Ellipsometry: Alborg Niversity

A ALBORG U NIVERSITY
Institute of Physics and Nanotechnology

Pontoppidanstræde 103 - 9220 Aalborg Øst - Telephone 96 35 92 15
TITLE:
Ellipsometry
SYNOPSIS:
This project concerns measurement of the re-
fractive index of various materials and mea-
PROJECT PERIOD: surement of the thickness of thin films on sili-
September 1st - December 21st 2004 con substrates by use of ellipsometry. The el-
lipsometer used in the experiments is the SE
850 photometric rotating analyzer ellipsome-
ter from Sentech.
THEME: After an introduction to ellipsometry and a
Detection of Nanostructures problem description, the subjects of polar-
ization and essential ellipsometry theory are
covered.
PROJECT GROUP: The index of refraction for silicon, alu-
116 minum, copper and silver are modelled us-
ing the Drude-Lorentz harmonic oscillator
model and afterwards measured by ellipsom-
etry. The results based on the measurements
GROUP MEMBERS: show a tendency towards, but are not ade-
Jesper Jung quately close to, the table values. The mate-
Jakob Bork rials are therefore modelled with a thin layer
of oxide, and the refractive indexes are com-
Tobias Holmgaard
puted. This model yields good results for the
Niels Anker Kortbek refractive index of silicon and copper. For
aluminum the result is improved whereas the
result for silver is not.
SUPERVISOR: The thickness of a thin film of SiO2 on a sub-
strate of silicon is measured by use of ellip-
Kjeld Pedersen
sometry. The result is 22.9 nm which deviates
from the provided information by 6.5 %.
The thickness of two thick (multiple wave-
NUMBERS PRINTED: 7 lengths) thin polymer films are measured.
The polymer films have been spin coated on
REPORT PAGE NUMBER: 70 substrates of silicon and the uniformities of
the surfaces are investigated. The two poly-
TOTAL PAGE NUMBER: 132 mer films are measured to be 3.32 and 4.17 µm
on average with a maximum deviation from
the mean across the surface of 3.0 % and 4.2 %
respectively.
Preface
This project is written by group 116 at the Institute of Physics and Nanotechnology, Aal-
borg University, in the period between September 1st 2004 and December 21st 2004. The
theme for the 7th semester project on Applied Physics is ”Detection of Nanostructures” un-
der which the project ”Ellipsometry” has been chosen.
The main report consists of three parts. The Preanalysis, the Ellipsometry Theory and the
Simulations and Experiments. The three parts begins with a short introduction and consists
of several chapters each starting with a description of the contents. The main report treats
subjects that have direct relevance to ellipsometry and the ellipsometric measurements per-
formed during this project.
In addition to the main report several subjects are treated in appendixes which can be
found after the bibliography. The appendixes treat subjects of general theory of light waves,
optics and properties of materials. Also included in the appendixes are some of the more
extensive calculations and test descriptions of the performed tests.
Enclosed on the inside of the back page is the project CD, containing test data, source
code etc. References to files on this CD e.g. Matlab programs, are written in square
brackets containing the path and filename with extension like this: [CD 2004, matlab/poly-
mer_thickness/film_thickness_polymer.m]. References to other sources are written in simi-
lar manner, where the author and year of publishing refer to the bibliography, where further
information about the source can be found. References to a source can be found at the end
of a section, where the use of the source ends. The reference to a source might also contain a
page number if necessary, e.g. [Azzam & Bashara 1977, p. 245]. A citation before a punctu-
ation means that the reference is to the sentence only. A citation after a punctuation means
that the reference is related to the whole paragraph.
Regarding the notation throughout the report it has been decided to use [Klein & Furtak
1986] as a general guideline, e.g. vectors will be boldfaced E and matrixes will be boldfaced
Arial A. In the cases that [Klein & Furtak 1986] does not cover the notation, the notation of
the used source is adapted. The symbol j will be used as the imaginary unit.
Aalborg University, Tuesday 21st 2004
III
Jesper Jung Jakob Bork
Tobias Holmgaard Niels Anker Kortbek
IV
Contents
Introduction 1
I Preanalysis 3
1 Introduction to Ellipsometry 5
2 Problem Description 7
II Ellipsometry Theory 9
3 Polarization of Light 11
3.1 Definition of Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Polarization Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Definition of the ellipsometric parameters ψ and Δ . . . . . . . . . . . . . . . . 16
4 Ellipsometer Systems 21
4.1 Description of an Ellipsometer . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Different Ellipsometer Configurations . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Determination of ψ and Δ with a Static Photometric RAE . . . . . . . . . . . . 25
4.4 Description of the Sentech SE 850 Ellipsometer . . . . . . . . . . . . . . . . . . 29
5 Calculations of Physical Properties 33

5.1 Index of Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Film Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.3 Film Thickness of a Thick Thin Film . . . . . . . . . . . . . . . . . . . . . . . . 36
III Simulations and Experiments 39
6 Simulation of the Refractive Index of Crystals 41

6.1 Silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.2 Copper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.3 Aluminum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.4 Silver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7 Discussion of Crystal Measurements 49
V
CONTENTS
7.1 Silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.2 Copper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.3 Aluminum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.4 Silver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
8 Discussion of SiO2 Film Measurements 59

8.1 Film Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
9 Discussion of Polymer Film Measurements 63

9.1 Film Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.2 Uniformity of Film Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
9.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Conclusion 67
Bibliography 69
Appendix
A The Electromagnetic Wave Model of Light 71
B Fresnel Reflection and Transmission Coefficients 77
C Three Phase Optical System 85
D The Jones Formalism 91
E Correlation between the Stokes Parameters and ψ and Δ 97
F Index of Refraction 101
G Index of Refraction and Band Structures of Crystals 109
H Test Report of Refractive Index of Crystals 119
I Test Report of the Thickness of an SiO2 Film on Si 123
J Test Report of the Thickness of a Polymer Film on Si 127
CD
/datasheets
/matlab
/maple
VI
Introduction
Optical measurement techniques in development and production are of great interest since
these techniques are normally non-invasive. Using these techniques involves no physical
contact with the surface and does not normally destruct the surface. This is a notable prop-
erty of a measurement technique on nanoscale. Only surfaces that are sensitive to bleaching
can be subject to damage by optical measurements. There are several optical measurement
techniques based on the reflection or transmission of light from a surface, among these are
interferometry, reflectometry and ellipsometry. There are three different types of ellipsom-
etry, namely scattering, transmission and reflection ellipsometry. This project concerns re-
flection ellipsometry only.
Ellipsometry normally requires some computer power to get results and therefore, the
technique has only recently become widely used, although it has been known and used
since Paul Drude proposed it over 115 years ago. [Poksinski 2003, p. 1]
Ellipsometry can be used to measure any physical property of an optical system that will
induce a change in polarization state of the incident light wave. This makes ellipsometry a
very versatile technology that is useful in many different applications.
One possible application is in the semiconductor industry. This industry often deals with
a thin layer of SiO2 on a silicon wafer used throughout production. In order to keep track
and effectively control the thickness of this film, process engineers can use ellipsometry
to measure the film thickness of selected sample wafers. Ellipsometry is known for the
high accuracy when measuring very thin film, with a thickness in the Ångström scale or
below. When measuring thicker films the technique becomes more complex and requires
more calculations.
Other possible applications of ellipsometry are determination of the refractive index, the
surface roughness or the uniformity of a sample and more. [Jawoollam 2004]
This project addresses the issue of ellipsometry as a way of determining physical prop-
erties of an optical system. A Sentech SE 850 ellipsometer has been put at disposal for the
group in order to perform test of various samples. This leads to the goal of the project
The goal of this project is to determine the complex index of refraction of several materials and the
thickness of films by ellipsometric measurements.
1
Part I
Preanalysis
The first part contains a short introduction to ellipsometry and the different terms used in connection
with the technique. All important terms mentioned in this part will be described in depth later in
the report. After this introduction, a problem description will line up the purposes and more specific
goals within this project.
3
Introduction to Ellipsometry 1
Ellipsometry is generally a non-invasive, non-destructive measurement technique to obtain
optical properties of a sample material by means of the reflected light waves. The technique
measures a relative change in polarization and is therefore not dependent on absolute in-
tensity as long as the absolute intensity is sufficient. This makes ellipsometric measurement
very precise and reproducible.
Ellipsometry uses the fact that linearly polarized light at an oblique incidence to a surface
changes polarization state when it is reflected. It becomes elliptically polarized, thereby the
name ”ellipsometry”. In some cases elliptically polarized light is used as the incident light
wave. The idea of ellipsometry is shown in general in Figure 1.1.
Linearly polarized
light wave
Ein
π-plane
Elliptically polarized
σ-plane light wave
kin π-plane
Plane Eout
of inc σ-plane
idenc
e kout
Surfa
ce
Figure 1.1: The general principle in ellipsometry. [Jawoollam 2004]
When a monochromatic, plane light wave is directed at a surface at oblique incidence,

the plane of incidence is defined as a plane perpendicular to the surface and containing the
vector which points in the direction of propagation of the light wave. This vector is called
the wavevector kin . Perpendicular to kin are the two mutually perpendicular vectors for the
electric field E and the magnetic field B of the light wave. The E-vector is chosen as the
vector defining the polarization of the light wave and is therefore the only one shown in
Figure 1.1. The E-vector is decomposed into two components, which are mutually perpen-
dicular and perpendicular to kin . The two components of E are respectively parallel and
perpendicular to the plane of incidence as seen in Figure 1.1. The vectors are named from
their German names, ”Parallel” and ”Senkrecht”, and are from this given the corresponding
Greek letters π and σ, respectively.
The incident light wave is linearly polarized. Polarization will be described in depth
later, but for now the π- and σ-component of E can be seen as oscillating with an amplitude
5
CHAPTER 1. INTRODUCTION TO ELLIPSOMETRY
and mutual phase causing the endpoint of E to move in a straight line in the plane of the
π- and σ-components. When the light wave reflects off the surface, the polarization changes
to elliptical polarization. This means that the amplitude and mutual phase of the π- and
σ-component of E are changed causing the endpoint of E to move in an ellipse.
The form of the ellipse can be measured by a detector and data processing can relate
this to the ellipsometric parameters ψ and Δ. The ellipsometric parameters can be related
to the reflection coefficients of the light polarized parallel and perpendicular to the plane of
incidence ρπ and ρσ , respectively. The relation is the basic equation in ellipsometry and is
given by the complex ratio ρ of the two reflection coefficients
ρπ
ρ= = tan(ψ)e jΔ (1.1)
ρσ
The ellipsometric parameters ψ and Δ are given by a measurement with an ellipsometer and
the two reflection coefficients are functions of the complex refractive index of the material.
Ellipsometry is often used to measure the thickness of thin films on top of a substrate.
A simplified model of this is shown in Figure 1.2 where an incident light wave is reflected
off and transmitted through the surface of a thin film. If the refractive indexes of the film
and the substrate are known, it is possible to calculate the thickness d of the thin film by
ellipsometry. This application of ellipsometry is widely used to investigate materials and
surfaces.
Ein Eout
Air
Thin ﬁlm }d
Substrate
Figure 1.2: Illustration of a thin film on top of a crystal.
6
Problem Description 2
There are two overall objectives in this project. These are to determine the refractive index
of various materials and to determine the thickness of various films by ellipsometry. The
materials at hand in this project for measuring the index of refraction are silicon, aluminum,
copper and silver. The materials with a film on a substrate are silicon with a silicon dioxide
film and silicon with a polymer film.
The following requirements are to be met in this project.
1. Theoretical
(a) Modelling of the optical system under investigation in order to enable calculation
of refractive index and film thickness.
2. Simulation
(a) Simulation of the refractive index of silicon, aluminum, copper and silver.
3. Experiments
(a) Refractive Index — Use ellipsometry to measure the refractive index of silicon,
aluminum, copper and silver.
(b) Thickness of thin SiO2 film — Measurement of the thickness of a thin film of
silicon dioxide on a silicon wafer.
(c) Thickness and uniformity of polymer film — Measurement of the thickness of
two optically thick polymer films on substrates of silicon. Furthermore, several
measurements of the thickness of the polymers should be performed in order to
illustrate the uniformity of the surfaces.
All materials except the SiO2 and polymer films are provided by the Institute of Physics
and Nanotechnology at Aalborg University. The SiO2 film sample is a test wafer from Sen-
tech. The test wafer has a known film thickness. Two polymer films are imposed on sili-
con wafers by NanoNord A/S. The polymers are spin coated on the wafers and afterwards
baked at high temperature as prescribed by the manufacturer of the polymer. The poly-
mer is manufactured by HD MicroSystems and is called PI-5878G. The two samples differ
only in the angular speed of the spin coating which should yield different thicknesses of the
polymer. Estimates from NanoNord suggest that the polymers are 2 and 5 µm, but these
estimates are very loose.
7
Part II
Ellipsometry Theory
This part contains three chapters. The first concerns polarization of light, where elliptically polarized
light is of special interest. Also treated in this chapter is the correlation between the ellipsometric
parameters and the Fresnel reflection coefficients. The second chapter concerns ellipsometer systems.
In this chapter a description of different ellipsometer configurations is given. The ellipsometer used
in the tests is also described in this chapter. The last chapter in this part concerns calculation of
refractive index and film thickness by use of measured ellipsometric parameters.
9
Polarization of Light 3
This chapter describes the polarization of light. Three types of polarization will be introduced. These
are linearly, circularly and elliptically polarized light. The descriptions of these polarization types will
be limited to treat monochromatic plane waves only. Apart from this, a description of unpolarized
light will be given. Finally, the ellipsometric parameters ψ and Δ will be introduced.
This chapter is mainly based on [Klein & Furtak 1986, pp. 585-596] and the definitions derived
in Appendix A.
3.1 Definition of Polarization

From the description of monochromatic plane light waves, treated in Appendix A, it can be
seen that light consists of an electric field E and a magnetic field B. The connection between
these and the direction of propagation is given by (A.28) and rewritten here
k×E
B= (3.1)
ω
where the direction of propagation is the direction of k. It is seen that electromagnetic waves
are transverse waves, i.e. E and B are mutually perpendicular and perpendicular to k. As
a consequence, E can point in any direction perpendicular to k. Thus E has two degrees of
freedom, i.e. it is ”free” to move in a 2-dimensional coordinate system. This can be seen in
opposition to longitudinal waves, which are bound to point in the direction of propagation.
This extra degree of freedom implies the existence of different polarization states, which in
the following will be divided into some basic types. But first, some general definitions will
be stated.
The polarization direction of light is defined as the direction of E. When E is known, B
can readily be deduced, direct or indirect from Maxwell’s equations, e.g. from (3.1).
In the following, a right-handed system of coordinates is used, where the z-axis is defined
as the direction of propagation. Thus, the E-field can be described as a linear combination
of an x- and y-component
E(z,t) = Ex x̂ + Ey ŷ (3.2)
where
Ex (z,t) = Ax cos(ωt − kz + φx ) (3.3a)

Ey (z,t) = Ay cos(ωt − kz + φy ) (3.3b)
as described in (A.24).
11
CHAPTER 3. POLARIZATION OF LIGHT
If only the polarization state is of interest, the temporal and spatial dependencies can be
omitted. Thus, by using the Jones formalism, which is described in Appendix D, (3.3) can
be expressed by a Jones vector

Ex Ax e jφx
E= = (3.4)
Ey Ay e jφy
as described in (D.6) and (D.7). (3.4) entirely describes the polarization of light. Whether the
light is described using the Jones formalism (3.4) or with the temporal and spatial informa-
tion included (3.3), the essential parameters are the relative phase φ defined as
φ = φy − φx (3.5)
and the relative amplitude, which is a relation between Ax and Ay .
3.2 Polarization Types

In the following some basic polarization types will be defined.
3.2.1 Linearly Polarized Light

Linearly polarized (LP) light1 is the most straightforward example of polarized light. Light
is linearly polarized when E(z) and E(t) oscillates on a bounded straight line projected in
the xy-plane, with the center at (0, 0). This occurs when
φ = ±pπ , for p = {0, 1, 2, · · · } (3.6)
which imply
Ay
Ey = ± Ex (3.7)
Ax
Hence, Ey is a linear function of Ex , and vice versa, as both Ax and Ay are constants. An illus-
tration of the projection of LP light in the xy-plane can be found in Figure 3.1(a). A depiction
of E with respect to z, where the time is held constant, can be found in Figure 3.1(b).
Note
The requirement to the phase stated in (3.6) is only requisite when both Ax = 0 and Ay = 0. If
Ax = 0 or Ay = 0, the light is linearly polarized.
3.2.2 Circularly Polarized Light

Light is circularly polarized (CP) when E, with respect to t as well as z, defines a circle
projected in the xy-plane. Thus, the following phase-relation must hold
π
φ= ± pπ , for p = {0, 1, 2, · · · } (3.8)
2
1 Linearly polarized light is also denoted plane polarized light.
12
3.2. POLARIZATION TYPES
y
y Ay
Ay z
Ax x
x
6
4
2
0
0
Ax
(a) Depiction of E in the xy-plane. (b) Illustration of E(z).
Figure 3.1: Illustration of linearly polarized light for φ = π and Ax = Ay . Ex

and Ey are illustrated by the dark and light grey curve, whereas the total
electric field is illustrated by the black curve.
and the following amplitude relation must be fulfilled

Ax = Ay = 0 (3.9a)
or
Ax = −Ay = 0 (3.9b)
but the latter is not taken into account in the following, as (3.9b) can be expressed using
(3.9a) with a extra phase difference of π. It is seen that the direction of rotation projected in
the xy-plane depends of φ. The two directions are denoted RCP (Right Circularly Polarized)2
and LCP (Left Circularly Polarized) respectively. LCP occurs when E rotates clockwise with
respect to z when viewed along the negative direction of the z-axis. This can be seen in Fig-
ure 3.2(a). A spatial depiction of LCP light with respect to z can be found in Figure 3.2(b).
When LCP light is described with respect to t, the rotation will consequently be in the coun-
terclockwise direction. LCP occurs when φ = −π/2 ± 2pπ, where p = {0, 1, 2, · · · }. Naturally
the direction of rotation of RCP is opposite to LCP both with respect to z and t. RCP occur
when φ = π/2 ± 2pπ, where p = {0, 1, 2, · · · }.
The fact that E defines a circle in the xy-plane when (3.8) and (3.9) are met, can be seen
from the following. If φ = ±π/2, and Ax = Ay = A, then (3.3) can be expressed as
Ex = A cos(ωt − kz + φx ) (3.10a)
π
Ey = A cos(ωt − kz + φx ± )
2
= ∓A sin(ωt − kz + φx ) (3.10b)
2 The
name RCP origins from the appearance of a normal screw, where the spiral groove has the same shape
as RCP light with respect to z if the screw is placed in the z-axis [Klein & Furtak 1986, p. 588].
13
y
Ay
z
y
1
Ay
Ay x
0
Ax 1 x
(a) Depiction of E(z) in the xy-plane. (b) Illustration of E(z).
Figure 3.2: Illustration of LCP light (φ = −π/2) and Ax = Ay . Ex and Ey are

illustrated by the dark and light grey curve, whereas the total electric field
are illustrated by the black curve.
Adding the squared of (3.10a) to the squared of (3.10b) yields

Ex2 + Ey2 = (A cos(ωt − kz + φx ))2 + (∓A sin(ωt − kz + φx ))2
= A2 (cos2 (ωt − kz + φx ) + sin2 (ωt − kz + φx ))
= A2 (3.11)
where it is observed that (3.11) is the representation of a circle with the center in (0, 0).
3.2.3 Elliptically Polarized Light

If E with respect to z and t describes an ellipse projected in the xy-plane, the light is denoted
elliptically polarized (EP). First, a simple description of EP light is considered.
Description of Elliptically Polarized Light Starting From Circularly Polarized Light

Starting from the description of CP light, the restriction to the phase given by (3.8) is kept,
whereas the amplitude relation given by (3.9) is discarded. This is done in order to allow
Ax = Ay . Ax and Ay must however still be nonzero. Similarly as in (3.11) it is seen that
Ex2 Ey2
+ =1 (3.12)
A2x A2y
which is the description of an ellipse with the major and minor axis along the x- and y-axis.
14
3.2. POLARIZATION TYPES
General description of Elliptically Polarized Light

In general, no restrictions to the relation between the amplitudes Ax and Ay or the phase
difference φ exist for EP light. The general case of EP light can then be stated using (3.3) as
Ex = Ax cos(ωt − kz) (3.13a)

Ey = Ay cos(ωt − kz + φ) (3.13b)
which can be written as

Ex
= cos(ωt − kz) (3.14a)
Ax
Ey
= cos(ωt − kz) cos(φ) − sin(ωt − kz) sin(φ) (3.14b)
Ay
Multiplying (3.14a) with cos(φ) and subtracting the result from (3.14b) yields
E y Ex
+ cos(φ) = − sin(ωt − kz) sin(φ) (3.15)
Ay Ax

= − sin(φ) 1 − cos2 (ωt − kz) (3.16)
Squaring this, results in
Ey2 Ex2 Ex Ey
2
+ 2 cos2 (φ) − 2 cos(φ) = [1 − cos2 (ωt − kz)] sin2 (φ) (3.17)
Ay Ax Ax Ay
Substituting the squared of (3.14a) into (3.17) yields

2
Ey2 Ex2 Ex Ey Ex
+ cos (φ) − 2
2
cos(φ) = 1 − sin2 (φ) (3.18)
A2y A2x Ax Ay Ax
or
Ey2 Ex2 Ex Ey
2
+ 2 −2 cos(φ) + cos2 (φ) = 1 (3.19)
Ay Ax Ax Ay
which defines an ellipse in the xy-plane. It is seen from (3.19) that if φ = pπ, for
p = {· · · , −2, −1, 0, 1, 2, · · ·}, then
Ay
Ey = ((−1) p ) Ex (3.20)
Ax
which, as expected results in the definition of LP light. Similarly if φ = pπ/2, then
Ey2 Ex2
+ =1 (3.21)
A2y A2x
which is EP light with the major axis along the x- or y-axis; or if Ax = Ay it is CP light. Thus,
LP and CP light are both special cases of EP light. It is clear that the ellipse described in
the xy-plane will be inscribed in a rectangle given by Ax and Ay . An illustration of EP light
projected in the xy-plane can be seen in Figure 3.3.
15
Ax
Ay
z x
Figure 3.3: Illustration of elliptically polarized light.
3.2.4 Unpolarized Light

Unpolarized light is the term used for light that is not polarized in any defined pattern as
the ones stated above. For unpolarized light, the electric field vector fluctuates in a random
pattern. If E is divided into components described as in (3.2), Ex and Ey will be incoherent.
That is, the phase relation of the components will be random. Furthermore, as the field
vector fluctuates randomly, the mean value of the magnitude of the field will be the same in
all directions perpendicular to the direction of propagation. Thus A2x = A2y .
3.3 Definition of the ellipsometric parameters ψ and Δ

ψ and Δ angles will in the following be defined as quantities describing the reflected light,
when linearly polarized light is incident on a surface. A depiction of the orientations of the
coordinate systems for the incident and the reflected E-field in relation to the surface can be
seen in Figure 3.4. Using the definition of the Jones vector, a new term χ is defined as the
ratio between the components in the Jones vector, namely
Ey
χ= (3.22)
Ex
The surface can be viewed as a system with the incident E-field Ei as the input and the
reflected E-field Eo as the output. This is illustrated in Figure 3.5 in terms of χ. The term of
interest is the relation describing the optical system S, which is given as χi /χo , thus
Eiy
χi Eix Eiy Eox
= = (3.23)
χo Eoy Eix Eoy
Eox
16
3.3. DEFINITION OF THE ELLIPSOMETRIC PARAMETERS ψ AND Δ
zo
Eo yo
xo
Et
Ei zi
xi
yi
Surface
Figure 3.4: Illustration showing the orientation of the coordinate systems

relative to the sample surface. Ei is the incident or input E-field, Eo is the
reflected or output E-field and Et is the transmitted E-field. y is parallel
with the surface.
χi χo
S
Figure 3.5: Input χi and output χo to an optical system S.
Rewriting this expression using the Jones vector, (3.4) yields
χi Aiy e jφiy Aox e jφox

