened by the wave vector k
e
and the unit vector o. Consequently,
the electric polarization vector of the E-wave is given by
E
e
= (k
e
o) H. (2.29)
The magnetic polarization vector H
e
is also tangent to the optical indicatrix and per-
pendicular with respect to k
e
, o and H.
The optical indicatrix is again depicted in Figure 2.5 including the electric polariza-
tion vectors of the O-waves and the E-waves. Since the polarization vectors of E and
H are tangent to the optical indicatrix for both the O-wave and the E-wave, the time-
averaged Poynting vector must be perpendicular to the optical indicatrix according to
Equation 2.15. As a result, the direction of the time-averaged Poynting vector is given
by
S = H. (2.30)
The polarization vectors of an electromagnetic plane wave are dened for a given di-
rection of propagation for both the O-wave and the E-wave. In the next section, we
investigate the amplitude and phase of the electromagnetic eld vectors.
12 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
2.6 Polarization ray tracing
Let us consider a plane boundary that forms the interface between two dierent trans-
parent media. We allow these media to have either homogeneous isotropic or homoge-
neous uniaxially anisotropic properties. This gives rise to four dierent kinds of inter-
faces, namely isotropic-isotropic, isotropic-anisotropic, anisotropic-isotropic and nally
anisotropic-anisotropic. We want to determine the electromagnetic eld of the trans-
mitted and reected waves as a function of the incident wave for each possible type of
interface. This concept is usually referred to as polarization ray tracing [13][14]. In
order to do this, we apply boundary conditions at the interface. The boundary condi-
tions demand that across the boundary the tangential components of E and H should
be continuous. In an isotropic medium, these boundary conditions are applied to two
independent modes. One mode is s-polarized, which means that the electric polarization
vector component is perpendicular to the plane of incidence. The plane of incidence is
dened by the normal n to the boundary and the direction of propagation

k. The other
mode is p-polarized, which means that the electric polarization vector component is in
the plane of incidence.
4
In anisotropic media, the boundary conditions are applied to
the O-wave and the E-wave separately.
In what follows, a method for solving the electromagnetic eld will be discussed. We
work out the case for an anisotropic-anisotropic interface and obtain the other three
cases as limit cases.
2.6.1 Anisotropic-anisotropic interface
In isotropic media, a wave can be regarded as the superposition of an s-polarized wave
and a p-polarized wave. These waves both propagate in the same direction and are
linearly polarized. Similarly, in anisotropic media, the O-wave and the E-wave are
linearly polarized but propagate independently along dierent directions.
Figure 2.6 shows the refracted and reected waves at an anisotropic-anisotropic in-
terface. In general, there are two reected waves, namely an O-wave and an E-wave,
indicated by R
ro
and R
re
, respectively. Similarly, there is a transmitted O-wave and a
transmitted E-wave, indicated by T
to
and T
te
, respectively.
The energy ux in each individual wave depends on the material properties and
the incident wave. The energy ux is determined by the electromagnetic eld which
can be calculated by applying the boundary conditions. In order to apply the boundary
conditions, we need the eld vectors dened in Equations 2.11 and 2.12. For convenience,
we choose the position vector r at the origin of the principal coordinate system, which
is exactly at the boundary surface (r=(0,0,0)).
The polarization vectors are determined by the equations in Section 2.5. The vectors
E
o
and E
e
will be normalized, yielding

E
o
and

E
e
, respectively. The vectors H
o
and H
e
can be calculated using

E
o
,

E
e
and Equation 2.13. Note that H
o
and H
e
are therefore not
normalized. In a similar way,

E
i
represents the normalized incident electric polarization
vector and H
i
represents the incident magnetic polarization vector. It is very important
to realize that

E
i
should represent a polarization eigen mode of the incident medium.
For this type of interface, this means that

E
i
represents either an O-wave or an E-wave.
4
The s-polarized and p-polarized waves are also known as transverse electric (TE) and transverse
magnetic (TM) waves, respectively.
c _ Koninklijke Philips Electronics N.V. 2007 13
TN-2007/00025 Unclassied
Figure 2.6: Refraction and reection at an anisotropic-anisotropic interface.
In an isotropic medium,

E
i
is always a polarization eigen mode of the medium so that

E
i
can be chosen arbitrarily.
In addition, we dene the Fresnel coecients as the complex amplitudes of the eld
vectors given by (see also Equation 2.10)
a = Ae
i
. (2.31)
It is the set of Fresnel coecients a
to
, a
te
, a
ro
and a
re
that needs to be solved.
2.6.2 Wave vectors in the plane of incidence
In order to calculate

E
o
and

E
e
, we need to determine k
o
and k
e
rst. In addition, if
we want to calculate k
o
and k
e
, we must rst know the incident wave vector k
i
and
the tangential wave vector k
tang
, which is the wave-vector component parallel to the
boundary surface. It is possible to derive these vectors from the incident time-averaged
Poynting vector S
i
), which is, just like

