Sie sind auf Seite 1von 139

Neutrino Flavor Mixing

Mark Pinckers
Master Thesis in Theoretical Physics
Under Supervision of Dr. T. Prokopec
Institute for Theoretical Physics, Utrecht University

March 6, 2012
2
Contents

1 Introduction 5
1.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Neutrino Mixing Theory 11


2.1 Two Flavors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Three Flavors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3 Experiments 19
3.1 Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Laboratory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Solar Neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Cosmology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5 Results from Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Fermionic Propagators 31
4.1 Feynman Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.1 Contour Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.1.2 Spatial Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1.3 Massless Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5 Dirac Flavor Mixing 43


5.1 Set Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.1.1 Chirality and Helicity Decomposition . . . . . . . . . . . . . . . . . 45
5.2 Statistical Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2.1 Calculating the Statistical Propagator . . . . . . . . . . . . . . . . 49
5.2.2 Initial Expectation Values . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Time Translation Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.1 Example State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3.2 General Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

6 Majorana Fermions 65
6.1 Neutrinoless Double Beta Decay . . . . . . . . . . . . . . . . . . . . . . . . 66
6.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3
4 CONTENTS

6.2.1 Helicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.2.2 Rest Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.2.3 Boost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.2.4 General Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.2.5 Excited States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2.6 Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

7 Majorana Flavor Mixing 87


7.1 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.1.1 Diagonalizing the Equations of Motion . . . . . . . . . . . . . . . . 89
7.2 Statistical Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.2.1 Initial Correlators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.2 Time Translational Invariance . . . . . . . . . . . . . . . . . . . . . 97
7.2.3 General Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

8 Discussion 105
8.1 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
8.1.1 Laboratory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.1.2 Thermal Distributions . . . . . . . . . . . . . . . . . . . . . . . . . 109

Appendices 115

A Conventions 115

B Diagonalizing the Mass Matrix 117


B.1 General Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
B.2 Symmetric Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

C Kinetic Description 123


C.1 Wigner Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
C.2 Helicity and Chirality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
C.3 Wigner Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
C.4 Flow Equations in the Diagonal Mass Basis . . . . . . . . . . . . . . . . . 131

D Statistical Description of the Density Matrix 133


D.1 Spin Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Chapter 1

Introduction

The history of neutrino physics goes back to 1930, when the neutrino was first theoreti-
cally postulated by Wolfgang Pauli. They were needed to ensure energy conservation in
radioactive decays. In the decay of a neutron, a proton and an electron are created. These
electrons were observed to have a range of energies, which let to the formulation of the
neutrino to make this range of energies possible. The neutrino has therefore no charge.
This also gave the first weight limit on the neutrino, a limit which is still being refined.
Nowadays neutrinos are considered to be an integral part of the standard model in which
the weak interactions are embedded. The importance of the neutrino for our model of
the universe was only realized in the last few decades. According to current models,
neutrinos are one of the most abundant forms of matter in the universe. Neutrinos are
also involved in nucleosynthesis, therefore participating in the creation of all heavy atoms,
with speculations about its earlier importance still ongoing.
The next step was to observe these neutrinos. They interact very lightly with matter,
which makes observing them a delicate job. The main idea is to try to let the reverse
decay reaction take place. A neutrino combines with a proton to form a neutron and
an electron. To do this, a beam of neutrinos enters an area of dense protons, and then
one tries to observe an electron and a neutron. Consequently, this neutrino is referred to
as an electron neutrino. When the incoming neutrino is a muon neutrino, instead of an
electron a muon particle will be created. There is a third neutrino, the tau neutrino which
corresponds to the tau particle and these three different neutrinos are referred to as the
three flavors. It has been experimentally found that there are three neutrino flavors, which
is confirmed by the standard model, however some theories beyond the Standard Model
predict the existence of additional ”sterile” neutrinos, which are either very heavy or very
weakly interacting.
Aside from artificial sources, neutrinos are also produced in nature. This includes for
example the creation of neutrinos in the sun, in the atmosphere and in supernovas. These
neutrinos are observed in different experiments, which still uncover more about the funda-

5
6 CHAPTER 1. INTRODUCTION

mental nature and give stricter bounds on for instance the mass. In laboratories stricter
bounds on the mass are still being researched. There are two main types of experiments
for these bounds:

• Kinematic searches

These processes are allowed for both massive and massless neutrinos, and the idea
is to compare the outcome with calculations for massive neutrinos and massless
neutrinos. When the outcome is a function of the neutrino mass, this also gives
an estimation of the mass. Examples included nuclear β-decay, pion decay and tau
decay.

• Exclusive tests

Exclusive tests are only allowed if the neutrino is massive. The most well-known
example is neutrino oscillations, and also include neutrino decays, observation of
certain electromagnetic properties (such as e.g. a magnetic moment) and neutrinoless
β-decay.

The better results that were obtained, the lower the upper bound for the mass of the
neutrino became. It became a possibility that neutrinos could actually be massless. In
1957 Pontecorvo was the first to come up with the idea of neutrino mixing [4]. Should
neutrinos be massive, neutrino mixing can be observed. Neutrino mixing changes the
flavor of the neutrino into another flavor, which depends on the mass difference between
the different flavors.

Recently, these oscillations have been an effective tool for evidence of neutrino masses and
mixing. Usually the flux of a distinct flavor of neutrinos is measured and then compared to
a measurement of the same flux over large distances. It is found that neutrinos of a distinct
flavor appear to disappear, which can be explained through oscillations. This means they
have oscillated into a different flavor. Two examples include the electron neutrinos in the
atmosphere and muon neutrinos within the sun and from the sun to the Earth. It is more
difficult to find this effect in a laboratory.

An important feature of neutrino oscillations is that it is impossible for massless neutrinos


to oscillate, and therefore the observation of neutrino mixing implies that neutrinos have a
mass. In the Standard Model neutrinos are massless, it is actually impossible to construct
a gauge invariant renormalizable mass term for them in the Standard Model and thus there
is no neutrino mixing in the Standard Model. Massive neutrinos can therefore tell us about
physics beyond the Standard Model.

Apart from the mass of the neutrino, it is know that is spin 1/2 particle, in units of ~, and
has no electric charge. Including the quarks in the different flavors of neutrinos, we can
write the different families of matter as
7

(u, d, e, νe ) (1.1)
(c, s, µ, νµ ) (1.2)
(t, b, τ, ντ ). (1.3)

As mentioned briefly, in the Standard Model neutrinos are predicted to be massless. This
would make neutrinos the only fermion which is massless, all the other fermions in the Stan-
dard Model are known to have mass. Especially theorists favored the massless hypothesis,
as the Standard Model was very successful in calculating other processes such as current
interactions of neutrinos. Experiments showed however that neutrinos can oscillate, which
serves as convincing evidence that neutrino do have mass.
The Standard Model only contains left-handed neutrino field (νL ), but there is no real
fundamental reason why there is no right-handed neutrino field (νR ). Should this right-
handed neutrino field be present, it could have given the neutrinos a mass through the Higgs
mechanism. The photon for instance has its masslessness guaranteed by the conservation
of a gauge symmetry, for a neutrino such a symmetry is not present, thus allowing it to be
massive in principle. In most unified theories the neutrino indeed does have a mass.
A different theoretical question is related to the magnitude of the mass of the neutrino.
The neutrino mass is very small compared to the mass of charged fermions. The Standard
Model also does not have the answer several other mass gaps, e.g. the hierarchy problem.
The mass difference between charged neutrinos in the same family is by far not as large
as the difference for the neutrino. The mass of the electron is 0.511 MeV, and the mass of
the up and down quarks is about one order of magnitude larger. The mass of the electron
neutrino is however at least five orders of magnitude smaller. We see the same pattern in
the other two families.
Thirdly there is still uncertainty whether neutrinos are Dirac fermions or Majorana fermions.
Majorana particles have the unique property that they are their own anti particle. All other
fermions are Dirac particles, because they have a charge and can therefore not be their
own anti particle. For neutrinos this obvious restriction does not apply, because of their
chargelessness. This is the open question in which we will be mostly be interested in this
thesis.
Currently experiments are underway that try to determine the nature of the neutrino. The
most popular method is to try to detect neutrinoless double beta decay, a process that is
only possible of the neutrino is a Majorana fermion. For a Dirac fermion only double beta
decay with the emittance of neutrinos is possible, such that the detection of neutrinoless
double beta decay would immediately lead to an answer to the question of the nature of
neutrinos.
We will try to construct a different experiment which can also test the nature of the
neutrino. In order to do so, we will use a new framework to discuss flavor oscillations.
8 CHAPTER 1. INTRODUCTION

Currently, when evaluating flavor mixing always pure initial states are considered. We will
however extend this to also include mixed initial states. This will give more complicated
physical situations. We can now for instance construct a state that does not exhibit flavor
oscillations. This special state is not the same for Majorana and Dirac fermions, thus
providing us with the possibility to distinguish between them. Finally, we try to construct
an experiment in which this difference is apparent. This experiment will be substantially
different from current efforts to detect the nature of neutrinos.

1.1 Outline

In this thesis first the current state of neutrino mixing theory will be discussed in chapter
2. This is the standard mixing theory which can be found in many references discussing
neutrino physics. This will first be done for two neutrino flavors and next for three flavors,
for which several relevant limits will be discussed.
In chapter 3 recent experimental data will be summarized which puts bounds on possible
theories for the neutrino. In general all experiments can be classified as appearance and
disappearance experiments. There are three main classes of experiments, namely atmo-
spherical, solar and laboratory experiments. The most important results of recent years
will be presented. These put bounds on the possible mass differences between neutrinos
and mixing angles. From cosmological arguments a bound on the total sum of the masses
can be calculated. In literature this bound has further been improved.
Chapter 4 will focus on the dynamics of fermions. We start out with discussing the different
relevant propagators. The propagator for fermions in position space will be solved in D-
dimensions. This will be done in the massive case, and the massless will be obtained in 4
dimensions, showing that our result agrees with well known results from literature. Since
neutrinos are also fermions, this result is directly applicable to neutrinos.
In chapter 5 we will apply an approach with the two point functions that originate from the
Schwinger Keldysh formalism to the problem of oscillating Dirac neutrinos. The advantage
of this approach is the allowance of mixed initial states. In order to do so, we need to
diagonalize the mass in flavor space and helicity states, both of which were discussed
earlier. We will look at the possibility of constant states, that do not oscillate in flavor
space although the mass matrix still has off diagonal components.
In chapter 6 Majorana fermions will be introduced. Their defining properties are discussed
and the experimental search for their existence in the form of Majorana neutrinos is
discussed. We will describe their dynamics and quantize the classical solution.
Finally flavor oscillations for Majorana neutrinos will be discussed in chapter 7. Here we
will apply the same approach as for the Dirac neutrinos in the previous chapter. The goal
is to see if there is any difference in their behavior in flavor mixing, which will be discussed
1.1. OUTLINE 9

in the discussion chapter at the end of this thesis. These differences between Majorana
and Dirac neutrinos could lead to a new way to construct an experiment in which we can
determine the nature of neutrinos.
10 CHAPTER 1. INTRODUCTION
Chapter 2

Neutrino Mixing Theory

In order to properly treat neutrino dynamics and flavor oscillations in specific, we first
discuss the current state of research. We will give a short overview of current treatments
on mixing of the mass eigenstates of neutrinos, also dubbed neutrino oscillations.
Before understanding neutrino mixing we have to consider the process in which neutrinos
are produced. The electron neutrino (νe ) is defined as the particle that is coupled to
the electron through the charged weak current. The same goes for the tau neutrino (ντ )
which is coupled to the tau particle and the muon neutrino (νµ ) which is coupled to the
muon particle. In the case that neutrinos have a mass, which we consider experimentally
verified as will be discussed in chapter 3, these particles do not necessarily have to be mass
eigenstates. Instead of a single mass eigenstates, we can say in general that the neutrinos
are in a superposition of mass eigenstates.
There are two different eigenstates for neutrinos. On the one hand we have flavor eigen-
states, on the other hand there are mass eigenstates. When a neutrino is produced it is
in a flavor eigenstate, which is a superposition of mass eigenstates. The other way round,
we can say that a neutrino with a certain mass is in a superposition of different flavor
eigenstates and can therefore couple to different leptons. This is comparable to the mixing
of quarks, and a similar oscillation is found with neutral kaons for instance, and has also
been experimentally verified in this case.
We consider a neutrino beam that has been created together with an anti lepton ˆl. We
denote the neutrino as νl . Now we can expand the created neutrino by a weak interaction
in a combination of mass eigenstates

X
|νl i = Vli |νi i. (2.1)
i

The matrix Vli is a unitary matrix that represents the neutrino mixing. In the literature
this matrix is denoted by U , however we will use the letter U for a rotation matrix in

11
12 CHAPTER 2. NEUTRINO MIXING THEORY

later chapters. Should this matrix be complex, we would have an implied CP-violation.
In general there is no reason why it should be real. This matrix is referred to as the
Pontecorvo-Maki-Nakagawa-Sakata matrix (PMNS matrix) [38]. Now we assume that all
the 3-momentum p~ of the different components are equal. Because of the different masses,
every component has a different energy which is given by the relativistic energy-momentum
relation
q
Ei = p~2 + m2i . (2.2)

Next, we assume that the beam evolves as a plane wave. This implies that the neutrinos
do not decay. When the neutrinos do decay, the analysis is substantially different. So we
assume in this case that

X
|νl (t)i = Vli e−iEi t |νi (0)i. (2.3)
i

Because the different energies (Ei ) give rise to a different behavior when time progresses
for different mass eigenstates, we see that the superposition also changes through time.
We can write the probability that a state νl0 occurs in the beam that originated as νl as

X
hνl0 |νl (t)i = hνβ |Vβl† 0 e−iEα t Vlα |να i (2.4)
α,β
X
= e−iEi t Vli Vl∗0 i . (2.5)
i

Here we used the orthogonality of the mass eigenstates. For t → 0 this reduces to δll0 as
expected. We can now also write out the general probability for finding a certain νl0 in a
beam of νl particles.

Pll0 (t) = |hνl0 |νl (t)i|2 (2.6)


X
= |Vlα Vl∗0 α Vlβ∗ Vl0 β | cos((Eα − Eβ )t − φll0 αβ ), (2.7)
α,β

and here we used the shorthand

φll0 αβ = arg(Vlα Vl∗0 α Vlβ∗ Vl0 β ). (2.8)

Since the mass of neutrinos is found to be relatively small in some cases, such as solar
neutrinos, they travel at extremely relativistic speeds. The energy momentum relation can
be approximated in this case as
2.1. TWO FLAVORS 13

m2i
Ei ≈ |~p| + . (2.9)
2|~p|

Also, the distance x that is traveled by the beam can replace the time t that has passed
since the beam was created. Here we used the convention that c = 1 and ~ = 1. In this
approximation we find for the probability

 
X 2πx
Pll0 (x) = |Ulα Ul∗0 α Ulβ

Ul0 β | cos − φll0 αβ . (2.10)
α,β
Lαβ

Here we defined a new scale


Lαβ = , (2.11)
∆m2αβ
and here we defined the mass difference intuitively as

∆m2αβ = m2α − m2β . (2.12)

The Lαβ is the oscillation length of the problem. They give a natural distance scale for this
process and give us an indication for what sort of scales we should look in experiments to
see any effects. If the distance x traveled is an integral multiple of the oscillation length,
then the original beam is again obtained. At any other distance the results are not trivial
and mixing will take place.
In this quantum mechanical picture it is generally assumed that the initial state is a pure
flavor state. It is this assumption that we will try to relax in the later chapters of this
thesis.
The dependence on the mass-squared difference between states ∆m2αβ instead of depen-
dence on the actual mass of a state, already signals an important restriction on many
experiments. Most experiments are only sensitive to a mass-squared difference between
states, not the actual value of the mass of a certain neutrino state.

2.1 Two Flavors

In the simplest case there are two different Dirac flavors. Now the matrix V only depends
on one parameter and there is only a single mass-squared difference between these two
states
14 CHAPTER 2. NEUTRINO MIXING THEORY

 
cos ϑ sin ϑ
U= . (2.13)
− sin ϑ cos ϑ

We see that there are no complex phases in this matrix, which shows us that in the case
of two Dirac flavor neutrinos there can be no CP phase.
The transition probabilities now take the specific form

∆m2
 
2 2
Pl,l0 = δl,l0 − (δl,l0 − 1) sin 2ϑ sin x . (2.14)
4E

We omitted the indices on the mass-squared difference, since they are trivial in this
case. From an experiment one can therefore restrict the mass-squared difference based
on measurements of the angle ϑ. Note that the probability is actually invariant to taking
∆m2 → −∆m2 , i.e. redefining ν1 ↔ ν2 . Therefore it is impossible to determine which of
the components is the heavier one in a vacuum experiment. In experiments the transition
probability is measured, and usually the results are interpreted in this framework of only
two neutrino species.

Figure 2.1: A single neutrino starting out in the blue flavor (i.e. the species with probability
1 at L/E = 0), oscillating between two different flavors.

We can picture this as in figure (2.1), where the probability of a single neutrino in the blue
flavor propagates is plotted. The frequency of this oscillation is controlled by the mass
difference between the two species. The angle ϑ controls the amplitude of the oscillation,
i.e. the minimum of the blue probability and the maximum of the red probability.
For Majorana neutrinos the most general form of the mixing matrix in the case of only
two flavors is

e−iρ sin ϑ
 
cos ϑ
U= . (2.15)
−eiρ sin ϑ cos ϑ
2.2. THREE FLAVORS 15

The difference with the Dirac case is the appearance of the complex phase ρ, which shows
that in this case there can in principle be CP mixing. However, when calculating the
mixing probabilities, the additional phase factor ρ drops out and exactly the same result
as for the Dirac neutrinos is obtained.

2.2 Three Flavors


When can redo the analysis with three neutrino species instead of the two flavors in the
previous section. For simplicity we assume that CP is conserved, which results in a real
mixing matrix V . In general we can write the oscillation probability as

∆m2αβ
X  
2
Pl,l0 (x) = δl,l0 − 4 Vlα Vl0 α Vlβ Vl0 β sin x . (2.16)
α>β
4E

Here we also used the fact that V is orthogonal. In certain approximations much simpler
forms can be obtained. An example is when two of the three masses are very close together
∆m2 x
[5], we assume 2E12  1, which gives

∆m232
 
2
Pl,l0 (x) = δl,l0 − 4Vl3 Vl0 3 (δl,l0 − Vl3 Vl0 3 ) sin x . (2.17)
4E

In an experiment where we would like to measure the flux of νe for example, this gives the
following probability for νe surviving after production and traveling for a distance x.

∆m232
 
2 2
Pνe ,νe0 (x) = 1 − sin (2ϑ13 ) sin x . (2.18)
4E

We can immediately see that by comparing this result with (2.14) the survival probability
completely agrees if we make the following two identification: the mixing angle is given by
ϑ13 and the mass difference squared ∆m232 .
The final case we will discuss is when one of the neutrino species is much lighter than the
other two:

∆m232 x ∆m231 x
1  1. (2.19)
2E 2E
The transition probabilities are now given by

∆m2 12
 
2
Pl,l0 (x) = δl,l0 − 2Ul3 Ul0 3 (δl,l0 − Ul3 Ul0 3 ) − 4Ul1 Ul0 1 Ul2 Ul0 2 sin x , (2.20)
4E
16 CHAPTER 2. NEUTRINO MIXING THEORY

and for the probability for an electron neutrino to survive the propagation is given by

∆m221
 
4 2 2
Pνe ,νe0 (x) = cos ϑ13 1 − sin (2ϑ12 ) sin ( x) + sin4 ϑ13 . (2.21)
4E
We made several assumptions in these derivations, and one can of course improve upon
these. We have not taken into account the uncertainty principle since we assumed that the
neutrino was produced with a definite value for the momentum. Consider that when this
would be a good approximation, the location of production would not be known. If the
location is unknown in the order of the oscillation length, the distance traveled x does not
have any meaning. Instead of the definite momentum, one should consider a wave packet
to represent the neutrino [6].
We can give a small quantitative estimate that shows that this induced error is however
very small. We said that the uncertainty should be much smaller than the oscillation
length, δx  L. Using the uncertainty principle we can make an estimation of the bounds
on the momentum uncertainty:

δp 1 ∆m2
 = . (2.22)
p Lp 4πE 2
We will see in the next section that characteristic values for E are at least 1 MeV and for
∆m2 smaller than 1 eV. This gives an order of magnitude for the left hand side of at most
10−12 , which shows that the uncertainty is very small and thus that the approximation of
not treating the neutrino as a wave packet is a very good one.
There have also been calculations where the creation of the neutrino, the propagation and
the detection were considered in a field theoretical framework [7]. This gives an expression
for the probability of detection which is equal to (2.10), and therefore we will not go in
any deeper into this derivation.
The most general form for V in the case of three neutrino flavors (electron, muon and tau)
including complex phases is given by

0 s13 e−iδ
     iα /2 
1 0 0 c13 c12 s12 0 e 1 0 0
V = 0 c23 s23   0 1 0  −s12 c12 0  0 eiα2 /2 0

0 −s23 c23 −s13 e 0 c13 0 0 1 0 0 1
(2.23)
s13 e−iδ
   
c12 c13 s12 c13 eiα1 /2 0 0
iδ iδ iα2 /2
= −s12 c23 − c12 s23 s13 e
 c12 c23 − s12 s23 s13 e s23 c13   0 e 0 .
iδ iδ
s12 s23 − c12 c23 s13 e −c12 s23 − s12 c23 s13 e c23 c13 0 0 1
(2.24)
2.2. THREE FLAVORS 17

Here we used the notation sin(ϑij ) = sij and cos(ϑij ) = cij . Secondly there are also three
phases in this matrix expression. The phase δ is a CP-mixing phase which can be present.
The other two phases (α1 and α2 ) are Majorana phases. They can only be present if the
neutrino has a Majorana nature. Should neutrinos have a Dirac nature, these two phases
have to be set to zero. Experimental efforts have been usually directed at measuring
and improving the bounds on the mixing angles ϑij and the mass differences between two
different flavors ∆mij .
To finish the discussion of the quantum mechanical picture of mixing, it should be noted
that everywhere a pure initial state is assumed. This pure state then starts to oscillate in
flavor space, however initially it is a pure single flavor state. We will relax this requirement
in the later chapters and see what changes. The picture will be more complicated then the
overview presented here.
18 CHAPTER 2. NEUTRINO MIXING THEORY
Chapter 3

Experiments

Neutrino mixing has been confirmed by experiments. These experiments also give bounds
on the mass of neutrinos and the mixing angles. Secondly there are cosmological implica-
tions linked to different properties of the neutrino. Note for instance that a neutrino mass
could have influenced the expansion of the universe and plays a role in structure formation.
We discuss both experimental results and results from cosmological data. These two types
of data are a relevant addition to each other, as experiments typically only give values for
the mass difference between different species and the cosmological data gives bounds on
the sum of the masses of all three species. In experiments where one wants to look for
neutrino oscillations, there are two possible effects that can be observable. For both effects,
the distance from the source should not be an integral multiple of the oscillation length,
as discussed before.
Firstly, there are disappearance events:

Pll (x) < 1, (3.1)

as the name suggests, one looks for a reduction in the flux of neutrinos that were produced
in a certain reaction and are no longer present after propagating for a certain distance x.
When the initial flux is known, this can then be compared to the flux at a distance x.
This shows that some neutrinos have disappeared, from which neutrino oscillations can be
deduced.
Secondly, there are appearance events:

Pll0 (x) > 0, (3.2)

here one looks out for the appearance of a neutrino species that was not produced initially.

19
20 CHAPTER 3. EXPERIMENTS

Consider a reaction in which for instance νµ is produced. After propagating for a distance
x the flux of νe is no longer equal to zero. This suggests that during the propagation of the
νµ , this beam oscillated into νe , also leading to the conclusion that apparently neutrinos
oscillate.
These differences between the two classes of experiments have several implications. The
advantage of an appearance experiment is that a single event can already be meaningful.
When the single appearance of a neutrino is measured, this can already give an implication
of neutrino oscillations. In a disappearance experiment one is forced to measure the entire
flux of neutrinos, and then detect a significant difference from the expected flux.
Another difference is that in an appearance experiment only a single transition can be
measured. If one is trying to measure the oscillation from a beam of νe into νµ and for
some reason this rate is very low, but the νe into ντ rate is high, an appearance experiment
would be in trouble. Measuring the decline of the rate of νe would still be enough to argue
that oscillations have been found. We can say that an appearance experiment is sensitive
for a oscillation of the type νl → νl0 , and a disappearance experiment measures νl → νX ,
where νX is the sum of all the other flavors.
This statement that disappearance experiments measure the sum of oscillations into all
other flavors has some unforeseen implications. Some theories Beyond the Standard Model
(BSM), predict that there exist additional species of neutrinos, so-called sterile neutrinos.
These neutrinos do not interact through gauge bosons of the standard model. When one of
the three familiar neutrino species oscillates into this fourth species, it is no longer directly
measurable, since it has an extremely small cross section. However, a disappearance
experiment could in principle show that a certain amount of neutrinos have oscillated
away from the original flux.
All experiments that have been performed have the same parameter upon which they
depend, the figure of merit [8]:

E
m̄2 = . (3.3)
x

This fraction can be influenced in experiments, other parameters from the Lagrangian such
as mass differences and mixing angles are fixed ”by nature”. The oscillation probabilities
(2.10) can be rewritten to depend on the figure of merit

∆m2αβ
X  
Pll0 (x) = |Ulα Ul∗0 α Ulβ

Ul0 β | cos − φll0 αβ . (3.4)
α,β
2m̄2

This figure of merit dictates to which mass square difference the experiment is sensitive,
an experiment can only detect oscillations if m̄2 < ∆m2αβ . This shows the importance of
a small figure of merit, which can be accomplished by a long distance from the source or
3.1. ATMOSPHERE 21

a highly energetic neutrinos. For some experiments characteristic values for m̄2 are given
below.

Table 3.1: Figure of merit for several neutrino oscillation experiments


Source x in cm E in MeV m̄2 in eV2
Short baseline 102 1 10−2
Long baseline 107 10 10−6
3 3
Accelerators 10 10 1
Atmosphere 107 104 10−3
Solar 1011 1 10−11

Next we will discuss the experiments denoted in table 3.1. We can split them up in
three main categories: laboratory experiments, atmospherical experiments and solar exper-
iments. Next we will discuss some important aspects of all three categories of experiments.

3.1 Atmosphere
In 1998 the first evidence that was model independent was found at the Super-Kamiokande
atmospherical neutrino experiment. This detector is based on water Cherenkov detectors,
with a total mass of water of 50 kilotons. The following reaction takes place in order to
detect the neutrinos (both electron and muon neutrinos can be detected):

νl + N →l− + X (3.5)
ν̄l + N →l+ + X. (3.6)

In these reactions N is the target which is hit by the neutrino νl . A lepton with the same
flavor l is produced and an atom or ion X depending on the target. The second reaction
is the equivalent for an impacting anti neutrino instead of a neutrino. Only upgoing
muons can be identified as being produced by neutrinos, downgoing muons can also have
different origins. There are two different events: either all energy can be deposited inside
the detector and the particle will come to a halt, or the particle can escape the detector,
hitting the rock surrounding it. The energies of the neutrinos that produce the muons that
come to a stop is about 10 GeV, the through going particles are created by neutrinos with
an energy of on average 100 GeV.
Neutrino production in the atmosphere is mainly caused by the decay of pions and muons.
Comparing the ratio measured at the detector with the ratio in which they are produced,
oscillations have been shown to exist. Secondly they also measured the zenith angle
distribution. Neutrinos that travel exactly upward have traveled 13000 kilometers, whereas
22 CHAPTER 3. EXPERIMENTS

neutrinos coming in almost downward have traveled a much shorter distance. This gives
a distribution for different angles and allows an approximation for the mixing angles and
mass differences.

3.2 Laboratory

Laboratory experiments use neutrino beams originating from either a nuclear reactor or
an accelerator.

The conventional experiment with an accelerator creates a neutrino beam through the
decay of pions. Protons are accelerated at a target resulting in pions, which can decay into
four different particles muon and electron neutrinos and anti neutrinos:

p + target → π ± + X
π+ → µ + + νµ
µ+ → e+ + νe + ν̄µ
π− → µ− + ν̄µ
µ− → e− + ν̄e + νµ
.

The characteristic distances involved in experiments were of the order of a hundred meters,
these are referred to as short baseline experiments, however in order to obtain better results
the baseline was increased and the first long baseline experiments were started.

An example of a long baseline experiment is the K2K accelerator. Here neutrinos were
created at the KEK accelerator in Japan and detected in the Super-Kamiokande detector,
discussed previously, 250 kilometers away. In the beam at the accelerator νµ were isolated
and they constitute the majority of the beam that was send in the direction of the detector.
There the flux of this beam was measured, and indeed this rate was lower than the rate at
the accelerator. The bounds found were not very strict, however this experiment was the
first experiment with laboratory produced neutrinos that confirmed neutrino oscillations.

