The Inflationary Universe

Bachelor of Science Thesis in Physics, 15 credits

Author:
Benjamin Cavcic

Supervisor:
Daniel Panizo Pérez

Subject reader:
Ulf Danielsson

June 12, 2023

Department of Physics and Astronomy

Division of Theoretical Physics
Uppsala University
 Abstract

Astrophysical observations of the cosmic microwave background point to in-

consistencies in the standard model of cosmology, and a primordial accelerated
expansion of the universe known as inflation has been suggested as a solution.
Unfortunately, observational evidence of inflation is lacking, and there exists
hundreds of models that populate the inflationary landscape. In this thesis,
we explore three of these and see what constraints are set on them in order
to account for observations. We find that two of the models have regimes of
trans-planckian nature, while the third leads to a non-invertible equation.

Sammanfattning

Astronomiska observationer av den kosmiska bakgrundsstrålningen tyder

på bristfälligheter i standardkosmologin, vilket har lett till förslaget om en ac-
celererande expansion i de tidigaste skeden av universum känd som inflation.
I avsaknaden av observationella bevis finns det numera hundratals inflation-
smodeller, och i detta arbete kommer vi att rikta fokuset mot tre av dessa av
vilka två visar sig överstiga transplanckianska värden medan den sista leder
till en ekvation som inte är inverterbar.

ii
 In other words, the so-called involuntary circulation of your
blood is one continuous process with the stars shining. If you find
out it’s you who circulates your blood, you will at the same mo-
ment find out that you are shining the sun. Because your physical
organism is one continuous process with everything else that’s go-
ing on. Just as the waves are continuous with the ocean, your body
is continuous with the total energy system of the cosmos, and it’s
all you.
Alan Watts

iii
Contents
1 Introduction 1

2 Review 2
2.1 Tensor Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Flat Spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 General Relativity 4
3.1 Covariant Derivatives and Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Parallel Transport and Geodesic Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Energy-momentum Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3.1 Perfect Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4 Riemann Tensor, Ricci Tensor and Ricci Scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.5 Bianchi Identities and Einstein Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.6 Einstein Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.7 Einstein-Hilbert Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4 Cosmology 19
4.1 FLRW Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Friedmann Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.1 Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.3 Model Universes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3.1 The Universe Without a Cosmological Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3.2 Flat Universe Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.4 Standard Cosmology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.1 Hot Big Bang Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.2 ΛCDM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4.3 Hubble Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5 Introduction to Inflation 32
5.1 The Horizon Problem and Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 The Inflaton Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3 Slow-roll Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

6 Case Studies 41
6.1 Chaotic Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.1.1 The Vacuum Catastrophe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.2 Natural Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.3 Valley Hybrid Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

7 Conclusion 47
7.1 Ethical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

A Mathematical Derivations 49

B Tensor Components for the FLRW Metric 51

iv
 Bachelor of Science Thesis in Physics B. Cavcic

1 Introduction
Below is a picture of the cosmic microwave background (CMB), a snapshot of the universe when it was
380,000 years old. Prior to this moment, photons were locked to the dense plasma of baryons and unable
to travel freely through space. Later, the primordial plasma was gradually cooled by the expansion of
the universe, and at some moment photons were freed from baryons, marking the transition from an
opaque to a transparent universe. The instant of baryon-photon decoupling is forever imprinted on the
cosmic microwave background, and mappings of the CMB reveal a spectularly uniform universe, with
temperature fluctuations being in the order of only 0.1 microkelvins [1]. This is in direct violation of the
standard model of cosmology; at the time of the CMB, the universe was simply too young for all points
to establish uniformity to such a degree [2]. This has lead to the theory of inflation, a short-lived period
of accelerated expansion in the early universe that solves the so-called horizon problem of cosmology.

Figure 1: The cosmic microwave background, as mapped following a nine-year data run by WMAP [3].

Naturally, one would like to know whether the theory of inflation is correct, and if so, find an ap-
propriate model that accounts for astrophysical observations. Indeed, there exists hundreds of different
inflationary scenarios, and ultimately observations need to constrain these to a single option. Luckily,
testing the theory of inflation is a more attainable task than ever, thanks to highly accurate space ob-
servatories such as the Planck satellite and WMAP1 . However, convincing evidence of inflation is yet
to be found, and as such the inflationary landscape remains populated by a rich amount of hypothetical
models.
In this thesis, we will introduce the classical theory of inflation, and explore three different models
to see what constraints they set on the inflaton, the hypothetical field giving rise to inflation. At the
1
Wilkinson Microwave Anistropy Probe.

1
 Bachelor of Science Thesis in Physics B. Cavcic

same time, we aim for a text that is readable at an undergraduate level, where essential topics such as
general relativity and standard cosmology are usually yet to be introduced. Therefore, a substantial
part of the thesis (§3 and §4) will be devoted to explaining these theories in fine detail, only later giving
way to the topic of inflation.
The thesis can be summarized as an attempt to answer the following questions:

1 What is the standard model of cosmology and how was it developed?

2 What is the horizon problem in standard cosmology, and how is it solved by inflation?

3 In the three models that will be explored (chaotic inflation, natural inflation and valley hybrid
inflation), what constraints are set on the inflaton field?

By knowing the answers to these questions, the reader will be well-versed in the language of physical
cosmology, which includes terminology such as the Friedmann-Robertson-Lemaître metric, the Fried-
mann equations, standard cosmology, the multiverse and fine-tuning. Additionally, it will set the stage
for a quantum-mechanical treatment of inflation, where minute quantum perturbations lay the seed for
the large-scale structure of the universe.

2 Review
Before introducing general relativity, we will have to briefly review the notion of a tensor and the
Minkowski metric.

