23.05.2014 Views

Towards a covariant formulation of electromagnetic wave polarization

Towards a covariant formulation of electromagnetic wave polarization

Towards a covariant formulation of electromagnetic wave polarization

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Towards</strong> a<br />

<strong>covariant</strong> <strong>formulation</strong> <strong>of</strong><br />

<strong>electromagnetic</strong> <strong>wave</strong><br />

<strong>polarization</strong><br />

by<br />

Mikael Ol<strong>of</strong>sson<br />

Swedish Institute <strong>of</strong> Space Physics<br />

Uppsala-division<br />

Box 537<br />

751 21 Uppsala<br />

IRF Scientific Report 273<br />

January 2001


i<br />

Abstract<br />

A <strong>covariant</strong> <strong>formulation</strong> <strong>of</strong> <strong>electromagnetic</strong> <strong>wave</strong> <strong>polarization</strong> is introduced. It<br />

is shown that 36 parameters are needed to fully describe the <strong>polarization</strong> <strong>of</strong> a<br />

partially coherent <strong>wave</strong>. The 36 parameters becomes more understandable when<br />

they are organized according to their tensor nature. An attempt is made to find<br />

the largest set <strong>of</strong> tensor constituents, which uniquely determines the 36 <strong>polarization</strong><br />

parameters. The <strong>formulation</strong> is based on an irreducible representation<br />

<strong>of</strong> <strong>electromagnetic</strong> <strong>wave</strong> <strong>polarization</strong>, using geometrical quantities with known<br />

transformation properties under certain groups. The groups we concentrate on<br />

are the Cartesian rotation group, SO(3), and the proper homogeneous Lorentz<br />

group, SO(3, 1). Specifically, the parameters are obtained by decomposing tensor<br />

products, <strong>of</strong> the electric and magnetic vector fields, ⃗ E and ⃗ B respectively,<br />

into scalars, vectors and tensors. A set <strong>of</strong> Lorentz <strong>covariant</strong> <strong>polarization</strong> tensors,<br />

the group SO(3, 1), are constructed from the <strong>electromagnetic</strong> field strength<br />

tensor and its dual tensor. These <strong>covariant</strong> tensors have the SO(3) parameters<br />

as components. The geometric quantities under SO(3) have a known physical<br />

interpretation, only two are not previously found in the literature. One application<br />

<strong>of</strong> these <strong>polarization</strong> parameters is to characterize the <strong>polarization</strong> <strong>of</strong> <strong>wave</strong>s<br />

in plasma, e.g., by determining the refractive index vector, ⃗n. The <strong>covariant</strong><br />

spectral density tensor is introduced and defined as the tensor product <strong>of</strong> the<br />

field strength tensor with it-self. The <strong>covariant</strong> spectral density tensor is expanded<br />

in a basis (the Dirac gamma matrices) and decomposed into a reducible<br />

tensor representation. It is found that this reducible representation consists <strong>of</strong><br />

known <strong>covariant</strong> quantities. Since <strong>wave</strong>s are considered, it is convenient to use<br />

the frequency domain which implies the use <strong>of</strong> complex spaces.


ii<br />

Preface<br />

This report is a Master thesis in physics at the University <strong>of</strong> Uppsala. The thesis<br />

was performed at the Swedish Institute <strong>of</strong> Space Physics, Uppsala division (IRF-<br />

U). Supervisors for this work have been Jan Bergman and Tobia Carozzi and<br />

examinator was pr<strong>of</strong>essor Bo Thidé.<br />

I would like to thank my supervisors Jan Bergman and Tobia Carozzi for<br />

suggesting this problem to me and for their help and tutoring in completing<br />

this thesis. I would also like to thank pr<strong>of</strong>essor Bo Thidé for giving me the<br />

opportunity to do my thesis at the Swedish Institute <strong>of</strong> Space Physics, Uppsala<br />

division (IRF-U) and for his help with correcting and completing this report.<br />

Thanks to all the personnel at IRF-U for a pleasant time.


Contents<br />

1 Introduction 1<br />

2 Background 3<br />

2.1 Covariant <strong>formulation</strong> <strong>of</strong> Maxwell equations . . . . . . . . . . . . 3<br />

2.1.1 Basic relativity and tensor operations . . . . . . . . . . . 3<br />

2.1.2 Microscopic Maxwell equations . . . . . . . . . . . . . . . 6<br />

2.1.3 Construction <strong>of</strong> the <strong>electromagnetic</strong> field strength tensor . 8<br />

2.1.4 Electromagnetic energy-momentum tensor . . . . . . . . . 9<br />

2.2 Generalized Stokes parameters . . . . . . . . . . . . . . . . . . . 11<br />

2.2.1 Spectral density matrix . . . . . . . . . . . . . . . . . . . 11<br />

2.2.2 Instantaneous Stokes’ parameters . . . . . . . . . . . . . . 11<br />

2.2.3 Three-dimensional <strong>polarization</strong> parameters . . . . . . . . 13<br />

2.2.4 Interpretation <strong>of</strong> the 3D <strong>polarization</strong> parameters . . . . . 13<br />

2.2.5 The Storey and Lefeuvre six-quasivector . . . . . . . . . . 15<br />

2.3 Irreducible representation . . . . . . . . . . . . . . . . . . . . . . 17<br />

2.3.1 Complex vector algebra . . . . . . . . . . . . . . . . . . . 17<br />

2.3.2 The <strong>electromagnetic</strong> field . . . . . . . . . . . . . . . . . . 18<br />

2.3.3 The product E ⃗ ⊗ B ⃗ . . . . . . . . . . . . . . . . . . . . . 20<br />

2.4 Group theory <strong>of</strong> transformation matrices . . . . . . . . . . . . . . 22<br />

2.4.1 The general linear group and subgroups . . . . . . . . . . 22<br />

2.4.2 Physical use <strong>of</strong> the groups . . . . . . . . . . . . . . . . . . 23<br />

2.4.3 Irreducible tensor representation <strong>of</strong> Lie groups . . . . . . 23<br />

3 Covariant spectral density tensor 27<br />

3.1 Covariant spectral density tensor . . . . . . . . . . . . . . . . . . 27<br />

3.2 Pauli spin matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 28<br />

3.3 Dirac gamma matrices . . . . . . . . . . . . . . . . . . . . . . . . 28<br />

3.4 Decomposition <strong>of</strong> τ αβγδ . . . . . . . . . . . . . . . . . . . . . . . 29<br />

4 Irreducible representation <strong>of</strong> <strong>polarization</strong> 31<br />

4.1 The irreducible parameters under SO(3) . . . . . . . . . . . . . . 31<br />

4.1.1 Σ, the sum . . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />

4.1.2 ∆, the difference . . . . . . . . . . . . . . . . . . . . . . . 33<br />

4.1.3 χ, the real part . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

4.1.4 ψ, the imaginary part . . . . . . . . . . . . . . . . . . . . 34<br />

4.2 Table <strong>of</strong> the parameters . . . . . . . . . . . . . . . . . . . . . . . 34<br />

4.3 Normalization <strong>of</strong> the S, V, T parameters . . . . . . . . . . . . . . 35<br />

4.3.1 Example: the refractive index . . . . . . . . . . . . . . . . 35<br />

iii


iv<br />

CONTENTS<br />

4.4 Stokes parameters from the S, V, T parameters . . . . . . . . . . 36<br />

D Covariant <strong>polarization</strong> tensors 61<br />

D.1 The X tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62<br />

D.2 The Y tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63<br />

D.3 The K and K tensors . . . . . . . . . . . . . . . . . . . . . . . . 64<br />

6 Reducible representation <strong>of</strong> τ αβγδ 45<br />

6.1 The <strong>covariant</strong> spectral density tensor . . . . . . . . . . . . . . . . 45<br />

6.2 The irreducible tensors . . . . . . . . . . . . . . . . . . . . . . . . 46<br />

6.3 Reducible representation . . . . . . . . . . . . . . . . . . . . . . . 47<br />

7 Conclusions 49<br />

A Generalized <strong>polarization</strong> parameters 51<br />

B Parameters from Dirac gamma matrices 53<br />

C E, ⃗ B ⃗ component form <strong>of</strong> S, V, T parameters 57<br />

C.1 Σ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57<br />

C.2 ∆ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58<br />

C.3 χ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58<br />

C.4 ψ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58<br />

D Covariant <strong>polarization</strong> tensors 61<br />

D.1 The X tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62<br />

D.2 The Y tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63<br />

D.3 The K and K tensors . . . . . . . . . . . . . . . . . . . . . . . . 64


Chapter 1<br />

Introduction<br />

A fundamental principle in constructing a physical law is to define it with equations<br />

that are independent <strong>of</strong> the choice <strong>of</strong> coordinates. This way <strong>of</strong> formulating<br />

physics is known as a <strong>covariant</strong> <strong>formulation</strong>. The word <strong>covariant</strong> relates to tensors<br />

which are geometrical objects defined by their transformation properties.<br />

The simplest example is a scalar, being a rank zero tensor, which is known to<br />

be invariant for any choice <strong>of</strong> coordinate system.<br />

The purpose <strong>of</strong> this theoretical work is to find a manifestly <strong>covariant</strong> set <strong>of</strong><br />

parameters describing classical <strong>electromagnetic</strong> <strong>wave</strong> <strong>polarization</strong>. Traditionally,<br />

<strong>electromagnetic</strong> <strong>wave</strong> <strong>polarization</strong> is characterized using Stokes’ parameters<br />

(see e.g. Stokes [22] and Born & Wolf [3]). Considering Stokes’ parameters,<br />

which are manifestly <strong>covariant</strong> in the two-dimensional representation (this is<br />

shown in Section 2.4, using the SO(2) group), it is not immediately apparent<br />

how they transform under the Lorentz group. Furthermore it is not clear how<br />

Stokes’ parameters handle <strong>wave</strong>s which are not simply quasi-monochromatic<br />

transverse <strong>wave</strong>s.<br />

The approach used in this thesis is that <strong>of</strong> classical macroscopic <strong>electromagnetic</strong><br />

field <strong>wave</strong>s and not the photonic or quantum mechanical picture. In<br />

this context, we define <strong>polarization</strong> as the second order statistical correlation<br />

between electric and magnetic field components. Therefore we compare the<br />

components <strong>of</strong> the field pairwise to see how they are related to each other, i.e.,<br />

one has quadratic components. Invariants are generally quadratic quantities in<br />

some way, for example the length <strong>of</strong> a vector (using the Euclidean three dimensional<br />

space with the ordinary metric) is invariant under rotation but the<br />

three different components <strong>of</strong> the vector are changed under a rotation <strong>of</strong> coordinates.<br />

Many <strong>of</strong> the calculations presented in this work are not restricted to<br />

<strong>electromagnetic</strong> fields but can be applied to a general field.<br />

As we consider the <strong>wave</strong> properties <strong>of</strong> electric and magnetic fields, it is convenient<br />

to use the frequency domain representation, i.e., the Fourier transform<br />

<strong>of</strong> the <strong>wave</strong>. When using the frequency domain representation <strong>of</strong> the <strong>wave</strong> it<br />

becomes necessary to use complex spaces. In a complex space, one ends up with<br />

36 correlation parameters with two components combined into the form AB † ,<br />

where A and B are two <strong>of</strong> the six field components (A and B can be the same<br />

component).<br />

After formulating <strong>polarization</strong> using the <strong>covariant</strong> formalism, the main emphasis<br />

is on combining these 36 quantities to obtain parameters which transform<br />

1


2 CHAPTER 1. INTRODUCTION<br />

according to certain transformation rules (the number <strong>of</strong> degrees <strong>of</strong> freedom will<br />

not change). These parameters are obtained by means <strong>of</strong> an irreducible tensor<br />

representation <strong>of</strong> the spectral density matrix.<br />

Chapter 2 gives an introduction and a background where some useful concepts<br />

are introduced: Stokes’ parameters and their three dimensional generalization<br />

(see Barut [5] and Carozzi, Karlsson & Bergman [6]) are introduced and<br />

the Storey and Lefeuvre [23] representation <strong>of</strong> the <strong>electromagnetic</strong> field is discussed.<br />

A general decomposition <strong>of</strong> second rank tensors, in three dimensions, is<br />

introduced and applied to a general field. The relativistic meaning <strong>of</strong> covariance<br />

and the tensor algebra <strong>of</strong> relativity is presented, together with the construction<br />

<strong>of</strong> the <strong>electromagnetic</strong> field strength tensor (<strong>of</strong>ten referred to as just the field<br />

tensor). In Chapter 3 we introduce and define the <strong>covariant</strong> spectral density<br />

tensor. This tensor is expanded in a basis consisting <strong>of</strong> the Dirac gamma matrices.<br />

In Chapter 4 a new set <strong>of</strong> parameters giving a <strong>covariant</strong> representation <strong>of</strong><br />

the <strong>electromagnetic</strong> field with geometrical quantities is presented. The covariance<br />

here refers to real rotations, SO(3), and the space is the three dimensional<br />

complex Euclidean space. Chapter 5 contains a construction <strong>of</strong> relativistically<br />

<strong>covariant</strong> tensors in which the new parameters are found as entries. The group<br />

we use here is the real proper homogeneous Lorentz group, SO(3, 1). In Chapter<br />

6 the <strong>covariant</strong> spectral density tensor is decomposed into a reducible tensor<br />

representation, consisting <strong>of</strong> irreducible tensors and a reducible rest tensor. All<br />

tensors in this representation are manifestly <strong>covariant</strong> and the irreducible tensors<br />

are identified with known physical quantities. In Chapter 7 the thesis is<br />

summarized and concluded.


Chapter 2<br />

Background<br />

2.1 Covariant <strong>formulation</strong> <strong>of</strong> Maxwell equations<br />

In this Section we introduce some basic concepts from the theory <strong>of</strong> relativity<br />

and a <strong>covariant</strong> <strong>formulation</strong> <strong>of</strong> Maxwell equations. In the process we derive<br />

the field strength tensor, which is a fundamental object in electromagnetism,<br />

that will be used later in this paper. With the field strength tensor we express<br />

Maxwell’s equations as <strong>covariant</strong> 1 tensor equations. We will use Lorentz<br />

notation with the proper four-vector x µ = (ct, x, y, z). This notation together<br />

with the Lorentz scalar product, Equation (2.11), is <strong>of</strong>ten called Lorentz space<br />

(Barut [1]).<br />

2.1.1 Basic relativity and tensor operations<br />

In relativity one considers the concept <strong>of</strong> Lorentz invariance. The Lorentz transformations<br />

are the transformations that keep the expression<br />

c 2 ∆t 2 − ∆x 2 − ∆y 2 − ∆z 2 , (2.1)<br />

invariant. The ∆ refers to a difference between two vectors, which gives the<br />

invariance under translations. These transformations form a group, which is<br />

usually divided into different parts.<br />

The transformations satisfying Equation (2.1) include translations and form<br />

the inhomogeneous Lorentz group, which is also called the Poincaré group (see<br />

Barut [1] or Tung [24]). The transformations which keep the expression<br />

c 2 t 2 − x 2 − y 2 − z 2 , (2.2)<br />

invariant form the inhomogeneous Lorentz group and does not include translations.<br />

All homogeneous Lorentz transformations, denoted by L, have det L =<br />

±1, see Barut [1]. Transformations with det L = 1 are the proper Lorentz transformations<br />

and the transformations with det L = −1 are the improper Lorentz<br />

transformations, which include reflections. The proper Lorentz transformations<br />

1 The covariance refers to a specific group, in this case the Lorentz group (see Section<br />

2.4). Lorentz <strong>covariant</strong> quantities and equations are <strong>of</strong>ten referred to as proper quantities and<br />

equations.<br />

3


4 CHAPTER 2. BACKGROUND<br />

form a subgroup, the improper transformations does not. 2 In the literature,<br />

the proper homogeneous Lorentz group is <strong>of</strong>ten referred to as just the Lorentz<br />

group. A proper four-vector is a vector that transforms like x µ = (ct, x, y, z), µ =<br />

0, 1, 2, 3, under Lorentz transformations.<br />

A space similar to Lorentz space is the Minkowski space. In Minkowski<br />

space, the time coordinate is imaginary and a four-vector is written as x µ =<br />

(x, y, z, ict), where µ = 1, 2, 3, 4. The Lorentz transformations are in this space<br />

orthogonal transformations that preserves the ordinary Cartesian scalarproduct.<br />

The reason for not using the Minkowski space is that it is very complicated to<br />

add a curvature to this space.<br />

Considering Lorentz space, every object is a tensor. A scalar is a rank zero<br />

tensor, a vector is a tensor <strong>of</strong> rank one and so on. When dealing with tensors<br />

there are a number <strong>of</strong> tools and operations at our disposal. These are called<br />

permissible tensor operation since the results are new tensor components. The<br />

operations are (Schutz [21]) 3 :<br />

1. Linear combinations <strong>of</strong> tensors <strong>of</strong> the same type, meaning that the tensors<br />

have the same number <strong>of</strong> <strong>covariant</strong> (subscripts) indices and same number<br />

<strong>of</strong> contravariant (superscripts) indices.<br />

2. Multiplication <strong>of</strong> components <strong>of</strong> two tensors (the tensor product 4 ) gives<br />

a new tensor <strong>of</strong> type equal to the sum <strong>of</strong> the types.<br />

3. Contraction on a pair <strong>of</strong> indices. (only between <strong>covariant</strong> (lower) and<br />

contravariant (upper) indices)<br />

4. Covariant differentiation, see Dirac [8] or Schutz [21], defined for a <strong>covariant</strong><br />

vector as (using Einstein sum convention 5 )<br />

and for a contravariant vector as<br />

A µ;ν ≡ ∂A µ<br />

∂x ν − Γα µνA α , (2.3)<br />

A µ ;ν ≡ ∂Aµ<br />

∂x ν + Γµ ανA α . (2.4)<br />

For tensors <strong>of</strong> higher rank, a Γ term appears for each index, with a minus<br />

sign for a <strong>covariant</strong> index and a plus sign for a contravariant index. The<br />

<strong>covariant</strong> derivative <strong>of</strong> a rank two tensor will be<br />

T µ ν;σ ≡ ∂T µ ν<br />

∂x σ + Γµ ασT α ν − Γ α νσT µ α . (2.5)<br />

The Γ’s are called the Christ<strong>of</strong>fel symbols and are defined later in this<br />

section.<br />

2 The group theory <strong>of</strong> the Lorentz groups are further discussed in Section 2.4<br />

3 For further information on tensors and tensor algebra see e.g. Schutz [21], Arfken &<br />

Weber [10], Barut [1] [18] or Misner, Thorne & Wheeler [7].<br />

4 Also called the outer product or the direct product, but these terms are sometimes used<br />

for other types <strong>of</strong> products.<br />

5 The Einstein sum convention is that in a term you sum over repeated indices. Note that<br />

in general relativity you can only sum over a <strong>covariant</strong> index and a contravariant index, i.e.,<br />

contraction.


2.1. COVARIANT FORMULATION OF MAXWELL EQUATIONS 5<br />

The metric, g µν , is a symmetric tensor, which describes the curvature <strong>of</strong> space,<br />

and is used to ’raise’ and ’lower’ indices. This metric is constant under <strong>covariant</strong><br />

differentiation. The metric we use is the flat metric 6<br />

⎛<br />

⎞<br />

1 0 0 0<br />

g µν = g µν = ⎜0 −1 0 0<br />

⎟<br />

⎝0 0 −1 0 ⎠ . (2.6)<br />

0 0 0 −1<br />

We are also allowed to use the four dimensional Levi-Civita symbol or the<br />

totally anti-symmetric tensor 7 (see e.g. Goldstein [11] or Misner, Thorne &<br />

Wheeler [7]). With the flat metric, g µν , it is defined by<br />

ɛ αβγδ = −ɛ αβγδ<br />

ɛ 0123 = 1<br />

ɛ αβγδ = [αβγδ] =<br />

{<br />

1 αβγδ even permutation <strong>of</strong> 0,1,2,3,<br />

−1 αβγδ odd permutation <strong>of</strong> 0,1,2,3,<br />

(2.7)<br />

where [αβγδ] is the permutation symbol. Generalizing the Levi-Civita tensor<br />

to an arbitrary metric it is defined as<br />

ɛ αβγδ = √ −g[αβγδ], (2.8)<br />

where<br />

g = det g µν . (2.9)<br />

The contravariant and mixed Levi-Civita tensors are then formed by raising the<br />

indices. Note that the sign convention <strong>of</strong> ɛ 0123 varies and some authors define<br />

it as -1, which just changes the sign <strong>of</strong> ɛ αβγδ and ɛ αβγδ .<br />

The differential invariant distance can, in our flat space be written<br />

ds 2 = g µν dx µ dx ν = c 2 dt 2 − dx 2 − dy 2 − dz 2 = c 2 dτ 2 . (2.10)<br />

In the last step here we have introduced the proper time, τ, a term used in<br />

relativity and it is the time experienced by the particle in question and not<br />

by an observer <strong>of</strong> the particle. Note that the coordinate time, t, is the time<br />

experienced by the observer defining the coordinate system and also defining<br />

the metric.<br />

The scalar product in Lorentz space is defined according to contraction. The<br />

procedure is the following: a vector a µ , multiplied with a vector b µ , gives a rank<br />

two tensor, which via contraction gives a scalar. This is also the <strong>covariant</strong> trace<br />

<strong>of</strong> a tensor, the formula is<br />

a µ b ν = c µν , g µν c µν = c µ µ = a 0 b 0 − a 1 b 1 − a 2 b 2 − a 3 b 3 . (2.11)<br />

6 The strong equivalence principle (see Schutz [21]) states that ’any physical law which can<br />

be expressed in tensor notation in special relativity has exactly the same form in a locally<br />

inertial frame <strong>of</strong> curved space-time’, i.e. if we find a tensor equation using the flat metric, it<br />

will be the same in a curved space time but with another metric.<br />

7 The totally anti-symmetric tensor will be further discussed in Section 2.4, when the group<br />

theory <strong>of</strong> the Lorentz group is considered.


6 CHAPTER 2. BACKGROUND<br />

For a tensor <strong>of</strong> rank two, the trace is defined as contraction <strong>of</strong> the indices, which<br />

is a permissible tensor operation. In contrast, the ordinary trace is not a permissible<br />

operation. The four-gradients, ∂ µ (contravariant) and ∂ µ (<strong>covariant</strong>),<br />

are defined as<br />

∂ µ ≡<br />

∂<br />

∂x µ = ( ∂<br />

∂ct , ∂<br />

∂x , ∂ ∂y , ∂ ). (2.12)<br />

∂z<br />

An <strong>of</strong>ten used notation is ∂ µ A = A, µ (compare with the <strong>covariant</strong> derivative,<br />

A ;µ ).<br />

In <strong>covariant</strong> differentiation, Equation (2.3), the Γ α µν comes in to assure covariance.<br />

These Γ’s are called Christ<strong>of</strong>fel symbols and are defined as<br />

Γ γ βµ = 1 2 gαγ (g αβ,µ + g αµ,β − g µβ,α ). (2.13)<br />

In flat space, these Christ<strong>of</strong>fel symbols vanish everywhere and from Equation<br />

(2.3) we have<br />

A ,µ = A ;µ . (2.14)<br />

To find a <strong>covariant</strong> form <strong>of</strong> an equation one needs to find proper four-vectors<br />

and (or) proper tensors that, by performing the permissible operations on them,<br />

leads to the equation in question. This is what we will do with the Maxwell<br />

equations in the next section.<br />

2.1.2 Microscopic Maxwell equations<br />

In vacuum the Maxwell equations in the general differential form do not include<br />

the ⃗ H and ⃗ D fields. These are the so-called microscopic Maxwell equations (see<br />

Arfken & Weber [10], Wangsness [25] or Jackson [12]) and have the following<br />

form<br />

∇ · ⃗E = ρ ε 0<br />

, (2.15)<br />

∇ · ⃗B = 0, (2.16)<br />

∇ × ⃗ E = − ∂ ∂t ⃗ B, (2.17)<br />

∇ × ⃗ B = µ 0<br />

⃗j + 1 c 2 ∂<br />

∂t ⃗ E, (2.18)<br />

where ⃗j is the current density and ρ is the charge density. It is now convenient<br />

to use Lorentz gauge, which is the condition<br />

∇ · ⃗A + ɛµ ∂φ<br />

∂t<br />

+ µσφ = 0, (2.19)<br />

where σ is the conductivity, ɛ is the permittivity and µ is the permeability.<br />

In vacuum (σ = 0, ɛ = ɛ 0 , µ = µ 0 ), using ɛ 0 µ 0 = 1/c 2 , the Lorentz condition<br />

reduces to<br />

∇ · ⃗A + 1 ∂φ<br />

c 2 = 0. (2.20)<br />

∂t<br />

Here we have introduced the electric potential, φ, and the magnetic vector potential,<br />

A. ⃗ The electric and magnetic field can be calculated from the potentials<br />

as<br />

⃗B = ∇ × A, ⃗ (2.21)


2.1. COVARIANT FORMULATION OF MAXWELL EQUATIONS 7<br />

⃗E = −∇ · φ − 1 ∂A<br />

⃗<br />

c ∂t . (2.22)<br />

By inserting Equation (2.21) and Equation (2.22) into the inhomogeneous Maxwell<br />

equations (Eq. (2.15) and Eq. (2.18)) and using the Lorentz condition for vacuum<br />

(Eq. (2.20)) we get two separated equations for the potentials,<br />

for the magnetic vector potential, and<br />

∇ 2 ⃗ A −<br />

1<br />

c 2 ∂ 2 ⃗ A<br />

∂ 2 t = −µ 0 ⃗ j, (2.23)<br />

∇ 2 φ − 1 c 2 ∂ 2 φ<br />

∂ 2 t = − ρ ɛ 0<br />

, (2.24)<br />

for the electric scalar potential. The differential operator in the above equations<br />

is known as the d’Alembertian (with a minus sign) and is a fundamental operator<br />

in electrodynamics and is known to be invariant (see e.g. Wagnsness [25], Arfken<br />

& Weber [10] or Jackson [12]). Using four-vector notation it it can be written:<br />

□ 2 = 1 c 2 ∂ 2<br />

∂ 2 t − ∇2 = ∂ 2 = ∂ µ ∂ µ . (2.25)<br />

From Equation (2.23) and Equation (2.24) we get the following definitions <strong>of</strong><br />

four-vectors,<br />

A µ = (φ, c ⃗ A), (2.26)<br />

J µ = (ρ,⃗j/c), (2.27)<br />

which are the four-potential and the four-current, respectively. We insert the c<br />

to get the correct dimension. 8 Then we can write Equations (2.23) and (2.24)<br />

together as a four-vector differential equation<br />

∂ 2 A µ = µ 0 cJ µ . (2.28)<br />

To see that this is a tensor equation, i.e. showing that A µ and J µ are proper<br />

four-vectors, it suffices to show that J µ is a proper four-vector. This is because<br />

the d’Alembertian is an invariant scalar (see e.g. Jackson [12] or Barut [1]).<br />

Then, by the quotient rule (see Schutz [21], Arfken & Weber [10]), A µ is a<br />

four-vector if J µ is a four-vector. Since an electric charge element de = ρd 3 x is<br />

invariant we compare this with the four-dimensional volume element d 4 x. We<br />

see that ρ must transform the same way as dx 0 . Consider J 1 , we have<br />

J 1 = ρv x = ρ dx1<br />

dt<br />

= cρ dx1<br />

d(ct) = J 0 dx1<br />

dx 0 . (2.29)<br />

Since J 0 transforms as dx 0 , the equation above shows that J 1 transforms as dx 1<br />

and thus J 2 and J 3 transform as dx 2 and dx 3 , respectively. So, Equation (2.28)<br />

is a proper tensor equation and A µ and J µ are proper four-vectors. However, A µ<br />

is not a measurable quantity and it also depends on the gauge we have chosen.<br />

What we would like to have is a <strong>covariant</strong> expression that involves ⃗ E and ⃗ B<br />

since these are the quantities we measure and they are gauge invariant, which<br />

is not the case with the potential.<br />

8 The potentials can be defined differently and different units can be used, and thus gives<br />

a different field strength tensor, see e.g. Jackson [12] or Wangsness [25].