In the KamLAND experiment compelling evidence for neutrino oscillations were obtained.
Electron anti neutrinos produced in nuclear reactors in Japan were measured in an old mine.
The emission rates are known, which allows for a comparison with the measurements done
in the experiment. Without oscillations, 2179 detections were expected, however only 1609
events were measured during five years. The angle and the mass difference between νµ and
νe can be estimated from these results, which will be discussed later in this chapter.
3.3. SOLAR NEUTRINOS 23

3.3 Solar Neutrinos

In the Sun mainly electron neutrinos are produced. This is largely done through the pp
cycle. In this cycle deuterium is produced according to the following reaction:

p + p → d + e+ + νe . (3.7)

Other important cycles for current experiments are the 7 Be and 8 B reactions, given by

e− +7 Be →7 Li + νe (3.8)
8
B →8 Be + e+ + νe . (3.9)

The so-called Standard Solar Model (SSM) predicts a total rate of 6.4 · 1010 cm−2 s−1 of
neutrinos produced in the sun.
The start of solar neutrino experiments was the Homestake experiment, a radiochemical
experiment by Davis et al, and a Nobel prize has been awarded for this experiment. The
experiments has ran in total from 1968 until 1994, located in the Homestake mine in order
to reduce cosmic ray background. It uses the following reaction to detect neutrinos:

νe +37 Cl → e− +37 Ar. (3.10)

The produced 37 Ar is radioactive. This decay is measured, giving numbers on the flux
of neutrinos. The energy threshold of this reaction is too high to capture the pp chain
neutrinos, the main flux measured is coming from the 8 B reaction. The SSM then predicts
a rate of

RSSM = 8.1 ± 1.3 SNU. (3.11)

The unit SN U stands for a solar neutrino neutrino unit (10−36 events per atom per second),
a commonly used unit in these experiments. The measurements however only give a rate
of
RCl = 2.56 ± 0.16 ± 0.16 SNU. (3.12)

This is significantly smaller than predicted. Note that these predictions do not include
neutrino oscillations.
24 CHAPTER 3. EXPERIMENTS

Similar radiochemical experiments include GNO-GALLEX and SAGE experiments. These


detectors were based on a different detection reaction, allowing them to also measure the
pp chain:

νe +71 Ga → e− +71 Ge. (3.13)

Secondly, they performed test experiments with an intense neutrino beam of known in-
tensity to check if the predicted and measured flux matched, and this was positive. Both
experiments have produced results that are in good agreement with each other:

RGALLEX−GN O =67.5 ± 5.1 SNU (3.14)


RSAGE =70.8 ± 5.3 ± 3.5 SNU. (3.15)

The predicted rate is given by

RGa = 128 ± 8 SNU. (3.16)

Again this rate is without taking any neutrino oscillations into account.
The Kamiokande and Super-Kamiokande have also measured solar neutrinos, having the
advantage that is possible to measure the direction of the neutrinos, whereas this is im-
possible for the radiochemical experiments. This gives data on for instance the day-night
asymmetry.
It was discussed that the fluxes measured by the previous solar neutrino experiments are
smaller than predicted by the SSM, which can be explained with neutrino oscillations. The
SNO (Sudbury Neutrino Observatory) solar neutrino experiment was the first experiment
to actually supply evidence for this claim without depending on any model [20]. The
difference between the SNO and other neutrino experiments is that the SNO measures
three different processes:

νe + d →e− + p + p (3.17)
νx + d →νx + p + n (3.18)
νx + e →νx + e. (3.19)

These processes are referred to as respectively the CC, the NC and the ES processes. These
three processes were measured successively with the SNO observatory. From the first two
process (CC and NC) the following rates were found
3.4. COSMOLOGY 25

−2 −1
ΦCC 6
νe = 1.68 ± 0.06 ± 0.08 · 10 cm s (3.20)
−2 −1
ΦN C
νe,µ,τ = 4.94 ± 0.21 ± 0.36 · 10 cm s 6
(3.21)

The CC process is only sensitive to electron neutrinos, whereas the NC process is sensitive
to the sum of the three species. From the difference in these fluxes we can conclude that
on the way from the sun to Earth some electron neutrinos are transformed into the other
species. This is the model independent evidence for neutrino oscillations. The ES process
measures a linear combination of the neutrino species, which gives additional data for
determining oscillation parameters.
The total flux from the NC process is in agreement with the SSM, which shows no indication
for additional sterile neutrinos. Should there be additional sterile neutrinos, this rate should
be lower as some of the produced neutrinos would have oscillated into these sterile neutrino
states.

3.4 Cosmology
By considering the early big bang universe, bounds can be obtained on the neutrino mass.
We will not present a full thorough derivation, but show how one can obtain a bound on
the mass of neutrinos from data on the contents of the universe.
The entropy for ultra relativistic particles is given by

2π 2
S= g∗s (kT )3 V. (3.22)
45

The volume (V ) scales as a3 , the scale factor of the universe cubed. We need a new
quantity: g∗s , which is related to the effective number of relativistic degrees of freedom for
the involved particles. This is given by

 3   X  3
X Ti 7 Ti
g∗s = gi + gi . (3.23)
bosons
T 8 fermions T

The new variables are:T , the temperature of the plasma that is in equilibrium, Ti the
temperature of the species i and gi , the effective number of relativistic degrees of freedom
for species i. As long as the Hubble expansion rate (H = ȧ/a) is smaller than the
interaction rate of the reactions that maintain this equilibrium, the neutrinos will be in a
thermodynamical equilibrium. We will try to look at the moment when neutrinos are no
longer in thermodynamical equilibrium. As the universe expands the temperature drops
26 CHAPTER 3. EXPERIMENTS

and at a certain moment the rate of the reaction is equal to the Hubble expansion rate.
This temperature is referred to as the freeze-out temperature and is given at the moment
when

Γ
∼1 (3.24)
H

for a reaction rate Γ and the Hubble rate H. Next is find expressions for both these
quantities. We know that the early universe is dominated by ultra-relativistic particles
and the Hubble parameter is given by


r
8πG
H= ρ ∼ G(kT )2 . (3.25)
3

The interaction rate is given by

Γ = σvn. (3.26)

Here σ is the cross section of the reaction, v the relative speed and n the number density of
the particles. First we know v ' 1, we are dealing with ultra-relativistic particles. Secondly,
n ∼ (kT )3 for ultra relativistic particles. To find the cross section, we are interested in the
cross section of the reactions

e+ + e−
νx + ν̄x (3.27)
νx + e±
νx + e± (3.28)
ν̄x + e±
ν̄x + e± . (3.29)

Here νx reppresent the different neutrino species. Thermally averaging the cross sections
of these reactions gives us an expression for the cross section:

σ ∼ G2F (kT )2 , (3.30)

and thus we find for the interaction rate:

Γ ∼ G2F (kT )5 . (3.31)

Here we denoted the Stefan-Boltzmann constant with k. Now we are ready to calculate
the freeze-out temperature from the ratio of the expansion rate and the reaction rate:
3.4. COSMOLOGY 27

 3
Γ kT
∼ G2F MP (kT )3 ∼ . (3.32)
H 1 MeV

Hence on this temperature the neutrinos decouple from the photons on the one hand and
the electrons and the positrons on the other hand, which are still in thermal equilibrium.
When the temperature drops further, e± annihilate and are no longer in thermal equilib-
rium. This heats up the photon sea, but does not affect the neutrino temperature. Using
entropy conservation and the fact that g∗s = 11/2 before and g∗s = 2 after decoupling, we
have the following relation between the temperature before and after e± decoupling:

 1/3
Tafter 11 abefore
= . (3.33)
Tbefore 3 aafter

The neutrinos are not effected by the decoupling, thus

Tνafter abefore
= . (3.34)
Tνbefore aafter

And we know that Tνbefore = Tbefore , which leads us to conclude that

 1/3
4
Tν = Tγ . (3.35)
11

This relates the current temperature of the CMB (Tγ ) to the temperature of the background
neutrino radiation (Tν ). Note that the CMB has been experimentally confirmed, but the
neutrino background radiation has not been found yet. The CMB has a temperature of
Tγ = 2.725K, leading to a neutrino temperature of Tν = 1.945K.
This enables to calculate the Gerstein-Zeldovich bound on the sum of the masses of the
neutrino species. The number density of the neutrinos, which scale is if they are still
relativistic, is given by:

9ζ(3)
nν = (kTν )3 . (3.36)
2π 2

We know the temperature of the neutrino sea from (3.35), which leads us to the following
number density:

nν = 336 cm−3 . (3.37)

The large abundance of neutrinos in the universe is shown by comparing this number to
28 CHAPTER 3. EXPERIMENTS

the baryon density: nB = 2.5 · 10−7 cm−3 . Only the number density of the photons is
slightly larger: nγ = 410.5cm−3 , neutrinos and photons therefore constitute numerically
the largest category of particles in the universe.

If we assume that the neutrino masses are close together (it is quasi-degenerate, m1 ' m2 ' m3 )
we write the density parameter of the neutrinos as

P
i mi nν
Ων ' . (3.38)
3ρc

3H 2
Here ρc is the critical density, ρc = 8πG
. Numerically calculating this density parameter
we find

P
i mi
Ων ' . (3.39)
94h2 eV

This allows us to easily find a limit on the sum of the mass of the neutrinos, if we for
instance assume that the density of the neutrinos cannot be larger than the density of dark
matter. They can of course be a part of dark matter. Using recent data ΩDM ' 0.23 and
h ' 0.70 [37].

X
mi ≤ 10 eV. (3.40)
i

And the mass of a single neutrino species is approximately mi ≤ 3.3 eV, considering the
relatively small mass squared differences between the different
P flavors. This bound can
still be improved and in literature values are found of i mi ≤ 1.6 P eV [16]. By adding
data of the galaxy red shift surveys the bound can be improved, i mi ≤ 0.87 eV [14].
Using
P also Lyman-α data and other data, the most strict bound is obtained in [15] of
i mi ≤ 0.30 eV. All these cited results are at 95% confidence level. Note that there
is not a general agreement about these bounds, different papers cite different bounds.
Especially about the Lyman-α data is disagreement, because of the strong dependence on
what analysis of this data is used [15].

Recently
P [39] has constructed a new mass bound on the sum of the neutrino masses:
i mi ≤ 0.26 eV by using the Sloan digital sky survey data and combining it with WMAP
data and Hubble telescope measurements of the Hubble parameter. The problem with
these results is that the procedure of combining different data sets can lead to not yet
understood systematic errors and that there is a model P dependence in these results. For
the most conservative model dependence, a bound of i mi ≤ 0.36 eV is cited. Note that
in these cases the Lyman-α is not needed to obtain the bounds.
3.5. RESULTS FROM EXPERIMENTS 29

3.5 Results from Experiments


The following are the best known bounds on the mass difference and mixing angles param-
eters:
The three mixing angles are given by
• ϑ13 < 10.3◦ [31]
• ϑ12 = 33.9◦ ± 2.4◦ [32]
• ϑ23 = 45◦ ± 7◦ [33]
Note that there is only an upper bound on ϑ13 . This angle can also still be zero, it has
not been verified to actually exist. This would imply that neutrinos cannot mix from the
electron flavor directly to the tau flavor however it is possible that the electron neutrino
first mixes to the muon flavor and then to the tau flavor. It has been recently discussed
[40] that this angle is very likely to be non-zero based on a global neutrino data analysis.
For the mass differences we have the following
• ∆m221 = 8.0 ± 0.6 · 10−5 eV2 [32]
• ∆m231 ≈ ∆m232 = 2.43 ± 0.13 · 10−3 eV2 [34]
Here it should be noted that these are only mass differences squared and that the sign (and
thus the hierarchy) is not known.
In addition to the mixing angles and the mass differences there were also three CP violating
phases in the general expression for the mixing matrix. One angle (δ) can in principle be
present for both Dirac and Majorana neutrinos. The other two (α1 and α2 ) are linked to
the Majorana nature and are automatically zero in the case of a Dirac neutrino. Neither
of the three angles have been measured and are all unknowns at present.
There are also experiments underway that depend on the absolute mass scale. One of
the most promising experiments is the KATRIN experiment. Analyzing the beta decay of
tritium a bound on the electron neutrino mass is hoped to be found. Because of the low
energy involved in the beta decay of tritium, this particular set up is chosen. In about one
of a trillion of the decays the neutrino is emitted with almost no kinetic energy, leading to
the possibility to derive a direct prediction for the mass.
Later we will discuss neutrinoless double beta decay. This decay is sensitive to the so-called
effective Majorana neutrino mass, given by

X
|mββ | = mi Vei ≤ 0.3 ∼ 1 eV. (3.41)
i

The matrix elements of the PMNS matrix (V ) are summed over, the index i denotes
30 CHAPTER 3. EXPERIMENTS

the different mass eigenstates. This bound is only applicable if the neutrino is indeed a
Majorana particle. The final result of these experiments still depends on the choice of the
matrix elements [17].
Chapter 4

Fermionic Propagators

We will begin our analysis of neutrino with the propagators for fermions. First of all we
can define four two point functions. These four two point functions originate from the
Schwinger-Keldysh formalism. Note this is done specifically for fermionic operators, for
scalar field operators the story is different. Here we use that ψ̂ is the usual fermionic field
operator that satisfies the anticommutation relation at equal time

{ψ̂, ψ̂ † } = 1. (4.1)

The two point functions are defined as

iS ++ (t, t0 ) = hT [ψ̂(t)ψ̂ † (t0 )]i (4.2)


iS +− (t, t0 ) = −hψ̂ † (t0 )ψ̂(t)i (4.3)
iS −+ (t, t0 ) = hψ̂(t)ψ̂ † (t0 )i (4.4)
iS −− (t, t0 ) = hT̄ [ψ̂(t)ψ̂ † (t0 )]i. (4.5)

Time ordering and anti-time ordering are defined by T and T̄ :

iS ++ (t, t0 ) = θ(t − t0 )iS −+ (t, t0 ) + θ(t0 − t)iS +− (t, t0 ) (4.6)


iS −− (t, t0 ) = θ(t0 − t)iS −+ (t, t0 ) + θ(t − t0 )iS +− (t, t0 ). (4.7)

Here we can identify the well known Feynman propagator iS ++ = iSF and the anti
Feynman propagator iS −− . The other two two point functions are referred to as the
Wightman functions (iS +− and iS −+ ).

31
32 CHAPTER 4. FERMIONIC PROPAGATORS

From this definition it is obvious that

iS ++ (t, t0 ) + iS −− (t, t0 ) = iS −+ (t, t0 ) + iS +− (t, t0 ) (4.8)


iS ++ (t, t0 ) − iS −− (t, t0 ) = sign(t − t0 )(iS −+ (t, t0 ) − iS +− (t, t0 )), (4.9)

this identity also holds in the scalar case. The definition of the two point functions is
actually done in such a way that this is true.
Because of the time ordered nature of iS ++ (t, t0 ) = iS ++ (t0 , t) we also know that

θ(t − t0 )iS −+ (t, t0 ) + θ(t0 − t)iS +− (t, t0 ) = θ(t0 − t)iS −+ (t0 , t) + θ(t − t0 )iS +− (t0 , t) (4.10)

and thus

iS −+ (t, t0 ) = iS +− (t0 , t). (4.11)

These Green functions can be written in matrix form

iS (t, t0 ) iS +− (t, t0 )
 ++ 
0
iG(t, t ) = (4.12)
iS −+ (t, t0 ) iS −− (t, t0 )

satisfying

 
iγ 0 ∂0 + i~γ · ~k − m iG(t, t0 ) = iσ 3 δ(t − t0 ). (4.13)

By taking linear combinations of the two point functions we can define advanced and
retarded propagators:

iS r (t, t0 ) = iS ++ (t, t0 ) − iS +− (t, t0 ) (4.14)


iS a (t, t0 ) = iS ++ (t, t0 ) − iS −+ (t, t0 ). (4.15)

Causal (spectral) and the statistical propagator (two point function) can be expressed as
4.1. FEYNMAN PROPAGATOR 33

1 1
ρψ (t, t0 ) = (iS −+ (t, t0 ) − iS +− (t, t0 )) = h{ψ̂(t), ψ̂ † (t0 )}i (4.16)
2 2
1 1
Fψ (t, t0 ) = (iS −+ (t, t0 ) + iS +− (t, t0 )) = h[ψ̂(t), ψ̂ † (t0 )]i. (4.17)
2 2

At equal time the causal propagator is fixed by the commutation relation to give the correct
limit. This is again similar to the bosonic case, at equal time the causal propagator is fixed
by (anti)commutation relations and the statistical propagator is still to be determined.
The final property is the KMS condition, which represents the anti-periodicity the fermions
we are dealing here with have, opposed to the periodicity of bosons,

iS −+ (t − iβ, t0 ) = −iS +− (t0 , t). (4.18)

4.1 Feynman Propagator


Next we will try to solve the Feynman propagator in D dimensions for Dirac fermions. The
propagator in Fourier (momentum) space is given in many textbooks, e.g. [9], however we
are going to solve it in position space in this chapter.
We start with the general definition that the Dirac propagator in D dimensions should
satisfy:
(i∂/ − m)iSF (x, x0 ) = iδ D (x − x0 ). (4.19)

By expanding the propagator SF (x, x0 ) in momentum space we see that

dD p +ip(x−x0 )
Z
(i∂/ − m) e iSF (p) = iδ D (x − x0 ) (4.20)
(2π)D
dD p +ip·(x−x0 )
Z
e (p/ − m)iSF (p) = iδ D (x − x0 ). (4.21)
(2π)D

Using the Fourier transform of the Dirac delta function, we can immediately read of the
expression for the propagator in momentum space:

1
iSF (p) = . (4.22)
p/ − m
34 CHAPTER 4. FERMIONIC PROPAGATORS

This is the solution of the Dirac propagator in momentum space. We can obtain an
expression in position space if we Fourier transform this back to position space. This gives
us the following:

dD p +ip·(x−x0 )
Z
0 1
iSF (x, x ) = D
e (4.23)
(2π) (p/ − m)
D
d p +ip·(x−x0 ) p/ + m
Z
= e . (4.24)
(2π)D p2 − m2

4.1.1 Contour Integration

Now we need to perform the integral over all the momenta to find an expression that no
longer contains any integrals that are not evaluated. First we perform a contour integration
over p0 to deal with the poles that this expression has. To pull out the p/ + m factor, we
rewrite the previous expression as:

dD p +ip·(x−x0 )
Z
0 ∂
µ 1
iSF (x, x ) = (iγ + m) e . (4.25)
∂xµ (2π) D p − m2
2

It is clear that this expression has poles at

1 1 −1
= = 0 ,
p2 −m2 0 2 2
−(p ) + Ep (p − Ep )(p0 + Ep )

realizing from the definitions we are employing here that

0 0 (x0 −x00 ) ~0
e+ip·(x−x ) = e−ip e+i~p·(~x−x ) .

0
Now we have to decide how to close the contour. For x0 > x00 , eip·(x−x ) → 0 for p0 → −i∞,
so we close the lower half of the plane. The residue at p0 = Ep is then
0 00 )
2πeiEp (x −x
2Ep

For x0 < x00 , we have to close the upper half of the plane, and the residue is
0 −x00 )
−2πe−iEp (x
2Ep
4.1. FEYNMAN PROPAGATOR 35

at p0 = −Ep . There arises an extra minus from the counterclockwise contour. Therefore
we have for the integral in (4.25):

0 00
dD−1 p i~p·(~x−x~0 ) eiEp (x −x )
Z
0 ∂
iSF (x, x ) = (iγ µ
+ m) e θ(x0 − x00 )
∂xµ (2π)D−1 2Ep
0 00
dD−1 p i~p·(~x−x~0 ) e−iEp (x −x )
Z

+ (iγµ
+ m) e θ(x00 − x0 ). (4.26)
∂xµ (2π)D−1 2Ep

Now for notational convenience we define ∆x = x − x0 and using that Ep2 = p~2 + m2 :


i p~2 +m2 (∆x0 )
dD−1 p i~p·(∆x)
Z
∂ ~ e
iSF (x, x0 ) = (iγ µ µ + m) e p θ(∆x0 )
∂x (2π)D−1 2 p~2 + m2

Z D−1 −i p~2 +m2 (∆x0 )
µ ∂ d p i~ ~
p·(∆x) e
+ (iγ + m) e p θ(−∆x0 ). (4.27)
∂xµ (2π)D−1 2 p~2 + m2

From this equation we can read off iSF (~k, t, t0 ), using iγ µ ∂x∂ µ = ~γ · ~k − iγ 0 ∂t

:

0
∂ eiE~k (t−t )
iSF (~k, t, t0 ) = (~γ · ~k − iγ 0 + m) θ(t − t0 )
∂t 2E~k
0
∂ e−iE~k (t−t ) 0
+ (~γ · ~k − iγ 0 + m) θ(t − t). (4.28)
∂t 2E~k

4.1.2 Spatial Integrals

The spatial parts of the integral over p are left in (4.27). For this we need the identity
from [21] equation number (75):

Z
dD−1 p i~p·~x 2
Z ∞ J D−3 (pk~xk)
D−1
e f (k~pk) = D−1 dp pD−2 2
D−3 f (k~pk). (4.29)
(2π) (4π) 2 0 ( 12 pk~xk) 2

This identity can be shown to hold by first integrating it in spherical coordinates:


36 CHAPTER 4. FERMIONIC PROPAGATORS


dD−1 p i~p·~x
Z Z Z
dp
e f (k~pk) = f (k~pk) dΩD−2 ei~p·~x (4.30)
(2π)D−1 (2π) D−1
Z0 ∞ Z π
dp
= D−1
pD−2 f (k~pk)ΩD−3 dθ sinD−3 θeik~pkk~xk cos θ (4.31)
0 (2π) 0
Z ∞ Z 1
dp D−4
= D−1
pD−2 f (k~pk)ΩD−3 dy(1 − y 2 ) 2 eik~pkk~xky . (4.32)
0 (2π) −1

Note that in the second line we choose p~ = k~pkẑ. The ΩD−3 is the surface of a D − 3-
dimensional sphere, given by
2π (D−2)/2
ΩD−3 =
Γ( D−2
2
)
Evaluating the integral, indeed gives the identity. Note that we used that f (k~pk) only
depends on the magnitude of p~, such that in spherical coordinates it only depends on the
radial coordinate and does not come into play in the integration over the angles.
We can apply the identity to carry out the integral (4.27):


~
!
J D−3 (pk∆xk) p2 +m2 (∆x0 )
ei
Z
∂ dp
iSF (x, x0 ) =(iγ µ µ + m) D−1 pD−2 2
θ(∆x0 )
~ D−3
p
∂x (4π) 2 ( 12 pk∆xk) 2 2
2 p +m 2

~ 0
!

Z
dp J D−3 (pk∆xk) 2 2
e−i p +m (∆x )
+(iγ µ + m) D−1 p
D−2 2
θ(−∆x0 ).
~ D−3
p
∂xµ (4π) 2 ( 12 pk∆xk) 2 2
2 p +m 2

(4.33)

To perform this integral, standard integrals involving Bessel functions are necessary. We
have the following identity from Gradshteyn and Ryzhik (6.596.10) [3]:

√ 2 µ−ν−1
Z ∞ p
2 µ uν v − u2 (2)

Jν (ux)Hµ(2) (v 2 2 2 ν+1
x + y )(x +y ) 2 x dx = µ Hµ−ν−1 (y v 2 − u2 ).
0 v y
(4.34)

With the requirements √<[µ] < <[ν], <[ν] > −1, u > 0,√v > 0 and y > 0. There is another
requirement that arg[ v 2 − u2 ] = 0 for v > u and arg[ v 2 − u2 ] = − π2 for u > v. It turns
out to be easier to write this expression in terms of Hankel functions, these are denoted as
(2)
Hij . Plug in the definition for H 1 :
2

r
(1),(2) 2 e±iz
H1 (z) = . (4.35)
2 πz ±i
4.1. FEYNMAN PROPAGATOR 37

In general we have the following identity for the complex conjugate of Hankel functions:
∗ (2)
Hν(1) (z) = Hν ∗ (z ∗ )
such that in the case that is relevant for us
 ∗
(1) (2)
H 1 (z) = H 1 (z).
2 2

In order to solve equation (4.33) we first write the complex conjugate of it. Next we
can apply this identity to solve the complex conjugate of the integral and finally complex
conjugate back to get what we need. The part of the integral we are about to solve we
define as I for clarity:


~
!
Z i p2 +m2 (∆x0 )
J D−3 (pk∆xk)
dp
D−2 e
I= D−1 p
2
D−3
p θ(∆x0 )
(4π) 2 (p) 2 p2 + m2

~
!
Z J D−3 (pk∆xk) −i p2 +m2 (∆x0 )
dp D−2 e
+ D−1 p
2
D−3
p θ(−∆x0 ). (4.36)
2
p +m 2
(4π) 2 (p) 2

The complex conjugate of I reads


~
!
Z J D−3 (pk∆xk) −i p2 +m2 (∆x0 )
∗ dp D−2 e
I = D−1 p 2
D−3
p θ(∆x0 )
2
p +m 2
(4π) 2 (p) 2

~ 0
!
Z
dp J D−3 (pk∆xk) 2 2
ei p +m (−∆x )
D−2
+ D−1 p
2
D−3
p θ(−∆x0 ) (4.37)
2
p +m 2
(4π) 2 (p) 2
r

π∆x0
Z p 1 D−1
=−i ~
dp J D−3 (pk∆xk)H (2)
(∆x 0
p2 + m2 )(p2 + m2 )− 4 p 2
θ(∆x0 )
µ
0
2 2
Z ∞ r
~ (2) 0
p
2 2 − 1 D−1 π∆x0
−i dp J D−3 (pk∆xk)Hµ (−∆x p2 + m2 )(p + m ) 4 p 2 θ(−∆x0 ).
0
2 2
(4.38)

This is in the right form to apply the identity from Gradshteyn and Ryzhik:

! 2−D
D−3 ~ 2
(∆x0 )2 − k∆xk
4
(2)
 q 
∗ ~
I = − ik∆xk 2 H 2−D ~ 2 θ(∆x0 )
m (∆x0 )2 − k∆xk
m 2

! 2−D
D−3 ~ 2
(∆x0 )2 − k∆xk
4
(2)
 q 
~
− ik∆xk 2 H 2−D ~ 2 θ(−∆x0 ).
m (∆x0 )2 − k∆xk
m 2

(4.39)
38 CHAPTER 4. FERMIONIC PROPAGATORS

This integral solution has the following requirements in this case: D > 4 (so we have to
~ > 0, ∆x0 > 0 and m > 0.
analytically continue the result for other dimensions), k∆xk
Now we need to take a closer look at the requirement for the argument of the interval. For
the argument to satisfy
~ 2] = − π
q
arg[ (∆x0 )2 − k∆xk
2
we need to shift (∆x0 )2 to (∆x0 + i)2 , when ∆x0 > 0. If ∆x0 < 0, the shift should be
to (∆x0 − i)2 . Remember that the argument has to be zero when ∆x0 > k∆xk, ~ which is
satisfied since  is small, and the argument of the square root of a positive number plus
a small imaginary piece indeed goes to zero. When ∆x0 < k∆xk, ~ we have the square
root of a negative number plus an imaginary piece. The sign of the imaginary piece tells
us in which quadrant of the complex plane we are. For a small positive imaginary piece
the argument is π/2, for a negative piece −π/2. This is necessary to make the integral
convergent. After we include this subtle point, we can again take the complex conjugate
to find the integral we need for solving the propagator:

! 2−D
D−3 ~ 2
(∆x0 − i)2 − k∆xk
4
(1)
 q 
~
I =ik∆xk 2 H 2−D ~ 2 θ(∆x0 )
m (∆x0 − i)2 − k∆xk
m 2

! 2−D
D−3 ~ 2
(∆x0 + i)2 − k∆xk
4
(1)
 q 
~
+ ik∆xk 2 H 2−D ~ 2 θ(−∆x0 ).
m (∆x0 + i)2 − k∆xk
m 2

(4.40)

We are ready to insert the integral back into (4.33)

! 2−D
∂ −2−D 2−D ~ 2
(∆x0 − i)2 − k∆xk
4

iSF (x, x0 ) =(−i)(iγ µ µ + m)2 2 π 2


∂x m2
q
(1) ~ 2 ) θ(∆x0 )
× H 2−D (m (∆x0 − i)2 − k∆xk
2
! 2−D
~ 2
(∆x0 + i)2 − k∆xk
4
µ ∂ −2−D 2−D
+(−i)(iγ + m)2 2 π 2 (4.41)
∂xµ m2
q
(1) ~ 2 ) θ(−∆x0 ).
× H 2−D (m (∆x0 + i)2 − k∆xk
2

~ 2 − (|∆x0 | − i)2 , we incorporate the convergence factor:


If we now define ∆x2++ ≡ k∆xk
4.1. FEYNMAN PROPAGATOR 39

  2−D
0 ∂ µ 1 2π∆x++ 2
(1)
iSF (x, x ) = (−i)(iγ + m) H 2−D (m∆x++ ). (4.42)
∂xµ 4 m 2

4.1.3 Massless Limit

To calculate the massless limit directly from this final expression, we have to take the limit
m → 0. To do this, we have to expand the Hankel function. First we express the Hankel
function as a sum of Bessel functions, and then apply the limit:

iπν
−ie− 2  − iπν iπν

Hν(1) (z) = e 2 Jν (z) − J−ν (z)e 2 . (4.43)
sin πν

We know for the Bessel function expansion around z = 0:

(z/2)ν (z/2)2
 
2
Jν = 1− + O(z ) , (4.44)
Γ(ν + 1) (ν + 1)

such that

iπν
−ie− 2 (z/2)ν (z/2)2 (z/2)−ν (z/2)2
    
− iπν iπν
Hν(1) (z) = e 2 1− −e 2 1−
sin (πν) Γ(ν + 1) (ν + 1) Γ(−ν + 1) (−ν + 1)
(4.45)

2−D
To show that this result is finite in D = 4, we plug in ν = 2
and show that there is no
divergence around D = 4:

iπ(D−2)
(1) ie− 4
H 2−D (z) =
2 sin (π(D − 2)/2)
2−D
! D−2
!!

iπ(D−2) (z/2) 2 (z/2)2 iπ(D−2) (z/2) 2 (z/2)2
× e 4 1 − 2−D −e 4 1 − D−2 .
Γ( 2−D
2
+ 1) ( 2 + 1) Γ( D−2
2
+ 1) ( 2 + 1)
(4.46)

D−2
The second term on the last line goes as (z/2) 2 , which is linear for D = 4. The first
2−D
term goes as (z/2) 2 , which diverges. Rewrite this to isolate the divergence:
40 CHAPTER 4. FERMIONIC PROPAGATORS

iπ(D−2)  
(1) ie 2 D−2
H 2−D (z) = Γ
2 π 2
  D−2 ! !
2 2 (z/2)2 −
iπ(D−2) Γ(− D−4
2
)  z D−2  (z/2)2
× 1+ D−4
−e 2 1− .
z 2
Γ(D − 2) 2 D/2
(4.47)

We can directly read of the zeroth order term, now we show that for the first order in z
the divergence cancels

D−4
!
 z 2 2 e−iπ e−iπ 2 (1 − D−4
2
ψ(1)) 2
+ (1 + (D − 4) log (z/2)) (4.48)
2 D−4 Γ(2)(1 + D−4
2
ψ(2)) D−4
 z 2
= (iπ + ψ(1) + ψ(2) − log z/2) + O(D − 4), (4.49)
2

such that we find for D → 4 in the massless case:

iπ(D−2)     D−2
(1) ie 2 D−2 2 2 iz
H 2−D (z) = Γ − (iπ + ψ(1) + ψ(2) − log (z/2)) (4.50)
2 π 2 z π2
D→4 −i 2
→ . (4.51)
π z

The − πi z2 (iπ + ψ(1) + ψ(2) − log (z/2) term is the first order correction to a small mass in
D = 4. Plug this final approximation in (4.42) to find:

 −1
0 ∂ 1 2π∆x++
µ −i 2
iSF (x, x ) =(−i)iγ µ
(4.52)
∂x 4 m π m∆x++
∂ 1 1
=iγ µ µ 2 . (4.53)
∂x 4π ∆x2++

We can check that this limit is correct by going back to (4.33) and calculate the massless
case from here onwards (before calculating the final radial integral). The massless version
of this equation reads:
4.1. FEYNMAN PROPAGATOR 41

~
!