2.1 Tensor Algebra

In differential geometry, tensors are defined through vectors and dual vector spaces; if V is a vector
space, it has a corresponding dual vector space usually denoted V∗ . A dual vector, or one-form, is an
element of the dual vector space, and a map such that [4]

ω(aV + bW ) = aω(V ) + bω(W ) ∈ R, (2.1)

where a, b ∈ R and V, W ∈ V.
In other words, a dual vector is a linear map from vectors to real numbers. Similarly, tensors map
vectors and dual vectors to real numbers; let V be a vector space and V∗ a corresponding dual vector
space. Then, a tensor T of rank (m, n) is a multilinear map such that [5]

T : V∗ ⊗ · · · ⊗ V∗ ⊗ V ⊗ · · · ⊗ V → R, (2.2)
| {z } | {z }
m copies n copies

2
 Bachelor of Science Thesis in Physics B. Cavcic

where the symbol ⊗ denotes the cartesian product. Tensors are generalizations of vectors; a type (1,0)
tensor is a regular vector, whereas a type (0,1) tensor is a dual vector.
Any vector V ∈ V and dual vector V ∗ ∈ V∗ can be decomposed into a linear combination of basis
vectors,
n
X n
X
∗
V = µ
V êµ =: V êµ , µ
V = Vµ∗ θ̂µ =: Vµ∗ θ̂µ , (2.3)
µ=0 µ=0

where n is the dimension of both vector spaces, V µ , Vµ∗ the components of the (dual) vector, and êµ , θ̂µ
basis vectors. Notice how we defined a shorthand notation for the sum. This is known as the Einstein
summation convention [6], and is used extensively in general relativity. Due to this convention, it is also
common to loosely denote a given vector V as simply V µ , in which case it is known as a contravariant
vector. On the other hand, dual vectors are denoted as Vµ and are referred to as covariant vectors [4].
In light of this new notation, we can introduce the transformation law for tensors [4],
′ ′
′ ′ ∂xµ1 ∂xµn ∂xν1 ∂xνm µ1 ...µn
T µ1 ...µn ν1′ ...νm
′ = . . . ′ . . . ′ T ν1 ...νm . (2.4)
∂xµ1 ∂xµn ∂xν1 ∂xνm
There is an important operation for tensors known as contraction: if a tensor has the same index up and
down, it may be implicitly summed over according to the Einstein summation convention, yielding a
new tensor. For example,
Aρµρ = Aµ . (2.5)
Although the index ρ appears on the left hand side, it is implicitly summed over and therefore does not
need to be included on the right hand side. The contracted version Aµ is a perfectly valid tensor since it
satisfies the tensor transformation law. On the contrary, contraction with two up or two down indices
is not permissible, as the resulting tensor does not transform tensorially.
Contraction may also occur when multiplying two tensors, such as in the following case:

C ν µ = Aνρ Bρµ . (2.6)

2.2 Flat Spacetime

In special relativity, it is postulated that the laws of physics are the same in all inertial reference frames
and that the speed of light is constant and independent of relative motion to the source. This leads to

3
 Bachelor of Science Thesis in Physics B. Cavcic

the Lorentz transformations [7] from a frame S to another frame S ′ ,

′

 t = γ(x − vt),
x′ = γ(t − vx/c2 ),

(2.7)


 y ′ = y,
 ′
z = z,

where v is the velocity of S ′ relative to S, c the speed of light and γ := (1 − v 2 /c2 )−1/2 . The Lorentz
transformations imply the following invariant quantity:

ds2 = −dt2 + dx2 + dy 2 + dz 2 = ηµν dxµ dxν , (2.8)

where ηµν = diag(−1, 1, 1, 1) is known as the Minkowski metric. It is the metric for flat spacetime, and
can be used to raise/lower tensor indices, for example

ηµν Aν = Aµ . (2.9)

3 General Relativity
Out of the four fundamental forces, gravity is the most dominant at the largest scales of the universe.
Therefore, the study of cosmology is heavily reliant on the underlying gravitational theory that is being
used. Currently, the best description of gravity is general relativity (GR), a theory developed by Albert
Einstein in the early 20th century. It is considered to be the successor of Newtonian gravity, which
has already served as a powerful theory for describing a variety of phenomena. The necessity for an
extended theory, however, became apparent with incoming observations that directly disagreed with
the Newtonian picture. An important example is the peculiar precession of planet Mercury’s orbit
around the sun; in 1859, Le Verrier published a paper detailing the precession of the perihelion of
Mercury’s orbit, and found that the results were inconsistent with those predicted by Newton’s theory
[8]. Another example is the deflection of light around massive objects, which can not be appropriately
accounted for by Newtonian gravity as the theoretical calculations are off by a factor of two [9]. On the
other hand, Einstein’s theory not only manages to explain these anomalies, but also introduces aspects
of gravity that are not contained in Newton’s theory. Some of these include the notion of curvature,
which is one of the key elements of physical cosmology.
To see how GR differs from Newtonian gravity, we start off with Newton’s universal law of gravi-
tation in the form of Poisson’s equation [4],

∇ · g = 4πGρ, (3.1)

4
 Bachelor of Science Thesis in Physics B. Cavcic

where g is the gravitational field, G a constant and ρ the mass density. We interpret the equation as
follows: mass density generates a gravitational field, just like electric charge density creates an electric
field. Within the Newtonian framework, the preceding statement is perfectly valid. However, the need
for a more intrinsic description was made apparent by Einstein, as we will now see.
Assume that any massive object has two types of masses, an inertial mass mi and a gravitational
mass mg , with the gravitational force acting only on the latter. According to Newton’s second law, we
have
Fg = mg g = mi a. (3.2)
In principle, mi need not equal mg . An immediate consequence if they were equal, however, is that
all objects would fall at the same rate regardless of their mass, which becomes apparent when setting
mi = mg in the previous equation:
g = a. (3.3)
In experiments with balls rolling down inclined planes, Galileo Galilei managed to show that objects
do indeed fall at the same rate [4], validating the above equation. Therefore, we are brought to the
conclusion that the motion of free-falling objects is the same in a gravitational field as in uniformly
accelerated reference frames. This is known as the weak equivalence principle (WEP) [4].
In essence, the weak equivalence principle is simply the statement that mi = mg . However, special
relativity requires a statement a more general description; according to the WEP, the sum of the inertial
(i) (i) (g) (g)
and gravitational mass of a proton and electron should be equal, i.e. mp + mn = mp + mn . A
hydrogen atom, which is a composition of a proton and electron, has an inertial mass smaller than this
due to the fact that some of the rest energy of the two atoms is converted into binding energy. In that
case, the WEP fails since
m(i) (i) (g)
p + mn − mbind ̸= mp + mn .
(g)
(3.4)
In spite of this, the hydrogen atom falls at the same rate as protons and electrons, hence the inertial
and gravitational masses must be the same. The only way to make such an equivalence true again, is
to require that the gravitational field couples to electromagnetism in such a way that the gravitational
mass equals the left hand side of the equation. This lead Einstein to formulate a generalized version of
the WEP:

Einstein Equivalence Principle (EEP)

In local regions of spacetime, the laws of physics reduce to those of special relativity, and no
experiment involving gravity can be conducted to detect a gravitational field.