E
i
, an input vector of the model. In isotropic
media, the incident wave vector is k
i
= nS
i
), where S
i
) is normalized and k
i
is in units
of

c
. For anisotropic media, things are more complicated, since the optical indicatrix
is not a sphere, but the combination of a sphere and an ellipsoid of revolution. If the
incident wave is an ordinary wave, the incident wave vector yields k
io
= n
o
S
i
). If the
incident wave is an extraordinary wave, the incident wave vector can be calculated by
using the geometrical properties of the ellipsoid of revolution. According to Equation
2.30, there is exactly one position on the ellipsoid of revolution where the normalized
incident Poynting vector and H are parallel. This means that the dot product of S
i
)
and H reads
H S
i
) = [H[. (2.32)
We will use this expression to solve for the extraordinary wave vector k
ie
. In addition,
we demand that the end point of k
ie
lies on the surface H = 0. These two conditions
lead to an expression for the vector components of k
ie
given by
k
2
iej
=
n
4
e
S
ij
)
2
n
2
e
+ (n
2
o
n
2
e
)S
iz
)
2
, j = x, y, z. (2.33)
14 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Remember that these results are only valid in the principal coordinate system. If we
study k and S) more closely, we will nd that their vector components always have
the same sign in the principal coordinate system. Moreover, if we choose n
o
= n
e
= n,
Equation 2.33 reduces to k
i
= nS
i
), which applies for isotropic media.
The tangential wave vector k
tang
can be calculated by subtracting the normal com-
ponent from the incident wave vector, yielding
k
tang
= k
i
(k
i
n) n. (2.34)
The transmitted and reected wave vectors are determined by the demand that the
tangential components of the wave vectors should be continuous across the boundary.
This demand is also known as Snells law and in a generalized form given by
k
i
n = k n, (2.35)
where k is a refracted or reected wave vector. For the calculation of the refracted or
reected wave vectors we again use the geometrical properties of the optical indicatrix
in the principal coordinate system.
For the ordinary wave, it is easily derived that the expression for the refracted or
reected wave vector is given by
k
o
= k
tang

n
2
o
[k
tang
[
2
n. (2.36)
In Equation 2.36, the plus sign is applied to transmitted waves (indicated by index t)
and the minus sign is applied to reected waves (indicated by index r). For isotropic
media, we are allowed to use Equation 2.36 if we replace the ordinary index of refraction
n
o
by n.
Similarly, the extraordinary wave vector is given by
k
e
= k
tang
+ n, (2.37)
where is a constant. The constant can be determined by demanding that the endpoint
of the wave vector k
e
lies on the surface H = 0. As a result, is given by
=
B

B
2
4AC
2A
, (2.38)
with
A =
n
2
z
n
2
o
+
n
2
x
+ n
2
y
n
2
e
, (2.39)
B =
2k
tangz
n
z
n
2
o
+
2k
tangx
n
x
+ 2k
tangy
n
y
n
2
e
, (2.40)
C =
k
2
tangz
n
2
o
+
k
2
tangx
+ k
2
tangy
n
2
e
1. (2.41)
Again, the plus sign is applied to transmitted waves and the minus sign is applied to
reected waves. If n
o
= n
e
= n, reads
=

n
2
[k
tang
[
2
. (2.42)
By using equations 2.33, 2.34, 2.36 and 2.37, we can calculate all wave vectors that
are needed for the calculation of

E
o
and

E
e
. Note that all wave vectors, together with
n, are in the same plane, which is the plane of incidence.
c _ Koninklijke Philips Electronics N.V. 2007 15
TN-2007/00025 Unclassied
2.6.3 Fresnel equations
Since now

E
o
,

E
e
, H
o
and H
e
are known, the only thing left to do is solving the
Fresnel coecients. Before we can apply the boundary conditions, we need to dene two
orthogonal vectors t
s
and t
p
tangential to the interface given by
t
s
= k
i
n (2.43)
and
t
p
= n t
s
. (2.44)
From equations 2.43 and 2.44 it is clear that t
s
is the polarization vector of an s-polarized
wave (perpendicular with respect to the plane of incidence) and t
p
is a polarization vec-
tor of a p-polarized wave (which lies in the plane of incidence). The boundary conditions
are applied for both the s-component vectors and the p-component vectors of the elec-
tromagnetic eld vectors. Application of the boundary conditions yields four linear
equations. These equations are also called the Fresnel equations [15][16][17][18] and are
given by
t
s
(a
to

E
to
+ a
te

E
te
) = t
s
(

E
i
+ a
ro

E
ro
+ a
re

E
re
),
t
p
(a
to

E
to
+ a
te

E
te
) = t
p
(

E
i
+ a
ro

E
ro
+ a
re

E
re
),
t
s
(a
to
H
to
+ a
te
H
te
) = t
s
(

H
i
+ a
ro
H
ro
+ a
re
H
re
), (2.45)
t
p
(a
to
H
to
+ a
te
H
te
) = t
p
(

H
i
+ a
ro
H
ro
+ a
re
H
re
).
Note that the only unknowns in the equations are the Fresnel coecients a
to
, a
te
, a
ro
and a
re
. These equations can be rewritten as a linear matrix equation given by

t
s


E
to
t
s


E
te
t
s


E
ro
t
s


E
re
t
p


E
to
t
p


E
te
t
p


E
ro
t
p


E
re
t
s
H
to
t
s
H
te
t
s
H
ro
t
s
H
re
t
p
H
to
t
p
H
te
t
p
H
ro
t
p
H
re

a
to
a
te
a
ro
a
re

t
s


E
i
t
p


E
i
t
s


H
i
t
p


H
i

. (2.46)
This matrix equation can be solved by any of the standard methods as e.g. described in
Numerical Recipes [19].
The advantage of Equation 2.46 is that it is also applicable to the remaining types
of interfaces. Consider for example an isotropic-anisotropic interface. For this type of
interface, a
ro
and a
re
need to be replaced by a
rs
and a
rp
, respectively. In other words,
in isotropic media, a wave is split into an s-polarized wave and a p-polarized wave, while
both waves propagate along the same direction. In addition,