8 CHAPTER 2. BACKGROUND<br />

2.1.3 Construction <strong>of</strong> the <strong>electromagnetic</strong> field strength<br />

tensor<br />

The relation between the potentials and the electric and magnetic fields are<br />

given by in Equations (2.21) and (2.22), so we first rewrite these equations<br />

using the four-vector potential, A µ . Considering the i:th component <strong>of</strong> the field<br />

we have<br />

⃗B = ∇ × ⃗ A ⇒ cB i = ∂Ak<br />

∂x j − ∂Aj<br />

∂x k = ∂j A k − ∂ k A j (i, j, k) = (1, 2, 3), (2.30)<br />

and<br />

⃗E = −∇φ − ∂ A ⃗<br />

∂t ⇒ E i = ∂A0<br />

∂x i − ∂Ai<br />

∂x 0 = ∂i A 0 − ∂ 0 A i (i = 1, 2, 3). (2.31)<br />

Equations (2.30) and (2.31), are proper tensor equations since ∂ µ and A µ are<br />

proper four-vectors. Thus we have the equations on the form we want and<br />

proceed by defining an antisymmetric rank two tensor, F µν , with elements given<br />

by the equations above (Eqs. (2.30) and (2.31)). The tensor is found to be<br />

⎛<br />

⎞<br />

0 −E 1 −E 2 −E 3<br />

F µν ≡ ∂ µ A ν − ∂ ν A µ = ⎜ E 1 0 −cB 3 cB 2<br />

⎟<br />

⎝ E 2 cB 3 0 −cB 1<br />

⎠ . (2.32)<br />

E 3 −cB 2 cB 1 0<br />

This is the field strength tensor which is not uniquely defined. The form <strong>of</strong><br />

the components depends on the choice <strong>of</strong> space (Lorentz or Minkowski space)<br />

and the choice <strong>of</strong> units. 9 Note that an invariant four-vector multiplied with a<br />

constant is again an invariant four-vector. Next step is to rewrite the Maxwell<br />

equations in terms <strong>of</strong> the field strength tensor, but first we define the dual tensor<br />

<strong>of</strong> F µν , in the following way<br />

⎛<br />

⎞<br />

0 −cB 1 −cB 2 −cB 3<br />

F αβ = 1 2 ɛαβγδ F γδ = 1 2 ɛαβγδ g γµ g δν F µν = ⎜ cB 1 0 E 3 −E 2<br />

⎟<br />

⎝ cB 2 −E 3 0 E 1<br />

⎠ .<br />

cB 3 E 2 −E 1 0<br />

(2.33)<br />

Now using the field strength tensor and its dual we can rewrite the Maxwell<br />

equations in the <strong>covariant</strong> form for the flat metric, g µν from Equation (2.6),<br />

which is valid in special relativity. We have the continuity equation<br />

∂ µ J µ = ∂J µ<br />

= 0, (2.34)<br />

∂x<br />

µ<br />

the inhomogeneous Maxwell equations (Eqs. (2.15) and (2.18))<br />

∂ µ F µν = µ 0 cJ ν , (2.35)<br />

and the homogeneous Maxwell equations (Eqs. (2.16) and (2.17))<br />

∂ µ F µν = 0. (2.36)<br />

9 For some other variants <strong>of</strong> the field strength tensor see Barut [1] or Wangsness [25].


2.1. COVARIANT FORMULATION OF MAXWELL EQUATIONS 9<br />

To generalize these equations to an arbitrary metric we need to replace the<br />

derivative with the <strong>covariant</strong> derivative (see Eq. (2.3)). We the obtain a manifestly<br />

<strong>covariant</strong> form <strong>of</strong> Maxwell’s equations. The continuity equation<br />

the inhomogeneous Maxwell equations<br />

and the homogeneous Maxwell equations<br />

J µ ;µ = 0, (2.37)<br />

F µν<br />

;µ = µ 0 cJ ν , (2.38)<br />

F µν<br />

;µ = 0. (2.39)<br />

According to the strong equivalence principle (SEP), see Schutz [21], we can<br />

replace the four-gradient, ∂ µ , with the <strong>covariant</strong> derivative if we have the equations<br />

as proper tensor equations in special relativity, i.e. using the flat metric<br />

g µν . In the form above (Eqs. (2.34), (2.35) and (2.36)), Maxwell equations are<br />

proper tensor equations and thus we can just replace the ordinary derivative<br />

with the <strong>covariant</strong> derivative.<br />

Invariant quadratic quantities from the field strength tensor<br />

Using the permissible tensor operations on the field strength tensor tensor and<br />

its dual, it is possible to construct invariant scalars. This is done by contracting<br />

the tensor product <strong>of</strong> the field strength tensor with itself or its dual tensor.<br />

From this we get the invariant scalars<br />

and<br />

F µν F ∗ µν = −2c ⃗ E · ⃗B † − 2c ⃗ B · ⃗E † = −4cRe[ ⃗ E · ⃗B † ] (2.40)<br />

F µν F ∗ µν = 2(c 2 ⃗ B 2 − ⃗ E 2 ). (2.41)<br />

The first invariant is the real part <strong>of</strong> the scalar product <strong>of</strong> ⃗ E and ⃗ B, and the<br />

second is recognised as the Lagrangian, which corresponds to the free energy <strong>of</strong><br />

an <strong>electromagnetic</strong> <strong>wave</strong> in vacuum.<br />

2.1.4 Electromagnetic energy-momentum tensor<br />

This is a good time to introduce another proper rank two tensor , the <strong>electromagnetic</strong><br />

energy-momentum tensor 10 . The fields are real valued and we first<br />

define the energy density E and the Poynting vector ⃗ S to be<br />

E = 1 2 ( ⃗ E 2 + c 2 ⃗ B 2 ), (2.42)<br />

⃗S = ⃗ E × c ⃗ B. (2.43)<br />

In terms <strong>of</strong> the field strength tensor, the <strong>electromagnetic</strong> energy-momentum<br />

tensor can be defined as (Dirac [8])<br />

T µν = −F µ α F να + 1 4 gµν F αβ F αβ . (2.44)<br />

10 Also called the stress-energy tensor or the canonical stress tensor, and is a fundamental<br />

object in general relativity (see Schutz [21], Dirac [8], Barut [1], Jackson [12] or Misner, Thorne<br />

& Wheeler [7]).


10 CHAPTER 2. BACKGROUND<br />

With the energy density, E, and the Poynting vector, S, the energy-momentum<br />

tensor can be written as<br />

⎛<br />

⎞<br />

E S 1 S 2 S 3<br />

T µν = ⎜S 1 −E1 2 − c 2 B1 2 + E −E 1 E 2 − c 2 B 1 B 2 −E 1 E 3 − c 2 B 1 B 3<br />

⎟<br />

⎝S 2 −E 2 E 1 − c 2 B 2 B 1 −E2 2 − c 2 B2 2 + E −E 2 E 3 − c 2 B 2 B 3<br />

⎠ .<br />

S 3 −E 3 E 1 − c 2 B 3 B 1 E 3 E 1 − c 2 B 3 B 1 −E3 2 − c 2 B3 2 + E<br />

(2.45)<br />

The components T mn (m, n = 1, 2, 3) together form the Maxwell stress tensor<br />

(see Section 2.3 Equation (2.95)).


2.2. GENERALIZED STOKES PARAMETERS 11<br />

2.2 Generalized Stokes parameters<br />

In this Section we give the definitions <strong>of</strong> the two-dimensional Stokes parameters<br />

and the generalized <strong>polarization</strong> parameters introduced by T. Carozzi, R. Karlsson<br />

and J. Bergman [6]. These <strong>polarization</strong> parameters are a generalization <strong>of</strong><br />

the Stokes parameters into three-dimensions. A similar set <strong>of</strong> parameters were<br />

introduced by Brosseau [5]. The Poynting vector and the Maxwell stress tensor<br />

are introduced and defined. Storey and Lefeuvre’s [23] six dimensional <strong>electromagnetic</strong><br />

quasi-vector and their representation <strong>of</strong> the quadratic quantities <strong>of</strong> an<br />

<strong>electromagnetic</strong> field are also discussed.<br />

2.2.1 Spectral density matrix<br />

Consider an arbitrary vector field, ⃗ f(⃗r, t). The field is a superposition <strong>of</strong> all<br />

<strong>wave</strong>s and stationary fields at the point ⃗r. The spectral components <strong>of</strong> the field<br />

⃗f(⃗r, t), one obtains by taking the Fourier transform, ⃗ F (⃗r, ω), <strong>of</strong> the field, defined<br />

as<br />

⃗F (⃗r, ω) =<br />

∫ ∞<br />

−∞<br />

⃗f(⃗r, t)e iωt dt. (2.46)<br />

The vector, ⃗ F , represents a complex valued vector field which can be separated<br />

into three complex amplitudes (the x, y, z components), or into three real<br />

amplitudes (F 1 , F 2 , F 3 ) and their real phases (δ 1 , δ 2 , δ 3 ), in the following way<br />

⎛<br />

⃗F (⃗r, ω) = ⎝ F ⎞ ⎛<br />

⎞<br />

x(⃗r, ω) F 1 (⃗r, ω)e δ1(⃗r,ω)<br />

F y (⃗r, ω) ⎠ = ⎝F 2 (⃗r, ω)e δ2(⃗r,ω) ⎠ . (2.47)<br />

F z (⃗r, ω) F 3 (⃗r, ω)e δ3(⃗r,ω)<br />

The spectral density matrix, S d (⃗r, ω), is constructed by taking the tensor product<br />

between F ⃗ = F ⃗ (⃗r, ω), and its Hermitian conjugate, F ⃗ † = F ⃗ † (⃗r, ω). This<br />

gives<br />

⎛<br />

⎞<br />

S d (⃗r, ω) = F ⃗ ⊗ F ⃗ F x F † x ∗ F x Fy ∗ F x Fz<br />

∗<br />

= ⎝F y Fx ∗ F y Fy ∗ F y Fz<br />

∗ ⎠ , (2.48)<br />

F z Fx ∗ F z Fy ∗ F z Fz<br />

∗<br />

where F ∗ denotes complex conjugate. We will assume (⃗r, ω)-dependence and<br />

the notation S d = S d (⃗r, ω) will be used in the following.<br />

2.2.2 Instantaneous Stokes’ parameters<br />

Stokes’ parameters describe the <strong>polarization</strong> state <strong>of</strong> a transverse <strong>electromagnetic</strong><br />

<strong>wave</strong> in the <strong>polarization</strong> plane. One usually assumes that the <strong>wave</strong> propagates<br />

in the z-direction and the <strong>polarization</strong> ellipse lies in the xy-plane (see<br />

Brosseau [4]). Stokes parameters can be written (using the notation 11 <strong>of</strong> R.<br />

11 The notation for Stokes’ parameters is not universal. The notation <strong>of</strong> Stokes [22] was<br />

(A,B,C,D), another notation is (S 0 , S 1 , S 2 , S 3 ) used by e.g. Jackson [12] and Brosseau [4].


12 CHAPTER 2. BACKGROUND<br />

Karlsson [13], Born and Wolf [3])<br />

I = F 2 x + F 2 y , (2.49)<br />

Q = F 2 x − F 2 y , (2.50)<br />

U = 2Re[F y F ∗ x ] = F x F ∗ y + F y F ∗ x , (2.51)<br />

V = 2Im[F y F ∗ x ] = i ( F x F ∗ y − F y F ∗ x<br />

)<br />

, (2.52)<br />

where I denotes the intensity <strong>of</strong> the field, Q and U describe the linear <strong>polarization</strong><br />

and V is related to circular <strong>polarization</strong>. 12 In the two-dimensional case the<br />

spectral density tensor, Equation (2.48), is reduced to<br />

( )<br />

Fx Fx S d = ∗ F x Fy<br />

∗<br />

F y Fx<br />

∗ F y Fy<br />

∗ . (2.53)<br />

The two-dimensional spectral density matrix can be decomposed using the unit<br />

matrix, I 2 , and the generators <strong>of</strong> the special unitary symmetry group, SU(2),<br />

which are recognized as the Pauli spin matrices, σ i , i = 1, 2, 3. These can be<br />

written<br />

σ 1 =<br />

( 0 1<br />

1 0<br />

) ( 0 −i<br />

, σ 2 =<br />

i 0<br />

) ( 1 0<br />

, σ 3 =<br />

0 −1<br />

) ( 1 0<br />

, I 2 =<br />

0 1<br />

(2.54)<br />

Writing the spectral density matrix, S d , as a linear combination <strong>of</strong> I 2 and the<br />

Pauli spin matrices, Equation (2.54), one identifies the coefficients as the Stokes<br />

parameters, I, Q, U and V. The spectral density matrix, S d , can then be written<br />

in the following form<br />

S d = 1 )<br />

(II 2 + Uσ 1 + Vσ 2 + Qσ 3 = 1 ( )<br />

I + Q U − iV<br />

. (2.55)<br />

2<br />

2 U + iV I − Q<br />

This is a convenient way to express the spectral density matrix, S d , since the<br />

Stokes parameters are just scalars and the Pauli spin matrices are unitary with<br />

determinant equal to ±1. This procedure can be generalized into three dimensions,<br />

and leads to a similar relation.<br />

For a deterministic <strong>wave</strong> we have<br />

)<br />

.<br />

I 2 = U 2 + Q 2 + V 2 , (2.56)<br />

which means that the <strong>wave</strong> is totally polarized and hence one <strong>of</strong> the parameters<br />

is abundant for a complete description <strong>of</strong> the state <strong>of</strong> <strong>polarization</strong>. In reality<br />

we can not have a deterministic <strong>wave</strong> and must therefore integrate over a small<br />

interval (bandwidth). The relation above then becomes<br />

I 2 ≥ U 2 + Q 2 + V 2 . (2.57)<br />

This means that for a deterministic (or a fully polarized <strong>wave</strong>) <strong>wave</strong> we do<br />

not need all the parameters to describe the <strong>polarization</strong> <strong>of</strong> the <strong>wave</strong>, but for a<br />

partially polarized <strong>wave</strong> all four parameters are needed.<br />

12 For a longer discussion on Stokes parameters see Brosseau [4] or Born & Wolf [3].


2.2. GENERALIZED STOKES PARAMETERS 13<br />

2.2.3 Three-dimensional <strong>polarization</strong> parameters<br />

We want to obtain a more general form <strong>of</strong> Equation (2.53) applicable to an<br />

arbitrary <strong>wave</strong> in three dimensions (to use Stokes parameters it is necessary to<br />

know the direction <strong>of</strong> propagation). To achieve this we expand the full spectral<br />

density matrix, S d , using the unit matrix, I 3 , and the generators <strong>of</strong> the SU(3)<br />

symmetry group, λ i , (i = 1, . . . , 8), as given by Gell-Mann and Ne’eman [17]<br />

(see Appendix A Equation (A.2)). The spectral density matrix is then formed<br />

as a linear combination <strong>of</strong> these matrices in the following way<br />

S d = 1 3 Λ 0I 3 + 1 8∑<br />

Λ i λ i =<br />

2<br />

i=1<br />

⎛<br />

1<br />

3<br />

⎜<br />

Λ 0 + 1 2 Λ 3 + 1<br />

1<br />

⎝ 2 Λ 1 + i 2 Λ 2<br />

1<br />

2 Λ 4 + i 2 Λ 5<br />

2 √ 3 Λ 8<br />

1<br />

2 Λ 1 − i 2 Λ 2<br />

1<br />

3 Λ 0 − 1 2 Λ 3 + 1<br />

1<br />

2 Λ 6 + i 2 Λ 7<br />

2 √ 3 Λ 8<br />

1<br />

2 Λ 4 − i 2 Λ ⎞<br />

5<br />

1<br />

2 Λ 6 − i 2 Λ ⎟<br />

7 ⎠ , (2.58)<br />

1<br />

3 Λ 0 − √ 1<br />

3<br />

Λ 8<br />

where Λ i , i = 0, 1, . . . 8, are the generalized Stokes parameters. The trace <strong>of</strong><br />

this matrix, S d , is Λ 0 which is the energy density <strong>of</strong> the field. For the two<br />

dimensional case this is the Stokes parameter I. The Λ i are found by comparing<br />

Equation (2.48) with Equation (2.58), the explicit expressions for the Λ i are<br />

found in Appendix A. In the case <strong>of</strong> a transverse <strong>wave</strong> propagating in the z-<br />

direction, i.e. F z = 0, the spectral density matrix, Equation (2.58), reduces to<br />

two-dimensional form<br />

⎛<br />

⎞<br />

S d = 1 Λ 0 + Λ 3 Λ 1 − iΛ 2 0<br />

⎝Λ 1 + iΛ 2 Λ 0 − Λ 3 0⎠ , (2.59)<br />

2<br />

0 0 0<br />

where the Λ i have reduced to the Stokes parameters as in Equation (2.53). This<br />

is just a rotation <strong>of</strong> the coordinate system into the one <strong>of</strong> the <strong>wave</strong> with the<br />

z-axis being the direction <strong>of</strong> propagation.<br />

There are different normalizations <strong>of</strong> the generalized <strong>polarization</strong> parameters.<br />

Brosseau [5] uses 1 3 instead <strong>of</strong> 1 2<br />

, which we have used, in front <strong>of</strong> the SU(3)<br />

generators in Equation (2.58). If one uses 1 2<br />

, Equation (2.59) (transverse <strong>wave</strong><br />

propagating in the z-direction) reduces to Stokes parameters, Equation (2.55),<br />

which is not the case if one uses 1 3 .<br />

2.2.4 Interpretation <strong>of</strong> the 3D <strong>polarization</strong> parameters<br />

Every tensor can be decomposed into its symmetric and anti-symmetric parts,<br />

Arfken & Weber [10]. Thus, Equation (2.58) can be written<br />

S d = S S d + S A d = 1 2(<br />

Sd + S T d<br />

) 1( + Sd − S T )<br />

d . (2.60)<br />

2<br />

The superscript T denotes a transposed tensor. From Equation (2.58) it is easily<br />

seen that the symmetric part, Sd S, and anti-symmetric part, SA d<br />

, are given by<br />

⎛<br />

1<br />

3<br />

Sd S ⎜<br />

Λ 0 + 1 2 Λ 3 + 1<br />

2 √ Λ 1<br />

3<br />

8<br />

2 Λ 1<br />

1<br />

2 Λ ⎞<br />

4<br />

1<br />

= ⎝<br />

2 Λ 1<br />

1<br />

3 Λ 0 − 1 2 Λ 3 + 1<br />

2 √ Λ 1<br />

3<br />

8 2 Λ ⎟<br />

6 ⎠ (2.61)<br />

1<br />

2 Λ 1<br />

4<br />

2 Λ 1<br />

6<br />

3 Λ 0 − √ 1<br />

3<br />

Λ 8


14 CHAPTER 2. BACKGROUND<br />

and<br />

⎛<br />

0 − i<br />

Sd A 2 Λ 2 − i 2 Λ ⎞<br />

5<br />

= ⎝ i<br />

2 Λ 2 0 − i 2 Λ ⎠<br />

7 , (2.62)<br />

i<br />

2 Λ i<br />

5 2 Λ 7 0<br />

respectively. To any anti-symmetric rank two tensor, in three dimensions 13 a<br />

dual pseudo-vector is associated (see Arfken & Weber [10]). This dual pseudovector<br />

is defined by<br />

V i = 1 2 ɛ ijkT jk , (2.63)<br />

where ɛ ijk is the totally anti-symmetric tensor, also known as the Levi-Civita<br />

tensor. 14 The dual pseudo vector to Sd<br />

A is then<br />

⃗V ′ = i(−Λ 7 , Λ 5 , −Λ 2 ). (2.64)<br />

The totally anti-symmetric tensor, ɛ ijk , contracted with a vector <strong>of</strong> ones, (1,1,1)<br />

becomes<br />

⎛<br />

0 1<br />

⎞<br />

−1<br />

ɛ ijk (1, 1, 1) k = ⎝−1 0 1 ⎠ . (2.65)<br />

1 −1 0<br />

We see that it has a different sign convention compared to λ 7 and λ 2 and<br />

therefore we get the minus sign on them (see Appendix A). This vector ⃗ V ′ is<br />

an imaginary vector and for later use we would like to have a real vector so<br />

following Carozzi, Karlsson and Bergman [6] we therefore define a vector ⃗ V by<br />

which is real. The components <strong>of</strong> ⃗ V are<br />

⃗V ≡ (Λ 7 , −Λ 5 , Λ 2 ). (2.66)<br />

Λ 7 = i(F y F ∗ z − F z F ∗ y ) = 2Im(F z F ∗ y ) = −2Im(F y F ∗ z ),<br />

Λ 5 = i(F x F ∗ z − F z F ∗ x ) = 2Im(F z F ∗ x ) = −2Im(F x F ∗ z ),<br />

Λ 2 = i(F x F ∗ y − F y F ∗ x ) = 2Im(F y F ∗ x ) = −2Im(F x F ∗ y ).<br />

(2.67a)<br />

(2.67b)<br />

(2.67c)<br />

Recalling that ⃗ V is a pseudo-vector and that a crossproduct between two polar<br />

(ordinary) vectors or two axial (pseudo) vectors is a pseudo-vector, we can from<br />

the form <strong>of</strong> the components conclude that ⃗ V can be defined by<br />

⃗V = i ⃗ F × ⃗ F † . (2.68)<br />

The complex vector F ⃗ and its complex conjugate F ⃗ † form a <strong>polarization</strong> plane<br />

in space, see e.g. Lindell [14] or Karlsson [13], and because V ⃗ = iF ⃗ × F ⃗ † it is<br />

perpendicular to both F ⃗ and F ⃗ † and thus parallel or anti-parallel to the normal<br />

<strong>of</strong> the <strong>polarization</strong> plane (see Karlsson [13] or Lindell [14]). Λ 2 , Λ 5 and Λ 7 are<br />

the three-dimensional generalization <strong>of</strong> circular <strong>polarization</strong>. Comparing the<br />