Note the addition of the word local; indeed, the equivalence principle only makes sense in local
regions of spacetime, since inhomogeneities in the gravitational field lead to tidal forces. The earth’s
gravitational field, for example, is directed radially towards the center and the radial motion of free-

5
 Bachelor of Science Thesis in Physics B. Cavcic

falling particles will be apparent if they are separated by a significant distance.

The EEP matters because it suggests that there is no such thing as a gravitational field in local
spacetime. Instead, Einstein suggested that gravity manifests in the curvature of spacetime, caused by
the energy density of matter as described by the Einstein field equations,

8πG
Gµν = Tµν .
c4
This is the GR equivalent of Newton’s law of gravity (3.1). The left term, as we will see, describes the
curvature of spacetime while the right term is related to the matter energy-density.
The purpose of this section is to understand what the left and right hand sides of Einstein’s equations
mean, as we will later want to solve Einstein’s equations for the whole universe. To begin with, we will
look at vectors and how they transform in curved spaces, which amounts to sections §3.1 and §3.2.
In §3.3, the energy-momentum tensor Tµν that appears on the right side of Einstein’s equation will be
described in physical terms and an explicit derivation for a perfect fluid will be shown in §3.3.1. Later in
§3.4-§3.6 we will use the concepts from previous sections to relate curvature to matter, which will lead
to the Einstein field equations. Finally, in §3.7 we will present the Lagrangian formulation of general
relativity.

3.1 Covariant Derivatives and Connections

At its heart, general relativity is a theory that extends the notion of flat spacetime to curved. In this
sense, it is the generalization of special relativity. It is therefore necessary to describe curvature math-
ematically, and the first step towards doing so is to consider how vectors change in curved coordinates.
In flat geometry, partial derivatives of vectors transform as tensors. This is no longer the case for
′
curved spaces. To see this, consider a vector V µ in a frame S with basis xµ , and another vector V µ in
′ ′ ′
frame a S ′ with basis xµ = xµ (xµ ). The partial derivative of V µ in S ′ is given by
 ν′   ν′ 
∂ ν′ ∂ ∂x λ ∂xσ ∂ ∂x
µ′V = µ′ λ
V = µ ′ σ
Vλ
∂x ∂x ∂x ∂x ∂x ∂xλ
 2 ν′ (3.5)
ν′ 2 ν′ ν′
∂xσ λ σ σ λ

∂ x ∂x ∂V ∂x ∂ x ∂x ∂x ∂V
= Vλ+ = V λ + µ′ ,
∂xµ′ ∂xσ ∂xλ ∂xλ ∂xσ ∂x′µ ∂xσ ∂xλ ∂x ∂xλ ∂xσ

where we used the transformation law for tensors in the first equality and the chain rule in the second.
In flat space, the second partial derivative in the first term of the last equality vanishes. On the
other hand, this term is generally non-zero for curved spaces. Therefore, to make the vector transform
as a tensor, we need to get rid of that first term. We propose the simplest solution: a new differential

6
 Bachelor of Science Thesis in Physics B. Cavcic

operator consisting of a regular partial derivative plus a correction term,

∇µ V ν := (∂µ + Γ)V ν (3.6)

To preserve free indices, we need Γ to atleast include a covariant (down) index. It will, however, prove
useful to have a pair of covariant indices. In that case, we obtain the following construction:

∇µ V ν := ∂µ V ν + Γνµλ V λ . (3.7)

The operator ∇µ is known as the covariant derivative [4]. If we want (3.7) to be a tensor, we need to
know what constraints it sets on the extra term Γνµλ . First of all, we require that the covariant derivative
of a vector ∇µ V ν obeys the transformation law for tensors (2.4),
′ ′ ′
ν′ ∂xµ ∂xν ν ∂xµ ∂xν ν ∂xµ ∂xν ν λ
∇ V
µ′ = ∇µ V = ∂µ V + µ′ Γ V . (3.8)
∂xµ′ ∂xν ∂xµ′ ∂xν ∂x ∂xν µλ
On the other hand, each term of the covariant derivative must transform individually,
 ν′ 
ν′ ν′ ν′ λ′ ∂ ∂x ν ν′ λ′
∇µ′ V = ∂µ′ V + Γµ′ λ′ V = ′ V + Γ µ ′ λ′ V
∂xµ ∂xν
(3.9)
′ ′ ′
∂xµ ∂ 2 xν ν ∂xµ ∂xν ν ν ′ ∂x
λ
= V + ∂ µ V + Γ µλ′ ′ V λ.
∂xµ′ ∂xµ ∂xν ∂xµ′ ∂xν ∂xλ
′
In the last step, we applied the chain rule, ∂µ′ = (∂xµ /∂xµ )∂µ . Equating (3.8) and (3.9), we find
′ ′ ′
∂xµ ∂xν ν λ ∂xµ ∂ 2 xν ν ′ ∂x
λ

µ′ ν
Γµλ V = µ′ µ ν
V + Γµ′ λ′ λ V λ .
ν
(3.10)
∂x ∂x ∂x ∂x ∂x ∂x
′
We can now multiply by ∂xλ /∂xλ on both sides and let ν → λ on the first term on the right hand side
to get
′ ′
∂xλ ∂xµ ∂xν ν λ ∂xλ ∂xµ ∂ 2 xν ′

λ ′ µ′ ν
Γµλ V = λ ′ µ ′ µ λ
V λ + Γνµ′ λ′ V λ . (3.11)
∂x ∂x ∂x ∂x ∂x ∂x ∂x
Requiring that the expression must hold true for all V λ , we obtain a transformation law,

′ ′
′ ∂xλ ∂xµ ∂xν ν ∂xλ ∂xµ ∂ 2 xν
Γνµ′ λ′ = Γ − . (3.12)
∂xλ ∂xµ ∂xν µλ ∂xλ′ ∂xµ′ ∂xµ ∂xλ
′ ′

7
 Bachelor of Science Thesis in Physics B. Cavcic

Generally, Γλµν is known as a connection, and it tracks the change of the basis vectors when a vector is
moved. Notice, however, that we did not obtain a general expression for the connection; the covariant
derivative of a vector is a tensor as long as the relation above holds true. On the other hand, an ex-
plicit definition in terms of the metric can be obtained if two additional properties are forced upon the
connection [4]:
1 torsion-free: Γρµν = Γρνµ ,

2 metric compatibility: ∇ρ gµν = 0.