= (3.24)
χo Aix e jφix Aoy e jφoy
Aiy j(φiy −φix ) Aox j(φox −φoy )
= e e (3.25)
Aix Aoy
If the incident light is linearly polarized with φi = 0 and Aix = Aiy , then (3.25) is given as
χi Aox j(φox −φoy )

= e (3.26)
χo Aoy
which only contains information for the elliptically polarized reflected light. For the re-
flected light, the parameter ψ is defined in order to satisfy the following
Aox
tan ψ = (3.27)
Aoy
This is illustrated in Figure 3.6.
17
yo
Axo
ψ Ayo
zo xo
Figure 3.6: Illustration of ψ, which is defined for reflected elliptically po-

larized light.
Furthermore, the parameter Δ is defined as
Δ = φox − φoy (3.28)
Using (3.27) and (3.28), (3.26) can be expressed as
χi
= tan(ψ)e jΔ (3.29)
χo
which is a general ellipsometer equation [Azzam & Bashara 1977, p. 259].
3.3.1 Connecting ψ and Δ to ρπ and ρσ
The Fresnel reflection coefficients are introduced in Appendix B as the reflected amount
of the E-field in proportion to the incident amount. This is viewed either parallel (π) or
perpendicular (σ) to the plane of incidence as

Eoπ

ρπ = (3.30a)
Eiπ

Eoσ

ρσ = (3.30b)
Eiσ
where all E-vectors are Jones vectors. By fixing the xy-coordinate system to the sample
surface so that x is parallel to the plane of incidence and y is perpendicular to the plane of
18
3.3. DEFINITION OF THE ELLIPSOMETRIC PARAMETERS ψ AND Δ
incidence, (3.23) can be rewritten to

χi Eiy Eox
= (3.31)
χo Eix Eoy
|Eiσ | |Eoπ |
= (3.32)
|Eiπ | |Eoσ |
|Eoπ |
|Eiπ |
= |Eoσ |
(3.33)
|Eiσ |
ρπ
= (3.34)
ρσ
Inserting this expression into (3.29) yields

ρπ
= tan(ψ)e jΔ (3.35)
ρσ
which correlates the ellipsometric parameters to the Fresnel reflection coefficient of a sur-
face. This correlation is utilized throughout the rest of the report to derive expressions for
e.g. the refractive index of a material as a function of ψ and Δ.
19
Ellipsometer Systems 4
This chapter concerns the operational principle of ellipsometers. First a general introduction to ellip-
someters is given which explains the different components encountered in an ellipsometer. After this
introduction to ellipsometers, two different ellipsometer configurations are described, namely the null
and the photometric ellipsometer. After this, the problem of measuring the ellipsometric parameters
ψ and Δ with a photometric rotating analyzer ellipsometer is treated. Finally a description of the
Sentech SE 850 ellipsometer used in this project is given.
4.1 Description of an Ellipsometer

Ellipsometry is generally defined as the task of measuring the state of polarization of a wave.
In the case of an optical system the wave of interest would be a light wave. Although the po-
larization state of a light wave itself can be of interest, in reflection ellipsometry the change
in polarization is the essential issue. This change in polarization as the light is reflected at
a surface boundary is caused by difference in Fresnel reflection coefficients as described in
Appendix B. These coefficients are different for π and σ polarized light. A general ellipsome-
ter configuration is depicted in Figure 4.1. As can be seen from the figure an ellipsometer
σ σ
αC π π
S
αA
αP
C
A
P D
L
Figure 4.1: Illustration of a general ellipsometer setup. Light is emitted

from the source L, passes through the linear polarizer P and the compen-
sator C before it is reflected at the surface boundary S. After reflection
the light again passes a linear polarizer denoted the analyzer A before it
reaches the detector D. [Azzam & Bashara 1977, p. 159]
generally consists of six parts:
The light source which emits circularly or unpolarized light. This can be either a laser or
some type of lamp. A laser has the advantage of emitting very intense and well col-
limated light which produces a very small spot size on the sample. It is however not
possible to use a laser to perform spectroscopic measurements as the laser contains
21
CHAPTER 4. ELLIPSOMETER SYSTEMS
only one wavelength. However a lamp made of e.g. Xenon emits light at many differ-
ent wavelengths enabling spectroscopic measurement.
The linear polarizer which converts the incoming light to linearly polarized light. The rota-
tional azimuth angle of the polarizer relative to the direction of the π linear eigenpolar-
ization is denoted αP in the figure. This angle is the angle from the plane of incidence
to the transmission axis of the polarizer.
The compensator or linear retarder, retards the two perpendicular components of the elec-
trical vector by different amounts thus alternating the polarization state of the wave.
The azimuth angle of the compensator αC is measured relative to the direction of the
π eigenpolarization.
The surface where a fraction of the light wave is transmitted and another is reflected due
to the Fresnel reflection and transmission coefficients ρπ , ρσ , τπ and τσ as described in
Appendix B.
The analyzer is a linear polarizer at a rotational azimuth angle αA relative to the π direction
of the linear eigenpolarization.
The detector measures the intensity of the light from the analyzer. The detector can be any
device able to measure the intensity of a light wave.
Upon making ellipsometric measurements of a surface the rotational angles of the polarizer,
the compensator and the analyzer and the degree of retardation in the compensator must be
known in order to determine the ellipsometric parameters ψ and Δ. There exists a variety of
ways to perform the task of determining the ellipsometric parameters. In the next section
the principles behind two such methods are described.
4.2 Different Ellipsometer Configurations

In this section two general ellipsometer configurations are described; the null and the pho-
tometric ellipsometer.
4.2.1 Null Ellipsometer

The null ellipsometer was historically the first ellipsometer, to be constructed. An illustra-
tion of the general structure of the null ellipsometer and the polarization state of the light
between the components is shown in Figure 4.2. The principle behind this ellipsometer type
is to minimize the intensity of the light wave at the detector. This is done by adjusting the
rotational azimuth angle of the polarizer P, the compensator C and the analyzer A. As illus-
trated in the figure the source emits unpolarized light, which is made linearly polarized by
the polarizer. By adjusting the azimuth angle of the polarizer and compensator the light can
be made linearly polarized after reflection at the surface boundary. By adjusting the azimuth
angle of the analyzer in order to achieve a perpendicular orientation relative to the linearly
polarized wave the light intensity at the detector is minimized or ”nulled”. The ellipso-
metric parameters can then be calculated. As the degree of retardation in the compensator
22
4.2. DIFFERENT ELLIPSOMETER CONFIGURATIONS
P D
C A
Sample
Figure 4.2: A null ellipsometer. The components of the ellipsometer are

illustrated by encircled letters. The polarization state between the com-
ponents is illustrated above. The black dot between the analyzer and the
detector illustrates that the light wave intensity has been nulled.
is dependent on wavelength it is not possible to perform spectroscopic measurements in a

large range of wavelengths with a null ellipsometer.
By use of the Jones matrix formalism the system matrix for a null ellipsometer can be
found.1 The input to this system matrix must be a Jones vector describing the light wave at
the source Eπσ
Lo . The superscript shows that the Jones vector is defined relative to the π and σ
directions i.e. parallel and perpendicular to the plane of incidence and perpendicular to the
direction of propagation. The subscript shows that it is the Jones vector at the light source
output. The output Jones vector of the source is the same as the input Jones vector to the
polarizer i.e. Eπσ πσ
Pi = ELo . The Jones vector at the output of the polarizer is then given as
Eπσ πσ
Po = R(−P)TP R(P)ELo
te
(4.1)
where R(P) is a Jones matrix that rotates coordinate system from πσ to te, which is an abbre-
viation for transmission extinction referring to the fact that a polarizer has a transmission
and an extinction axis. TteP is the Jones matrix for the polarizer. R(−P) rotates the Jones
vector back to the πσ-coordinate system.
With the Jones vector at the output of the polarizer given, the Jones vector at the output
of the compensator is expressed as
πσ
= R(−C)TC R(C)Eπσ
fs
ECo Po (4.2)
fs
where TC is the Jones matrix of the compensator. Again the Jones vector is rotated to the
coordinate system of the compensator, which is denoted f s for fast-slow, referring to the
fact that a compensator has a fast and a slow axis. The Jones vector at the output side of the
surface can be expressed as
πσ πσ
Eπσ
So = TS ECo (4.3)
1 See Appendix D for an explanation of the Jones matrix formalism.
23
where Tπσ
S is the Jones matrix of the surface. Finally the Jones vector at the output of the
analyzer and hence at the detector is given as
πσ
Ao = TA R(A)ESo
te
Ete (4.4)
where TteA is the Jones matrix of the analyzer. There is no rotational matrix after the analyzer
that transforms the Jones vector back to the π, σ system of coordinates. This is because the
light detector, in the absence of errors is insensitive to polarization. Thus Eo = Ete
Ao
By combining (4.1), (4.2), (4.3) and (4.4), an expression for the Jones vector of the light
wave at the detector as a function of the Jones vector of the light wave at the source can be
derived
πσ fs πσ
Eo = Tte
A R(A)Ts R(−C)TC R(C − P)TP R(P)ELo
te
(4.5)
This equation describes a null ellipsometer where a compensator has been placed before the
surface, but a compensator can also be placed after the surface. In that case the equation
describing the system will be
fs πσ πσ
Eo = Tte
A R(A −C)TC R(C)Ts R(−P)TP R(P)ELo
te
(4.6)
The light wave intensity measured at the detector Io is then given as the multiplication of
Eo with its Hermitian adjoint E†o [Röseler 1990, p. 60]. The Hermitian adjoint of a matrix is
defined as the complex conjugate of the transpose of the matrix i.e.
Io = E† E (4.7)
In both cases the only unknown is the Jones matrix for the surface Tπσ
S , which can be found
if the output intensity is ”nulled” and the rotational angles of the polarizer, compensator,
analyzer, the relative phase retardation of the compensator and the angle of incidence are
known.
4.2.2 Photometric Ellipsometer

In photometric ellipsometry one or more conditions are varied while the light intensity at the
detector is measured. This is unlike null ellipsometry, as it is not the means of a photometric
ellipsometer to have zero light intensity at the detector. Thus the output of photometric
ellipsometry measurements are light intensity values at a number of prescribed conditions.
The varied conditions could be the rotational azimuth angle of the polarizer, compensator
or analyzer, the relative retardation of the compensator or the angle of incidence. In most
cases the varied condition is the angle of the polarizer or analyzer, and thus only these two
cases are considered in the following.
Unlike the null ellipsometer the photometric ellipsometer does not necessarily include
a retarding element. This has the apparent advantage of making spectroscopic measure-
ments possible as the polarizers generally are achromatic over a wider spectral range than
retarders. Other advantages include that polarizers are relatively easy to construct com-
pared to compensators and that they are easy to align within a system. On the other hand
a disadvantage is that the system looses sensitivity when Δ is near 0 or 180◦ , which will be
described further in the next section.
24
4.3. DETERMINATION OF ψ AND Δ WITH A STATIC PHOTOMETRIC RAE
L D
P A
Sample
Figure 4.3: A photometric ellipsometer. The components of the ellipsome-

ter are illustrated by encircled letters. The polarization state between the
components is illustrated above. The arrows on the circle that illustrates
the analyzer shows that it is rotating.
The general structure of a photometric rotating analyzer ellipsometer (RAE) is shown in

Figure 4.3. Also shown in the figure is the polarization state between the different compo-
nents of the ellipsometer. The light source emits unpolarized light, which is linearly polar-
ized by the polarizer. After reflection at the surface boundary the polarization state of the
light wave is changed from linearly polarized to elliptical polarized. The analyzer is rotated
and the light intensity is measured at different rotational azimuth angles of the analyzer. The
general principle behind a rotating analyzer photometric ellipsometer is thus to measure the
intensity at different analyzer rotational angles, and from these measurements calculate the
ellipsometric parameters ψ, Δ. It is also possible to vary the angle of the polarizer in which
case the ellipsometer will be denoted a photometric rotating polarizer ellipsometer (RPE).
The operation characteristics of the RAE and the RPE are basically the same, but some dis-
advantages/disadvantages exist for both configurations. The RPE requires the source to
be totally unpolarized in order to perform accurate measurements. Correspondingly the
RAE requires photodetectors that are insensitive to polarization in order to minimize errors.
[Röseler 1990], [Jawoollam 2004]
Static and Dynamic Photometric Ellipsometers
As mentioned the light intensity is measured when either the angle of the polarizer or the
analyzer is varied in a photometric ellipsometer in order to measure the ellipsometric para-
meters ψ and Δ. This variation can be done in one of two different ways. One is to measure
the light intensity at predetermined fixed azimuthal positions. This method is denoted static
photometric ellipsometry. The other is to periodically vary the azimuth angle of either or
both the analyzer and polarizer with time. The detected signal is then Fourier-analyzed in
order to determine ψ and Δ. In the next section a description of a photometric RAE is given
by use of the Jones matrix formalism. [Azzam & Bashara 1977, pp. 255-260]
4.3 Determination of ψ and Δ with a Static Photometric RAE

The ellipsometer available in this project is a photometric RAE ellipsometer and thus this
section concerns this type of ellipsometer only. This ellipsometer utilizes static analyzer an-
25
gles in the determination of the ellipsometric parameters in the ultra violet (UV) and visible
(VIS) region. As the tests performed in this project are done in the UV-VIS area the static
method of determining ψ and Δ is treated. The determination of the ellipsometric para-
meters ψ and Δ with a photometric RAE is performed by measuring the intensity of the
reflected light at three or more analyzer angles and making a calculation of the ellipsometric
parameters from these light intensity values. The principles underlying these calculations
are explained in this section. In the following the light waves are considered being mono-
chromatic plane waves.
4.3.1 Description of an RAE by use of the Jones Matrix Formalism

Measurement of the ellipsometric parameters of a sample is illustrated in Figure 4.4. The
Sample
(ρπ,ρσ)
Eσ
δσ ψ Eiσ
Eoσ θ0
Eπ
δπ α1
Eiπ Polarizer
Analyzer α2
Eoπ
Detector Source
Figure 4.4: Illustration of ellipsometry performed with a photometric RAE

without a compensator. [Röseler 1990, p. 73]
light from the source becomes linearly polarized at the fixed polarizer. The Jones vector of
the light wave after the polarizer Eπσ
i is

πσ Eiπ Ei cos(α1 )
Ei = = (4.8)
Eiσ Ei sin(α1 )
where Ei is the magnitude of the Jones vector Eπσ

i and α1 is the azimuth angle of the polarizer
measured from the direction of the π eigenpolarization.
The light is reflected by the surface, which in Jones notation corresponds to multiplica-
tion by the Jones matrix of the surface

πσ ρπ 0
Ts = (4.9)
0 ρσ
Next the Jones vector of the light wave after the surface must be rotated to the coordinate
system of the analyzer by the Jones transform matrix

cos(α2 ) sin(α2 )
R(α2 ) = (4.10)
− sin(α2 ) cos(α2 )
26
4.3. DETERMINATION OF ψ AND Δ WITH A STATIC PHOTOMETRIC RAE
With the Jones vector given in the coordinates system of the analyzer the Jones matrix of the
analyzer is given by

1 0
TA =
te
(4.11)
0 0
as the analyzer is considered ideal. [Röseler 1990, pp. 60-63], [Azzam & Bashara 1977, p. 76]
The Jones vector at the detector Ete
o can be expressed as
πσ πσ
o = TA R(α2 )Ts Ei
te
Ete (4.12)

1 0 cos(α2 ) sin(α2 ) ρπ 0 Ei cos(α1 )
= (4.13)
0 0 − sin(α2 ) cos(α2 ) 0 ρσ Ei sin(α1 )

cos(α2 )ρπ cos(α1 )Ei + sin(α2 )ρσ sin(α1 )Ei
= (4.14)
0

cos(α2 )Eπ + sin(α2 )Eσ
= (4.15)
0
where Eπ = ρπ cos(α1 )Ei and Eσ = ρσ sin(α1 )Ei .
4.3.2 Light Wave Intensity at the Detector

The light wave intensity at the detector Io is given as
Io = E†o Eo (4.16)

∗ ∗
cos(α2 )Eπ + sin(α2 )Eσ
= cos(α2 )Eπ + sin(α2 )Eσ 0 (4.17)
0
= cos2 (α2 )Eπ Eπ∗ + sin2 (α2 )Eσ Eσ∗ + cos(α2 ) sin(α2 ) (Eπ Eσ∗ + Eσ Eπ∗ ) (4.18)
where the te notation is omitted. This expression for the intensity can be rewritten by utiliz-
ing the following trigonometric identities [Råde & Westergren 1998, p.124]
1
cos(α2 ) sin(α2 ) = sin(2α2 ) (4.19a)
2
1 − cos(2α2 )
sin2 (α2 ) = (4.19b)
2
1 + cos(2α 2)
cos2 (α2 ) = (4.19c)
2
The intensity is then given as
1 1 1
Io = (Eπ Eπ∗ + Eπ Eπ∗ cos(2α2 )) + (Eσ Eσ∗ − Eσ Eσ∗ cos(2α2 )) + (Eπ Eσ∗ + Eσ Eπ∗ ) sin(2α2 ) (4.20)
2 2 2
1
= Eπ Eπ∗ + Eσ Eσ∗ + (Eπ Eπ∗ − Eσ Eσ∗ ) cos(2α2 ) + (Eπ Eσ∗ + Eσ Eπ∗ ) sin(2α2 ) (4.21)
2
1
= s0 + s1 cos(2α2 ) + s2 sin(2α2 ) (4.22)
2
where the three Stokes parameters s0 = Eπ Eπ∗ + Eσ Eσ∗ , s1 = Eπ Eπ∗ − Eσ Eσ∗ and s2 = Eπ Eσ∗ + Eσ Eπ∗
are introduced. [Röseler 1990, p. 74]
27
4.3.3 Determination of ψ and Δ with a Static Photometric RAE
From (4.22) expressions for the light wave intensity at different analyzer angles can be cal-
culated. The light intensity at four specific values of α2 in steps of 45◦ is
1
Io (0◦ ) = (s0 + s1 ) (4.23a)
2
1
Io (45◦ ) = (s0 + s2 ) (4.23b)
2
◦ 1
Io (90 ) = (s0 − s1 ) (4.23c)
2
1
Io (−45◦ ) = (s0 − s2 ) (4.23d)
2
The determination of the ellipsometric parameters ψ and Δ requires only three analyzer
angle setting e.g. α2 = 0◦ , α2 = 45◦ and α2 = 90◦ . Additional measurements would be redun-
dant, but due to practical imperfections in the ellipsometer they might increase the precision
of the determined parameter values.
The Stokes parameters are connected to the measured light wave intensities at the de-
tector due to (4.23). The Stokes parameters are furthermore connected to the ellipsometric
parameters in Appendix E due to (E.26). Combining these equations yields
−s1 1
(s0 − s1 ) − 12 (s0 + s1 ) Io (90◦ ) − Io (0◦ )
cos(2ψ ) = = 21 = (4.24)
s0 2 (s0 − s1 ) + 2 (s0 + s1 )
1 Io (90◦ ) + Io (0◦ )
and
s2 s 0 + s2 2Io (45◦ )
sin(2ψ ) cos(Δ) = = = (4.25)
s0 s0 Io (90◦ ) + Io (0◦ )
where the new variable ψ is given by the relation
tan(ψ)
tan(ψ ) = (4.26)
tan(α1 )
In the case where the polarizer angle is 45◦ i.e. α1 = 45◦ the relation reduces to
tan(ψ ) = tan(ψ) (4.27)
and hence ψ = ψ.
As can be seen from (4.24) and (4.25) the ellipsometric parameters can, as mentioned
above, be calculated from only three measurements of the light wave intensity, however
some limitations are present. Δ is determined in the region 0◦ ≤ Δ ≤ 180◦ only. In the region
of cos(Δ) ≈ 1 the determined Δ can be very inaccurate because a small variation in cos(Δ)
causes a large variation in the determined Δ. These problems can be minimized by intro-
ducing a retarder in the system. [Röseler 1990, p. 76]
28
4.4. DESCRIPTION OF THE SENTECH SE 850 ELLIPSOMETER
4.4 Description of the Sentech SE 850 Ellipsometer

The ellipsometer used for experiments in this project is the spectroscopic ellipsometer SE 850
from Sentech. This is a photometric RAE ellipsometer that utilizes both static and dynamic
measurements. The SE 850 is computer controlled via the Sentech software ”Spectraray”.
The ellipsometer has a range of wavelength from 350 nm to 1700 nm.
During this project it has not been possible to use the NIR part of the ellipsometer due
to software failure. This means that the range of wavelength is limited to 350 nm - 850 nm
in all measurements.
4.4.1 Functional Description

A block scheme showing the structure of the SE 850 is depicted in Figure 4.5. The ellip-
Input Box Output Box

UV/VIS NIR
light detector
source
Source Polarizer (Compen- Aperture Sample Stage Aperture Rotating Detector
alternator sator) control control analyzer alternator
NIR light UV/VIS

source detector
Control and data

processing computer
with GUI
Figure 4.5: A block scheme of the SE 850. Control signals are illustrated
with dashed lines, the light waves travelling through air are illustrated by
fully drawn lines, and the light waves travelling through optical fibers are
illustrated by thick fully drawn lines.
someter is centered around the control and data processing computer. This computer also
has a graphical user interface (GUI) in order for the user to initiate the ellipsometric mea-
surement and perform data processing of the measured data. The control signals from and
to the computer are shown as dashed lines. These output signals are control signals for
the choice of source and detector, control signal for the compensator if this is to be used in
the measurement and a control signal for the rotating analyzer. The input signals are the
measured intensities from the detectors. The fully drawn lines in the figure illustrate light
waves travelling through air. The light wave is propagated through an optical fiber between
the sources and the source alternator. The same is the case between the detector alternator
and the detectors. The optical fibers are illustrated by thick fully drawn lines. Note that
the compensator is enclosed by brackets as this component only is utilized in some special
cases. Further specifications of the SE 850 are listed below.
4.4.2 Specifications
The specifications for the SE 850 are found at Sentech’s web page [Spectroscopic Ellipsome-
ter SE 850 2004].
29
UV/VIS Light Source: The light source used for UltraViolet/Visible (UV/VIS) measure-
ments is a 75 W xenon lamp. With this light source measurements in the spectrum
from 350 nm to 850 nm can be performed.
NIR Light Source: A halogen lamb is used for measurements in the Near InfraRed (NIR)
range between 850 nm to 1700 nm.
UV/VIS Detector: A photodiode array with 1024 elements is used to detect the light inten-
sity in the UV/VIS range. This unit is placed in the control computer cabinet.
NIR Detector: A Fourier Transform InfraRed (FT-IR) photodetector is used in the NIR range.
This unit is placed in the output box of the ellipsometer after the source alternator.
Polarizer: The polarizer is fixed at a rotational azimuth angle of 45◦ .
Analyzer: The azimuth angle of the analyzer is variable and is controlled by the spectraray
software running on the computer.
Compensators: Computer controlled super achromatic retarder for UV/VIS spectral range.
(Optional)
Goniometer: The angle of incidence is controlled by a manual goniometer which has a

range from 30◦ to 90◦ with a step size of 5◦ .
Sample Stage: The sample stage is manually controlled. Possible adjustment parameters
are the height and inclination of the sample stage. Furthermore it is possible to control
the azimuth angle of the sample stage with a resolution of 1◦ and the translational
position in one dimension of the sample in the plane of incidence with a resolution of
10 µm.
Apertures A manual aperture control is placed on the input and output side of the sample
stage. With this component it is possible to adjust the spot size and hence the intensity
of the light wave.
A picture of the SE 850 ellipsometer is shown in Figure 4.6. The box to the left is the input
box with the source alternator, the polarizer, the compensator and the aperture control. The
box to the right is the output box with the aperture control, the analyzer, detector alternator
and the NIR detector. A picture of the two boxes without the cover can be seen in Figure 4.7.
30
4.4. DESCRIPTION OF THE SENTECH SE 850 ELLIPSOMETER
Figure 4.6: The Sentech SE 850 ellipsometer.
(a) The input box of the ellipsometer.. (b) The output box of the ellipsometer.
Figure 4.7: The input and output units of the ellipsometer without the cov-
ers.
31
Calculations of Physical
Properties 5
This chapter describes the calculation of the complex refractive index and the film thickness of a thin
film. Furthermore, the method of calculating the thickness of an optically thick film is described.
5.1 Index of Refraction