Z
dp J D−3 (pk∆xk) eip(∆x
0)

=iγ µ µ D−1 pD−2 2


θ(∆x0 )
∂x (4π) 2 ~ D−3
( 12 pk∆xk) 2 2p
~
!

Z
dp J D−3 (pk∆xk) e−ip(∆x
0)
µ D−2
−iγ D−1 p 2
θ(−∆x0 ). (4.54)
∂xµ (4π) 2 ~ D−3
( 12 pk∆xk) 2 2p

Now we can apply:


(2α)ν Γ(ν + 12 )
Z
ν −βp
p Jν (p α)e =√ 1 . (4.55)
0 π(α2 + β 2 )ν+ 2

From 6.623.1 in [3]. This holds for ν > − 12 , which means for us D > 2. More importantly,
also <[α] > 0. This is necessary to ensure the convergence of the exponential function,
but note that we have a completely imaginary argument for our exponential (ip∆x0 ). We
need to add a small imaginary part to ∆x0 , ∆x0 → ∆x0 + i. Now we can apply (4.55)
and find:

∂ Γ( D2 − 1) 1
iSF (x, x0 ) =iγ µ θ(∆x0 )
~ 2 + (i∆x0 + )2 ) D2 −1
∂xµ 4π D/2 ((k∆xk)
∂ Γ( D2 − 1) 1
+ iγ µ θ(−∆x0 ) (4.56)
~ 2 + (i∆x0 − )2 ) D2 −1
∂xµ 4π D/2 ((k∆xk)
∂ Γ( D2 − 1) 1
=iγ µ D θ(∆x0 )
µ
∂x 4π D/2 ~ 2 0 2
((k∆xk) − (∆x − i) ) 2 −1

∂ Γ( D2 − 1)
µ 1
+ iγ D θ(−∆x0 ). (4.57)
µ
∂x 4π D/2 ~ 2 0 2
((k∆xk) − (∆x + i) ) 2 −1

~ 2 − (|∆x0 | − i)2 , we incorporate the convergence factor. This


If we define ∆x2++ ≡ k∆xk
allows us to combine the dependence on θ(∆x0 ) in this new quantity:

0 ∂ Γ( D2 − 1)
µ 1
iSF (x, x ) = iγ µ D/2 D . (4.58)
∂x 4π (∆x++ ) 2 −1
2

Comparing this with (4.53) leads us to conclude that this agrees with the earlier calculated
solution in the massless limit for D = 4.
42 CHAPTER 4. FERMIONIC PROPAGATORS
Chapter 5

Dirac Flavor Mixing

In this chapter we will solve a model of two fermionic species of a Dirac nature with
correlators. This procedure allows us to consider mixed states instead of only pure states.
The two point functions involved originate from the Schwinger Keldysh formalism. This
correlator approach was first discussed for a quantum mechanical system of two coupled
scalar harmonic oscillators in [22]. In this paper the approach is explained and compared
to the master equation approach to decoherence for bilinear coupled simple harmonic
oscillators. It is shown that the master equation approach suffers from trouble with an
unbounded entropy growth at later times, where the correlator approach does not have
this problem.

We will calculate the different correlators involved in our system to solve for the evolution
of the system. This will be done for a general Dirac fermion, and should therefore also
hold in the case of a neutrino.

The strategy is first to construct the statistical propagator for the system, and then extract
the correlators from this statistical propagator. There may be other routes to calculating
these correlators, however we choose this one because in other, more complicated, physical
systems it is often possible to calculate the statistical propagator perturbatively. This may
not be the case for other methods of obtaining the correlators. This derivation will be
done for initial conditions that also allow initial entanglements. Note that this method of
is similar to the one in [22], and similarly the expression for the statistical propagator is
exact, we do not have to use any form of perturbation expansion.

First the equations of motion will be obtained, after which the mass terms can be diag-
onalized. In this diagonal basis the equations of motion can be solved much more easily,
because the different flavor states will decouple. The different flavor states will be assumed
to mix through off diagonal states in the mass matrix, so a diagonal mass matrix will not
lead to any mixing. After solving in this diagonal basis, we can rotate back to the original
basis, in which mixing is present and look at the results.

43
44 CHAPTER 5. DIRAC FLAVOR MIXING

5.1 Set Up
In this model we consider the following Lagrangian, which consists out of a kinetic part
and a Dirac mass term:
L = iψ̄ ∂/ψ − ψ̄L M ψR − ψ̄R M † ψL . (5.1)

The flavor mixing is included in the mass term. We take a two by two mass term (M ) in
flavor space and the off diagonal terms give rise to flavor mixing. The dimensionality of
the mass matrix dictates the number of flavors considered. In a full model for neutrino
oscillations, one would like to include a third flavor, however for simplicity we refrain from
doing so. Most mixing experiments also take only two flavors into account because in
experimental set ups the third flavor can often safely be neglected. This number of flavors
also effects the structure of the spinor ψ, which will be discussed later.
Using the projection operators, which are also stated in appendix A for completeness,
defined by

1 − γ5
ψL = ψ (5.2)
2
1 + γ5
ψR = ψ (5.3)
2
1 + γ5
ψ̄L = ψ̄ (5.4)
2
1 − γ5
ψ̄R = ψ̄ , (5.5)
2

the Lagrangian can now be rewritten. We can eliminate the explicit left handed and
right handed parts. The equations of motion for this Lagrangian can be found by varying
the Lagrangian with ψ̄. Also, we decompose the mass matrix in a hermitian and anti
hermitian part. This was also done in the previous chapter where the equations of motion
were discussed with a kinetic approach. The equations of motion are now given by

1 + γ5 1 − γ5
 

i∂/ − (M +M ) ψ=0 (5.6)
2 2
5 5
 
1 + γ 1 − γ
iγ ∂t − ~γ · ~k − (M
0
+M †
) ψ=0 (5.7)
2 2
5 5
 
1 + γ 1 − γ
∂t + iγ ~γ · ~k + iγ (M
0 0
+M †
) ψ = 0. (5.8)
2 2

Here we transformed the spacial coordinates to momentum space, by first writing out the
5.1. SET UP 45

term including ∂/. This was done to allow us to identify the helicity in these equations of
motion.

5.1.1 Chirality and Helicity Decomposition

We can decompose the spinor in chirality and helicity spinors. For a more complete
discussion on the concepts of chirality and helicity refer to section C.2. This decomposition
gives us a helicity eigenspinor ξh and left- and right-handed chirality spinors ψLh and ψRh :

X ψRh 
ψ= ⊗ ξh (5.9)
ψLh
h

ˆ
Next, we can also see the helicity factor coming out of the equations: ĥ = ~k · ~γ γ 0 γ 5 . We
find the following two equations

(i∂t + hk)ψRh − M † ψLh = 0 (5.10)


(i∂t − hk)ψLh − M ψRh = 0. (5.11)

At this point we need to use the diagonalization procedure of the mass matrix again by a
rotation in flavor space. This is further specified in appendix B.1. The rotation matrices
UL and UR are given by

 
cos θ̄ sin θ̄eiω̄
UR = −iω̄ (5.12)
− sin θ̄e cos θ̄
 
cos θ sin θeiω
UL = . (5.13)
− sin θe−iω cos θ

The two equations can be diagonalized with the help of (B.1) to find

d
(i∂t + hk)ψRh − Md† ψLh
d
=0 (5.14)
d d
(i∂t − hk)ψLh − Md ψRh = 0. (5.15)

Here we defined the spinors in this diagonal mass base as


46 CHAPTER 5. DIRAC FLAVOR MIXING

d
ψLh ≡ UL† ψLh (5.16)
d
ψRh ≡ UR† ψRh , (5.17)

which confirms that UL† and UR† respectively operate on the left- and right-handed spinors.
The mass matrix in the diagonal basis is given by

UL† M UR = Md

and
UR† M † UL = MD†
If we separate the left- and right-handed spinors by combing the two equations we find the
following two quadratic equations:

∂t2 + k 2 + |Md |2 ψRh


 d
=0 (5.18)
∂t2 + k 2 + |Md |2 ψLh
d

=0. (5.19)

These two equations look like the scalar case in [22], however we will see later that equation
(5.16) will give us additional constraints that are not present in the scalar case. The
quadratic equations can be written as

d2 d
Ψ + Ωd Ψd = 0, (5.20)
dt2

where we defined the matrix Ωd as the following:


 2 
ω1 0 0 0
 0 ω22 0 0 
Ωd = 
 0 0 ω12 0 
 (5.21)
0 0 0 ω22
 2 
k + (|Md |11 )2 0 0 0
 0 k 2 + (|Md |22 )2 0 0 
= . (5.22)
 0 0 k 2 + (|Md |11 )2 0 
2 2
0 0 0 k + (|Md |22 )

The two new frequencies ω12 ≡ k 2 + (|Md |11 )2 and ω22 ≡ k 2 + (|Md |22 )2 were defined to
improve the clarity in order to solve the system. The spinor Ψd is given by:
5.1. SET UP 47

 d 
 d  ψ1Rh
d 
ψRh  ψ2Rh
Ψd = d = d .
 (5.23)
ψLh ψ1Lh
d
ψ2Lh

For completeness, this spinor is only half of the story. The second half has an opposite
helicity, but is in every other aspect equal. We will only continue the discussion of the
aforementioned spinor, since the other half can always be acquired by taking the helicity to
the opposite sign. Equation (5.20) is now a decoupled second order differential equation.
Its solution, with several constants, can be written as

 d
  
ψ1Rh Â1Rh cos(ω1 t) + B̂1Rh sin(ω1 t)
d 
ψ2Rh Â2Rh cos(ω2 t) +ˆˆB2Rh sin(ω2 t)
 
 d = . (5.24)
ψ1Lh  
 Â1Lh cos(ω1 t) + B̂1Lh sin(ω1 t) 
d
ψ2Lh Â2Lh cos(ω2 t) + B̂2Lh sin(ω2 t)

The new constant operators (Âij , B̂ij ) we introduced in these general solutions are not
independent, they still have to abide to (5.14). There are two main differences with the
scalar case, here there are more initial constants to fix and there is second equation, (5.14),
which also has to be taken into account. This allows us to express the B̂ij as functions of
Âij and other constants we already acquired such as the ωi :

(Md )∗11 A1L − hk Â1Rh


B̂1Rh =
iω1
(Md )11 A1R + hk Â1Lh
B̂1Lh =
iω1
ˆˆB2Rh (Md )∗22 A2L − hk Â2Rh
=
iω2
(Md )22 Â2Rh + hk Â2Lh
B̂2Lh = . (5.25)
iω2

Now we go back to the general solution to apply these extra derived constraints to find
48 CHAPTER 5. DIRAC FLAVOR MIXING

 (M )∗ Â
d 11 1Lh − hk Â1Rh

d
ψ1Rh = Â1Rh cos(ω1 t) + sin(ω1 t)
iω1
 (M )∗ Â
d 22 2Lh − hk Â2Rh

d
ψ2Rh = Â2Rh cos(ω2 t) + sin(ω2 t)
iω2
 (M ) Â
d 11 1Rh + hk Â1Lh

d
ψ1Lh = Â1Lh cos(ω1 t) + sin(ω1 t)
iω1
 (M ) Â
d 22 2Rh + hk Â2Lh

d
ψ2Lh = Â2Lh cos(ω2 t) + sin(ω2 t). (5.26)
iω2

These four equations solve the system completely when the initial conditions are specified
through fixing the Aij constants. The elimination of the Bij constants assures that the
given equations indeed also abide to the Dirac equations of motion.

5.2 Statistical Propagator


Next step is to solve for the statistical propagator of the system. In order to do so, we have
to use several two point functions [36]. These four two point functions originate from the
Schwinger-Keldysh formalism that was discussed previously. We concluded that we could
express the statistical propagator (two point function) as

1 1
Fψ (t, t0 ) = (iS −+ (t, t0 ) + iS +− (t, t0 )) = h[ψ̂(t), ψ̂ † (t0 )]i. (5.27)
2 2

We will proceed to calculate the statistical propagator for different states. By using the
definition of the statistical propagator we can make a connection with the number density.
The statistical propagator of for instance the left handed part of the first flavor is

1
F1Lh (t; t0 ) = h[ψ̂1Lh (t0 ), ψ̂1Lh (t)† ]i. (5.28)
2

The particle number operator is given by


N̂1Lh (t) = ψ̂1Lh (t)ψ̂1Lh (t0 ). (5.29)

Note that at equal time the statistical propagator can be expressed as a function of the
expectation value of the particle number operator of that specific state, which gives us a
physical interpretation of the statistical propagator.
5.2. STATISTICAL PROPAGATOR 49

nij (t) = hN̂ij (t)i = hψ̂i (t)† ψ̂j (t)i (5.30)

The statistical propagator in terms of the average particle number is given by

Fij (t, t) = δij − 2nij (t) (5.31)

We will also encounter statistical propagators of mixed states, such as

1
F1L1R (t; t0 ) = h[ψ̂1Lh (t0 ), ψ̂1Rh (t)† ]i. (5.32)
2

At equal time this is a function of the expectation value of the particle number operator
of a mixed state, in this example of the state combining the left and right handed chirality
of the first flavor.

5.2.1 Calculating the Statistical Propagator

We can now solve for the statistical propagator of the right handed first species. The
goal of this discussion will be to relate different expectation values with the statistical
propagator. This will allow us to apply the solutions we have found previously for the
system and express the statistical propagator in terms of initial constants.
The statistical propagator of the first flavor and the right handed part is

1
F1R1R (t; t0 ) = h[ψ̂1Rh (t0 ), ψ̂1Rh (t)† ]i (5.33)
2
1h
= cos2 θ̄h[ψ̂1Rhd
(t0 ), ψ̂1Rh
d
(t)† ]i + sin2 θ̄h[ψ̂2Rh
d
(t0 ), ψ̂2Rh
d
(t)† ]i
2  i
+ cos θ̄ sin θ̄ h[ψ̂1Rhd
(t0 ), ψ̂2Rh
d
(t)† ]ie−iω̄ + h[ψ̂2Rh
d
(t0 ), ψ̂1Rh
d
(t)† ]ieiω̄ . (5.34)

The statistical propagator is then given by one half times the expectation value of the
commutator of the solution in the original frame. Next step is to plug in the rotation
of these components of the spinor to the components in the diagonal frame. We do this
because in the diagonal frame we have solved the equations that govern the motion of these
spinorial components in the previous section. The explicit form of this rotation is given by
(5.16). This gives us using the unitarity of the rotation
50 CHAPTER 5. DIRAC FLAVOR MIXING

d
ψRh = UR ψRh (5.35)
† d † †
ψRh = (ψRh ) UR , (5.36)

which reads in component form as

    d 
ψ1Rh cos θ̄ sin θ̄eiω̄ ψ1Rh
=
ψ2Rh − sin θ̄e−iω̄ cos θ̄ d
ψ2Rh
 
† †
 d † d †
 cos θ̄ − sin θ̄eiω̄
ψ1Rh ψ2Rh = (ψ1Rh ) (ψ2Rh ) . (5.37)
sin θ̄e−iω̄ cos θ̄

Next we write the statistical propagator out with the general solution that was obtained.
This general solution does not have any initial constraints built in, we are still free to fix
these. For the right handed statistical propagator for the first species we find

cos2 h
F1R1R (t; t0 ) = θ̄ h[Â1Rh , †1Rh ]i cos(ω1 t) cos(ω1 t0 ) + h[B̂1Rh , B̂1Rh

]i sin(ω1 t) sin(ω1 t0 )
2

h[Â1Rh , B̂1Rh ]i h[B̂1Rh , †1Rh ]i i
+ cos(ω1 t) sin(ω1 t0 ) + cos(ω1 t0 ) sin(ω1 t)
2 2
2 h
sin
+ θ̄ h[Â2Rh , †2Rh ]i cos(ω2 t) cos(ω2 t0 ) + h[ˆˆB2Rh ,ˆˆB2Rh

]i sin(ω2 t) sin(ω2 t0 )
2

h[Â2Rh ,ˆˆB2Rh ]i 0 h[B̂2Rh , †2Rh ]i 0
i
+ cos(ω2 t) sin(ω2 t ) + cos(ω2 t ) sin(ω2 t)
2 2
sin 2θ̄ h
+ h[Â1Rh , †2Rh ]i cos(ω1 t0 ) cos(ω2 t)e−iω̄ + h[Â1Rh , B̂2Rh †
]i cos(ω1 t0 ) sin(ω2 t)e−iω
4
+ h[B̂1Rh , †2Rh ]i sin(ω1 t0 ) cos(ω2 t)e−iω̄ + h[B̂1Rh , B̂2Rh †
]i sin(ω1 t0 ) sin(ω2 t)e−iω

+ h[Â2Rh , †1Rh ]i cos(ω1 t) cos(ω2 t0 )eiω̄ + h[Â2Rh , B̂1Rh ]i cos(ω1 t) sin(ω2 t0 )eiω
i
† 0 iω̄ † 0 iω
+ h[B̂2Rh , Â1Rh ]i sin(ω1 t) cos(ω2 t )e + h[B̂2Rh , B̂1Rh ]i sin(ω1 t) sin(ω2 t )e .
(5.38)

And here we can apply (5.14) to eliminate B̂iR . These B̂ij constants are after all not for us
to fix, they are already determined in terms of the Âij components by the Dirac equation:
5.2. STATISTICAL PROPAGATOR 51

F1R1R (t; t0 ) =
cos2 h †

θ̄ h[Â1Rh , Â1Rh ]i cos(ω1 t) cos(ω1 t0 )
2
h|~k| 0 0 0 |~k|2 0

+ (− cos ω1 t sin ω1 t + cos ω1 t sin ω1 t ) + 2 sin ω1 t sin ω1 t
iω1 ω1
2
|(Md )11 |
+ h[Â1Lh , †1Lh ]i sin ω1 t sin ω1 t0
ω12
h|~k| (M ∗ )11
+ h[Â1Lh , †1Rh ]i(cos ω1 t + sin ω1 t) d sin ω1 t0
iω1 iω1
~
h|k| (M ∗ )11 i
− h[Â1Rh , †1Lh ]i(cos ω1 t0 − sin ω1 t0 ) d sin ω1 t
iω1 iω1
2 h
sin 
+ θ̄ h[Â2Rh , †2Rh ]i cos(ω2 t) cos(ω2 t0 )
2
hk |~k|2 
+ (− cos ω2 t sin ω2 t0 + cos ω2 t0 sin ω2 t0 ) + 2 sin ω2 t sin ω2 t0
iω2 ω2
2
|(Md )22 |
+ h[Â2Lh , †2Lh ]i sin ω2 t sin ω2 t0
ω22
h|~k| (M ∗ )22
+ h[Â2Lh , †2Rh ]i(cos ω2 t + sin ω2 t) d sin ω2 t0
iω2 iω2
h|~k| (M ∗ )22 i
− h[Â2Rh , †2Lh ]i(cos ω2 t0 − sin ω2 t0 ) d sin ω2 t
iω2 iω2
sin 2θ̄ h  h|~k| h|~k|
+ h[Â1Rh , †2Rh ]i cos(ω2 t) cos(ω1 t0 ) − sin ω1 t0 cos ω2 t + cos ω1 t0 sin ω2 t
4 iω1 iω2
|~k|2 
+ sin ω1 t0 sin ω2 t e−iω̄
ω1 ω2

 h|~k| h|~k|
+ h[Â2Rh , Â1Rh ]i cos(ω1 t) cos(ω2 t0 ) − sin ω2 t0 cos ω1 t + cos ω2 t0 sin ω1 t
iω2 iω1
~
|k| 2 
+ sin ω2 t0 sin ω1 t eiω̄
ω2 ω1


0 (Md )22 0 h|~k|(Md )22  −iω̄
+ h[Â1Rh , Â2Lh ]i − cos ω1 t sin ω2 t − sin ω1 t sin ω2 t e
iω2 ω1 ω2
 (M ∗ )22 h|~k|(Md )∗22  iω̄
+ h[Â2Lh , †1Rh ]i cos ω1 t sin ω2 t0 d − sin ω1 t sin ω2 t0 e
iω2 ω1 ω2
 (M ∗ )11 h|~k|(Md )∗11  −iω̄
+ h[Â1Lh , †2Rh ]i sin ω1 t0 cos ω2 t d − sin ω1 t0 sin ω2 t e
iω1 ω1 ω2
 (Md )11 h|~k|(Md )11  iω̄
+ h[Â2Rh , †1Lh ]i − cos ω2 t0 sin ω1 t − sin ω2 t0 sin ω1 t e
iω1 ω1 ω2


0 (Md )∗11 (Md )22  −iω̄
− h[Â1Lh , Â2Lh ]i sin ω1 t sin ω2 t e
ω1 ω2
 (Md )∗22 (Md )11  iω̄ i
− h[Â2Lh , †1Lh ]i sin ω1 t sin ω2 t0 e . (5.39)
ω1 ω2
52 CHAPTER 5. DIRAC FLAVOR MIXING

We see that we obtain an expression in terms of expectation values for the initial constants
hÂijh i. Here we see why these initial constants can indeed be operators, as we see that we
are taking expectation values of them.
Next, for the left handed particles of the first species we can do a likewise analysis, which
gives us

1
F1L1L (t; t0 ) = h[ψ̂1Lh (t0 ), ψ̂1Lh (t)† ]i (5.40)
2
1h
= cos2 θh[ψ̂1Lh d
(t0 ), ψ̂1Lh
d
(t)† ]i + sin2 θh[ψ̂2Lh
d
(t0 ), ψ̂2Lh
d
(t)† ]i
2  i
+ cos θ sin θ h[ψ̂1Lh d
(t0 ), ψ̂2Lh
d
(t)† ]ie−iω + h[ψ̂2Lh
d
(t0 ), ψ̂1Lh
d
(t)† ]ieiω . (5.41)

In the last line we use the rotation for the left handed species, given by

d
ψLh = UL ψLh (5.42)
† d † †
ψLh = (ψLh ) UL . (5.43)

This allows us again to express the statistical propagator in terms of the diagonal compo-
nents, for which we have solved the system:

cos2 h
F1L1L (t; t0 ) = θ h[Â1Lh , †1Lh ]i cos(ω1 t) cos(ω1 t0 ) + h[B̂1Lh , B̂1Lh

]i sin(ω1 t) sin(ω1 t0 )
2

h[Â1Lh , B̂1Lh ]i 0 h[B̂1Lh , †1Lh ]i i
+ cos(ω1 t) sin(ω1 t ) + cos(ω1 t0 ) sin(ω1 t)
2 2
sin2 h
+ θ h[Â2Lh , †2Lh ]i cos(ω2 t) cos(ω2 t0 ) + h[B̂2Lh , B̂2Lh†
]i sin(ω2 t) sin(ω2 t0 )
2

h[Â2Lh , B̂2Lh ]i h[B̂2Lh , †2Lh ]i i
+ cos(ω2 t) sin(ω2 t0 ) + cos(ω2 t0 ) sin(ω2 t)
2 2
sin 2θ h
+ h[Â1Lh , †2Lh ]i cos(ω1 t) cos(ω2 t0 )e−iω + h[Â1Lh , B̂2Lh †
]i cos(ω1 t) sin(ω2 t0 )e−iω
4
+ h[B̂1Lh , †2Lh ]i sin(ω1 t) cos(ω2 t0 )e−iω + h[B̂1Lh , B̂2Lh †
]i sin(ω1 t) sin(ω2 t0 )e−iω
+ h[Â2Lh , †1Lh ]i cos(ω1 t) cos(ω2 t0 )eiω + h[Â2Lh , B̂1Lh

]i cos(ω1 t) sin(ω2 t0 )eiω
i
† 0 iω † 0 iω
+ h[B̂2Lh , Â1Lh ]i sin(ω1 t) cos(ω2 t )e + h[B̂2Lh , B̂1Lh ]i sin(ω1 t) sin(ω2 t )e .
(5.44)

We eliminate the B̂iL components to obtain the following form for the statistical propagator
5.2. STATISTICAL PROPAGATOR 53

F1L1L (t; t0 ) =
cos2 h †

θ h[Â1Lh , Â1Lh ]i cos(ω1 t) cos(ω1 t0 )
2
h|~k| 0 0 0 |~k|2 0

− (− cos ω1 t sin ω1 t + cos ω1 t sin ω1 t ) + 2 sin ω1 t sin ω1 t
iω1 ω1
2
|(Md )11 |
+ h[Â1Rh , †1Rh ]i sin ω1 t sin ω1 t0
ω12
h|~k| (Md )11
+ h[Â1Rh , †1Lh ]i(cos ω1 t − sin ω1 t) sin ω1 t0
iω1 iω1
~
h|k| (Md )11 i
− h[Â1Lh , †1Rh ]i(cos ω1 t0 + sin ω1 t0 ) sin ω1 t
iω1 iω1
2 h
sin 
+ θ h[Â2Lh , †2Lh ]i cos(ω2 t) cos(ω2 t0 )
2
h|~k| |~k|2 
− (− cos ω2 t sin ω2 t0 + cos ω2 t0 sin ω2 t0 ) + 2 sin ω2 t sin ω2 t0
iω2 ω2
2
|(Md )22 |
+ h[Â2Rh , †2Rh ]i sin ω2 t sin ω2 t0
ω22
h|~k| (Md )22
+ h[Â2Rh , †2Lh ]i(cos ω2 t − sin ω2 t) sin ω2 t0
iω2 iω2
h|~k| (Md )22 i
− h[Â2Lh , †2Rh ]i(cos ω2 t0 + sin ω2 t0 ) sin ω2 t
iω2 iω2
sin 2θ h  h|~k| h|~k|
+ h[Â1Lh , †2Lh ]i cos(ω2 t) cos(ω1 t0 ) + sin ω1 t0 cos ω2 t − cos ω1 t0 sin ω2 t
4 iω1 iω2
|~k|2 
+ sin ω1 t0 sin ω2 t e−iω
ω1 ω2

 h|~k| h|~k|
+ h[Â2Lh , Â1Lh ]i cos(ω1 t) cos(ω2 t0 ) + sin ω2 t0 cos ω1 t − cos ω2 t0 sin ω1 t
iω2 iω1
~
|k| 2 
+ sin ω2 t0 sin ω1 t eiω
ω2 ω1


0 (Md∗ )22 0 h|~k|(Md )22  −iω
+ h[Â1Lh , Â2Rh ]i − cos ω1 t sin ω2 t + sin ω1 t sin ω2 t e
iω2 ω1 ω2
 (Md )22 h|~k|(Md )∗22  iω
+ h[Â2Rh , †1Lh ]i cos ω1 t sin ω2 t0 + sin ω1 t sin ω2 t0 e
iω2 ω1 ω2
 (Md )11 h|~k|(Md )∗11  −iω
+ h[Â1Rh , †2Lh ]i sin ω1 t0 cos ω2 t + sin ω1 t0 sin ω2 t e
iω1 ω1 ω2
 (M ∗ )11 h|~k|(Md )11  iω
+ h[Â2Lh , †1Rh ]i − cos ω2 t0 sin ω1 t d + sin ω2 t0 sin ω1 t e
iω1 ω1 ω2


0 (Md )∗11 (Md )22  −iω
− h[Â1Rh , Â2Rh ]i sin ω1 t sin ω2 t e
ω1 ω2
 (Md )∗22 (Md )11  iω i
− h[Â2Rh , †1Rh ]i sin ω1 t sin ω2 t0 e . (5.45)
ω1 ω2
54 CHAPTER 5. DIRAC FLAVOR MIXING