The meaning of a torsion-free connection will be explained in §3.4, while metric compatibility is more of
an assumption that will first be exploited in §3.7. The above restrictions lead to the Christoffel connection,
given by [4]
1
Γρµν = g ρσ (∂µ gνσ + ∂ν gσµ − ∂σ gµν ) . (3.13)
2
The associated components of the Christoffel connection are known as Christoffel symbols. Although
the final results will be dependent on the Christoffel connection, we will assume an arbitrary connection
for now, as it will lead to more general results than if the Christoffel connection is immediately assumed.
For an arbitrary (n,m) tensor, the covariant derivative is given by (see Appendix A)

∇ρ T µ1 µ2 ...µn ν1 ν2 ...νm
. (3.14)
= ∂ρ T µ1 µ2 ...µn ν1 ν2 ...νm + Γµρσ1 T σµ2 ...µn ν1 ν2 ...νm + · · · − Γσρν1 T µ1 µ2 ...µn σν2 ...νm − . . .

Especially relevant is the case of taking two covariant derivatives of a vector; in that case, we are taking
the covariant derivative of a (1,1) tensor, which leads to

∇µ ∇ν V ρ = ∇µ (∇ν V ρ ) = ∂µ (∇ν V ρ ) + Γρµσ (∇ν V σ ) − Γλµν (∇λ V ρ ). (3.15)

In curved spaces, the covariant derivate puts the partial derivative out of business in describing in-
finitesimal changes of tensors. In the next section, we will see how vectors, or equivalently rank 1
tensors, behave differently in curved spaces when being changed by a small amount.

3.2 Parallel Transport and Geodesic Equation

In flat space, vectors can be be moved around freely, allowing them to be subtracted, added and multi-
plied with each other. This is possible because vectors in flat space are only defined by the displacement
of two points, see Figure 2 for an illustrative example. Of course, vectors can be moved around under
the condition that they remain parallel; rotating the vector, for example, transforms the components
into a different vector. The act of moving a vector while keeping it parallel is known as parallel transport
[6].

8
 Bachelor of Science Thesis in Physics B. Cavcic

3 ∆y

2 ∆y
∆x
1
∆x
x
1 2 3 4 5

Figure 2: In flat space, vectors are defined by the displacement of two points. As demonstrated in the
figure, the vector (1,2,0) can be obtained by the subtraction of points (2,3,0) and (1,1,0), or (5,4,0) and (4,2,0),
or any other subtraction of points that equal the vector (∆x, ∆y). Therefore, vectors can be transported
freely in flat space as long as they remain parallel.

In curved spaces, transporting vectors is complicated by the fact that they transform when moved
along some path, see Figure 3. Mathematically, this is because the basis vectors themselves are changing.
Two vectors at two different points on a curved space can therefore not be compared, hence vector
operations like addition and subtraction are not well-defined. The only time when such notions can be
applied is in a local neighbourhood of a point on the curved space, since the curvature looks flat near
this point. Therefore, we may perform flat-space operations on vectors like addition and subtraction as
long as they are constrained to a local neighbourhood of the point. With this in mind, the best we can
do to mimick parallel transport in non-flat spaces is to demand that a local (i.e. infinitesimal) change of
a vector along some path does not alter its components. Formally, a vector V µ is defined to be parallel

Figure 3: Left: vectors do not change direction when parallel transported along flat spaces. Right: vectors
do change direction when parallel transported along curved spaces.

9
 Bachelor of Science Thesis in Physics B. Cavcic

transported along a curve γ when its covariant derivative is zero along that path, i.e.

∇γ V µ = 0 . (3.16)

Having extended the notion of parallel transport to curved spaces, we can easily derive the equation for
a geodesic, the shortest path between two points. In flat space, this is intuitively known to be a straight
line, but on a curved surface like a sphere they are great circles. Since space looks locally flat on a curved
surface, a geodesic can be traced if you promise to always locally move in the same direction. Therefore,
a geodesic is defined as follows: let γ: xµ (λ) be some path parametrized by λ. The tangent vector to
this path is then given by dxµ /dλ. The path γ is a geodesic if the tangent is parallel transported along
γ,
dxµ dxµ dxµ
∇γ = ∇µ = 0. (3.17)
dλ dλ dλ
Inserting the definition of covariant derivative (3.9), we get

d 2 xµ ρ
µ dx dx
σ
+ Γ ρσ =0 . (3.18)
dλ2 dλ dλ

This is known as the geodesic equation. It does not only give the shortest path on a curved surface,
but is also a powerful equation for finding Christoffel symbols instead of using the explicit expression
involving the metric (3.13). See Appendix B for more details.

3.3 Energy-momentum Tensor

The Poisson equation in Newtonian gravity (3.1) suggests that the presence of a mass density generates
a gravitational field. Somehow, this notion needs to mimicked in general relativity, but in terms of
energy and not mass. The correct object that replaces the mass density ρ in Poisson’s equation is the
energy-momentum (EM) tensor Tµν , defined as follows [6]:

Tµν := Flux of 4-momentum pµ through a surface of constant xν . (3.19)

The reason for defining the EM tensor in terms of momentum and not energy becomes apparent when
checking the units:
momentum kg · m/s J
∼ 2
= 3 = Pa. (3.20)
time · area s·m m
This means that each component of the energy-momentum tensor can be interpreted as either momen-
tum flux, energy density or pressure. For example, the component T11 indicates a momentum flux in the

10
 Bachelor of Science Thesis in Physics B. Cavcic

 
T00 T01 T02 T03
T T11 T12 T13 
 10 
 T20 T21 T22 T23
 

T30 T31 T32 T33

Figure 4: Element interpretation of the energy-momentum tensor. In red: energy density, green: energy
“flow“ in a particular direction, orange: pressure, purple: stress.

x-direction through the yz plane, i.e. flux orthogonal to the surface, which is understood as pressure.
For an off-diagonal term like T12 , there is a momentum flux in the x-direction through the xz-plane,
which can be interpreted as shear stress rather than pressure. As for the components that include any
0 index, such as T00 or T01 , the interpretation changes; from special relativity, we know that the p0
component of the 4-momentum is equal to E/c2 , where E is the total energy. Hence, T00 is equal to the
energy density. On the other hand, off-diagonal terms in the EM tensor that include a 0 index such as T01
are interpreted as a flow of energy in a particular direction. For a complete summary of all components,
see Figure 4.
There are two important properties of the energy-momentum tensor:

1 symmetry: Tµν = Tνµ .