This section describes the relation between the ellipsometric parameters ψ and Δ measured
with the ellipsometer and the complex index of refraction ñ1 . Figure 5.1 shows the basic
setup in order to calculate the refractive index. The basic equation for an ellipsometer found
θ0
Medium (0)
Medium (1)
θ1
Figure 5.1: Reflection and transmission of an incident light wave at a sur-

face boundary.
in (3.35) on page 19 contains the Fresnel reflection coefficients ρπ and ρσ which are deduced
in Appendix B. In this appendix, they are given by (B.30a) and (B.25) as
ñ1 cos(θ0 ) − ñ0 cos(θ1 )

ρπ = (5.1)
ñ1 cos(θ0 ) + ñ0 cos(θ1 )
ñ0 cos(θ0 ) − ñ1 cos(θ1 )
ρσ = (5.2)
ñ0 cos(θ0 ) + ñ1 cos(θ1 )
where cos(θ1 ) can be found via Snell’s law and the trigonometric identity as
2
ñ0
cos(θ1 ) = 1 − sin2 (θ0 ) (5.3)
ñ1
Here, ñ1 is the complex refractive index of medium 1, ñ0 is the complex refractive index
of the ambient, θ0 is the angle of incidence and θ1 is the unknown angle of transmission.
Inserting (5.1), (5.2) and (5.3) into (3.35) and solving for ñ1 yields

1 − 4 sin (θ0 ) tan (ψ) e + 2 tan (ψ) e + tan (ψ) e ñ0 sin(θ0 )
2 jΔ jΔ 2 jΔ
ñ1 = (5.4)
cos(θ0 ) [1 + tan (ψ) e jΔ ]
33
CHAPTER 5. CALCULATIONS OF PHYSICAL PROPERTIES
The data from the ellipsometer are values of ψ and Δ as a function of wavelength. Using
(5.4), these data can be used to calculate the complex index of refraction as a function of
wavelength.
5.2 Film Thickness

When ellipsometric measurements are performed on a three phase optical system consisting
of an ambient-film-substrate structure, it is possible to determine the thickness of the film, if
the refractive indexes for the three media are known. This section concerns relating the film
thickness to the ellipsometric parameters ψ and Δ. An ambient-film-substrate optical system
is depicted in Figure 5.2 The incident light wave from the ellipsometer strikes the surface
θ0
Ambient (0)
Film (1) d θ1
Substrate (2)
θ2
Figure 5.2: Illustration of an ambient-film-substrate optical system. The

incident wave is partially reflected and partially transmitted.
boundary between ambient and the film at an angle of θ0 , which will also be the angle of
the reflected wave due to Snell’s law. Reflection and transmission of a polarized wave due
to the surface boundaries in a three phase optical system is treated in Appendix C. In this
appendix the total reflection coefficients of σ and π polarized light are found to be
ρ01,σ + ρ12,σ e− j2β

Pσ = (5.5a)
1 + ρ01,σ ρ12,σ e− j2β
ρ01,π + ρ12,π e− j2β
Pπ = (5.5b)
1 + ρ01,π ρ12,π e− j2β
due to (C.29a) and (C.29b). P is the Greek letter capital ρ. The reflection coefficients in these
equations are given in (C.30). An expression for β is given in (C.16).
5.2.1 Relation Between Ellipsometric Parameters and Film Thickness

(5.5a) and (5.5b) can be related to the ellipsometric parameters due to (3.35) where the re-
flection coefficients in a two-phase optical system ρπ and ρσ are replaced by the reflection
coefficients of a three-phase optical system Pπ and Pσ
Pπ
P= = tan(ψ)e jΔ (5.6)
Pσ
34
5.2. FILM THICKNESS
where the parameter P is introduced as the complex reflection ratio [Azzam & Bashara 1977,
p. 288]. By inserting the expressions for Pσ and Pπ the following is given
1 ρ01,π + ρ12,π e− j2β 1 + ρ01,σ ρ12,σ e− j2β
P = Pπ · = · (5.7)
Pσ 1 + ρ01,π ρ12,π e− j2β ρ01,σ + ρ12,σ e− j2β
ρ12,π ρ01,σ ρ12,σ e− j4β + (ρ01,π ρ01,σ ρ12,σ + ρ12,π ) e− j2β + ρ01,π
= (5.8)
ρ01,π ρ12,π ρ12,σ e− j4β + (ρ01,π ρ12,π ρ01,σ + ρ12,σ ) e− j2β + ρ01,σ
This is an equation of 11 parameters, where the two ellipsometric parameters ψ and Δ are
related to nine real parameters. These parameters are the real and imaginary parts of the
complex refractive indexes, ñ0 , ñ1 , ñ2 , the angle of incidence θ0 , the free-space wavelength
of the incident light wave λ and the film thickness d. If a set of ellipsometric parameters are
measured at a given angle of incidence and a given wavelength the thickness of the film is
the only unknown, assuming that the refractive indexes of the ambient, film and substrate
are known. Thus by solving (5.8) for d, the film thickness of a sample can be determined.
5.2.2 Solving for the Film Thickness

(5.8) can be rewritten to
AX 2 + BX +C
P= (5.9)
DX 2 + EX + F
where A = ρ12,π ρ01,σ ρ12,σ , B = ρ01,π ρ01,σ ρ12,σ + ρ12,π , C = ρ01,π , D = ρ01,π ρ12,π ρ12,σ ,
E = ρ01,π ρ12,π ρ01,σ + ρ12,σ , F = ρ01,σ and X = e− j2β . Rearrangement of this equation yields
(PD − A)X 2 + (PE − B)X + (PF −C) = 0 (5.10)
which is a complex quadratic equation with the solution

−(PE − B) ± (PE − B)2 − 4(PD − A)(PF −C)
X= (5.11)
2(PD − A)
If the refractive index for the film is known, two analytical solutions to this equation exist,
namely X1 and X2 .
If the refractive index for the film is not known, but is assumed real i.e. ñ1 = n1 , solutions
to (5.11) can be found by iteration. In this iteration procedure n1 is varied until the condition
|X| = 1 is satisfied. As X = e− j2β , where β is given by (C.16) on page 87 it is given that |X|
must equal 1. With the determined value of n1 a value for X is also given.
With X determined from either of the methods, it is possible to calculate the film thick-
ness due to
X = e− j2β (5.12)
d
ln(X) = − j4π ñ1 cos(θ1 ) (5.13)
λ
j ln(X)λ
d= (5.14)
4πñ1 cos(θ1 )
where X is either of the previously calculated solutions to (5.11). Obviously only one so-
lution for the film thickness is valid, which should be real and positive. In the presence of
errors the calculated thickness may be complex. In this case the solution with the smaller
imaginary part should be chosen.
35
CHAPTER 5. CALCULATIONS OF PHYSICAL PROPERTIES
5.3 Film Thickness of a Thick Thin Film

In cases where the film is transparent i.e. the refractive index is purely real, multiple solu-
tions for the film thickness exist. This can be seen from (5.12) which can be rewritten to
X = e− j4π λ ñ1 cos(θ1 ) = e− j2π( D )

d d
(5.15)
where
λ
D= (5.16)
2ñ1 cos(θ1 )
. From this equation it can be seen that X is a periodic function of d with a period of D, thus
D is denoted the film thickness period. The film thickness period is a function of the angle of
incidence, the wavelength of the light in free-space and the refractive indexes of the ambient
and the film. The complete solution for the film thickness is then given as
d = d0 + mD (5.17)
where d0 is the solution found in (5.14), which is called the standard solution and m is either
0 or a natural number i.e. m = {0, 1, 2, ..}. Without knowledge of the range of the thickness
in advance it can thus prove difficult to explicitly determine the film thickness if the film is
non-absorbing. The next section treats this subject. [Azzam & Bashara 1977, pp. 283-317]
5.3.1 Solving for the Film Thickness of a Thick Thin Film

When the film thickness d exceeds the film thickness period D, interference in the reflected
light will appear, as the different components1 of the reflected wave will be in phase at some
wavelengths and in counter phase at other wavelengths. This will result in ψ and Δ angles
that vary between positive and negative interference with a period that is dependent on
wavelength. An example of this is given in Figure 5.3 where Δ is plotted as a function of
wavelength. It is emphasized that the graph serves as an illustration only, and that it is not
an actual experimental result. From a Δ-spectrum as this, it is possible to calculate the film
thickness, as the distance between the local maxima are determined by the angle of incidence
θ0 , the refractive index of the ambient ñ0 , the film ñ1 , the substrate ñ2 , the wavelength of the
light λ and the film thickness d. Usually the only unknown is the film thickness which can
then be calculated. This is done by determining the wavelengths of two adjacent peaks. At
these two wavelengths the standard solution d0 and the thickness period D can be calculated
as described above. Two expressions for the film thickness can be set up due to (5.17). One
at the first maximum at λ = λ0
d = d00 + m0 D0 (5.18)
and one at the next at λ = λ1 .
d = d01 + m1 D1 (5.19)
1 See Figure 5.2
36
5.3. FILM THICKNESS OF A THICK THIN FILM
140
120
100
Delta [Degrees]
80
60
40
20
500 520 540 560 580 600 620 640 660 680 700
Wavelength [nm] λ0 λ1
Figure 5.3: Illustration of interference as the film thickness exceeds the

film thickness period. It can be seen that the distance between to adja-
cent peaks increases as the wavelength increases. Two adjacent peaks has
been marked as λ0 and λ1 .
As the thickness of the film is not dependent of wavelength, d must be the same in both
equations. Subtraction of (5.18) from (5.19) yields
0 = d01 − d00 + m1 D1 − m0 D0 (5.20)
The factors m0 and m1 are then the only unknowns. If the measured Δ spectrum has sufficient
resolution to enable determination of all peaks, i.e. it is certain that there are no peaks
between λ0 and λ1 it can be reasoned that
m0 = m1 + 1 (5.21)
With a relation between m0 and m1 (5.20) can be rewritten to
0 = d01 − d00 + m1 D1 − (m1 + 1)D0 (5.22)

D0 + d00 − d01
m1 = (5.23)
D1 − D0
The only unknown parameter in (5.23) is m1 and thus the value for this can be calculated.
In the absence of errors m1 is a natural number. Inserting the expression for m1 into (5.19)
yields
D0 + d00 − d01
d = d01 + D1 (5.24)
D1 − D0
from which the film thickness can be directly calculated.
37
Part III
Simulations and Experiments
This part contains simulations of the refractive index of silicon, aluminum, copper and silver. It
also describes the results of the tests performed with the SE 850 ellipsometer. These tests include
measurement of the refractive index of the same four materials that were simulated, measurement of
the film thickness of a silicon dioxide film coated on a silicon wafer and measurement of the thickness
and uniformity of an optically thick polymer film coated on a silicon wafer.
39
Simulation of the Refractive
Index of Crystals 6
This chapter contains simulations of the refractive index of silicon, copper, aluminum and silver de-
rived from the Drude-Lorentz harmonic oscillator model in Appendix F. There are unknown parame-
ters in some of the simulations, which are then found by fitting to table values shown in Appendix H.
The simulations are performed with wavelengths between 350 nm and 820 nm, which
corresponds to the spectrum of the experiments.
The Drude-Lorentz model for a classical, forced, damped harmonic oscillator is explained
in depth in Appendix F. The model sees the electrons in the material as oscillating with
regard to the nucleus. This yields a mathematical expression for the complex dielectric co-
efficient ε̃(ω) given as
ω2p
ε̃(ω) = 1 + (6.1)
ω20 − ω2 − jγω
from (F.40). ω0 is the resonance frequency of the undamped oscillator, γ is the damping
coefficient given as γ = 1/τ where τ is a table value for the specific material yielding the mean
time between collisions of the electrons and ω p is the plasma frequency defined as

Ne2
ωp = (6.2)
ε0 m
where N is a table value for the electron density in the material, e is the electron charge, ε0 is
the free-space permittivity and m is the mass of the electron. The relation between (6.1) and
the complex refractive index ñ(ω) is given as

ñ(ω) = ε̃(ω) (6.3)
The Drude-Lorentz model is used to simulate resonance phenomena in the refractive
index, where ω0 is the frequency corresponding to the resonance. When simulating these
resonance peaks, it is necessary to use another amplification factor instead of ω p in order to
achieve a proper fit to the table values of the refractive index. This amplification factor is in
the following Drude-Lorentz models denoted A.
In case of metals, Drude’s free-electron model is used to simulate the refractive index of
the materials. This model is given as the Drude-Lorentz model with ω0 = 0 yielding
ω2p
ε̃(ω) = 1 − (6.4)
ω(ω + jγ)
from (F.43).
If the refractive index of the simulated material contains several resonance peaks, a
model that is made up of a sum of several dielectric coefficients must be used. When sim-
ulating metals with resonance peaks, a model that is made up of a sum of Drude’s free-
electron model and the Drude-Lorentz model must be used.
41
CHAPTER 6. SIMULATION OF THE REFRACTIVE INDEX OF CRYSTALS
6.1 Silicon
Silicon is a semiconductor and the simulation of n and K is done only by means of the
Drude-Lorentz model. The intrinsic electron concentration for silicon at room temperature
is estimated from [Kittel 1986, p. 184] to 1010 electrons per cm3 . This is very low considering
that there are approximately 5.0 · 1022 Si-atoms per cm3 . This entails that the effect from
the plasma frequency can be entirely neglected in this model. Instead, the amplification-
factor A is used to fit the amplitude of the resonant peak to the table values. To further fit
the amplitude of the resonance peak, a phase factor is introduced in A. The phase factor
can move the peak upwards or downwards in the spectrum and determines if the peak is
positive or negative in relation to the original curve.
Table values and [Palik 1998, III: p. 531-534] suggest that there are two oscillators affect-
ing the index of refraction in the vicinity of the simulation spectrum due to the two band
gaps on 3.38 eV and 4.27 eV. The resonance peak at ω0 related to the band gap on 3.38 eV
can be seen in the simulation spectrum, whereas the other band gap is outside the spectrum
and therefore not affecting the simulation. The frequency corresponding to the band gap
affecting the spectrum can be found as
Eg
ω0 = (6.5)
h̄
where h̄ is Planck’s constant divided by 2π and ω0 is the resonance frequency for the un-
damped oscillator related to the band gap from Eg . The relation between any frequency ω
and the wavelength λ is given by
2πc
ω= (6.6)
λ
where c is the speed of light in vacuum. This yields a resonance peak in the model at
λ = 366.9 nm. The model then becomes

A
ñ = 1 + 2 (6.7)
ω0 − ω2 − jγω
where the damping coefficient γ will be estimated. The effect of γ is related to the height and
width of the resonance peak. It is stated that γ ω0 [Reitz et al. 1993, p. 500] but by manual
iteration, the peak of the simulation is fitted to the table values of the refractive index of
silicon. The values of A, γ and ω0 are given as
A = 1.2 · 1032 e− j44.8 s−2
γ = 7.4 · 1014 s−1

ω0 = 5.13 · 1015 s−1
The simulation of (6.7) with the parameters given can be seen in Figure 6.1. n is the real
part of (6.7) and k is the imaginary.
The model provides a simple understanding of the effect of the band gap yielding the
given resonance frequency.
42
6.2. COPPER
Silicon
7
n table values
k table values
n simulation
6 k simulation
4
n and k
0
350 400 450 500 550 600 650 700 750 800
Wave length [nm]
Figure 6.1: A manual fit to table values of n and k for silicon. [CD 2004,
matlab/refractive_index_simulation/drude_free_electron_model.m]
6.2 Copper
Copper is the first of the metals which are simulated. For the metals, Drude’s free-electron
model is mainly used to simulate the index of refraction due to the free electrons in the
material. For copper, there is however also an effect from two resonance peaks.
The plasma frequency for copper is located outside the simulation spectrum at 115 nm.
There is resonance at 5 eV due to interband transition in the conduction bands and resonance
due to a transition from the d-band to the conduction band is yielding a peak at approxi-
mately 2 eV. This can be seen in Appendix G. The resonance at 5 eV mainly affects the real
part of the refractive index in wavelengths up to approximately 600 nm. The resonance at
2 eV can be seen as a little ”flip” in the imaginary part of the complex index of refraction
around 580 nm in Figure 6.2.
The simulation model used for copper is then given as the square root of the sum of a
free-electron model and two resonance models

ωp A2 A3
ñ = A1 1 − +2+ 2 + (6.8)
ω(ω + jγ1 ) ω0,2 − ω2 − jγ2 ω ω20,3 − ω2 − jγ3 ω
The values of the parameters are shown in Table 6.1. To fit the amplitude of the resonance
peaks, phase factors are introduced in A2 and A3 .
43
Copper
6
n table values
k table values
n simulation
k simulation
5
4
n and k
0
350 400 450 500 550 600 650 700 750 800
Wave length [nm]
Figure 6.2: A manual fit to table values of n and k for copper. [CD 2004,
1 2 2
A [s−2 ] 0.60 4.0 · 1030 e− j90 2.7 · 1032 e j0.4
γ [s−1 ] 3.70 · 1013 5.2 · 1014 9.3 · 1015
ω0 [s−1 ] 0 3.30 · 1015 7.60 · 1015
ω p [s−1 ] 1.64 · 1016 · ·
Table 6.1: Estimated values for the parameters involved in the simulation
of the refractive index of copper. A1 is without unit.
6.3 Aluminum
Aluminum is also a metal and Drude’s free-electron model is mainly used to simulate the
index of refraction due to the free electrons in the material. There is however also a small
effect from a resonance peak at 1.55 eV. This implies that a square root of the sum of two
complex dielectric functions should be taken. One accounting for the plasma frequency
without resonance and one relating to the resonance

ωp A2
ñ = A1 1 − +1+ 2 (6.9)
ω(ω + jγ1 ) ω0 − ω2 − jγ2 ω
First, an estimate for the plasma frequency can be found by inserting the number of
electrons per unit volume N = 18.1 · 1022 cm−3 into (F.36) which yields a plasma frequency
44
6.3. ALUMINUM
at ω p = 2.40 · 1016 s−1 . This corresponds to a wavelength of λ p = 78.5 nm, which is outside
the spectrum range of the simulation. The plasma frequency does however still influence
the spectrum. A scaling factor A1 is introduced in order to achieve a proper fit to the table
values of the index of refraction.
Next, an estimate of the damping coefficient γ must be calculated. For aluminum τ is
given as τ = 0.8 · 10−14 s. The corresponding damping can be used in the free-electron model,
but must be larger when used in the expression for the resonance peak.
The amplification-factor A2 used in the resonance model is also manually fitted to the
table values of the index of refraction. The values used in the simulation of the refractive
index of aluminum can be seen in Table 6.2.
1 2
A [s−2 ] 0.65 9.0 · 1031
γ [s−1 ] 1.25 · 1014 8.75 · 1014
ω0 [s−1 ] 0 2.35 · 1015
ω p [s−1 ] 2.40 · 1016 ·
Table 6.2: Estimated values for the parameters involved in the simulation
of the refractive index of aluminum. A1 is without unit.
The simulation is seen in Figure 6.3 together with the table values of n and k. Again, the
Aluminum
9
n table values
k table values
8 n simulation
k simulation
5
n and k
0
350 400 450 500 550 600 650 700 750 800
Wave length [nm]
Figure 6.3: A manual fit to table values of n and k for aluminum. [CD 2004,
45
small derivation in especially n implies other effects that are not accounted for in the model
or simply just that the manual fit is not optimal. The model does however explain the form
and to some extend the size of the curve.
6.4 Silver
The simulation of the index of refraction of silver can be done fairly simple. There are no
resonance peaks that affect the simulation spectrum, so the model is simply described by
means of the free-electron model as

ωp
ñ = A 1 − (6.10)
ω(ω + jγ)
Using a normal table value of the electron density yields a plasma frequency that does
not fit the model of the refractive index to the table values. This is because the effective elec-
tron density becomes lower because the d-electrons are shielding the free electrons. Taking
this effect into account causes a plasma frequency at the wavelength λ = 326 nm. The damp-
ing coefficient can be found by using the table value τ = 4.0 · 10−14 . An amplification factor
is used to fit the simulation to the table values. The values of the parameters are given as
A = 5.3
γ = 2.5 · 1013 s−1