E
ro
,

E
re
, H
ro
and H
re
are
replaced by

E
rs
,

E
rp
, H
rs
and H
rp
, respectively. Of course, equations 2.26 and 2.29 no
longer apply for the reected waves. Instead, we dene the electric polarization vector
of the s-polarized wave

E
rs
=
k
r
n
[k
r
n[
(2.47)
and the electric polarization vector of the p-polarized wave is dened as

E
rp
=

E
rs
k
r
[

E
rs
k
r
[
. (2.48)
16 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
For an anisotropic-isotropic interface, equations 2.47 and 2.48 apply as well provided
that the index r is replaced by the index t, resulting in the set of Fresnel coecients
a
ts
, a
tp
, a
ro
and a
re
. For isotropic-isotropic interfaces, equations 2.47 and 2.48 apply for
both the index r and the index t, yielding a
ts
, a
tp
, a
rs
and a
rp
.
2.6.4 Total reection
According to Equation 2.31, the Fresnel coecients are complex numbers with amplitude
A and argument . Hence, according to Equation 2.46, the incident electric eld vector

E
i
or the polarization vectors can be complex too, while t
s
and t
p
are always real.
If = N with NZ, the incident wave is linearly polarized and

E
i
is real. If

E
i
is
complex and = (2N + 1)

2
with NZ, the incident wave is circularly polarized. All
the remaining values for correspond to elliptically polarized waves. However, it is not
possible to dene such incident waves in anisotropic media, since the incident wave must
be a polarization eigen mode of the medium. The latter should either be a linearly
polarized O-wave or a linearly polarized E-wave.
According to Equation 2.19, polarization vectors in non-absorbing anisotropic media
are also real. However, polarization vectors can be complex if the corresponding wave
vectors (see Equation 2.26 and Equation 2.29) are complex vectors
5
. Let us investigate
the case for which a wave vector is complex. Consider Equation 2.36. If [k
tang
[ > n
o
,
we can rewrite Equation 2.36 as
k
o
= k
tang
ik
I
n, (2.49)
with
k
I
=

[k
tang
[
2
n
2
o
. (2.50)
For reected waves, [k
tang
[ never exceeds n
o
. Hence, only the plus sign applies in Equa-
tion 2.49. Physically, this means that the wave is totally reected. Nevertheless the
electromagnetic eld in the second medium does not disappear, although there is no
longer a ow of energy across the boundary. If we substitute Equation 2.49 into Equa-
tion 2.11, we obtain
E =

Ee
k
I
nr
e
i(k
tang
rt)
. (2.51)
Equation 2.51 represents a wave which is propagating along the interface in the plane of
incidence and which decreases exponentially with the normal distance from the interface.
This type of wave is called an evanescent wave [1]. The eective depth of penetration is
in the order of
c
n
=

2
, which is of the order of a wavelength. The wave described by
Equation 2.51 applies for waves both in isotropic and anisotropic media in the case of
total reection.
It can be concluded that the Fresnel coecients are complex when total reection
occurs.
2.6.5 Transmittance and reectance
By using the theory in the previous sections, we are now able to calculate the electro-
magnetic eld at the interface. For example, the electromagnetic eld of a reected wave
5
A complex vector is a vector whose elements are complex numbers, i.e. v = (v
x
+iv

x
, v
y
+iv

y
, v
z
+
iv

z
) = (v
x
, v
y
, v
z
) + i(v

x
, v

y
, v

z
).
c _ Koninklijke Philips Electronics N.V. 2007 17
TN-2007/00025 Unclassied
in an isotropic medium is given by
E
r
= (a
rs

E
rs
+ a
rp

E
rp
)e
i(k
r
rt)
, H
r
= (a
rs
H
rs
+ a
rp
H
rp
)e
i(k
r
rt)
, (2.52)
while the electromagnetic eld of a transmitted extraordinary wave is given by
E
te
= a
te

E
te
e
i(k
te
rt)
, H
te
= a
te
H
te
e
i(k
te
rt)
. (2.53)
From the electromagnetic eld, we can calculate the time-averaged Poynting vector using
Equation 2.15.
Next to the electromagnetic eld itself, we are also interested in the energy ux of
each individual transmitted or reected wave, with respect to the incident energy ux
of the incident wave. In order to investigate this, we apply the law of conservation of
energy in the direction of the normal n. For an anisotropic-anisotropic interface, this
yields
n S
i
) = n S
to
) + n S
te
) n S
ro
) n S
re
). (2.54)
This equation can be transformed into
n S
to
)
n S
i
)
+
n S
te
)
n S
i
)

n S
ro
)
n S
i
)