2 conservation of energy: ∇ρ Tµν = 0.

For a further discussion on these properties, see [4] or [6].

3.3.1 Perfect Fluid

In cosmology, matter is commonly assumed to take the shape of a perfect fluid, meaning it has no shear
stresses, viscosity or heat conduction and is completely isotropic. Owing to the previous discussion of
the energy-momentum tensor, we conclude that all off-diagonal elements in the EM tensor must then
be zero and the pressure equal in all three spatial directions. This leads to
 
ρ 0 0 0
 
0 P 0 0
Tµν =
0
, (3.21)
 0 P 0
0 0 0 P

11
 Bachelor of Science Thesis in Physics B. Cavcic

ρ
V
∇ν∇µ
∇ν
[∇µ , ∇ν ]V ρ

∇µ ∇
∇µ νV ρ

∇µ

∇ν
Figure 5: The commutator of two covariant derivates, visualized. As the vector V ρ takes two opposite
paths, it transforms differently. The displacement is given by the commutator, and increases with stronger
curvature.

where ρ is the energy-density and P the isotropic pressure, both as measured in the frame that is moving
with the fluid. Since we are in the rest frame of the fluid, we have U µ = (1, 0, 0, 0). In flat space, the
matrix can be rewritten in the following tensorial form:

Tµν = (ρ + P )Uµ Uν + P ηµν , (3.22)

where ηµν is the Minkowski metric with signature (−, +, +, +). To obtain the EM tensor for curved
spaces, we replace ηµν with a general metric gµν ,

Tµν = (ρ + P )Uµ Uν + P gµν . (3.23)

3.4 Riemann Tensor, Ricci Tensor and Ricci Scalar

How is curvature characterized and quantified? In the previous section, we saw that the transforma-
tion of a parallel-transported vector is path-dependent in curved spaces. This can be used to describe
curvature.
Consider a vector first transported along a direction xµ , then xν . Then let the same vector take
the opposite path, i.e. xν then xµ . If the space is curved, the result will give two vectors pointing
in different directions, see Figure 5. The larger the curvature, the greater the seperation between the

12
 Bachelor of Science Thesis in Physics B. Cavcic

vectors. Therefore, curvature can be characterized by the commutator of two covariant derivatives,

[∇µ , ∇ν ]V ρ = (∇µ ∇ν − ∇ν ∇µ )V ρ = ∇µ (∇ν V ρ ) − ∇ν (∇µ V ρ )

= ∂µ (∇ν V ρ ) + Γρµσ (∇ν V σ ) − Γλµν (∇λ V ρ ) − (µ ↔ ν)

(3.24)
= ∂µ ∂ν V ρ + ∂µ (Γρνσ V σ ) + Γρµσ (∂ν V σ ) + Γρµσ Γσνλ V λ

− Γλµν (∂λ V ρ ) − Γλµν Γρµσ V σ − (µ ↔ ν) ,

where (µ ↔ ν) stands for all preceding terms, but with switched indices. Writing out all terms, we find

[∇µ , ∇ν ]V ρ =

= ∂µ ∂ν V ρ + ∂µ (Γρνσ ) V σ + Γρνσ ∂µ V σ + Γρµσ ∂ν V σ + Γρµσ Γσνλ V λ − Γλµν ∂λ V ρ − Γλµν Γρλσ V σ

− ∂ν ∂µ V ρ − ∂ν (Γρµσ )V σ − Γρσµ ∂ν V σ − Γρσν ∂µ V σ − Γρνσ Γσµλ V λ + Γλνµ ∂λ V ρ + Γλνµ Γρλσ V σ

= ∂µ Γρνσ − ∂ν Γρµσ + Γρµλ Γλνσ − Γρνλ Γλµσ V σ − 2Γσ[µν] ∇σ V ρ = Rσµν ρ

V σ − Tµνσ
∇σ V ρ ,


(3.25)

where T is known as the torsion tensor [4], defined as

σ
Tµν := 2(Γσµν − Γσνµ ), (3.26)

and
ρ
Rσµν := ∂µ Γρνσ − ∂ν Γρµσ + Γρµλ Γλνσ − Γρνλ Γλµσ , (3.27)

is the Riemann tensor [4].

In general relativity, it is most common to work with the Christoffel connection, which assumes
σ
torsion-free spaces (see §3.1). “Torsion-free“ means that Tµν = 0, which according to (3.26) is equivalent
to having a symmetric connection,
Γρµν = Γρνµ . (3.28)
With no torsion, (3.25) reduces to
[∇µ , ∇ν ]V ρ = Rσµν
ρ
V ρ. (3.29)
Assuming the Christoffel connection (3.13), Three symmetries can be found in the Riemann tensor [4],
two of which are antisymmetries,

Rρσµν = −Rρσνµ , (3.30)

Rρσµν = −Rσρµν , (3.31)

13
 Bachelor of Science Thesis in Physics B. Cavcic

and skew symmetry

Rρσµν = Rµνρσ . (3.32)
The first symmetry can be easily seen since the Riemann tensor is the commutator if there is no torsion.
Symmetries like these reduce the amount of independent components of the Riemann tensor, which will
immediately prove useful.

Ricci Tensor and Ricci Scalar

The Riemann tensor, being of rank 4, is incompatible with the rank 2 energy-momentum tensor. This
is problematic if we want to relate matter to curvature, and the only way of reconciling this fact is to
contract the Riemann tensor to a rank 2 tensor. We therefore define the Ricci tensor [4],

ρ
Rµν := Rµρν . (3.33)

The choice of contracting the second covariant index may seem arbitrary, but if the Riemann tensor
is equipped with the Christoffel connection, we can use the symmetries of the Riemann tensor (3.30)-
(3.32) to infer that there is only one independent contraction. Contracting with the first covariant index
would give a vanishing Ricci tensor, whereas contraction with the third covariant index is related to the
definition above in the following way:
ρ ρ
Rσµρ = −Rσρµ = −Rσµ . (3.34)

Hence, there is only one independent contraction of the Riemann tensor.