ω p = 5.77 · 1015 s−1
The simulation graph is shown in Figure 6.4.
46
6.4. SILVER
Silver
6
n table values
k table values
n simulation
k simulation
5
4
n and k
0
350 400 450 500 550 600 650 700 750 800
Wave length [nm]
Figure 6.4: A manual fit to table values of n and k for silver. [CD 2004,
47
Discussion of Crystal
Measurements 7
This chapter presents the data processing and the discussion of the measurements performed with the
ellipsometer on the silicon, copper, aluminum and silver crystal. The purpose of the experiment is to
determine the refractive index of the crystals.
Figure 7.1 and 7.2 depicts the real and the imaginary part of the calculated refractive
index as a function of wavelength together with table values for the four crystals. For further
reference on the measurement procedure see Appendix H. By contemplating Figure 7.1(a)
it can be seen that there is a relatively good correspondence between the table values and
the calculated values of the refractive index of silicon. However it can be seen that the
imaginary part of refractive index is higher than the table values and the real is lower over
the measured spectrum.
Figure 7.1(b) depicts the calculated values of the refractive index of copper together with
the table values. The figure shows that the calculated values of both the real and the imag-
inary part are lower than the table values. The figure furthermore shows that even though
the calculated values are lower than the table values they still have the same tendency, over
the measured spectrum, as the table values.
In Figure 7.2(a) the calculated refractive index of aluminum together with the table val-
ues are depicted. It can be seen that the calculated values are relatively far from the table
values but that they have the same tendency as a function of wavelength.
Figure 7.2(b) shows the table and the calculated values in the case of silver. In this case
the calculated values are also relatively far from the table values. The imaginary part of the
calculated refractive index for silver has the same tendency as the table values. In the lower
part of the spectrum the calculated values fit the table values well. The real part however,
does not have the same tendency as a function of wavelength. This is because the table
values decreases as the wavelength is increasing and the calculated values increases as the
wavelength is increasing.
In order to improve these results a new model of calculation is considered. The new
model of calculation will used the same measured data as the old. In the new model it is
assumed that the surfaces of the crystals are oxidized. It is assumed that there is a thin
layer of silicon dioxide SiO2 on the silicon surface, a thin layer of cuprous oxide Cu2 O on
the copper surface, a thin layer of aluminum oxide Al2 O3 on the aluminum surface and
a thin layer of silver oxide Ag2 O on the silver surface. It is required that the indexes of
refraction of the four oxide layers are known. The index of refraction of silicon dioxide,
cuprous oxide and aluminum oxide are found in [Palik 1998]. The index of refraction of
silver oxide is found in [Yin et al. 2001]. The unknown variables in the new model are the
index of refraction of the crystal below the oxide layer and the thickness of the oxide layer.
As it is the refractive index of the crystal below the oxide layer which is to be determined,
the calculations will be made as an iterative process where the thickness of the oxide layer is
changed. In order to calculate the refractive index of a material below a thin film where the
49
CHAPTER 7. DISCUSSION OF CRYSTAL MEASUREMENTS
Silicon
7
n
k
4
n and k
0
350 400 450 500 550 600 650 700 750 800
Wavelength [nm]
(a)
Copper
6
n
k
4
n and k
0
350 400 450 500 550 600 650 700 750 800
Wavelength [nm]
(b)
Figure 7.1: The calculated real n and imaginary k part of the refractive
index as a function of wavelength together with table values of the re-
fractive index of silicon and copper. The real part of the table values
are dots and the imaginary part are asterisks. Table values are found
in [Palik 1998] and in [Klein & Furtak 1986]. [CD 2004, matlab/sur-
face_test/RefractiveIndex.m]
50
Aluminum
9
n
k
8
5
n and k
0
350 400 450 500 550 600 650 700 750 800
Wavelength [nm]
(a)
Silver
6
n
k
4
n and k
0
350 400 450 500 550 600 650 700 750 800
Wavelength [nm]
(b)
Figure 7.2: The calculated real n and imaginary k part of the refractive
index as a function of wavelength together with table values of the refrac-
tive index of aluminum and silver. The real part of the table values are dots
and the imaginary part are asterisks. Table values are found in [Palik 1998].
[CD 2004, matlab/surface_test/RefractiveIndex.m]
51
refractive index and thickness of the film are known, (5.6) and (5.8) are combined to yield
ρ12,π ρ01,σ ρ12,σ e− j4β + (ρ01,π ρ01,σ ρ12,σ + ρ12,π ) e− j2β + ρ01,π
tan(ψ)e jΔ = (7.1)
ρ01,π ρ12,π ρ12,σ e− j4β + (ρ01,π ρ12,π ρ01,σ + ρ12,σ ) e− j2β + ρ01,σ
Given the refractive indexes of air n0 and of the oxide layer ñ1 , the angle of incidence θ0 , the
wavelength of the light in free-space λ and Snell’s law it is possible to calculate ρ01,σ and ρ01,π
for the given wavelength. From the measurements, ψ and Δ are connected to the wavelength
λ. Substituting the expressions for ρ12,π and ρ12,σ and β given in Appendix C into (7.1) yields
an equation in two variables, namely the thickness d of the oxide layer and the index of
refraction of the crystal below the oxide layer ñ2 . In order to solve this equation for ñ2 the
thickness of the oxide layer must be guessed or an iterative calculation, where the thickness
is changed, must be performed. When a thickness is decided upon, (7.1) can be solved in
order to find the refractive index of the crystal below the film for a given wavelength.
7.1 Silicon
A series of calculations where ten different thicknesses of a silicon dioxide film are imposed
on the silicon surface and (7.1) is solved for ñ2 over a wide range of wavelengths are per-
formed. The interval of the thickness is from 1 nm to 10 nm with a jump in thickness of 1 nm.
Only the best fit of the ten different thicknesses is depicted in Figure 7.3. The figure shows
the real and imaginary part of the calculated refractive index of silicon without a film and
with a 4 nm silicon dioxide film imposed on the silicon surface together with the table val-
ues. It can be seen from the figure that in order to move the refractive index of silicon closer
to the table values, a 4 nm silicon dioxide film can be imposed on the silicon surface. This
indicates that there indeed is a film of silicon dioxide or a material with similar refractive
index on the silicon surface. From the figure it can be seen that, due to the low resolution
of points where the refractive index is calculated, the curve of the real part of the calculated
refractive index is edged and non-fluctuating.
52
7.1. SILICON
Silicon
7
n
k
4
SiO2 4 nm
n and k
SiO2 0 nm
SiO2 0 nm
SiO2 4 nm
0
350 400 450 500 550 600 650 700 750 800 850 900
Wavelength [nm]
Figure 7.3: Real and imaginary part of the calculated refractive index of
silicon without a film and with a 4 nm silicon dioxide film imposed on the
silicon surface together with the table values. [CD 2004, matlab/refrac-
tive_index_si/refractive_index_si.m]
53
7.2 Copper
A series of calculations where ten different thicknesses of a cuprous oxide film are imposed
on the copper surface and (7.1) is solved for ñ2 over a wide range of wavelengths are per-
formed. The interval of the thickness is from 1 nm to 10 nm with a jump in thickness of 1
nm. Only the best fit of the ten different thicknesses is depicted in Figure 7.4. The figure
Copper
6
n
k
Cu2O 8 nm
5
Cu2O 0 nm
4
n and k
Cu2O 8 nm
Cu2O 0 nm
0
350 400 450 500 550 600 650 700 750 800 850 900
Wavelength [nm]
copper without a film and with an 8 nm cuprous oxide film imposed on
the copper surface together with the table values. [CD 2004, matlab/re-
fractive_index_cu/refractive_index_cu.m]
shows the real and imaginary part of the calculated refractive index of copper without a
film and with an 8 nm cuprous oxide film imposed on the copper surface together with the
table values. The calculated refractive index of copper without a film is drawn in order to
compare the new model with the old model. It can be seen from the figure that in order to
move the refractive index of copper closer to the table values, an 8 nm cuprous oxide film
can be imposed on the copper surface. This indicates that there indeed is a film of cuprous
oxide or a material with similar refractive index on the copper surface.
54
7.3. ALUMINUM
7.3 Aluminum
A series of calculations where ten different thicknesses of an aluminum oxide film are im-
posed on the aluminum surface and (7.1) is solved for ñ2 over a wide range of wavelengths
are performed. The interval of the thickness is from 1 nm to 10 nm with a jump in thick-
ness of 1 nm. The results for three essential thicknesses besides the model without film are
depicted in Figure 7.5. The figure shows the results of the calculations where no film and
Aluminum
9
n
k
Al2O3 9 nm
8
Al2O3 7 nm
Al2O3 5 nm
7
Al2O3 0 nm
5
n and k
4
Al2O3 9 nm
Al2O3 7 nm
3 Al2O3 5 nm
Al2O3 0 nm
2
0
350 400 450 500 550 600 650 700 750 800 850 900
Wavelength [nm]
aluminum without a film, with a 5, 7 and 9 nm aluminum oxide film on
top of the aluminum surface together with the table values. [CD 2004,
matlab/refractive_index_al/refractive_index_al.m]
a 5, 7 and 9 nm thick aluminum oxide layer are imposed on the aluminum surface. As can
be seen from the figure there is a clear tendency of a better result when an aluminum oxide
film is imposed on the surface. In order to fit the real part best it can be seen that a 5 nm alu-
minum oxide film must be imposed on the aluminum surface. However in order to fit the
imaginary part of the refractive index best, a 9 nm aluminum oxide film must be imposed.
This indicates that the new model is still not sufficient to get a good fit of both the real and
55
the imaginary part of the refractive index of aluminum.

An explanation could be that it is not pure aluminum oxide that forms on the aluminum
surface. Thus the refractive index of the film currently used in the calculations is not correct.
7.4 Silver
Figure 7.6 depicts the result of the calculation of ñ2 where the thickness of a silver oxide
layer is 0, 5 and 10 nm. Only the result of a 10 nm, a 5 nm and a 0 nm thick silver oxide
Silver
6
n
k
Ag2O 10 nm
5
Ag2O 5 nm
Ag2O 0 nm
4
n and k
1 Ag2O 10 nm
Ag2O 5 nm
Ag2O 0 nm
0
300 400 500 600 700 800 900
Wavelength [nm]
silver without a film, with a 5 and 10 nm silver oxide film on top of the
aluminum surface together with the table values. [CD 2004, matlab/re-
fractive_index_ag/refractive_index_ag.m]
film are depicted in the figure. It can be seen from the figure that when a silver oxide layer is
imposed there is a lower correspondence between the calculated values and the table values.
The best fit to the table value of the refractive index is when no silver oxide film is imposed
on the surface of the crystal.
56
7.5. SUMMARY
The assumption that it is only Ag2 O is probably not adequate. Using another refractive
index could to some extend remedy the problem. Generally, a silver crystal structure tends
to grow argentite Ag2 S which has a different refractive index.
7.5 Summary
In general it can be observed that there is a better fit when the new model is used instead
of the old model, where no oxide layer is imposed on the surface. The best fits are in the
case of silicon and copper, which indicates that it is in fact a good approximation to assume
that it is silicon dioxide and cuprous oxide that forms on the respective surfaces. The small
deviations from the table values are considered within the range of test uncertainties. In the
case of the aluminum, the model with the oxide layer makes a better fit, but a deviation from
the table values still exists.
From the result of silver, it can be concluded that neither the model with pure silver nor
the model with a layer of Ag2 O imposed on the silver surface are correct. A model with a
different film layer, e.g. Ag2 S must be tested. Due to difficulties obtaining table values of
the refractive index of Ag2 S, this has not been tested.
57
Discussion of SiO2 Film
Measurements 8
This chapter presents and discusses the measurement of film thickness of an SiO2 film on an Si
surface using the SE 850 ellipsometer. The procedure of making the measurements are presented in
Appendix I, and the method of calculating the thickness of the film is presented in Chapter 5.
The measurements that underlie the calculation of film thickness are the same type as
in Chapter 7. That is, it is based on measurements of ψ and Δ angles. A depiction of ψ and
Δ angles as a function of wavelength for a measurement on the Si wafer with an SiO2 film
can be found in Figure I.1 on page 124. It is noted that the Δ angles fluctuate in the higher
wavelength region.
8.1 Film Thickness

The thickness is calculated by means of the method described in Section 5.2.
As the ellipsometer measure ψ and Δ angles for approximately every 1 nm within the
described spectrum, it is necessary to have the corresponding values of the refractive index
for Si and SiO2 in order to be able to calculate the thickness for each measured wavelength.
Thus, an interpolation of table values is performed. Table values for Si and SiO2 are from
[Palik 1998, I: pp. 563-565] and [Index of Refraction Values and Photonics Calculations 2004]
respectively. The interpolation is carried out as a simple linear interpolation and the re-
turned curves are generally smooth.
Graphical illustrations of the calculated thicknesses of the film as functions of the wave-
length can be found in Figure 8.1(a) and Figure 8.1(b). The first figure presents the thickness
result in a large scale, whereas the second figure zooms in, in order to reveal the curvature
of the results.
The actual thickness does obviously not alter with respect to the wavelength. Hence, the
calculated thickness would ideally be a constant. This is however not the case. Generally,
some fluctuations, which are most pronounced for the larger wavelength, can be seen in
Figure 8.1(a) and 8.1(b). This can be traced back to the fluctuations in the measurement
of Δ angles in the same wavelength area, as this is the only parameter that does fluctuate.
This is similar to Si which is seen in Figure H.2(a) on page 121 and the fluctuations are
caused by low light intensities for the mentioned wavelengths. For a graphical confirmation,
comparison between Figure I.1 and 8.1(a) can be made.
The appearance of the curve for the small wavelength region is affected by fluctuations
in the table values of the imaginary part of the refractive index for Si. Si fluctuates slightly
especially in the region from approximately 400 to 500 nm.
It is estimated that the result is most reliable in the region where the measured para-
meters and the table values are most stable. Thus, it should be in the region from approxi-
mately 500 to 600 nm. In this region, the thickness does however have a tendency to increase
59
CHAPTER 8. DISCUSSION OF SIO2 FILM MEASUREMENTS
Film thickness of SiO2 on Si

30
25
20
Film thickness [nm]
15
10
0
350 400 450 500 550 600 650 700 750 800 850
Wavelength [nm]
(a) Plot showing the uniformity
Film thickness of SiO2 on Si

23.3
23.2
23.1
Film thickness [nm]
23
22.9
22.8
22.7
22.6
350 400 450 500 550 600 650 700 750 800 850
Wavelength [nm]
(b) Plot showing the heterogeneity
Figure 8.1: the thickness of the SiO2 film on a Si surface as a function of

wavelength. Note that the fluctuations in the larger wavelength region
resemble the fluctuation in the values for Δ. [CD 2004, matlab/thickness/-
film_thickness_sio_2.m]
60
8.1. FILM THICKNESS
slightly. Nevertheless, it is estimated that a thickness of approximately 22.9 nm is a good es-

timate of the actual thickness. This value does also coincide with an average over the entire
spectrum. Furthermore, this estimated value coincides with the automatic fit of the thick-
ness by the Spectraray II software, which yields a thickness of 22.8 nm. From this, it can
be concluded that the model is adequate. The manufacturer of the film does however state
that the thickness of the film is 21.5±0.3 nm. This corresponds to a deviation of the mea-
sured value from this value of 6.5 %. An explanation to the difference between the thickness
stated by the manufacturer and the measured/calculated value can be that either the ellip-
someter is not calibrated accurately or the thickness of the film is not as specified by the
manufacturer.
61
Discussion of Polymer Film
Measurements 9
This chapter presents the discussion of the measurements performed on the two wafers with a PI-
5878G polymer film imposed on the surface. The test report of the experiments is found in Appendix J.
The first subject of the discussion will be the experiment where the thickness of the two polymer films
is calculated as a function of wavelength. The second subject of discussion is the experiment where
the uniformity of the thickness of the polymer films is investigated.
In the following it should be possible to distinguish between the two polymer films.
They are spin coated with the same angular velocity, but for different time intervals. The
wafer where the polymer is spin coated for the shortest time interval will be the thickest
and will therefore in the following be referred to as the thick polymer. The wafer where the
polymer is spin coated in the longest time interval will be the thinnest and will therefore in
the following be referred to as the thin polymer.
Both polymers are optically thick, that is, they are several times thicker than the wave-
length of the incident light.
9.1 Film Thickness

Figure 9.1 shows the calculated thickness as a function of wavelength for the thick and the
thin polymer. In the ideal case the thicknesses of the two polymer films should be inde-
Polymer Thick Polymer Thin

7.5 4.4
7
4.2
6.5
4
Thickness [um]
Thickness [um]
3.8
5.5
3.6
5
3.4
4.5
4 3.2
450 500 550 600 650 700 750 800 450 500 550 600 650 700 750 800
Wavelength [nm] Wavelength [nm]
(a) Thickness of the thick polymer at the middle, as (b) Thickness of the thin polymer at the middle, as
a function of wavelength. a function of wavelength.
Figure 9.1: Polymer thickness at the middle of the wafer, as a func-

tion of wavelength. [CD 2004, matlab/polymer_thickness/film_thick-
ness_polymer_spectrum.m]
63
CHAPTER 9. DISCUSSION OF POLYMER FILM MEASUREMENTS
pendent of the wavelength λ. That is, the thickness of the film should be constant for all
wavelengths. By contemplating Figure 9.1 it can be seen that the calculated film thickness
of both the thick and the thin polymer film is not constant over wavelength. The figures
show that the calculated polymer film thickness is larger in the lower part of the wavelength
range. The thicknesses of the films are calculated by means of the interference pattern in the
ψ-Δ spectrum. By contemplating the ψ-Δ spectrums of the polymer films in Appendix J, it is
observed that the interference patterns in the lower part of the wavelength interval are not
as ambiguous as in the upper part. This is a possible reason for the larger thickness in the
lower part of the wavelength interval. A possible explanation of the variations in the calcu-
lated thickness could be that the value of the refractive index of the polymer is not constant
1.78 as specified in [MicroSystems 2003]1 . If white light is incident on an isolated part of the
film, i.e. a part without the wafer, it can be observed that the transmitted part of the light
has a tendency to contain a greater part of yellow, compared to the rest of the spectrum. The
reflected part does however not contain a relative larger part of the blue and red spectrum,
i.e. the red and blue part of the spectrum is not reflected more than the yellow part. Thus,
light in the red and blue part of the spectrum is absorbed more than it is for the yellow part.
This again results in the existence of an imaginary part of the refractive index different from
zero and dependent of the wavelength.
9.2 Uniformity of Film Thickness

Figure 9.2 shows the calculated thickness over the axes, as defined in Appendix J, for both
the thin and the thick polymer film. The figure shows the thickness of the two polymer

6 6
x u
y v
5 5
4 4
Thickness [um]
Thickness [um]
3 3
2 2
1 1
0 0
−40 −30 −20 −10 0 10 20 30 40 −40 −30 −20 −10 0 10 20 30 40
Position [mm] Position [mm]
(a) Thickness of the thick polymer film measured (b) Thickness of the thin polymer film measured
over the x- and the y-axis as depicted in Appendix J over the u- and the v-axis as depicted in Appendix J
Figure 9.2: Uniformity of thickness of the two polymer films. [CD 2004,
matlab/polymer_thickness/film_thickness_polymer.m]
films in a full scale. It can be seen from the figures that there are no notable variations in the
thickness as a function of place on the surface. It is furthermore seen that, as expected the
1 This is the only available material stating the refractive index of the polymer
64
9.2. UNIFORMITY OF FILM THICKNESS
thick polymer film is thicker than the thin polymer film. Zooming in on the thick polymer
film yields the plot shown in Figure 9.3. The figure shows the variations in the thickness as
Polymer Thick
4.25
x
y
4.2
4.15
Thickness [um]
4.1
4.05
3.95
−40 −30 −20 −10 0 10 20 30 40
Position [mm]
Figure 9.3: Thickness of the thick polymer, zoomed in. [CD 2004, mat-
lab/polymer_thickness/film_thickness_polymer.m]
a function of the position. It can be seen that the film is thickest at the middle of the wafer
and gets thinner at the edge. This makes sense because the polymer film is spin coated on
the wafer. This is because the velocity of the edge of the wafer is higher than the velocity
near the center of the wafer during the spin coating. The mean value of the thickness of the
thick polymer film Ethick is calculated to
Ethick = 4.17 µm (9.1)
The maximum deviation from the mean value for the thick polymer film is calculated to
Smax,thick = 4.2 % (9.2)
Zooming in on the thin film yields the plot in Figure 9.4. Furthermore it can be seen that
the thin polymer film has a vague tendency of being thickest at the middle of the wafer and
thinner towards the edge of the wafer. This is however not as noticeable as it is for the thick
65
CHAPTER 9. DISCUSSION OF POLYMER FILM MEASUREMENTS
Polymer Thin
3.38
u
v
3.36
3.34
3.32
Thickness [um]
3.3
3.28
3.26
3.24
3.22
3.2
−40 −30 −20 −10 0 10 20 30 40
Position [mm]
Figure 9.4: Thickness of the thin polymer, zoomed in.[CD 2004, mat-
lab/polymer_thickness/film_thickness_polymer_thin.m]
film. The mean value of the thickness of the thin polymer film Ethin is calculated to
Ethin = 3.32 µm (9.3)
The maximum deviation from the mean value for the thin polymer film is calculated to
Smax,thin = 3.0 % (9.4)
It is noted that both the wafers, with the imposed film, show a ring interference pattern.
This indicates that the thickness of the film is a function of the distance from the center.
9.3 Summary
In general the calculated polymer film thickness changes as a function of wavelength, where
the thickness is larger in the lower part of the wavelength spectrum. The polymer film thick-
ness as a function of position on the wafer is almost constant with a maximum deviation of
4.2 % from the mean. Both wafers tend to be slightly thicker at the middle than at the edge.
66
Conclusion
The purpose of this project is to determine the complex index of refraction and the film
thickness of different materials by use of ellipsometry.
The index of refraction as a function of wavelength is determined for four crystals from
ellipsometric measurements of ψ and Δ. The crystals are silicon, copper, aluminum and
silver. First, the index of refraction is calculated from a model where the crystals are consid-
ered pure. This gives results which are more or less equal to the respective table values. For
silicon, copper and aluminum the tendency for the refractive index depicted as a function
of the wavelength is the same as for the table values, but displaced slightly in magnitude.
Silver has a different tendency for the real part of the refractive index.
In order to improve the results, the models are expanded to include thin layers of oxide.
This improves the individual results considerably. Especially the refractive indexes for sil-
icon and copper are hereafter very close to the table values. The small deviation from the
table values is considered to be within a margin of experimental uncertainty. For aluminum
and silver the results still do not coincide with table values. Optimizing the ambient-film-
substrate model by changing the film material could improve the result of the calculated
refractive indexes of the substrates. A silver surface tends to grow argentite and not silver
oxide. This is not treated due to difficulties retrieving table values for the refractive index.
The tendency in the refractive index of the materials is theoretically explained by Drude-
Lorentz models and band gap structures of the crystals. By using these models it has proven
possible to simulate the refractive index of the four crystals within a reasonable margin.
The thickness of three films is calculated from measurements of ψ and Δ. The thickness
of a layer of silicon dioxide on a substrate of silicon is determined from ψ-Δ measurements
to be 22.9 nm. This value coincides with the automatic fit of the thickness by the Spectraray
II software, which yields 22.8 nm. From this it can be concluded that the model is adequate.
The manufacturer of the film states however that it is 21.5 ± 0.3 nm. An explanation to the
difference can be that either the ellipsometer is not calibrated accurately or the thickness of
the film is not as specified.
The thickness of two polymer films is calculated. The two films consist of the same
type of polymer, but with different thickness. Compared to the silicon dioxide film, the
calculations of the polymer film are made by use of additional considerations due to the
interference pattern of ψ and Δ. These calculations yield a film thickness of 3.32 and 4.17 µm.
There are no table values to confirm these results, but the results are accepted by NanoNord
A/S. The results are therefore considered to be satisfying.
The uniformity of the polymer film is calculated by means of 34 individual measure-
ments on each of the two films. The maximum deviations from the mean values of the
thickness are 3.0 % and 4.2 % for the thin and the thick polymer film, respectively. The thick
polymer film shows a slight tendency to decrease in thickness towards the edge, which can
be explained by the higher speed at the edge during the spin coating of the film. Generally,
the two films can be taken as uniform.
The goal of this project has been to measure the refractive index of several materials and
the thickness of films imposed on a substrate. This goal is achieved with satisfying results.
67
Bibliography
Azzam, R. M. A. & Bashara, N. M. [1977], Ellipsometry and Polarized Light, 1st ed, North-
Holland Publishing company. ISBN 0-7204-0694-3.
CD [2004], ‘Enclosed cd by group 116’.
Classical Light [2004], http://www-gap.dcs.st-and.ac.uk/~history/

HistTopics/Light_1.html.
Ehrenreich, H. & Philipp, H. R. [1962], Optical Properties of Ag and Cu, General Electric Re-
search Laboratory Schenectady, New York.
Eric W. Weisstein, Geometric Series [2004], http://mathworld.wolfram.com/

GeometricSeries.html. Geometric Series. From MathWorld–A Wolfram Web Re-
source.
Index of Refraction Values and Photonics Calculations [2004], http://www.luxpop.

com/.
Jawoollam [2004], http://www.jawoollam.com/Tutorial_index.html.
Kittel, C. [1986], Introduction to Solid State Physics, 6th ed, John Wiley and Sons Inc. ISBN
0-471-87474-4.
Klein, M. V. & Furtak, T. E. [1986], Optics, 2nd ed, John Wiley and Sons Inc. ISBN 0-471-
87287-0.
Kreyszig, E. [1999], Advanced Engineering Mathematics, 8th ed, John Wiley and Sons Inc. ISBN
0-471-33328-0.
Merriam-Webster Online [2004], http://www.m-w.com/.
MicroSystems, H. [2003], PI-5878G Wet Etch Polyimide, 1st ed, HD MicroSystems.
Palik, E. D. [1998], Handbook of Optical Constants of Solids, Academic Press. ISBN 0-12-544420-
6.
Poksinski, M. [2003], Total Internal Reflection Ellipsometry, 1st ed, Institute of Technology,
Linköping University. Licentiate Thesis No. 1016.
Råde, L. & Westergren, B. [1998], Mathematics Handbook for Science and Engineering, 4th ed,
Studentlitteratur. ISBN 91-44-00839-2.
69
BIBLIOGRAPHY
Reitz, J. R., Milford, F. J. & Christy, R. W. [1993], Foundations of Electromagnetic Theory, 4th ed,
Addison-Wesley Publishing Company. ISBN 0-201-52624-7.
Rohlfing, M., Krüger, P. & Pollmann, J. [1993], Quasiparticle Band-Structure for C, Si, Ge, GeAs
and SiC using Gaussian-Orbital Sets, Institut für Theoretische Physik II - Festkörper-
physik, Universität Münster Germany.
Röseler, A. [1990], Infrared Spectroscopic Ellipsometry, 1st ed, Akademie-Verlag Berlin. ISBN
3-05-500623-2.
Spectroscopic Ellipsometer SE 850 [2004], http://www.sentech.com/se850.htm.
Yin, Y., Li, Z.-Y., Zhong, Z. & Gates, B. [2001], Synthesis and Charactization of stable aqueous
dispersions of silver nanoparticles through the Tollens process, 1st ed, Journal of Materials
Chemistry.
70
The Electromagnetic Wave
Model of Light A
The purpose of this appendix is to show that light can be described as electromagnetic waves through
Maxwell’s equations. Furthermore this appendix presents a solution to the wave equations describing
light.
It has been known for a long time that light has wave properties, but it was not before
Maxwell published his equations that the question of ”what was waving ?” was solved.
In 1862 Maxwell found out that electromagnetic waves are related to light when he real-
ized that they travel at the same speed in free-space. In 1864 Maxwell wrote a paper in
which he stated: ”This velocity is so nearly that of light that it seems we have strong reason to
conclude that light itself is an electromagnetic disturbance in the form of waves propagated through
the electromagnetic field according to electromagnetic laws” [Classical Light 2004]. It was then
that Maxwell realized that light itself can be described as an electromagnetic field in form
of waves propagating through space and time. In 1873 Maxwell published his paper, Elec-
tricity and Magnetism, where he had developed four partial differential equations, know as
Maxwell’s equations, which completely describes the classical electromagnetic wave theory.
It was therefore assumed that light could be described through Maxwell’s equations.
Maxwell’s Equations of Light