n S
re
)
n S
i
)
= 1. (2.55)
Each term on the left side of Equation 2.55 represents either the transmittance T or the
reectance R. Consequently, Equation 2.55 can be written as
T
o
+ T
e
+ R
o
+ R
e
= 1, (2.56)
with
T
o
= [
n S
to
)
n S
i
)
[, T
e
= [
n S
te
)
n S
i
)
[, R
o
= [
n S
ro
)
n S
i
)
[, R
e
= [
n S
re
)
n S
i
)
[. (2.57)
For the remaining types of interfaces, we obtain similar results. The dierence is that
for an isotropic-anisotropic interface, we use T
o
, T
e
and R. For an anisotropic-isotropic
interface, we use R
o
, R
e
and T. For isotropic-isotropic interfaces, we simply use R and
T.
2.7 OpIn model
Starting with the Maxwell equations, we derived a model based on geometrical optics,
which enables us to calculate the transmitted and reected waves at an homogeneous
anisotropic-anisotropic interface. In addition, we have come to the conclusion that the
optical properties of isotropic media are included by the optical modelling of uniaxially
anisotropic media. Hence, the model is applicable to all four possible types of interface.
The model makes use of the geometrical properties of the optical indicatrix rather
frequently. For this reason, if we refer to the model in the coming sections, we henceforth
call it the OpIn model (referring to the optical indicatrix).
Figure 2.7 shows a scheme with the input parameters and the output parameters of
the OpIn model. The input parameters contain the indices of refraction n
o
, n
e
and the
optical axis o of the media on either side of the interface. The normal n to the interface,
18 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Figure 2.7: Scheme containing the input and output parameters of the OpIn model.
the time-averaged incident Poynting vector S
i
) and the complex amplitude

E
i
of the
incident electric eld vector are also input parameters.
The output of the OpIn model contains the electromagnetic eld of the transmitted
and reected waves, including the wave vectors, the complex amplitudes of the electric
and magnetic eld vectors and the time-averaged Poynting vectors. In addition, the
OpIn model can determine the polarization state and phase shift from the complex am-
plitudes of the electric or magnetic eld vectors. The phase shift is dened as the change
in phase of an s-polarized, p-polarized, ordinary or extraordinary wave. The polarization
state is determined by the phase dierence between an s-polarized and p-polarized wave.
Finally, the model can calculate the relative energy ux of each individual wave in terms
of transmittance T and reectance R.
The OpIn model can dene just one interface, while ray tracing programs in general
can dene whole optical systems. Therefore, the OpIn model can be regarded as a
fundamental building block. Consequently, it provides us an opportunity to extend the
range of applications of existing ray tracing programs. In Chapter three, we will discus
ray tracing performed by a ray-tracing algorithm based on the OpIn model. In addition,
we present some typical results which will be discussed thoroughly.
c _ Koninklijke Philips Electronics N.V. 2007 19
TN-2007/00025 Unclassied
20 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Section 3
Numerical ray tracing of
anisotropic media
Ray tracing is a technique to determine the actual path of individual rays of light through
an optical system. In a homogeneous medium, each ray can be traced, independently,
using the OpIn model described in Chapter 2. With the help of a ray-tracing algorithm,
the necessary calculations yielding the changes in direction of a ray is done more easily
and quickly.
In the previous chapter, both the terms rays and waves have been used when referring
to plane waves. In order to clarify any obscurities in the use of these terms, we apply
the following denitions.
A monochromatic plane wave is a constant-frequency wave whose wave fronts are
innite parallel planes of constant amplitude and phase. The wave vector is perpen-
dicular to these wave fronts at any position on the wave fronts and denes the (local)
direction of propagation of a plane wave. The path which describes the propagation of
a plane wave is dened as a ray. A ray is therefore an articial line which describes the
orthogonal trajectory of a (plane)
1
wave through a medium or optical system.
In contrast with isotropic media, the Poynting vectors and wave vectors in anisotropic
media have dierent directions. Therefore, it is important to realize that the ray path
of the Poynting vector is dierent from the ray path of the wave vector.
In this chapter, we give a brief description of the structure of the ray-tracing algo-
rithm based on the OpIn model. Additionally, we present and evaluate some typical
results. Finally, we compare the results of a test case with similar results of the com-
mercial ray-tracing program ASAP.
3.1 Ray-tracing algorithm
The structure of the algorithm designed for ray tracing is depicted in Figure 3.1 and
Figure 3.2. Figure 3.1 shows the selection procedure for determining the type of in-
terface depending on the input parameters n
o
and n
e
. Subsequently, the algorithm
schematically depicted in Figure 3.2 shows individual treatment for each possible type
1
The denition of a ray is not restricted to plane waves, but applies to any kind of wave, for example
a spherical wave.
c _ Koninklijke Philips Electronics N.V. 2007 21
TN-2007/00025 Unclassied
Figure 3.1: Scheme for determining the type of interface.
Figure 3.2: Scheme of the algorithm based on the OpIn model.
22 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
of interface. The ray tracing algorithm is designed in this way, so that errors can be
detected suciently. For example, if an error occurs in the branch for isotropic-isotropic
interfaces, modications can be made without harming the functionality of the other
branches.
Again, we emphasize the fact that the formulas for the wave vectors dened in the
OpIn model are only valid in the principal coordinate system. Therefore, the input
vectors S
i
), o and n, dened in a global coordinate system, need to be transformed
to the principal coordinate system before the formulas of section 2.6.2 can be applied.
When the wave vectors have been calculated, they will be transformed back to the global
coordinate system, together with the input vectors. In this way, it is possible to trace
rays in the global coordinate system and simultaneously maintain the simplicity of the
formulas of the OpIn model in the principal coordinate system.
Subsequently, the complex amplitudes and the time-averaged Poynting vectors are
calculated, using the wave vectors and the input vectors. Finally, the transmittance and
reectance are calculated and, optionally, the polarization states and phase shifts.
Remarkably, Equation 2.56 appears to be highly indicative of errors. If the sum
of the transmittance and reectance diers from unity, some error must have occurred
in one of the calculations. If we consider the sequence of equations used in the OpIn
model, it is not dicult to understand why Equation 2.56 is so sensitive to an error in
the calculations. Apparently, each parameter which needs to be calculated depends on
another, directly or indirectly. For example, the incident wave vector determines the tan-
gential and refracted or reected wave vectors. These wave vectors then determine the
polarization vectors. The latter determine the Fresnel coecients, which determine the
Poynting vectors and nally the transmittance and reectance through equation 2.55.
Because of this successive parameter dependence, any error occurring in the calculations
results in an error in equation 2.56. Hence, the calculations are correct if the sum of the
transmittance and reectance is exactly 1.
In the next section, we use the ray-tracing algorithm to reproduce some typical results
of geometrical optics in order to gain condence in the use of the OpIn model.
3.2 Phase shift and reectance of internally reected waves
Consider a glass-air interface where the glass has an index of refraction n
glass
= 1.5 and
air an index of refraction n
air
= 1. For every incident wave in the glass, there exists a
transmitted wave and an internally reected wave. We use the OpIn model to calculate
the phase shift of the internally reected waves as a function of the angle of incidence.
Figure 3.3 shows the phase shift of the s-polarized (red curve) and p-polarized waves
(green curve) as a function of the angle of incidence
i
.
From Figure 3.3 we notice two special angles of incidence. At
i
= 33.7