Although we have reduced the Riemann tensor to second order, it is not enough to directly relate
the Ricci tensor to the energy-momentum tensor. For example, the expression Rµν = Tµν is not valid,
since taking the covariant derivative on both sides would give ∇ρ Rµν = 0, which is generally not true.
It is therefore necessary to find an expression with the Ricci tensor that vanishes when differentiated.
To this end, we will have to define another important object, the Ricci scalar, defined as the trace of the
Ricci tensor,
R := g µν Rµν . (3.35)

The Ricci scalar is a measure of curvature in the given metric; it is positive for closed curvature such as
a sphere, negative for open, saddle-like curvatures and zero for flat space. For example, the Ricci scalar
for a sphere is 2/r2 , which is strictly positive and vanishing for increasing radius, coinciding with the
fact that a sphere appears flatter the larger it gets for an observer on the surface.

14
 Bachelor of Science Thesis in Physics B. Cavcic

a
∇e Rbcd
c e
Figure 6: Cyclic permutation of indices in the covariant derivative of the Riemann tensor.

3.5 Bianchi Identities and Einstein Tensor

In the previous section, we hinted at the need for an expression that involves the Ricci tensor, but
vanishes upon differentiation in order to conserve energy. A pair of such expressions exist, known as
the Bianchi identities.
To begin with, we take note of the fact that at any point in space, it is possible to define locally inertial
coordinates for which the first derivatives of the metric vanish2 . This is because space looks locally flat,
and we can therefore define flat coordinates at any point. If we assume the Christoffel connection, the
Christoffel symbols will also be zero according to (3.13). This can also be understood intuitively as the
connection tracks the change of basis vectors, which are not altered in flat space. The Riemann tensor
then reduces to
ρ
Rσµν = ∂µ Γρνσ − ∂ν Γρµσ . (3.36)
Note that although the Christoffel symbols vanish, the derivatives of them are not necessarily zero.
Differentiating the Riemann tensor with respect to some coordinate xλ , we get
ρ
∇λ Rσµν = ∂λ ∂µ Γρνσ − ∂λ ∂ν Γρµσ . (3.37)

Cyclically permuting the indices according to Figure 6, we find

ρ
∇µ Rσνλ = ∂µ ∂ν Γρλσ − ∂µ ∂λ Γρνσ , (3.38)
ρ
∇ν Rσλµ = ∂ν ∂λ Γρµσ − ∂ν ∂µ Γρλσ . (3.39)

The sum of equations (3.37)-(3.39) gives

2
For a proof, see [4].

15
 Bachelor of Science Thesis in Physics B. Cavcic

ρ ρ ρ
∇λ Rσµν + ∇µ Rσνλ + ∇ν Rσλµ =0 , (3.40)

which is known as the differential Bianchi identity. There is a contracted form of this equation which
we can obtain by first letting ρ = µ, giving
ρ
∇λ Rσν + ∇ρ Rσνλ − ∇ν Rσλ = 0. (3.41)

Recall from (3.34) that a sign change occurs when contracting the third covariant index, hence the minus
sign in the third term.
Multiplying by g σν leads to
ρν
∇λ R + ∇ρ Rνλ − ∇ν Rλν = 0,
(3.42)
⇔ ∇λ R − ∇ρ Rλρ − ∇ν Rλν = 0.

We may now relabel ν → µ, arriving at

∇λ R − 2∇ρ Rλρ = 0 . (3.43)

The above equation is known as the contracted Bianchi identity.

Einstein Tensor
Using that ∇µ = gµν ∇ν , the contracted Bianchi identity (3.43) can be rewritten (with some relabeling)
as
1
∇µ Rµν − ∇ν R = 0, (3.44)
2
or equivalently
1
∇µ (Rµν − Rgµν ) = ∇µ Gµν = 0, (3.45)
2
where
1
Gµν := Rµν − Rgµν (3.46)
2

is known as the Einstein tensor [4]. As promised, we have derived a tensor expression that includes the
Ricci tensor and Ricci scalar, and vanishes when differentiated.

16
 Bachelor of Science Thesis in Physics B. Cavcic

3.6 Einstein Field Equations

In the previous sections, we put great effort into finding a description of curvature that is compati-
ble with the energy-momentum tensor. We arrived at the Einstein tensor, which satisfies each of the
following conditions:

1 rank: 2,

2 symmetry: Gµν = Gνµ ,

3 conservation: ∇ρ Gµν = 0,

These three conditions allow for the Einstein tensor to be set equal to the energy-momentum tensor,

Gµν = κTµν , (3.47)

where κ is a normalization constant. As is shown in [4], κ can be found by considering the above
equation in the Newtonian limit, where particles are slow-moving (relative to the speed of light) and
the gravitational field is weakly perturbed. Poisson’s equation (3.1) can then be used to infer that κ =
8πG/c4 [4], where G is the universal gravitational constant and c the speed of light.
We may now announce the Einstein field equations,

8πG
Gµν = Tµν . (3.48)
c4

These are the full set of equations that describe how spacetime is curved in the presence of matter.
Provided a metric, any object in 4D spacetime can be traced by solving the 10 equations3 contained in
(3.48).
From now on, we will adopt the common convention that c is set to 1.

3.7 Einstein-Hilbert Action

The Lagrangian approach to classical mechanics is a powerful method for obtaining the equations of
motion [10]. The same thing can be said about general relativity, and in this section we will show that
the action whose variation gives the Einstein field equations is the Einstein-Hilbert action, given by
Z p  
1
S= −|g| R + LM dn x, (3.49)
16πG
3
While the Einstein field equations contain 16 components in four dimensions, symmetry of both the Einstein tensor
and energy-momentum tensor reduces the number of independent ones to 10.