Maxwell’s equations are four partial differential equations, which describes how the electric
and the magnetic fields are coupled in an electromagnetic field waving through space and
time. Maxwell’s equations are in general given as
∇·D = ρ (A.1a)
∇·B = 0 (A.1b)
∂B
∇×E = − (A.1c)
∂t
∂D
∇ × H = (J + ) (A.1d)
∂t
where
D = ε0 E + P (A.2a)
B = µ0 (M + H) (A.2b)
where D is the electric displacement, H is the magnetic intensity, E is the electric field, B
is the magnetic field, J is the current density, P is the polarization, M is the magnetization,
ρ is the charge density, ε0 is the electric permittivity of free-space (vacuum) and µ0 is the
magnetic permeability of free-space.
71
APPENDIX A. THE ELECTROMAGNETIC WAVE MODEL OF LIGHT
In order to show that light is an electromagnetic wave, it is assumed that light is propa-
gating through free-space. In free-space there are no charge distributions and therefore no
charges which can be polarized or magnetized. That is, P = 0, M = 0, J = 0 and ρ = 0. This
yields
D = ε0 E (A.3a)
B
H= (A.3b)
µ0
which reduces Maxwell’s equations for light propagating through free-space to
∇·E = 0 (A.4a)
∇·B = 0 (A.4b)
∂B
∇×E = − (A.4c)
∂t
∂E
∇ × B = µ0 ε0 (A.4d)
∂t
Maxwell’s equations for light can also be written in integral form which has the form
E · dA = 0 (A.5a)
s
B · dA = 0 (A.5b)
s
−∂
E · dr = B · dA (A.5c)
k ∂t
∂
B · dr = µ0 ε0 E · dA (A.5d)
k ∂t
To show that Maxwell’s equations of light leads to wave equations, the curl is taken of both
sides of (A.4d)

∂E
∇ × ∇ × B = ∇ × ε0 µ0 (A.6)
∂t
Using a vector identity [Reitz et al. 1993, p. 20] on the left side and rearranging terms of the
right side of (A.6) yields
∂
∇(∇ · B) − ∇ · ∇B = µ0 ε0 (∇ × E) (A.7)
∂t
Now using (A.4b) and (A.4c)
∂2 B
∇2 B − µ0 ε0 =0 (A.8)
∂t 2
or
1 ∂2 B
∇2 B − =0 (A.9)
c2 ∂t 2
72
where c, the speed of light, is given as c = (µ0 ε0 )−1/2 . With similar manipulations an equation
describing the electric field can be obtained as
1 ∂2 E
∇2 E − =0 (A.10)
c2 ∂t 2
From (A.9) and (A.10) it is seen that Maxwell’s equations for light leads to two 3-dimensional
wave equations; one for the magnetic field and one for the electric field, both propagating
through space and time with the speed of light c in free-space.
Solving the Wave Equations of Light

In general the solutions of the wave equations of light, (A.9) and (A.10), are vector functions
of four variables namely, x, y, z and t. That is
E = E(r,t) (A.11)
B = B(r,t) (A.12)
where r is a 3-dimensional vector r = xx̂ + yŷ + zẑ where x̂, ŷ and ẑ are the unit vectors along
the respective coordinate axes. These general solutions are rather complex to calculate and
the problem is therefore simplified by assuming a plane wave and by placing the system of
coordinates such that the wave propagates along the z-axis. For a plane wave propagating
along the z-axis the disturbance over each xy-plane is constant and the wave equation of the
electric field in (A.10) 1 is therefore reduced to
∂2 E(z,t) 1 ∂2 E(z,t)
− 2 =0 (A.13)
∂z2 c ∂t 2
or
∂2 E(z,t) 2 ∂ E(z,t)
2
= c (A.14)
∂t 2 ∂z2
From (A.13) or (A.14) it can be seen that the equation for a plane wave propagating along the
z-axis is a 1-dimensional wave equation, which is a second order partial differential equation
in two variables namely z and t. This partial differential equation can be solved be means of
the Laplace transformation if the initial conditions are given. It is assumed that the electric
field at t = 0 is zero and that the derivative of the electric field with respect to time at t = 0 is
zero. 2 That is the initial conditions are
E(z, 0) = 0 (A.15a)

∂E
=0 (A.15b)
∂t t=0
1 The
wave equation of the magnetic field is reduced in a similar way.
2 The
solution of the partial differential of a plane wave propagating along the z-axis is given for the electric
field but can easily be found for the magnetic field by interchanging E with B.
73
APPENDIX A. THE ELECTROMAGNETIC WAVE MODEL OF LIGHT
Taking the Laplace transform with respect to t of (A.14) yields with reference to [Kreyszig
1999, p. 259 (2)]
2 2
∂ E(z,t) ∂E ∂ E(z,t)
L = s L {E(z,t)} − sE(z, 0) −
2
=c L
2
(A.16)
∂t 2 ∂t t=0 ∂z2
where s is the complex Laplace variable. The Laplace transform with respect to t of the
partial second order derivative of the electric field with respect to z is given as
∞ ∞
∂2 E(z,t) ∂2 E(z,t) ∂2 ∂2
L = e−st dt = e−st E(z,t)dt = L {E(z,t)} (A.17)
∂z2 0 ∂z2 ∂z2 0 ∂z2
Comparing (A.16), (A.17) and using the initial conditions from (A.15) yields
∂2 s2
EL (z, s) − EL (z, s) = 0 (A.18)
∂z2 c2
where EL (z, s) = L {E(z,t)}. Since this equation only contain derivatives with respect to z
it can be considered as an ordinary differential equation of EL (z, s), which has a general
solution given as
EL (z, s) = F(s)e−sz/c + G(s)esz/c (A.19)
Taking the inverse Laplace transform of (A.19) and making use of the Laplace shifting theo-
rem yields
z z
E(z,t) = f t − +g t + (A.20)
c c
Thus the general solution for a plane wave propagating along the z-axis is given by (A.20).
A special case of (A.20) is a harmonic traveling plane light wave where g = 0
z z
E(z,t) = f t − = E0 cos ω t − +φ (A.21)
c c
where E0 is a constant vector associated with the maximum size of the oscillation, φ is the
phase at z = 0 and t = 0, ω = 2πν and ν is the frequency in hertz. If the normal of the wave-
plan is not along the z-axis but pointing in an arbitrary direction along the unit vector ŝ the
monochromatic3 plane wave yields

r · ŝ r · ŝ
E(r,t) = f t − = E0 cos ω t − +φ (A.22)
c c
where r is a vector pointing from the origin of the coordinate system to an arbitrary point
on the plan with the normal vector ŝ and E0 is a constant vector equal to the maximum
oscillation. If the wavevector k is introduced as
ω
k= ŝ (A.23)
c
3 Radiation of a single wavelength, that is ω is constant over time.
74
the monochromatic plane wave is given as

E(r,t) = E0 cos [ωt − k · r + φ] = E0 Re e j(ωt−k·r+φ) (A.24)
In similar ways the solution to (A.9), the wave equation for the magnetic field, can be found
as

B(r,t) = B0 Re e j(ωt−k·r+φ) (A.25)
Thus the solution to Maxwell’s equations for monochromatic plane light waves are given as
E = E0 e j(ωt−k·r+φ) (A.26a)
B = B0 e j(ωt−k·r+φ)
(A.26b)
where k is the wavevector, defined as in (A.23), along which the light waves propagates.
The ”Re” notation is dropped knowing that the physical fields are the real part of (A.26).
[Klein & Furtak 1986, pp. 22-23], [Kreyszig 1999, pp. 251-273]
Monochromatic Plane Light Waves

Substituting the solution of the monochromatic plane light wave from (A.26) into Maxwell’s
equation (A.4c) yields
− jk × E = − jωB (A.27)
or
k×E
B= (A.28)
ω
or by substituting with (A.23)
1
B = ŝ × E (A.29)
c
Thus by (A.28) a monochromatic plane light wave has an electric field perpendicular to
a magnetic fields which again propagates along a wavevector perpendicular to both the
electric and magnetic field. Furthermore by (A.29) the electric field and the magnetic field
are in phase. [Klein & Furtak 1986, pp. 46-47]
75
Fresnel Reflection and
Transmission Coefficients B
The purpose of this appendix is to describe the reflection and transmission of light waves at an in-
terface between two media. The interface is assumed to be smooth and abrupt, and the two media at
the interface are assumed homogeneous, linear and isotropic. The incident light is divided into two
different polarization states, of which one is analyzed further using among other things Maxwell’s
equations. This appendix is based on [Klein & Furtak 1986, pp. 71-80].
When an electromagnetic wave of light is incident on a boundary between two materials
with different index of refraction, a situation as the one depicted in Figure B.1 occur. The
surface
reﬂected light
transmied light
θ’’
θ’
ñ y ñ’ z
θ
incident light
Figure B.1: Situation where incident light, in the xz-plane is propagating

towards a surface, in the xy-plane. After interaction with the surface, part
of the light is transmitted and part of it is reflected.
surface lies in the xy-plane and the plane of incident is the xz-plane. Part of the wave is
transmitted and part of it is reflected. It can readily be proven which direction the reflected
and transmitted wave will have by use of Fermat’s principle or Huygens’ construction.
The focus in this appendix will, however, be on a derivation of the reflection coefficient ρ
defined as Ere f lected /Eincident and the transmission coefficient τ defined as Etransmitted /Eincident .
77
APPENDIX B. FRESNEL REFLECTION AND TRANSMISSION COEFFICIENTS
Description of Light According to Maxwell’s Equations

Light can be considered to be electromagnetic waves, as described in Appendix A. It con-
sists of a magnetic and an electric field, B and E respectively. They are perpendicular to
each other and to the wavevector k, which is pointing in the direction of propagation. The
connection between B and E is for a monochromatic plane wave given as
k×E
B= (B.1)
ω
as described in (A.28). As E is perpendicular to k, the amplitudes of the electric and mag-
netic field in (B.1) can be expressed as
k
B= E (B.2)
ω
or in a media
ñ
B= E (B.3)
c
where ñ is the complex index of refraction and c is the speed of light in vacuum. The substi-
tution from (B.2) to (B.3) is carried out using the expression given by (F.5).
Definition of Polarization States

It is convenient to split the polarization state of the incoming wave into two components.
These are denoted π- and σ-polarized light1 . Each case has to fulfill (B.1).
π-polarized light: E is in the plane of incidence, i.e. it is parallel to the plane of incidence.
Consequently, B is perpendicular to the plane of incidence.
σ-polarized light: B lies in the plane of incidence, thus E is perpendicular to the plane.
An incident linearly polarized plane wave can be described by a linear combination of these
components.
Derivation of Boundary Conditions

In the following it will be deduced from Maxwell’s equation (A.4c) that the tangential com-
ponent of E is continuous across an interface between to media with different index of refrac-
tion. The situation is illustrated in Figure B.2. The figure is a two-dimensional illustration,
where the interface is marked as a dashed line. A rectangle has been inserted to assist in
deriving the condition. The rectangle has the length l0 , which is finite, yet small enough to
avoid a difference in the electric field along each of the individual sides. The width Δl of the
rectangle is infinitely small (Δl → 0).
The E-fields on the two sides are denoted E1 and E2 , with their respective tangential
component with respect to the interface denoted E1t and E2t .
1π
and σ are the Greek letters for p and s, where p refer to the term parallel, and s refer to the term senkrecht,
which is German for perpendicular [Azzam & Bashara 1977, p. 272].
78
Interface
ñ1
E1
ñ2
-l0
E1t
E2t
E2
l0
Δl
Figure B.2: Depiction of an interface between two media with different

index of refraction. A rectangle, which is placed across the interface, is
used to derive the tangential condition for the electric field.
(A.4c) is in integral form given by the following as described in Appendix A
∂
E · dr = − B · dA (B.4)
k ∂t
which is applied to the area described by the rectangle in Figure B.2. As the bounded area is
infinitely small due to Δl, the integral over dA is zero. Thus (B.4) is reduced to
E · dr = 0 (B.5)
k
where the closed curve integral can be divided into four components given by the sides of
the rectangle. As the two sides with the length Δl are infinitely small, their contributions to
the integral are likewise infinitely small. Thus the curve integral can be expressed by the
two parallel sides
l0 · E2 − l0 · E1 = 0 (B.6)
where the tangential parts of E1 and E2 can be inserted in order to skip the vector notation
l0 E2t − l0 E1t = 0 (B.7)
which results in the following
E1t = E2t (B.8)
By this, it can be concluded that the tangential part of E is continuous across an interface.
In a similar way it can be shown that the tangential component of B/µ is preserved across
an interface.
B1t B2t
= (B.9)
µ1 µ2
79
However, since µ ∼
= µ0 at optical frequencies for all materials [Reitz et al. 1993, p. 415], the
simplification
B1t = B2t (B.10)
is used instead of (B.9).

The boundary conditions for the normal component can likewise be derived as
ε1 E1n = ε2 E2n (B.11)

B1n = B2n (B.12)
Direction of Propagation
The plane of incidence is defined to be the xz-plane, thus
k = (kx , 0, kz ) (B.13a)
The components of the wavevector of the transmitted wave k and the wavevector of the
reflected wave k are given as
k = (kx , ky , kz ) (B.13b)

k = (kx , ky , kz ) (B.13c)
It is possible to express (B.8) and (B.10) by the use of the solutions to the wave equations
which states
E = E0 e j(ωt−k·r+φ) (B.14a)
B = B0 e j(ωt−k·r+φ)
(B.14b)
as described in (A.26). However, as the frequency is the same for both the incident, the
transmitted and the reflected wave, the term involving ω can be neglected. The same holds
for φ. Thus by using (B.14), with the notations given by (B.13), it is possible to express (B.8)
and (B.10) by the following

Et e− jxkx + Et e− j(xkx +yky ) = Et e− j(xkx +yky ) (B.15a)
− jxkx
Bt e + Bt e− j(xkx +yky ) = Bt e− j(xkx +yky ) (B.15b)
From (B.15) it follows that
ky = ky = 0 (B.16)
and
kx = kx = kx (B.17)
It can thus be concluded that both the transmitted and the reflected wave lies in the plane
of incidence, i.e. the xz-plane.
80
Orientation of E and B for σ-polarized Light
From (B.8) it is clear that
Ey + Ey = Ey (B.18)
and
Ex + Ex = Ex (B.19)
For σ-polarized light, the incident E-field is defined to be strictly perpendicular to the
plane of incidence, thus E = Ey . Since the incident light has no x- or z-component, so will
neither the reflected nor the transmitted. Thus E, E and E will all be directed along the
y-axis with the following amplitude relation according to (B.18)
E + E = E (B.20)
As the E-fields are strictly tangential to the interface, it can be concluded due to (B.1) that
B, B and B lies in the xz-plane. Thus, the x-component of B, B and B is the only tangential
component of the respective magnetic fields with respect to the interface. Hence from (B.10)
it can be concluded that
Bx + Bx = Bx (B.21)
Reflection and Transmission Coefficient for σ-polarized Light

The purpose of this section is to derive the reflected and transmitted part of σ-polarized
light, ρσ and τσ respectively. The situation where σ-polarized light is incident on a surface is
depicted in Figure B.3. The following trigonometrical coherences can be deduced from the
figure
Bx = −B cos(θ) (B.22a)
Bx
= −B cos(θ )
(B.22b)
Bx
= B cos(θ) (B.22c)
By inserting (B.22) into (B.21), it can be seen that
cos(θ)(B − B ) = cos(θ )B (B.23)
When applying (B.3) on (B.23) it yields
ñ cos(θ)(E − E ) = ñ cos(θ )E (B.24)
By substituting E in (B.24) with the expression given by (B.20) and isolating the term E /E,
an expression for the reflected component ρσ is derived
E ñ cos(θ) − ñ cos(θ )

ρσ = (B.25)
E ñ cos(θ) + ñ cos(θ )
81
k’’ B’’
E’’ k’
E’
B’
θ’’
θ’
ñ y ñ’ z
θ
k
E
B
Figure B.3: Depiction of σ-polarized light. [Klein & Furtak 1986, p. 75]
By using (B.20) in order to eliminate E in (B.24), an expression for the transmitted part of
the E-field can be obtained
E 2ñ cos(θ)
τσ = (B.26)
E ñ cos(θ) + ñ cos(θ )
It is, however, in some cases disadvantageous to have θ included in the expression for ρσ
and τσ , as θ can be rather difficult to measure. Consequently, cos(θ ) is replaced by another
term derived by Snell’s law. Snell’s law states that
ñ sin(θ) = ñ sin(θ ) (B.27)
When modifying (B.27) by means of simple geometrical manipulations, an expression for
cos(θ ) can be deduced
2
ñ
cos(θ ) = 1 − sin2 (θ) (B.28)
ñ
By inserting this expression for cos(θ ) into (B.25) and (B.26) respectively, ρσ and τσ can be
defined merely by ñ, ñ and θ
2
ñ cos(θ) − ñ 1 − ññ sin2 (θ)
ρσ = 2 (B.29a)
ñ cos(θ) + ñ 1 − ññ sin2 (θ)
2ñ cos(θ)
τσ = 2 (B.29b)

ñ cos(θ) + ñ 1 − ññ sin2 (θ)
82
Reflection and Transmission Coefficient for π-polarized Light
Similar calculations can be done for π-polarized light in order to obtain expressions for ρπ
and τπ . This leads to the following expressions
ñ cos(θ) − ñ cos(θ )

ρπ = (B.30a)
ñ cos(θ) + ñ cos(θ )
2ñ cos(θ)
τπ = (B.30b)
ñ cos(θ) + ñ cos(θ )
where again cos(θ ) can be replaced by the expression derived in (B.28).
83
Three Phase Optical System C
This appendix concerns reflection and transmission in an optical system of three phases consisting of
medium 0, medium 1 and medium 2. The objective is to relate the electrical field of an incident wave
to the electrical field of a resultant reflected and a resultant transmitted wave. The optical system is
illustrated in Figure C.1
E θ0 Er,1 Er,2 Er,3

Medium (0)
d θ1
Medium (1)
Medium (2)
Et,1 Et,2
θ2
Figure C.1: Illustration of a three phase optical system where an incident

wave E is being reflected and transmitted due to the boundaries between
the three media. [Azzam & Bashara 1977, p. 283]
In order to simplify the deduction of the relation between the incident wave and the
resultant reflected and transmitted wave some assumptions are made:
1. The system is considered to have parallel boundaries between each phase.
2. Medium 1 has lateral dimensions much larger than its thickness d, whereas medium 0
and 2 are of infinite thickness.
3. All three media are considered homogenous and optically isotropic.
4. Medium 1 is not an amplifying medium.
5. Monochromatic plane incident wave.

Each of these assumptions is elaborated in the next sections where the assumptions have
direct influence.
Reflection and Transmission of an Incident Wave

The three phase system depicted in Figure C.1 shows an incident wave propagating through
medium 0, until it makes contact with medium 1 at an angle of θ0 . The incident wave is
85
APPENDIX C. THREE PHASE OPTICAL SYSTEM
partially reflected at the boundary and partially transmitted through medium 1 at an angle
of θ1 . The transmitted part of the incident wave propagates through medium 1 until it makes
contact with the boundary between medium 1 and medium 2. At this boundary the wave is
again partially reflected and partially transmitted through medium 2 at an angle of θ2 . The
connection between the angles θ0 , θ1 and θ2 are given by Snell’s law
ñ0 sin(θ0 ) = ñ1 sin(θ1 ) = ñ2 sin(θ2 ) (C.1)
where ñ0 , ñ1 and ñ2 are the complex refractive indexes for medium 0, 1 and 2 respectively.
In the following, calculations and considerations are done without regard to the polar-
ization of the incident wave. It is assumed that the incident wave is either σ or π polarized.
The subject polarization is treated in the last section of this appendix.
Reflection
By considering Figure C.1 and using the method of calculating reflection and transmission
of light waves by use of Fresnel reflection and transmission coefficients (ρ, τ), as described
in Appendix B, an expression for the reflected wave can be deducted as
Er,1 = ρ01 · E (C.2a)

− jβ − jβ
Er,2 = τ01 e ρ12 e τ10 · E (C.2b)
Er,3 = τ01 e− jβ ρ12 e− jβ ρ10 e− jβ ρ12 e− jβ τ10 · E (C.2c)
where β denotes the phase change due to propagation of the light wave through medium 1.
Deduction of β
The cause of this phase change is illustrated in Figure C.2. The reflected waves Er,1 and Er,2
can be considered parts of a total reflected plane wave, hence the phases of the two reflected
waves must be compared in the points E and C. Wave Er,1 has a phase change due to the
E
Er,1
Ein θ0 D Er,2
z0
Medium (0) A
C
d z1
θ1
Medium (1)
y B
Medium (2)
Figure C.2: Illustration of the reflection of an incident monochromatic

plane wave Ein . The reflected waves Er,1 and Er,2 constitute a plane by
connecting the points E and C as illustrated in the figure.
86
propagation from A to E, whereas wave Er,2 has a phase change due to the propagation from
A to C. From Figure C.2 it can be seen that
|AE|
|AD| = (C.3)
2
|ABC|
|AB| = (C.4)
2
and thus the phase difference is considered in points D and B instead, as this will simplify
the calculations, while still yielding the same result.
As described in Appendix A a plane wave is described by the equation
E(z,t) = E0 cos (ω · t − k · z + ϕ) (C.5)
where k = 2π λ ñ, λ being the free-space wavelength of the incident wave. The phase of this
wave is denoted Φ = −k · z + ϕ.
The phase change β can be expressed as the difference in phase of the two waves Er,1 and
Er,2 in the points D and B respectively
β = ΦD − ΦB (C.6)
= k 1 · z 1 − k0 · z 0 (C.7)
2π
= (ñ1 · z1 − ñ0 · z0 ) (C.8)
λ
where z1 = |AB| and z0 = |AD|. By considering Figure C.2 it can be seen that
d
z1 = (C.9)
cos(θ1 )
d · sin(θ1 )
y = sin(θ1 ) · z1 = (C.10)
cos(θ1 )
d · sin(θ1 ) · sin(θ0 )
z0 = sin(θ0 ) · y = (C.11)
cos(θ1 )
Furthermore it is known by Snell’s law that
ñ1 · sin(θ1 )
ñ0 = (C.12)
sin(θ0 )
By making substitutions for z1 , z0 and ñ0 due to (C.9), (C.11) and (C.12) into (C.8) a
reduced expression for the phase change can be obtained

2π d ñ1 · sin(θ1 ) d · sin(θ0 ) · sin(θ1 )
β= ñ1 · − · (C.13)
λ cos(θ1 ) sin(θ0 ) cos(θ1 )

2π · d · ñ1 1 − sin2 (θ1 )
= (C.14)
λ cos(θ1 )
d
= 2π ñ1 cos(θ1 ) (C.15)
λ

d
= 2π ñ21 − ñ20 sin2 (θ0 ) (C.16)
λ
If medium 1 is absorbing β is complex i.e. β = β + jβ [Röseler 1990, p. 31].
87
Total Reflected Wave

Due to the second assumption made in the beginning of this appendix1 and (C.2) the total
reflected wave Er can be calculated by adding all the partial reflected waves given in (C.2)