, the phase
shift of the p-polarized wave changes abruptly from to 0. At this specic angle,
internally reected p-polarized waves vanish and the reected waves become solely s-
polarized. This angle is called Brewsters angle and is given by
p
= tan
1
(
n
air
n
glass
). At

i
= 41.8

, total internal reection occurs. According to Snells law, the critical angle
for total internal reection is
c
= sin
1
(
n
air
n
glass
).
In accordance with the arguments of Subsection 2.6.4, the phase shifts below 41.8

c _ Koninklijke Philips Electronics N.V. 2007 23


TN-2007/00025 Unclassied
Figure 3.3: Phase shift after reection at a glass-air interface as a function of the angle of
incidence
i
. The p indicates the phase shift for the p-polarized waves (in green) and the
s indicates the phase shift for the s-polarized waves (in red). The blue curve indicates
the phase dierence between the s-polarized and p-polarized waves.
Figure 3.4: The reectance R for a glass-air interface as a function of the angle of
incidence
i
. The reectance for the p-polarized wave is indicated in green and the
reectance for the s-polarized wave is indicated in red. The incident light is linearly
polarized under an angle of 45

with the plane of incidence.


24 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
represent real Fresnel coecients, while the phase shifts above 41.8

represent complex
Fresnel coecients.
Figure 3.3 also shows the phase dierence between the s-polarized and p-polarized
waves (blue curve). At an angle of about 53

, the phase dierence of a reected wave is

4
radians. Two consecutive internal reections at this angle produce a phase dierence
of

2
radians, resulting in a circularly polarized wave. This concept is utilized in the
Fresnel rhomb (see Appendix B.1).
Figure 3.4 shows the reectance R for a glass-air interface for incident light which is
linearly polarized under an angle of 45

with the plane of incidence. At an angle of 33.7

,
the reectance for the p-polarized waves vanish which is in accordance with Brewsters
law. Additionally, the critical angle for total internal reection yields 41.8

correspond-
ing to the critical angle of 41.8

depicted in Figure 3.3. Since the incident light is linearly


polarized under an angle of 45

with the plane of incidence, the reectance of each totally


reected component is R = 0.5. Hence, the total reectance for total internally reected
light yields 1.
Figure 3.3 exactly matches the gure in Pedrotti, page 415, Figure 20-6 [2]. As a
result, it can be concluded that the OpIn model can correctly reproduce results from geo-
metrical optics related topics as treated in standard books on optics [1][2][3]. Obviously,
this gives condence in the use of the model.
3.3 Phase shift and reectance of isotropic-anisotropic in-
terfaces
A second example of ray tracing in the geometrical optics approach for an articial
isotropic-anisotropic interface will be evaluated in this section. The index of refraction
of medium 1 is n
glass
= 1.5. The ordinary and extraordinary index of refraction of
medium 2 are n
o
= 1 and n
e
= 1.2, respectively. We dene the plane of incidence as
the yz-plane, the surface normal n = (0, 0, 1) and the optical axis o =
1
2

2(1, 0, 1). In
addition, we dene the incident light to be p-polarized.
Figure 3.5 shows the reectance of the reected waves. We can clearly see the
presence of s-polarized light in the reected light, despite the fact that the incident light
is p-polarized. Apparently, the anisotropic medium induces an s-component due to the
orientation of the optical axis. Obviously, if the optical axis would be in the plane of
incidence, the reected waves would be 100% p-polarized.
The phase shift of the reected waves is depicted in Figure 3.6. Both Figures 3.5
and 3.6 indicate Brewsters angle (
p
= 37.6

) and the critical angles for total internal


reection for the O-wave and the E-wave (
co
= 41.8

and
ce
= 53.2

).
An example of a real crystalline system is the mineral chiolite (Na
5
Al
3
F
14
). The
ordinary and extraordinary refractive indices of chiolite are n
o
= 1.349 and n
e
= 1.342
in a tetragonal crystal system [20]. Since n
o
> n
e
, the crystal is uniaxially negative.
Figure 3.7 shows the reectance and transmittance of a glass-chiolite interface. The in-
cident light is p-polarized and the optical axis is under an angle of 45

with the plane of


incidence. The optical axis induces s-polarized reected light close to the critical angles
for total internal reection. In contrast with positive uniaxially anisotropic media, the
critical angle
co
exceeds the critical angle
ce
.
c _ Koninklijke Philips Electronics N.V. 2007 25
TN-2007/00025 Unclassied
Figure 3.5: The reectance R for an isotropic-anisotropic (n
glass
= 1.5, n
o
= 1 and
n
e
= 1.2) interface as a function of the angle of incidence
i
. The incident light is
p-polarized and the optical axis is under an angle of 45

with the plane of incidence.