13 In four dimensions a dual second rank tensor is associated to an anti-symmetric second<br />

rank tensor (see Sections 2.1 and 2.4).<br />

14 The sign <strong>of</strong> the metric in the three-dimensional Euclidean space (the ordinary threedimensional<br />

space) is (1, 1, 1). This gives that the totally anti-symmetric tensor <strong>of</strong> this space<br />

is equal to the permutation symbol and contraction between indices <strong>of</strong> the same kind is allowed<br />

(compare with Sections 2.1 and 2.4)


2.2. GENERALIZED STOKES PARAMETERS 15<br />

expressions for the Λ i ’s, Equation (2.67), with Stokes’ parameter for circular<br />

<strong>polarization</strong>, V, we have the relation (see Carozzi, Karlsson and Bergman [6])<br />

|V| =<br />

√<br />

Λ 2 7 + Λ2 5 + Λ2 2 = |⃗ V |. (2.69)<br />

We define left- and right-handed <strong>polarization</strong> as seen by the <strong>wave</strong> (opposite to<br />

an observer looking into the <strong>wave</strong>). For a right-handed polarized <strong>wave</strong>, V ⃗ will<br />

be parallel to the direction <strong>of</strong> propagation and anti-parallel for a left-handed<br />

polarized <strong>wave</strong>. 15 Because measurements is generally made in one point, i.e.<br />

we have the field components in one point, we really have two possibilities with<br />

opposite signs (opposite direction <strong>of</strong> propagation). To get a correct sign on V ⃗<br />

one needs further information about the direction <strong>of</strong> propagation <strong>of</strong> the <strong>wave</strong>.<br />

2.2.5 The Storey and Lefeuvre six-quasivector<br />

In order to obtain a straightforward description <strong>of</strong> the <strong>electromagnetic</strong> field with<br />

all possible combinations <strong>of</strong> the <strong>electromagnetic</strong> field components, Storey and<br />

Lefeuvre introduced a six dimensional quasivector, for the field components,<br />

defined as<br />

⃗E = (E 1 , E 2 , E 3 , cB 1 , cB 2 , cB 3 ). (2.70)<br />

For a continuum <strong>of</strong> plane <strong>wave</strong>s Storey and Lefeuvre introduces a generalized<br />

electric field, e m i (⃗r, t), where the i-th component is the real part <strong>of</strong> the evolution<br />

e m i (⃗r, t) = Re[e i exp i(ωt − ⃗ k · ⃗r)]. (2.71)<br />

Here e i is the complex amplitude. The 6 × 6 matrix they define is constructed<br />

as<br />

a ij = e ie ∗ j<br />

δ . (2.72)<br />

These are constant since the complex amplitude is constant for the plane <strong>wave</strong><br />

in Equation (2.71). The six-vector, Equation (2.70), is not a vector, in the<br />

geometrical meaning, more than in the sense that it has six entries. The first<br />

three components, the electric part, together form the E ⃗ vector (polar vector)<br />

and the last three form the B ⃗ vector which is a pseudo-vector (axial vector). So,<br />

the six-vector defined here has three components transforming as a polar vector<br />

and three components transforming as a pseudo-vector. 16 Anyhow, using this<br />

six-vector we obtain a spectral density matrix<br />

⎛<br />

⎞<br />

E1 2 E 1 E2 ∗ E 1 E3 ∗ E 1 B1 ∗ E 1 B2 ∗ E 1 B3<br />

∗ E 2 E1 ∗ E2 2 E 2 E3 ∗ E 2 B1 ∗ E 2 B2 ∗ E 2 B ∗ 3<br />

S d =<br />

E 3 E1 ∗ E 3 E2 ∗ E3 2 E 3 B1 ∗ E 3 B2 ∗ E 3 B3<br />

∗ ⎜B 1 E1 ∗ B 1 E2 ∗ B 1 E3 ∗ B1 2 B 1 B2 ∗ B 1 B3<br />

∗ . (2.73)<br />

⎟<br />

⎝B 2 E1 ∗ B 2 E2 ∗ B 2 E3 ∗ B 2 B1 ∗ B2 2 B 2 B3<br />

∗ ⎠<br />

B 3 E1 ∗ B 3 E2 ∗ B 3 E3 ∗ B 3 B1 ∗ B 3 B2 ∗ B3<br />

2<br />

As seen from its structure, this matrix is symmetric in the real part and antisymmetric<br />

in the imaginary part. It has 36 different entries <strong>of</strong> which 21 are<br />

15 For more information on the ⃗ V and complex vector analysis see Lindell [14] [15].<br />

16 More information on polar and axial vectors are found in Arfken & Weber [10]


16 CHAPTER 2. BACKGROUND<br />

real and 15 imaginary. Since the six-vector, ⃗ E, is not a geometrical quantity<br />

and do not have the convenient transformation properties <strong>of</strong> a rank one tensor,<br />

this matrix is no more than a table <strong>of</strong> the 36 correlation components. Even so,<br />

these 36 correlation components are those components we want to combine into<br />

geometrical quantities to get a more physical representation <strong>of</strong> <strong>polarization</strong>.


2.3. IRREDUCIBLE REPRESENTATION 17<br />

2.3 Irreducible representation in terms <strong>of</strong> geometrical<br />

objects<br />

A quite natural way <strong>of</strong> representing a complex vector field in quadratic components<br />

is to take the tensor product <strong>of</strong> the field and its hermitian conjugate. The<br />

tensor product between any two three-dimensional vectors is defined as<br />

⎛<br />

⃗A ⊗ B ⃗ † = ⎝ A ⎞<br />

1B1 ∗ A 1 B2 ∗ A 1 B3<br />

∗<br />

A 2 B1 ∗ A 2 B2 ∗ A 2 B3<br />

∗ ⎠ , (2.74)<br />

A 3 B1 ∗ A 3 B2 ∗ A 3 B3<br />

∗<br />

where † denotes Hermitian conjugate and ∗ denotes complex conjugate. This<br />

tensor product is the fundamental object in this chapter and we will decompose<br />

it into different geometrical objects. Since it is constructed from the tensor<br />

product <strong>of</strong> two vectors it is also clear that it is a tensor. This matrix is <strong>of</strong>ten<br />

called the correlation matrix. The symmetry used here is that <strong>of</strong> the rotation<br />

group, SO(3) (see Arfken & Weber [10], Tung [24] or Barut & Raczka [18]), since<br />

we are in the complex Euclidean three dimensional space and we are restricting<br />

the rotations to real rotations.<br />

2.3.1 Complex vector algebra<br />

In the spectral domain, the Fourier transform <strong>of</strong> the <strong>electromagnetic</strong> field, we<br />

have a complex vector field. For a good introduction to complex algebra, the<br />

paper[14] and book [15] by Lindell is recommended. The complex vector algebra,<br />

with notations, used in this work is presented here.<br />

The Hermitian conjugate <strong>of</strong> a scalar is the same as the complex conjugate,<br />

and for a vector<br />

⃗F = ⃗a + i ⃗ b ⇒ ⃗ F † = ⃗a − i ⃗ b, (2.75)<br />

it is also the same. The main rule is that in in a multiplication <strong>of</strong> components,<br />

we have the conjugation on the second factor. This gives the scalar product <strong>of</strong><br />

⃗A and ⃗ B, to be<br />

scalar( ⃗ A, ⃗ B) = ⃗ A · ⃗B † , (2.76)<br />

and the cross product<br />

cross( ⃗ A, ⃗ B) = ⃗ A × ⃗ B † . (2.77)<br />

For a matrix or tensor, the Hermitian conjugate is conjugation and transponation,<br />

A † = (A T ) ∗ . In this work we use the products defined in a real space and<br />

write out the conjugate, ⃗ A · ⃗B † .<br />

Following Lindell [14], we have some useful relations for complex vectors.<br />

For determining if a vector is zero, we have<br />

⃗A · ⃗A † = 0 ⇒ ⃗ A = 0, (2.78)<br />

⃗A × ⃗ B = 0, ⃗ A · ⃗ B † = 0 ⇒ ⃗ A = 0 or ⃗ B = 0. (2.79)<br />

If the cross product <strong>of</strong> two vectors is zero, we have the relation<br />

⃗A × ⃗ B = 0, ⃗ A ≠ 0 ⇒ ∃α ∈ C, ⃗ B = α ⃗ A, (2.80)


18 CHAPTER 2. BACKGROUND<br />

and if the scalar product is zero, we have<br />

⃗A · ⃗B = 0, A ≠ 0 ⇒ ∃ ⃗ C, ⃗ B = ⃗ C × ⃗ A. (2.81)<br />

With complex field ⃗ A, the condition for linear <strong>polarization</strong> is<br />

⃗A × ⃗ A † = 0, (2.82)<br />

having ⃗ A = ⃗a+i ⃗ b, this becomes ⃗a× ⃗ b = 0. The condition for circular <strong>polarization</strong><br />

is<br />

⃗A · ⃗A = 0, (2.83)<br />

which means that |Im[A]| = |Re[A]|, i.e. the complex and real part are <strong>of</strong> equal<br />

size. This section gives the main features <strong>of</strong> complex vector algebra, which<br />

we are using in this work. Complex vector algebra is used extensively when<br />

considering <strong>electromagnetic</strong> <strong>wave</strong>s, but is generally not thoroughly explained in<br />

the literature.<br />

2.3.2 The <strong>electromagnetic</strong> field<br />

For the complex electric or the magnetic field separately ( F ⃗ = F ⃗ (⃗r, ω) is either<br />

⃗E or B) ⃗ the tensor product is usually called the spectral density matrix. We<br />

have<br />

⎛<br />

⎞<br />

⃗F ⊗ F ⃗ F 1 F † 1 ∗ F 1 F2 ∗ F 1 F3<br />

∗<br />

= ⎝F 2 F1 ∗ F 2 F2 ∗ F 2 F3<br />

∗ ⎠ . (2.84)<br />

F 3 F1 ∗ F 3 F2 ∗ F 3 F3<br />

∗<br />

This matrix contains 9 parameters which is the number <strong>of</strong> degrees <strong>of</strong> freedom<br />

needed to completely describe the <strong>wave</strong> field. Since ⃗ F is a complex vector we<br />

can separate its real and imaginary parts. With this in mind one can easily<br />

show that Equation (2.84) is symmetric in the real part and anti-symmetric in<br />

the imaginary part. We have that the spectral density matrix consists <strong>of</strong> six real<br />

(the components on the diagonal are real) and three imaginary components, a<br />

total <strong>of</strong> 9 different degrees <strong>of</strong> freedom.<br />

This way <strong>of</strong> describing the <strong>wave</strong> is not very convenient since the parameters<br />

are in general not very physical and are not invariant under coordinate transformations.<br />

A better way is to write the spectral density matrix as a sum <strong>of</strong><br />

geometric quantities. Recall from the previous chapter, with ⃗ F = ⃗a + i ⃗ b, the<br />

quantities<br />

I F = ⃗ F · ⃗F † = a 2 + b 2 = | ⃗ F | 2 (2.85)<br />

and<br />

⃗V = i ⃗ F × ⃗ F † = i(⃗a + i ⃗ b) × (⃗a − i ⃗ b) = 2⃗a × ⃗ b. (2.86)<br />

I F is the energy density <strong>of</strong> the field in question. This V ⃗ is parallel (or antiparallel<br />

depending on whether the <strong>wave</strong> is left or rigthand polarized) to the<br />

direction <strong>of</strong> propagation <strong>of</strong> the <strong>wave</strong>. Note that E ⃗ is a polar vector and B ⃗ is an<br />

axial vector, but E ⃗ × E ⃗ † and B ⃗ × B ⃗ † are both axial vectors, since B ⃗ × B ⃗ † is a<br />

product <strong>of</strong> two axial vectors.<br />

The trace <strong>of</strong> the spectral density matrix is equal to the energy density <strong>of</strong> the<br />

field, which is an invariant scalar. From Equation (2.63) we know that we can


2.3. IRREDUCIBLE REPRESENTATION 19<br />

associate a vector to the anti-symmetric part <strong>of</strong> the matrix. This means that<br />

we can write the spectral density matrix in the following way<br />

⃗F ⊗ ⃗ F † = 1 3 Iδ ij + iɛ ijk v k + Q ij , (2.87)<br />

where I is a real scalar, ⃗v is the real dual pseudo-vector (associated to the antisymmetric<br />

imaginary part, see Eq. (2.63) ) and Q is a real symmetric tensor (the<br />

anti-symmetric part is contained in ɛ ijk v k ). This way <strong>of</strong> representing a tenor is<br />

termed an irreducible tensor representation (see e.g. Sakurai [20] and Arfken<br />

& Weber [10]). Since I and Q are real valued we know that ⃗v must contain the<br />

imaginary part. Hence,<br />

⎛<br />

⎞<br />

0 Im[F 1 F2 ∗ ] Im[F 1 F3 ∗ ]<br />

ɛ ijk v k = ⎝Im[F 2 F1 ∗ ] 0 Im[F 2 F3 ∗ ] ⎠ , (2.88)<br />

Im[F 3 F1 ∗ ] Im[F 3 F2 ∗ ] 0<br />

which gives<br />

⃗v = ( Im[F 2 F ∗ 3 ], Im[F 3 F ∗ 1 ], Im[F 1 F ∗ 2 ] ) . (2.89)<br />

To get ⃗v in terms <strong>of</strong> ⃗ F , we look at one component <strong>of</strong> ⃗ V , say V 3 (see Eq. (2.66))<br />

V 3 = i(F 1 F ∗ 2 − F 2 F ∗ 1 ) = i2i(b 1 a 2 − a 1 b 2 ) = −2Im(F 1 F ∗ 2 ) = 2Im(F 2 F ∗ 1 ). (2.90)<br />

And similarly for the other two components. We want ⃗v to be the imaginary<br />

part and thus we get<br />

⃗V = 2 ( Im(F 3 F ∗ 2 ), Im(F 1 F ∗ 3 ), Im(F 2 F ∗ 1 ) ) = −2⃗v. (2.91)<br />

Rewriting ⃗v in terms <strong>of</strong> ⃗ F we have the following expression for ⃗v<br />

⃗v = i [<br />

(F3 F2 ∗ − F 2 F3 ∗ ), (F 1 F3 ∗ − F 3 F1 ∗ ), (F 2 F1 ∗ − F 1 F2 ∗ ) ]<br />

2<br />

= ( − Im(F 3 F ∗ 2 ), −Im(F 1 F ∗ 3 ), −Im(F 2 F ∗ 1 ) ) = − i 2 ⃗ F × ⃗ F † . (2.92)<br />

Comparing ⃗v with V ⃗ , we have that ⃗v defines left-handed and right-handed circular<br />

<strong>polarization</strong> as seen by an observer receiving the <strong>wave</strong> (standing in the<br />

direction <strong>of</strong> propagation), the size <strong>of</strong> this vector, ⃗v is 1 π<br />

times the area <strong>of</strong> the <strong>polarization</strong><br />

ellipse (see Lindell [14]). The relation between ⃗v and Stokes parameter<br />

V is (compare with Equation (2.69))<br />

√<br />

|V| = 2 v1 2 + v2 2 + v2 3 = |2⃗v|. (2.93)<br />

By subtraction we get Q from Equation (2.87), and we have<br />

⎛<br />

F1 2 − 1 3 I 1<br />

2 (F 1F2 ∗ + F 2 F ∗ 1<br />

1 )<br />

2 (F ⎞<br />

3F1 ∗ + F 1 F3 ∗ )<br />

Q = ⎝ 1<br />

2 (F 1F2 ∗ + F 2 F1 ∗ ) F2 2 − 1 3 I 1<br />

2 (F 3F2 ∗ + F 2 F3 ∗ ) ⎠ . (2.94)<br />

1<br />

2 (F 3F1 ∗ + F 1 F3 ∗ 1<br />

)<br />

2 (F 3F2 ∗ + F 2 F3 ∗ ) F3 2 − 1 3 I<br />

We see that ⃗v contains the imaginary part <strong>of</strong> the matrix ⃗ F ⊗ ⃗ F † and that Q is<br />

real valued and symmetric.


20 CHAPTER 2. BACKGROUND<br />

Maxwell stress tensor<br />

A tensor similar to Q is the Maxwell stress tensor (see Wangsness [25], Jackson<br />

[12] and Marion & Heald [16]), for real fields defined as<br />

T ij = ɛ 0<br />

(<br />

Ei E j − 1 2 ⃗ E 2 δ ij<br />

)<br />

+<br />

1<br />

µ 0<br />

(<br />

Bi B j − 1 2 ⃗ B 2 δ ij<br />

)<br />

=<br />

⎛<br />

E1 2 − 1 2E ⃗ 2 ⎞<br />

E 1 E 2 E 1 E 3<br />

ɛ 0<br />

⎝ E 2 E 1 E2 2 − 1 ⃗ 2E 2 E 2 E 3<br />

⎠ +<br />

E 3 E 1 E 3 E 2 E3 2 − 1 ⃗ 2E 2<br />

The trace <strong>of</strong> this symmetric tensor is<br />

⎛<br />

B<br />

1<br />

1 2 − 1 2B ⃗ 2 ⎞<br />

B 1 B 2 B 1 B 3<br />

⎝ B<br />

µ 2 B 1 B2 2 − 1 ⃗ 2B 2 B 2 B 3<br />

⎠ . (2.95)<br />

0<br />

B 3 B 1 B 3 B 2 B3 2 − 1 ⃗ 2B 2<br />

T r(T ij ) = − ɛ 0<br />

2 ⃗ E 2 − 1<br />

2µ 0<br />

⃗ B 2 , (2.96)<br />

which is minus the energy density <strong>of</strong> the <strong>electromagnetic</strong> field.<br />

2.3.3 The product ⃗ E ⊗ ⃗ B<br />

We now form the tensor product <strong>of</strong> E ⃗ and B ⃗ † , which is<br />

⎛<br />

⃗E ⊗ B ⃗ † = ⎝ E ⎞<br />

1B1 ∗ E 1 B2 ∗ E 1 B3<br />

∗<br />

E 2 B1 ∗ E 2 B2 ∗ E 2 B3<br />

∗ ⎠ . (2.97)<br />

E 3 B1 ∗ E 3 B2 ∗ E 3 B3<br />

∗<br />

This matrix has a total <strong>of</strong> 18 components and is a tensor since it is constructed<br />

from two vectors. Since it is a tensor and we can write any tensor as a sum <strong>of</strong><br />

its symmetric and anti-symmetric parts we follow the same procedure as we did<br />

with the field ⃗ F in the previous section, Equation (2.87), and we get<br />

I EB = E 1 B ∗ 1 + E 2 B ∗ 2 + E 3 B ∗ 3,<br />

(2.98a)<br />

⃗v EB = i 2[<br />

(E3 B ∗ 2 − E 2 B ∗ 3), (E 1 B ∗ 3 − E 3 B ∗ 1), (E 2 B ∗ 1 − E 1 B ∗ 2) ] , (2.98b)<br />

⎛<br />

E 1 B1 ∗ − 1 3 I EB<br />

Q EB = ⎝ 1<br />

2 (E 1B2 ∗ + E 2 B1) ∗ E 2 B2 ∗ − 1 3 I EB<br />

1<br />

2 (E 1B2 ∗ + E 2 B1) ∗ 1<br />

2 (E ⎞<br />

3B1 ∗ + E 1 B3)<br />

∗<br />

1<br />

2 (E 3B2 ∗ + E 2 B3)<br />

∗ ⎠ . (2.98c)<br />

1<br />

2 (E 3B1 ∗ + E 1 B3) ∗ 1<br />

2 (E 3B2 ∗ + E 2 B3) ∗ E 3 B3 ∗ − 1 3 I EB<br />

When considering reflections, these quantities are different from the decomposition<br />

<strong>of</strong> the tensor product <strong>of</strong> the fields with themselves (see previous Section,<br />

decomposition <strong>of</strong> Equation (2.84)). In Equation (2.98), we have that I EB is a<br />

pseudo-scalar (odd under spatial inversion) and Q EB is a pseudo-tensor (odd<br />

under spatial inversion). The vector, ⃗v EB , is a polar vector which is odd under<br />

spatial inversion. 17 Spatial inversion, or reflection, is an example <strong>of</strong> an improper<br />

17 For definitions and transformation properties <strong>of</strong> pseudo-quantities, see Goldstein [11],<br />

Arfken & Weber [10] or Jackson [12].


2.3. IRREDUCIBLE REPRESENTATION 21<br />

rotation, which changes the coordinate system from right-handed to left-handed<br />

or vice versa.<br />

All the components in the equations above (Eq. (2.98)) are complex so we<br />

separate the real and imaginary parts according to<br />

⃗E ⊗ ⃗ B † = Re[ ⃗ E ⊗ ⃗ B † ] + iIm[ ⃗ E ⊗ ⃗ B † ]. (2.99)<br />

These two matrices, the real and imaginary parts, are two real matrices and<br />

can be divided in the same way just taking real and imaginary parts <strong>of</strong> the<br />

EB-components separately, and we have<br />

This is a separation we will use later on.<br />

The Poynting vector<br />

Re[ ⃗ E ⊗ ⃗ B † ] = 1 2( ⃗E ⊗ ⃗ B † + ⃗ B ⊗ ⃗ E †) , (2.100)<br />

iIm[ ⃗ E ⊗ ⃗ B † ] = 1 2( ⃗E ⊗ ⃗ B † − ⃗ B ⊗ ⃗ E †) . (2.101)<br />

The vector ⃗v EB (Eq. (2.98b)) is very similar to the Poynting vector. For real<br />

valued vector fields, the Poynting vector is defined as<br />

⃗S = ⃗ E × ⃗ H, (2.102)<br />

where H ⃗ is the magnetic H-field (see Wagnsness [25] or Jackson [12]). Compared<br />

to the vector ⃗v EB , the Poynting vector is recognised as a polar vector. The<br />

Poynting vector, S, is the instantaneous energy flow and points in the direction<br />

<strong>of</strong> energy flow. In the complex case, using our notation with E ⃗ and B, ⃗ it becomes<br />

(<br />

)<br />

⃗S = cE ⃗ × B ⃗ † = c E 2 B3 ∗ − E 3 B2, ∗ E 3 B1 ∗ − E 1 B3, ∗ E 1 B2 ∗ − E 2 B1<br />

∗ . (2.103)<br />

The real part <strong>of</strong> Equation (2.103) is the instantaneous energy flow <strong>of</strong> the field<br />

and is varying with time. The time average <strong>of</strong> the energy flow, 〈 ⃗ S〉, is <strong>of</strong>ten<br />

more useful. Using the complex notation for a plane <strong>wave</strong> where ⃗ E and ⃗ B are<br />

the complex vector amplitudes this time average is very easy to calculate, we<br />

get<br />

〈 ⃗ S〉 = 1 2 cRe[ ⃗ E × ⃗ B † ]. (2.104)<br />

Comparing Equa-<br />

Which is just half <strong>of</strong> the real part <strong>of</strong> Equation (2.103).<br />

tion (2.103) with Equation (2.98b) we see that<br />

⃗S = 2i⃗v EB . (2.105)<br />

Thus, we have a simple relation between the complex Poynting vector and the<br />

⃗v EB vector.