Er = ρ01 + τ01 τ10 ρ12 e− j2β + τ01 τ10 ρ10 ρ212 e− j4β + τ01 τ10 ρ210 ρ312 e− j6β + ... · E (C.17)
P = ρ01 + τ01 τ10 ρ12 e− j2β + τ01 τ10 ρ10 ρ212 e− j4β + τ01 τ10 ρ210 ρ312 e− j6β + ... (C.18)
where P (uppercase ρ) denotes the total reflection ratio.
Transmission
The deduction of the total transmission ratio T (uppercase τ) is analogous to the deduction
of the total reflection ratio and hence will not be described. T is given by
T = τ01 τ12 e− jβ + τ01 τ12 ρ10 ρ12 e− j3β + τ01 τ12 ρ210 ρ212 e− j5β + ... (C.19)
Rewriting of Infinite Geometric Series

The two series describing P and T can be rewritten in order to achieve expressions that can
be used in the further analysis of the reflection and transmission of a light wave due to a
three phase optical system. As the principle behind the rewriting is the same, only the total
reflected amplitude will be treated. First some manipulations are performed
P = ρ01 + τ01 τ10 ρ12 e− j2β + τ01 τ10 ρ10 ρ212 e− j4β + τ01 τ10 ρ210 ρ312 e− j6β + ... (C.20)
− j2β
= ρ01 + τ01 τ10 ρ12 e · Pn (C.21)
where Pn is an infinite geometric series expressed by

n k n
Pn = ∑ ρ10 ρ12 e− j2β = 1 + ρ10 ρ12 e− j2β + ρ210 ρ212 e− j4β + ... + ρ10 ρ12 e− j2β (C.22)
k=0
n
= ∑ qk = 1 + q + q2 + ... + qn (C.23)
k=0
with q = ρ10 ρ12 e− j2β . The task of rewriting the total reflected amplitude now consists of
simplifying (C.23). Multiplication by q yields
qPn = q + q2 + q3 + ... + qn+1 (C.24)
Subtracting (C.24) from (C.23) gives
(1 − q)Pn = 1 − qn+1
1 − qn+1
Pn = (C.25)
1−q
1 Due
to the much larger lateral dimensions of medium 1, compared to the thickness, an almost infinite
number of partial waves will be reflected
88
Due to the fourth assumption in the beginning of this appendix |ρ10 ρ12 e− j2β | < 1 and thus
|q| < 1 which entails that
1 − q∞+1 1
P ≡ P∞ = → for n → ∞ (C.26)
1−q 1−q
[Kreyszig 1999, p. 736], [Eric W. Weisstein, Geometric Series 2004]
By substituting this limit value for Pn into (C.21) a simplified expression for P is accom-
plished
τ01 τ10 ρ12 e− j2β

P = ρ01 +
1 − ρ10 ρ12 e− j2β
ρ01 − ρ01 ρ10 ρ12 e− j2β + τ01 τ10 ρ12 e− j2β
=
1 − ρ10 ρ12 e− j2β
ρ01 + ρ12 e− j2β
= (C.27)
1 + ρ01 ρ12 e− j2β
where the relations ρ10 = −ρ01 and τ01 τ10 = 1 − ρ01 2 have been used [Klein & Furtak 1986, p.
80].
Similar rewriting can be done of the total transmission ratio T given by (C.19) which
yields
τ01 τ12 e− jβ
T= (C.28)
1 + ρ01 ρ12 e− j2β
The results in (C.27) and C.28 are the total complex reflection and transmission coeffi-
cients of either σ or π polarization.
σ and π Polarized Incident Wave

The information of polarization of the wave is readily restored by adding a subscript to the
deducted expressions for P and T
ρ01,σ + ρ12,σ e− j2β

Pσ = (C.29a)
1 + ρ01,σ ρ12,σ e− j2β
ρ01,π + ρ12,π e− j2β
Pπ = (C.29b)
1 + ρ01,π ρ12,π e− j2β
τ01,σ τ12,σ e− jβ
Tσ = (C.29c)
1 + ρ01,σ ρ12,σ e− j2β
τ01,π τ12,π e− jβ
Tπ = (C.29d)
1 + ρ01,π ρ12,π e− j2β
With these four equations it is possible to calculate the reflection and transmission of
any kind of polarized incident light wave. The incident light wave can be divided into a
component perpendicular to the plane of incidence (σ component) and a component paral-
lel to the plane of incidence (π component) and the reflection and transmission of each of
89
these components can then be treated separately. The β value is not different for the σ and π
components due to the third assumption in the beginning of the appendix. However the re-
flection and transmission coefficients (ρ, τ) are. This is treated in Appendix B. The reflection
coefficients are listed in (C.30)
ñ0 cos(θ0 ) − ñ1 cos(θ1 )

ρ01,σ = (C.30a)
ñ0 cos(θ0 ) + ñ1 cos(θ1 )
ñ1 cos(θ1 ) − ñ2 cos(θ2 )
ρ12,σ = (C.30b)
ñ1 cos(θ1 ) + ñ2 cos(θ2 )
ñ1 cos(θ0 ) − ñ0 cos(θ1 )
ρ01,π = (C.30c)
ñ1 cos(θ0 ) + ñ0 cos(θ1 )
ñ2 cos(θ1 ) − ñ1 cos(θ2 )
ρ12,π = (C.30d)
ñ2 cos(θ1 ) + ñ1 cos(θ2 )
and the transmission coefficients are listed in (C.31)
2ñ0 cos(θ0 )
τ01,σ = (C.31a)
ñ0 cos(θ0 ) + ñ1 cos(θ1 )
2ñ1 cos(θ1 )
τ12,σ = (C.31b)
ñ1 cos(θ1 ) + ñ2 cos(θ2 )
2ñ0 cos(θ0 )
τ01,π = (C.31c)
ñ1 cos(θ0 ) + ñ0 cos(θ1 )
2ñ1 cos(θ1 )
τ12,π = (C.31d)
ñ2 cos(θ1 ) + ñ1 cos(θ2 )
[Azzam & Bashara 1977, pp. 283-286]
90
The Jones Formalism D
This appendix describes the Jones Formulation including the Jones vector of a monochromatic plane
light wave and the Jones matrix of an optical device. Furthermore the Jones matrix of a cascade
coupled optical device and Jones matrixes of different optical devices are described.
The Jones Vector

In order to describe the polarization of a monochromatic plane light wave the entire solution
to the wave equation, given in (A.26a) on page 75, is not necessary. As a substitute the
Jones vector, which is a more concise mathematical description of the wave, is introduced.
The Jones vector is a vector which only contains information about the amplitude and the
phase of the field, hence it contains information about the polarization state of the field.
The derivation of the Jones vector starts with the equation of the monochromatic plane light
wave propagating along the z-axis which is deduced in (A.21) and repeated here
z
E(z,t) = E0 cos ω t − +φ (D.1)
c
or written in components
z z
E(z,t) = Ax cos ω t − + φx x̂ + Ay cos ω t − + φy ŷ (D.2)
c c
where Ax and Ay are the amplitudes of the harmonic oscillation of the electric field along the
x-axis and the y-axis respectively. φx and φy are the phases of the components. The first step
towards the Jones vector is to set up the components in a vector

Ex (z,t) Ax cos
ω t − cz + φx
E(z,t) = = (D.3)
Ey (z,t) Ay cos ω t − cz + φy
Rewriting (D.3) by means of the complex exponential function and knowing that the physi-
cal field is the only real part of the complex E-field yields
jφx

jωt − j ωz A x e
E(z,t) = e e c (D.4)
Ay e jφy
Regarding the polarization state of the wave the temporal information has no influence and
can therefore be suppressed. This yields
jφx

− j ωz A x e
E(z) = e c (D.5)
Ay e jφy
91
APPENDIX D. THE JONES FORMALISM
Furthermore the spatial information is of no interest when only the polarization is consid-
ered and can therefore also be suppressed which yields

Ax e jφx
E= (D.6)
Ay e jφy
The vector in (D.6) is known as the Jones vector of the monochromatic plane light wave. The
Jones vector can also be written as

E
E= x (D.7)
Ey
where Ex = Ax e jφx and Ey = Ay e jφy . As can bee seen the Jones vector contains complete infor-
mation about the amplitudes and the phases of the components of the electrical field, hence
it is a vector that describes the polarization of the light. [Azzam & Bashara 1977, pp. 13-15]
The Jones Matrix

In order to describe the modification of the polarization of a monochromatic plane light
wave by an optical device, a Jones matrix of the optical system is introduced. The Jones ma-
trix of an optical system relates the polarization of an incident monochromatic plane light
wave to the polarization of an outgoing wave. That is, the Jones matrix relates the Jones
vector of the incident wave to the Jones vector of the outgoing wave. A schematic drawing
of this can be seen in Figure D.1. It is illustrated that the polarization of the incident wave,
y y‘
x x‘
z z‘
Optical device
(T)
Ei Eo
Figure D.1: The polarization of a wave is changed from Ei to Eo by an opti-

cal device. The two coordinate systems (x, y, z) and (x , y , z ) are associated
with the incident and the outgoing plane waves respectively. The direc-
tions z and z are parallel to the two wavevectors k and k , which need not
to be parallel.
described by its Jones vector Ei , is changed due to interaction with the optical device. The
polarization of the outgoing wave is described by the Jones vector Eo . The two coordinate
systems (x, y, z) and (x , y , z ) are associated with the incident and the outgoing waves respec-
tively. The directions z and z are parallel to the two wavevectors k and k of the incident
and outgoing waves respectively. z and z need not to be parallel. It is seen that change in
polarization of the incident light wave due to the interaction with the optical device corre-
sponds to a change of coordinate system from (x, y, z) to (x , y , z ) of the incident Jones vector.
92
That is a change of basis of the incident Jones vector. A change of basis is, in the absence of
non-linearities, given as
Eo = T Ei (D.8)
where T is the change of coordinate matrix from (x, y, z) to (x , y , z ). T is also called the Jones
matrix of the optical system. (D.8) describes the law of interaction of the incident wave and
the optical system as a matrix transformation of the incident Jones vector. Writing (D.8) i
terms of the components of the Jones vectors and the Jones matrix yields

Eox T11 T12 Eix
= (D.9)
Eoy T21 T22 Eiy
or

Eox T11 Eix + T12 Eiy
= (D.10a)
Eoy T21 Eix + T22 Eiy
It can be seen from (D.10) that the components of the Jones vector of the outgoing wave are
given as a linear combination of the components of the Jones vector of the incident wave.
If a cascade of N optical systems is placed in the path of the incident wave, the polariza-
tion of the outgoing wave is given as
Eo = TN TN−1 · · · T1 Ei (D.11)
or
Eo = TC Ei (D.12)
where
TC = TN TN−1 · · · T1 (D.13)
(D.12) shows that a cascade of N optical system can be replaced by one system with the
overall Jones matrix given as TC . Note that the order of the Jones matrixes in (D.11) and
(D.13) is of importance.
Jones Matrices of Basic Optical Devices

In this section the Jones matrixes of some basic optical devices will be presented. The
retarder, the polarizer, the rotator and the reflector. The first three mentioned are of the
transmission-type where the polarization of the wave is changed as the wave is propagated
through the media. The reflector is a reflection-type medium by which the wave is modified
by reflection of the surface of the medium. In the subsequent it is assumed that the light
wave propagates in the direction of the z-axis.
93
APPENDIX D. THE JONES FORMALISM
Retarder
The Jones matrix describing the effects on a plane wave propagating through an isotropic
optical device with real refractive index n over a distance d is given as
− j2πnd/λ
e 0
T= (D.14)
0 e− j2πnd/λ
It can be seen from (D.14) that the plane wave is retarded by 2πnd/λ when the wave has
travelled the distance d in the device. The device is called an isotropic retarder. Now in-
troducing a medium which is not isotropic but instead uniaxially1 linearly birefringent. If
the light wave is travelling perpendicular to the optic axis and if the light wave is linearly
polarized in a direction parallel to the optic axis then the light wave will be subject to a re-
fractive index ne . If the light wave is linearly polarized in a direction orthogonal to the optic
axis the wave will be subject to another refractive index, namely no . As the light wave is
travelling in the z-direction and the optical axes is in the x-direction, the refractive index in
the x-direction is given as ne and the refractive index in the y-direction is given as no . That
is, the Jones matrix of the medium of thickness d is given as
− j2πn d/λ
e e 0
T= (D.15)
0 e− j2πno d/λ
A medium described by the Jones matrix of in (D.15) is a linear retarder. The x-components
and y-components of the Jones vector will be retarded differently by a linear retarder.
Polarizer
If a medium is uniaxially linearly dichroic 2 , a wave travelling parallel to the optic axis
will be attenuated different then a wave travelling orthogonal to the optic axis. The wave
polarized parallel to the optic axis will be attenuated with attenuation coefficient Ke and the
wave polarized orthogonal to the optic axis will be attenuated with attenuation coefficient
Ko . The Jones matrix of a uniaxially linearly dichroic medium is given as
−2πK d/λ
− j2πnd/λ e e 0
T=e (D.16)
0 e−2πKo d/λ
A medium of thickness d described by the Jones matrix of (D.16) is called a linear partial
polarizer because unpolarized light will be partially polarized after interaction with the
medium. A special case of the linear partial polarizer is the ideal linear polarizer were
Ke = 0 and Ko = ∞ which is described by the Jones matrix

− j2πnd/λ 1 0
T=e (D.17)
0 0
1 Thereis only one optic axis in the medium. An optic axis is a line in a doubly refracting medium that is
parallel to the direction in which all components of plane-polarized light travel with the same speed. [Merriam-
Webster Online 2004]
2 Dichroism is the property of absorbing one of two plane-polarized components of light more that the other.
[Merriam-Webster Online 2004]
94
Rotator
A medium in which linearly polarized light is still linearly polarized light, but the direc-
tion of the variation of the electric field changes as the light wave propagates through the
medium is called an optical rotator. The Jones matrix of the optical rotator is given as

− j2πnd/λ cos αd − sin αd
R=e (D.18)
sin αd cos αd
Where α is the rotation of the polarization per part length and d is the thickness of the
material. With the definition of the sign as in (D.18) the rotation of the polarization by the
medium will be clockwise if α is positive and counter clockwise if α is negative. If the light
is elliptically polarized the major axis of the ellipse is rotated by the optical rotator but the
ratio between the major and the minor axes of the ellipse is not changed.
Reflector
If the incident Jones vector is described in the π, σ coordinate system the Jones matrix of a
reflector is given as

ρ 0
T= π (D.19)
0 ρσ
Where ρπ and ρσ are the Fresnel reflection coefficients discussed in Appendix B. [Azzam &
Bashara 1977, pp. 61-79]
95
Correlation between the
Stokes Parameters and ψ and Δ E
This appendix concerns the relation between the three Stokes parameters s0 , s1 , s2 and the ellipsomet-
ric parameters ψ and Δ. This relation is deducted by manipulating the expression for the measured
light intensity at the detector of the ellipsometer Io .
The light intensity measured at the detector is given by (4.18) on page 27, which can be
rewritten by inserting expressions for Eπ , Eπ∗ , Eσ and Eσ∗
Eπ = ρπ cos(α1 )Ei (E.1a)

Eπ∗ = ρ∗π cos(α1 )Ei∗ (E.1b)
Eσ = ρσ sin(α1 )Ei (E.1c)
Eσ∗ = ρ∗σ sin(α1 )Ei∗ (E.1d)
Io thus becomes
Io = cos2 (α2 )Eπ Eπ∗ + sin2 (α2 )Eσ Eσ∗ + cos(α2 ) sin(α2 ) (Eπ Eσ∗ + Eσ Eπ∗ ) (E.2)
= ρπ ρ∗π cos2 (α1 ) cos2 (α2 )Ei Ei∗ + ρσ ρ∗σ sin2 (α1 ) sin2 (α2 )Ei Ei∗
+ (ρπ ρ∗σ Ei Ei∗ + ρσ ρ∗π Ei Ei∗ ) sin(α1 ) cos(α1 ) sin(α2 ) cos(α2 ) (E.3)
1
= Ii ρπ ρ∗π cos2 (α1 ) cos2 (α2 ) + ρσ ρ∗σ sin2 (α1 ) sin2 (α2 )
2
+ (ρπ ρ∗σ + ρσ ρ∗π ) sin(α1 ) cos(α1 ) sin(α2 ) cos(α2 ) (E.4)
where Ei Ei∗ = 12 Ii , Ii being the intensity of the light at the source. From this expression it can
be seen that the two polarizers are equivalent. The equations are thus the same for an RAE
and an RPE.
(E.4) can be rewritten by utilizing the trigonometric relations [Råde & Westergren 1998,
p. 124]
1
cos(α2 ) sin(α2 ) = sin(2α2 ) (E.5a)
2
1 − cos(2α2 )
sin2 (α2 ) = (E.5b)
2
1 + cos(2α 2)
cos2 (α2 ) = (E.5c)
2
97
APPENDIX E. CORRELATION BETWEEN THE STOKES PARAMETERS AND ψ AND Δ
The intensity thus becomes

1
Io = Ii ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 ) + cos(2α2 )(ρπ ρ∗π cos2 (α1 ) − ρσ ρ∗σ sin2 (α1 ))
4
+ (ρπ ρ∗σ + ρσ ρ∗π ) sin(α1 ) cos(α1 ) sin(2α2 ) (E.6)

1 ρπ ρ∗π cos2 (α1 ) − ρσ ρ∗σ sin2 (α1 )
= Ii (ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 )) 1 + cos(2α2 )
4 ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 )

(ρπ ρ∗σ + ρσ ρ∗π ) sin(α1 ) cos(α1 )
+ sin(2α2 ) (E.7)
ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 )
ρπ
The two fractions in this expression can be rewritten by using the relation tan(ψ)e jΔ = ρσ .
The first fraction becomes
ρπ ρ∗π cos2 (α1 ) ρ ρ∗ sin2 (α ) tan2 (ψ)
ρπ ρ∗π cos2 (α1 ) − ρσ ρ∗σ sin2 (α1 ) ρσ ρ∗σ sin2 (α1 )
− ρσ ρσ∗ sin2 (α1 ) tan2 (α1 )
−1
σ σ
= 1
= tan2 (ψ)
(E.8)
ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 ) ρπ ρ∗π cos2 (α1 )
+
ρσ ρ∗σ sin2 (α1 )
+1
ρσ ρ∗σ sin2 (α1 ) ρσ ρ∗σ sin2 (α1 ) tan2 (α1 )
Introduction of a new parameter ψ that satisfies the equation tan(ψ ) = tan(ψ)

tan(α1 ) reduces (E.8)
to
1 − tan2 (ψ ) cos2 (ψ ) + sin2 (ψ ) − tan2 (ψ )
− = − (E.9)
1 + tan2 (ψ ) 1 + tan2 (ψ )
cos2 (ψ ) − (1 − cos2 (ψ )) tan2 (ψ )
=− (E.10)
1 + tan2 (ψ )
cos2 (ψ ) − sin2 (ψ ) tan2 (ψ )
=− (E.11)
1 + tan2 (ψ )
(cos2 (ψ ) − sin2 (ψ ))(1 + tan2 (ψ ))
=− (E.12)
1 + tan2 (ψ )
= −(cos2 (ψ ) − sin2 (ψ )) (E.13)

= − cos(2ψ ) (E.14)
Correspondingly the second fraction in (E.7) can be rewritten
− jΔ

tan(α1 ) tan(ψ)e + tan(ψ)e
1 jΔ
(ρπ ρ∗σ + ρσ ρ∗π ) sin(α1 ) cos(α1 ) 2 tan(ψ )
= = cos(Δ) (E.15)
ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin (α1 )
2 tan (ψ)
2
+1 1 + tan2 (ψ )
2tan (α)
This can again be rewritten

2 tan(ψ ) 2 tan(ψ )(cos2 (ψ ) + sin2 (ψ ))
cos(Δ) = cos(Δ) (E.16)
1 + tan2 (ψ ) 1 + tan2 (ψ )
2 sin(ψ ) cos(ψ ) + 2 tan(ψ ) sin2 (ψ )
= cos(Δ) (E.17)
1 + tan2 (ψ )
2 sin(ψ ) cos(ψ )(1 + tan2 (ψ ))
= cos(Δ) (E.18)
1 + tan2 (ψ )
= cos(Δ)2 sin(ψ ) cos(ψ ) (E.19)

= sin(2ψ ) cos(Δ) (E.20)
98
By substituting (E.14) and (E.20) into (E.7) the intensity at the detector is given as
1
Io = Ii (ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 )) 1 − cos(2α2 ) cos(2ψ )
4
+ sin(2α2 ) sin(2ψ ) cos(Δ) (E.21)
By realizing that
1
Ii (ρπ ρ∗π cos2 (α1 ) + ρσ ρ∗σ sin2 (α1 )) = Eπ Eπ∗ + Eσ Eσ∗ = s0 (E.22)
2
(E.21) can be expressed as
1

s0 1 − cos(2ψ ) cos(2α2 ) + sin(2ψ ) cos(Δ) sin(2α2 ) (E.23)
2
This equation can be compared to another expression for Io namely (4.22) on page 27
1
Io = [s0 + s1 cos(2α2 ) + s2 sin(2α2 )] (E.24)
2
1 s1 s2
= s0 1 + cos(2α2 ) + sin(2α2 ) (E.25)
2 s0 s0
By doing so two important relations between the ellipsometric parameters and the Stokes
parameters arises
s1
cos(2ψ ) = − (E.26a)
s0
s2
sin(2ψ ) cos(Δ) = (E.26b)
s0
Remembering that tan(ψ ) = tan(α

tan(ψ)
1)
these equations make it possible to calculate the ellipso-
metric parameters ψ and Δ if the three Stokes parameters s0 , s1 and s2 are known.
99
Index of Refraction F
This appendix will explain the mathematical foundations of the refractive index and give a simple
understanding of the interpretation of it. The concept of permittivity is introduced by means of the
Fourier transform of the electric field E and the electric displacement D. Then, the Drude-Lorentz
theory will be used to model the microscopic nature of the material and thereby give an explanation
of the complex index of refraction.
Defining the Index of Refraction

The term ”refraction” deals with the deflection of electromagnetic waves e.g. light from a
straight path when passing obliquely from one medium to another, see Figure F.1. The index
of refraction is a means of describing an optical property of a medium. If the medium is fully
transparent, the index of refraction is purely real and denoted n. If the medium is absorbing
light, the index of refraction becomes complex and is denoted ñ.
z-axis
θ0
Air
ñ
θ1
Figure F.1: Refraction at an oblique angle of incidence into a medium with

a complex index of refraction ñ.
The mathematical foundation is given by the Maxwell equations. By using (A.10) in

linear, isotropic matter instead of in free-space, we get the wave-equation for an electrical
field
∂2 E
∇2 E − µε =0 (F.1)
∂t 2
Here µ and ε are complex quantities and not as in the free-space equations, where they are
named the free-space permeability µ0 and permittivity ε0 and are both real numbers. The
solution for (F.1) is given by (A.26a) and rewritten here
E(r,t) = E0 e j(ωt−k·r+ϕ) (F.2)
101
APPENDIX F. INDEX OF REFRACTION
Substituting (F.2) into (F.1) yields

k2 = µεω2 (F.3)
where k is the wave-number and ω is the angular frequency of the light. We then define the
complex index of refraction
µε
ñ2 (n − jK )2 µεc2 = (F.4)
µ0 ε0
such that the dispersion relation for the medium is given by [Reitz et al. 1993, p. 426]
ω
k = ñ (F.5)
c
where
1
c= √ (F.6)
µ0 ε0
Notice that the imaginary part K used in (F.4) is not directly related to the wave-number k
used in the dispersion relation in (F.5). The complex index of refraction is sometimes also
written as
ñ (n − jK ) = n(1 − jκ) (F.7)
where κ (kappa) is called the absorption index [Azzam & Bashara 1977, p. 270]. To show the
effect of the two real optical constants n and K , the dispersion relation (F.5) is inserted into
(F.2), assuming the direction of propagation is along the z-coordinate axis yields
E(z,t) =E0 e j(ωt+ϕ) e− jkz (F.8)
j(ωt+ϕ) − jñ ωc z
=E0 e e (F.9)
ω
j(ωt−n ωc z+ϕ)
=E0 e−K cz e (F.10)
ω
As can be seen from (F.10), the first term E0 e−K c z is a function of z and affects only the
amplitude of the wave. The wave is attenuated in the medium with this factor as the wave
moves along the z-axis. E0 is constant and the term K ωc is called the attenuation constant.
The reciprocal δ = Kcω is called the skin depth and is a measurement of the distance to which
the field is at 1e of the original value.
The velocity in a medium is found by evaluating the solution to the wave-equation (F.10).
When measuring velocity, the total phase can be taken as constant.
ω
ωt − n z + ϕ =constant (F.11)
c
This just means that the velocity is defined as following a certain value through the wave
propagation. The velocity v can then be found by isolating z and taking the derivative with
respect to time
dz c
v= = (F.12)
dt n
So far, the complex index of refraction has been defined as a constant. This is a crude sim-
plification, as it is very frequency-dependent. This dependency relates to the permittivity.
[Klein & Furtak 1986, p. 68-70]
102
Permittivity as a frequency-dependent concept
When dealing with permittivity, the Fourier transform can be used to understand the de-
pendency on frequency in the medium. The Fourier transform of a function (a signal, field,
excitation etc.) is defined as
1 ∞
XF (ω) = √ X(t)e jωt dt (F.13)
2π −∞
where X(t) is the time-dependent signal, XF (ω) is the Fourier transform of this signal, ω
√
is the angular frequency, t is the time and 1/ 2π is introduced to restore symmetry of the
transform. The inverse Fourier transform is likewise defined as
1 ∞
X(t) = √ XF (ω)e− jωt dω (F.14)
2π −∞
One of the fundamental preconditions for the media that are under investigation is that
the medium is linear. When this holds, the medium can be considered to be a system with an
input, response and output function to which the convolution theorem can be applied. This
also implies that the system is causal, but all real physically systems are causal. Causality is
that the cause comes before the effect. In a physical system this implies that the output only
appears after the input.
The convolution theorem relates the input function X(t), the response function R(t) and
the output function Y (t) as
1 ∞
Y (t) = √ R(t − τ)X(τ)dτ (F.15)
2π −∞
If the convolution theorem can be applied to the medium, the Fourier transform of the input,
response and output function are denoted XF (ω), RF (ω) and YF (ω), respectively, and are
related as
YF (ω) = RF (ω)XF (ω) (F.16)
This relation is a consequence of the Fourier transform. It shows that if the output function
can be found as the convolution of the input and response function in the time domain,
it can conversely be found as the multiplication of the Fourier transform of the individual
functions in the frequency domain. If the convolution is mathematically difficult to achieve,
the functions can be Fourier transformed individually, multiplied in the frequency domain
and then inverse Fourier transformed.
The consequences of (F.16) are important because it shows that if the input function is an
impressed field in the form of a monochromatic wave, the resulting field (the output func-
tion) is proportional to the impressed field, because RF (ω) is constant. The proportionality
factor is, however, frequency dependent if the full spectrum is considered.
Another consequence of the Fourier transform is that the relation between the input and
output function can be written as a linear, inhomogeneous differential equation of N’th order
with constant coefficients, namely
N
dn
∑ an dt n
Y (t) = bX(t) (F.17)
n=0
103
where an and b are constant coefficients and N is the order of the differential equation. Gen-
erally, the right-hand side of (F.17) can also be written as a sum of derivatives of X(t) each
with a coefficient bn , but this will not be used in the following. By use of the definition for
the inverse Fourier transformation, the response function can be written as
b
RF (ω) = (F.18)
∑Nn=0 an (− jω)n
which will be useful later when recognizing the response function.