5.2.2 Initial Expectation Values

We can derive a set of initial expectation values, which express the expectation values in
terms of the initial constants. Note that because we expressed B̂ij in Âij in (5.25), we only
need to set the initial conditions for Âij , and then B̂ij follows automatically.
This is in contrast with the scalar case, where also the initial momenta had to be fixed.
This was necessary to determine the equivalence of the B̂ij constants. However, here we
have the extra demands on these by the Dirac equation, summarized in (5.25), which fixes
these constants, as was discussed earlier. The following initial correlators for the right
handed correlators fix the right handed part of our system completely:


h[ψ̂1Rh (t0 ), ψ̂1Rh (t0 )]i = cos2 θ̄h[ˆˆA1Rh ,ˆˆA†1Rh ]i + sin2 θ̄h[ˆˆA2Rh ,ˆˆA†2Rh ]i
 
+ sin θ̄ cos θ̄ h[ˆˆA1Rh ,ˆˆA†2Rh ]ie−iω̄ + h[ˆˆA2Rh ,ˆˆA†1Rh ]ieiω̄ (5.46)

h[ψ̂2Rh (t0 ), ψ̂2Rh (t0 )]i = cos2 θ̄h[ˆˆA2Rh ,ˆˆA†2Rh ]i + sin2 θ̄h[ˆˆA1Rh ,ˆˆA†1Rh ]i
 
ˆ ˆ † −iω̄ ˆ ˆ † iω̄
− sin θ̄ cos θ̄ h[ˆA1Rh ,ˆA2Rh ]ie + h[ˆA2Rh ,ˆA1Rh ]ie (5.47)

h[ψ̂1Rh (t0 ), ψ̂2Rh (t0 )]i = cos2 θ̄h[ˆˆA1Rh ,ˆˆA†2Rh ]i − sin2 θ̄h[ˆˆA2Rh ,ˆˆA†1Rh ]e2iω̄ i
 
+ sin θ̄ cos θ̄eiω̄ −h[ˆˆA1Rh ,ˆˆA†1Rh ]i + h[ˆˆA2Rh ,ˆˆA†2Rh ]i . (5.48)

And for the left handed correlators we have:



h[ψ̂1Lh (t0 ), ψ̂1Lh (t0 )]i = cos2 θh[ˆˆA1Lh ,ˆˆA†1Lh ]i + sin2 θh[ˆˆA2Lh ,ˆˆA†2Lh ]i
 
+ sin θ cos θ h[ˆˆA1Lh ,ˆˆA†2Lh ]ie−iω + h[ˆˆA2Lh ,ˆˆA†1Lh ]ieiω (5.49)

h[ψ̂2Lh (t0 ), ψ̂2Lh (t0 )]i = cos2 θh[ˆˆA2Lh ,ˆˆA†2Lh ]i + sin2 θh[ˆˆA1Lh ,ˆˆA†1Lh ]i
 
− sin θ cos θ h[ˆˆA1Lh ,ˆˆA†2Lh ]ie−iω + h[ˆˆA2Lh ,ˆˆA†1Lh ]ieiω (5.50)

h[ψ̂1Lh (t0 ), ψ̂2Lh (t0 )]i = cos2 θh[ˆˆA1Lh ,ˆˆA†2Lh ]i − sin2 θh[ˆˆA2Lh ,ˆˆA†1Lh ]e2iω i
 
iω ˆ ˆ † ˆ ˆ †
+ sin θ cos θe −h[ˆA1Lh ,ˆA1Lh ]i + h[ˆA2Lh ,ˆA2Lh ]i . (5.51)

By reviewing these equations it should be kept in mind that we assume that we know from
the mass matrix the quantities θ, θ̄, ω and ω̄. The only unknowns are the expectation values
5.2. STATISTICAL PROPAGATOR 55

for the initial constants. Previously we have discussed that these equal time commutators
can be seen as a function of the expectation value of the particle number operator.
These initial constants give upon inverting to express the constants as functions of the
initial conditions on the spinorial components. The right handed case:


h[ˆˆA1Rh ,ˆˆA†1Rh ]i = cos2 θ̄h[ψ̂1Rh (t0 ), ψ̂1Rh †
(t0 )]i + sin2 θ̄h[ψ̂2Rh (t0 ), ψ̂2Rh (t0 )]i
 
† †
− sin θ̄ cos θ̄ h[ψ̂1Rh (t0 ), ψ̂2Rh (t0 )]ie−iω̄ + h[ψ̂2Rh (t0 ), ψ̂1Rh (t0 )]ieiω̄ (5.52)

h[ˆˆA2Rh ,ˆˆA†2Rh ]i = sin2 θ̄h[ψ̂1Rh (t0 ), ψ̂1Rh †
(t0 )]i + cos2 θ̄h[ψ̂2Rh (t0 ), ψ̂2Rh (t0 )]i
 
† −iω̄ † iω̄
+ sin θ̄ cos θ̄ h[ψ̂1Rh (t0 ), ψ̂2Rh (t0 )]ie + h[ψ̂2Rh (t0 ), ψ̂1Rh (t0 )]ie (5.53)

h[ˆˆA1Rh ,ˆˆA†2Rh ]i = cos2 θ̄h[ψ̂1Rh (t0 ), ψ̂2Rh †
(t0 )]i + sin2 θ̄e2iω̄ h[ψ̂2Rh (t0 ), ψ̂1Rh (t0 )]i
 
† †
+ sin θ̄ cos θ̄eiω̄ h[ψ̂1Rh (t0 ), ψ̂1Rh (t0 )]i − h[ψ̂2Rh (t0 ), ψ̂2Rh (t0 )]i , (5.54)

and the left handed:

h[ˆˆA1Lh ,ˆˆA†1Lh ]i = cos2 θh[ψ̂1Lh (t0 ), ψ̂1Lh


† †
(t0 )]i + sin2 θh[ψ̂2Lh (t0 ), ψ̂2Lh (t0 )]i
 
† †
− sin θ cos θ h[ψ̂1Lh (t0 ), ψ̂2Lh (t0 )]ie−iω + h[ψ̂2Lh (t0 ), ψ̂1Lh (t0 )]ieiω (5.55)
h[ˆˆA2Lh ,ˆˆA†2Lh ]i = sin2 θh[ψ̂1Lh (t0 ), ψ̂1Lh
† †
(t0 )]i + cos2 θh[ψ̂2Lh (t0 ), ψ̂2Lh (t0 )]i
 
† −iω † iω
+ sin θ cos θ h[ψ̂1Lh (t0 ), ψ̂2Lh (t0 )]ie + h[ψ̂2Lh (t0 ), ψ̂1Lh (t0 )]ie (5.56)
h[ˆˆA1Lh ,ˆˆA†2Lh ]i = cos2 θh[ψ̂1Lh (t0 ), ψ̂2Lh
† †
(t0 )]i + sin2 θe2iω h[ψ̂2Lh (t0 ), ψ̂1Lh (t0 )]i
 
† †
+ sin θ cos θeiω h[ψ̂1Lh (t0 ), ψ̂1Lh (t0 )]i − h[ψ̂2Lh (t0 ), ψ̂2Lh (t0 )]i . (5.57)

The same can be done for mixed correlators, such as hψ̂1Lh (t0 )ψ̂2Rh (t0 )i. We did not
explicitly write out the statistical propagator for these mixed states, however we can obtain
them by carefully considering the initial correlators. As an example, we will calculate
hψ̂1Lh (t0 )ψ̂1Rh (t0 )i.
First we write out the rotation to the diagonal mass frame as always:

h
h[ψ̂1Lh (t0 ), ψ̂1Rh (t0 )† ]i = h[ψ1Lh
d d
(t0 ), ψ1Rh (t0 )† ]i cos θ cos θ̄ + h[ψ2Lh
d d
(t0 ), ψ1Rh (t0 )]† i sin θ cos θ̄eiω
i
d
+ h[ψ1Lh d
(t0 ), ψ2Rh (t0 )† ]i cos θ sin θ̄e−iω̄ + h[ψ2Lh
d d
(t0 ), ψ2Rh (t0 )† ]i sin θ sin θ̄eiω−iω̄ .
(5.58)
56 CHAPTER 5. DIRAC FLAVOR MIXING

d
Next we look at the explicit form of ψ1Lh (t0 ), which is given by Â1Lh as can be seen by
(5.26). This identification is only possible at initial time when equation (5.26) simplifies.
d
The same can be done for all the other ψijh (t0 ) components, to give

h
h[ψ̂1Lh (t0 ), ψ̂1Rh (t0 )† ]i = h[ˆˆA1Lh ,ˆˆA†1Rh ]i cos θ cos θ̄ + h[ˆˆA2Lh ,ˆˆA†1Rh ]† i sin θ cos θ̄eiω
i
ˆ ˆ † −iω̄ ˆ ˆ † iω−iω̄
+ h[ˆA1Lh ,ˆA2Rh ]i cos θ sin θ̄e + h[ˆA2Lh ,ˆA2Rh ]i sin θ sin θ̄e .
(5.59)

Analogously, the other three initial mixed correlators can be found, without having to
resort to calculating the complete statistical propagator for them. The other three are
given by

h
h[ψ̂1Lh (t0 ), ψ̂2Rh (t0 ) ]i = h[ˆˆA1Lh ,ˆˆA†2Rh ]i cos θ cos θ̄ + h[ˆˆA2Lh ,ˆˆA†2Rh ]† i sin θ cos θ̄eiω

i
− h[ˆˆA1Lh ,ˆˆA†1Rh ]i cos θ sin θ̄eiω̄ − h[ˆˆA2Lh ,ˆˆA†1Rh ]i sin θ sin θ̄eiω+iω̄
(5.60)
h
h[ψ̂2Lh (t0 ), ψ̂1Rh (t0 )† ]i = h[ˆˆA2Lh ,ˆˆA†1Rh ]i cos θ cos θ̄ − h[ˆˆA1Lh ,ˆˆA†1Rh ]† i sin θ cos θ̄e−iω
i
+ h[ˆˆA2Lh ,ˆˆA†2Rh ]i cos θ sin θ̄e−iω̄ − h[ˆˆA1Lh ,ˆˆA†2Rh ]i sin θ sin θ̄e−iω−iω̄
(5.61)
h
h[ψ̂2Lh (t0 ), ψ̂2Rh (t0 )† ]i = h[ˆˆA2Lh ,ˆˆA†2Rh ]i cos θ cos θ̄ + h[ˆˆA1Lh ,ˆˆA†2Rh ]† i sin θ cos θ̄e−iω
i
− h[ˆˆA2Lh ,ˆˆA†1Rh ]i cos θ sin θ̄e+iω̄ − h[ˆˆA1Lh ,ˆˆA†1Rh ]i sin θ sin θ̄e−iω+iω̄ .
(5.62)

In order to calculate for instance correlations of these mixed states, one would still like to
know the exact form of the entire statistical propagator. This can still be written out in
the same fashion as was done for F1L1L (t; t0 ) and F1R1R (t; t0 ), however for our purposes this
is sufficient, since we have the statistical propagator for the first species, which tells us its
complete evolution.

5.3 Time Translation Invariance


As a next step we would like to look for a solution that remains constant through time.
We will try to see if the initial conditions can be chosen in such a way that there are no
5.3. TIME TRANSLATION INVARIANCE 57

oscillations in the first species, and the system does not show any oscillatory behavior. The
off diagonal components of the mass matrix are still present, and therefore one could still
expect flavor mixing. However, we will try to see if there can be state where there is no
dependence in the time, such the state is invariant under time translations.

In order to do so, the statistical propagator for the right handed part of the first species
can be rewritten in terms of the average time (∆t) and the time difference (τ ):

∆t = t − t0 (5.63)
t + t0
τ= . (5.64)
2

Next we will look for the conditions such that the statistical propagator does not depend
on the time difference τ .

This will put certain bounds on the initial state of the system, the initial correlators. If
we can find a set up for these initial correlators when the statistical propagator does not
depend on τ , we have constructed a constant state which does not oscillate.

We also define two new frequencies which come out of these initial time and time difference
equations naturally, the sum of the two frequencies and the difference between them:

1
ω̄ = (ω1 + ω2 ) (5.65)
2
∆ω̄ = ω1 − ω2 . (5.66)

With these new quantities the resulting expressions simplify. This gives us in the case of
the right handed statistical propagator
58 CHAPTER 5. DIRAC FLAVOR MIXING

F1R1R (∆t; τ ) =
cos2 h 1
θ̄ h[Â1Rh , †1Rh ]i (cos 2ω1 τ + cos ω1 ∆t)
2 2
~
h|k| |~k|2 1 
+ sin ω1 ∆t + 2 (− cos 2ω1 τ + cos ω1 ∆t)
iω1 ω1 2
|(Md )11 |2 1
+ h[Â1Lh , †1Lh ]i (− cos 2ω1 τ + cos ω1 ∆t)
ω12 2
(Md∗ )11 1 h|~k|(Md∗ )11
+ h[Â1Lh , †1Rh ]i((sin 2τ ω1 − sin ∆tω1 ) − (− cos 2ω1 τ + cos ω1 ∆t) )
iω1 2 ω12
(M ∗ )11 1 h|~k|(Md∗ )11 i
− h[Â1Rh , †1Lh ]i((sin 2τ ω1 − sin ∆tω1 ) d − (cos 2ω1 τ + cos ω1 ∆t) )
iω1 2 ω12
sin2 h †
1
+ θ̄ h[Â2Rh , Â2Rh ]i (cos 2ω2 τ + cos ω2 ∆t)
2 2
~
h|k| |~k|2 1 
+ sin ω2 ∆t + 2 (− cos 2ω2 τ + cos ω2 ∆t)
iω2 ω2 2
|(Md )22 |2 1
+ h[Â2Lh , †2Lh ]i (− cos 2ω2 τ + cos ω2 ∆t)
ω22 2
(M ∗ )22 1 h|~k|(Md∗ )22
+ h[Â2Lh , †2Rh ]i((sin 2τ ω2 − sin ∆tω2 ) d − (− cos 2ω2 τ + cos ω2 ∆t) )
iω2 2 ω22
(M ∗ )22 1 h|~k|(Md∗ )22 i
− h[Â2Rh , †2Lh ]i((sin 2τ ω2 − sin ∆tω2 ) d − (cos 2ω2 τ + cos ω2 ∆t) )
iω2 2 ω22
sin 2θ̄ h
+ h[Â1Rh , †2Rh ]if1 (∆t, τ ) + [Â2Rh , †1Rh ]if2 (∆t, τ ) + [Â1Rh , †2Lh ]if3 (∆t, τ )
4
+ h[Â2Lh , †1Rh ]if4 (∆t, τ ) + h[Â1Lh , †2Rh ]if5 (∆t, τ ) + h[Â2Rh , †1Lh ]if6 (∆t, τ )
i
+ h[Â2Lh , †1Lh ]if7 (∆t, τ ) + h[Â1Lh , †2Lh ]if8 (∆t, τ ) . (5.67)

We defined the general functions fi (∆t, τ ) to keep the length of this expression in check.
For the following analysis their exact form is not important, only the fact that they all
depend on both τ and ∆t.

For the left handed statistical propagator we can do a likewise analysis, we also express it
in terms of τ and ∆t, where we use the same definitions ω̄ and ∆ω̄:
5.3. TIME TRANSLATION INVARIANCE 59

F1L1L (∆t; τ ) =
cos2 h 1
θ h[Â1Lh , †1Lh ]i (cos 2ω1 τ + cos ω1 ∆t)
2 2
~
h|k| |~k|2 1 
− sin ω1 ∆t + 2 (− cos 2ω1 τ + cos ω1 ∆t)
iω1 ω1 2
|(Md )11 |2 1
+ h[Â1Rh , †1Rh ]i (− cos 2ω1 τ + cos ω1 ∆t)
ω12 2
(Md∗ )11 1 h|~k|(Md∗ )11
+ h[Â1Rh , †1Lh ]i((sin 2τ ω1 − sin ∆tω1 ) + (− cos 2ω1 τ + cos ω1 ∆t) )
iω1 2 ω12
† (Md∗ )11 1 h|~k|(Md∗ )11 i
− h[Â1Lh , Â1Rh ]i((sin 2τ ω1 − sin ∆tω1 ) + (cos 2ω1 τ + cos ω1 ∆t) )
iω1 2 ω12
sin2 h 1
+ θ h[Â2Lh , †2Lh ]i (cos 2ω2 τ + cos ω2 ∆t)
2 2
~
h|k| |~k|2 1 
− sin ω2 ∆t + 2 (− cos 2ω2 τ + cos ω2 ∆t)
iω2 ω2 2
|(Md )22 |2 1
+ h[Â2Rh , †2Rh ]i (− cos 2ω2 τ + cos ω2 ∆t)
ω22 2
(M ∗ )22 1 h|~k|(Md∗ )22
+ h[Â2Rh , †2Lh ]i((sin 2τ ω2 − sin ∆tω2 ) d + (− cos 2ω2 τ + cos ω2 ∆t) )
iω2 2 ω22
(M ∗ )22 1 h|~k|(Md∗ )22 i
− h[Â2Lh , †2Rh ]i((sin 2τ ω2 − sin ∆tω2 ) d + (cos 2ω2 τ + cos ω2 ∆t) )
iω2 2 ω22
sin 2θ h
+ h[Â1Lh , †2Lh ]ig1 (∆t, τ ) + h[Â2Lh , †1Lh ]ig2 (∆t, τ ) + h[Â1Lh , †2Rh ]ig3 (∆t, τ )
4
+ h[Â2Rh , †1Lh ]ig4 (∆t, τ ) + h[Â1Rh , †2Lh ]ig5 (∆t, τ ) + h[Â2Lh , †1Rh ]ig6 (∆t, τ )
i
† †
+ h[Â2Rh , Â1Rh ]ig7 (∆t, τ ) + h[Â1Rh , Â2Rh ]ig8 (∆t, τ ) . (5.68)

Analogous to the functions fi (∆t, τ ) we have defined the functions gi (∆t, τ ) which also all
depend on τ and ∆t.

As we explained earlier, next we want the part depending on the average time (τ ) to vanish.
With this step we assume that the statistical propagator does not depend on the average
time and remains constant through the evolution.

First we will look at the right handed statistical propagator for the first flavor. For the
terms with cos2 θ̄ we find from the cos 2ω1 τ terms:
60 CHAPTER 5. DIRAC FLAVOR MIXING

!
|~k|2 |(Md )11 |2
h[Â1Rh , †1Rh ]i 1 − 2 − h[Â1Lh , †1Lh ]i
ω1 ω12
h|~k|(Md )∗11  † †

+ h[ Â1Lh , Â 1Rh ]i + h[ Â1Rh , Â 1Lh ]i =0 (5.69)
ω12

and the sin 2ω1 τ terms give from the same right handed statistical propagator of the first
flavor:

(Md )∗11 (Md )∗11


h[Â1Lh , †1Rh ]i = h[Â1Rh , †1Lh ]i , (5.70)
iω1 iω1

which leads immediately to

h[Â1Lh , †1Rh ]i = h[Â1Rh , †1Lh ]i. (5.71)

Therefore h[Â1Rh , †1Lh ]i must be hermitian. In the first condition we find that

 
(Md )11 h[Â1Lh , †1Lh ]i − h[Â1Rh , †1Rh ]i = 2h|~k|h[Â1Lh , †1Rh ]i. (5.72)

Note that we find from the left handed propagator exactly the same two conditions.
From the terms proportional to sin2 θ̄ we find with the same procedure the following
conditions:

h[Â2Lh , †2Rh ]i = h[Â2Rh , †2Lh ]i (5.73)


 
(Md )22 h[Â2Lh , Â2Lh ]i − h[Â2Rh , Â2Rh ]i = 2h|~k|h[Â2Lh , †2Rh ]i.
† †
(5.74)

All the other mixed correlators, such as h[Â1Lh , †2Rh ]i must vanish. When all four condi-
tions (5.71, 5.72, 5.73 and 5.74) hold, the statistical propagator will not oscillate in time.

5.3.1 Example State

An example of a choice of initial conditions is when we choose


5.3. TIME TRANSLATION INVARIANCE 61

h[Â1Lh , †1Rh ]i = 0 (5.75)


h[Â2Lh , †2Rh ]i = 0, (5.76)

resulting in

h[Â1Lh , †1Lh ]i = h[Â1Rh , †1Rh ]i = C1h (5.77)


h[Â2Lh , †2Lh ]i = h[Â2Rh , †2Rh ]i = C2h . (5.78)

This gives us the following correlators for the initial states:


h[ψ̂1Lh (t0 ), ψ̂1Lh (t0 )]i = cos2 θ C1h + sin2 θ C2h (5.79)
† 2 2
h[ψ̂1Rh (t0 ), ψ̂1Rh (t0 )]i = cos θ̄ C1h + sin θ̄ C2h (5.80)
† 2 2
h[ψ̂2Lh (t0 ), ψ̂2Lh (t0 )]i = sin θ C1h + cos θ C2h (5.81)
† 2 2
h[ψ̂2Rh (t0 ), ψ̂2Rh (t0 )]i = sin θ̄ C1h + cos θ̄ C2h (5.82)
† iω
h[ψ̂1Lh (t0 ), ψ̂2Lh (t0 )]i = sin θ cos θ (C2h − C1h )e (5.83)
† iω̄
h[ψ̂1Rh (t0 ), ψ̂2Rh (t0 )]i = sin θ̄ cos θ̄ (C2h − C1h )e . (5.84)

Therefore, if we can set up the system with the following initial expectation values, we have
a system of two fermionic species that will not oscillate, although they have a mass mixing
term that enables them to do so. If this mechanism indeed describes neutrino oscillations,
a system prepared in this exact way should not show any neutrino oscillations.

5.3.2 General Case

We can also construct a general solution, this leads to a more complicated formulation
then the previous examples state where we chose the constant in such a way to simplify
the final state. For the general case we have the following four conditions:

h[Â1Lh , †1Rh ]i = h[Â1Rh , †1Lh ]i (5.85)


 
(Md )11 h[Â1Lh , †1Lh ]i − h[Â1Rh , †1Rh ]i = 2h|~k|h[Â1Lh , †1Rh ]i (5.86)
h[Â2Lh , †2Rh ]i = h[Â2Rh , †2Lh ]i (5.87)
 
(Md )22 h[Â2Lh , †2Lh ]i − h[Â2Rh , †2Rh ]i = 2h|~k|h[Â2Lh , †2Rh ]i. (5.88)
62 CHAPTER 5. DIRAC FLAVOR MIXING

First we introduce the following notation:

D1Lh = h[Â1Lh , †1Lh ]i (5.89)


D1Rh = h[Â1Rh , †1Rh ]i (5.90)
D1M h = h[Â1Lh , †1Rh ]i (5.91)
D2Lh = h[Â2Lh , †2Lh ]i (5.92)
D2Rh = h[Â2Rh , †2Rh ]i (5.93)
D2M h = h[Â2Lh , †2Rh ]i (5.94)
and the conditions on the initial state now tell us that D1M h and D2M h must be real and
have to satisfy

(Md )11
D1M H = (D1Lh − D1Rh ) (5.95)
2h|~k|
(Md )22
D2M H = (D2Lh − D2Rh ) . (5.96)
2h|~k|
This immediately implies that we can only have the freedom to set D1Lh , D1Rh , D2Lh and
D2Rh .
For the initial physical state we find that this must satisfy


h[ψ̂1Lh (t0 ), ψ̂1Lh (t0 )]i = cos2 θ D1Lh + sin2 θ D2Lh (5.97)
† 2 2
h[ψ̂1Rh (t0 ), ψ̂1Rh (t0 )]i = cos θ̄ D1Rh + sin θ̄ D2Rh (5.98)
† 2 2
h[ψ̂2Lh (t0 ), ψ̂2Lh (t0 )]i = sin θ D1Lh + cos θ D2Lh (5.99)

h[ψ̂2Rh (t0 ), ψ̂2Rh (t0 )]i = sin2 θ̄ D1Rh + cos2 θ̄ D2Rh (5.100)

h[ψ̂1Lh (t0 ), ψ̂2Lh (t0 )]i = sin θ cos θ (D2Lh − D1Lh )eiω (5.101)

h[ψ̂1Rh (t0 ), ψ̂2Rh (t0 )]i = sin θ̄ cos θ̄ (D2Rh − D1Rh )eiω̄ (5.102)

h[ψ̂1Lh (t0 ), ψ̂1Rh (t0 )]i = cos θ cos θ̄D1M h + sin θ sin θ̄eiω−iω̄ D2M h (5.103)

h[ψ̂2Lh (t0 ), ψ̂2Rh (t0 )]i = cos θ cos θ̄D2M h + sin θ sin θ̄e−iω+iω̄ D1M h (5.104)

h[ψ̂1Lh (t0 ), ψ̂2Rh (t0 )]i = − cos θ sin θ̄eiω̄ D1M h + sin θ cos θ̄eiω D2M h (5.105)

h[ψ̂2Lh (t0 ), ψ̂1Rh (t0 )]i = − cos θ̄ sin θe−iω D1M h + sin θ̄ cos θe−iω̄ D2M h . (5.106)

This is the general case of the state that does not exhibit any oscillations. Note that there
are many more initial mixed states that were not present in the example state.
5.3. TIME TRANSLATION INVARIANCE 63

It should also be possible to do a likewise analysis in Wigner space with a kinetic descrip-
tion. This is treated in Appendix C. Note that this analysis is not completed, however in
principle it should lead to the same final results as were found here.
For a more general discussion of these results, we refer to the last chapter. First we will
also solve the Majorana case so that we can make a comparison between the two, which
will also be done in the last chapter.
64 CHAPTER 5. DIRAC FLAVOR MIXING
Chapter 6

Majorana Fermions

It was mentioned that the main difference between a Dirac fermion and a Majorana fermion
is the fact that a Majorana fermion is its own anti particle. In this chapter we will uncover
the dynamics of Majorana fermions, by further expanding the implications of this defining
property of a Majorana fermion. First it is noted that a Majorana fermion cannot have
any charge. Besides neutrinos, all other fermions have a Dirac nature as far as is known
since they have a charge. Therefore, the following considerations can probably only be
relevant for neutrinos, as they are the only current candidates for Majorana particles.
Should neutrinos indeed be Majorana fermions, they cannot have an additive quantum
number, and the total lepton number is no longer conserved. With the presence of neutrino
mixing, the only true global symmetry of the Standard Model is B − L. Here B stands for
the Baryon number, one third of the number of quarks minus the number of anti quarks
and L for the Lepton number, the number of leptons minus the number of anti leptons.
Majorana neutrinos would allow for processes such as neutrinoless double beta decay, which
violate B −L. This is one of the most fundamental problems in physics currently. Whether
neutrinos are Majorana or Dirac particles will also tell us about the origin of the neutrino
mass and rule out or strengthen several theories beyond the standard model.
Before discussing the Majorana solutions, it is instructive to take a close look at the different
spinors that appear in physics when dealing with fermions. Dirac particles are described
by a four component complex spinor. Massless fermions are described by a two component
complex spinor [9], which are named Weyl spinors. From the discussed experiments it was
clear that neutrinos do have a mass and are therefore not described by Weyl spinors. The
third possibility is that a neutrino is a Majorana particle, which means they are described
by a Majorana spinor.
A Weyl spinor has only two components because it is massless and travels at the speed of
light. It is impossible for an observer to go to a frame in which the Weyl particle has its
helicity flipped, which implies that one needs only half the components that are needed
to describe the general Dirac particle. A Majorana spinor also has only two components,

65
66 CHAPTER 6. MAJORANA FERMIONS

because the particle is its own anti particle. A particle and an anti particle are defined
through a certain quantum number, however in this case the quantum number is no longer
conserved, so it is impossible to distinguish between these two. For a Majorana particle,
the particle is its own antiparticle.

6.1 Neutrinoless Double Beta Decay


As was discussed earlier, neutrino mixing experiments with pure initial states do not
give an information about the nature of the neutrino, Dirac or Majorana, as the mixing
probabilities are equal for both Majorana and Dirac neutrinos. Different experiments
are needed to reveal this. The most commonly suggested experiment is to measure the
neutrinoless double beta decay (0νββ). This decay reaction is only possible if the neutrino
is a Majorana particle, since it violates lepton number conservation. See figure (6.1).
A different example of a process that requires virtual Majorana neutrinos and violate the
total lepton number [13]:

K + → π − + µ+ + µ+ K + → π − + e + + e + K + → π − + µ+ + e + . (6.1)

This represents the decay of a kaon. The following bound was obtained on the Majorana
mass: |Mµµ | ≤ 4 · 104 MeV, the sensitivity of this processes is much too small to be
able to do any relevant predictions. The same goes for almost all other processes that
involve virtual Majorana neutrinos. The notable exception is neutrinoless double beta
decay (0νββ):

(A, Z) → (A, Z + 2) + e− + e− . (6.2)

This process should only be possible for a limited number of even-even nuclei (nuclei with
an even number of protons and neutrons) when the mass of the (A, Z) nucleus is larger
than the mass of the (A, Z + 2) nucleus.
In experiments the effective Majorana mass mββ is the parameter that appears in the total
decay rate. This Majorana mass is given by:

X
mββ = Vli2 mi . (6.3)
i

From other experiments we already have estimations of the mixing angles and the mass
squared differences. This leaves several possible Majorana masses. Usually the following
three spectra for the neutrino masses are considered, and we will follow this convention:
6.1. NEUTRINOLESS DOUBLE BETA DECAY 67

• Hierarchy of the neutrino masses

m1  m2  m3 (6.4)

This is the same spectrum as the other leptons obey. The Majorana mass is in this
case bounded by
 q q 
|mββ | ≤ sin ϑ12 ∆m12 + sin ϑ13 ∆m23 = 5.3 · 10−3 eV.
2 2 2 2
(6.5)

Here we used the aforementioned data discussed in chapter 3. This is only a bound,
because there is still a Majorana phase difference appearing in the final mass. This
bound is however much smaller than the maximal sensitivity of all current and pro-
posed experiments. Should this spectrum be the spectrum of the neutrino Majorana
masses, then we will not be able to detect them in the near future.
• Inverted hierarchy of the neutrino masses

m3  m1 < m2 (6.6)

For this hierarchy we can again calculate the bounds on the Majorana mass:

q q
cos 2ϑ12 |∆m213 | ≤ |mββ | ≤ |∆m213 | (6.7)
1.8 · 10−2 eV ≤ |mββ | ≤ 4.9 · 10−2 eV. (6.8)

Again, this is only a bound because of the Majorana phase difference. In this case the

Figure 6.1: In the left Feynman graph we picture conventional double beta decay with the
production of two neutrinos, in the right graph neutrinoless double beta decay
68 CHAPTER 6. MAJORANA FERMIONS

lower bound is not zero, which is important for experiments. Prospective experiments
are expected to reach these energies, and should therefore be able to test if the
inverted hierarchy of Majorana neutrinos is indeed a correct model.