22 CHAPTER 2. BACKGROUND<br />

2.4 Group theory <strong>of</strong> transformation matrices<br />

In this section we give an introduction to the group theory <strong>of</strong> the Lie groups<br />

used in this thesis. The definitions <strong>of</strong> the groups are given together with physical<br />

interpretations. We also introduce the mathematical concept <strong>of</strong> irreducible<br />

tensor representation and apply it to the groups SO(2), SO(3) and SO(3, 1).<br />

2.4.1 The general linear group and subgroups<br />

In this thesis there are a number <strong>of</strong> different symmetry groups consisting <strong>of</strong> operations,<br />

such as rotation or reflection, on a real or complex vector space. These<br />

groups can be thought <strong>of</strong> as matrices (see Fieseler [9] or Arfken & Weber [10]),<br />

i.e. transformation matrices. These groups are continuous groups or Lie groups.<br />

For more information on these groups, see Barut & Raczka [18] or Tung [24].<br />

The first group, the general linear group, is the group <strong>of</strong> all invertible matrices,<br />

and can be expressed as<br />

GL n (K) = {A ∈ K n,n ; det A ≠ 0}, (2.106)<br />

where K, in our case, is either the real numbers, R, or the complex numbers,<br />

C.<br />

A subgroup to GL n (K) is the the Special linear group, and is defined as<br />

SL n (K) = {A ∈ GL n (K); det A = 1} = {A ∈ K n,n ; det A = 1}. (2.107)<br />

This group leaves any totally anti-symmetric tensor <strong>of</strong> rank n invariant.<br />

The orthogonal group is the subgroup <strong>of</strong> GL n (K) given by<br />

O(n) = {A ∈ GL n (R); A⃗x · A⃗y = ⃗x · ⃗y ∀⃗x, ⃗y ∈ R n }<br />

= {A ∈ GL n (R); A T A = I n }. (2.108)<br />

This is the set <strong>of</strong> all linear transformations that leaves the Euclidean metric in n<br />

dimensions, δ µν (n) , invariant. This notation can be extended to any metric tensor<br />

with signature (n + , n − ), and the group is O(n + , n − ). For the flat metric, g µν<br />

(Eq. (2.6)), with signature (1, −1, −1, −1), we have the group O(1, 3), which is<br />

called the orthogonal group <strong>of</strong> signature (1, −1, −1, −1).<br />

The special orthogonal group, SO(n), or SO(n + , n − ), is the subgroup <strong>of</strong><br />

O(n) which admits the totally anti-symmetric tensor, ɛ [n] (the [n] on ɛ [n] is the<br />

number <strong>of</strong> suffixes, the rank, <strong>of</strong> the tensor), as an additional invariant tensor.<br />

This group is the intersection<br />

SO(n) = SL n (C) ∩ O(n), (2.109)<br />

<strong>of</strong> the special linear group and the orthogonal group. An important difference<br />

between the O(n) and SO(n) groups is that the O(n) groups include improper<br />

rotations (reflection, inversion), which is not included in the SO(n) groups because<br />

<strong>of</strong> the detA = 1 condition. This absence <strong>of</strong> reflections gives the SO(n)<br />

groups the extra invariant tensor ɛ [n] , the totally anti-symmetric tensor.<br />

Similar to the orthogonal group, O(n), which is real, we have in a complex<br />

vector space the unitary group, defined as<br />

U(n) = {A ∈ GL n (C); A⃗x · (A⃗y) † = ⃗x · ⃗y † ∀⃗x, ⃗y ∈ C n }<br />

= {A ∈ GL n (C); A † A = AA † = I n }. (2.110)


2.4. GROUP THEORY OF TRANSFORMATION MATRICES 23<br />

This is the group <strong>of</strong> linear unitary transformations that leave a metric, δ α β ,<br />

invariant. Here we also have the possibility <strong>of</strong> different signature and it it is<br />

written in the the same way as for O(n + , n − ). The special unitary group is a<br />

subgroup <strong>of</strong> U(n), given by<br />

SU(n) = SL n (C) ∩ U(n). (2.111)<br />

This is a subgroup <strong>of</strong> U(n), which admits a totally anti-symmetric tensor <strong>of</strong><br />

rank n as an invariant tensor. As in the real case, the orthogonal groups,<br />

the unitary group, U(n), includes reflections, while the special unitary group,<br />

SU(n), do not have reflections, but allows the totally anti-symmetric tensor as<br />

as invariant. Similar to U(n + , n − ), we can have a signature <strong>of</strong> the metric tensor,<br />

with the same notation, SU(n + , n − ).<br />

2.4.2 Physical use <strong>of</strong> the groups<br />

The groups considered in this work are SO(2), SU(2), SO(3), SU(3), SO(3, 1)<br />

and SU(3, 1).<br />

The generators <strong>of</strong> the first group, SU(2), are known as the Pauli spin matrices,<br />

σ ν , given by Equation (2.54). The group SU(2) is isomorphic to the<br />

next group, SO(3), which is the rotation group in three-dimensional Euclidean<br />

space. The SU(3) group is similar to SO(3), but is a complex group, consisting<br />

<strong>of</strong> complex rotations in C 3 . The generators <strong>of</strong> SU(3) are found in Appendix A<br />

and are used to extract the generalized <strong>polarization</strong> parameters.<br />

The proper homogeneous Lorentz group, with the flat metric g µν , given<br />

by Equation (2.6), is the group SO(3, 1) and the generalization to a complex<br />

vector space is the group SU(3, 1). Note that the signature <strong>of</strong> the metric is<br />

(1, −1, −1, −1), which really is the same as (−1, 1, 1, 1) considering Lorentz<br />

transformations, i.e. SU(3, 1) = SU(1, 3).<br />

If we enlarge the groups SO(3, 1) and SU(3, 1) to include reflections, we<br />

have the groups O(3, 1) and U(3, 1), respectively. These groups includes the<br />

improper Lorentz transformations, which have det A = −1 . The proper Lorentz<br />

transformations, together with translations, forms the Poincare group (proper<br />

inhomogeneous Lorentz group), a group <strong>of</strong>ten encountered in the subject <strong>of</strong> the<br />

Lorentz group.<br />

The groups we are considering are SU(2), SO(3), SU(3), SO(3, 1) and SU(3, 1).<br />

The special unitary and special orthogonal groups have their respective metric<br />

tensor and the totally anti-symmetric tenor as invariant tensors. Using these invariant<br />

tensors we can decompose tensors <strong>of</strong> the groups into lower rank tensors.<br />

The unitary groups will include complex rotations, given by spherical tensors<br />

(see Sakurai [20]), which can be interpreted as a change in the phase <strong>of</strong> the <strong>wave</strong>.<br />

Restricting the rotations to only be real rotations <strong>of</strong> the Cartesian coordinate<br />

axes, we can use the groups SO(3) and SO(3, 1) on complex vector spaces.<br />

2.4.3 Irreducible tensor representation <strong>of</strong> Lie groups<br />

In Section 2.3 we made a decomposition <strong>of</strong> a tensor product, ⃗ A ⊗ B, which is a<br />

rank two tensor, into a scalar (rank zero tensor), a vector (rank one tensor) and<br />

a tensor (rank two tensor). The tensors produced in this process are irreducible,<br />

which means that they can not be written as a product <strong>of</strong> lower rank tensors


24 CHAPTER 2. BACKGROUND<br />

using the metric tensor, g µν , and the totally anti-symmetric tensor, ɛ [n] . In<br />

the rotation group, SO(3), the irreducible tensor representation <strong>of</strong> a general<br />

second rank tensor is just the decomposition done in Section 2.3, with the metric<br />

g ab = δ ab , the Kronecker delta. The decomposition is given by<br />

where<br />

and<br />

A ij = 1 3 Θδ ij + 1 2 ɛ ijkω k + σ ij , (2.112)<br />

θ = A i i, (2.113)<br />

ω i = ɛ ijk A jk (2.114)<br />

σ ij = 1 2 (A ij + A ji ) − 1 3 g ijA k k (2.115)<br />

are the trace, twice the dual vector and the symmetric part <strong>of</strong> A ij , respectively.<br />

Here we considered SO(3), but this decomposition is the same for SU(3) in the<br />

complex vector space, which we have done in Section 2.3 (where we use a vector<br />

algebra approach), but under SU(3) this representation is not irreducible (see<br />

Section 2.4.2). The constant, 1 2 , in the anti-symmetric term, ωk , is dependent<br />

on how one defines the dual vector (compare with Equation (2.87)).<br />

Considering the groups SO(3, 1), real vector space, and SU(3, 1), complex<br />

vector space, the representations are the equivalent (compare with SU(3) ↔<br />

SO(3) above) but SO(3, 1) is real and SU(3, 1) is complex. In the four-dimensional<br />

groups, SO(3, 1) and SU(3, 1), the invariant tensors are a rank two tensor, the<br />

metric tensor, and a rank four tensor, the totally anti-symmetric tensor. For the<br />

flat metric, g µν , with signature (1, −1, −1, −1), and the definition <strong>of</strong> complex<br />

scalar product, ⃗a · ⃗b † , SO(3, 1) and SU(3, 1) are the proper Lorents groups for<br />

real and complex vector spaces, respectively. The decomposition for a general<br />

rank two tensor, A µν , is given by<br />

A µν = 1 4 Sg µν + 1 2 ɛ µναβT (a)αβ + T µν , (2.116)<br />

where S is a scalar, recognised as the trace <strong>of</strong> A. T (a) is a dual tensor and<br />

contains the anti-symmetric part <strong>of</strong> A, ɛ is antisymmetric. T is a tensor and<br />

is the symmetric part <strong>of</strong> A. Comparing with the three-dimensional case, Equation<br />

(2.112), S, T (a) and T are given by<br />

and<br />

S = A µν g µν , (2.117)<br />

T (a)<br />

µν = ɛ µναβ A αβ (2.118)<br />

T µν = 1 2 (A µν + A νµ ) − 1 4 g µνA α α. (2.119)<br />

Comparing with Equation (2.112), there are no vectors, i.e. rank one tensors,<br />

in Equation (2.116). This is due to the fact that we have an even number <strong>of</strong><br />

dimensions and the invariant tensors has even rank. This gives that a traceless


2.4. GROUP THEORY OF TRANSFORMATION MATRICES 25<br />

rank two tensor is irreducible under SO(3, 1) but can be split into its symmetric<br />

and anti-symmetric parts, T and T (a) in Equation (2.116) respectively. Since<br />

contraction is over two indices we do not get any odd rank tensors. Note, this<br />

does not mean that proper four-vectors are forbidden but we can not construct<br />

them directly using the invariant tensors <strong>of</strong> the group. To achieve the odd rank<br />

tensors we have to consider the pertinent physical equations.<br />

Before moving on, there are some remarks concerning index notation <strong>of</strong><br />

tensors under different groups. In the group SO(3) we have a diagonal metric<br />

with signature (1, 1, 1), i.e. the metric is the Kronecker delta. With this metric<br />

there is no difference between a <strong>covariant</strong> (lower) and contravariant (upper)<br />

index. I.e. the relation A ij... = A ij... is valid for every tensor and the relation<br />

holds for randomly mixed indices. In contrast to SO(3) we have SO(3, 1) with a<br />

metric signature (1, −1, −1, −1). For a general second rank tensor we have that<br />

A µν ≠ A µν and A µ ν ≠ Aµ ν , where the last relation is the difference between<br />

raising (or lowering) the first or the second index. With mixed indices the<br />

metric tensor is just the Kronecker delta, which can be convenient to use in<br />

certain calculations. In decomposition and construction <strong>of</strong> lower rank tensors it<br />

is favourable to use non-mixed tensors since it allows direct permutation <strong>of</strong> the<br />

indices.<br />

This decomposition can be generalized for tensors <strong>of</strong> higher rank and will<br />

include pseudo-scalars and higher order pseudo-tensors, but to obtain an irreducible<br />

form we must form all possible invariant tensors, using products <strong>of</strong> the<br />

invariant tensors, g µν and ɛ αβγδ , and lower rank tensors. The number <strong>of</strong> terms<br />

will increase dramatically with higher rank and it is very difficult to determine<br />

what each tensor has as its components. To make it simple we can do a reducible<br />

representation with fewer terms, where it is much easier to identify each tensor.<br />

Considering a general rank four tensor, A αβγδ , we can get it in a representation<br />

similar to a tensor product <strong>of</strong> a rank two representation, Equation (2.116). This<br />

gives the following decomposition<br />

A αβγδ = Sg αβ g γδ + (1) T µν g αβ ɛ γδµν + (2) T γδ g αβ<br />

+ (3) T µν ɛ αβµν g γδ + (4) T µν (5) T ψχ ɛ αβµν ɛ γδψχ<br />

+ (6) T µν (7) T γδ ɛ αβµν + (8) T αβ g γδ + (9) T αβ (10) T µν ɛ γδµν + U αβγδ , (2.120)<br />

where S is a scalar, the n T ’s are rank two tensors that are either symmetric<br />

or anti-symmetric and U is a rank four tensor. This representation is generally<br />

not an irreducible representation. For this to be a proper tensor representation<br />

the S, T and U tensors have to be proper tensors, and for a general rank four<br />

tensor, they are difficult to determine. The technique we use is to construct<br />

lower rank tensors from the original tensor, A, and use these lower rank tensors<br />

in the decomposition. We construct the tensors<br />

and<br />

S = A αβγδ g αβ g γδ ,<br />

S (p) = A αβγδ ɛ αβγδ<br />

(2.121)<br />

T αβ = A αβγδ g γδ − 1 S. (2.122)<br />

4


26 CHAPTER 2. BACKGROUND<br />

Using these tensors to decompose A αβγδ , we obtain<br />

A αβγδ = 1 16 Sgαβ g γδ + 1 4! S(p) ɛ αβγδ + 1 2 T αβ g γδ + Ψ αβγδ , (2.123)<br />

where Ψ is a reducible rank four tensor. The tensor T αβ is a rank two tensor and<br />

thus can be reduced according to Equation (2.116). The irreducible tensors are<br />

scalar, pseudo-scalar, symmetric traceless rank two tensor and anti-symmetric<br />

rank two tensor.<br />

To do a decomposition <strong>of</strong> higher rank tensors, one makes a representation<br />

consisting <strong>of</strong> these irreducible tensors and products there<strong>of</strong>, iterating the symmetry<br />

and anti-symmetry between indices. Comparing this with the symmetries<br />

<strong>of</strong> the tensor we want to decompose, it will be easier to construct the proper<br />

tensors in the representation.<br />

Example: Irreducible tensor representation <strong>of</strong> the two-dimensional<br />

spectral density matrix<br />

A simple example <strong>of</strong> irreducible tensor representation is to take the two dimensional<br />

spectral density tensor, Equation (2.53), which is a tensor under the<br />

group SO(2), and decompose it into an irreducible representation. In SO(2)<br />

we have the invariant metric tensor, δ αβ , and the totally anti-symmetric tensor,<br />

ɛ αβ . For a general rank two tensor, A αβ , the irreducible representation will be<br />

A αβ = 1 2 Sδ αβ + 1 2 S(p) ɛ αβ + T αβ , (2.124)<br />

where we have S as the trace, S = A αβ δ αβ , and the anti-symmetric part, S (p) =<br />

A αβ ɛ αβ , which is a pseudo-scalar. Performing this on the two dimensional<br />

spectral density matrix, S d (Eq. (2.53)), we have<br />

( )<br />

F1 F1 S d = ∗ F 1 F2<br />

∗<br />

F 2 F1 ∗ F 2 F2<br />

∗ = 1 2 (F 1 2 + F2 2 )δ αβ + 1 2 (F 1F † 2 − F 2F † 1 )ɛ αβ+<br />

( F<br />

2<br />

1 − F2 2 F 1 F2 ∗ + F 2 F1<br />

∗<br />

1<br />

2<br />

F 1 F † 2 + F 2F ∗ 1 F 2 2 − F 2 1<br />

)<br />

. (2.125)<br />

Comparing this representation with Stokes’ parameters, Equation (2.49), we see<br />

that it can be written<br />

(<br />

F1 F †<br />

S d =<br />

1 F 1 F † )<br />

2<br />

F 2 F † 1 F 2 F † = 1<br />

2<br />

2 Iδ αβ − i 2 Vɛ αβ + 1 ( ) Q U<br />

, (2.126)<br />

2 U −Q<br />

where the energy density, I, is the invariant scalar and the degree <strong>of</strong> circular<br />

<strong>polarization</strong>, V, is the invariant pseudo scalar. The last two Stokes parameters,<br />

corresponding to linear <strong>polarization</strong> along the axis (Q) and linear <strong>polarization</strong> in<br />

45 ◦ angle to the axis (U), together form a traceless symmetric rank two tensor.<br />

This is the irreducible tensor representation <strong>of</strong> the two-dimensional spectral<br />

density tensor under SO(2).


Chapter 3<br />

The <strong>covariant</strong> spectral<br />

density tensor and a generic<br />

decomposition scheme<br />

In this chapter we introduce the <strong>covariant</strong> spectral density tensor, τ αβγδ ≡<br />

F αβ (F γδ ) ∗ , which is a manifestly Lorentz <strong>covariant</strong> rank four tensor, containing<br />

all quadratic components <strong>of</strong> the <strong>electromagnetic</strong> field. An alternative way<br />

<strong>of</strong> constructing a <strong>covariant</strong> spectral density tensor is to consider A µ A ν (see<br />

Barut [1]), where A µ is the <strong>electromagnetic</strong> four-potential (see Section 2.1.2),<br />

The four-potential is however not a measurable physical quantity and it also<br />

depends on the choice <strong>of</strong> gauge. Since our approach is to consider measurable<br />

parameters <strong>of</strong> <strong>polarization</strong> (just as Stokes’ parameters) the seemingly more complicated<br />

rank four tensor τ αβγδ is preferable. One way to expand the <strong>covariant</strong><br />

spectral density tensor is to use the Dirac gamma matrices. In this chapter<br />

two different definitions <strong>of</strong> the Dirac gamma matrices is presented and we use<br />

the one that comes from the flat metric g µν (see Section 2.1, Equation (2.6)) to<br />

extract the 36 parameters needed for a complete representation <strong>of</strong> the field. The<br />

technique here is similar to that in Section 2.2, where a three-dimensional tensor<br />

<strong>of</strong> rank two is decomposed in terms <strong>of</strong> a basis <strong>of</strong> matrices. In this Chapter we<br />

use the original four Dirac gamma matrices and expand them into a complete<br />

16 dimensional matrix basis. The <strong>covariant</strong> spectral density tensor is expanded<br />

in this basis.<br />

3.1 Covariant spectral density tensor<br />

To have all quadratic <strong>electromagnetic</strong> field components collected into one object,<br />

we introduce and define the <strong>covariant</strong> spectral density tensor, τ αβγδ , by<br />

τ αβγδ ≡ F αβ (F γβ ) ∗ . (3.1)<br />

The definition for real fields is the same but the complex conjugate is then<br />

abundant. This is the deterministic version where we know the pertinent field<br />

completely. For a physical field, where one considers a certain bandwidth, this<br />

27


28 CHAPTER 3. COVARIANT SPECTRAL DENSITY TENSOR<br />

tensor becomes<br />

τ αβγδ ≡ 〈F αβ (F γβ ) ∗ 〉, (3.2)<br />

where the average is taken separately on each component. When performing calculations<br />

we can ignore the average, 〈 〉, and do the analysis for a deterministic<br />

<strong>wave</strong>.<br />

The deterministic approach, in contrast to consider a stochastical <strong>wave</strong>, is<br />

used throughout this thesis due to the fact that the average can be defined<br />

differently. To keep the generality we therefore do not explicitly define the<br />

〈 〉-operator. 1<br />

3.2 Pauli spin matrices<br />

The gamma matrices are usually written in terms <strong>of</strong> the Pauli spin matrices<br />

(σ 1 , σ 2 , σ 3 ) and the 2 × 2 unit matrix (I 2 ). The Pauli matrices comes from the<br />

non-relativistic treatment <strong>of</strong> spin- 1 2<br />

particles and they are defined as<br />

σ 1 =<br />

( 0 1<br />

1 0<br />

) ( 0 −i<br />

, σ 2 =<br />

i 0<br />

) ( 1 0<br />

, σ 3 =<br />

0 −1<br />

) ( 1 0<br />

, I 2 =<br />

0 1<br />

These are the well known Pauli spin matrices from quantum mechanics.<br />

3.3 Dirac gamma matrices<br />

)<br />

.<br />

(3.3)<br />

The gamma matrices arise in the relativistic treatment <strong>of</strong> spin- 1 2<br />

particles and is<br />

the four dimensional generalization <strong>of</strong> the Pauli matrices. There are two different<br />

ways <strong>of</strong> defining the gamma matrices, one comes from using the Minkowski<br />

notation, (x, y, z, ict), and the second one from using Lorentz space with the<br />

flat metric, g µν (Eq. (2.6)).<br />

The first notation (used by Sakurai [19]) is the one originally used by Dirac,<br />

who defined the matrices as<br />

( )<br />

( )<br />

0 −iσµ<br />

I2 0<br />

γ µ =<br />

, µ = 1, 2, 3, γ<br />

iσ µ 0<br />

4 =<br />

. (3.4)<br />

0 −I 2<br />

One also defines a fifth matrix as<br />

γ 5 = γ 1 γ 2 γ 3 γ 4 (3.5)<br />

Here the γ µ ’s are Hermitian for µ=1,2,3,4,5. The second notation (used by for<br />

example Bjorken & Drell [2], Arfken & Weber [10]) one gets when using the<br />

metric g µν (see Eq. (2.6)), Lorentz notation, and has the following form<br />

(<br />

γ µ 0 σ<br />

µ<br />

=<br />

−σ µ 0<br />

The fifth matrix in this case looks as<br />

)<br />

, µ = 1, 2, 3 , γ 0 =<br />

( )<br />

I2 0<br />

. (3.6)<br />

0 −I 2<br />

γ 5 = γ 5 = iγ 0 γ 1 γ 2 γ 3 . (3.7)<br />

1 Compare with quantum mechanics, where the 〈 〉-operator is completely different from<br />

the classical operator (see e.g. Sakurai [19] [20]).


3.4. DECOMPOSITION OF τ αβγδ 29<br />

These matrices have the properties<br />

γ µ γ ν + γ ν γ µ = 2g µν µ, ν = 0, 1, 2, 3, (3.8)<br />

γ 0† = γ 0 , γ k† = −γ k , k = 1, 2, 3. (3.9)<br />

With these matrices we can form a 16-dimensional algebra, using the unit matrix<br />

and the symmetric combinations <strong>of</strong> the matrices. It is a complete basis<br />

with convenient Lorentz transformation properties. These 16 matrices are constructed<br />

in the following way 2<br />

I 4 , γ 5 = iγ 0 γ 1 γ 2 γ 3 , γ µ , γ 5 γ µ , σ µν = i(γ µ γ ν − γ ν γ µ )/2. (3.10)<br />

Considering Lorentz transformations, the first four tensors, γ µ , µ = 0, 1, 2, 3,<br />

transforms as a vector and γ 5 γ µ transforms as an axial vector (pseudo-vector).<br />

The six anti-symmetric combinations, σ µν transforms as a tensor. 3<br />

3.4 Decomposition <strong>of</strong> τ αβγδ<br />

We want to extract the quadratic forms <strong>of</strong> the components in the <strong>covariant</strong><br />

spectral density tensor, τ αβγδ = F αβ (F γδ ) ∗ . This tensor contains all correlation<br />

components and is Lorentz <strong>covariant</strong>. 4<br />

Here we use the notation that γ 4 is the unit 4 × 4 matrix and perform the<br />

operations<br />

F µν (F αβ ) ∗ γ n µνγ n αβ , n = 0, 1, . . . , 15. (3.11)<br />

This gives a total <strong>of</strong> 256 scalars (16 × 16) but only 36 are non-zero (all 36 are<br />

found in appendix B). These 36 nonzero are the combinations <strong>of</strong> γ 1 , γ 3 , γ 6 , γ 8 , γ 11<br />

and γ 14 , which are the antisymmetric gamma matrices. We make the contractions<br />

F µν γ n µν, which yields<br />

F µν γ 1 µν = −2(E 3 + cB 3 ),<br />

F µν γ 3 µν = −2(E 2 + cB 2 ),<br />

F µν γ 6 µν = 2(E 2 − cB 2 ),<br />

F µν γ 8 µν = 2i(−E 1 + cB 1 ),<br />

F µν γ 11<br />

µν = 2(−E 3 + cB 3 ),<br />

F µν γ 14<br />

µν = −2i(E 1 + cB 1 ).<br />

(3.12)<br />

Performing this operation with the symmetric gamma matrices gives zero since<br />

the field strength tensor is antisymmetric and has zeroes as its diagonal elements.<br />

Our 36 quadratic parameters are then the combinations <strong>of</strong> any two <strong>of</strong> the linear<br />

expressions in Equation (3.12), remembering that in general, ab ∗ ≠ a ∗ b.<br />

Rewriting the field strength tensor as a sum, using the Dirac gamma matrices,<br />

we obtain<br />

2F µν = −(E 3 + cB 3 )γ 1 − (E 2 + cB 2 )γ 3 + (E 2 − cB 2 )γ 6<br />

− i(−E 1 + cB 1 )γ 8 + (−E 3 + cB 3 )γ 11 + i(E 1 + cB 1 )γ 14 . (3.13)<br />

2 All 16 gamma matrices are found in appendix B where they are written out in matrix<br />

form.<br />

3 More information on the transformation properties <strong>of</strong> Dirac gamma matrices is found in<br />

Bjorken & Drell [2], Sakurai [19] and Arfken & Weber [10].<br />

4 The <strong>covariant</strong> spectral density tensor is explored and used more extensively in Chapter 6.


30 CHAPTER 3. COVARIANT SPECTRAL DENSITY TENSOR<br />

We have here a method <strong>of</strong> constructing 36 linearly independent parameters<br />

<strong>of</strong> the <strong>electromagnetic</strong> field but it is not an easy and direct method, since the<br />

parameters we gain are not collected into different geometrical quantities. On<br />

the good side we are just using the field strength tensor and the Dirac gamma<br />

matrices, which have known transformation properties under the Lorentz group.