Figure 3.6: Phase shift after reection of p-polarized light at an isotropic-anisotropic
interface (n
glass
= 1.5, n
o
= 1, n
e
= 1.2, n = (0, 0, 1) and o =
1
2

2(1, 0, 1)).
26 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Figure 3.7: Reectance and transmittance of a glass-chiolite interface for p-polarized
light (n
glass
= 1.5, n
o
= 1.349, n
e
= 1.342, n = (0, 0, 1) and o =
1
2

2(1, 0, 1)).
The graphs like the ones depicted in Figures 3.5, 3.6 and 3.7 are easily and quickly
obtained when we use the ray tracing algorithm based on the OpIn model. In the next
section, we will compare similar results with results from the commercial ray tracing
program ASAP and discuss possible dierences.
3.4 ASAP as a bench mark
ASAP (Advanced Systems Analysis Program) is a software package suitable for sim-
ulating the physics of optical systems [21]. For the last twenty years, ASAP (among
other) has been the industry standard in optical engineering software. For this reason,
ASAP can be considered a fair bench mark for the evaluation of novel developments in
the modelling of optical systems. In this section, we introduce a simple test case and
compare the results obtained by ASAP and the OpIn model.
Consider an isotropic-anisotropic interface. Medium 1 has an index of refraction
n = 1.5 and medium 2 has an ordinary index of refraction n
o
= 1 and an extraordinary
index of refraction n
e
= 1.5. The optical axis is under an angle of 45

with the surface


normal n and the incident light is linearly polarized under an angle of 45

with the plane


of incidence. For this conguration, we will investigate the transmittance and reectance
as a function of the angle of incidence.
Figure 3.8 shows the results obtained by ASAP version 2005. The transmittance for
the ordinary wave vanishes at 41.8

, which is exactly the critical angle for total internal


reection at a glass air interface as described in section 3.2. However, the curves which
represent the transmittance of the extraordinary waves and the reectance show some
c _ Koninklijke Philips Electronics N.V. 2007 27
TN-2007/00025 Unclassied
Figure 3.8: The results for the transmittance (T) and reectance (R) at an isotropic-
anisotropic interface, according to ASAP version 2005.
Figure 3.9: The results for the transmittance (T) and reectance (R) at an isotropic-
anisotropic interface, according to ASAP version 2006.
28 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Figure 3.10: The results for the transmittance (T) and reectance (R) at an isotropic-
anisotropic interface, according to the OpIn model. The red, green and blue curves
represent the ordinary, extraordinary and reected waves, respectively. The black line
represents the sum of T
o
, T
e
and R.
peculiar features. Moreover, the sum of the transmittance and reectance vanishes for
a range of angles between 50

and 60

. The latter is physically impossible, suggesting


that ASAP has some deciencies in the modelling of anisotropic media.
Figure 3.9 shows the results obtained by ASAP version 2006. We can immediately
conclude that the results of versions 2005 and 2006 are not consistent. Moreover, the
results are not correct, since we detect an interval of zero intensity between 50

and 60

.
The results for the transmittance and reectance according to the OpIn model are
depicted in Figure 3.10. According to the gure, the critical angle
co
for total internal
reection of the ordinary wave is 41.8