• Quasi-degenerate neutrino mass spectrum

m1 ' m2 ' m3 (6.9)

This case is more complicated than the previous two. The mass bounds are given
by:

cos(2ϑ12 )mmin ≤ |mββ | ≤ mmin (6.10)


|mββ | ≤ mmin ≤ 2.8|mββ |. (6.11)

This bound still depends on mmin , the lightest neutrino mass. Depending on how
light it actually is, the Majorana mass would be detectable in an experiment.

Experiments that try to measure the 0νββ rate have an additional constraint. Only
the product of the Majorana mass and the nuclear matrix elements is measurable in an
experiment, and this nuclear matrix element is difficult to determine, which results in a
large uncertainty.

Several experiments have already been performed on 0νββ. The IGEX and the Heidelberg-
Moscow experiment have measured the decay of 76 Ge, giving the following range:

|mββ | ≤ 0.3 ∼ 1 eV. (6.12)

Some researchers from the Heidelberg-Moscow group have claimed to have found a signifi-
cant result for a positive signal of a Majorana mass with a mass of |mββ | = 0.2−0.5 eV [17],
however this claim is being challenged by many people in the particle physics community
and is definitely not generally accepted [18].

A number of new experiments are planned in the hope to find a signal for a Majorana mass.
The major candidates to shed more light on this puzzle are the GERDA , the CUORE, the
EXO and the NEMO experiments. These groups hope to reach a sensitivity of up to a few
hundreds of electron volts. This should definitely be enough to test the inverse hierarchy
spectrum of Majorana masses, and show if the claims of [17] are indeed correct.
6.2. THEORY 69

6.2 Theory

We will examine the Majorana mass term and see to which equations of motion it abides.
We will solve these and show to quantize the theory.
The Majorana condition is in literature often written as [27] [28]

ΨM = ΨcM

Here the ΨM is a four spinor that represents a Majorana particle. The small c superscript
denotes the CP conjugate of this spinor. Writing out this CP-conjugate we can summarize
the Majorana condition as:

ψM = C(ψ̄M )T . (6.13)

The notation can be confusing, the C in this equation is the charge conjugation matrix,
note the difference between c and C. The explicit form is given in Appendix A. Note
that this form depends on which convention is being used. Next we can write out this
requirement on the four spinor to find:

 
φ
ψ= . (6.14)
iσ 2 φ∗

In this notation φ is a two spinor, such that ψ, as discussed earlier, is a four spinor,
analogous to the four spinor for the general Dirac equation. We will denote all four
component spinors as ψ and all two component spinors with φ. Notice that the system is
fully described by the two component spinor φ, however for some normalization purposes
it will still be worthwhile to return to the four component description. The equations of
motion are then given in terms of φ:

(~σ · ∇ − ∂t )φ − mσ2 φ∗ = 0 (6.15)


(~σ · ∇ + ∂t )σ2 φ∗ − mφ = 0. (6.16)

The second equation can be obtained by taking the complex conjugate of the first one and
multiplying by appropriate constants.
Next we assume the following ansatz for the two component spinor φ, from which the
complex conjugate φ∗ automatically follows. It amounts to expanding the spinor in positive
and negative Fourier modes and taking the complex conjugates from these modes:
70 CHAPTER 6. MAJORANA FERMIONS

d4 k
Z
0 ~ 0 ~
φ= 4
Ae−ik t+ik·~x + Beik t−ik·~x (6.17)
(2π)
Z 4
d k ∗ ik0 t−i~k·~x 0 ~
φ∗ = 4
Ae + B ∗ e−ik t+ik·~x . (6.18)
(2π)

A and B have two components, they are given by


 
A1
A=
A2

and  
B1
B=
B2
From this we can infer

0 ~ 0 ~
(−i~k · ~σ B − ik 0 B − mσ 2 A∗ )eik t−ik·~x + (i~k · ~σ A + ik 0 A − mσ 2 B ∗ )e−ik t+ik·~x = 0. (6.19)

Here we can read off two equations that A and B have to satisfy:

−i~k · ~σ B − ik 0 B − mσ 2 A∗ =0 (6.20)
i~k · ~σ A + ik 0 A − mσ 2 B ∗ =0. (6.21)

These two are equivalent and solving the classical system of motion amounts to finding the
correct expressions for A and B.

6.2.1 Helicity

It is possible to solve the classical system with helicity eigenstates. First it is clear that
we can set

k̂ · ~σ Ah = hAh (6.22)
k̂ · ~σ B h = −hB h . (6.23)

Here we have two choices, we can either set h = 1 or h = −1, which represent two different
6.2. THEORY 71

solutions. For the first choice, A is proportional to the positive helicity eigenstates and B
to the negative eigenstate. The second choice, h = −1, amounts to the swapped choice.
These helicity eigenstates (ξ+ and ξ− ) are explicitly given by:

 
1 k1 − ik2
ξ+ = q (6.24)
|~k| − k3
2|~k|(|~k| − k3 )
 
1 −k1 + ik2
ξ− = q . (6.25)
|~k| + k3
2|~k|(|~k| + k3 )

The way we defined them here, they are orthogonal and properly normalized to unity

ξi† ξj = δij

h
Now Ah+ and B+ are given by

s
k 0 − |~k|
1
Ah+ = ξ+
2 k0
s
0 + |~
r
h 1 k k| k1 + ik2
B+ = −ξ− 0
. (6.26)
2 k k1 − ik2

This solution can be verified by plugging it in the equations motion directly. The last
h
square root in B+ is a phase, we chose to put it in B, however we could have also included
it in Ah+ . A second solution is given by the other choice for h, leading to Ah− and B−
h
. We
use the same helicity eigenstates to arrive at

s
1 k 0 + |~k|
Ah− = ξ−
2 k0
s
k 0 − |~k|
r
h 1 k1 + ik2
B− = ξ+ . (6.27)
2 k0 k1 − ik2

h
Again, the phase is included in B− .
Writing out φh we find
72 CHAPTER 6. MAJORANA FERMIONS

 s s 
~ ~
4
r
k0
− |k| −ik·x 0
1 k + |k| k1 + ik2 ik·x 
Z
dk  1
φh1 = 4
ξ+ 0
e − ξ− e (6.28)
(2π) 2 k 2 k0 k1 − ik2
 s s 
~ ~
4
r
0
d k  1 k + |k| −ik·x 0
1 k − |k| k1 + ik2 ik·x 
Z
φh2 = 4
ξ− 0
e + ξ+ e . (6.29)
(2π) 2 k 2 k0 k1 − ik2

Both these solutions are normalized to

1
(φh1 )† φh1 = (φh2 )† φh2 = . (6.30)
2

This normalization of the two spinor also fixes the normalization for the four spinor:

(ψ1h )† ψ1h = (ψ2h )† ψ2h = 1. (6.31)

A general solution can be obtained by adding the two independent solutions:

d4 k
Z
h h −ik·x h ∗ ik·x h −ik·x h ∗ ik·x

φ = µ e − ν λ e + ν e + µ λ e . (6.32)
(2π)4

Here we defined the phase r


k1 − ik2
λ=
k1 + ik2
and we also defined

s
k 0 − |~k|
ξ+ = µh
2k 0
s
k 0 + |~k|
ξ− = ν h. (6.33)
2k 0

We prefer to look on for a different solution, because it is singular in the rest frame. Since
these are helicity eigenstates, they are not well defined for ~k → 0. We can however derive
a different solution that still holds in this limit.
6.2. THEORY 73

6.2.2 Rest Frame

In this section we will derive a different solution that does behave well in the rest frame.
We go to the rest frame of the system, the spacial momenta give zero, such that we are
left with the following equation:

0 0 0 0
− ik 0 Beik t + ik 0 Ae−ik t − mσ 2 B ∗ e−ik t − mσ 2 A∗ eik t = 0. (6.34)

0
In equation 6.34 the terms in front of eik t give us:

− ik 0 B = mσ 2 A∗ (6.35)

0
and the terms with e−ik t give

ik 0 A = mσ 2 B ∗ . (6.36)

Combining these two gives us the mass shell condition,

(k 0 )2 = m2

This gives us in the rest frame the following two identities where we define a frequency ω:

k 0 = ±ω (6.37)
|ω| = m. (6.38)

Now we can eliminate the constant B with the components B1 and B2 , by constructing
the relationship between them and the A constants:

m
B1 = A∗2 (6.39)
k0
m
B2 = −A∗1 0 . (6.40)
k

The solution for φ in the rest frame is now given by

dk 0
Z    ∗  
A1 −ik0 t A2 ik0 t
φ= e + e . (6.41)
2π A2 −A∗1
74 CHAPTER 6. MAJORANA FERMIONS

The vacuum solution satisfies the following properties, the scalar density is zero:
ψ̄ψ = 0
Switching from the two component quantity φ to the four component ψ and vice versa is
done through (6.14). To normalize the given expression, we can require that the density
per unit volume in the rest frame is equal to unity
ψ†ψ = 1
which gives us

1
|A1 |2 + |A2 |2 = . (6.42)
4

Next we are allowed to construct in the vacuum the following two spinors, where we require
for the first A1 = − 12 , |A2 | = 0 and for the second |A1 | = 0 combined with A2 = 21 , which
gives

   
0 1 −1 −ik0 t 1 0 ik0 t
φ− (k ) = e + e (6.43)
2 0 2 1
   
0 1 0 −ik0 t 1 1 ik0 t
φ+ (k ) = e + e . (6.44)
2 1 2 0

Both are solutions for the mode functions that satisfy the equations of motion considering:

dk 0
Z
φ= φ± (k 0 ). (6.45)

The labels + and − are explained by determining the spin of the solution. Remember that
we also had found solutions that were expressed in terms of helicity eigenspinors. We can
express the vectors    
1 0
and
0 1
as linear combinations of the helicity spinors ξ± :

q q
  ~ ~
|k| − k3 (|k| + k3 ) |~k| + k3 (|~k| − k3 )
1
= ξ+ + q ξ− (6.46)
0
q
~
2|k|(k1 − ik2 ) ~
2|k|(k1 − ik2 )
  s~ s
0 |k| − k3 |~k| + k3
= ξ+ − ξ− . (6.47)
1 2|~k| 2|~k|
6.2. THEORY 75

This shows that the new rest frame solution is a linear combination of the previously found
solution which was build out of helicity eigenstates ξ± .
In the four component formalism this is found to be after applying (6.14):

   
−1 0
1  0  −ik0 t 1 1
  0
ψ− (k 0 ) =  e +   eik t (6.48)
2  1  2 0 
0 1
   
0 1
1  1  −ik t
 0 1  0
 eik0 t .
ψ+ (k 0 ) =  e +  (6.49)
2  0  2 1

−1 0

The spin in a general direction ~n can be found by solving for the eigenvector equation
nµ Σµ ψi = ψi where Σ0 = I and
 
~ = 1 ~σ 0
Σ .
2 0 ~σ

Using that ψ− is normalized to unity, we see ψ− ψ− ψ− = ψ− , therefore


n(−) µ
µ Σ = φ− φ− (6.50)

and explicit calculation yields

n(−) 0 0
µ = (1, − cos 2k t, − sin 2k t, 0). (6.51)

This shows that the spin of the positive spinor is oscillating in the xy plane. It is not
constant, but it evolves through time with a period of π/k 0 . We defined this as the positive
spinor, since the rotation of the spin is positive in a right handed, standard Cartesian
coordinate system.
For the second positive spinor we find

n(+) 0 0
µ = (1, cos 2k t, − sin 2k t, 0). (6.52)

This spin is also oscillating in the xy plane, however it is oscillating in the opposite direction
76 CHAPTER 6. MAJORANA FERMIONS

(−)
Figure 6.2: The red line represents the spin in the xy plane of nµ with the time t on the
(+)
vertical axis and the green line the spin of nµ

with the same period as the negative spinor, hence this is the positive spinor. In figure
(6.2) both the spins of the states are pictured.
Note that we have a different interpretation for the spin of this solution then in [35], which
we believe is physically incorrect.
These spinors are normalized in such a way that they are orthonormal. This should allow
us to define the following projection operators:


P− = v − v − (6.53)

P+ = v+ v+ . (6.54)

They indeed satisfy Pi2 = Pi , P12 = P1 and P22 = P2 , which is required of a projection
operator. This completes the discussion of the solutions in the rest frame.

6.2.3 Boost

These results can be boosted to give rise to a solution that also holds in the general frame
of reference, not only in the rest frame. To do this we apply a boost to the rest frame
6.2. THEORY 77

solution.
A general Lorentz boost is defined as:

~
Λ(~k) = e−i η k̂·K . (6.55)

~ is the generator of the Lorentz boost, in our case given


Here η gives the rapidity, and K
by

− 2i ~σ
 
~ = 0
K i . (6.56)
0 2

This allows us to write the boost as

  
η − k̂ · ~
σ 0
Λ(~k) = exp (6.57)
2 0 k̂ · ~σ
   
1 0 −k̂ · ~σ 0
= cosh(η/2) + sinh(η/2) . (6.58)
0 1 0 k̂ · ~σ

The rapidity has the following properties:

k0 |~k|
cosh(η) = sinh(η) = . (6.59)
m m

Which allow us to write with the help of the half angle formulas for the hyperbolic functions

k0 + m |~k|
cosh(η/2) = p sinh(η/2) = p (6.60)
2m(k 0 + m) 2m(k 0 + m)

and we find for the boost

!
1 k 0
+ m − ~k · σ 0
Λ(~k) = p (6.61)
2m(k 0 + m) 0 k 0 + m + ~k · σ
 0 
k + m − k3 k1 − ik2 0 0
1  k1 + ik2 k 0 + m + k3 0 0 
=p  0
.
2m(k 0 + m)  0 0 k + m + k3 k1 − ik2 
0 0 k1 + ik2 k 0 + m − k3
(6.62)
78 CHAPTER 6. MAJORANA FERMIONS

This boost is normalized such that

Λ(~k)Λ(−~k) = 1. (6.63)

Which makes physically sense; boosting to a frame with a certain momentum ~k and then
boosting back with −~k should return the original quantity.

6.2.4 General Frame

The solution in a general frame is now given by applying this boost to the rest frame
solution:

ψ± (~k) = Λ(~k)ψ± (k 0 ). (6.64)

This gives the following two solutions:

   0  
−k + ik k + m − k
φ+ (~k) = 1 2
e −ik·x
+ 3
eik·x
∗N (6.65)
k 0 + m + k3 −k1 − ik2
 0    
~ −k − m + k3 −ik·x −k1 + ik2 ik·x
φ− (k) = e + e ∗ N. (6.66)
k1 + ik2 k 0 + m + k3
p
Here the normalization factor is given by N = 1/(2 2m(k 0 + m)). These two solutions
follow from the ansatz (6.17) which leads to

 
−k1 + ik2
A+ =N = µb
k 0 + m + k3
 0 
k + m − k3
B+ =N = νb
−k1 − ik2
 0 
−k − m + k3
A− =N = −ν b
k1 + ik2
 
−k1 + ik2
B− =N = µb . (6.67)
k 0 + m + k3

They can be verified directly by plugging them into the equations of motion, which shows
that they are indeed correct solutions to this problem. To do this, we also need the
generalized mass shell condition:
6.2. THEORY 79

m2 = (k 0 )2 − ~k 2 . (6.68)

The mass shell condition is also obtained when one tries to solve the positive and negative
frequency part of equation (6.19). In order for the determinant of the system to vanish,
mass shell is again found.
This allows us to express ω, the frequency, as
q
k = ± ~k 2 + m2 = ±ω.
0
(6.69)

We will need this frequency later in the analysis in order to define vacuum and excited
states. This shows that we can choose two different shells for k 0 , a positive shell: ω and a
negative shell: k 0 = −ω.
The four component equivalent of the solution is given by

   0  
−k1 + ik2 k + m − k3
 k 0 + m + k3  −ik·x  −k1 − ik2  ik·x 
ψ+ (~k) = N 
 −k1 + ik2  e
  +
 k 0 + m + k3  e 
  (6.70)
−k 0 − m + k3 k1 + ik2
 0    
−k − m + k3 −k1 + ik2
 k1 + ik2  −ik·x k 0 + m + k3  ik·x 
ψ− (~k) = N 
 k 0 + m + k3  e
  +
 k1 − ik2  e  .
  (6.71)
k1 + ik2 k 0 + m − k3

Conserved Current

In the rest frame we were able to normalize ψ † ψ, however this is not possible for the
boosted solution, ψ † ψ is space and time dependent. We have chosen to normalize it by
first normalizing the rest frame and then using a properly normalized boost. For the
solution in terms of helicity eigenstates this was not the case, they could be normalized
by setting (ψ h )† ψ h = 1. The choice to find a solution that is also valid in the rest frame,
comes at the cost of introducing extra difficulties with the conserved current. From the
conserved Noether current we find the continuity equation:

∂µ (ψ̄γ µ ψ) = 0 (6.72)
∂t j 0 = −∂i j i . (6.73)
80 CHAPTER 6. MAJORANA FERMIONS

In the rest frame this is trivially satisfied, there is no time dependence in ψ † ψ = j 0 , as is


the case with the helicity eigenstates solution, and there is also no space dependence, so
everything vanishes upon differentiating. For the boosted solution this is different however.
We will examine the ∂µ (ψ̄+ γ µ ψ+ ) case in detail.

First we notice from the defining property (6.14), which simplifies the necessary algebra,
immediately that


ψ+ ψ+ = φ†+ φ+ + φT+ φ∗+ (6.74)

and calculation yields after applying the solution we found for φ+ :


ψ+ = 8|N |2 (k 0 + m) k 0 + (k1 + ik2 )e2ik·x − (k1 − ik2 )e−2ik·x .

ψ+ (6.75)

The time derivative of this quantity, which is the probability density, is given by


ψ+ ) = 8|N |2 (k 0 + m)ik 0 (k1 + ik2 )e2ik·x + (k1 − ik2 )e−2ik·x .

∂t (ψ+ (6.76)

The other three spacial derivatives of the respective currents ∂i (ψ̄+ γ i ψ+ ) should give the
negative of this result:

∂x j 1 = 4|N |2 ik1 (−(k1 + ik2 )2 − (k 0 + m)2 + k32 )e2ik·x


+ ((k1 + ik2 )2 + (k 0 + m)2 − k32 )e−2ik·x

(6.77)
∂y j = 4|N | k2 (−(k1 + ik2 ) + (k + m) − k32 )e2ik·x
2 2 2 0 2

+ (−(k1 − ik2 )2 + (k 0 + m)2 − k32 )e−2ik·x



(6.78)
∂z j 3 = 8|N |2 ik3 −k32 (k1 + ik2 )e2ik·x + k32 (k1 − ik2 )e−2ik·x .

(6.79)

Adding these three derivatives of the current indeed gives minus the time derivative of the
probability density, thereby satisfying with the continuity equation. For the ψ− we have a
similar case. This convinces us that we have indeed found correct solutions to the classical
Majorana system.
6.2. THEORY 81

6.2.5 Excited States

From these two classical solutions we can build a general solution in order to quantize the
system. First we construct the vacuum state, which is found by taking the positive solution
for k 0 = ω:

X
φvac = α~k,± φ± |k0 =ω (6.80)
±
    
h −k1 + ik2 −iωt+i~k·~
x ω + m − k3 iωt−i~k·~x
=N α~k,+ e + e
ω + m + k3 −k1 − ik2
    i
−ω − m + k3 −iωt+i~k·~x −k1 + ik2 iωt−i~k·~
x
+ α~k,− e + e . (6.81)
k1 + ik2 ω + m + k3

The vacuum state is found by projecting k 0 on the positive shell by setting ω = k 0 (the left
moving wave), excited states are found by projecting −ω = k 0 (the right moving wave).
The next step is to also include the excited states, which are given by

X
φexc = β~k,± φ± |k0 =−ω (6.82)
±
    
h −k1 + ik2 iωt+i~k·~
x −ω + m − k3 −iωt−i~k·~x
=N β~k,+ e + e
−ω + m + k3 −k1 − ik2
    i
ω − m + k3 iωt+i~k·~x −k1 + ik2 −iωt−i~k·~
x
+ β~k,− e + e . (6.83)
k1 + ik2 −ω + m + k3

We can clean up our notation a bit by using the µ and ν states defined previously. We
will add a subscript + for the positive frequency states and a subscript − for the negative
states:

 
−k1 + ik2
N = µ+ (6.84)
ω + m + k3
 
ω + m − k3
N = ν+ (6.85)
−k1 − ik2
 
−k1 + ik2
N = µ− (6.86)
−ω + m + k3
 
−ω + m − k3
N = ν− . (6.87)
−k1 − ik2
82 CHAPTER 6. MAJORANA FERMIONS

These simplify the notation for these states. The full expression that we need to capture
the characteristics of the system is the sum of both the vacuum states and the excited
states, given the constraint for the normalization of this solution

X
φ= α~k,± φ± |k0 =ω + β~k,± φ± |k0 =−ω (6.88)
±
h i h i
~ ~ ~ ~
=α~k,+ µ+ e−iωt+ik·~x + ν+ eiωt−ik·~x + α~k,− −ν+ e−iωt+ik·~x + µ+ eiωt−ik·~x
h i h i
iωt+i~k·~
x −iωt−i~k·~
x iωt+i~k·~
x −iωt−i~k·~
x
+ β~k,+ µ− e + ν− e + β~k,− −ν− e + µ− e . (6.89)

To properly normalize this general solution, we have to set

|α~k,± |2 + |β~k,± |2 = 1. (6.90)

6.2.6 Quantization

In order to quantize the system, we can first look at equations (6.20) and (6.21). These tell
us that the operator preceding A and B ∗ should be the same, as is the operator preceding
A∗ and B. This enables us to associate creation (â†k,± ) and annihilation operators (âk,± )
with the quantities φ and φ∗ . This is the vacuum state, as derived in the previous section:

d3~k X
Z
~ ~
φ̂(t, ~x) = 3
A± e−i(ωt−k·~x) â~k,± + B± ei(ωt−k·~x) â~†k,± . (6.91)
(2π) ±

For the complex conjugate we find then as a result:

d3~k X ∗ i(ωt−~k·~x) †
Z
∗ ∗ −i(ωt−~k·~
x)
φ̂ (t, ~x) = 3
A± e â~k,± + B± e â~k,± . (6.92)
(2π) ±

These two decompositions indeed satisfy the restrictions given by the equations of motion.
We can use either of the two solutions we have constructed before, the helicity eigenstates
(6.26,6.27) and the boosted rest frame solutions (6.67). In principle, both solutions are
valid, we have showed that they are linear combination of each other. However, one should
keep in mind that the helicity solution does not have a rest frame and the boosted rest
frame solution has a time dependent probability density, which gives extra complexity
when normalizing the solution.
The creation and annihilation operators satisfy the canonical anticommutation relation:
6.2. THEORY 83

{â~k,i , â~†l,j } = δ(~k − ~l)δij . (6.93)

Here the indices i and j are spinor indices. The two other anticommutators give zero:

{â~k,i , â~l,j } = 0 (6.94)


{â~†k,i , â~†l,j } = 0. (6.95)

In the four component formalism the creation and annihilation operators take a matrix
form
!
â~k,± 0
Â~k,± = . (6.96)
0 â~†k,±

This is the only way to ensure that we obtain the right relation between the factors in
front of A and B ∗ , as mentioned earlier. For the operator ψ̂ we then find

d3~k X
Z     
A± −i(ωt−~k·~
x) B± † i(ωt−~k·~
x)
ψ̂(t, ~x) = ∗ Âk,± e + Â k,± e . (6.97)
(2π)3 ± iσ 2 B± iσ 2 A∗±

The anticommutation relation remains the same for the four component creation annihi-
lation operators

{Â~k,i , †~l,j } = δ(~k − ~l)δi,j . (6.98)

We still have the ability to switch between the two formalisms, and it is clear that the
operators have a simpler form in the two component formalism. We know from normal
Dirac quantization that the anticommutation relation for the creation/annihilation opera-
tors implies the anticommutation relation for the fields

{ψ̂α (t, ~x), ψ̂β† (t, ~y )} = δ 3 (~x − ~y )δαβ . (6.99)

This essential property has to be checked for the solutions we obtained, to ensure that we
have applied the correct quantization procedure. First, we note that:

{ψ̂α (t, ~x), ψ̂β† (t, ~y )} = 2Re{φ̂α (t, ~x), φ̂†β (t, ~y )}, (6.100)
84 CHAPTER 6. MAJORANA FERMIONS

Which allows us to work with the two component spinors φ̂. In order to evaluate {φ̂i (~x), φ̂†j (~y )}
we first write φ̂ in terms of the previously defined µ and ν. We are dealing with vacuum
states here, so we will suppress the subscript indicating the shell, all quantities are on the
positive shell. Instead of writing out the components, this anticommutator is a two by two
matrix:

d3~k h
Z
~ ~
φ̂(t, ~x) = 3
α~k+ (â~k,+ µ~k e−i(ωt−k·~x) + â~†k,+ ν~k ei(ωt−k·~x)
(2π)
i
~ ~
+ α~k− (â~k,− (−ν~k )e−i(ωt−k·~x) + â~†k,− µ~k ei(ωt−k·~x) ) . (6.101)

Using this expansion to evaluate the anticommutator, we first realize that we can drop all
terms proportional to anticommutators that give zero. This also includes terms such as
{â~†k,+ , â~l,− } with opposite spin indices. This leaves us with

d3~k d3~l h
Z
~ ~ 0

{φ̂(t, ~x), φ̂ (t, ~y )} = 3 3
|α~k+ |2 µ~k µ~†l ei(k·~x−l·~y)−it(ω−ω ) {â~k,+ , â~†l,+ }
(2π) (2π)
~ ~ 0
+ |α~k+ |2 ν~k ν~l† e−i(k·~x−l·~y)+it(ω−ω ) {â~†k,+ , â~l,+ }
~ ~ 0
+ |α~k− |2 ν~k ν~l† ei(k·~x−l·~y)−it(ω−ω ) {â~k,− , â~†l,− }
i
~ ~ 0
+ |α~k− |2 µ~k µ~†l e−i(k·~x−l·~y)+it(ω−ω ) {â~†k,− , â~l,− } . (6.102)

Using the anticommutation relation for the creation/annihilation operators, we integrate


over ~l, which gives us ~k = ~l through the delta function and ω = ω 0 :

d3~k h
Z
~ ~

{φ̂(t, ~x), φ̂ (t, ~y )} = 3
|α~k+ |2 (µ~k µ~†k eik·(~x−~y) + ν~k ν~k† e−ik·(~x−~y) ) (6.103)
(2π)
i
~ ~
+ |α~k− |2 (µ~k µ~†k e−ik·(~x−~y) + ν~k ν~k† eik·(~x−~y) ) . (6.104)

In order to solve this, we set |α~k+ |2 = |α~k− |2 to find

d3~k
Z h i
† 2 i~k·(~
x−~
y) † † † †
{φ̂(t, ~x), φ̂ (t, ~y )} = |α~ | e µ~k µ~k + ν−~k ν−~k + µ−~k µ−~k + ν~k ν~k . (6.105)
(2π)3 k
6.2. THEORY 85

At this point we have to choose the helicity solution for µ and ν (6.33) or the boosted
solution (6.67). First we will treat the helicity solution. In this case we find that

d3~k |α~k |2 i~k·(~x−~y) h


Z


{φ̂(t, ~x), φ̂ (t, ~y )} = e (ω − |~k|)ξ+ ξ+
(2π)3 4ω
i
† † †
+ (ω + |~k|)ξ+ ξ+ + (ω − |~k|)ξ− ξ− + (ω + |~k|)ξ− ξ− . (6.106)

The phase λ drops out of this equation, since |λ| = 1. To proceed, we use the relation

† †
ξ− ξ− + ξ+ ξ+ = 12×2 (6.107)

to find

d3~k |α~k |2 i~k·(~x−~y)


Z

{φ̂(t, ~x), φ̂ (t, ~y )} = e 12×2 (6.108)
(2π)3 2
|α|2
= δ(~x − ~y )12×2 , (6.109)
2

thus proving the relation (6.99), for the constraint |α|2 = 1.