Chapter 4<br />

Irreducible representation<br />

<strong>of</strong> SO(3) <strong>covariant</strong><br />

<strong>polarization</strong><br />

In Section 2.3 we defined the tensor product between two fields and represented<br />

it as a sum <strong>of</strong> a scalar, a vector and a tensor, using SO(3) symmetry. In this<br />

chapter we use this to define a new set <strong>of</strong> parameters describing the <strong>electromagnetic</strong><br />

field. The number <strong>of</strong> linearly independent components for a total<br />

description <strong>of</strong> the fields are 36. Comparing with Storey and Lefeuvre [23], this<br />

is the number <strong>of</strong> quadratic terms in the product ( E, ⃗ B) ⃗ ⊗ ( E ⃗ † , B ⃗ † ).<br />

4.1 The irreducible parameters under SO(3)<br />

The three-dimensional rotation group, SO(3) (see Section 2.4), is just rotation<br />

<strong>of</strong> coordinates in three-dimensional Euclidean space. The generators <strong>of</strong> SO(3)<br />

are rotation around each <strong>of</strong> the coordinate axis. The geometrical objects <strong>of</strong> this<br />

group, SO(3), and the ordinary Cartesian three-dimensional space are scalars,<br />

vectors and tensors. 1 To label the parameters in a convenient way we use the<br />

letters S, V, T , which are mnemonics for Scalar, V ector and T ensor. The general<br />

decomposition scheme can be written<br />

F ⊗ F † = S ⊕ V ⊕ T. (4.1)<br />

This is a mathematical way <strong>of</strong> displaying the decomposition into different subspaces<br />

<strong>of</strong> geometrical objects and that the subspaces are closed under SO(3)<br />

transformations.<br />

The superscripts used to label the parameters are Σ, ∆, χ, ψ. The labels Σ<br />

and ∆ denotes the autocorrelated parameters involving only multiplication <strong>of</strong> ⃗ E<br />

with itself and ⃗ B with itself. Σ stands for the sum between electric and magnetic<br />

components and ∆ is the difference. The last two labels, χ and ψ, denotes the<br />

cross-correlated parameters, where the components only contain multiplication<br />

1 For more information on SO(3) and similar groups, good sources are Arfken & Weber [10],<br />

Goldstein [11] and Tung [24].<br />

31


32CHAPTER 4. IRREDUCIBLE REPRESENTATION OF POLARIZATION<br />

between an electric and a magnetic quantity. The label χ corresponds to the sum<br />

which means that it is the real part. The label ψ corresponds to the difference<br />

and thus the imaginary part. The parameters are defined as<br />

Σ : ⃗ E ⊗ ⃗ E † + ⃗ B ⊗ ⃗ B † = 1 3 δ ijS Σ + iɛ ijk V Σ<br />

k + T Σ ij , (4.2)<br />

∆ : ⃗ E ⊗ ⃗ E † − ⃗ B ⊗ ⃗ B † = 1 3 δ ijS ∆ + iɛ ijk V ∆<br />

k + T ∆ ij , (4.3)<br />

χ : ⃗ E ⊗ ⃗ B † + ⃗ B ⊗ ⃗ E † = 1 3 δ ijS χ + iɛ ijk V χ<br />

k + T χ ij , (4.4)<br />

ψ : ⃗ E ⊗ ⃗ B † − ⃗ B ⊗ ⃗ E † = 1 3 δ ijS ψ + iɛ ijk V ψ<br />

k + T ψ ij . (4.5)<br />

The number <strong>of</strong> components and degrees <strong>of</strong> freedom are distributed as<br />

Adding up to a total <strong>of</strong> 36 components.<br />

4.1.1 Σ, the sum<br />

parameter label S V i T ij Sum<br />

Σ 1 3 5 9<br />

∆ 1 3 5 9<br />

χ 1 3 5 9<br />

ψ 1 3 5 9<br />

total 4 12 20 36<br />

The first three parameters S Σ , V Σ and T Σ are labelled with Σ to denote the<br />

sum. The first parameter, S Σ , given by<br />

S Σ = ⃗ E 2 + c 2 ⃗ B 2 = E 2 1 + E 2 2 + E 2 3 + c 2 B 2 1 + c 2 B 2 2 + c 2 B 2 3, (4.6)<br />

is a real scalar and corresponds to the total energy density <strong>of</strong> the fields. The<br />

second parameter, ⃗ V Σ , is a real pseudo-vector that holds the imaginary part <strong>of</strong><br />

the sum. This is easy to see since it is a crossproduct (pseudo-vector) and the<br />

components have the form A − A † . It has the following form<br />

⃗V Σ = ( Im[E 2 E ∗ 3 + c 2 B 2 B ∗ 3], Im[E 3 E ∗ 1 + c 2 B 3 B ∗ 1], Im[E 1 E ∗ 2 + c 2 B 1 B ∗ 2] )<br />

= − i 2( ⃗E × ⃗ E † + c 2 ⃗ B × ⃗ B<br />

† ) . (4.7)<br />

In the case <strong>of</strong> transverse <strong>wave</strong>s, V ⃗ is parallel or anti-parallel to the direction <strong>of</strong><br />

propagation. The third parameter is<br />

⎛<br />

⎞<br />

E1 2 + c 2 B1 2 − 1 3 SΣ Re[E 1 E2 ∗ + c 2 B 1 B2] ∗ Re[E 1 E3 ∗ + c 2 B 1 B3]<br />

∗ Tij Σ =<br />

⎜Re[E 1 E2 ∗ + c 2 B 1 B2] ∗ E2 2 + c 2 B2 2 − 1 3<br />

⎝<br />

SΣ Re[E 2 E3 ∗ + c 2 B 2 B3]<br />

∗ ⎟<br />

⎠ .<br />

Re[E 1 E3 ∗ + c 2 B 1 B3] ∗ Re[E 2 E3 ∗ + c 2 B 2 B3] ∗ E3 2 + c 2 B3 2 − 1 3 SΣ (4.8)<br />

This parameter, Tij Σ , is similar to the Maxwell stress tensor, see Equation (2.95),<br />

the only difference being the 1 3 on the SΣ ’s on the diagonal. This tensor is<br />

traceless (it is made traceless in the definition), this is not the case with the<br />

Maxwell stress tensor, which has the trace equal to the energy density.


4.1. THE IRREDUCIBLE PARAMETERS UNDER SO(3) 33<br />

4.1.2 ∆, the difference<br />

The next label, ∆, corresponds to the difference between the tensor products.<br />

The first one, S ∆ , is the difference in energy<br />

S ∆ = ⃗ E 2 − c 2 ⃗ B 2 = E 2 1 + E 2 2 + E 2 3 − c 2 B 2 1 − c 2 B 2 2 − c 2 B 2 3, (4.9)<br />

and is recognised as the Lagrangian for a free <strong>electromagnetic</strong> field, which is a<br />

known invariant scalar. 2 The vector, ⃗ V ∆ , is given by<br />

⃗V ∆ = ( Im[E 2 E ∗ 3 − c 2 B 2 B ∗ 3], Im[E 3 E ∗ 1 − c 2 B 3 B ∗ 1], Im[E 1 E ∗ 2 − c 2 B 1 B ∗ 2] )<br />

= − i 2( ⃗E × ⃗ E † − c 2 ⃗ B × ⃗ B<br />

† ) . (4.10)<br />

Recalling the V ⃗ defined in Section 2.4, V ⃗ Σ is the difference V ⃗ E − V ⃗ B , i.e. a<br />

measure <strong>of</strong> the difference <strong>of</strong> the <strong>polarization</strong> ellipses. Here we get the tensor to<br />

be<br />

⎛<br />

⎞<br />

E1 2 − c 2 B1 2 − 1 3 S∆ Re[E 1 E2 ∗ − c 2 B 1 B2] ∗ Re[E 1 E3 ∗ − c 2 B 1 B3]<br />

∗ Tij ∆ =<br />

⎜Re[E 1 E2 ∗ − c 2 B 1 B2] ∗ E2 2 − c 2 B2 2 − 1 3<br />

⎝<br />

S∆ Re[E 2 E3 ∗ − c 2 B 2 B3]<br />

∗ ⎟<br />

⎠ .<br />

Re[E 1 E3 ∗ − c 2 B 1 B3] ∗ Re[E 2 E3 ∗ − c 2 B 2 B3] ∗ E3 2 − c 2 B3 2 − 1 3 S∆ (4.11)<br />

It is real, symmetric and traceless as it should be. Again all parameters are real<br />

and V ⃗ ∆ is the imaginary part. Rewriting Tij Σ and T ij ∆ in the same manner as<br />

the Maxwell stress tensor (Eq. (2.95)) we have<br />

T Σ ∆<br />

ij = ( E i E j − 1 3 ⃗ E 2 δ i j ) ± c 2( B i B j − 1 3 ⃗ B 2 δ i j ) , (4.12)<br />

where the plus-sign corresponds to Tij Σ and the minus to T ij ∆ , and it is easy to<br />

see the similarity with the Maxwell stress tensor.<br />

4.1.3 χ, the real part<br />

The label χ corresponds to the real part <strong>of</strong> the cross-correlated components,<br />

i.e. the components containing EB ∗ and BE ∗ parts. The scalar S χ is just two<br />

times the real part <strong>of</strong> the trace <strong>of</strong> Equation (2.97), the correlation matrix for ⃗ E<br />

and ⃗ B, and is given by<br />

S χ = c ⃗ E · ⃗B † + c ⃗ B · ⃗E † = 2Re[c ⃗ E · ⃗B † ], (4.13)<br />

The vector ⃗ V χ corresponds to the antisymmetric imaginary part <strong>of</strong> the correlation<br />

matrix and the imaginary part <strong>of</strong> the Poynting vector and is given by<br />

(<br />

)<br />

⃗V χ = c Im[E 2 B3 ∗ + B 2 E3], ∗ Im[E 3 B1 ∗ + B 3 E1], ∗ Im[E 1 B2 ∗ + B 1 E2]<br />

∗<br />

= −c i 2( ⃗E × ⃗ B † + ⃗ B × ⃗ E †) . (4.14)<br />

2 The Lagrangian is really half <strong>of</strong> this (see Barut [1]) but the 1 is just a constant that can<br />

2<br />

be ommited since we are just interested in covariance <strong>of</strong> the parameters.


34CHAPTER 4. IRREDUCIBLE REPRESENTATION OF POLARIZATION<br />

The tensor, T χ ij , is the symmetric real part <strong>of</strong> Equation (2.97) made traceless<br />

with S χ , and is written<br />

⎛<br />

⎞<br />

2cRe[E 1 B1] ∗ − 1 3 Sχ cRe[E 1 B2 ∗ + E 2 B1] ∗ cRe[E 1 B3 ∗ + E 3 B1]<br />

∗ T χ ij = ⎜cRe[E 1 B2 ∗ + E 2 B1] ∗ 2cRe[E 2 B2] ∗ − 1 3<br />

⎝<br />

Sχ cRe[E 2 B3 ∗ + E 3 B2]<br />

∗ ⎟<br />

⎠ .<br />

cRe[E 1 B3 ∗ + E 3 B1] ∗ cRe[E 2 B3 ∗ + E 3 B2] ∗ 2cRe[E 3 B3] ∗ − 1 3 Sχ<br />

(4.15)<br />

4.1.4 ψ, the imaginary part<br />

The last label, ψ, is a bit different. The scalar S ψ is two times the imaginary<br />

part <strong>of</strong> the trace <strong>of</strong> the correlation matrix ( ⃗ E ⊗ ⃗ B † ), hence<br />

S ψ = c ⃗ E · ⃗B † − c ⃗ B · ⃗E † = 2icIm[ ⃗ E · ⃗B † ]. (4.16)<br />

The vector, V ψ , contains the antisymmetric real part and is given by<br />

(<br />

)<br />

⃗V ψ = −ic Re[E 2 B3 ∗ − E 3 B2], ∗ Re[E 3 B1 ∗ − E 1 B3], ∗ Re[E 1 B2 ∗ − E 2 B1]<br />

∗<br />

= −c i 2( ⃗E × ⃗ B † − ⃗ B × ⃗ E †) . (4.17)<br />

This is also the real part <strong>of</strong> the complex Poynting vector, Equation (2.103). The<br />

tensor T ψ ij holds the symmetric imaginary part and is written<br />

⎛<br />

⎞<br />

2icIm[E 1 B1] ∗ − 1 3 Sψ icIm[E 1 B2 ∗ + E 2 B1] ∗ icIm[E 1 B3 ∗ + E 3 B1]<br />

∗ T ψ ij = ⎜icIm[E 1 B2 ∗ + E 2 B1] ∗ 2icIm[E 2 B2] ∗ − 1 3<br />

⎝<br />

Sψ icIm[E 2 B3 ∗ + E 3 B2]<br />

∗ ⎟<br />

⎠ .<br />

icIm[E 1 B3 ∗ + E 3 B1] ∗ icIm[E 2 B3 ∗ + E 3 B2] ∗ 2icIm[E 3 B3] ∗ − 1 3 Sψ<br />

The total complex Poynting vector is then given by<br />

(4.18)<br />

⃗S = i ⃗ V ψ + i ⃗ V χ . (4.19)<br />

4.2 Table <strong>of</strong> the parameters<br />

There is a total <strong>of</strong> 12 parameters in the form <strong>of</strong> scalars, vectors and tensors.<br />

Here is a table <strong>of</strong> the parameters for a better overview:


4.3. NORMALIZATION OF THE S, V, T PARAMETERS 35<br />

S Σ<br />

S ∆<br />

Total energy density<br />

Lagrangian for a free <strong>electromagnetic</strong> field<br />

S χ Re[cE ⃗ · ⃗B † ]<br />

S ψ Im[cE ⃗ · ⃗B † ]<br />

⃗V Σ<br />

⃗V ∆<br />

⃗V χ<br />

⃗V ψ<br />

Tij<br />

Σ<br />

Tij<br />

∆<br />

T χ ij<br />

Sum <strong>of</strong> the ⃗v vectors (see Section 2.3) for E ⃗ and B. ⃗ Sum <strong>of</strong> the<br />

<strong>polarization</strong> ellipses.<br />

Difference <strong>of</strong> the ⃗v vectors for E ⃗ and B. ⃗ Difference<br />

<strong>of</strong> the <strong>polarization</strong> ellipses<br />

Imaginary part <strong>of</strong> the complex Poynting vector.<br />

Real part <strong>of</strong> the complex Poynting vector.<br />

Similar to Maxwell stress tensor but traceless.<br />

Similar to Tij Σ but corresponds to the difference <strong>of</strong> the tensors for E ⃗ and B ⃗<br />

Not found in the literature. Real part <strong>of</strong> a cross-correlated<br />

version <strong>of</strong> Maxwell stress tensor.<br />

T ψ ij<br />

Not found in the literature. The imaginary part to T χ ij<br />

The last two are <strong>of</strong> special interesting since to the best <strong>of</strong> our knowledge they<br />

have not previously been defined or used in the literature.<br />

4.3 Normalization <strong>of</strong> the S, V, T parameters<br />

The parameter S Σ , the total energy density, is the only parameter that is a<br />

positive real number for any <strong>electromagnetic</strong> field. Using s, v and t as mnemonics<br />

for the normalized parameters, we have<br />

s Σ = SΣ<br />

S Σ = 1,<br />

⃗v Σ = ⃗ V Σ<br />

S Σ ,<br />

t Σ ij = T Σ ij<br />

S Σ ,<br />

s∆ = S∆<br />

S Σ ,<br />

⃗v∆ = ⃗ V ∆<br />

S Σ ,<br />

t∆ ij = T ∆ ij<br />

S Σ ,<br />

sχ = Sχ<br />

S Σ ,<br />

⃗vχ = ⃗ V χ<br />

S Σ ,<br />

tχ ij = T χ ij<br />

S Σ ,<br />

sψ = Sψ<br />

S Σ , (4.20)<br />

⃗vψ = ⃗ V ψ<br />

S Σ , (4.21)<br />

tψ ij = T ψ ij<br />

S Σ . (4.22)<br />

These are normalized versions where the components <strong>of</strong> the parameters have a<br />

maximum absolute value <strong>of</strong> one.<br />

4.3.1 Example: the refractive index<br />

In this section we use some complex vector algebra. A good source <strong>of</strong> information<br />

on this subject is Lindell [14]. This example illustrates how to determine<br />

the refractive index vector ⃗n which is then expressed in terms <strong>of</strong> the parameters<br />

S, V and T . The refractive index is defined by<br />

⃗n ≡ c ω ⃗ k. (4.23)<br />

The Fourier transformed version <strong>of</strong> Faraday’s law is given by<br />

⃗ k × ⃗ E = ω ⃗ B ⇔ ⃗n × ⃗ E = c ⃗ B. (4.24)


36CHAPTER 4. IRREDUCIBLE REPRESENTATION OF POLARIZATION<br />

The condition <strong>of</strong> a divergence free magnetic field is easily seen from this equation.<br />

By taking the scalar product with ⃗n, from the left we obtain<br />

⃗n · (⃗n × ⃗ E) = c⃗n · ⃗B ⇒ ⃗n · ⃗B = 0, (4.25)<br />

where we have used the vector relations ⃗a · ( ⃗ b × ⃗c) = (⃗a × b) · ⃗c and ⃗n × ⃗n = 0.<br />

We also multiply with ⃗n † , from the left, hence<br />

⃗n † · (⃗n × ⃗ E) = c⃗n † · ⃗B ⇒ ⃗n † · ⃗B = 0. (4.26)<br />

The condition for the last equation to be zero is that ⃗n † ×⃗n = 0. This is fulfilled<br />

if ⃗n can be written ⃗n = n⃗r, where n is a complex scalar and ⃗r is a real vector.<br />

This means that the refractive index is the same in all direction, which is true<br />

for a homogeneous medium. To get an expression for ⃗n we use Faraday’s law<br />

cross multiplied from the right with ⃗ B †<br />

(⃗n × ⃗ E) × ⃗ B † = c ⃗ B × ⃗ B † , (4.27)<br />

which leads to<br />

−⃗n( ⃗ B † · ⃗E) + ⃗ E( ⃗ B † · ⃗n) = c ⃗ B × ⃗ B † . (4.28)<br />

If we restrict ourself to the case ⃗n × ⃗n † = 0 (homogeneous medium), the second<br />

term on the left hand side vanishes and we are left with<br />

which for ⃗ E · ⃗B † ≠ 0 can be rewritten as<br />

−⃗n( ⃗ E · ⃗B † ) = c ⃗ B × ⃗ B † , (4.29)<br />

B<br />

⃗n = −c ⃗ × B ⃗ †<br />

. (4.30)<br />

⃗E · ⃗B<br />

†<br />

If we compare this expression with the parameters defined earlier in this chapter<br />

we have<br />

⃗B × ⃗ B † = i<br />

c 2 ( ⃗V Σ − ⃗ V ∆) , (4.31)<br />

and<br />

Inserting this in the expression for ⃗n we get<br />

⃗E · ⃗B † = 1 2c(<br />

S χ + S ψ) . (4.32)<br />

⃗n = −2 i( ⃗ V Σ − ⃗ V ∆)<br />

S χ + S ψ . (4.33)<br />

Given the S, V and T parameters this is a simple calculation.<br />

4.4 Stokes parameters from the S, V, T parameters<br />

In the two dimensional case, with E 3 = 0 and only considering the electric field<br />

(putting ⃗ B = 0), we have Stokes parameters. In terms <strong>of</strong> the S, V, T parameters


4.4. STOKES PARAMETERS FROM THE S, V, T PARAMETERS 37<br />

they can be written<br />

I = F 2 x + F 2 y = S Σ , (4.34)<br />

Q = F 2 x − F 2 y = T Σ 11 − T Σ 22, (4.35)<br />

U = 2Re[F y F ∗ x ] = 2T Σ 12, (4.36)<br />

V = 2Im[F y F ∗ x ] = −2V Σ<br />

3 , (4.37)<br />

where Σ can be replaced by ∆, since the magnetic field, ⃗ B , is zero. It is similar<br />

for the <strong>polarization</strong> <strong>of</strong> the magnetic field. The two dimensional spectral density<br />

matrix (Eq. (2.53)) becomes<br />

( )<br />

Fx Fx S d = ∗ F x Fy<br />

∗<br />

F y Fx<br />

∗ F y Fy<br />

∗ = 1 2<br />

( )<br />

S Σ + T11 Σ − T22 Σ 2T12 Σ + i2V3<br />

Σ<br />

2T12 Σ − i2V3 Σ S Σ − T11 Σ + T22<br />

Σ , (4.38)<br />

where we have put in the Stokes parameters in the S, V, T parameter form. We<br />

can simplify this to<br />

( )<br />

T<br />

Σ<br />

S d = 11 T12 Σ + iV3<br />

Σ<br />

T12 Σ − iV3 Σ T22<br />

Σ . (4.39)


38CHAPTER 4. IRREDUCIBLE REPRESENTATION OF POLARIZATION


Chapter 5<br />

Covariant <strong>polarization</strong><br />

tensors in E,B-component<br />

form<br />

In Chapter 5 we saw that the S, V, T parameters can be found in Lorentz <strong>covariant</strong><br />

tensors. These tensors were constructed according to<br />

where<br />

is the metric,<br />

and<br />

1 X µν = F αµ (F αδ ) ∗ g δν , (5.1)<br />

2 X µν = F αµ (F αδ ) ∗ g δν , (5.2)<br />

3 X µν = F αµ (F αδ ) ∗ g δν , (5.3)<br />

4 X µν = F αµ (F αδ ) ∗ g δν , (5.4)<br />

⎛<br />

⎞<br />

1 0 0 0<br />

g µν = g µν = ⎜0 −1 0 0<br />

⎟<br />

⎝0 0 −1 0 ⎠ , (5.5)<br />

0 0 0 −1<br />

⎛<br />

⎞<br />

0 −E 1 −E 2 −E3<br />

F µν = ⎜E 1 0 −cB 3 cB 2<br />

⎟<br />

⎝E 2 cB 3 0 −cB 1<br />

⎠ (5.6)<br />

E 3 −cB 2 cB 1 0<br />

⎛<br />

⎞<br />

0 E 1 E 2 E 3<br />

F µν = ⎜−E 1 0 −cB 3 cB 2<br />

⎟<br />

⎝−E 2 cB 3 0 −cB 1<br />

⎠ (5.7)<br />

−E 3 −cB 2 cB 1 0<br />

are the contravariant and <strong>covariant</strong> field strength tensors, respectively, and<br />

⎛<br />

⎞<br />

0 −cB 1 −cB 2 −cB 3<br />

F µν = ⎜cB 1 0 E 3 −E 2<br />

⎟<br />

⎝cB 2 −E 3 0 E 1<br />

⎠ (5.8)<br />

cB 3 E 2 −E 1 0<br />

39


40 CHAPTER 5. COVARIANT POLARIZATION TENSORS<br />

and<br />

⎛<br />

⎞<br />

0 cB 1 cB 2 cB 3<br />

F µν = ⎜−cB 1 0 E 3 −E 2<br />

⎟<br />

⎝−cB 2 −E 3 0 E 1<br />

⎠ (5.9)<br />

−cB 3 E 2 −E 1 0<br />

are the corresponding dual tensors.<br />

5.1 The X tensors<br />

In E, B-component form the n X µν tensors are given by<br />

1 X µν =<br />

[−E 1 E ∗ 1 − E 2 E ∗ 2 − E 3 E ∗ 3, −E 2 cB ∗ 3 + E 3 cB ∗ 2, E 1 cB ∗ 3 − E 3 cB ∗ 1, −E 1 cB ∗ 2 + E 2 cB ∗ 1]<br />

[−cB 3 E ∗ 2 + cB 2 E ∗ 3, E 1 E ∗ 1 − c 2 B 3 B ∗ 3 − c 2 B 2 B ∗ 2, E 1 E ∗ 2 + c 2 B 2 B ∗ 1, E 1 E ∗ 3 + c 2 B 3 B ∗ 1]<br />

[cB 3 E ∗ 1 − cB 1 E ∗ 3, E 2 E ∗ 1 + c 2 B 1 B ∗ 2, E 2 E ∗ 2 − c 2 B 3 B ∗ 3 − c 2 B 1 B ∗ 1, E 2 E ∗ 3 + c 2 B 3 B ∗ 2]<br />

[−cB 2 E ∗ 1 +cB 1 E ∗ 2, E 3 E ∗ 1 +c 2 B 1 B ∗ 3, E 3 E ∗ 2 +c 2 B 2 B ∗ 3, E 3 E ∗ 3 −c 2 B 2 B ∗ 2 −c 2 B 1 B ∗ 1]<br />

(5.10)<br />

2 X µν =<br />

[−c 2 B 1 B ∗ 1−c 2 B 2 B ∗ 2−c 2 B 3 B ∗ 3, −cB 3 E ∗ 2+cB 2 E ∗ 3, cB 3 E ∗ 1−cB 1 E ∗ 3, −cB 2 E ∗ 1+cB 1 E ∗ 2]<br />

[−E 2 cB ∗ 3 + E 3 cB ∗ 2, c 2 B 1 B ∗ 1 − E 3 E ∗ 3 − E 2 E ∗ 2, E 2 E ∗ 1 + c 2 B 1 B ∗ 2, E 3 E ∗ 1 + c 2 B 1 B ∗ 3]<br />

[E 1 cB ∗ 3 − E 3 cB ∗ 1, E 1 E ∗ 2 + c 2 B 2 B ∗ 1, c 2 B 2 B ∗ 2 − E 3 E ∗ 3 − E 1 E ∗ 1, E 3 E ∗ 2 + c 2 B 2 B ∗ 3]<br />

[−E 1 cB ∗ 2 + E 2 cB ∗ 1, E 1 E ∗ 3 + c 2 B 3 B ∗ 1, E 2 E ∗ 3 + c 2 B 3 B ∗ 2, c 2 B 3 B ∗ 3 − E 2 E ∗ 2 − E 1 E ∗ 1]<br />