. In addition, the gure shows a second critical


angle
ce
for the extraordinary wave and has a value of 58.2

. More importantly, the


total intensity equals 1 for all angles of incidence. As a whole, Figure 3.10 shows more
solid results than the results represented by ASAP 2005 and ASAP 2006.
It can be concluded that the results of ASAP show a number of defects, while the
OpIn model produces plausible results.
c _ Koninklijke Philips Electronics N.V. 2007 29
TN-2007/00025 Unclassied
30 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Section 4
Conclusions
In Chapter 2, we introduced the OpIn model which can calculate the electromagnetic
eld of refracted or reected waves in the geometrical optics approach at homogeneous
uniaxially anisotropic interfaces. In Section 2.5, we derived concise expressions for the
electric and magnetic polarization vectors in anisotropic media. In Section 2.6, we de-
rived a method to calculate the Fresnel coecients (i.e. the complex amplitudes of
the electric and magnetic eld vectors). The OpIn model can calculate phase shifts,
transmittance and reectance. Finally, we have come to the conclusion that the optical
properties of isotropic media are included by the OpIn model.
A ray-tracing algorithm based on the OpIn model is introduced in Chapter 3. The
ray-tracing algorithm is applied to a test case, concerning an isotropic-isotropic inter-
face. It appears that the phase shift and reectance of reected waves at a glass-air
interface exactly match the results claimed in standard books on geometrical optics.
A second example of an isotropic-anisotropic interface produces plausible results. In a
third test case, another isotropic-anisotropic interface is investigated. Results obtained
by the OpIn model are compared with results obtained by the commercial ray-tracing
program ASAP. It appears that the results obtained by ASAP show a number of defects,
while the OpIn model produces plausible and solid results. Altogether, these results give
condence in the use of the OpIn model and the corresponding ray-tracing algorithm.
Optical design and analysis software programs like ASAP [21], CODE V [22], ZE-
MAX [23], FRED [24], Lightools [25] and TracePro [26] all claim to be able to ray trace
uniaxially anisotropic media. Surprisingly, ASAP cannot full this claim completely. It
would be interesting to see whether these optical engineering programs can full their
claim or not. This is a useful assignment for the near future.
Being able to ray trace homogeneous uniaxially anisotropic media, we go one step
further and develop a model which can ray trace inhomogeneous uniaxially anisotropic
media, where the direction of the optical axis depends on the position inside the medium.
This will not only extend the functionality of the existing OpIn model, but it will also be
a unique tool for the optical modelling of inhomogeneous uniaxially anisotropic media
in contrast with existing optical engineering software.
c _ Koninklijke Philips Electronics N.V. 2007 31
TN-2007/00025 Unclassied
32 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Acknowledgement
I would like to thank Dick de Boer, Joseph Braat, Thomas Kraan, Jan Stromer, Gilles
Vissenberg, Fetze Pijlman, Frank van Abeelen, Wilbert IJzerman and Siebe de Zwart
for useful discussions.
c _ Koninklijke Philips Electronics N.V. 2007 33
TN-2007/00025 Unclassied
34 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
References
[1] M. Born and E. Wolf, Principles of Optics, sixth edition, Pergamon Press, chapter
3, 109, chapter 16, 665(1986)
[2] F.L. Pedrotti and L.S. Pedrotti, Introduction to optics, second edition, Prentice-Hall
International (1996)
[3] E. Hecht, Optics, third edition, Addison-Wesley (1998)
[4] P. Yeh and C. Gu, Optics of Liquid Crystal Displays, John Wiley & Sons (1999)
[5] M. Sluijter, Ray-Optics Analysis of Switchable Auto-Stereoscopic Lenticular-Based
2D/3D Displays, Technical Note TN-2005/01134, (2005)
[6] R.A.M. Hikmet, T. van Bommel and T.C. Kraan, Switchable Liquid Crystal Graded
Refractive Index Arrays for Beam Control, Technical Note PR-TN 2006/00841,
(2006)
[7] J.D. Jackson, Classical Electrodynamics, third edition, John Wiley & Sons (1999)
[8] M. Kline and I.W. Kay, Electromagnetic Theory and Geometrical Optics, Pure and
Applied Mathematics Series, Vol. XII, John Wiley & Sons, (1965)
[9] Yury A. Kravtsov and Yury I. Orlov, Geometrical optics of inhomogeneous media,
Springer-Verlag (1990)
[10] P. Yeh and A. Yariv, Electromagnetic propagation in periodic stratied media. II.
Birefringence, phase matching, and x-ray lasers, J. Opt. Soc. Am., Vol. 67, No. 4,
438 (1977)
[11] L. Fletcher, The optical indicatrix and the transmission of light in crystals, Oxford
University Press, London, (1892)
[12] James G. Lunney and Denis Weaire, The ins and outs of conical refraction, Euro-
physics News, Vol. 37, No. 3, 26 (2006)
[13] James D. Trollinger, Jr., Russel A. Chipman and Donald K. Wilson, Polarization
ray tracing in birefringent media, Optical Engineering, Vol. 30, No. 4, 461-466,
(1991)
[14] Russel A. Chipman, Mechanics of polarization ray tracing, Optical Engineering,
Vol. 34, No. 6, 1636-1645, (1995)
c _ Koninklijke Philips Electronics N.V. 2007 35
TN-2007/00025 Unclassied
[15] Maria C. Simon and Liliana I. Perez, Reection and tranmission coecients in
uniaxial crystals, Journal of Modern Optics, Vol. 38, No. 3, 503-518 (1991)
[16] Stephen C. McClain, Lloyd W. Hillman and Russel A. Chipman, Polarization ray
tracing in anisotropic optically active media. I Algorithms, J. Opt. Soc. Am. A, Vol.
10, No. 11, 2371 (1993)
[17] Stephen C. McClain, Lloyd W. Hillman and Russel A. Chipman, Polarization ray
tracing in anisotropic optically active media. II Theory and physics, J. Opt. Soc.
Am. A, Vol. 10, No. 11, 2383 (1993)
[18] Horst Greiner, Ray tracing algorithms for modelling light propagation in birefringent
media, PFL-Aachen Report 1632/2003, (2003)
[19] W. Press et al., Numerical recipes in FORTRAN, The art of Scientic Computing,
second edition, Cambridge University Press, (1992)
[20] David R. Lide, Handbook of Chemistry and Physics, 86
th
edition, Taylor & Francis
Group, 4-141 - 4-147, (2005-2006)
[21] ASAP, www.breault.com, (2006)
[22] CODE V, www.opticalres.com, (2006)
[23] ZEMAX, www.zemax.com, (2006)
[24] FRED, www.photonengr.com/fred/, (2007)
[25] Lightools, www.lightools.com, (2007)
[26] TracePro, www.lambdares.com , (2007)
36 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
Appendix A
A.1 Plane waves in (in)homogeneous media
The Maxwell equations combined with the material equations can be transformed into