17
 Bachelor of Science Thesis in Physics B. Cavcic

where |g| is the determinant of the metric, R the Ricci scalar and LM a Lagrangian density describing a
matter field. To begin with, we split S into S = S0 + SM , where
Z p
1
S0 := −|g|R dn x, (3.50)
16πG
Z p
SM := −|g|LM dn x. (3.51)

Varying S0 , we find Z h p
1 p i
δS0 = Rδ −|g| + −|g|δR dn x. (3.52)
16πG
p
In Appendix A, we show that the variation of −|g| and R are given by
p 1p
δ −|g| = − −|g|gµν δg µν , (A.12)
2
δR = Rσν δg σν + ∇µ [g σν δ(Γµνσ ) − g σµ δ(Γµµσ )]. (A.18)

Inserting (A.12) and (A.18) into (3.52), we find

Z  
1 1 p
δS0 = Rµν − Rgµν −|g|δg µν dn x
16πG 2
Z (3.53)
1 p
+ ∇µ [g σν δ(Γµνσ ) − g σµ δ(Γµµσ )] −|g| dn x.
16πG
The second term is an integral over the covariant divergence of a vector, which will only yield boundary
terms according to Stokes’ theorem [4]. It is known as the Gibbons-Hawking-York (GHY) boundary term
and is zero unless there is a boundary in the spacetime [11]. The GHY term is an object of interest when,
for example, computing the entropy of a black hole [12]. In our case, however, we assume that there is
no boundary so the second integral vanishes. We are left with only the first term,
Z  
1 1 p
δS0 = Rµν − Rgµν −|g|δg µν dn x. (3.54)
16πG 2

Setting δS0 = 0 to find stationary points and requiring that the equation holds for any δg µν , we obtain
1
Rµν − Rgµν = 0. (3.55)
2

18
 Bachelor of Science Thesis in Physics B. Cavcic

We see that varying S0 gives the Einstein field equations (3.48) with no matter. To obtain the full form,
we vary the Einstein-Hilbert action (3.49), which gives
Z "   p #
1 1 1 δ( −|g|LM ) p
δS = Rµν − Rgµν + p µν
−|g|δg µν dn x. (3.56)
16πG 2 −|g| δg

Once again, setting δS = 0 and requiring that the integral vanishes for all δg µν , we obtain
  p
1 1 1 δ( −|g|LM )
Rµν − Rgµν + p = 0. (3.57)
16πG 2 −|g| δg µν

Defining the energy-momentum tensor to be

p
2 δ( −|g|LM )
Tµν := − p , (3.58)
−|g| δg µν

we get
1
Rµν − Rgµν = 8πGTµν , (3.59)
2
which is the full form of the Einstein field equations, as promised.
The Einstein-Hilbert action is a powerful tool to obtain the equations of motion for a given matter
field. This will later prove useful when describing the physics of inflation.

4 Cosmology
In the following pages, we will solve the Einstein field equations for large-scale cosmological structure,
allowing for a precise account of the evolution of the universe. We will begin by introducing a metric
that is applicable to the whole universe (§4.1), which in turn leads to the Friedmann equations §(4.2)
that describe cosmological evolution. Later in §4.3, we will solve the Friedmann equations for partical
setups. This is enough to outline standard cosmology, which amounts to the final subsection (§4.4).
After this section, we will have gathered enough information to highlight inconsistencies in the
standard model of cosmology, which will ultimately lead to the theory of inflation.

4.1 FLRW Metric

In order to describe the large-scale evolution of the universe, it is necessary to solve the Einstein field
equations, which in turn require a metric that accurately captures the geometry of the universe at
the largest scales. The most simplifying assumption is the cosmological principle, which states that

19
 Bachelor of Science Thesis in Physics B. Cavcic

Figure 7: The three possible geometries for a homogeneous and isotropic universe. Left: Hyperbolic(
κ = −1). Middle: Flat (κ = 0). Right: Spherical (κ = +1).

the universe is homogeneous4 and isotropic5 . Although this assumption is the foundation of standard
cosmology, it is worth noting that there is research opposing the cosmological principle [13], [14]. In
such a universe, there are topologically only three realizable geometries (Figure 7),

1 hyperbolic (open),

2 flat,

3 spherical (closed).

What this means mathematically is that all points in spacetime satisfy the equation given by the assigned
geometry. For example, in a 3D embedding of a 2D sphere all points are described by the equation

r2 = a2 . (4.1)

We can generalize this to a 3D embedding in 4D spacetime,

t2 ± r2 = ±a2 , (4.2)

where the plus (+) sign describes a closed, spherical universe and the minus (−) is for an open, hyper-
bolic universe. The metric is subsequently given by

dℓ2 = dt2 ± dr2 . (4.3)

4
Homogeneous - translational invariance
5
Isotropic - rotational invariance

20
 Bachelor of Science Thesis in Physics B. Cavcic

By convention, we redefine our coordinates in the following way:

r → ar, t → at. (4.4)

This will turn (4.2) and (4.3) into

dℓ2 = a2 (dt2 + dr2 ), (4.5)

t2 ± r2 = ±1. (4.6)

Differentiating (4.6), we find t dt = ∓ r · dr = r dr, which means that we can replace dt in (4.5), giving
 2 2 
2 2 r dr 2
dℓ = a + dr . (4.7)
1 ∓ r2

Using spherical coordinates, we find

 2
r + (1 ∓ r2 ) 2

2 2 2 2 2 2 2
dℓ = a dr + r dθ + r sin θ dϕ . (4.8)
1 ∓ r2

To replace the ∓ sign, we set r2 → κr2 , where κ = −1, 0, +1. We then have
 
2 2 1 2 2 2 2 2 2
dℓ = a dr + r dθ + r sin θ dϕ . (4.9)
1 − κr2

In special relativity, the metric of flat spacetime is commonly written as

ds2 = −dt2 + dx2 + dy 2 + dz 2 . (4.10)

Borrowing the (−,+,+,+) signature convention, we construct the Friedmann-Lemaître-Robertson-Walker

(FLRW) metric,

 
2 2 2 1
ds = −dt + a(t) dr2 + r2 dθ2 + r2 sin2 θ dϕ2 , (4.11)
1 − κr2

where a(t) is known as the scale factor, a time-dependent parameter that describes the size of the uni-
verse relative to some reference point. The convention is that the scale factor of the universe as mea-
sured today (t = t0 ) is equal to 1, i.e. a(t0 ) := a0 = 1.
There are a few important things to note about the FLRW metric. First of all, the coordinates are
comoving coordinates; if the expansion of the universe froze today, the physical coordinates measured

21
 Bachelor of Science Thesis in Physics B. Cavcic

would be equal to the comoving coordinates. To obtain the physical distance at any other time, the
comoving coordinates are multiplied with the scale factor, for example a(t)r gives the radial distance
to some point at time t. Secondly, κ is known as the curvature parameter, and is equal to −1 for a
hyperbolic, 0 for a flat, and +1 for a spherical universe. Setting κ = 0, we see that the FLRW metric
simply reduces to the flat metric in spherical coordinates.