(5.11)<br />

3 X µν =<br />

[−cB 1 E ∗ 1−cB 2 E ∗ 2−cB 3 E ∗ 3, −c 2 B 2 B ∗ 3+c 2 B 3 B ∗ 2, c 2 B 1 B ∗ 3−c 2 B 3 B ∗ 1, −c 2 B 1 B ∗ 2+c 2 B 2 B ∗ 1]<br />

[E 3 E ∗ 2 − E 2 E ∗ 3, cB 1 E ∗ 1 + E 3 cB ∗ 3 + E 2 cB ∗ 2, cB 1 E ∗ 2 − E 2 cB ∗ 1, cB 1 E ∗ 3 − E 3 cB ∗ 1]<br />

[−E 3 E ∗ 1 + E 1 E ∗ 3, cB 2 E ∗ 1 − E 1 cB ∗ 2, cB 2 E ∗ 2 + E 3 cB ∗ 3 + E 1 cB ∗ 1, cB 2 E ∗ 3 − E 3 cB ∗ 2]<br />

[E 2 E ∗ 1 − E 1 E ∗ 2, cB 3 E ∗ 1 − E 1 cB ∗ 3, cB 3 E ∗ 2 − E 2 cB ∗ 3, cB 3 E ∗ 3 + E 2 cB ∗ 2 + E 1 cB ∗ 1]<br />

(5.12)<br />

4 X µν =<br />

[−E 1 cB ∗ 1 − E 2 cB ∗ 2 − E 3 cB ∗ 3, −E 3 E ∗ 2 + E 2 E ∗ 3, E 3 E ∗ 1 − E 1 E ∗ 3, −E 2 E ∗ 1 + E 1 E ∗ 2]<br />

[c 2 B 2 B ∗ 3 −c 2 B 3 B ∗ 2, E 1 cB ∗ 1 +cB 3 E ∗ 3 +cB 2 E ∗ 2, −cB 2 E ∗ 1 +E 1 cB ∗ 2, −cB 3 E ∗ 1 +E 1 cB ∗ 3]<br />

[−c 2 B 1 B ∗ 3+c 2 B 3 B ∗ 1, −cB 1 E ∗ 2+E 2 cB ∗ 1, cB 3 E ∗ 3+E 2 cB ∗ 2+cB 1 E ∗ 1, −cB 3 E ∗ 2+E 2 cB ∗ 3]<br />

[c 2 B 1 B ∗ 2−c 2 B 2 B ∗ 1, −cB 1 E ∗ 3+E 3 cB ∗ 1, −cB 2 E ∗ 3+E 3 cB ∗ 2, cB 2 E ∗ 2+E 3 cB ∗ 3+cB 1 E ∗ 1]<br />

(5.13)


5.2. THE Y TENSORS 41<br />

5.2 The Y tensors<br />

The Y tensors, constructed by the combinations<br />

Σ Y µν = 1 2<br />

( 1<br />

X µν + 2 X µν )<br />

,<br />

χ Y µν = 1 ( 3<br />

X µν + 4 X µν )<br />

,<br />

2<br />

∆ Y µν = 1 ( 1<br />

X µν − 2 X µν )<br />

(5.14)<br />

,<br />

2<br />

ψ Y µν = 1 ( 3<br />

X µν − 4 X µν )<br />

,<br />

2<br />

and are then found to be<br />

Σ Y µν = 1 2<br />

[−E 1 E ∗ 1−E 2 E ∗ 2−E 3 E ∗ 3−cB 1 cB ∗ 1−cB 2 cB ∗ 2−cB 3 cB ∗ 3, −cB 3 E ∗ 2+cB 2 E ∗ 3−E 2 cB ∗ 3+E 3 cB ∗ 2,<br />

E 1 cB ∗ 3 − E 3 cB ∗ 1 + cB 3 E ∗ 1 − cB 1 E ∗ 3, −E 1 cB ∗ 2 + E 2 cB ∗ 1 − cB 2 E ∗ 1 + cB 1 E ∗ 2]<br />

[−cB 3 E ∗ 2+cB 2 E ∗ 3−E 2 cB ∗ 3+E 3 cB ∗ 2, E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2+cB 1 cB ∗ 1−E 3 E ∗ 3−E 2 E ∗ 2,<br />

E 2 E ∗ 1 + cB 1 cB ∗ 2 + E 1 E ∗ 2 + cB 2 cB ∗ 1, E 3 E ∗ 1 + cB 1 cB ∗ 3 + E 1 E ∗ 3 + cB 3 cB ∗ 1]<br />

[E 1 cB ∗ 3 − E 3 cB ∗ 1 + cB 3 E ∗ 1 − cB 1 E ∗ 3, E 2 E ∗ 1 + cB 1 cB ∗ 2 + E 1 E ∗ 2 + cB 2 cB ∗ 1,<br />

E 2 E ∗ 2−cB 3 cB ∗ 3−cB 1 cB ∗ 1+cB 2 cB ∗ 2−E 3 E ∗ 3−E 1 EC 1 , E 3 E ∗ 2+cB 2 cB ∗ 3+E 2 E ∗ 3+cB 3 cB ∗ 2]<br />

[−E 1 cB ∗ 2 + E 2 cB ∗ 1 − cB 2 E ∗ 1 + cB 1 E ∗ 2, E 3 E ∗ 1 + cB 1 cB ∗ 3 + E 1 E ∗ 3 + cB 3 cB ∗ 1,<br />

E 3 E ∗ 2+cB 2 cB ∗ 3+E 2 E ∗ 3+cB 3 cB ∗ 2, E 3 E ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+cB 3 cB ∗ 3−E 2 E ∗ 2−E 1 E ∗ 1]<br />

(5.15)<br />

χ Y µν = 1 2<br />

[−cB 1 E ∗ 1−cB 2 E ∗ 2−cB 3 E ∗ 3−E 1 cB ∗ 1−E 2 cB ∗ 2−E 3 cB ∗ 3, −cB 2 cB ∗ 3+cB 3 cB ∗ 2−E 3 E ∗ 2+E 2 E ∗ 3,<br />

cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3, −cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2]<br />

[E 3 E ∗ 2−E 2 E ∗ 3+cB 2 cB ∗ 3−cB 3 cB ∗ 2, cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1+cB 3 E ∗ 3+cB 2 E ∗ 2,<br />

cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2, cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3]<br />

[−E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1, cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1,<br />

cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1+cB 3 E ∗ 3+cB 2 E ∗ 2, cB 2 E ∗ 3−E 3 cB ∗ 2−cB 3 E ∗ 2+E 2 cB ∗ 3]<br />

[E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1, cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1,<br />

cB 3 E ∗ 2−E 2 cB ∗ 3−cB 2 E ∗ 3+E 3 cB ∗ 2, cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1+cB 3 E ∗ 3+cB 2 E ∗ 2]<br />

(5.16)


42 CHAPTER 5. COVARIANT POLARIZATION TENSORS<br />

∆ Y µν = 1 2<br />

[−E 1 E ∗ 1−E 2 E ∗ 2−E 3 E ∗ 3+cB 1 cB ∗ 1+cB 2 cB ∗ 2+cB 3 cB ∗ 3, cB 3 E ∗ 2−E 2 cB ∗ 3−cB 2 E ∗ 3+E 3 cB ∗ 2,<br />

cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3, cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1]<br />

[cB 2 E ∗ 3−E 3 cB ∗ 2−cB 3 E ∗ 2+E 2 cB ∗ 3, E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+E 3 E ∗ 3+E 2 E ∗ 2,<br />

− cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2, −E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1]<br />

[cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1, E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1,<br />

E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+E 3 E ∗ 3+E 2 E ∗ 2, −cB 2 cB ∗ 3+cB 3 cB ∗ 2−E 3 E ∗ 2+E 2 E ∗ 3]<br />

[cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2, cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3,<br />

E 3 E ∗ 2−E 2 E ∗ 3+cB 2 cB ∗ 3−cB 3 cB ∗ 2, E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+E 3 E ∗ 3+E 2 E ∗ 2]<br />

(5.17)<br />

ψ Y µν = 1 2<br />

[−cB 1 E ∗ 1−cB 2 E ∗ 2−cB 3 E ∗ 3+E 1 cB ∗ 1+E 2 cB ∗ 2+E 3 cB ∗ 3, E 3 E ∗ 2−E 2 E ∗ 3−cB 2 cB ∗ 3+cB 3 cB ∗ 2,<br />

cB 1 cB ∗ 3 − cB 3 cB ∗ 1 − E 3 E ∗ 1 + E 1 E ∗ 3, −cB 1 cB ∗ 2 + cB 2 cB ∗ 1 + E 2 E ∗ 1 − E 1 E ∗ 2]<br />

[E 3 E ∗ 2−E 2 E ∗ 3−cB 2 cB ∗ 3+cB 3 cB ∗ 2, cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2−E 1 cB ∗ 1−cB 3 E ∗ 3−cB 2 E ∗ 2,<br />

cB 2 E ∗ 1 − E 1 cB ∗ 2 + cB 1 E ∗ 2 − E 2 cB ∗ 1, cB 3 E ∗ 1 − E 1 cB ∗ 3 + cB 1 E ∗ 3 − E 3 cB ∗ 1]<br />

[cB 1 cB ∗ 3 − cB 3 cB ∗ 1 − E 3 E ∗ 1 + E 1 E ∗ 3, cB 2 E ∗ 1 − E 1 cB ∗ 2 + cB 1 E ∗ 2 − E 2 cB ∗ 1,<br />

cB 2 E ∗ 2+E 3 cB ∗ 3+E 1 cB ∗ 1−cB 3 E ∗ 3−E 2 cB ∗ 2−cB 1 E ∗ 1, cB 3 E ∗ 2−E 2 cB ∗ 3+cB 2 E ∗ 3−E 3 cB ∗ 2]<br />

[−cB 1 cB ∗ 2 + cB 2 cB ∗ 1 + E 2 E ∗ 1 − E 1 E ∗ 2, cB 3 E ∗ 1 − E 1 cB ∗ 3 + cB 1 E ∗ 3 − E 3 cB ∗ 1,<br />

cB 3 E ∗ 2−E 2 cB ∗ 3+cB 2 E ∗ 3−E 3 cB ∗ 2, cB 3 E ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1−cB 2 E ∗ 2−E 3 cB ∗ 3−cB 1 E ∗ 1]<br />

(5.18)<br />

5.3 The K and K tensors<br />

The K and K tensors are formed by removing the trace from ∆ Y µν and χ Y µν ,<br />

respectively, and are given by<br />

K µν = 1 2<br />

[0, cB 3 E ∗ 2 − E 2 cB ∗ 3 − cB 2 E ∗ 3 + E 3 cB ∗ 2, cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3,<br />

cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1]<br />

[cB 2 E ∗ 3 − E 3 cB ∗ 2 − cB 3 E ∗ 2 + E 2 cB ∗ 3, 0, −cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2,<br />

− E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1]<br />

[cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1, E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1, 0,<br />

− cB 2 cB ∗ 3 + cB 3 cB ∗ 2 − E 3 E ∗ 2 + E 2 E ∗ 3]<br />

[cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2, cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3,<br />

E 3 E ∗ 2 − E 2 E ∗ 3 + cB 2 cB ∗ 3 − cB 3 cB ∗ 2, 0] (5.19)


5.3. THE K AND K TENSORS 43<br />

and<br />

K µν = 1 2<br />

[0, −cB 2 cB ∗ 3 + cB 3 cB ∗ 2 − E 3 E ∗ 2 + E 2 E ∗ 3, cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3,<br />

− cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2]<br />

[E 3 E ∗ 2 − E 2 E ∗ 3 + cB 2 cB ∗ 3 − cB 3 cB ∗ 2, 0, cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2,<br />

cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3]<br />

[−E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1, cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1, 0,<br />

cB 2 E ∗ 3 − E 3 cB ∗ 2 − cB 3 E ∗ 2 + E 2 cB ∗ 3]<br />

[E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1, cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1,<br />

respectively.<br />

cB 3 E ∗ 2 − E 2 cB ∗ 3 − cB 2 E ∗ 3 + E 3 cB ∗ 2, 0], (5.20)


44 CHAPTER 5. COVARIANT POLARIZATION TENSORS


Chapter 6<br />

Reducible representation <strong>of</strong><br />

the <strong>covariant</strong> spectral<br />

density tensor<br />

In Chapter 3 we introduced the <strong>covariant</strong> spectral density tensor, τ αβγδ , as<br />

the tensor product <strong>of</strong> the field strength tensor with its complex conjugate,<br />

F αβ (F γδ ) ∗ . This tensor contains all correlation components <strong>of</strong> the <strong>electromagnetic</strong><br />

field and it is manifestly Lorentz <strong>covariant</strong>. One would like to decompose<br />

the <strong>covariant</strong> spectral density tensor in an irreducible tensor representation (see<br />

Sections 2.3 and 2.4). To keep it simple one can begin with a reducible representation,<br />

which can be expanded into an irreducible representation. In this<br />

Chapter we decompose the <strong>covariant</strong> spectral density tensor, into a reducible representation<br />

consisting <strong>of</strong> irreducible tensors The irreducible tensors are identified<br />

as known scalars and tensors. The <strong>covariant</strong> <strong>polarization</strong> tensors in Chapter 5<br />

are reconstructed directly from the <strong>covariant</strong> spectral density tensor.<br />

6.1 The <strong>covariant</strong> spectral density tensor<br />

As shown in Chapter 3, it is the possible to construct a rank four tensor, τ αβγδ =<br />

F αβ (F γδ ) ∗ , which consists <strong>of</strong> quadratic components. This tensor is the <strong>covariant</strong><br />

spectral density tensor, and it contains all quadratic correlation components <strong>of</strong><br />

the <strong>electromagnetic</strong> field.<br />

Since τ αβγδ is the tensor product <strong>of</strong> <strong>of</strong> the field strength tensor with itself,<br />

it has some convenient symmetries and zeroes:<br />

τ αβγδ = (τ γδαβ ) ∗ , (6.1)<br />

τ αβγδ = −τ βαγδ = −τ αβδγ = τ βαδγ , (6.2)<br />

τ αβγδ = 0 , if α = β or(and) γ = β. (6.3)<br />

45


46 CHAPTER 6. REDUCIBLE REPRESENTATION OF τ αβγδ<br />

6.2 The irreducible tensors<br />

The decomposition we make is in terms <strong>of</strong> proper Lorentz tensors, i.e. the<br />

tensors are manifestly <strong>covariant</strong> under Lorentz transformations. The decomposition<br />

is similar to Equation (2.123) but here we construct the tensors in the<br />

decomposition from the original tensor. By considering which tensors we can<br />

construct using τ αβγδ and the invariant tensors <strong>of</strong> the group SO(3, 1) (see Section<br />

2.4 Equation (2.44)) we obtain a decomposition <strong>of</strong> the tensor. From the<br />

<strong>covariant</strong> spectral density we form the following tensors<br />

(p) S = τ αβγδ ɛ αβγδ ,<br />

S = τ αβγδ g αγ g βδ ,<br />

T αγ = τ αβγδ g βδ − 1 4 Sgαγ ,<br />

(6.4)<br />

(p) T µν = τ αβγδ g αγ ɛ βδµν ,<br />

where S is an ordinary scalar, (p) S is a pseudo-scalar, T αγ and (p) T µν are tensor<br />

and pseudo tensor, respectively. We can construct several similar tensors, but<br />

due to the symmetries <strong>of</strong> τ αβγδ several <strong>of</strong> them will be zero and for simplicity,<br />

we decompose τ αβγδ into a reducible representation.<br />

Calculating the scalars, S and (p) S, in the representation, they are found to<br />

be<br />

S = −2(E 2 − c 2 B 2 ),<br />

(p) S = 4( ⃗ E · ⃗B † + ⃗ B · ⃗E † ).<br />

(6.5)<br />

These are recognised as the Lagrangian for a free <strong>electromagnetic</strong> field and the<br />

real part <strong>of</strong> the scalar product ⃗ E · ⃗B † , respectively, both known invariants. The<br />

first tensor, T αγ , is the complex <strong>electromagnetic</strong> energy-momentum tensor (see<br />

Section 2.1.4 Equation (2.44)). The second tensor, (p) T µν , is found to be −2K µν ,<br />

a tensor introduced in the previous Chapter (Chapter 5, Equation (??)).<br />

In Chapter 5 we introduced ten <strong>covariant</strong> <strong>polarization</strong> tensors, the X and Y<br />

tensors together with the two tenors K and K. These tensors can be constructed<br />

from the <strong>covariant</strong> spectral density tensor directly, using the metric tensor, g µν<br />

and the totally anti-symmetric tensor, ɛ µναβ . In terms <strong>of</strong> the <strong>covariant</strong> spectral<br />

density tensor, 3 X νδ and 4 X νδ (Eq. (??)) are given by<br />

and<br />

3 X νδ = F αν (F αµ ) ∗ g µδ = 1 2 g µγɛ µν<br />

αβ τ αβγδ (6.6)<br />

4 X νδ = F αν (F αµ ) ∗ g µδ = 1 2 g αµɛ µν<br />

γδ τ αβγδ . (6.7)<br />

From these expressions we get the expressions for χ Y νδ (plus sign) and ψ Y νδ<br />

(minus sign), and we have<br />

χ νδ 1( ψ Y = 3<br />

X νδ ± 4 X νδ ) 1 =<br />

2<br />

4 τ αβγδ( g µγ ɛ µν<br />

αβ ± g αµɛ µν )<br />

γδ . (6.8)<br />

For 1 X and 2 X, the expressions are<br />

1 X νδ = F αν F ∗ αµg µδ = τ ανβδ g αβ (6.9)


6.3. REDUCIBLE REPRESENTATION 47<br />

and<br />

2 X νδ = F αν Fαµg ∗ µδ = 1 4 ɛαν α ′ β ′ɛ α δ<br />

γ ′ σ ′τ α′ β ′ γ ′ σ ′ , (6.10)<br />

where the prime, ” ′ ” denotes dummy indices. The expression for the corresponding<br />

Y tensors is<br />

Σ νδ 1( ∆ Y = 3<br />

X νδ ± 4 X νδ ) 1 =<br />

2<br />

2 τ α′ β ′ γ ′ δ ′( gβ ν ′gδ δ ′g α ′ γ ′ ± 1 4 ɛαν α ′ β ′ɛ α δ )<br />

δ ′ γ . (6.11)<br />

′<br />

These equations are examples <strong>of</strong> the variety <strong>of</strong> tensors one is able to construct,<br />

just using the metric tensor and the totally anti-symmetric tensor and combine<br />

them into new tensors.<br />

6.3 Reducible representation<br />

In this Chapter we use the <strong>covariant</strong> spectral density tensor with purely contravariant<br />

components. The tensors and decompositions can clearly be altered<br />

into any mixed or <strong>covariant</strong> form using the metric, g µν . Having all indices in<br />

the same position (upper or lower) makes it simple to combine the elements in a<br />

correct manner, in contrast to a mixed representation where one must be careful<br />

in which order one has the indices. Furthermore it is only in the purely <strong>covariant</strong><br />

or contravariant notation one can form symmetric and anti-symmetric parts<br />

directly by adding and subtracting the transposed tensor (see Section 2.4). One<br />

reason to use a mixed representation is that the metric reduces to the Kronecker<br />

delta, g µ ν = δ µ ν , which can be more convenient in some cases. 1<br />

The tensors in Equation (6.4) are, due to their construction, manifestly<br />

<strong>covariant</strong> and irreducible under SO(3, 1). With these tensor we can decompose<br />

τ αβγδ in terms <strong>of</strong> irreducible tensors and a rest tensor, which will be reducible.<br />

We have<br />

τ αβγδ = F αβ (F γδ ) ∗ =<br />

1<br />

16 Sgαγ g βδ + 1 4! S(p) ɛ αβγδ + 1 (1) T αγ g βδ + 1 (2) T µν g αγ ɛ βδµν + Ψ αβγδ . (6.12)<br />

4 8<br />

The rest tensor Ψ will be a proper rank four tensor since all other tensors are<br />

proper tensors, and it is formed by a linear combination <strong>of</strong> proper tensors.<br />

Since τ αβγδ is constructed from the field strength tensor it is antisymmetric<br />

in the first two indices and also in the last two, but is symmetric between the<br />

first two indices and the last two indices. By looking at the symmetries we can<br />

see how we can decompose Ψ further, using similar tensors but permuting the<br />

indices. 2 This process <strong>of</strong> performing a simple decomposition to get a reduced<br />

rest tensor, can be applied to a general rank four tensor.<br />

The <strong>covariant</strong> spectral density tensor is a basic and fundamental object<br />

considering <strong>polarization</strong>. Because <strong>of</strong> its construction the components <strong>of</strong> this<br />

tensor only contains one correlation component, and from it one can construct<br />

a variety <strong>of</strong> tensors consisting <strong>of</strong> different components.<br />

1 See Sections 2.1 and 2.4 for more details on mixed form.<br />

2 For further info and examples <strong>of</strong> this ”index gymnastics” see Misner, Thorne & Wheeler [7]<br />

(”the big black book”) or Schutz [21].


48 CHAPTER 6. REDUCIBLE REPRESENTATION OF τ αβγδ


Chapter 7<br />

Conclusions<br />

The purpose <strong>of</strong> this work was to derive <strong>covariant</strong> a representation for the <strong>polarization</strong><br />

properties <strong>of</strong> an <strong>electromagnetic</strong> field. I.e. the parameters in the<br />

representation should have known transformation properties. The basic idea<br />

has been to make a decomposition <strong>of</strong> each <strong>of</strong> the correlation matrices into a<br />

scalar, a vector and a tensor, an irreducible tensor representation under SO(3).<br />

Performing this decomposition, one gets parameters with components very similar<br />

to the generalized <strong>polarization</strong> parameters, defined by Carozzi, Karlsson<br />

and Bergman [6] who used the generators <strong>of</strong> the SU(3) symmetry group (see<br />

Gell-Mann et. al. [17]).<br />

The representation <strong>of</strong> the <strong>electromagnetic</strong> field given by Storey and Lefeuvre<br />

[23] uses a six dimensional ”quasi-vector” which gives a 6×6 spectral density<br />

matrix, with 36 correlation parameters. The problem with this representation<br />

is that the six-vector is not a genuine vector (see Section 2.2) and that it does<br />

not give a geometrical interpretation for the field.<br />

The decomposition in Chapter 3 has some similar problems. We obtain 36<br />

parameters but they are not combined in such a way that one can use them to<br />

get a geometrical interpretation <strong>of</strong> how the field behaves.<br />

The new set <strong>of</strong> <strong>polarization</strong> parameters introduced and defined in Chapter<br />

4 does not exhibit these problems. These parameters represent well known geometrical<br />

quantities, <strong>covariant</strong> under the group SO(3), and they have separately<br />

been found in the literature, i.e. they have a known physical meaning. E.g., the<br />

Poynting vector and the total energy density <strong>of</strong> the field. Two <strong>of</strong> the parameters<br />

have not been found in the literature. These are the tensors T χ and T ψ . These<br />

parameters has so-far no physical interpretation but can be thought <strong>of</strong> like two<br />

cross-correlated Maxwell stress tensors.<br />

A set <strong>of</strong> Lorentz <strong>covariant</strong> tensors, i.e. <strong>covariant</strong> under the group SO(3, 1),<br />

are constructed in Chapter 5, the Y tensors. These tensors have the S, V, T<br />

parameters as components and the tensors are valid in any curved space-time.<br />

In Chapter 3 we introduce and define the <strong>covariant</strong> spectral density tensor, a<br />

manifestly Lorentz <strong>covariant</strong> rank four tensor, consisting <strong>of</strong> all quadratic correlation<br />

components. The Y tensors can be constructed from the <strong>covariant</strong> spectral<br />

density tensor, τ αβγδ (Chapters 3 and 6), using the algebra <strong>of</strong> general relativity.<br />

The <strong>covariant</strong> spectral density tensor is a more fundamental object than the<br />

Y tensors and for a further investigation <strong>of</strong> <strong>polarization</strong> τ αβγδ is a favourable<br />

object to consider.<br />

49


50 CHAPTER 7. CONCLUSIONS<br />

The new parameters defined in this work give a representation that consists<br />

<strong>of</strong> SO(3)-<strong>covariant</strong>, and in some cases relativistically <strong>covariant</strong>, parameters that<br />

give a full description <strong>of</strong> the <strong>polarization</strong> state <strong>of</strong> an <strong>electromagnetic</strong> <strong>wave</strong>.<br />

From the relativistic tensors in Chapter 5 we deduce how the S, V, T parameters<br />

transform under proper Lorentz transformations.<br />

In the calculations we have considered deterministic fields, which means that<br />

we have a field which is completely known. In measurements one considers bandlimited<br />

stochastical fields, i.e., one considers the correlation in a bandwidth, in<br />

contrast to the deterministic fields considered in this work. In this process <strong>of</strong> averaging<br />

over a bandwidth, The formulas for S, V, T parameters are not affected<br />

which means that the parameters are directly applicable to measurements.<br />

A direct continuation <strong>of</strong> this work would be to find a physical interpretation<br />

<strong>of</strong> T χ and T ψ , the previously unknown tensors, and to calculate the parameters<br />

for different <strong>wave</strong>s with known dispersion relations. This would make the parameters<br />

useful in identification <strong>of</strong> <strong>wave</strong> modes in space plasma or other media<br />

supporting <strong>electromagnetic</strong> <strong>wave</strong> propagation. Also lacking is the irreducble<br />

representation <strong>of</strong> the <strong>covariant</strong> spectral density tensor and calculation <strong>of</strong> all<br />

non-zero components in this representation, which would be convenient for a<br />

deeper investigation <strong>of</strong> the gravitational effect on <strong>wave</strong> <strong>polarization</strong>. An interesting<br />

question for the future is if it is possible to detect gravitational <strong>wave</strong>s by<br />

<strong>polarization</strong> effects in space plasma.