If we use the other solution, we start again from (6.105), and now use (6.67). We need the
following two outer products:

 
k12 + k22 (−k1 + ik2 )(ω + m + k3 )
µ~k µ~†k=N 2
(6.110)
(−k1 − ik2 )(ω + m + k3 ) (ω + m + k3 )2
 
† 2 (−ω − m + k3 )2 (k1 − ik2 )(−ω − m + k3 )
ν~k ν~k = N . (6.111)
(k1 + ik2 )(−ω − m + k3 ) k12 + k22

After setting |α~k+ |2 = |α~k− |2 we see that this also leads to

d3~k |α~k |2 i~k·(~x−~y)


Z

{φ̂(t, ~x), φ̂ (t, ~y )} = e 12×2 (6.112)
(2π)3 2
|α~ |2
= k δ(~x − ~y )12×2 , (6.113)
2
86 CHAPTER 6. MAJORANA FERMIONS

which gives the correct solution and the same constraint on the constant |α~k |2 = 1 as in
the case with the helicity eigenstates. This is an overall phase factor which can also be
included in µ and ν, as can be seen from the Majorana equations of motion.
In total, this gives us the following correct expansion for φ̂ in the vacuum that satisfies all
the necessary conditions:

d3~k
Z h
−i(ωt−~k·~ ~
φ̂(t, ~x)vac = 3
α k â~k,+ µ~k e
~
x)
+ â~†k,+ ν~k λ∗ ei(ωt−k·~x)
(2π)
i
~ ~
− â~k,− ν~k e−i(ωt−k·~x) + â~†k,− µ~k λ∗ ei(ωt−k·~x) . (6.114)

This completes our discussion of quantizing the Majorana field. We have obtained two
different solutions for the classical equations of motion for a Majorana field. Both these
solutions were quantized.
Chapter 7

Majorana Flavor Mixing

We can modify our model for the mass mixing to include a Majorana mass term instead of
a Dirac mass term. Again, we will try to solve the system with the help of the correlators.
Only now our starting point will be a Majorana fermion instead of a Dirac fermion.
Previously we have discussed the classical equations of motion for a Majorana fermion
and how we can appropriately quantize them.

Similarly to the Dirac case we will first look at the full equations of motion. These can be
simplified by first diagonalizing the mass matrix by a rotation in flavor space. Secondly
we can use the fact that the general state can be expanded in helicity eigenstates for the
spinorial components. This gives us a simpler representation of the equations of motion
which we can solve. This solution in the diagonal flavor frame is in terms of some constant
operators. The constant operators in terms of which the solution is expressed can be set
by us. We can express the initial conditions in such a way that they are a function of
the newly introduced operators. These initial conditions can be inverted such that we can
derive the constant operators for any given initial physical state.

After finding this general solution, we will try to see if we can set the initial conditions
in such a way that there are no oscillations in the resulting statistical propagator. Again,
rewriting the time variables in terms of average time and time difference will proof to be
vital. This gives us a set of conditions. These conditions allow us to construct an initial
state in which the oscillations are eliminated. With the help of the expressions for the
initial state we can go a step further and specify an initial physical state that will not
exhibit any flavor oscillations.

This will give us the tools to compare the two different natures for neutrinos. A Dirac
fermion can show different behavior in this flavor mixing scenario than a Majorana neu-
trino.

87
88 CHAPTER 7. MAJORANA FLAVOR MIXING

7.1 Equations of Motion


We start with the equations of motion for a Majorana fermion, familiar from chapter 6 in
which we solved these for a single particle. The equations of motion are given by:

~ − ∂t )φ − M σ2 φ∗ = 0
(~σ · ∇
(~σ · ∇~ + ∂t )σ2 φ∗ − M ∗ φ = 0. (7.1)

Since we are looking at flavor oscillations, M is a 2 by 2 matrix in flavor space, and φ has
two components in flavor space and another two in spinor space. For the solution with
only a single flavor present, φ was also a two component spinor. This is different from
the previous discussion of Majorana fermions where we solved the equations of motion for
a single particle, without any flavor mixing. The two equations are not independent, by
assuming the mass shell condition and taking the complex conjugate of the first equation,
the second can be obtained.
Upon Fourier transforming we find
(i~σ · ~k − ∂t )φ − M σ2 φ∗ = 0
(i~σ · ~k + ∂t )σ2 φ∗ − M ∗ φ = 0, (7.2)

M can be any complex symmetric matrix. We write the components of it in the standard
form:
 
M11 M12
M= . (7.3)
M12 M22

In order to solve the equations in flavor space, we first diagonalize M . This can be done
with a unitary matrix U † = U −1 :

U M U T = Md . (7.4)

The matrix Md is a diagonal matrix created by applying the rotation U . This procedure
is discussed explicitly in appendix B.2. Summarizing this discussion, the general form of
the rotation matrix U is given by:

cos θ − sin θe−iω


 
iϕ/2
U =e (7.5)
sin θeiω cos θ.
7.1. EQUATIONS OF MOTION 89

Three variables are introduced: two phases (ϕ and ω) and an angle θ. All three are fixed
by the mass matrix. Thus if the mass matrix is known, these three variables can also be
considered as known variables.

7.1.1 Diagonalizing the Equations of Motion

With the angles and phases obtained from the diagonalization procedure, we have found
the two dimensional matrix that diagonalizes a general symmetric mass matrix. We can
use these to diagonalize the equations of motion using U † U = 1 and U ∗ U T = 1:

(i~σ · ~k − ∂t )φd − Md σ2 (φd )∗ = 0


(i~σ · ~k + ∂t )σ2 (φd )∗ − Md∗ φd = 0, (7.6)

here we used the definitions:

U φ ≡ φd U ∗ φ∗ ≡ (φd )∗ . (7.7)

The rotation matrix U works in flavor space, whereas the σ2 matrix is in spinor space,
therefore they commute. Also φd still has two components in spinor space. This in contrast
to the Dirac case, where we were able to separate the different helicities, however because
in solutions to the Majorana equations helicities are mixed, we are not able to do that. We
can however write the spinor φd as a sum over the two helicities:

X
φd = φdh ξh . (7.8)
h

Using the following form for the helicity spinors ξh :

|~k| + k3
 
1
ξ+ = q (7.9)
k1 + ik2
2|~k|(|~k| + k3 )
 
1 −k1 + ik2
ξ− = q , (7.10)
|~k| + k3
2|~k|(|~k| + k3 )
90 CHAPTER 7. MAJORANA FLAVOR MIXING

which satisfy the following identities:

(~k · ~σ )ξh = |~k|hξh (7.11)


† †
ξ+ ξ+ = ξ− ξ− = 1 (7.12)
† †
ξ+ ξ− = ξ− ξ+ = 0 (7.13)
2 ∗
iσ ξh = −hξ−h . (7.14)

Therefore we obtain the following:

X
iσ 2 (φd )∗ = h(φd−h )∗ ξh . (7.15)
h

We can enter these two expressions for φd and iσ 2 (φd )∗ into the equations of motion to find

(ih|~k| − ∂t )φdh + ihMd (φd−h )∗ = 0


(ih|~k| + ∂t )(φd−h )∗ − ihMd∗ φdh = 0. (7.16)

Using the second equation to express the time dependence of the first equation, we obtain
the following quadratic form:

(|~k|2 + ∂t2 + |Md |2 )φdh = 0. (7.17)

The solution to this quadratic equation is well known:

   
φd1h Â1h cos(ω1 t) + B̂1h sin(ω1 t)
φdh = = . (7.18)
φd2h Â2h cos(ω2 t) + B̂2h sin(ω2 t)

Note that there are still two helicities, h = 1 and h = −1. The definition of the frequency
ωi is given by:

ωi2 ≡ |~k|2 + |(Md )ii |2 . (7.19)

Next we go back to the original equation of motion in equation (7.16) to make sure that
the solution obtained for the quadratic equation also satisfies this equation of motion. We
will fix the constant B̂ih in order to obtain this:
7.2. STATISTICAL PROPAGATOR 91

ih(|~k|Â1h + (Md )11 †1−h )


B̂1h = (7.20)
ω1
ih(|~k|Â2h + (Md )22 †2−h )
B̂2h = . (7.21)
ω2

When we fix the Bi according to the above, the general solution to the quadratic form of
the equations, will also hold for the equations of motion.
This gives us for the full solution in the diagonal frame:

 
 d  ih(|~k|Â1h +(Md )11 †1−h )
φ1h Â1h cos(ω1 t) + sin(ω1 t)
φdh = d = ω1
ih(|~k|Â2h +(Md )22 †2−h )
. (7.22)
φ2h  cos(ω t) + sin(ω2 t)
2h 2 ω2

This is a full solution where we still have the freedom to choose the two component constant
operators Âih . We will try to choose them in such a way that there are no oscillations in
the flavor.

7.2 Statistical Propagator

We can do this through determining the statistical propagator associated with these solu-
tions. The statistical propagator for the first flavor is given by:

1
F11h (t; t0 ) = h[φ̂1h (t0 ), φ̂†1h (t)]i. (7.23)
2

This is same formalism discussed earlier in the case of the Dirac fermions. With the help
of the statistical propagator we can look at properties of different states, such as states
that oscillate and states that do not oscillate.
We express φ̂ in the diagonal frame by using the inverse rotation for φ̂ and φ̂† :

cos θφ̂d1h + sin θe−iω φ̂d2h


   
φ̂1h −iϕ/2
=e (7.24)
φ̂2h − sin θeiω φ̂d1h + cos θφ̂d2h
φ̂†1h φ̂†2h = eiϕ/2
 
cos θ(φ̂d1h )† + sin θeiω (φ̂d2h )† − sin θe−iω (φ̂d1h )† + cos θ(φ̂d2h )† . (7.25)
92 CHAPTER 7. MAJORANA FLAVOR MIXING

The statistical propagator of the first flavor in the diagonal frame, where we have found
the general solution, is given by

1h 2
F11h (t; t0 ) = cos θh[φ̂d1h (t0 ), (φ̂d1h (t))† ]i + sin2 θh[φ̂d2h (t0 ), (φ̂d2h (t))† ]i
2  i
+ cos θ sin θ eiω h[φ̂d1h (t0 ), (φ̂d2h (t))† ]i + e−iω h[φ̂d2h (t0 ), (φ̂d1h (t))† ]i . (7.26)

In the diagonal frame we have solved for φ̂d1h (t) and φ̂d2h (t) in terms of the initial constant
operators, so we can apply the earlier found solution here:
7.2. STATISTICAL PROPAGATOR 93

1 h  ih|~k|
F11h (t; t0 ) = cos2 θ h[Â1h , †1h ]i cos(ω1 t0 ) cos(ω1 t) − cos(ω1 t0 ) sin(ω1 t)
2 ω1
~
ih|k| k 2 
+ cos(ω1 t) sin(ω1 t0 ) + 2 sin(ω1 t0 ) sin(ω1 t)
ω1 ω1
 ih(Md )∗11 |~k|(Md )∗11 
+ h[Â1h , Â1−h ]i − cos(ω1 t0 ) sin(ω1 t) + sin(ω1 t) sin(ω1 t0 )
ω1 ω1
 ih(Md )11 |~k|(Md )11 
+ h[†1−h , †1h ]i cos(ω1 t) sin(ω1 t0 ) + sin(ω1 t) sin(ω1 t0 )
ω1 ω1
2
|(M )
d 11 | i
+ h[†1−h , Â1−h ] sin(ω1 t) sin(ω1 t0 )
ω12
1 h  ih|~k|
+ sin2 θ h[Â2h , †2h ]i cos(ω2 t0 ) cos(ω2 t) − cos(ω2 t0 ) sin(ω2 t)
2 ω2
ih|~k| k 2 
+ cos(ω2 t) sin(ω2 t0 ) + 2 sin(ω2 t0 ) sin(ω2 t)
ω2 ω2
 ih(Md )∗22 |~k|(Md )∗22 
+ h[Â2h , Â2−h ]i − cos(ω2 t0 ) sin(ω2 t) + sin(ω2 t) sin(ω2 t0 )
ω2 ω2
~
0 ih(Md )22 0 |k|(Md )22
 
† †
+ h[Â2−h , Â2h ]i cos(ω2 t) sin(ω2 t ) + sin(ω2 t) sin(ω2 t )
ω2 ω2
2
|(Md )22 | i
+ h[†2−h , Â2−h ] 2
sin(ω2 t) sin(ω2 t0 )
ω2
1 h  ih|~k|
+ cos θ sin θ eiω h[Â1h , †2h ]i cos(ω1 t0 ) cos(ω2 t) − cos(ω1 t0 ) sin(ω2 t)
2 ω2
ih|~k| k1 
+ cos(ω2 t) sin(ω1 t0 ) + sin(ω1 t0 ) sin(ω2 t)
ω1 ω2 ω1
 ih(Md )∗22 |~k|(Md )∗22 
+ eiω h[Â1h , Â2−h ]i − cos(ω1 t0 ) sin(ω2 t) + sin(ω2 t) sin(ω1 t0 )
ω2 ω2 ω1
~
0 ih(Md )11 0 |k|(Md )11
 
iω † †
+ e h[Â1−h , Â2h ]i cos(ω2 t) sin(ω1 t ) + sin(ω2 t) sin(ω1 t )
ω1 ω2 ω1

(M )
d 22 (M )
d 11
+ eiω h[†1−h , Â2−h ] sin(ω2 t) sin(ω1 t0 )
ω2 ω1

 ih|~k|
+ e h[Â2h , Â1h ]i cos(ω2 t0 ) cos(ω1 t) − cos(ω2 t0 ) sin(ω1 t)
−iω
ω1
~
ih|k| k 2 
+ cos(ω1 t) sin(ω2 t0 ) + sin(ω2 t0 ) sin(ω1 t)
ω2 ω1 ω2
ih(Md )∗11 ~ ∗ 
0 |k|(Md )11

−iω 0
+ e h[Â2h , Â1−h ]i − cos(ω2 t ) sin(ω1 t) + sin(ω1 t) sin(ω2 t )
ω1 ω1 ω2
 ih(Md )22 ~
|k|(Md )22 
+ e−iω h[†2−h , †1h ]i cos(ω1 t) sin(ω2 t0 ) + sin(ω1 t) sin(ω2 t0 )
ω2 ω1 ω2

(Md )11 (Md )22 i
+ e−iω h[†2−h , Â1−h ] sin(ω1 t) sin(ω2 t0 ) . (7.27)
ω1 ω2
94 CHAPTER 7. MAJORANA FLAVOR MIXING

We see that the resulting expression is a function of hÂih i. Because we explicitly defined
the constant  as operators, it makes sense to take an expectation value. This is the main
reason why these constants are indeed operators.

We can go through exactly the same procedure to find for the statistical propagator of the
second species:

1
F22 (t; t0 ) = h[φ̂2h (t0 ), φ̂†2h (t)]i
2
1h 2
= cos θh{φ̂d2h (t0 ), (φ̂d2h (t))† }i + sin2 θh{φ̂d1h (t0 ), (φ̂d1h (t))† }i
2  i
− cos θ sin θ eiω h{φ̂d1h (t0 ), (φ̂d2h (t))† }i + e−iω h{φ̂d2h (t0 ), (φ̂d1h (t))† }i . (7.28)

And now we use the solution found earlier for the diagonal frame:
7.2. STATISTICAL PROPAGATOR 95

1 h  ih|~k|
F22h (t; t0 ) = cos2 θ h[Â2h , †2h ]i cos(ω2 t0 ) cos(ω2 t) − cos(ω2 t0 ) sin(ω2 t)
2 ω2
~
ih|k| k 2 
+ cos(ω2 t) sin(ω2 t0 ) + 2 sin(ω2 t0 ) sin(ω2 t)
ω2 ω2
 ih(Md )∗22 |~k|(Md )∗22 
+ h[Â2h , Â2−h ]i − cos(ω2 t0 ) sin(ω2 t) + sin(ω2 t) sin(ω2 t0 )
ω2 ω2
 ih(Md )22 |~k|(Md )22 
+ h[†2−h , †2h ]i cos(ω2 t) sin(ω2 t0 ) + sin(ω2 t) sin(ω2 t0 )
ω2 ω2
2
|(M )
d 22 | i
+ h[†2−h , Â2−h ] sin(ω2 t) sin(ω2 t0 )
ω22
1 h  ih|~k|
+ sin2 θ h[Â1h , †1h ]i cos(ω1 t0 ) cos(ω1 t) − cos(ω1 t0 ) sin(ω1 t)
2 ω1
ih|~k| k 2 
+ cos(ω1 t) sin(ω1 t0 ) + 2 sin(ω1 t0 ) sin(ω1 t)
ω1 ω1
 ih(Md )∗11 |~k|(Md )∗11 
+ h[Â1h , Â1−h ]i − cos(ω1 t0 ) sin(ω1 t) + sin(ω1 t) sin(ω1 t0 )
ω1 ω1
~
0 ih(Md )11 0 |k|(Md )11
 
† †
+ h[Â1−h , Â1h ]i cos(ω1 t) sin(ω1 t ) + sin(ω1 t) sin(ω1 t )
ω1 ω1
2
|(Md )11 | i
+ h[†1−h , Â1−h ] 2
sin(ω1 t) sin(ω1 t0 )
ω1
1 h  ih|~k|
− cos θ sin θ eiω h[Â1h , †2h ]i cos(ω1 t0 ) cos(ω2 t) − cos(ω1 t0 ) sin(ω2 t)
2 ω2
ih|~k| k1 
+ cos(ω2 t) sin(ω1 t0 ) + sin(ω1 t0 ) sin(ω2 t)
ω1 ω2 ω1
 ih(Md )∗22 |~k|(Md )∗22 
+ eiω h[Â1h , Â2−h ]i − cos(ω1 t0 ) sin(ω2 t) + sin(ω2 t) sin(ω1 t0 )
ω2 ω2 ω1
~
0 ih(Md )11 0 |k|(Md )11
 
iω † †
+ e h[Â1−h , Â2h ]i cos(ω2 t) sin(ω1 t ) + sin(ω2 t) sin(ω1 t )
ω1 ω2 ω1

(M )
d 22 (M )
d 11
+ eiω h[†1−h , Â2−h ] sin(ω2 t) sin(ω1 t0 )
ω2 ω1

 ih|~k|
+ e h[Â2h , Â1h ]i cos(ω2 t0 ) cos(ω1 t) − cos(ω2 t0 ) sin(ω1 t)
−iω
ω1
~
ih|k| k 2 
+ cos(ω1 t) sin(ω2 t0 ) + sin(ω2 t0 ) sin(ω1 t)
ω2 ω1 ω2
ih(Md )∗11 ~ ∗ 
0 |k|(Md )11

−iω 0
+ e h[Â2h , Â1−h ]i − cos(ω2 t ) sin(ω1 t) + sin(ω1 t) sin(ω2 t )
ω1 ω1 ω2
 ih(Md )22 ~
|k|(Md )22 
+ e−iω h[†2−h , †1h ]i cos(ω1 t) sin(ω2 t0 ) + sin(ω1 t) sin(ω2 t0 )
ω2 ω1 ω2

(Md )11 (Md )22 i
+ e−iω h[†2−h , Â1−h ] sin(ω1 t) sin(ω2 t0 ) . (7.29)
ω1 ω2
96 CHAPTER 7. MAJORANA FLAVOR MIXING

This gives us the statistical propagator for the first and the second flavor in terms of the
initial constant operators and tells us the exact time evolution of them.

7.2.1 Initial Correlators

We would like to relate certain initial physical states with a form for the initial operators.
In order to be able to do so, we first look at the initial commutators for different states:

h[φ̂1h (t0 ), φ̂†1h (t0 )]i = cos2 θh[Â1h , †1h ]i + sin2 θh[Â2h , †2h ]i
 
−iω † iω †
+ sin θ cos θ e h[Â2h , Â1h ]i + e h[Â1h , Â2h ]i (7.30)
h[φ̂2h (t0 ), φ̂†2h (t0 )]i = cos2 θh[Â2h , †2h ]i + sin2 θh[Â1h , †1h ]i
 
− sin θ cos θ e−iω h[Â2h , †1h ]i + eiω h[Â1h , †2h ]i (7.31)
h[φ̂1h (t0 ), φ̂†2h (t0 )]i = cos2 θh[Â1h , †2h ]i − sin2 θh[Â2h , †1h ]ie−2iω
 
+ sin θ cos θe−iω h[Â2h , †2h ]i − h[Â1h , †1h ]i . (7.32)

Now we have the initial physical states in terms of the initial constant. Specifying the
initial constants corresponds to the physical state described by the commutators above.
Remember that the commutators at equal time, which is the case here, can be interpreted
as a function of the particle density of the given state.
Next we invert these expressions to obtain expressions for the constants in terms of the
initial physical states:

h[Â1h , †1h ]i = cos2 θh[φ̂1h (t0 ), φ̂†1h (t0 )]i + sin2 θh[φ̂2h (t0 ), φ̂†2h (t0 )]i
− sin θ cos θ(h[φ̂1h (t0 ), φ̂†2h (t0 )]ieiω + h[φ̂2h (t0 ), φ̂†1h (t0 )]ie−iω ) (7.33)
h[Â2h , †2h ]i = sin 2
θh[φ̂1h (t0 ), φ̂†1h (t0 )]i
+ cos θh[φ̂2h (t0 ), φ̂†2h (t0 )]i
2

+ sin θ cos θ(h[φ̂1h (t0 ), φ̂†2h (t0 )]ieiω + h[φ̂2h (t0 ), φ̂†1h (t0 )]ie−iω ) (7.34)
h[Â1h , †2h ]i = cos2 θh[φ̂1h (t0 ), φ̂†2h (t0 )]i + sin2 e−2iω θh[φ̂2h (t0 ), φ̂†1h (t0 )]i
 
+ sin θ cos θe−iω h[φ̂1h (t0 ), φ̂†1h (t0 )]i − h[φ̂2h (t0 ), φ̂†2h (t0 )]i . (7.35)

This enables to find the explicit form of the commutators of the initial operators given the
initial physical state.
7.2. STATISTICAL PROPAGATOR 97

Should we be presented with a certain initial physical state, we can use the above expres-
sions to determine the appropriate constants for this state. Next we can use the complete
expression for the statistical operator to determine the time evolution of that specific state.
In this way, we have completely solved the time evolution of any general state that can be
prepared.

7.2.2 Time Translational Invariance

Next we look at states that show time translational invariance. In order to do so, we first
rewrite the statistical propagator in terms of the average time and the time difference:

∆t = t − t0 (7.36)
t + t0
τ= . (7.37)
2

Here we naturally encounter the following two new frequencies:

1
ω̄ = (ω1 + ω2 ) (7.38)
2
∆ω̄ = ω1 − ω2 (7.39)

and the statistical propagator of the first flavor is given by


98 CHAPTER 7. MAJORANA FLAVOR MIXING

1 h 1
F11h (∆t; τ ) = cos2 θ h[Â1h , †1h ]i (cos 2ω1 τ + cos ω1 ∆t)
2 2
ih|~k| |~k|2 1 
− sin ω1 ∆t + 2 (− cos 2ω1 τ + cos ω1 ∆t)
ω1 ω1 2
ih(Md∗ )11 1 h|~k|(Md∗ )11
− h[Â1h , Â1−h ]i((sin 2τ ω1 − sin ∆tω1 ) + (cos 2ω1 τ + cos ω1 ∆t) )
ω1 2 ω12
(M ∗ )11 1 h|~k|(Md∗ )11
+ h[†1−h , †1h ]i((sin 2τ ω1 − sin ∆tω1 ) d + (− cos 2ω1 τ + cos ω1 ∆t) )
iω1 2 ω12
|(Md )11 |2 1 i
+ h[†1−h , Â1−h ] (− cos 2ω 1 τ + cos ω 1 ∆t)
ω12 2
1 h  1
+ sin2 θ h[Â2h , †2h ]i (cos 2ω2 τ + cos ω2 ∆t)
2 2
~
ih|k| |~k|2 1 
− sin ω2 ∆t + 2 (− cos 2ω2 τ + cos ω2 ∆t)
ω2 ω2 2
ih(Md∗ )22 1 h|~k|(Md∗ )22
− h[Â2h , Â2−h ]i((sin 2τ ω2 − sin ∆tω2 ) + (cos 2ω2 τ + cos ω2 ∆t) )
ω2 2 ω22
(M ∗ )22 1 h|~k|(Md∗ )22
+ h[†2−h , †2h ]i((sin 2τ ω2 − sin ∆tω2 ) d + (− cos 2ω2 τ + cos ω2 ∆t) )
iω2 2 ω22
|(Md )22 |2 1 i
+ h[†2−h , Â2−h ] (− cos 2ω 2 τ + cos ω 2 ∆t)
ω22 2
1 h
+ cos θ sin θ h[Â1h , †2h ]if1 (∆t, τ ) + h[Â1h , Â2−h ]if2 (∆t, τ )
2
+ h[†1−h , †2h ]if3 (∆t, τ ) + h[†1−h , Â2−h ]if4 (∆t, τ ) + h[Â2h , †1h ]if5 (∆t, τ )
i
+ h[Â2h , Â1−h ]if6 (∆t, τ ) + h[†2−h , †1h ]if7 (∆t, τ ) + h[†2−h , Â1−h ]if8 (∆t, τ ) .
(7.40)

We defined several functions fi (∆t, τ ) which all depend on both τ and ∆t. Their exact
form will not important in the following analysis, only their dependence.

For the second flavor we find:


7.2. STATISTICAL PROPAGATOR 99

1 h 1
F22h (∆t; τ ) = cos2 θ h[Â2h , †2h ]i (cos 2ω2 τ + cos ω2 ∆t)
2 2
~
ih|k| |~k|2 1 
− sin ω2 ∆t + 2 (− cos 2ω2 τ + cos ω2 ∆t)
ω2 ω2 2
ih(Md∗ )22 1 h|~k|(Md∗ )22
− h[Â2h , Â2−h ]i((sin 2τ ω2 − sin ∆tω2 ) + (cos 2ω2 τ + cos ω2 ∆t) )
ω2 2 ω22
(M ∗ )22 1 h|~k|(Md∗ )22
+ h[†2−h , †2h ]i((sin 2τ ω2 − sin ∆tω2 ) d + (− cos 2ω2 τ + cos ω2 ∆t) )
iω2 2 ω22
† |(Md )22 |2 1 i
+ h[Â2−h , Â2−h ] (− cos 2ω2 τ + cos ω2 ∆t)
ω22 2
1 2 h †
1
+ sin θ h[Â1h , Â1h ]i (cos 2ω1 τ + cos ω1 ∆t)
2 2
~
ih|k| |~k|1 1 
− sin ω1 ∆t + 2 (− cos 2ω1 τ + cos ω1 ∆t)
ω1 ω1 2
ih(Md∗ )11 1 h|~k|(Md∗ )11
− h[Â1h , Â1−h ]i((sin 2τ ω1 − sin ∆tω1 ) + (cos 2ω1 τ + cos ω1 ∆t) )
ω1 2 ω12
(M ∗ )11 1 h|~k|(Md∗ )11
+ h[†1−h , †1h ]i((sin 2τ ω1 − sin ∆tω1 ) d + (− cos 2ω1 τ + cos ω1 ∆t) )
iω1 2 ω12
|(Md )11 |2 1 i
+ h[†1−h , Â1−h ] (− cos 2ω 1 τ + cos ω 1 ∆t)
ω12 2
1 h
+ cos θ sin θ h[Â2h , †1h ]ig1 (∆t, τ ) + h[Â2h , Â1−h ]ig2 (∆t, τ )
2
+ h[†2−h , †1h ]ig3 (∆t, τ ) + h[†2−h , Â1−h ]ig4 (∆t, τ ) + h[Â1h , †2h ]ig5 (∆t, τ )
i
+ h[Â1h , Â2−h ]ig6 (∆t, τ ) + h[†1−h , †2h ]ig7 (∆t, τ ) + h[†1−h , Â2−h ]ig8 (∆t, τ ) .
(7.41)

Again we have used the shorthand gi (∆t, τ ) for several functions.


Next we look at states when these new propagators do not depend on τ , such that they do
not oscillate in time. This gives us the following conditions from assuming that the states
from F11h (∆t; τ ) proportional to cos2 θ and cos 2ω1 τ give:

!
|~k| |(Md )11 |2
h[Â1h , †1h ]i 1− 2 − h[†1−h , Â1−h ]i
ω1 ω12
h|~k|(Md∗ )11  † †

− h[ Â1h , Â1−h ]i + h[ Â1−h , Â1h ]i = 0. (7.42)
ω12
100 CHAPTER 7. MAJORANA FLAVOR MIXING

Secondly we have from the terms with sin 2ω1 τ :

h[Â1h , Â1−h ]i = h[†1−h , †1h ]i. (7.43)

This simplifies the first requirement to


 
(Md )11 h[Â1h , †1h ]i − h[†1−h , Â1−h ]i = 2h|~k|h[Â1h , Â1−h ]i. (7.44)

From the sin2 θ terms we find analogously

h[Â2h , Â2−h ]i = h[†2−h , †2h ]i (7.45)


 
(Md )22 h[Â2h , †2h ]i − h[†2−h , Â2−h ]i = 2h|~k|h[Â2h , Â2−h ]i. (7.46)

Any state that fits into these two conditions, will have a constant statistical propagator
and thus the statistical propagator will not oscillate with the average time. It is clear that
the correlations of different helicity states of a flavor have to be dependent.
All other mixed correlators can be set to zero:

h[Â1h , †2h ]i =0
h[Â1h , Â2−h ]i =0
h[†1−h , †2h ]i =0
h[†1−h , Â2−h ]i =0
h[Â2h , †1h ]i =0
h[Â2h , Â1−h ]i =0
h[†2−h , †1h ]i =0
h[†2−h , Â1−h ]i =0. (7.47)

This is the most general description we can give of a state that will not have any oscillations.