The theory mentioned here can be used on real media, with some assumptions. When
impressing an electric field E on a linear time-invariant dielectric medium (LTI-system), the
electric displacement D is coupled to E via the convolution theorem from (F.15) as
1 ∞
D(r,t) = √ R(r, τ)E(r,t − τ)dτ (F.19)
2π 0
Due to causality, the integral needs only to be evaluated from 0 to ∞. The response function
does not depend on E and the relationship between D and E is assumed spatially local. This
means that D only depends on E at the same point in space. If R is assumed independent
of r, meaning that R only changes at an interface between two media, then (F.19) can be
reduced to
1 ∞
D(t) = √ R(τ)E(t − τ)dτ (F.20)
2π 0
which is identical with the convolution theorem in (F.15). It can hereby be seen that the
relation between the D- and E-field, called the permittivity ε, in the time domain is not just
dependant on the time t, but also of all past times because of τ. This can be interpreted as a
time delay, as the D-field at a given time does not only depend on the E-field at this time, but
on all previously values of the E-field. However, in the frequency domain the permittivity is
a proportionality factor dependant only on the frequency ω. [Reitz et al. 1993, pp. 484-489]
This yields the constitutive equation for a linear, isotropic dielectric
D(ω) = ε̃(ω)ε0 E(ω) (F.21)
where ε̃(ω) is the dimensionless complex dielectric coefficient. By using the definition of the
electric displacement given in [Reitz et al. 1993, p. 107]
D(ω) = ε0 E(ω) + P(ω) (F.22)
where P(ω) is the electric polarization of the dielectric material, another constitutive equa-
tion is given as
P(ω) = χ(ω)E(ω) (F.23)
where
χ(ω) = (ε̃(ω) − 1)ε0 (F.24)
104
is the electric susceptibility. The susceptibility can completely describe the electrical behav-
ior of the material, but often ε̃ is used instead. It relates the complex index of refraction as
[Reitz et al. 1993, p. 425]
ε̃(ω) = (ñ(ω))2 (F.25)
where
ε̃(ω) = εre (ω) + jεim (ω) (F.26)
giving

1
n(ω) = εre (ω) + (εre (ω)) + (εim (ω))
2 2 (F.27)
2

1
K (ω) = −εre (ω) + (εre (ω)) + (εim (ω))
2 2 (F.28)
2
This is the basis for introducing a model, in which the susceptibility relates the complex
index of refraction to some fundamental properties of the electrons in the material.
The Drude-Lorentz Model

This section relates the macroscopic optical constants with the microscopic electric-dipole
excitations in the material. This is done through the Drude-Lorentz harmonic oscillator
model, which was stated in the beginning of the 20th century and the electrons are therefore
seen as classical particles. The idea is to replace the atoms and molecules in a material
by a set of harmonically bound electrons, oscillating at some resonant frequency ω0 . The
term ”harmonically bound” indicates that the electrons are bound to an equilibrium by a
linear restoring force, as in a harmonic spring. The electrons are modeled as damped by a
damping device. Furthermore, a field called the molecular field Em (t) is impressed on the
electrons, yielding the driving force eEm (t). This yields a forced, damped harmonic oscillator
to model the motion of the electrons. Even though quantum mechanics is not considered
in this model, it still works well when dealing with conductivity-related terms, but fails to
model e.g. heat capacity.
The standard second-order linear differential equation representing the one-dimensional
motion of the forced, damped harmonic oscillator can be found by Newton’s second law
d2x
Fres = m (F.29)
dt 2
dx d2x
−Cx(t) − G + eEm (t) = m 2 (F.30)
dt dt
where x(t) is the displacement of the electron from an equilibrium, m is the mass of the
electron, −Cx(t) is the linear restoring force (spring force), −G dx
dt is the damper force and
eEm (t) is the driving force due to the impressed field Em (t). An illustration of this can be
seen in Figure F.2. By introducing a damping constant as γ = G m and the square of the natural
105
x(t)
C eEm(t)
-e G
m
Figure F.2: Illustration of a forced, damped harmonic oscillating electron.

C is the spring constant, G is the damping constant, Em (t) is the impressed
molecular field, e− and m are the electron charge and mass, respectively
and x(t) is the displacement of the electron from its equilibrium.
frequency of the undamped oscillator as ω20 = C

m, the fundamental differential equation of
this model appears as
d2x dx eEm (t)

2
+ γ + ω20 x = (F.31)
dt dt m
For free electrons, the restoring force is not present which yields ω0 = 0. If the displacement
of the electrons in the material is static, the derivatives of x(t) is 0 and the natural frequency
of the undamped oscillator can be found as1
e2
ω20 = (F.32)
4πε0 mR30
where R0 is the atomic radius. Two substitutions are needed to eliminate the mechanical
displacement x(t) and the molecular field Em (t). First, the polarization is calculated by the
number of charged particles N per unit volume times the dipole moment ex(t) due to the
displacement x(t) of the charge e
P(t) = Nex(t) (F.33)
An illustration of this can be seen in Figure F.3.

Second, the molecular field is calculated. It is produced by all external sources and all
polarized molecules in the material, with the exception of the one molecule under consider-
ation [Reitz et al. 1993, pp. 127-131]. In isotropic material this yields
ν
Em (t) = E(t) + P(t) (F.34)
ε0
where E(t) is the E-field of the wave and ν is a constant that effects the local field.
The two equations (F.33) and (F.34) are now used to rewrite (F.31) into
d2P dP
2
+γ + (ω20 − νω2p )P(t) = ε0 ω2p E(t) (F.35)
dt dt
where the plasma frequency ω p is given as
Ne2
ω2p = (F.36)
ε0 m
1 By using (5-12) in [Reitz et al. 1993, pp. 131-132] concerning induced dipoles
106
Em(t)
Charge
displacement
due to Em(t)
e- Standard
electron path
Nucleus
e-
+
x(t)
Figure F.3: An illustration of the dipole moment created due to the driving
field Em (t). The field causes the electron charge to be displaced and thereby
creating an electrical dipole.
As described in (F.16), this inhomogeneous second-order differential equation can be written

as
PF (ω) = RF (ω)EF (ω) (F.37)
where RF (ω) is the response function that represents the system and relates E to P. From
(F.18) the response function can be easily calculated and from (F.23) it is seen that this re-
sponse function can be regarded as the susceptibility
ε0 ω2p
χF (ω) = RF (ω) = (F.38)
ω20 − ω2 − jγω − νω2p
Rewriting this while including the definition of the electric susceptibility from (F.24) yields
ε̃(ω) − 1 ω2p
= (F.39)
1 + ν(ε̃(ω) − 1) ω20 − ω2 − jγω
which is the general relation between the dielectric coefficient of the material, and thereby
the index of refraction, and the microscopic properties of the material. From this relationship
it can also be seen that ε̃(ω) is complex and is a function of the frequency.
Some assumption to this result has to be made. Different particles in different distances
to their equilibrium have different restoring forces. This consideration leads to the problem
that they have different natural frequencies and damping. Restricting the term ”particles”
to electrons, thereby assuming no ions in the material, allow the mass and charge to be fixed.
To simplify the model, it is restricted to valence electrons, thereby making the distance to
equilibrium fixed.
Furthermore, ν is set to 0, which reduces (F.39) to
ω2p
ε̃(ω) − 1 = (F.40)
ω20 − ω2 − jγω
107
which is often used to model most metals. When plotting this, the real and imaginary part
is drawn independently. The real and imaginary parts of (F.40) are
ω2p (ω20 − ω2 )
εre (ω) = 1 + (F.41a)
(ω20 − ω2 )2 + (γω)2
ω2p γω
εim (ω) = (F.41b)
(ω20 − ω2 )2 + (γω)2
Knowing that the square of the complex refractive index ñ(ω) is the complex dielectric coef-
ficient ε̃(ω), yields that the real n and imaginary K part of the complex index of refraction
can be calculated as

ω2p
n(ω) = Re 1+ 2 (F.42a)
ω0 − ω2 − jγω

ω2p
K (ω) = Im 1+ 2 (F.42b)
ω0 − ω2 − jγω
This model can be simplified in metals, where the valence electrons are free to move around
and contribute to dc conduction. This is called the Drude free-electron model. That the va-
lence electrons are free to move means that these electrons are not affected by any restoring
force. Putting C = 0 in (F.30) causes ω0 = 0 in (F.40) which then can be calculated as
ω2p
ε̃(ω) = 1 − (F.43)
ω(ω + jγ)
where it is enough to estimate the plasma frequency ω p and the damping coefficient γ in
the simulation of the complex index of refraction. These parameters can be calculated from
material parameters given in various optical books. The plasma frequency is as already
stated in (F.36) given as
Ne2
ω2p = (F.44)
ε0 m
and the damping coefficient γ is given as
1
γ= (F.45)
τ
where τ is a time constant for the decay of a current with no driving field. This will be used
in the simulation of the metals.
108
Index of Refraction and Band
Structures of Crystals G
The purpose of this appendix is to explain the coherence between the index of refraction and the
energy band structure of a solid. First the origin of energy gabs in solids is explained by means of
the nearly free electron model and the Bragg diffraction condition. Afterwards the band structures of
silicon and copper are presented and related to the calculated refractive indexes of silicon and copper
in Appendix H.
Energy Bands in Solids

In order to describe the energy band structure of a solid the nearly free electron model is
used. In the nearly free electron model the electrons are described as waves propagating
through the crystal. The probability of finding an electron at a certain location can be calcu-
lated by means of the single-particle time independent Schrödinger equation

h̄2 2
− ∇ +V (r) ψ(r) = Eψ(r) (G.1)
2m
where h̄ is Planck’s constant divided with 2π, m is the mass of the electron and E is the
energy of the electron, V (r) is the potential energy, ∇ is the nabla operator, ψ(r) is a solution
to the wave equation and r is a position vector. The energy E of the electron is the sum of
the kinetic energy and potential energy
h̄2 2 p2 1
E =− ∇ +V (r) = +V (r) = mv2 +V (r) (G.2)
2m 2m 2
where v is the velocity of the electron and p is the linear momentum of the electron given as
p = − jh̄∇ (G.3)
To solve the Schrödinger equation we must impose some boundary conditions. The bulk
properties are not very sensitive to the boundary conditions. The bulk means everything
else than the last few atomic layers on the surface. The most convenient conditions will be
used which are the periodic boundary conditions. The solutions to the Schrödinger equation
that satisfy periodic boundary conditions are in the form of traveling plane waves
ψ(r) = e jk·r (G.4)
where the components of the wavevector k are given as

2πn
kx = ky = kz = (G.5)
L
109
APPENDIX G. INDEX OF REFRACTION AND BAND STRUCTURES OF CRYSTALS
where n is a positive or negative integer and L is the period, that is
ψ(x + L, y + L, z + L) = ψ(x, y, z) (G.6)
The probability density of an electron is given as
ψ∗ (r)ψ(r) = |ψ(r)|2 (G.7)
For a traveling wave the probability density is given as
e jk·r e− jk·r = 1 (G.8)
(G.8) shows that the charge density of a traveling electron is constant.

When a wave propagates in a crystal it is disturbed by Bragg reflection. At Bragg reflec-
tion, no wavelike solutions to the Schrödinger equation exist.
Bragg Diffraction Condition

In order to set up the condition of Bragg reflection a few important concepts of crystals must
be introduced. The lattice of a crystal is defined by means of three translation vectors a1 , a2
and a3 such that the set of points r defined as
r = u1 a1 + u2 a2 + u3 a3 (G.9)
for all integers u1 , u2 and u3 define the lattice. A translation in the lattice is defined by means
of a translation vector
T = u1 a1 + u2 a2 + u3 a3 (G.10)
All points in a lattice can be connected by means of T. A lattice is a regular periodic array
of points in 3 dimensions. The lattice is a mathematical construction. Given a lattice and a
basis of atoms, the crystal is given when the basis of atoms is attached to every point in the
lattice. Figure G.1 shows an illustration of the lattice, the crystal and the basis.
In order to set up the Bragg diffraction condition an expression for the electron number
density n(r) is necessary. It is used that the electron number density is periodic in r with the
period T. That is
n(r + T) = n(r) (G.11)
An arbitrary periodic function can be described as an infinite sum of complex exponential

function in a complex Fourier series. This yields
n(r) = ∑ nG e jG·r (G.12)

G
where G is a set of vectors such that the electron number density is invariant under all crystal
translations T.
Two lattices are connected to a crystal, the direct lattice and the reciprocal lattice. The
reciprocal lattice is defined as the Fourier transform of the lattice. The reciprocal lattice is
110
(a) The lattice. (b) The crystal structure is the lattice with the basis
of atoms attached to all lattice points.
(c) The basis atoms, in this case two

ions.
Figure G.1: Illustration of a lattice, a crystal and a basis in 2 dimensions.

[Kittel 1986, p. 6]
called the Fourier space of the crystal. Points in the reciprocal lattice are described by G. G
is given as
G = v1 b1 + v2 b2 + v3 b3 (G.13)
where v1 , v2 and v3 are arbitrary integers and b1 , b2 and b3 are the axis vectors. The coherence
between the axis vectors in the lattice and the axis vectors in the reciprocal lattice is given
from [Kittel 1986, p. 33] as
a2 × a3
b1 = 2π (G.14a)
a1 · a2 × a3
a3 × a1
b2 = 2π (G.14b)
a1 · a2 × a3
a1 × a2
b3 = 2π (G.14c)
a1 · a2 × a3
The Bragg diffraction condition is set up by means of Figure G.2 Contemplating the
figure yields that the difference in phase factor of waves scattered of volume elements r apart
is given as e j(k1 −k2 )·r . The electron density number of the volume element n(r) is proportional
to the amplitude of a wave scattered from the volume element dV . The total amplitude A of
the scattered wave is therefore given as
A= n(r)e j(k1 −k2 )·r dV (G.15)
which by (G.12) yields
A = ∑ nG e j(G−Δk)·r dV (G.16)
G
where
Δk = k2 − k1 (G.17)
111
Crystal specimen
dV
r k2
k1 α
Figure G.2: Illustration of Bragg scattering. The difference in phase factor

of waves scattered from volume elements r apart is given as e j(k1 −k2 )·r . The
length in part that the incident wave must travel longer to the point r then
to the point o is given as sin αr. This give rise to a phase 2π sin αr/λ which
is equal to k1 · r. The phase difference from r and o of the scattered wave
k2 is −k2 · r. The overall phase difference is therefore given as (k1 − k2 ) · r
which yields a difference in phase factor of e j(k1 −k2 )·r . [Kittel 1986, p. 35]
From (G.16) it is seen that when Δk = G then A = V nG where V is the volume of the crystal.
When Δk differs significantly from any G, A is negligibly small and amplitude of the electric
and magnetic field vector in the electromagnetic wave scattered is therefore negligibly small.
That is, a wave is only scattered when the Bragg diffraction condition is satisfied namely
Δk = G (G.18)
It is seen from (G.18) that it is the set of reciprocal lattice vectors that determines the possible
reflections. When an electron is scattered the energy h̄ω is conserved. Thus the frequency of
the incident wave is equal to the emergent wave ω1 = ω2 or ck1 = ck2 or k1 = k2 or k12 = k22 .
Combining this with (G.18) and (G.17) yields
k1 + G = k2 ⇒ (k1 + G)2 = k22 ⇒ (k1 + G)2 = k12 (G.19)
[Kittel 1986, 29-36]
Origin of Energy Gabs

In order to describe the origin of energy gabs a 1-dimensional linear crystal of lattice con-
stant a is used. When the Bragg diffraction condition is met, no wavelike solutions to the
Schrödinger equation exist. The Bragg condition is given in (G.19) which in 1-dimension
yields
1 nπ
(k + G)2 = k2 ⇒ k = ± G ⇒ k = ± (G.20)
2 a
where G = (n2π)/a. The Bragg reflection condition is met for k = ±(nπ)/a. The zone between
±π/a is called the first Brillouin zone. When a wavevector of a traveling wave is equal to
the Bragg condition k = ±π/a, a wave traveling to the right is Bragg reflected to travel to
the left and a wave traveling left is Bragg reflected to travel to the right. All the following
112
Bragg reflections will reverse the direction of the wave and a standing wave will therefore
be formed. Two standing waves can arise from the two traveling waves. One of them is the
wave traveling to the left and the other is the one traveling to the right, namely
ψ(x) = e(± jπx/a) (G.21)
The two possible standing waves can arise as
ψ1 (x) = e( jπx/a) + e(− jπx/a) = 2 cos(πx/a) (G.22a)
ψ2 (x) = e( jπx/a) − e(− jπx/a) = 2 j sin(πx/a) (G.22b)
Where the standing waves both are a sum of a right and a left traveling wave. The probabil-
ity densities of the two possible standing waves are given as
|ψ1 (x)|2 = 4 cos2 (πx/a) (G.23a)
|ψ2 (x)|2 = 4 sin2 (πx/a) (G.23b)
From (G.23) it can be seen that the two different standing waves pile up electrons at different
zones. The potential energy of a traveling electron in the field of the positive ion cores in a
periodic lattice is not constant over distance and the two standing waves will therefore have
different values of potential energy. Figure G.3 shows the potential energy of a conducting
electron in a field of ion cores of a 1-dimensional linear lattice together with the distribution
of probability density for the two possible standing waves and a traveling wave. It can be
U, potential energy
Probability density
|ψ2(x)|2 |ψ1(x)|2
Traveling
wave
x
Ion core
a
Figure G.3: The potential energy of a conducting electron in a field of

ion cores of a 1-dimensional linear lattice together with the distribution
of probability density for the two possible standing waves and a traveling
wave. [Kittel 1986, p. 162]
seen from Figure G.3 that when the average potential energy over the three charge distri-
butions is calculated, a different answer arises. The average potential energy of |ψ1 (x)|2 is
lower than the average potential energy of the traveling wave. The average potential energy
of |ψ2 (x)|2 is higher than the average potential energy of the traveling wave. The energy gab
Eg is given as the difference in average potential energy of |ψ2 (x)|2 and |ψ1 (x)|2 , that is
Eg =< V (x)|ψ2 (x)|2 > − < V (x)|ψ1 (x)|2 > (G.24)
where <> denotes the time-average value. [Kittel 1986, pp. 159-163]
113
Wave Equation of the Electron

(G.24) describes the energy gab if the wavevector k is at k = ±π/a. In order to describe the
wave equation of an electron in a general 1-dimensional periodic potential at general values
of k the 1-dimensional version of single-particle time independent Schrödinger equation
from (G.1) is used. Furthermore it is used that the potential energy is invariant under any
crystal translation V (x) = V (x + a). Because of this the potential energy may be expanded as
a Fourier series in the reciprocal lattice scalars G
V (x) = ∑ VG e jGx (G.25)

G
The 1-dimensional version of single-particle time independent Schrödinger equation now

yields

h̄2 d 2
2m dx2 ∑
− + VG e jGx ψ(x) = Eψ(x) (G.26)
G
The solution ψ(x) to (G.26) can also be expanded as a Fourier series in the permitted wavevec-
tors k given by (G.5)
ψ(x) = ∑ C(k)e jkx (G.27)

k
Substituting (G.27) into (G.26) yields
h̄2
∑ 2m k2C(k)e jkx + ∑ ∑ C(k)VG e j(G+k)x = E ∑ C(k)e jkx (G.28)
k k G k
Rearranging terms yields

h̄2 k2
2m
−E ∑ C(k)e jkx + ∑ VG ∑ C(k − G)e jkx = 0 (G.29)
k G k
or
(λk − E)C(k) − ∑ VGC(k − G) = 0 (G.30)

G
where
h̄2 k2
λk = (G.31)
2m
(G.30) is called the central equation, and it presents a set of linear equations. The number
of equations is given as the number coefficients C. When all the C’s are determined the
wavefunction in (G.27) is given as
ψ(x) = ∑ VGC(k − G)e j(k−G)x (G.32)

G
114
In order to show how (G.30) can be used, an approximate solution near the first Bragg re-
flection condition is calculated. First a wavevector exactly at the Bragg reflection condition
is considered. That is k2 = (1/2G)2 according to (G.20). This yields that the important coef-
ficients in the solution ψ(x) is C(1/2G) and C(−1/2G), all the other coefficients are neglected.
The solution to the Schrödinger equation is approximated as
ψ(x) = C(1/2G)e j /2Gx +C(−1/2G)e j(− /2G)x
1 1
(G.33)
If it is assumed that the potential energy contains only a single Fourier component, here
denoted V , two equations are now given according to the central equation (G.30), namely
(λk − E)C(1/2G) +VC(−1/2G) =0 (G.34)
(λk − E)C(−1/2G) +VC(1/2G) =0 (G.35)
where C(−3/2G) is transferred into first Brillouin zone by adding a suitable reciprocal lattice
vector of 2G [Kittel 1986, p. 173]. The two equations have solutions if the determinant of the
coefficients is zero