Appendix A<br />

Generalized <strong>polarization</strong><br />

parameters and generators<br />

<strong>of</strong> SU(3) symmetry group<br />

Given the spectral density matrix for a vector field ⃗ F (using the notation x, y, z<br />

instead <strong>of</strong> 1,2,3)<br />

⎛<br />

⎞<br />

S d = F ⃗ ⊗ F ⃗ F x F † x ∗ F x Fy ∗ F x Fz<br />

∗<br />

= ⎝F y Fx ∗ F y Fy ∗ F y Fz<br />

∗ ⎠ .<br />

F z Fx ∗ F z Fy ∗ F z Fz<br />

∗<br />

(A.1)<br />

The spectral density matrix can be written as a linear combination, using the<br />

generators <strong>of</strong> the SU(3) symmetry group, λ i , and the three dimensional unit<br />

matrix, I 3 , as a basis. In the form given by Gell-Mann [17], these matrices are<br />

given by<br />

⎛<br />

I 3 = ⎝ 1 0 0 ⎞ ⎛<br />

0 1 0⎠ , λ 1 = ⎝ 0 1 0 ⎞ ⎛<br />

1 0 0⎠ , λ 2 = ⎝ 0 −i 0 ⎞<br />

i 0 0⎠ ,<br />

0 0 1 0 0 0 0 0 0<br />

⎛<br />

λ 3 = ⎝ 1 0 0 ⎞ ⎛<br />

0 −1 0⎠ , λ 4 = ⎝ 0 0 1 ⎞ ⎛<br />

0 0 0⎠ , λ 5 = ⎝ 0 0 −i ⎞<br />

0 0 0 ⎠ ,<br />

0 0 0 1 0 0 i 0 0<br />

(A.2)<br />

⎛<br />

λ 6 = ⎝ 0 0 0 ⎞ ⎛<br />

0 0 1⎠ , λ 7 = ⎝ 0 0 0 ⎞ ⎛<br />

0 0 −i⎠ , λ 8 = √ 1 ⎝ 1 0 0 ⎞<br />

0 1 0 ⎠ .<br />

0 1 0 0 i 0<br />

3 0 0 −2<br />

51


52 APPENDIX A. GENERALIZED POLARIZATION PARAMETERS<br />

Denoting the coefficients in this expansion by Λ i , we obtain<br />

S d = 1 3 I 3 + 1 8∑<br />

Λ i λ i =<br />

2<br />

i=1<br />

⎛<br />

1<br />

3<br />

⎜<br />

Λ 0 + 1 2 Λ 3 + 1<br />

1<br />

⎝ 2 Λ 1 + i 2 Λ 2<br />

1<br />

2 Λ 4 + i 2 Λ 5<br />

2 √ 3 Λ 8<br />

1<br />

2 Λ 1 − i 2 Λ 2<br />

1<br />

3 Λ 0 − 1 2 Λ 3 + 1<br />

1<br />

2 Λ 6 − i 2 Λ 7<br />

2 √ 3 Λ 8<br />

1<br />

2 Λ 4 − i 2 Λ ⎞<br />

5<br />

1<br />

2 Λ 6 − i 2 Λ ⎟<br />

7 ⎠ ,<br />

1<br />

3 Λ 0 − √ 1<br />

3<br />

Λ 8<br />

(A.3)<br />

where the Λ i ’s are<br />

Λ 0 = F 2 x + F 2 y + F 2 z ,<br />

Λ 1 = F x F ∗ y + F y F ∗ x = 2Re[F x F ∗ y ],<br />

(A.4)<br />

(A.5)<br />

Λ 2 = i [ F x F ∗ y − F y F ∗ x<br />

]<br />

= −2Im[Fx F ∗ y ], (A.6)<br />

Λ 3 = F 2 x − F 2 y ,<br />

Λ 4 = F x F ∗ z + F z F ∗ x = 2Re[F x F ∗ z ],<br />

(A.7)<br />

(A.8)<br />

Λ 5 = i [ F x F ∗ z − F z F ∗ x<br />

]<br />

= −2Im[Fx F ∗ z ], (A.9)<br />

Λ 6 = F y F ∗ z + F z F ∗ y = 2Re[F y F ∗ z ],<br />

(A.10)<br />

Λ 7 = i [ F y Fz ∗ − F z Fy ∗ ]<br />

= −2Im[Fy Fz ∗ ], (A.11)<br />

Λ 8 = √ 1 [<br />

F<br />

2<br />

x + F 2 ] 2<br />

y − √3 Fz 2 .<br />

3<br />

(A.12)<br />

These Λ i ’s are the generalized <strong>polarization</strong> parameters.


Appendix B<br />

Parameters from Dirac<br />

gamma matrices<br />

In Chapter 3, the Dirac gamma matrices are used to obtain a decomposition <strong>of</strong><br />

the <strong>electromagnetic</strong> field, which is represented by the field strength tensor<br />

⎛<br />

⎞<br />

0 −E 1 −E 2 −E 3<br />

F µν = ⎜E 1 0 −cB 3 cB 2<br />

⎟<br />

⎝E 2 cB 3 0 −cB 1<br />

⎠ .<br />

E 3 −cB 2 cB 1 0<br />

(B.1)<br />

In Lorentz space ((ct, x, y, z) notation) the original four gamma matrices are<br />

given by<br />

γ 0 =<br />

( )<br />

I2 0<br />

0 −I 2<br />

( )<br />

γ i 0 σ<br />

i<br />

=<br />

−σ i , i = 1, 2, 3 , (B.2)<br />

0<br />

where σ i are the Pauli spin matrices and I 2 is the two dimensional unit matrix,<br />

Equation (3.3). These are the four matrices which appear in the relativistic<br />

treatment <strong>of</strong> a spin - 1 2 particle.<br />

The four gamma matrices are then expanded into a 16 dimensional basis<br />

defined as<br />

I 4 , γ 5 = iγ 0 γ 1 γ 2 γ 3 , γ µ , γ 5 γ µ , σ µν = i(γ µ γ ν − γ ν γ µ )/2, (B.3)<br />

where µ, ν = 0, 1, 2, 3. Explicitly, the original four gamma matrices are<br />

⎛<br />

⎞ ⎛<br />

⎞<br />

1 0 0 0<br />

0 0 0 1<br />

γ 0 = ⎜0 1 0 0<br />

⎟<br />

⎝0 0 −1 0 ⎠ , γ1 = ⎜ 0 0 1 0<br />

⎟<br />

⎝ 0 −1 0 0⎠ ,<br />

0 0 0 −1<br />

−1 0 0 0<br />

⎛<br />

⎞ ⎛<br />

⎞<br />

0 0 0 −i<br />

0 0 1 0<br />

γ 2 = ⎜ 0 0 i 0<br />

⎟<br />

⎝ 0 i 0 0 ⎠ , γ3 = ⎜ 0 0 0 −1<br />

⎟<br />

⎝−1 0 0 0 ⎠ .<br />

−i 0 0 0<br />

0 1 0 0<br />

(B.4)<br />

53


54 APPENDIX B. PARAMETERS FROM DIRAC GAMMA MATRICES<br />

The four dimensional unit matrix and the γ 5 are given by<br />

⎛ ⎞ ⎛ ⎞<br />

1 0 0 0<br />

0 0 1 0<br />

I 4 = ⎜0 1 0 0<br />

⎟<br />

⎝0 0 1 0⎠ , γ5 = ⎜0 0 0 1<br />

⎟<br />

⎝1 0 0 0⎠ .<br />

0 0 0 1<br />

0 1 0 0<br />

(B.5)<br />

And the ten constructed matrices, γ 5 γ µ and σ µν , are found to be<br />

⎛<br />

⎞<br />

⎛<br />

⎞<br />

0 0 −1 0<br />

0 −1 0 0<br />

γ 6 = γ 5 γ 0 = ⎜0 0 0 −1<br />

⎟<br />

⎝1 0 0 0 ⎠ , γ7 = γ 5 γ 1 = ⎜−1 0 0 0<br />

⎟<br />

⎝ 0 0 0 1⎠ ,<br />

0 1 0 0<br />

0 0 1 0<br />

⎛<br />

⎞<br />

⎛<br />

⎞<br />

0 i 0 0<br />

−1 0 0 0<br />

γ 8 = γ 5 γ 2 = ⎜−i 0 0 0<br />

⎟<br />

⎝ 0 0 0 −i⎠ , γ9 = γ 5 γ 3 = ⎜ 0 1 0 0<br />

⎟<br />

⎝ 0 0 1 0 ⎠ ,<br />

0 0 i 0<br />

0 0 0 −1<br />

⎛ ⎞<br />

⎛<br />

⎞<br />

0 0 0 i<br />

0 0 0 1<br />

γ 10 = σ 01 = ⎜0 0 i 0<br />

⎟<br />

⎝0 i 0 0⎠ , γ11 = σ 02 = ⎜ 0 0 −1 0<br />

⎟<br />

⎝ 0 1 0 0⎠ , (B.6)<br />

i 0 0 0<br />

−1 0 0 0<br />

⎛<br />

⎞<br />

⎛<br />

⎞<br />

0 0 i 0<br />

1 0 0 0<br />

γ 12 = σ 03 = ⎜0 0 0 −i<br />

⎟<br />

⎝i 0 0 0 ⎠ , γ13 = σ 12 = ⎜0 −1 0 0<br />

⎟<br />

⎝0 0 1 0 ⎠ ,<br />

0 −i 0 0<br />

0 0 0 −1<br />

⎛<br />

⎞<br />

⎛ ⎞<br />

0 i 0 0<br />

0 1 0 0<br />

γ 14 = σ 13 = ⎜−i 0 0 0<br />

⎟<br />

⎝ 0 0 0 i⎠ , γ15 = σ 23 = ⎜1 0 0 0<br />

⎟<br />

⎝0 0 0 1⎠ .<br />

0 0 −i 0<br />

0 0 1 0<br />

The 36 <strong>polarization</strong> parameters are obtained by the contractions<br />

τ αβµν γ n αβγ m µν = F αβ (F µν ) ∗ γ n αβγ m µν,<br />

(B.7)<br />

and they are found to be:<br />

1<br />

4 F αβ (F µν ) ∗ γ 1 γ 1 = E 3 E3 ∗ + E 3 cB3 ∗ + cB 3 E3 ∗ + cB 3 cB3, ∗ (B.8)<br />

1<br />

4 F αβ (F µν ) ∗ γ 3 γ 3 = E 2 E2 ∗ + E 2 cB2 ∗ + cB 2 E2 ∗ + cB 2 cB2, ∗ (B.9)<br />

1<br />

4 F αβ (F µν ) ∗ γ 6 γ 6 = E 2 E2 ∗ − E 2 cB2 ∗ − cB 2 E2 ∗ + cB 2 cB2, ∗ (B.10)<br />

1<br />

4 F αβ (F µν ) ∗ γ 8 γ 8 = −E 1 E1 ∗ + E 1 cB1 ∗ + 4cB 1 E1 ∗ − cB 1 cB1, ∗ (B.11)<br />

1<br />

4 F αβ (F µν ) ∗ γ 11 γ 11 = E 3 E3 ∗ − E 3 cB3 ∗ − cB 3 E3 ∗ + cB 3 cB3, ∗ (B.12)<br />

1<br />

4 F αβ (F µν ) ∗ γ 14 γ 14 = −E 1 E1 ∗ − E 1 cB1 ∗ − cB 1 E1 ∗ − cB 1 cB1, ∗ (B.13)


1<br />

4 F αβ (F µν ) ∗ γ 1 γ 3 = E 3 E2 ∗ + E 3 cB2 ∗ + cB 3 E2 ∗ + cB 3 cB2, ∗ (B.14)<br />

1<br />

4 F αβ (F µν ) ∗ γ 1 γ 6 = −E 3 E2 ∗ − E 3 cB2 ∗ + cB 3 E2 ∗ + cB 3 cB2, ∗ (B.15)<br />

1<br />

4 F αβ (F µν ) ∗ γ 1 γ 8 = i(E 3 E1 ∗ − E 3 cB1 ∗ + cB 3 E3 ∗ − cB 3 cB1), ∗ (B.16)<br />

1<br />

4 F αβ (F µν ) ∗ γ 1 γ 11 = E 3 E3 ∗ − E 3 cB3 ∗ + cB 3 E3 ∗ − cB 3 cB3, ∗ (B.17)<br />

1<br />

4 F αβ (F µν ) ∗ γ 1 γ 14 = i(E 3 E3 ∗ + E 3 cB3 ∗ + cB 3 E3 ∗ + cB 3 cB3), ∗ (B.18)<br />

1<br />

4 F αβ (F µν ) ∗ γ 3 γ 6 = −E 2 E2 ∗ + E 2 cB2 ∗ − cB 2 E2 ∗ + cB 2 cB2, ∗ (B.19)<br />

1<br />

4 F αβ (F µν ) ∗ γ 3 γ 8 = i(E 2 E1 ∗ − E 2 cB1 ∗ + cB 2 E1 ∗ − cB 2 cB1), ∗ (B.20)<br />

1<br />

4 F αβ (F µν ) ∗ γ 3 γ 11 = E 2 E3 ∗ − E 2 cB3 ∗ + cB 2 E3 ∗ − cB 2 cB3, ∗ (B.21)<br />

1<br />

4 F αβ (F µν ) ∗ γ 3 γ 14 = i(E 2 E1 ∗ + E 2 cB1 ∗ + cB 2 E1 ∗ + cB 2 cB1), ∗ (B.22)<br />

1<br />

4 F αβ (F µν ) ∗ γ 6 γ 8 = i(−E 2 E1 ∗ + E 2 cB1 ∗ + cB 2 E1 ∗ − cB 2 cB1), ∗ (B.23)<br />

1<br />

4 F αβ (F µν ) ∗ γ 6 γ 11 = −E 2 E3 ∗ + E 2 cB3 ∗ + cB 2 E3 ∗ − cB 2 cB3, ∗ (B.24)<br />

1<br />

4 F αβ (F µν ) ∗ γ 6 γ 14 = i(−E 2 E1 ∗ − E 2 cB1 ∗ + cB 2 E1 ∗ + cB 2 cB1), ∗ (B.25)<br />

1<br />

4 F αβ (F µν ) ∗ γ 8 γ 11 = i(E 1 E3 ∗ − E 1 cB3 ∗ − cB 1 E3 ∗ + cB 1 cB3), ∗ (B.26)<br />

1<br />

4 F αβ (F µν ) ∗ γ 8 γ 14 = −E 3 E3 ∗ + E 3 cB3 ∗ + cB 3 E3 ∗ + cB 3 cB3, ∗ (B.27)<br />

1<br />

4 F αβ (F µν ) ∗ γ 11 γ 14 = iE 3 E1 ∗ + E 3 cB1 ∗ − cB 3 E1 ∗ − cB 3 cB1, ∗ (B.28)<br />

1<br />

4 F αβ (F µν ) ∗ γ 14 γ 11 = i(E 1 E3 ∗ − E 1 cB3 ∗ + cB 1 E3 ∗ − cB 1 cB3), ∗ (B.29)<br />

1<br />

4 F αβ (F µν ) ∗ γ 14 γ 8 = −E 1 E1 ∗ + E 1 cB1 ∗ − cB 1 E1 ∗ + cB 1 cB1, ∗ (B.30)<br />

1<br />

4 F αβ (F µν ) ∗ γ 14 γ 6 = i(−E 1 E2 ∗ + E 1 cB2 ∗ − cB 1 E2 ∗ + cB 1 cB2), ∗ (B.31)<br />

1<br />

4 F αβ (F µν ) ∗ γ 14 γ 3 = i(E 1 E2 ∗ + E 1 cB2 ∗ + cB 1 E2 ∗ + cB 1 cB2), ∗ (B.32)<br />

1<br />

4 F αβ (F µν ) ∗ γ 14 γ 1 = i(E 1 E3 ∗ + E 1 cB3 ∗ + cB 1 E3 ∗ + cB 1 cB3), ∗ (B.33)<br />

1<br />

4 F αβ (F µν ) ∗ γ 11 γ 8 = i(E 3 E1 ∗ − E 3 cB1 ∗ − cB 3 E1 ∗ + cB 3 cB1), ∗ (B.34)<br />

1<br />

4 F αβ (F µν ) ∗ γ 11 γ 6 = −E 3 E2 ∗ + E 3 cB2 ∗ + cB 3 E2 ∗ − cB 3 cB2, ∗ (B.35)<br />

55


56 APPENDIX B. PARAMETERS FROM DIRAC GAMMA MATRICES<br />

1<br />

4 F αβ (F µν ) ∗ γ 11 γ 3 = E 3 E2 ∗ + E 3 cB2 ∗ − cB 3 E2 ∗ − cB 3 cB2, ∗ (B.36)<br />

1<br />

4 F αβ (F µν ) ∗ γ 11 γ 1 = E 3 E3 ∗ + E 3 cB3 ∗ − cB 3 E3 ∗ − cB 3 cB3, ∗ (B.37)<br />

1<br />

4 F αβ (F µν ) ∗ γ 8 γ 6 = i(−E 1 E2 ∗ + E 1 cB2 ∗ + cB 1 E2 ∗ − cB 1 cB2), ∗ (B.38)<br />

1<br />

4 F αβ (F µν ) ∗ γ 8 γ 3 = i(E 1 E2 ∗ + E 1 cB2 ∗ − cB 1 E2 ∗ − cB 1 cB2), ∗ (B.39)<br />

1<br />

4 F αβ (F µν ) ∗ γ 8 γ 1 = i(E 1 E3 ∗ + E 1 cB3 ∗ − cB 1 E3 ∗ − cB 1 cB3), ∗ (B.40)<br />

1<br />

4 F αβ (F µν ) ∗ γ 6 γ 3 = −E 2 E2 ∗ − E 2 cB2 ∗ + cB 2 E2 ∗ + cB 2 cB2, ∗ (B.41)<br />

1<br />

4 F αβ (F µν ) ∗ γ 6 γ 1 = −E 2 E3 ∗ − E 2 cB3 ∗ + cB 2 E3 ∗ + cB 2 cB3, ∗ (B.42)<br />

1<br />

4 F αβ (F µν ) ∗ γ 3 γ 1 = E 2 E3 ∗ + E 2 cB3 ∗ + cB 2 E3 ∗ + cB 2 cB3. ∗ (B.43)


Appendix C<br />

⃗E, ⃗ B component form <strong>of</strong> the<br />

S, V, T parameters<br />

In Chapter 4 we defined and labelled the S, V, T parameters in the following<br />

way<br />

Σ : ⃗ E ⊗ ⃗ E † + ⃗ B ⊗ ⃗ B † = 1 3 δ ijS Σ + iɛ ijk V Σ<br />

k + T Σ ij , (C.1)<br />

∆ : ⃗ E ⊗ ⃗ E † − ⃗ B ⊗ ⃗ B † = 1 3 δ ijS ∆ + iɛ ijk V ∆<br />

k + T ∆ ij , (C.2)<br />

χ : ⃗ E ⊗ ⃗ B † + ⃗ B ⊗ ⃗ E † = 1 3 δ ijS χ + iɛ ijk V χ<br />

k + T χ ij , (C.3)<br />

ψ : ⃗ E ⊗ ⃗ B † − ⃗ B ⊗ ⃗ E † = 1 3 δ ijS ψ + iɛ ijk V ψ<br />

k + T ψ ij . (C.4)<br />

Here we list the parameters in component form.<br />

C.1 Σ<br />

S Σ = E 2 + c 2 B 2 = E 1 E ∗ 1 + E 2 E ∗ 2 + E 3 E ∗ 3 + c 2 B 1 B ∗ 1 + c 2 B 2 B ∗ 2 + c 2 B 3 B ∗ 3, (C.5)<br />

(<br />

⃗V Σ = − i 2<br />

E 2 E3 ∗ −E 3 E2 ∗ +c 2 B 2 B3 ∗ −c 2 B 3 B2, ∗ E 3 E1 ∗ −E 1 E3 ∗ +c 2 B 3 B1 ∗ −c 2 B 1 B3<br />

∗<br />

)<br />

, E 1 E2 ∗ − E 2 E1 ∗ + c 2 B 1 B2 ∗ − c 2 B 2 B1<br />

∗ , (C.6)<br />

Tij Σ =<br />

[E1+c 2 2 B1− 2 1 3 SΣ , 1 2 (E 1E2+E ∗ 2 E1+c ∗ 2 B 1 B2+c ∗ 2 B 2 B1), ∗ 1 2 (E 1E3+E ∗ 3 E1+c ∗ 2 B 1 B3+c ∗ 2 B 3 B1)]<br />

∗<br />

[ 1 2 (E 1E2+E ∗ 2 E1+c ∗ 2 B 1 B2+c ∗ 2 B 2 B1), ∗ E2+c 2 2 B2− 2 1 3 SΣ , 1 2 (E 2E † 3 +E 3E † 2 +c2 B 2 B3+c ∗ 2 B 3 B2)]<br />

∗<br />

[ 1 2 (E 1E3+E ∗ 3 E1+c ∗ 2 B 1 B3+c ∗ 2 B 3 B1), ∗ 1 2 (E 2E3+E ∗ 3 E2+c ∗ 2 B 2 B3+c ∗ 2 B 3 B2), ∗ E3+B 2 3− 2 1 3 SΣ ].<br />

(C.7)<br />

57


58 APPENDIX C. ⃗ E, ⃗ B COMPONENT FORM OF S, V, T PARAMETERS<br />

C.2 ∆<br />

S ∆ = E 2 −c 2 B 2 = E 1 E ∗ 1 +E 2 E ∗ 2 +E 3 E ∗ 3 −c 2 B 1 B ∗ 1 −c 2 B 2 B ∗ 2 −c 2 B 3 B ∗ 3, (C.8)<br />

(<br />

⃗V ∆ = − i 2<br />

E 2 E3 ∗ −E 3 E2 ∗ −c 2 B 2 B3+c ∗ 2 B 3 B2, ∗ E 3 E1 ∗ −E 1 E3 ∗ −c 2 B 3 B1+c ∗ 2 B 1 B3<br />

∗<br />

)<br />

, E 1 E2 ∗ − E 2 E1 ∗ − c 2 B 1 B2 ∗ + c 2 B 2 B1<br />

∗ , (C.9)<br />

Tij ∆ =<br />

[E1−c 2 2 B1− 2 1 3 S∆ , 1 2 (E 1E2+E ∗ 2 E1−c ∗ 2 B 1 B2−c ∗ 2 B 2 B1), ∗ 1 2 (E 1E3+E ∗ 3 E1−c ∗ 2 B 1 B3−c ∗ 2 B 3 B1)]<br />

∗<br />

[ 1 2 (E 1E2+E ∗ 2 E1−c ∗ 2 B 1 B2−c ∗ 2 B 2 B1), ∗ E2−c 2 2 B2− 2 1 3 S∆ , 1 2 (E 2E3+E ∗ 3 E2−c ∗ 2 B 2 B3−c ∗ 2 B 3 B2)]<br />

∗<br />

[ 1 2 (E 1E3+E ∗ 3 E1−c ∗ 2 B 1 B3−c ∗ 2 B 3 B1), ∗ 1 2 (E 2E3+E ∗ 3 E2−c ∗ 2 B 2 B3−c ∗ 2 B 3 B2), ∗ E3−c 2 2 B3− 2 1 3 S∆ ].<br />

(C.10)<br />

C.3 χ<br />

S χ = c ⃗ E· ⃗B † +c ⃗ B· ⃗E † = c ( E 1 B ∗ 1 +E 2 B ∗ 2 +E 3 B ∗ 3 +B 1 E ∗ 1 +B 2 E ∗ 2 +B 3 E ∗ 3)<br />

, (C.11)<br />

(<br />

⃗V χ = −c i 2<br />

E 2 B3 ∗ − E 3 B2 ∗ + B 2 E3 ∗ − B 3 E2, ∗ E 3 B1 ∗ − E 1 B3 ∗ + B 3 E1 ∗ − B 1 E3<br />

∗<br />

)<br />

, E 1 B2 ∗ − E 2 B1 ∗ + B 1 E2 ∗ − B 2 E1<br />

∗ , (C.12)<br />

T χ ij =<br />

[cE 1 B1+cB ∗ 1 E1− ∗ 1 3 Sχ , c 2 (E 1B2+E ∗ 2 B1+B ∗ 1 E2+B ∗ 2 E1), ∗ c 2 (E 1B3+E ∗ 3 B1+B ∗ 1 E3+B ∗ 3 E1)]<br />