Example State

We can construct a state that is allowed according to these constants, since we have
expressions for different physical states in terms of the constant operators. First we choose
7.2. STATISTICAL PROPAGATOR 101

h[Â1h , Â1−h ]i = h[Â2h , Â2−h ]i = 0. (7.48)

This gives us the following constant correlators:

h[Â1h , †1h ]i = h[†1−h , Â1−h ]i = C1 (7.49)


h[Â2h , †2h ]i = h[†2−h , Â2−h ]i = C2 . (7.50)

The total number of variables we can still set ourselves is reduced to two. These constants
correspond to the following initial state:

h[φ̂1h (t0 ), φ̂†1h (t0 )]i = cos2 θh[Â1h , †1h ]i + sin2 θh[Â2h , †2h ]i
h[φ̂2h (t0 ), φ̂†2h (t0 )]i = cos2 θh[Â2h , †2h ]i + sin2 θh[Â1h , †1h ]i
 
h[φ̂1h (t0 ), φ̂†2h (t0 )]i = sin θ cos θe−iω h[Â2h , †2h ]i − h[Â1h , †1h ]i . (7.51)

We still have a choice for the helicity: positive or negative helicity. If we write out these
two options explicitly we see:

h[φ̂1+ (t0 ), φ̂†1+ (t0 )]i = cos2 θ C1 + sin2 θ C2


h[φ̂2+ (t0 ), φ̂†2+ (t0 )]i = cos2 θ C2 + sin2 θ C1
 
h[φ̂1+ (t0 ), φ̂†2+ (t0 )]i = sin θ cos θe−iω C2 − C1
h[φ̂1− (t0 ), φ̂†1− (t0 )]i = − cos2 θ C1 − sin2 θ C2
h[φ̂2− (t0 ), φ̂†2− (t0 )]i = − cos2 θ C2 − sin2 θ C1
 
h[φ̂1− (t0 ), φ̂†2− (t0 )]i = sin θ cos θe−iω C1 − C2 . (7.52)

The constants θ and ω are given by the form of the mass matrix. These six initial correlators
assure that the statistical propagator will remain constant, as was shown through the
calculation.

7.2.3 General Case

We can construct the general initial physical expectation values that will lead to no mixing.
Here we will not choose any of the constants to vanish, in contrast with the previous section.
This gives us:
102 CHAPTER 7. MAJORANA FLAVOR MIXING

h[φ̂1h (t0 ), φ̂†1h (t0 )]i = cos2 θh[Â1h , †1h ]i + sin2 θh[Â2h , †2h ]i
h[φ̂2h (t0 ), φ̂†2h (t0 )]i = cos2 θh[Â2h , †2h ]i + sin2 θh[Â1h , †1h ]i
 
† −iω † †
h[φ̂1h (t0 ), φ̂2h (t0 )]i = sin θ cos θe h[Â2h , Â2h ]i − h[Â1h , Â1h ]i
 
h[φ̂1h (t0 ), φ̂1−h (t0 )]i =e−iϕ h[Â1h , Â1−h ]i cos2 θ + h[Â2h , Â2−h ]i sin2 θe−2iω
 
h[φ̂2h (t0 ), φ̂2−h (t0 )]i =e−iϕ h[Â1h , Â1−h ]i sin2 θe−2iω + h[Â2h , Â2−h ]i cos2 θ
 
−iϕ iω −iω
h[φ̂1h (t0 ), φ̂2−h (t0 )]i =e cos θ sin θ − e h[Â1h , Â1−h ]i + e h[Â2h , Â2−h ]i
 
h[φ̂2h (t0 ), φ̂1−h (t0 )]i =e−iϕ cos θ sin θ − eiω h[Â1h , Â1−h ]i + e−iω h[Â2h , Â2−h ]i . (7.53)

Note that this shows directly that h[φ̂1h (t0 ), φ̂2−h (t0 )]i = h[φ̂2h (t0 ), φ̂1−h (t0 )]i. It is also
immediately clear that this can be more complicated than the example from the previous
section, as there are many more physical correlators present at the initial time.
We can simplify the notation by defining the following constants D:

h[Â1h , †1h ]i = D1+


h[†1−h , Â1−h ]i = D1−
h[Â2h , †2h ]i = D2+
h[†2−h , Â2−h ]i = D2−
h[Â1h , Â1−h ]i = D1∗
h[Â2h , Â2−h ]i = D2∗ . (7.54)

These are not all independent, but they satisfy

(Md )11
D1∗ = (D1+ − D1− ) (7.55)
2h|~k|
(Md )22
D2∗ = (D2+ − D2− ). (7.56)
2h|~k|

We have the freedom to fix four variables D1+ , D1− , D2+ and D2− . The other two (D1∗
and D2∗ ) are then automatically fixed. The physical initial conditions are now expressed
as
7.2. STATISTICAL PROPAGATOR 103

h[φ̂1h (t0 ), φ̂†1h (t0 )]i = cos2 θ D1+ + sin2 θ D2+


h[φ̂2h (t0 ), φ̂†2h (t0 )]i = cos2 θ D2+ + sin2 θ D1+
 
h[φ̂1h (t0 ), φ̂†2h (t0 )]i = sin θ cos θe−iω D2+ − D1+
h[φ̂1−h (t0 ), φ̂†1−h (t0 )]i = − cos2 θ D1− − sin2 θ D2−
h[φ̂2−h (t0 ), φ̂†2−h (t0 )]i = − cos2 θ D2− − sin2 θ D1−
 
† −iω
h[φ̂1−h (t0 ), φ̂2−h (t0 )]i = sin θ cos θe D1− − D2−
 
h[φ̂1h (t0 ), φ̂1−h (t0 )]i =e−iϕ D1∗ cos2 θ + D2∗ sin2 θe−2iω
 
−iϕ 2 −2iω 2
h[φ̂2h (t0 ), φ̂2−h (t0 )]i =e D1∗ sin θe + D2∗ cos θ
 
h[φ̂1h (t0 ), φ̂2−h (t0 )]i =e−iϕ cos θ sin θ − eiω D1∗ + e−iω D2∗
=h[φ̂2h (t0 ), φ̂1−h (t0 )]i. (7.57)

This is indeed the most general state that will not oscillate with τ , the average time. We
still have the freedom to choose several variables (four to be exact) which leads to ten
correlators that then are fixed. These ten correlators are functions of these four variables.
The constants ϕ and θ are again fixed by the mass matrix.
In the next chapter we will compare these results with the results found for the Dirac
fermions and analyze the differences and (possible) implications.
104 CHAPTER 7. MAJORANA FLAVOR MIXING
Chapter 8

Discussion

First we look back at what we have discussed in this thesis. We started with the overview
of the current state of the theory of flavor mixing. This is a quantum mechanical picture
of how a pure flavor state evolves as it propagates, which is a generally accepted picture.
From this framework different probabilities can be calculated to detect a neutrino in a
distinct flavor. The main shortcoming of this model is that only pure initial states are
considered. We have expanded the framework to also include mixed states.
Next we looked at the experimental effort that is currently going on and has been done
in order to verify the theory. Experiments have in general two different goals. The first is
to improve the values for mixing angles and the mass differences between different flavors.
Secondly there is still uncertainty about the nature of the neutrino, it can still be a Dirac
fermion or a Majorana fermion. These experiments try to observe neutrinoless double beta
decay. Should this be observed, then the neutrino is a Majorana particle, since this process
is only allowed for a Majorana fermion.
We started our own analysis with a discussion of the different propagators. For fermions
the definitions of these propagators can be substantially different than in the scalar case.
We also looked at the explicit form of the Feynman propagator in position space for a
fermion.
Subsequently we setup our model for flavor oscillations for Dirac fermions. In this setup
we explicitly allowed for mixed states through introducing initial expectation values of
all different commutators. The expectation values of commutators are related to number
densities of that particular state. This means that given a certain set of initial states, which
can also be mixed states, we have solved for the evolution of these states. In the process
of solving, two important notions are used. Firstly, the mass matrix can be diagonalized
in flavor space and secondly, we can use a helicity basis for the spinor structure. It turns
out that there is a certain group of initial conditions such that the system will not show
any oscillatory behavior. This done by looking for a set of initial conditions that cause the
statistical propagator to display time translational invariance.

105
106 CHAPTER 8. DISCUSSION

For Majorana fermions a similar analysis can be done. Again, with the help of expectation
values of different commutators we can give a general solution for the system which also
allows for mixed initial states. Given any initial set of expectation values of number
densities we can give the full evolution of the system. Again, the mass matrix has to be
diagonalized in flavor space and we use helicity states to describe the spinor structure of the
system. Finally, we can also construct a state that does not oscillate. This is the equivalent
of the analogous state found for the Dirac nature. It is now interesting to compare these
two states and see what this implies.

In the Dirac case the example state we constructed has several interesting properties, as
can be seen in (5.84). First of all we need initial correlations between the two different
flavors for both the left and right handed particles. This is the explicit form of an initial
mixed state and shows the relevance of the addition of allowing initially mixed states. The
possibility of initially mixed states opens up more physical possibilities, as is shown by this
result. Aside from the mixed states, we also need a defined density for the four pure states,
left and right handed particles of the first and second flavor. These are not independent.
There is a certain special relation between the initial densities, which also includes the
mixed states. There is still a dependence on parameters from the mass matrix, and we see
that complex phases in the mass matrix can also lead to complex expectation values for
the mixed correlators. As a final remark on the Dirac case there is no helicity mixing. We
can still set one of the helicity sectors completely to zero, which leads to a system in which
only a single helicity is present.

In the Majorana case we find a state that does not show any oscillatory behavior that is
similar to the Dirac state (7.52), however there are a few important differences. In the
Majorana case we again need correlations between the different flavors, there are two mixed
states that have non zero expectation values. However, the different helicity states are now
related. If we choose a certain state for the positive helicity, the negative helicity is fixed.
This is different from the Dirac case, where the helicities are strictly separated.

We have seen from the Majorana condition that a Majorana particle has half of the degrees
of freedom of a Dirac particle. This is also reflected in the number of degrees of freedom
in which we can set up our state that does not oscillate, where in the Dirac case there are
also twice as many free constants for us to choose.

8.1 Applications

The developed theoretical basis can be applied to a laboratory setting. We will discuss
several aspects of this possible application. Secondly, there is also information stored in
thermal sources of neutrinos. Two thermal sources are discussed, the cosmic neutrino
background and neutrino production in supernovae.
8.1. APPLICATIONS 107

8.1.1 Laboratory

Since we know the differences between Majorana and Dirac fermions in their behavior
in the case of mixed initial states, is there a way how we can use this difference in an
experiment to detect the nature of the neutrino?
In order to do so we will try to come up with a set up that will show oscillations in the case
of Dirac neutrinos, but will be constant in the case of neutrinos having a Majorana nature.
We have previously discussed that if we can create a Majorana state with the following
initial conditions, there will be no oscillations:

1 cos 2θ
h[φ̂1+h (t0 ), φ̂†1+h (t0 )]i = (C1 + C2 ) + (C1 − C2 )
2 2
1 cos 2θ
h[φ̂2+h (t0 ), φ̂†2+h (t0 )]i = (C1 + C2 ) − (C1 − C2 )
2 2
1  
h[φ̂1+h (t0 ), φ̂†2+h (t0 )]i = sin 2θe−iω C2 − C1
2
† 1 cos 2θ
h[φ̂1−h (t0 ), φ̂1−h (t0 )]i = − (C1 + C2 ) − (C1 − C2 )
2 2
1 cos 2θ
h[φ̂2−h (t0 ), φ̂†2−h (t0 )]i = − (C1 + C2 ) + (C1 − C2 )
2 2
† 1 −iω
 
h[φ̂1−h (t0 ), φ̂2−h (t0 )]i = − sin 2θe C2 − C1 . (8.1)
2

Here we defined the following:

h[Â1h , †1h ]i = h[†1−h , Â1−h ]i = C1 (8.2)


h[Â2h , †2h ]i = h[†2−h , Â2−h ]i = C2 . (8.3)

Note that this has to be done for both h = 1 and h = −1. So if we can produce two beams
of neutrinos that have these properties, each of them with a pure helicity and superimpose
them, there should be no oscillations in the Majorana case. If neutrinos are Dirac particles,
there will be oscillations, because we need a particular combination of left handed and right
handed particles to turn off the oscillations for Dirac particles (5.84).
Is it actually feasible to create such a beam in which neutrino have this particular state?
It is possible to create a beam of electrons and muons with a fixed helicity, and should be
possible to create from this a beam of neutrinos with only one helicity present. Creating
the mixed state will be more difficult. When a neutrino is produced in a reaction, it is
always created in a flavor eigenstate. The question is, how can we tune the neutrino beam
of two different flavors to have the exact mixed state as specified by the theory.
108 CHAPTER 8. DISCUSSION

The need for the off diagonal state can be circumvented by choosing the constants C1 and
C2 in such a fashion that the mixed commutator is zero: C1 = C2 . For pure states we can
identify the number density for a particular state as:

h[φ̂ih (t0 ), φ̂†ih (t0 )]i = 1 − 2ni,h (t0 ). (8.4)

The states of the two flavors are now given by

h[φ̂1+ (t0 ), φ̂†1+ (t0 )]i = 1 − 2n1,+ (t0 ) (8.5)


= h[φ̂2+ (t0 ), φ̂†2+ (t0 )]i = 1 − 2n2,+ (t0 ); (8.6)

for the positive helicity sector and for the negative helicity we find:

h[φ̂1− (t0 ), φ̂†1− (t0 )]i = 1 − 2n1,− (t0 ) (8.7)


= h[φ̂2− (t0 ), φ̂†2− (t0 )]i = 1 − 2n2,− (t0 ). (8.8)

Combining these two we find that

X
1 = ni+ (t0 ) + ni− (t0 ) = nih (t0 ), (8.9)
h

which is an expression for the sum of the initial number densities. Secondly, we also need
that the densities of the two flavors are equal:

n1h (t0 ) = n2h (t0 ). (8.10)

In order to get nih (t0 ) ≥ 21 , which must be true for some states in this system, population
inversion of the neutrino states needs to occur. Any Majorana neutrino system in this
state will not exhibit any neutrino oscillations.

Outlook

As a concluding remark, we also have to consider that neutrinos in nature have three
flavors instead of two. The complete calculation was done with only two flavors present.
In principle, an analogous calculation could be done for three flavors present, and that can
then show if the results are the same.
8.1. APPLICATIONS 109

A different application of this new framework is to take a closer look at CP phases for
neutrinos. As mentioned in the theoretical introduction, Dirac neutrinos can have a single
complex CP phase, whereas Majorana neutrinos can have up to three CP phases. It might
be possible to construct an initial state that will allow a measurement of these CP phases or
a physical state that is not allowed by the Dirac framework. Even a confirmation that two
CP phases are present, would lead to the conclusion that neutrinos are Majorana particles.
Aside from the time translational invariant state that we have constructed in order to come
up with a new experiment to test the nature of the neutrino, there can be other applications
where this framework can shed new light on the unknowns of neutrino physics.

8.1.2 Thermal Distributions

In nature neutrinos could be found in a thermal equilibrium. The number densities of


a thermal distribution are well known. We discuss two examples of neutrinos in a ther-
mal equilibrium that can provide additional information about mixing properties and CP
mixing: the CνB, Cosmic neutrino Background, and neutrinos produced at supernovae.
The CνB has a predicted neutrino number density given by the following Fermi-Dirac
distribution in chemical equilibrium:

1
ni = (8.11)
1 + e(Ei −µ)/(kT )
1
= . (8.12)
1 + e i /(kT )−ξ
E

In these expressions we see that the energy of the species is given by Ei , the chemical
potential by µ, the Boltzmann constant by k and the temperature by T . The second
expression is often used in literature [44]. Secondly, we defined ξ = µ/(kT ), the chemical
potential in units of kT or the degeneracy parameter.
Experimentally the following results have been found from observational data: the tem-
perature of the CνB is predicted to be T = 1.95 eV from CMB data. On the degeneracy
parameter the a quoted bound is given by −0.04 ≤ ξ ≤ 0.07, with a best fit for ξ = 0.0245
[42], note that some claim to have found evidence that ξ is nonzero. The different ξ
for every neutrino flavor are expected to be very similar because of oscillations between
different flavors [43]. The number densities in the case for the CνB can be calculated from
this data given the chemical potential and the temperature.
In the core of a supernova the idealized picture of a neutrino produced in a pure flavor
state which then starts to oscillate is no longer valid [45]. Because of the large density of
neutrinos, collisions have to be included for a complete picture. The very large density of
110 CHAPTER 8. DISCUSSION

neutrinos will cause them to collide, leading to a thermal equilibrium. The neutrinos will
then reach a thermal state.
The thermal state that is reached through collisions is diagonal in the mass basis. By using
the rotation matrices derived, we know that this state is characterized by the following
number densities in the flavor basis:

1 1
n11h = (nd11h + nd22h ) + cos(2θ)(nd11h − nd22h ) (8.13)
2 2
1 d 1
n22h = (n11h + nd22h ) − cos(2θ)(nd11h − nd22h ) (8.14)
2 2
1
n12h = sin(2θ)e−iω (nd22h − nd11h ) (8.15)
2

In the mass basis the two densities are given by nd11h and nd22h . Note that if there are no
oscillations in the mass basis, the expressions in the flavor basis are also constant through
time. To identify these states with the states discussed in the previous section, we note
that in this case the mixed correlator is not zero, but it has a definite value.
For example when the mass basis becomes diagonal if the collisions settle into an equi-
librium, the flavor basis densities are given by the above expressions. The first two
densities are the densities that represent the conventional pure states. The third state
is an interesting mixed state. If it is possible to measure the off diagonal component, a
measurement for a complex CP phase is possible. We see that the CP angle does not
appear in the diagonal states, which is confirmed by current measurements, however the
off diagonal density contains information about the complex phase.
If we know how to measure the two densities in the flavor basis, which is the basis in which
conventional measurements take place, we have expressions for the total particle number:
n11h + n22h = nd11h + nd22h (note that the total particle number is equal in both bases as
expected) and the difference of the particle number in the mass basis with an angle θ:
cos(2θ)(nd11h − nd22h ). If we would be able to measure the diagonal densities in the mass
basis through some mechanism, we would obtain only one new piece of information. The
total particle number will be obtained, however we will also be able to obtain the difference
in particle numbers, thus giving an indirect measurement of the angle θ.
If we can measure the energy spectrum of the thermal neutrinos, we can extract the
energies of the mass eigenstates from this. This would allow us to construct the different
absolute values of the masses in the diagonal basis. Combining this measurement with
the measurement for the mixing angle θ could give us information about the phase. This
can be possible because the angle θ depends on the absolute value of the difference of the
masses (|M11 − M22 |), meaning there can be a dependence on a phase present. Absolute
values for the masses in the flavor basis can be obtained by experiments such as KATRIN
and additional information in the case of Majorana neutrinos by measurements of the 0νββ
8.1. APPLICATIONS 111

decay. Combining different measurements of the masses and the mixing angles can give us
more information about the CP mixing angle.
Finally we will discuss the thermal state in the framework presented in chapter 7. It was
mentioned that the thermal case was an example of a case where other correlations are
also present. These correlations can explicitly be given by the two starting points that
there are no oscillations in the thermal state and that both the helicity states have equal
properties, there is no dependence on the helicity in the thermal state.
The first statement that there are no oscillations, implies that the thermal state must
obey the general conditions in 7.57. Secondly, the helicity independence gives us that
D1− = −D1+ and D2− = −D2+ . As a result we obtain a non zero Di∗ component, given
by:

(Md )11
D1∗ = D1+ (8.16)
h|~k|
(Md )22
D2∗ = D2+ (8.17)
h|~k|

We can explain the dependence on the helicity by looking at the definition of Di∗ . If
we change the order of the commutator, this is analogous to changing the helicity, both
operations give a minus sign.
The values found for Di∗ lead to non zero expectation values for several other commutators:

 
−iϕ 2 2 −2iω
h[φ̂1h (t0 ), φ̂1−h (t0 )]i =e D1∗ cos θ + D2∗ sin θe
 
h[φ̂2h (t0 ), φ̂2−h (t0 )]i =e−iϕ D1∗ sin2 θe−2iω + D2∗ cos2 θ
 
h[φ̂1h (t0 ), φ̂2−h (t0 )]i =e−iϕ cos θ sin θ − eiω D1∗ + e−iω D2∗
= h[φ̂2h (t0 ), φ̂1−h (t0 )]i. (8.18)

Comparing the theoretical expressions with the number densities in the diagonal mass basis
for the thermal state we see that we must have

D1+ = 1 − 2nd11h (8.19)


D2+ = 1 − 2nd22h . (8.20)

And thus the starred components are given by


112 CHAPTER 8. DISCUSSION

(Md )11
D1∗ = (1 − 2nd11h ) (8.21)
~
h|k|
(Md )22
D2∗ = (1 − 2nd22h ). (8.22)
~
h|k|

We can finally express the mixed correlators as a function of the number densities of the
first and second species:

 (M ) (Md )22 
d 11
h[φ̂1h (t0 ), φ̂1−h (t0 )]i =e−iϕ (1 − 2nd11h ) cos2 θ + (1 − 2nd22h ) sin2 θe−2iω
h|~k| h|~k|
 (M ) (Md )22 
d 11
h[φ̂2h (t0 ), φ̂2−h (t0 )]i =e−iϕ (1 − 2nd11h ) sin2 θe−2iω + (1 − 2nd22h ) cos2 θ
h|~k| h|~k|
1  (Md )11 (Md )22 
h[φ̂1h (t0 ), φ̂2−h (t0 )]i =e−iϕ sin 2θ − eiω (1 − 2nd11h ) + e−iω (1 − 2nd22h ) .
2 h|~k| h|~k|
(8.23)

This leads us to conclude that if neutrinos have a Majorana nature, these correlators
must be present in the thermal state. We have reached this conclusion by applying the
theory that was derived in this thesis on the thermal state where there are no oscillations.
The theory predicts that in order for the oscillations to vanish, these correlators between
different helicity states must be present.
Appendices

113
Appendix A

Conventions

The Pauli matrices are as usual given by

 
1 0 1
σ = σx = (A.1)
1 0
 
0 −i
σ 2 = σy = (A.2)
i 0
 
1 0
σ 3 = σz = . (A.3)
0 −1

These matrices all satisfy

σ i σ j = δ ij I + ijk σ k , (A.4)

here ijk is the antisymmetric Levi-Civita symbol.

The direct product is written as

(a ⊗ b)(c ⊗ d) = (ac) ⊗ (bd)

This allows us to write the γ-matrices as direct products of the Pauli matrices and the
identity matrix. We use the chiral representation in this thesis, which is

115
116 APPENDIX A. CONVENTIONS

 
0 1 0 −I
γ = −σ ⊗ I = (A.5)
−I 0
 
0 σi
γi 2 i
= iσ ⊗ σ = (A.6)
−σ i 0
 
I 0
γ5 3
=σ ⊗I (A.7)
0 −I
 
3 2 −iσ 2 0
C = −iσ ⊗ σ = , (A.8)
0 iσ 2

where the γ-matrices satisfy

{γ µ , γ ν } = 2g µν (A.9)

Here γ 0 is hermitian and γ i antihermitian. γ 5 anticommutes with all the other γ µ .


The charge conjugation matrix (C) satisfies the following properties:

C T = C † = −C (A.10)
CC † = C † C = I (A.11)
C 2 = I. (A.12)

The projection operators and their Dirac adjoint are given by

1 − γ5
ψL = ψ (A.13)
2
1 + γ5
ψR = ψ (A.14)
2
1 + γ5
ψ̄L = ψ̄ (A.15)
2
1 − γ5
ψ̄R = ψ̄ . (A.16)
2
Appendix B

Diagonalizing the Mass Matrix

In this appendix we will show how to perform a rotation in flavor space that diagonalizes
the mass matrix. The main advantage of this rotation is that the off diagonal terms reduce
to zero, therefore simplifying some relevant equations considerably. First a general mass
matrix will be considered, after which the case for the symmetric matrix, which appears
in the case of a Majorana neutrino, is discussed.

B.1 General Matrix


The first step is to find the rotation matrices that comply with the following equations:

UL† M UR = Md
UR† M † UL = Md† . (B.1)

In these equations the mass matrix is given by M . The diagonal mass matrix (Md ) can be
constructed by multiplying the original mass matrix (M ) with the two rotation matrices
UL and UR . These two rotation matrices represent the rotation in flavor space. From these
two equations we can infer:

UL† (M M † )UL = Md Md† = |Md |2 (B.2)


UR† (M † M )UR = Md† Md = |Md |2 . (B.3)

This shows the action of the two rotation matrices explicitly. It is shown in (5.16) that the

117
118 APPENDIX B. DIAGONALIZING THE MASS MATRIX

subscripts of the rotation matrix are not chosen at random. The UL matrix is associated
with the left handed spinor and UR with the right handed spinor, and analogously for the
densities as was shown in (C.52).
The mass matrix is a 2x2 matrix in this example, and can be given in component form in
general by:
 
m11 m12
M= . (B.4)
m21 m22

Writing out the matrix multiplication we get the following form for M M † :

m11 m∗21 + m12 m∗22


   
† a b |m11 |2 + |m12 |2
MM = = , (B.5)
b∗ c m21 m∗11 + m22 m∗12 |m21 |2 + |m22 |2

similarly,

m12 m∗11 + m22 m∗21


   
† α β |m11 |2 + |m21 |2
M M= = . (B.6)
β∗ γ m11 m∗12 + m21 m∗22 |m12 |2 + |m22 |2

We can directly see that both are hermitian expressions. In the discussion in section
C.4 is was mentioned that hermitian quantities such as M M † transform with a single
transformation matrix, and quantities that are not hermitian transform such as M under
a combination of the two transformation matrices. We can write the rotation matrices UL
and UR in the general form:

 
cos θ̄ sin θ̄eiω̄
UR = (B.7)
− sin θ̄e−iω̄ cos θ̄
 
cos θ sin θeiω
UL = . (B.8)
− sin θe−iω cos θ

These matrices are unitary, as can be verified easily from these forms. We can infer
from (B.3) and demanding the diagonal terms of the diagonal matrix indeed give zero the
following expressions for the angle θ:

2|β|
sin 2θ̄ = p (B.9)
4|β| + (γ − α)2
2

γ−α
cos 2θ̄ = p . (B.10)
4|β| + (γ − α)2
2
B.1. GENERAL MATRIX 119

Here we defined the absolute value of β as following:

β ≡ |β|eiω̄ , (B.11)

such that ω̄ = arg(m12 m∗11 + m∗21 m22 ). Similarly we find for the right handed rotation
matrix an expression that is very much similar to the left handed rotation case:

2|b|
sin 2θ = p (B.12)
4|b|2 + (c − a)2
c−a
cos 2θ = p . (B.13)
4|b|2 + (c − a)2

Analogously we defined

b ≡ |b|eiω . (B.14)

such that ω = arg(m12 m∗22 + m∗21 m11 ). The diagonal form of |Md |2 is now found out to be,
by multiplying out

γ−α
1 
† 2
(α + γ) − 2
cos 2θ − |β| sin 2θ 0
(M M )d = ,
0 1
2
(α + γ) + γ−α
2
cos 2θ + |β| sin 2θ
(B.15)

and equivalently

1 c−a

† 2
(a + c) − 2
cos 2θ̄ − |b| sin 2θ̄ 0
(M M )d = 1 . (B.16)
0 2
(a + c) + c−a
2
cos 2θ̄ + |b| sin 2θ̄

According to (B.3), these last two equations are equal, which can indeed be confirmed by
expressing them in the explicit components mij , this is not immediately obvious from the
form in which they are in the previous expression. However, we also need the diagonal
form of M , not only M 2 . So now we write from (B.1) for Md :
120 APPENDIX B. DIAGONALIZING THE MASS MATRIX

(Md )11 =m11 cos θ̄ cos θ − m12 sin θ̄ cos θe−iᾱ − m21 cos θ̄ sin θeiα + m22 sin θ̄ sin θei(α−ᾱ)
(Md )21 =m11 cos θ̄ sin θe−iα − m12 sin θ̄ sin θe−i(α+ᾱ) + m21 cos θ̄ cos θ − m22 sin θ̄ cos θe−iᾱ
(Md )12 =m11 sin θ̄ cos θeiᾱ − m12 sin θ̄ sin θei(α+ᾱ) + m21 cos θ̄ cos θ − m22 cos θ̄ sin θeiα
(Md )22 =m11 sin θ̄ sin θei(ᾱ−α) + m12 cos θ̄ sin θe−iα − m21 sin θ̄ cos θeiᾱ + m22 cos θ̄ cos θ.
(B.17)

By construction, the off diagonal terms (Md )21 and(Md )12 are zero. This can be checked by
explicitly calculating expressions for θ and ω that satisfy this to find out that they indeed
concur with the previously found identities.

B.2 Symmetric Matrix


In the case that the matrix M is symmetric, which is for instance the case for a Majorana
mass matrix, the diagonalization procedure is different. This can be done with a unitary
matrix U † = U −1 in the following way:

U M U T = Md . (B.18)

Note that we are not using the form U M U −1 , which is normally used in diagonalization
procedures. We will show that it is possible to diagonalize a matrix in the preceding form
with a unitary matrix U −1 = U † .
First we can write the matrix U as

cos θ − sin θe−iω


 
iϕ/2
U =e . (B.19)
sin θeiω cos θ

The unitarity can be checked immediately, note that the determinant of this matrix is not
necessarily one, but can also be a complex phase. The phases ω and ϕ, and the angle θ
are real numbers. Explicit calculation for the terms of the diagonal matrix (U M U T ) give
the following:

(Md )12 = (Md )21 = eiϕ sin θ cos θM11 eiω + cos2 θM12 − sin2 θM12 − sin θ cos θM22 e−iω


(B.20)
iϕ 2 −iω 2 −2iω

(Md )11 = e cos θM11 − 2 cos θ sin θM12 e + sin θM22 e (B.21)
(Md )22 = eiϕ cos2 θM22 + 2 cos θ sin θM12 eiω + sin2 θM11 e2iω

(B.22)
B.2. SYMMETRIC MATRIX 121

This leaves us with the following requirements for this to be a diagonal matrix:

eiϕ M11 sin θ cos θeiω + cos2 θM12 − sin2 θM12 − M22 sin θ cos θe−iω = 0

(B.23)
eiϕ sin θ cos θ M22 e−iω − M11 eiω = eiϕ cos2 θ − sin2 θ M12 .
 