λk − E V
=0 (G.36)
V λk − E
which yields the energy as
h̄2
E = λ ±V = (1/2G)2 ±V (G.37)
2m
It can be seen that the energy has two roots at k = ± πa .
In order to solve the wave equation of k near the Bragg diffraction condition, a two
component approximation to the solution is used again. Now given as
ψ(x) = C(k)eikx +C(k − G)e j(k−G)x (G.38)
All other components of the solution are neglected. By the central equation in (G.30) this
yields two equations
(λk − E)C(k) +VC(k − G) =0 (G.39)
(λk−G − E)C(k − G) +VC(k) =0 (G.40)
where C(k − 2G) is transferred into first Brillouin zone by adding a suitable reciprocal lattice
vector of 2G. These equations have solutions if the determinant is zero

λk − E V
=0 (G.41)
V λk−G − E
or
E 2 − E(λk−G + λk ) + λk−G λk −V 2 = 0 (G.42)
Solving this second order equation yields two roots of the energy, namely

1 1
E = (λk−G + λk ) ± (λk−G + λk )2 − λk−G λk +V 2 (G.43)
2 4
Figure G.4 shows a plot of the two roots of the energy as a function of the wavevector k. The
2
energies are plotted with h̄m = 1, G = 2 and V = 0.2. It can be seen from Figure G.4 that each
energy root describes an energy band. [Kittel 1986, pp. 167-179]
115
Second Band
1
Zone boundary
0.8
0.6
E
0.4
First Band
0.2
Free electron
0
0 0.5 1 1.5
k
Figure G.4: Plot of energy bands near the Bragg diffraction condition 1/2G.
The kinetic energy of the free electron is drawn in order to compare the en-
ergy levels of an electron in a periodic lattice with the energy of a free elec-
tron. After Figure 9 in [Kittel 1986, p. 176]. [CD 2004, matlab/Bandgab/en-
ergyband.m]
Band Structures and Refractive Indexes of Silicon and Copper

In order to related the refractive indexes to the band structures, the plots of the calculated
refractive indexes in appendix H is the starting point. By contemplating Figure H.2 on page
121, it can be seen that an abrupt change in the refractive index is found only in the case of
silicon and copper. As it is these abrupt changes of the refractive index which can be related
to the energy band structure of the materials, only the band structure of silicon and copper
will be related to the refractive indexes.
By contemplating Figure H.2(a) it can be seen that an abrupt change in both the real and
imaginary part of the refractive index for silicon can be found approximately at λ = 385 nm.
The energy of the photon corresponding to the wavelength λ = 385 nm can be found as
2πc
E = h̄ω = h̄ = 3.22 eV (G.44)
λ
where h̄ = 6.583 · 10−16 eVs is Planck’s constant divided by 2π and the speed of light is c =
3.0 · 108 m/s. The corresponding wavenumber k can be found as
2π
k= = 1.632 · 107 m−1 (G.45)
λ
By contemplating Figure H.2(b) the wavelength of the abrupt changes in the refractive index
for copper can approximately be found at λ = 580 nm. The corresponding photon energy
116
and wavenumber is given as
E = 2.17 eV (G.46)
−1
k = 1.083 · 10 m 7
(G.47)
The lattice constants of silicon and copper are given as
aSi = 0.543 nm (G.48)

aCu = 0.361 nm (G.49)
[Kittel 1986, p. 23]. The zone boundary of the first Brillouin zone in the 1-dimensional case
of silicon and copper is given as
π
kSi = = 5.79 · 109 m−1 (G.50)
aSi
π
kCu = = 8.70 · 109 m−1 (G.51)
aCu
This shows that the wavenumbers of the photons are small compared with the first zone
boundaries of the crystals. Figure G.5 shows the calculated band structures of silicon and
copper. When the band structures are plotted in k-space in the first Brillouin zone the k-
(a) Calculated electronic bands along lines (b) The calculated energy band of copper in the first Bril-
of high symmetry for Si in eV. Γ is at louin zone. Γ is at k = 0 [Ehrenreich & Philipp 1962].
k = 0 and all the other points in k-space
is inside the first Brillouin zone [Rohlfing
et al. 1993].
Figure G.5: The calculated band structures of silicon and copper.
values of the photons are small compared with the k-values of the zone boundaries. This
means that a transition of an electron between two bands, due to a photon, will be an al-
most vertical transition in k-space due to the conservation of wavevector. The conservation
117
of wavevector is due to the conservation of linear momentum because the linear momen-
tum is given as p = h̄k. In Figure G.5(a) the direct band gab1 of silicon is drawn and is
approximately equal to 3.4 eV. Comparing this energy with the energy of the photon of 3.22
eV yields a good correspondence. It is therefore seen that the abrupt change in refractive
index of silicon at λ = 385 nm can be connected to the direct energy gab in silicon of 3.4
eV. When the energy of the photon is equal to the band gab energy, the photon can excite
an electron into a higher energy band and the photon will thereby be absorbed or slowed.
When photons are slowed down or absorbed the refractive index is changed significantly
from unity.
Contemplating Figure G.5(b), the band structure of copper, shows that the energy scale is
in Rydberg. The Rydberg is an energy unit defined as the ground-state energy of an electron
in the hydrogen atom.
me e4
1 Ry = = 13.6 eV (G.52)
2h̄2
The energy of the photon corresponding to the wavelength λ = 580 nm in Rydberg is given
as
2.17eV
E= = 0.16 Ry (G.53)
13.6eV
On Figure G.5(b) the thresholds for transitions from the d-band, under the Fermi energy2 ,
to the conducting band are drawn. It can be seen that these energy gabs are approximately
of 0.16 Ry which is approximately equal to the energy of the photon, corresponding to the
wavelength of abrupt change in refractive index of copper. It is therefore concluded that the
abrupt change in refractive index of copper around the wavelength 580 nm can be connected
to the transitions from the d-band to the conducting band in the band structure of copper.
In general, it is observed that the abrupt changes in the refractive index at given wave-
lengths are closely related to the band gaps of the band structure of the materials. The band
gap energy can be directly converted to the wavelength of the corresponding change in the
refractive index.
1 The direct band gab is the band gab between the valence band and the conducting band at k = 0.
2 The Fermi energy of a material is the energy boundary between the occupied orbitals and the vacant orbitals
at 0 K.
118
Test Report of Refractive Index
of Crystals H
This appendix is the test report describing the experiment where the ψ and Δ angles of silicon, copper,
aluminum and silver are measured and the refractive indexes of these crystals are calculated. All
discussion of the test results and conclusions drawn about the test results are found in Chapter 7 in
the main report.
Purpose
The purpose of this test is to measure the ellipsometric parameters of several crystals by
means of an ellipsometer and from these calculate the refractive index as a function of the
wavelength. The calculated values are then compared to the table values of the refractive
indexes. The crystals under investigation are silicon, silver, copper and aluminum.
Equipment and Materials

The equipment used during the experiment is:
• SENTECH Instruments - UV-VIS-NIR spectroscopic ellipsometer SE 850
• SENTECH Instruments - Spectraray II Software
• Silicon sample
• Copper sample
• Aluminum sample
• Silver sample
Procedure
The experiment was performed on Skjernvej 4C in room 1.109 on the 20th of October 2004.
The ellipsometer was powered up, the xenon lamp was turned on, the incident angle was
set to 70◦ on the ellipsometer and the program Spectraray II was initiated. The sample was
placed on the ellipsometer sample stage and the sample stage was adjusted such that the
reflected light hit the detector inside the ellipsometer precisely, which was done by measur-
ing the intensity. This is important in order to get proper measurements. The ψ-Δ spectrum
was measured by means of Spectraray II from 350 nm to 825 nm with the UV-VIS setting in
Spectraray II. The data was stored on a floppy disk.
119
APPENDIX H. TEST REPORT OF REFRACTIVE INDEX OF CRYSTALS
Results
The data can be found on [CD 2004, /matlab/test_surface_201004]. The measured ψ-Δ spec-
trums for the surfaces used are shown in Figure H.1
Silicon Copper
180 140
Psi
Delta
160
Psi
Delta 120
140
120 100
Psi and Delta [Degrees]

100
80
80
60 60
40
40
20
0 20
350 400 450 500 550 600 650 700 750 800 850 350 400 450 500 550 600 650 700 750 800 850
(a) (b)
Aluminum Silver
160 130
Psi Psi
Delta Delta
120
140
110
120
100
90
100
80
80
70
60
60
50
40
40
20 30
350 400 450 500 550 600 650 700 750 800 850 350 400 450 500 550 600 650 700 750 800 850
(c) (d)
Figure H.1: ψ and Δ as a function of wavelength measured with Sen-

tech UV-VIS-NIR spectroscopic ellipsometer SE 850 on silicon, copper, alu-
minum and silver. [CD 2004, matlab/surface_test/RefractiveIndex.m]
It is seen from Figure H.1 that the curves of ψ and Δ, for all the surfaces, in general are
smooth curves. The only exception is the Δ angle of silicon which fluctuates at wavelengths
longer than 600 nm. None of them are subject to a considerable amount of random noise.
This is to be expected due to the nature of ellipsometric measurements.
120
Data Processing
In order to convert the measured ellipsometric parameters ψ and Δ into the refractive index
ñ the relation (5.4) on page 33 is used, namely

1 − 4 sin θ tan (ψ) e + 2 tan (ψ) e + tan (ψ) e n0 sin θ
2 jΔ jΔ 2 jΔ
ñ = (H.1)
cos θ [1 + tan (ψ) e jΔ ]
where θ is the incident angles, n0 is the refractive index of the ambient, in this case air.
Using (H.1), the refractive index is calculated as a function of wavelength by means of the
measured ψ-Δ spectrum. A plot of the real and the imaginary part of the refractive index as
a function of the wavelength for the four surfaces together with the table values of the real
and imaginary part of the refractive index is shown in Figure H.2 and H.3.
Silicon Copper
7 6
n n
k k
6
5
5
4
4
n and k
n and k
2
2
1
1
0 0
350 400 450 500 550 600 650 700 750 800 350 400 450 500 550 600 650 700 750 800
(a) (b)
Figure H.2: The calculated real n and imaginary k part of the refractive
index as a function of wavelength together with table values of the real
and imaginary part of the refractive index of silicon and copper. The real
part of the table values are dots and the imaginary part are asterisks. The
table values are found in [Palik 1998] and [Klein & Furtak 1986]. [CD 2004,
matlab/surface_test/RefractiveIndex.m]
121
APPENDIX H. TEST REPORT OF REFRACTIVE INDEX OF CRYSTALS
Aluminum Silver
9 6
n n
k k
8
5
7
6 4
5
n and k
n and k
3
4
3 2
2
1
1
0 0
350 400 450 500 550 600 650 700 750 800 350 400 450 500 550 600 650 700 750 800
(a) (b)
Figure H.3: The calculated real n and imaginary k part of the refractive
index as a function of wavelength together with table values of the real
and imaginary part of the refractive index of aluminum and silver. The
real part of the table values are dots and the imaginary part are aster-
isks. The table values are found in [Palik 1998]. [CD 2004, matlab/sur-
face_test/RefractiveIndex.m]
122
Test Report of the Thickness of
an SiO2 Film on Si I
This appendix is the test report describing the measurement of the silicon dioxide film thickness. All
discussion of the test results and conclusions drawn about the test results are found in Chapter 8 in
the main report.
Purpose
The purpose of this test is to measure the ellipsometric parameters in order to enable cal-
culation of the thickness of a silicon dioxide layer on a silicon substrate as a function of the
wavelength by means of an ellipsometer.

The equipment used during the experiment:
• SENTECH Silicon wafer with a silicon dioxide film
Procedure
The experiment was performed on Skjernvej 4C in room 1.109 on the 2nd of November 2004.
The ellipsometer was powered up, the xenon lamp was turned on, the incident angle was
set to 70◦ on the ellipsometer and the program Spectraray II was initiated. The sample was
placed on the ellipsometer sample stage and the sample stage was adjusted such that the
reflected light hit the detector inside the ellipsometer precisely, which was done by measur-
ing the intensity. This is important in order to get proper measurements. The ψ-Δ spectrum
was measured by means of Spectraray II from 350 nm to 825 nm with the UV-VIS setting in
Spectraray II. The data was stored on a floppy disk.
Results
The measured ψ-Δ spectrum for the sample used is shown in Figure I.1.
123
APPENDIX I. TEST REPORT OF THE THICKNESS OF AN SIO2 FILM ON SI
SiO film on Si
2
140
Psi
Delta
120
100
80
60
40
20
0
350 400 450 500 550 600 650 700 750 800 850
Wavelength [nm]
Figure I.1: ψ and Δ as a function of wavelength measured with Sentech

UV-VIS-NIR spectroscopic ellipsometer SE 850.
It is seen from the figure that the curves of ψ and Δ are relative smooth curves. Some
small fluctuations do however exist in the higher wavelength region for Δ.
The measurement is not subject to a considerable amount of random noise. This is to be
expected due to the nature of ellipsometric measurements.
Data Processing
The calculation of the SiO2 film thickness is done by utilizing the method described in Sec-
tion 5.2. In this calculation the angle of incidence, the free-space wavelength of the light,
the ellipsometric parameters ψ and Δ and the refractive indexes for air, SiO2 and Si must be
known in order to determine the film thickness. Values for the refractive index of Si and
SiO2 are found by use of [Palik 1998, I: pp. 563-565] and [Index of Refraction Values and
Photonics Calculations 2004] respectively.
Solving (5.14) leads to two solutions, both complex. The solution with a positive real
part and the smaller imaginary part is chosen. The result of these calculations is presented
in Figure I.2
124
Film thickness of SiO on Si
2
23.3
23.2
23.1
Film thickness [nm]
23
22.9
22.8
22.7
22.6
350 400 450 500 550 600 650 700 750 800 850
Wavelength [nm]
Figure I.2: The calculated thickness of the SiO2 film as a function of wave-
length. [CD 2004, matlab/thickness/film_thickness_sio_2.m]
125
Test Report of the Thickness of
a Polymer Film on Si J
This appendix is the test report describing the experiment where ψ and Δ of the polymer PI-5878G
from HD MicroSystems are measured and the thickness of the polymer is calculated. The polymer
was spin coated on a silicon wafer by NanoNord. Two silicon wafers with polymer film were made.
The two polymers were spin coated with the same angular velocity, namely 4000 RPM. The wafers
were however spin coated for different time periods, namely 1 and 5 minutes.
All discussion of the test results and conclusions drawn about the test results are found in Chap-
ter 9 in the main report.
Purpose
The purpose of this test is to determine the thickness, in the center of the sample, of the
two polymers by means of a ψ-Δ measurement performed with the ellipsometer. Another
purpose is to investigate the uniformity of the thickness of the two polymer films on the
wafers by means of a sequence of ψ-Δ measurements.

The equipment used during the experiment:
• Two silicon wafers with a spin coated PI-5878G polymer film
Procedure
The procedure of the two experiments is almost the same. In both experiments the ellip-
someter was powered up, the xenon lamp was turned on, the incident angle was set to 70◦
on the ellipsometer and the program Spectraray II was initiated. The sample was placed
on the ellipsometer sample stage and the sample stage was adjusted such that the reflected
light hit the detector inside the ellipsometer precisely, which was done by measuring the
intensity. This is important in order to get proper measurements. The rest of the procedure
of the two experiments is not alike and will therefore be described separately.
127
APPENDIX J. TEST REPORT OF THE THICKNESS OF A POLYMER FILM ON SI
Thickness
The experiment was performed on Skjernvej 4C in room 1.109 on the 9th of December 2004.
Two ψ-Δ spectrums were measured, one for each wafer. The ψ-Δ spectrums were measured
from 350 nm to 825 nm at the middle of the two polymers and they were stored on a floppy
disk.
Uniformity of Film Thickness

The experiment was performed on Skjernvej 4C in room 1.109 on the 12th of December 2004.
In this experiment 68 ψ-Δ spectrums were measured, 34 for each wafer. Figure J.1 depicts
wafer. The figure shows four axes, an x-axis, a y-axis, a u-axis and a v-axis. These are the
Figure J.1: The axes on the wafer along which the ψ-Δ spectrums are mea-
sured. In case of the thickest polymer film the spectrums are measured
along x and y. In case of the thinnest polymer film the spectrums are mea-
sured along u and v.
axes along which the ψ-Δ spectrums are measured. In the case of the silicon wafer with
the thickest polymer layer imposed the ψ-Δ spectrums were measured along the x-axis and
the y-axis. 17 spectrums were measured along each axis, with the positive direction along
the axis. The measurements were performed from -40 mm to 40 mm with a step of 5 mm
between each measurement. In the case of the thinnest polymer film the ψ-Δ spectrums were
measured along the u-axis and the v-axis. 17 spectrums were also measured along each axis,
with the positive direction along the axis. The measurements were again performed from
-40 mm to 40 mm with a step of 5 mm between each measurement.
All the ψ-Δ spectrums were measured from 350 nm to 825 nm by means of the UV-VIS
setting in Spectraray II. All the spectrums were stored on a floppy disk.
Results
In this section the ψ-Δ spectrums of the measurements are shown. In the experiment where
the uniformity of the two polymer films are tested there are 68 ψ-Δ spectrums and they will
128
therefore not be shown in this appendix. For reference to these spectrums see [CD 2004,
matlab/polymer_thickness/uniformity/]. In the experiment were the thicknesses of the
two polymer films is determined, only one ψ-Δ spectrum for each wafer were measured and
they are presented in the following subsection.
Thickness
The measured ψ-Δ spectrums of the polymer films on the silicon wafers are shown in Figure
J.2 and J.3. It can be seen from the figure that due to the large thickness of the films, an
Polymer Thick
150
Psi
Delta
100
50
0
350 400 450 500 550 600 650 700 750 800 850
Wavelength [nm]
Figure J.2: ψ-Δ spectrums of the thickest polymer film. [CD 2004, mat-
lab/polymer_thickness/psidelta_plotter.m]
interference pattern is seen in ψ and Δ. It can also be seen that the period between peaks is
larger for the thin film, indicating that it in fact is thinner.

See [CD 2004, matlab/polymer_thickness/uniformity/].
129
Polymer Thin
140
Psi
Delta
120
100
80
60
40
20
0
350 400 450 500 550 600 650 700 750 800 850
Wavelength [nm]
Figure J.3: ψ-Δ spectrums of the thinner polymer film. [CD 2004, mat-
lab/polymer_thickness/psidelta_plotter.m]
Data Processing
Thickness
The calculation of the two polymer film thicknesses is done by utilizing the method de-
scribed in Section 5.3. In this method the interference pattern in ψ and Δ is used to estimate
the thickness of the film imposed on the surface. In order to perform the calculation, the
index of refraction of both the polymer and the silicon wafer must be know. The index of re-
fraction of the polymer is found in [MicroSystems 2003], and the index of refraction of the sil-
icon wafer is found by use of [Index of Refraction Values and Photonics Calculations 2004].
The result of these calculations is presented in Figure J.4(a) for the thick polymer and in
Figure J.4(b) for the thin polymer. The ψ-Δ spectrum between 350 nm and 450 nm does not
contain adequate information to determine the thickness as a function of wavelength.

In order to determine the uniformity of the thickness of the two polymer films, the thickness
of all the points on the surface where the ψ-Δ spectrums are measured must be determined.
It is chosen to calculate the thickness in the wavelength interval from 650 nm to 825 nm.
The thickness in the point where the ψ-Δ spectrum is measured is then found as the average
value of the thickness over the wavelength interval. Figure J.5 depicts the thickness of the
thick polymer film measured over the x- and the y-axis as depicted on Figure J.1. Figure J.6
shows the thickness of the thin polymer film measured over the u- and the v-axis as shown
on Figure J.1.
130
7.5 4.4
7
4.2
6.5
4
Thickness [um]
Thickness [um]
6
3.8
5.5
3.6
5
3.4
4.5
4 3.2
450 500 550 600 650 700 750 800 450 500 550 600 650 700 750 800
(a) Thickness of the thick polymer at the middle, as (b) Thickness of the thin polymer at the middle, as
a function of wavelength. a function of wavelength.
Figure J.4: Polymer thickness at the middle of the wafer, as a func-

tion of wavelength. [CD 2004, matlab/polymer_thickness/film_thick-
ness_polymer_spectrum.m]
Polymer Thick
6
x
y
4
Thickness [um]
0
−40 −30 −20 −10 0 10 20 30 40
Position [mm]
Figure J.5: Thickness of the thick polymer film measured over the
x- and the y-axis as depicted in Figure J.1. [CD 2004, matlab/poly-
mer_thickness/film_thickness_polymer.m]
131
Polymer Thin
6
u
v
4
Thickness [um]
0
−40 −30 −20 −10 0 10 20 30 40
Position [mm]
Figure J.6: Thickness of the thin polymer film measured over the
u- and the v-axis as depicted in Figure J.1. [CD 2004, matlab/poly-
mer_thickness/film_thickness_polymer.m]
132

Ellipsometry: Alborg Niversity

Uploaded by

Copyright:

Available Formats

Ellipsometry: Alborg Niversity

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ellipsometry: Alborg Niversity

Uploaded by

Copyright:

Available Formats

A ALBORG U NIVERSITY

Institute of Physics and Nanotechnology

Aalborg University, Tuesday 21st 2004

Tobias Holmgaard Niels Anker Kortbek

5 Calculations of Physical Properties 33

III Simulations and Experiments 39

6 Simulation of the Refractive Index of Crystals 41

7 Discussion of Crystal Measurements 49

8 Discussion of SiO2 Film Measurements 59

9 Discussion of Polymer Film Measurements 63

A The Electromagnetic Wave Model of Light 71

B Fresnel Reflection and Transmission Coefficients 77

C Three Phase Optical System 85

D The Jones Formalism 91

E Correlation between the Stokes Parameters and ψ and Δ 97

F Index of Refraction 101

G Index of Refraction and Band Structures of Crystals 109

H Test Report of Refractive Index of Crystals 119

I Test Report of the Thickness of an SiO2 Film on Si 123

J Test Report of the Thickness of a Polymer Film on Si 127

Figure 1.1: The general principle in ellipsometry. [Jawoollam 2004]

When a monochromatic, plane light wave is directed at a surface at oblique incidence,

Figure 1.2: Illustration of a thin film on top of a crystal.

3.1 Definition of Polarization

Ex (z,t) = Ax cos(ωt − kz + φx ) (3.3a)

and the relative amplitude, which is a relation between Ax and Ay .

3.2 Polarization Types

3.2.1 Linearly Polarized Light

φ = ±pπ , for p = {0, 1, 2, · · · } (3.6)

3.2.2 Circularly Polarized Light

(a) Depiction of E in the xy-plane. (b) Illustration of E(z).

Figure 3.1: Illustration of linearly polarized light for φ = π and Ax = Ay . Ex

and the following amplitude relation must be fulfilled

(a) Depiction of E(z) in the xy-plane. (b) Illustration of E(z).

Figure 3.2: Illustration of LCP light (φ = −π/2) and Ax = Ay . Ex and Ey are

Adding the squared of (3.10a) to the squared of (3.10b) yields

3.2.3 Elliptically Polarized Light

Description of Elliptically Polarized Light Starting From Circularly Polarized Light

General description of Elliptically Polarized Light

Ex = Ax cos(ωt − kz) (3.13a)

which can be written as

Squaring this, results in

Substituting the squared of (3.14a) into (3.17) yields

Figure 3.3: Illustration of elliptically polarized light.

3.2.4 Unpolarized Light

3.3 Definition of the ellipsometric parameters ψ and Δ

Figure 3.4: Illustration showing the orientation of the coordinate systems

Figure 3.5: Input χi and output χo to an optical system S.

Rewriting this expression using the Jones vector, (3.4) yields

χi Aiy e jφiy Aox e jφox

χi Aox j(φox −φoy )

This is illustrated in Figure 3.6.

Figure 3.6: Illustration of ψ, which is defined for reflected elliptically po-

Furthermore, the parameter Δ is defined as

Δ = φox − φoy (3.28)

Using (3.27) and (3.28), (3.26) can be expressed as

which is a general ellipsometer equation [Azzam & Bashara 1977, p. 259].

3.3.1 Connecting ψ and Δ to ρπ and ρσ

incidence, (3.23) can be rewritten to

where the new variable ψ is given by the relation

tan(ψ ) = tan(ψ) (4.27)