∗<br />

[ c 2 (E 1B2+E ∗ 2 B1+B ∗ 1 E2+B ∗ 2 E1), ∗ cE 2 B2+cB ∗ 2 E2− ∗ 1 3 Sχ , c 2 (E 2B3+E ∗ 3 B2+B ∗ 2 E3+B ∗ 3 E2)]<br />

∗<br />

[ c 2 (E 1B3+E ∗ 3 B1+B ∗ 1 E3+B ∗ 3 E1), ∗ c 2 (E 2B3+E ∗ 3 B2+B ∗ 2 E3+B ∗ 3 E2), ∗ cE 3 B3+cB ∗ 3 E3− ∗ 1 3 Sχ ].<br />

(C.13)<br />

C.4 ψ<br />

S ψ = c ⃗ E· ⃗B † −c ⃗ B· ⃗E † = c ( E 1 B ∗ 1 +E 2 B ∗ 2 +E 3 B ∗ 3 −B 1 E ∗ 1 −B 2 E ∗ 2 −B 3 E ∗ 3)<br />

, (C.14)<br />

(<br />

⃗V ψ = −c i 2<br />

E 2 B3 ∗ − E 3 B2 ∗ − B 2 E3 ∗ + B 3 E2, ∗ E 3 B1 ∗ − E 1 B3 ∗ − B 3 E1 ∗ + B 1 E3<br />

∗<br />

)<br />

, E 1 B2 ∗ − E 2 B1 ∗ − B 1 E2 ∗ + B 2 E1<br />

∗ , (C.15)


C.4. ψ 59<br />

T ψ ij =<br />

[cE 1 B1−cB ∗ 1 E ∗ − 1 3 Sψ , c 2 (E 1B2+E ∗ 2 B1−B ∗ 1 E2−B ∗ 2 E1), ∗ c 2 (E 1B3+E ∗ 3 B1−B ∗ 1 E3−B ∗ 3 E1)]<br />

∗<br />

[ c 2 (E 1B2+E ∗ 2 B1−B ∗ 1 E2−B ∗ 2 E1), ∗ cE 2 B2−cB ∗ 2 E2− ∗ 1 3 Sψ , c 2 (E 2B3+E ∗ 3 B2−B ∗ 2 E3−B ∗ 3 E2)]<br />

∗<br />

[ c 2 (E 1B3+E ∗ 3 B1−B ∗ 1 E3−B ∗ 3 E1), ∗ c 2 (E 2B3+E ∗ 3 B2−B ∗ 2 E3−B ∗ 3 E2), ∗ cE 3 B3−cB ∗ 3 E3− ∗ 1 3 Sψ ].<br />

(C.16)


60 APPENDIX C. ⃗ E, ⃗ B COMPONENT FORM OF S, V, T PARAMETERS


Appendix D<br />

Covariant <strong>polarization</strong><br />

tensors in E,B-component<br />

form<br />

In Chapter 5 we saw that the S, V, T parameters can be found in Lorentz <strong>covariant</strong><br />

tensors. These tensors were constructed according to<br />

where<br />

is the metric,<br />

and<br />

1 X µν = F αµ (F αδ ) ∗ g δν , (D.1)<br />

2 X µν = F αµ (F αδ ) ∗ g δν , (D.2)<br />

3 X µν = F αµ (F αδ ) ∗ g δν , (D.3)<br />

4 X µν = F αµ (F αδ ) ∗ g δν , (D.4)<br />

⎛<br />

⎞<br />

1 0 0 0<br />

g µν = g µν = ⎜0 −1 0 0<br />

⎟<br />

⎝0 0 −1 0 ⎠ ,<br />

0 0 0 −1<br />

⎛<br />

⎞<br />

0 −E 1 −E 2 −E3<br />

F µν = ⎜E 1 0 −cB 3 cB 2<br />

⎟<br />

⎝E 2 cB 3 0 −cB 1<br />

⎠<br />

E 3 −cB 2 cB 1 0<br />

⎛<br />

⎞<br />

0 E 1 E 2 E 3<br />

F µν = ⎜−E 1 0 −cB 3 cB 2<br />

⎟<br />

⎝−E 2 cB 3 0 −cB 1<br />

⎠<br />

−E 3 −cB 2 cB 1 0<br />

(D.5)<br />

(D.6)<br />

(D.7)<br />

are the contravariant and <strong>covariant</strong> field strength tensors, respectively, and<br />

⎛<br />

⎞<br />

0 −cB 1 −cB 2 −cB 3<br />

F µν = ⎜cB 1 0 E 3 −E 2<br />

⎟<br />

⎝cB 2 −E 3 0 E 1<br />

⎠<br />

(D.8)<br />

cB 3 E 2 −E 1 0<br />

61


62 APPENDIX D. COVARIANT POLARIZATION TENSORS<br />

and<br />

⎛<br />

⎞<br />

0 cB 1 cB 2 cB 3<br />

F µν = ⎜−cB 1 0 E 3 −E 2<br />

⎟<br />

⎝−cB 2 −E 3 0 E 1<br />

⎠<br />

−cB 3 E 2 −E 1 0<br />

are the corresponding dual tensors.<br />

(D.9)<br />

D.1 The X tensors<br />

In E, B-component form the n X µν tensors are given by<br />

1 X µν =<br />

[−E 1 E ∗ 1 − E 2 E ∗ 2 − E 3 E ∗ 3, −E 2 cB ∗ 3 + E 3 cB ∗ 2, E 1 cB ∗ 3 − E 3 cB ∗ 1, −E 1 cB ∗ 2 + E 2 cB ∗ 1]<br />

[−cB 3 E ∗ 2 + cB 2 E ∗ 3, E 1 E ∗ 1 − c 2 B 3 B ∗ 3 − c 2 B 2 B ∗ 2, E 1 E ∗ 2 + c 2 B 2 B ∗ 1, E 1 E ∗ 3 + c 2 B 3 B ∗ 1]<br />

[cB 3 E ∗ 1 − cB 1 E ∗ 3, E 2 E ∗ 1 + c 2 B 1 B ∗ 2, E 2 E ∗ 2 − c 2 B 3 B ∗ 3 − c 2 B 1 B ∗ 1, E 2 E ∗ 3 + c 2 B 3 B ∗ 2]<br />

[−cB 2 E ∗ 1 +cB 1 E ∗ 2, E 3 E ∗ 1 +c 2 B 1 B ∗ 3, E 3 E ∗ 2 +c 2 B 2 B ∗ 3, E 3 E ∗ 3 −c 2 B 2 B ∗ 2 −c 2 B 1 B ∗ 1]<br />

(D.10)<br />

2 X µν =<br />

[−c 2 B 1 B ∗ 1−c 2 B 2 B ∗ 2−c 2 B 3 B ∗ 3, −cB 3 E ∗ 2+cB 2 E ∗ 3, cB 3 E ∗ 1−cB 1 E ∗ 3, −cB 2 E ∗ 1+cB 1 E ∗ 2]<br />

[−E 2 cB ∗ 3 + E 3 cB ∗ 2, c 2 B 1 B ∗ 1 − E 3 E ∗ 3 − E 2 E ∗ 2, E 2 E ∗ 1 + c 2 B 1 B ∗ 2, E 3 E ∗ 1 + c 2 B 1 B ∗ 3]<br />

[E 1 cB ∗ 3 − E 3 cB ∗ 1, E 1 E ∗ 2 + c 2 B 2 B ∗ 1, c 2 B 2 B ∗ 2 − E 3 E ∗ 3 − E 1 E ∗ 1, E 3 E ∗ 2 + c 2 B 2 B ∗ 3]<br />

[−E 1 cB ∗ 2 + E 2 cB ∗ 1, E 1 E ∗ 3 + c 2 B 3 B ∗ 1, E 2 E ∗ 3 + c 2 B 3 B ∗ 2, c 2 B 3 B ∗ 3 − E 2 E ∗ 2 − E 1 E ∗ 1]<br />

(D.11)<br />

3 X µν =<br />

[−cB 1 E ∗ 1−cB 2 E ∗ 2−cB 3 E ∗ 3, −c 2 B 2 B ∗ 3+c 2 B 3 B ∗ 2, c 2 B 1 B ∗ 3−c 2 B 3 B ∗ 1, −c 2 B 1 B ∗ 2+c 2 B 2 B ∗ 1]<br />

[E 3 E ∗ 2 − E 2 E ∗ 3, cB 1 E ∗ 1 + E 3 cB ∗ 3 + E 2 cB ∗ 2, cB 1 E ∗ 2 − E 2 cB ∗ 1, cB 1 E ∗ 3 − E 3 cB ∗ 1]<br />

[−E 3 E ∗ 1 + E 1 E ∗ 3, cB 2 E ∗ 1 − E 1 cB ∗ 2, cB 2 E ∗ 2 + E 3 cB ∗ 3 + E 1 cB ∗ 1, cB 2 E ∗ 3 − E 3 cB ∗ 2]<br />

[E 2 E ∗ 1 − E 1 E ∗ 2, cB 3 E ∗ 1 − E 1 cB ∗ 3, cB 3 E ∗ 2 − E 2 cB ∗ 3, cB 3 E ∗ 3 + E 2 cB ∗ 2 + E 1 cB ∗ 1]<br />

(D.12)<br />

4 X µν =<br />

[−E 1 cB ∗ 1 − E 2 cB ∗ 2 − E 3 cB ∗ 3, −E 3 E ∗ 2 + E 2 E ∗ 3, E 3 E ∗ 1 − E 1 E ∗ 3, −E 2 E ∗ 1 + E 1 E ∗ 2]<br />

[c 2 B 2 B ∗ 3 −c 2 B 3 B ∗ 2, E 1 cB ∗ 1 +cB 3 E ∗ 3 +cB 2 E ∗ 2, −cB 2 E ∗ 1 +E 1 cB ∗ 2, −cB 3 E ∗ 1 +E 1 cB ∗ 3]<br />

[−c 2 B 1 B ∗ 3+c 2 B 3 B ∗ 1, −cB 1 E ∗ 2+E 2 cB ∗ 1, cB 3 E ∗ 3+E 2 cB ∗ 2+cB 1 E ∗ 1, −cB 3 E ∗ 2+E 2 cB ∗ 3]<br />

[c 2 B 1 B ∗ 2−c 2 B 2 B ∗ 1, −cB 1 E ∗ 3+E 3 cB ∗ 1, −cB 2 E ∗ 3+E 3 cB ∗ 2, cB 2 E ∗ 2+E 3 cB ∗ 3+cB 1 E ∗ 1]<br />

(D.13)


D.2. THE Y TENSORS 63<br />

D.2 The Y tensors<br />

The Y tensors, constructed by the combinations<br />

Σ Y µν = 1 2<br />

( 1<br />

X µν + 2 X µν )<br />

,<br />

χ Y µν = 1 ( 3<br />

X µν + 4 X µν )<br />

,<br />

2<br />

∆ Y µν = 1 ( 1<br />

X µν − 2 X µν )<br />

(D.14)<br />

,<br />

2<br />

ψ Y µν = 1 ( 3<br />

X µν − 4 X µν )<br />

,<br />

2<br />

and are then found to be<br />

Σ Y µν = 1 2<br />

[−E 1 E ∗ 1−E 2 E ∗ 2−E 3 E ∗ 3−cB 1 cB ∗ 1−cB 2 cB ∗ 2−cB 3 cB ∗ 3, −cB 3 E ∗ 2+cB 2 E ∗ 3−E 2 cB ∗ 3+E 3 cB ∗ 2,<br />

E 1 cB ∗ 3 − E 3 cB ∗ 1 + cB 3 E ∗ 1 − cB 1 E ∗ 3, −E 1 cB ∗ 2 + E 2 cB ∗ 1 − cB 2 E ∗ 1 + cB 1 E ∗ 2]<br />

[−cB 3 E ∗ 2+cB 2 E ∗ 3−E 2 cB ∗ 3+E 3 cB ∗ 2, E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2+cB 1 cB ∗ 1−E 3 E ∗ 3−E 2 E ∗ 2,<br />

E 2 E ∗ 1 + cB 1 cB ∗ 2 + E 1 E ∗ 2 + cB 2 cB ∗ 1, E 3 E ∗ 1 + cB 1 cB ∗ 3 + E 1 E ∗ 3 + cB 3 cB ∗ 1]<br />

[E 1 cB ∗ 3 − E 3 cB ∗ 1 + cB 3 E ∗ 1 − cB 1 E ∗ 3, E 2 E ∗ 1 + cB 1 cB ∗ 2 + E 1 E ∗ 2 + cB 2 cB ∗ 1,<br />

E 2 E ∗ 2−cB 3 cB ∗ 3−cB 1 cB ∗ 1+cB 2 cB ∗ 2−E 3 E ∗ 3−E 1 EC 1 , E 3 E ∗ 2+cB 2 cB ∗ 3+E 2 E ∗ 3+cB 3 cB ∗ 2]<br />

[−E 1 cB ∗ 2 + E 2 cB ∗ 1 − cB 2 E ∗ 1 + cB 1 E ∗ 2, E 3 E ∗ 1 + cB 1 cB ∗ 3 + E 1 E ∗ 3 + cB 3 cB ∗ 1,<br />

E 3 E ∗ 2+cB 2 cB ∗ 3+E 2 E ∗ 3+cB 3 cB ∗ 2, E 3 E ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+cB 3 cB ∗ 3−E 2 E ∗ 2−E 1 E ∗ 1]<br />

(D.15)<br />

χ Y µν = 1 2<br />

[−cB 1 E ∗ 1−cB 2 E ∗ 2−cB 3 E ∗ 3−E 1 cB ∗ 1−E 2 cB ∗ 2−E 3 cB ∗ 3, −cB 2 cB ∗ 3+cB 3 cB ∗ 2−E 3 E ∗ 2+E 2 E ∗ 3,<br />

cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3, −cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2]<br />

[E 3 E ∗ 2−E 2 E ∗ 3+cB 2 cB ∗ 3−cB 3 cB ∗ 2, cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1+cB 3 E ∗ 3+cB 2 E ∗ 2,<br />

cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2, cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3]<br />

[−E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1, cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1,<br />

cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1+cB 3 E ∗ 3+cB 2 E ∗ 2, cB 2 E ∗ 3−E 3 cB ∗ 2−cB 3 E ∗ 2+E 2 cB ∗ 3]<br />

[E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1, cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1,<br />

cB 3 E ∗ 2−E 2 cB ∗ 3−cB 2 E ∗ 3+E 3 cB ∗ 2, cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1+cB 3 E ∗ 3+cB 2 E ∗ 2]<br />

(D.16)


64 APPENDIX D. COVARIANT POLARIZATION TENSORS<br />

∆ Y µν = 1 2<br />

[−E 1 E ∗ 1−E 2 E ∗ 2−E 3 E ∗ 3+cB 1 cB ∗ 1+cB 2 cB ∗ 2+cB 3 cB ∗ 3, cB 3 E ∗ 2−E 2 cB ∗ 3−cB 2 E ∗ 3+E 3 cB ∗ 2,<br />

cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3, cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1]<br />

[cB 2 E ∗ 3−E 3 cB ∗ 2−cB 3 E ∗ 2+E 2 cB ∗ 3, E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+E 3 E ∗ 3+E 2 E ∗ 2,<br />

− cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2, −E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1]<br />

[cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1, E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1,<br />

E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+E 3 E ∗ 3+E 2 E ∗ 2, −cB 2 cB ∗ 3+cB 3 cB ∗ 2−E 3 E ∗ 2+E 2 E ∗ 3]<br />

[cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2, cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3,<br />

E 3 E ∗ 2−E 2 E ∗ 3+cB 2 cB ∗ 3−cB 3 cB ∗ 2, E 1 E ∗ 1−cB 3 cB ∗ 3−cB 2 cB ∗ 2−cB 1 cB ∗ 1+E 3 E ∗ 3+E 2 E ∗ 2]<br />

(D.17)<br />

ψ Y µν = 1 2<br />

[−cB 1 E ∗ 1−cB 2 E ∗ 2−cB 3 E ∗ 3+E 1 cB ∗ 1+E 2 cB ∗ 2+E 3 cB ∗ 3, E 3 E ∗ 2−E 2 E ∗ 3−cB 2 cB ∗ 3+cB 3 cB ∗ 2,<br />

cB 1 cB ∗ 3 − cB 3 cB ∗ 1 − E 3 E ∗ 1 + E 1 E ∗ 3, −cB 1 cB ∗ 2 + cB 2 cB ∗ 1 + E 2 E ∗ 1 − E 1 E ∗ 2]<br />

[E 3 E ∗ 2−E 2 E ∗ 3−cB 2 cB ∗ 3+cB 3 cB ∗ 2, cB 1 E ∗ 1+E 3 cB ∗ 3+E 2 cB ∗ 2−E 1 cB ∗ 1−cB 3 E ∗ 3−cB 2 E ∗ 2,<br />

cB 2 E ∗ 1 − E 1 cB ∗ 2 + cB 1 E ∗ 2 − E 2 cB ∗ 1, cB 3 E ∗ 1 − E 1 cB ∗ 3 + cB 1 E ∗ 3 − E 3 cB ∗ 1]<br />

[cB 1 cB ∗ 3 − cB 3 cB ∗ 1 − E 3 E ∗ 1 + E 1 E ∗ 3, cB 2 E ∗ 1 − E 1 cB ∗ 2 + cB 1 E ∗ 2 − E 2 cB ∗ 1,<br />

cB 2 E ∗ 2+E 3 cB ∗ 3+E 1 cB ∗ 1−cB 3 E ∗ 3−E 2 cB ∗ 2−cB 1 E ∗ 1, cB 3 E ∗ 2−E 2 cB ∗ 3+cB 2 E ∗ 3−E 3 cB ∗ 2]<br />

[−cB 1 cB ∗ 2 + cB 2 cB ∗ 1 + E 2 E ∗ 1 − E 1 E ∗ 2, cB 3 E ∗ 1 − E 1 cB ∗ 3 + cB 1 E ∗ 3 − E 3 cB ∗ 1,<br />

cB 3 E ∗ 2−E 2 cB ∗ 3+cB 2 E ∗ 3−E 3 cB ∗ 2, cB 3 E ∗ 3+E 2 cB ∗ 2+E 1 cB ∗ 1−cB 2 E ∗ 2−E 3 cB ∗ 3−cB 1 E ∗ 1]<br />

(D.18)<br />

D.3 The K and K tensors<br />

The K and K tensors are formed by removing the trace from ∆ Y µν and χ Y µν ,<br />

respectively, and are given by<br />

K µν = 1 2<br />

[0, cB 3 E ∗ 2 − E 2 cB ∗ 3 − cB 2 E ∗ 3 + E 3 cB ∗ 2, cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3,<br />

cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1]<br />

[cB 2 E ∗ 3 − E 3 cB ∗ 2 − cB 3 E ∗ 2 + E 2 cB ∗ 3, 0, −cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2,<br />

− E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1]<br />

[cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1, E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1, 0,<br />

− cB 2 cB ∗ 3 + cB 3 cB ∗ 2 − E 3 E ∗ 2 + E 2 E ∗ 3]<br />

[cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2, cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3,<br />

E 3 E ∗ 2 − E 2 E ∗ 3 + cB 2 cB ∗ 3 − cB 3 cB ∗ 2, 0]<br />

(D.19)


D.3. THE K AND K TENSORS 65<br />

and<br />

K µν = 1 2<br />

[0, −cB 2 cB ∗ 3 + cB 3 cB ∗ 2 − E 3 E ∗ 2 + E 2 E ∗ 3, cB 1 cB ∗ 3 − cB 3 cB ∗ 1 + E 3 E ∗ 1 − E 1 E ∗ 3,<br />

− cB 1 cB ∗ 2 + cB 2 cB ∗ 1 − E 2 E ∗ 1 + E 1 E ∗ 2]<br />

[E 3 E ∗ 2 − E 2 E ∗ 3 + cB 2 cB ∗ 3 − cB 3 cB ∗ 2, 0, cB 1 E ∗ 2 − E 2 cB ∗ 1 − cB 2 E ∗ 1 + E 1 cB ∗ 2,<br />

cB 1 E ∗ 3 − E 3 cB ∗ 1 − cB 3 E ∗ 1 + E 1 cB ∗ 3]<br />

[−E 3 E ∗ 1 + E 1 E ∗ 3 − cB 1 cB ∗ 3 + cB 3 cB ∗ 1, cB 2 E ∗ 1 − E 1 cB ∗ 2 − cB 1 E ∗ 2 + E 2 cB ∗ 1, 0,<br />

cB 2 E ∗ 3 − E 3 cB ∗ 2 − cB 3 E ∗ 2 + E 2 cB ∗ 3]<br />

[E 2 E ∗ 1 − E 1 E ∗ 2 + cB 1 cB ∗ 2 − cB 2 cB ∗ 1, cB 3 E ∗ 1 − E 1 cB ∗ 3 − cB 1 E ∗ 3 + E 3 cB ∗ 1,<br />

respectively.<br />

cB 3 E ∗ 2 − E 2 cB ∗ 3 − cB 2 E ∗ 3 + E 3 cB ∗ 2, 0],<br />

(D.20)


66 APPENDIX D. COVARIANT POLARIZATION TENSORS


Bibliography<br />

[1] A. O. Barut. Electrodynamics and Classical Theory <strong>of</strong> Fields and Particles.<br />

Dover Publications, New York, 1980.<br />

[2] J. D. Bjorken and S. D. Drell. Relativistic Quantum Mechanics. McGraw-<br />

Hill Publishing Company, New York, 1964.<br />

[3] M. Born and E. Wolf. Principles <strong>of</strong> Optics. Pergamon Press, London, 1959.<br />

[4] C. Brosseau. Fundamentals <strong>of</strong> Polarized Light. John Wiley & Sons, New<br />

York.<br />

[5] C. Brosseau. Entropy <strong>of</strong> a classical stochastic <strong>wave</strong>field: Effect <strong>of</strong> <strong>polarization</strong>.<br />

Optik, 104(1):21–26, 1996.<br />

[6] T. Carozzi, R. Karlsson, and J. Bergman. Parameters characterizing<br />

<strong>electromagnetic</strong> <strong>wave</strong> <strong>polarization</strong>. Physical Review E, 61(2):2024–2028,<br />

Febuary 2000.<br />

[7] John Archibald Wheeler Charles W. Misner, Kip S. Thorne. Gravitation.<br />

W. H. Freeman and Company, New York, 1973.<br />

[8] P. A. M. Dirac. General Theory <strong>of</strong> Relativity. Princeton University Press,<br />

Princeton, New Jersey, 1996.<br />

[9] Karl-Heinz Fieseler. Algebra mn3. University <strong>of</strong> Uppsala, August 1998.<br />

[10] J. Weber G. B. Arfken, H. Mathematical Methods for Physicists. Academic<br />

Press, San Diego, 1995.<br />

[11] Herbert Goldstein. Classical Mechanics. Addison-Wesley Publishing Company,<br />

Reading, 1980.<br />

[12] J. D. Jackson. Classical Electrodynamics. John Wiley & Sons, New York,<br />

1975.<br />

[13] Roger Karlsson. Three-dimensional spectral stokes parameters. Master’s<br />

thesis, Institute <strong>of</strong> Technology, Uppsala University, April 1997.<br />

[14] Ismo Lindell. Complex vector algebra in <strong>electromagnetic</strong>s. Int. J. Elect.<br />

Enging Educ., 20:33–47, 1983.<br />

[15] Ismo V. Lindell. Methods for Electromagnetic Field Analysis, chapter 1,<br />

pages 1–16. Oxford University Press, New York, 1992.<br />

67


68 BIBLIOGRAPHY<br />

[16] Jerry B. Marion and Mark A. Heald. Classical Electromagnetic Radiation.<br />

Academic Press, Orlando, Florida, 1980.<br />

[17] M. Gell-Mann & Y. Ne’eman. The Eightfold Way. W. A. Benjamin Inc.,<br />

new York, 1964.<br />

[18] A. O. Barut & R. Raczka. Theory <strong>of</strong> Group Representations and Applications.<br />

World Scientific Publishing, Singapore, 1986.<br />

[19] J. J. Sakurai. Advanced Quantum Mechanics. Addison-Wesley Publishing<br />

Company, Reading.<br />

[20] J. J. Sakurai. Modern Quantum Mechanics. Addison-Wesley Publishing<br />

Company, Reading, Massachusetts, 1994.<br />

[21] Bernard F. Schutz. A First Course in General Relativity. Cambridge University<br />

Press, Cambridge, 1998.<br />

[22] G. G. Stokes. Trans. cambridge philos. soc. 9. In Mathematical and Physical<br />

Papers, Vol III, page 233. Cambridge University Press, 1901.<br />

[23] L. R. O. Storey and F. Lefeuvre. Theory for the interpretations <strong>of</strong> measurements<br />

<strong>of</strong> the six components <strong>of</strong> a random <strong>electromagnetic</strong> <strong>wave</strong> field in<br />

space. Space Research, 14:381–386, 1974.<br />

[24] Wu-Ki Tung. Group Theory in Physics. World Scientific Publishing, Singapore,<br />

1985.<br />

[25] Roald K. Wagnsness. Electromagnetic Fields. John Wiley & Sons, New<br />

York, 1986.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!