(B.24)

This notation suggests that we can set ϕ = −ArgM12 . This results in the following:

eiϕ sin θ cos θ M22 e−iω − M11 eiω = cos2 θ − sin2 θ |M12 |.
 
(B.25)

The right hand side of the equation is now completely real, therefore the left hand side
must also be real. This gives us the following requirement on ω:

Im eiϕ (M22 e−iω − M11 eiω ) = 0.



(B.26)

Expanding around the real and imaginary parts of M11 and M22 , we find the following
requirement for the phase ω:

ReM22 sin ϕ + ImM22 cos ϕ − ReM11 sin ϕ − ImM11 cos ϕ


tan ω = . (B.27)
ReM22 cos ϕ + ImM22 sin ϕ + ReM11 cos ϕ + ImM11 sin ϕ

Now we have the completely real equation:

(sin θ cos θ)|M22 − M11 | = cos2 θ − sin2 θ |M12 |,



(B.28)

which we can solve by the double angles

2|M12 |
sin 2θ = p 2
(B.29)
4|M12 | + |M11 − M22 |2

and

|M11 − M22 |
cos 2θ = p 2
. (B.30)
4|M12 | + |M11 − M22 |2

Note that these angles are indeed real, if we would not have defined the phases ϕ and ω
properly, we would have not been able to choose θ real.
122 APPENDIX B. DIAGONALIZING THE MASS MATRIX
Appendix C

Kinetic Description

In this chapter we will look at a kinetic description of flavor mixing. This is done in
the same spirit as in references [24]-[26]. Here the main goal was to propose models that
could lead to CP violations, thereby inducing baryogenesis. Here the goal is different, we
would like to understand the mixing process better through this kinetic description. In
this approach, the Wigner function is essential and will therefore also be treated.
First we split the mass matrix in a hermitian and an antihermitian part:

1
MH = (M + M † ) (C.1)
2
1
MA = (M − M † ). (C.2)
2i

The mass matrix is here of N by N size, we do not pose any constraints on the size here.
The off diagonal terms in this mass matrix cause the mixing of different flavors.
The Dirac equation now has a different form with these new hermitian and antihermitian
parts:

(i∂/ − MH − iγ 5 MA )ψ = 0. (C.3)

We define the Wigner function as:


Z
<
iS (k, x) = − d4 reik·r h0|ψ̄(x − r/2)ψ(x + r/2)|0i. (C.4)

Note that we have


iγ 0 S < = (iγ 0 S < )†

123
124 APPENDIX C. KINETIC DESCRIPTION

and is thus a Hermitian expression. The Dirac equation in this Wigner space, defined by
the Wigner transform of the Dirac equation, is given by


i i
(k/ + γ 0 ∂t − (MH + iγ 5 MA )e− 2 ∂t ∂k0 )S < = 0. (C.5)
2
ˆ
We note that the helicity operator ĥ = ~k · ~γ γ 0 γ 5 commutes with the Dirac equation in
Wigner space. We use the following notation for the unit vector pointing in the direction
of the momentum:
~kˆ = ~k/|~k|

C.1 Wigner Transformation

The Wigner transformation is often used when one tries to find a quantum mechanical
version of the Boltzmann equations [23]. These equations describe particle number densities
in phase space. However phase space is not a well defined object in quantum mechanics
because of the Heisenberg uncertainty principle, a particle cannot have both a well defined
momentum and position. The Wigner distribution is a distribution which could be called
a quantum mechanical version of the Boltzmann distribution in classical mechanics. It
is obtained when one tries to find a link between the wave function from Schrdinger’s
equation and a probability distribution in phase space. The Wigner transformation takes
an operator on a Hilbert space to a function on phase space:

Z ∞
s s
g(x, p) = ds eips/~ hx − | Ĝ |x + i. (C.6)
−∞ 2 2

An example of this transformation is the Wigner quasi-probability function, which is the


Wigner transformation of the density matrix. The evolution in time of the density matrix
is dictated by the Von Neumann equation:

∂ρ
i~ = [H, ρ]. (C.7)
∂t

And the Wigner transform of the Von Neumann equation is called Moyal’s evolution
equation:

∂W (x, p, t)
= −{{W (x, p, t) , H(x, p)}}. (C.8)
∂t
C.2. HELICITY AND CHIRALITY 125

Here we use the Moyal bracket {{, }}, which is defined by

 
~ ← → ← →
{{f, g}} = sin ( ∂ x ∂ p − ∂ p ∂ x ) g(x, p). (C.9)
2

In the limit ~ → 0 , the classical limit, this is simply a Poisson bracket, and the Moyal’s
evolution equation reduces to the classical Liouville equation.
The inverse of the Wigner transformation is the Weyl transformation. The Weyl transfor-
mation is a map from phase space to a Hilbert space operator. With this transformation,
quantum mechanics can be expressed in phase space, a method which was pioneered by
Moyal and Groenewold in the 1940s. Sometimes this map is referred to as the Weyl
quantization, however it is not a quantization scheme. It does not take a ’classical’ phase
space function to a ’quantum’ operator, it only changes the representation.

C.2 Helicity and Chirality

We see the helicity appearing naturally in these equations. Helicity is closely related to
chirality, two notions that are often confused, however are of such an importance that they
deserve some more explanation.
Helicity deals with the relative orientation of the momentum and the spin of a particle.
Helicity can be defined as the spin in the direction of motion, and is thus given by the
inner product of the spin with the unit vector of the momentum:

~σ · p~
ĥ ≡ . (C.10)
|p|

The eigenvalues of this operator are 1 and −1. The positive eigenstate is often referred to
as left-handed, and the negative eigenstate as right-handed. As we mentioned before, the
helicity operator commutes with the Dirac equation in this case, and therefore the helicity
is conserved. Under rotations, helicity is also conserved, this is ensured by the dot product
in the definition. Unfortunately, this is not case with boosts, they do not conserve helicity.
For a massive particle it is always possible to boost to such a frame in which the particle is
moving in the opposite direction, but the spin is not flipped. The helicity is thus reversed
in the boosted frame as opposed to the original frame. For a massive particle helicity is
therefore dependent on the observer. Because a massless particle travels at the speed of
light, this boost is only possible for a massive particle. For a massless particle the helicity
is Lorentz invariant.
126 APPENDIX C. KINETIC DESCRIPTION

Chirality is closely connected to the matrix γ 5 , which is characterized by the anticommu-


tation with the other gamma-matrices:

{γ 5 , γ µ } = 0. (C.11)

Which allows us to write from the anticommutation relations:


 
5 0 1 2 3 1 0
γ = iγ γ γ γ = . (C.12)
0 −1

This is fixed such that

(γ 5 )† = γ 5 (C.13)
(γ 5 )2 = I (C.14)

which allows us to write the projection operators as

1 − γ5
PL = (C.15)
2
1 + γ5
PR = . (C.16)
2

A Dirac spinor Ψ can be split up according to

Ψ = (PL + PR )Ψ = PL Ψ + PR Ψ = ΨL + ΨR , (C.17)

for ΨL we see

PL ΨL = ΨL (C.18)
PR ΨL = 0 (C.19)

and similar for PR . Under Lorentz transformations the chirality of a particle is not changed.
Chiral projections can therefore be Lorentz covariant. It is not conserved by a fermionic
particle, because it does not commute with the mass term in the Dirac equation.
Comparing helicity and chirality we can see that helicity is conserved for a free particle,
but is not Lorentz invariant and chirality is Lorentz invariant but is not conserved for a
free particle.
C.3. WIGNER REPRESENTATION 127

C.3 Wigner Representation

Keeping this notion of helicity in our mind we can go back to the flow equations. We are
allowed to make the following ansatz for the Wigner function:

1 ˆ
− iγ 0 S < = σ µ gµh ⊗ (I + h~k · ~σ ). (C.20)
4

The Pauli matrices are here denoted as σ µ , with the σ 0 matrix being the identity matrix.
Plugging this ansatz back into (C.5) provides all the data about the state of the fermion.
Now we need to extract this information in a more handsome form. In order to do this,
we multiply this by σ µ , take the trace, take the hermitian part and integrate over k0 . To
make life easier, we first write the gamma matrices as direct products of sigma matrices.
This is all done in Appendix A.

We multiply the ansatz (C.20) with γ 0 in order to be able to plug the ansatz in the Dirac
equation (C.5). Equation (C.20) now reads

1 ˆ
S < = (−iσ 1 g0h − ig1h + σ 3 g2h − σ 2 g3h ) ⊗ (I + h~k · ~σ ). (C.21)
4

Because we wrote the gamma matrices as a direct product, there is only one identity from
which we can calculate everything:

σ i σ j = δ ij I + ijk σ k

We find for the Dirac equation:

 ←

1 0 1 ~ 2 i 1 3 i
∂ ∂
0= −k (σ ⊗ I) − k(iσ ⊗ ~σ ) − (σ ⊗ I)∂t − (MH + iMA (σ ⊗ I))e 2 k t 0

4 2
ˆ
 
(−iσ 1 g0h − ig1h + σ 3 g2h − σ 2 g3h ) ⊗ (I + h~k · ~σ ) . (C.22)

We can multiply the direct products according to (a ⊗ b)(c ⊗ d) = (ac) ⊗ (bd), to give
128 APPENDIX C. KINETIC DESCRIPTION

!
1 0 1 ~k · ~σ
0 = (ik − ∂t ) g0h + σ 1 g1h + σ 2 g2h + σ 3 g3h ⊗ I + h

4 2 |~k|
1   
+ −iσ 3 g0h + σ 2 g1h − σ 1 g2h + ig3h ⊗ ~k · ~σ + h|~k|
4 !
MH − i ∂←t ∂k ~k · ~σ
e 2 0 iσ 1 g0h + ig1h + σ 3 g2h − σ 2 g3h ⊗ I + h

+
4 |~k|
!
MA − i ∂←t ∂k ~k · ~σ
e 2 0 −iσ 2 g0h − σ 3 g1h − ig2h + σ 1 g3h ⊗ I + h

+ . (C.23)
4 |~k|

This equation can be multiplied by σ µ after which we can take the trace to find equations
that govern the evolution of gµh . The trace of (C.23) is

← ←
1 i i
0 = (ik0 − ∂t )g0h + ig3h h|~k| + MH e− 2 ∂t ∂k0 ig1h − MA e− 2 ∂t ∂k0 ig2h . (C.24)
2

Since the Pauli matrices are traceless, all the terms with a σ i do not have a trace. We
subtract the Hermitian part to keep only the time derivative of g0h , and get rid of k0 . The

i
expansion of e− 2 ∂t ∂k0 stops after the first order term, the derivatives give zero. The first
order term gives a boundary term upon integrating over k 0 and we only have to keep the
order zero term, 1. In this case the h|~k| term drops out and we get

← ← ← ←
i i i i
0 = ġ0h + iMH e− 2 ∂t ∂k0 g1h − igh1 MH e+ 2 ∂t ∂k0 + iMA e− 2 ∂t ∂k0 g2h − ig2h MA e+ 2 ∂t ∂k0 . (C.25)

The h|~k| term does not drop out in every equation, the three remaining equations are:

← ← ← ←
i i i i
0 = ġ1h + 2h|~k|g2h + iMH e− 2 ∂t ∂k0 g0h − ig0h e+ 2 ∂t ∂k0 MH + MA e− 2 ∂t ∂k0 g3h + g3h e+ 2 ∂t ∂k0 MA
(C.26)
← ← ← ←
i i i i
0 = ġ2h − 2h|~k|g1h + MH e− 2 ∂t ∂k0 g3h + g3h e+ 2 ∂t ∂k0 MH − iMA e− 2 ∂t ∂k0 g0h + ig0h e+ 2 ∂t ∂k0 MA
(C.27)
← ← ← ←
i i i i
0 = ġ3h − MH e− 2 ∂t ∂k0 g2h − g2h e+ 2 ∂t ∂k0 MH − MA e− 2 ∂t ∂k0 g1h − g1h e+ 2 ∂t ∂k0 MA . (C.28)

Now we integrate over k 0 and we define


C.3. WIGNER REPRESENTATION 129

Z
dk0
fµh ≡ gµh . (C.29)

The four new fµh quantities have each an interpretation [25]. They all represent different
densities in Wigner space, namely:

• f0h is a charge density (it is the zeroth component of the vector current), analogous
to hψ̄γ0 ψi

• f1h is a scalar density, analogous to hψ̄ψi

• f2h is a pseudo scalar density, analogous to hψ̄γ 5 ψi

• f3h is a axial charge density, analogous to hψ̄γ 0 γ 5 ψi.

With the use of these four hermitian quantities the equations can be simplified to

f˙0h + i[MH , f1h ] + i[MA , f2h ] = 0 (C.30)


f˙1h + 2h|~k|f2h + i[MH , f0h ] − {MA , f3h } = 0
f˙2h − 2h|~k|f1h + {MH , f3h } + i[MA , f0h ] = 0
f˙3h − {MH , f2h } + {MA , f1h } = 0.

One can immediately see that the terms in every line are the terms in (C.23) that are
multiplied by the same Pauli matrix. If we look at the case of two different flavors, the
mass matrix is a two by two matrix. This already gives us sixteen equations when we write
them out in components. However using the hermiticity of f , (f˙21 )†µh = (f˙12 )µh , and four
equations become superfluous. These are the equations for (f˙12 )µh , and there are twelve
equations left. Writing out these twelve equations explicitly:
130 APPENDIX C. KINETIC DESCRIPTION

0 = (f˙11 )0h + i(MH )12 (f21 )1h − i(f12 )1h (MH )21
+ i(MA )12 (f21 )2h − (f12 )2h (MA )21 (C.31)
0 = (f˙12 )0h + i(MH )11 (f12 )1h + i(MH )12 (f22 )1h − i(f11 )1h (MH )12 − i(f12 )1h (MH )22
+ i(MA )11 (f12 )2h + i(MA )12 (f22 )2h − i(f11 )2h (MA )12 − i(f12 )2h (MA )22 (C.32)
0 = (f˙22 )0h + i(MH )21 (f12 )2h − i(f21 )2h (MH )12
+ i(MA )21 (f12 )2h − (f21 )2h (MA )12 (C.33)
0 = (f˙11 )1h + 2h|~k|(f11 )2h + i(MH )12 (f21 )0h − i(f12 )0h (MH )21
− 2(MA )11 (f11 )3h − (MA )12 (f21 )3h − (f12 )3h (MA )21 (C.34)
0 = (f˙12 )1h + 2h|~k|(f12 )2h + i(MH )11 (f12 )0h + i(MH )12 (f22 )0h − i(f11 )0h (MH )12
− i(f12 )0h (MH )22 − (MA )11 (f12 )3h − (MA )12 (f22 )3h − (f11 )3h (MA )12 − (f12 )3h (MA )22
(C.35)
0 = (f˙22 )1h + 2h|~k|(f22 )2h + i(MH )21 (f12 )0h − i(f21 )0h (MH )12
− 2(MA )22 (f22 )3h − (MA )21 (f12 )3h − (f21 )3h (MA )12 (C.36)
0 = (f˙11 )2h − 2h|~k|(f11 )1h + 2(MH )11 (f11 )3h + (MH )12 (f21 )3h + (f12 )3h (MH )21
+ i(MA )12 (f21 )0h − i(f12 )0h (MA )21 (C.37)
0 = (f˙12 )2h − 2h|~k|(f12 )1h + (MH )11 (f12 )3h + (MH )12 (f22 )3h + (f11 )3h (MH )12
+ (f12 )3h (MH )22 + i(MA )11 (f12 )0h + i(MA )12 (f22 )0h − i(f11 )0h (MA )12 − i(f12 )0h (MA )22
(C.38)
0 = (f˙22 )2h − 2h|~k|(f22 )1h + 2(MH )22 (f22 )3h + (MH )21 (f12 )3h + (f21 )3h (MH )12
+ i(MA )21 (f12 )0h − i(f21 )0h (MA )12 (C.39)
0 = (f˙11 )3h − 2(MH )11 (f11 )2h − (MH )12 (f21 )2h − (f12 )2h (MH )21
+ 2(MA )11 (f11 )1h + (MA )12 (f21 )1h + (f12 )1h (MA )21 (C.40)
0 = (f˙12 )3h − (MH )11 (f12 )2h − (MH )12 (f22 )2h − (f11 )2h (MH )12 − (f12 )2h (MH )22
+ (MA )11 (f12 )1h + (MA )12 (f22 )1h + (f11 )1h (MA )12 + (f12 )1h (MA )22 (C.41)
0 = (f˙22 )3h − 2(MH )22 (f22 )2h − (MH )21 (f12 )2h − (f21 )2h (MH )12
+ 2(MA )22 (f22 )1h + (MA )21 (f12 )1h + (f21 )1h (MA )12 . (C.42)

All these equations couple to each other, which shows how difficult solving them seems to
be. We need to simplify the equations before we can actually work with them.
C.4. FLOW EQUATIONS IN THE DIAGONAL MASS BASIS 131

C.4 Flow Equations in the Diagonal Mass Basis

We can do a rotation in flavor space to simplify (C.30). The exact details are given in
appendix B.1, the important fact is that we can diagonalize a two by two mass matrix
M by rotating it with two matrices UL and UR . The result is Md , a diagonal two by two
matrix:

UL† M UR = Md
UR† M † UL = Md† . (C.43)

From these two equations we can infer

UL† (M M † )UL = Md Md† = |Md |2 (C.44)


UR† (M † M )UR = Md† Md = |Md |2 . (C.45)

We can directly see that both are hermitian expressions. We can write the rotation matrices
UL , UR in the general form

 
cos θ sin θeiα
UL = (C.46)
− sin θe−iα cos θ
 
cos θ̄ sin θ̄eiᾱ
UR = . (C.47)
− sin θ̄e−iᾱ cos θ̄

And we have solved this for θ, θ̄ and α, ᾱ, see for the expressions for these four quantities
and further details appendix B.1.
Next we need to rewrite (C.30) with the help of the following new quantities

frh = f0h + f3h


flh = f0h − f3h
fN h = f1h + if2h . (C.48)

In terms of these new densities we find


132 APPENDIX C. KINETIC DESCRIPTION

0 =f˙rh + iM † fN h − ifN† h M (C.49)


0 =f˙lh + iM f † − ifN h M †
Nh (C.50)
0 =f˙N h − 2i|~k|hfN h + iM frh − iM flh . (C.51)

From the transformations we derived for the mass matrix (B.1) we see that we need to
multiply (C.49) with UR† from the left and UR from the right. For (C.56) we have to
use respectively UL† and UL . For (C.51) we have to use respectively UL† and UR . We
can insert the combinations UR UR† and UL UL† at will, since we are dealing with unitary
transformations. This gives us the following three transformations to the diagonal mass
basis for our newly introduced variables:

d
frh = UR† frh UR (C.52)
d
flh = UL† flh UL (C.53)
fNd h = UL† fN h UR . (C.54)

By comparing these to the transformations for the mass matrix we can see a pattern,
apparently hermitian quantities such as frh , flh and M M † transform with a single trans-
formation matrix. Quantities that are not hermitian transform under a combination of the
two transformation matrices, such as fN h and M in general. This shows why we needed
to rewrite (C.30), the new equations transform to the diagonal mass basis to

0 =f˙rh
d
+ i(M d )† fNd h − i(fNd h )† M d (C.55)
0 =f˙d + iM d (f d )† − if d (M d )†
lh Nh Nh (C.56)
0 =f˙Nd h − 2i|~k|hfNd h + iM frh
d d
− iM flh . (C.57)
Appendix D

Statistical Description of the Density


Matrix

In order to calculate physical quantities in the form of expectation values from these
solutions we have obtained, we can use the density matrix [46]. With the help of the
density matrix it will be easier to calculate expectation values. We will first describe a
density matrix from the very beginning and explain why it is very advantageous for us to
use in this case. In a classical system there is a phase space probability measure, and the
quantum version of this measure is the density matrix ρ̂. Density matrices are normally
introduced for a system which consists out of an ensemble or when the exact preparation of
the system is not known. the density matrix can be written in a general form by introducing
a complete set of states |ψi i

X
ρ̂ = cij |ψi ihψj |. (D.1)
ij

In order
P to fix the trace of the density matrix to unity, as is customary, the condition on
cij is cij = 1. This density matrix can exhibit block diagonal behavior, which is an
indication that the system can be simplified by considering only one block. The density
matrix is a Hermitian object, which has the advantage that we can diagonalize it. After
applying the diagonalization on a general density matrix, we find instead of the general
expression

X
ρ̂ = pi |ψi ihψi |. (D.2)
i

The diagonalization reduced the number of constants involved in the outer product of the
state, however they still sum to unity. We now have an interpretation for the coefficients
pi , they are a measure for probability corresponding to the state with which they are

133
134 APPENDIX D. STATISTICAL DESCRIPTION OF THE DENSITY MATRIX

associated.
With two examples we can show the implications of this density matrix formulation.
Consider a system in a pure state which is given by |ψi (t)i. We know from quantum
mechanics that the expectation value of an operator Ô is given by

hÔ(t)i = hψ(t)|Ô(t)|ψ(t)i. (D.3)

The density matrix is given in this example by the following product

ρ̂(t) = |ψ(t)ihψ(t)|. (D.4)

Since in this pure state formulation there is only a single state, the density matrix also
consists out of only this one part of the sum over all states. Now we can see that equivalently
to taking the expectation value one can trace over the density matrix:

Tr(ρ̂(t)Ô(t)) = hψ(t)|ψ(t)ihψ(t)|Ô(t)|ψ(t)i = hψ(t)|Ô(t)|ψ(t)i = hÔ(t)i. (D.5)

This density matrix obeys two identities which hold for pure states on general:

Tr(ρ̂(t)) = 1
Tr(ρ̂2 (t)) = 1.

For a mixed state, also referred to as an ensemble of quantum states, we can do a similar
analysis. Consider a set of states with which we associate constants cij in the same fashion
as described previously when introducing the density matrix. Now the probabilities that
an operator has a certain expectation value is given by

X
hÔ(t)i = cij hψj (t)|Ô(t)|ψi (t)i. (D.6)
ij

This is again analogous to calculating the trace over the density matrix times the operator
D.1. SPIN EXAMPLE 135

X
Tr(ρ̂(t)Ô(t)) = hψi (t)|ρ̂(t)Ô(t)|ψi (t)i (D.7)
i
X
= cjk hψi (t)|ψj (t)ihψk (t)|Ô(t)|ψi (t)i (D.8)
ijk
X
= cjk hψk (t)|Ô(t)|ψj (t)i (D.9)
jk

= hÔ(t)i. (D.10)

The advantage of this formulation of the density matrix is that it gives us a tool to calculate
an expectation value for an operator which is valid for both pure quantum states and mixed
quantum states, the result will be correct in both cases. The two identities for mixed states
are different however, we can evaluate these by applying the diagonalization transformation,
and in the diagonal basis we see that

X
Tr(ρ̂(t)) = pi = 1 (D.11)
i
X
2
Tr(ρ̂ (t)) = p2i < 1. (D.12)
i

This is not the same as in the case for the pure state, which is a sign that one should be
careful by interpreting the probabilities of a set of quantum states, since the properties can
be different for a set of pure quantum states and a set of mixed quantum states.

D.1 Spin Example


Consider a beam of spin- 21 particles with carry a certain spin. This beam can be represented
by the following two by two density matrix:

 
ρ11 ρ12
ρ= , (D.13)
ρ∗12 ρ22

where we know that ρ11 and ρ22 are real numbers because of the hermiticity condition on
the density matrix. In addition, ρ11 and ρ22 add to one in order to make sure that the
trace is equal to one. This leaves us three degrees of freedom in the density matrix. We
will show how we can relate these three degrees of freedom to the three different axis along
which we can measure the spin.
136 APPENDIX D. STATISTICAL DESCRIPTION OF THE DENSITY MATRIX

First we note that the spin is represented by the three Pauli matrices, and using the theory
in the previous section for the expectation value we see that the expectation value of the
spin is given by

hσi i = Tr(ρσi ) (D.14)

In component form we find for the expectation values of the three spin directions

hσx i = 2 Im(ρ12 ) (D.15)


hσy i = 2 Re(ρ12 ) (D.16)
hσz i = ρ11 − ρ22 (D.17)

In total this means we can express the density matrix as a function of spin expectation
values as
 
1 1 + hσz i hσx i − ihσy i
ρ= . (D.18)
2 hσx i + ihσy i 1 − hσz i

If we rotate the coordinate system of this problem to frame where the spin is in the ẑ 0
direction, we find
 
1 1 + hσz0 i 0
ρ= . (D.19)
2 0 1 − hσz0 i

In this frame there are no off diagonal terms. The difference between a pure and a mixed
state is now obvious, when the spin is either +1 or −1 the state is pure, otherwise the state
is mixed. If there is no polarization at all, hσz0 i = 0 and the state is maximally mixed.
Bibliography

[1] C. Itzykson and J.B. Zuber, Quantum Field Theory, 1980, New York, McGraw-Hill
[2] R.N. Mohapatra and P.B. Pal, Massive Neutrinos in Physics and Astrophysics, 2004,
Singapore, World Scientific, 3rd Ed.
[3] I. S. Gradshteyn and I. M. Ryzhik, Table of integrals, series and products, corrected
and enlarged edition, 1980, Academic Press, New York
[4] B. Pontecorvo, JETP 6, 429, 1957
[5] E.K. Akhmedov, ICTP Summer School 7 Jun-9 Jul, Trieste 1999
[6] C.Giunti, C.W. Kim and W. Lam, Physical Review D 44, 3635, 1991
[7] W. Grimus, P. Stockinger and S.Mohanty, Physical Review D 44, 013011, 1999
[8] S.M. Bilenky and B. Pontecorvo, Physical Reports 41, 225, 1978
[9] M.E. Peskin and D.V. Schroeder, An Introduction To Quantum Field Theory, 1995,
Westview Press
[10] T. Araki et al. [KamLAND Collaboration], Physical Review Letters 94, 8 081801-
081806, 2005
[11] M.C. Gonzalez-Garcia and Y. Nir, Rev. Mod. Phys. 75, 2003, 345
[12] A. Aguilar-Arevalo et al. [MiniBooNE Collaboration], Physical Review Letters 98,
2007, 231801.
[13] S. Bilenky, Introduction to the Physics of Massive and Mixed Neutrinos, 2010,
Springer-Verlag Berlin Heidelberg
[14] D.N. Spergel et al. [WMAP Coll.], arXiv.org:0603449 [astro-ph]
[15] A. Goobar, S. Hannestad, E. Mortsell and H. Tu, astro-ph/0602155.
[16] J. Lesgourgues, S. Pastor, Phys.Rept. 429, 2006, 307-379
[17] H.V. Klapdor-Kleingrothaus and I.V. Krivosheina, Modern Physics Letters A 21, 20,
1547-1566, 2006

137
138 BIBLIOGRAPHY

[18] C.E. Aalseth et al., Modern Physics Letters A 17, 1475-1478, 2002
[19] Super-Kamiokande Collaboration, Physical Review D 71, 112005, 2005
[20] SNO Collaboration, Physical Review Letters 101, 111301, 2008
[21] J. Koksma and T. Prokopec, Class. Quant. Grav. 26, 125003, 2009
[22] J. Koksma, T. Prokopec and M. Schmidt, arXiv:1012.3701v1 [quant-ph], 2010
[23] C. Zachos, D. Fairlie, and T. Curtright, Quantum Mechanics in Phase Space, 2005,
World Scientific, Singapore
[24] T. Prokopec, M.G. Schmidt and S. Weinstock, Annals Phys. 314, 208-265, 2004
[25] B. Garbrecht, T. Prokopec and M.S. Schmidt, Phys.Rev.Lett. 92, 061303, 2004
[26] T. Konstandin, T. Prokopec and M.S. Schmidt, Nucl.Phys. B 716, 373-400, 2005
[27] P. Langacker, Neutrino mass, Testing the Standard Model, Proceedings of the 1990
theoretical advanced study institute in elementary particle physics
[28] G. Senjanovic and R. Mohapatra, Phys. Rev. D 23, 165180, 1981
[29] K. Case, Phys. Rev., 107-1, 1957
[30] G. Scharf, Finite Quantum Electrodynamics, 1989, Springer-Verlag Berlin Heidelberg
[31] S. Eidelman et al., Physics Letters B 592, 2008
[32] B. Aharmim et al., Physical Review C 72, 055502, 2005
[33] Y. Ashie at al., Physical Review Letters 93, 101801, 2004
[34] P. Adamson et al, Physical Review Letters 101, 131802, 2008
[35] A. Aste, Symmetry 2, 1776-1809, 2010
[36] T. Prokopec, J. Weenink and M. Schmidt, The entropy of free fermionic systems, in
preparation
[37] D. Spergel et al. [WMAP Coll.], arXiv.org:0302209 [astro-ph]
[38] Z. Maki et al., Progress of Theoretical Physics, 28, 870, 1962
[39] R. de Putter et al., arXiv:1201.1909v1 [astro-ph], 2012
[40] G. Fogli et al., arXiv:1106.6028v2 [hep-ph], 2011
[41] L. Krauss et al., arXiv:1009.4666 [hep-ph], 2011, submitted to Physical Review D
[42] P. Serpico et al., Physical Review D, 71, 127301, 2005
[43] S. Pastor et al., Physical Letters, 102, 241302, 2009
[44] K. Abazajian et al., Physical Review D, 66, 013008, 2002
BIBLIOGRAPHY 139

[45] G. Raffelt, Stars as Laboratories for Fundamental Physics The Astrophysics of Neu-
trinos, Axions, and Other Weakly Interacting Particles, 1996, University of Chicago
Press
[46] K. Blum, Density Matrix Theory and Applications, 1981, Plenum Press

Das könnte Ihnen auch gefallen