2
E

c
2

2
E
t
2
+ (ln) E+(E ln) = 0, (A.1)
with a similar expression for the magnetic eld H [1]. If the medium is homogeneous,
ln = 0 and ln = 0. Hence, Equation A.1 reduces to

2
E

c
2

2
E
t
2
= 0. (A.2)
Equation A.2 is a standard equation of wave motion and suggest the existence of elec-
tromagnetic waves propagating with a velocity
v =
c

. (A.3)
The most simple solution for the electromagnetic wave is the time harmonic plane wave,
given by Equation 2.7.
If the medium is inhomogeneous [9] and non-magnetic, we nd ln = 0, but the
expression ln does not necessarily vanish. It can be shown that
ln
[n[
n
. (A.4)
If we assume that
|n|
n
1, then

2
E

c
2

2
E
t
2
0. (A.5)
We can assume that the wave eld may locally be represented by a quasi-plane wave
given by
E = E
0
(r)e
i(r)
. (A.6)
Knowing that
n
c
=
2n

0
1, we can conclude that Equation A.6 only applies if
[n[
n

2

, (A.7)
c _ Koninklijke Philips Electronics N.V. 2007 37
TN-2007/00025 Unclassied
with =

0
n
. This inequality can be more simplied by introducing a characteristic
length L. Let us dene the distance L as the distance over which the increment of n is
comparable with the value if n itself. Then, an estimate for L is L
n
|n|
. Substituting
this expression into Equation A.7 yields

2L
1. (A.8)
Hence, it can be concluded that the quasi-plane wave is a solution of the wave equation
if the properties of the medium change slowly over the wavelength.
A.2 Polarization eigen modes
In order to derive Equation 2.19 and 2.20, we consider Equation 2.16. By using a well-
known vector identity and Equation 2.9, we can rewrite Equation 2.16 as

2
c
2
n
2
[E(E

k)

k] +

2
c
2
E = 0. (A.9)
This equation can simply be written as
n
2
E

= D, (A.10)
where E

denotes the vector component of E perpendicular to



k in the plane of E and

k. From Equation A.9, we can deduce the vector components of E and these are given
by
E
i
=
n
2
(E

k)

k
i
n
2

i
, i = x, y, z, (A.11)
yielding Equation 2.19 apart from a factor n
2
E

k. Multiplying E
i
by

k
i
, adding the
resulting three equations and dividing the resulting expression by the common factor
E

k yields

k
2
x
n
2

x
+

k
2
y
n
2

y
+

k
2
z
n
2

z
=
1
n
2
. (A.12)
Substituting Equation 2.6 into Equation A.12 yields Equation 2.20.
A.3 Ordinary polarization vector
For uniaxially anisotropic media, the ordinary wave has an eective index of refrac-
tion n = n
o
. Equation A.11 does not apply since the denominators n
2

i
vanish.
Consequently, we consider Equation A.9, which reduces to
(E
o

k)

k = 0. (A.13)
Equation A.13 implies that the electromagnetic plane wave is transverse, which is also
the case in isotropic media. Equation A.13 provides more than one solution for the
38 c _ Koninklijke Philips Electronics N.V. 2007
Unclassied TN-2007/00025
polarization vector E
o
. We use Equation 2.18 to nd the right solution. By using
k
o
=

c
n
o

k, Equation 2.18 can be rewritten as

k
2
x

k
x

k
y

k
x

k
z

k
y

k
x

k
2
y

k
y

k
z

k
z

k
x

k
z

k
y
n
2
e
n
2
o

k
2
x

k
2
y

E
o
= 0. (A.14)
According to this equation, the ordinary electric polarization vector is given by
E
o
=

k
y

k
x
0

. (A.15)
A.4 Extraordinary polarization vector
Given Equation 2.27, the gradient vector operator applied to the scalar function H yields
H(k
x
, k
y
, k
z
) = (
2n

k
x
n
2
e
,
2n

k
y
n
2
e
,
2n

k
z
n
2
o
). (A.16)
This vector is by denition perpendicular with respect to the surface H = 0. The dot
product of E
e
and H is given by
H E
e
=
2n
(n
2
n
2
o
)(n
2
n
2
e
)
H = 0. (A.17)
Even if n = n
o
or n = n
e
, we apply lHopitals rule and still nd that the dot product of
Equation A.17 vanishes. From this we can conclude that the electric polarization vector
E
e
is tangent to the optical indicatrix.
c _ Koninklijke Philips Electronics N.V. 2007 39
TN-2007/00025 Unclassied
Appendix B
B.1 Fresnel rhomb
Figure B.1 shows the concept of a Fresnel rhomb. The incident light is linearly polarized
at an angle of 45

with the plane of incidence. After entering the rhomb at normal


incidence, the light is totally reected twice before leaving the rhomb again at normal
incidence. The two reections induce a phase dierence between the s-component and
p-component of the light. For glass with an index of refraction n
glass
= 1.5, the angle
depicted in Figure B.1 must be approximately 53

in order to induce a phase dierence


of

4
radians. As a result, two reections induce a phase dierence of

2
radians, yielding
circularly polarized light.
Figure B.1: The Fresnel rhomb produces circularly polarized light when the incident
light is linearly polarized under an angle of 45

with the plane of incidence.


40 c _ Koninklijke Philips Electronics N.V. 2007

Potrebbero piacerti anche