4.2 Friedmann Equations

Einstein’s equations put no constraints on the spacetime metric nor the matter content; they can only be
specified when such limitations are applied. As we will see in this section, two simplifying assumptions
about the universe will reduce Einstein’s set of ten equations to only two, known as the Friedmann
equations.
Starting with the Einstein field equations (3.48),
1
Rµν − Rgµν = 8πGTµν , (4.12)
2
assume that

1 the universe contains matter that can be modelled as a perfect fluid,

2 the universe is equipped with the FLRW metric.

As a consequence of the first postulate, we have from §3.3.1 that

Tµν = (ρ + P )Uµ Uν + P gµν . (4.13)

With the FLRW metric, we find the following non-zero Ricci components (see Appendix B):

ä äa + 2ȧ2 + 2κ
R00 = −3 R11 =
a 1 − κr2
R22 = r2 (äa + 2ȧ2 + 2κ) R33 = R22 sin2 θ,

and the Ricci scalar, " #

 2
ä ȧ κ
R=6 + + 2 . (4.14)
a a a

22
 Bachelor of Science Thesis in Physics B. Cavcic

In the µν = 00 case, (4.12) becomes

 2
ȧ 8πG κ
= ρ− 2 , (4.15)
a 3 a

which is known as the first Friedmann equation. The spatial components µν = ii all give the same
equation:      
ä ȧ κ
−2 − − 2 gii = 8πGP gii . (4.16)
a a a
Removing the metric from both sides and adding the first Friedmann equation (4.15) to (4.16), we obtain
the second Friedmann equation,
ä −4πG
= (ρ + 3P ) . (4.17)
a 3

Equations (4.15) and (4.17) are collectively known as the Friedmann equations, although the former is
usually referred to as the Friedmann equation.

4.2.1 Continuity Equation

The continuity equation in cosmology, also known as the fluid equation, is another differential equation
that describes the evolution of the matter energy-density. It will prove to be very useful, so we derive
it in this section.
We begin my multiplying (4.15) with a2 , giving

8πG 2
ȧ2 = ρa − κ. (4.18)
3
Next, differentiating with respect to time we obtain

8πG 2
2äȧ = (ρ̇a + 2ρȧa) (4.19)
3
ä 4πG
⇔ = (ρ̇a/ȧ + 2ρ). (4.20)
a 3
The left side is just the second Friedmann equation (4.17), hence

4πG 4πG
− (ρ + 3P ) = (ρ̇a/ȧ + 2ρ). (4.21)
3 3

23
 Bachelor of Science Thesis in Physics B. Cavcic

Rearranging, we obtain the continuity equation6 ,

ρ̇ + 3H(ρ + P ) = 0 , (4.22)

where H := ȧ/a is known as the Hubble parameter. This equation can be solved as a function of the
scale factor, first by rewriting the Hubble parameter as d ln a/dt, giving

d ln ρ
= −3(1 + w), (4.23)
d ln a
where we have defined w := P/ρ. Integrating, we find

ln ρ ∝ −3(1 + w) ln a, (4.24)

or
ρ ∝ a−3(1+w) . (4.25)

By solving the continuity equation, we have found how the energy density changes with the scale
factor. w is known as the equation of state parameter, since it is the proportionality constant in the
thermodynamic equation of state,
P = wρ. (4.26)
The equation of state parameter varies depending on the type of matter. In [15], it is shown that w = 0
for regular matter. A gas of photons, however, obeys an equation of state P = 1/3ρ, as demonstrated
in [16], hence w = 1/3 for radiation.

4.3 Model Universes

To determine the past and future evolution of the universe, it is only necessary to solve the Friedmann
equations. Doing this requires knowledge about the curvature parameter κ and the equation of state
parameter w, since the energy density is related to w by (4.25) and the pressure is related to the energy
density. In the last section, we claimed that w = 0 for matter and w = 1/3 for radiation. As we will
soon see, both of these fluids lead to a non-static universe. This was problematic for Einstein, who
longed for an immutable universe [1]. This lead to the introduction of a constant to the field equations,

Gµν + Λgµν = 8πGTµν , (4.27)

The continuity equation can also be derived from the first law of thermodynamics, dQ = dE + P dV . This is done in
6

[15].

24
 Bachelor of Science Thesis in Physics B. Cavcic

where Λ is known as the cosmological constant. It is specifically multiplied with the metric since it will
cancel nicely with the other terms in (4.16). Although the idea of a static universe was abandoned with
the discovery of cosmological expansion (see introduction to §4), it made a comeback decades later
with the additional discovery that the expansion of the universe was accelerating. Furthermore, adding
a constant to the Einstein field equations is related to inflation, as we will see in §5.
In this section, we will explore the different evolutionary scenarios in a universe without Λ. Then,
we will solve the Friedmann equations for a flat universe it will be made clear how Λ causes accelerated
expansion in a flat universe. Finally, we will present observational data that determine the spatial
geometry and matter distribution, allowing us to draw the ultimate fate of the universe.

4.3.1 The Universe Without a Cosmological Constant

Without the cosmological constant Λ, we may only consider two relevant fluids: matter and radiation.
Since the equation of state parameter w is non-negative for both of these, the following statements are
true:

1 ∂ρ/∂a < 0,

2 ρ > 0,

3 P > 0.

The first and second statements follows from (4.25), while the third is a consequence of the second due
to the equation of state (4.26). Now, recall the first Friedmann equation,
 2
ȧ 8πG κ
= ρ − 2. (4.15)
a 3 a

Together with the three statements above, we can infer the fate of the universe just by looking at the
curvature parameter κ. In the case of a flat universe (κ = 0), the time derivative of the scale factor will
always be positive but decreasing, since ρ → 0 as t → ∞. In the case of hyperbolic geometry (κ = +1),
the scale factor will never decrease and we get eternal expansion. Finally, in a closed geometry (k = +1),
the scale factor will reach an extrema which will be a maximum since ä < 0 according to the second
Friedmann equation (4.17). The results are illustrated in Figure 8.

4.3.2 Scale Factor Evolution for a Flat Universe

As we will see in §4.4, astronomical observations point to a flat universe. It is therefore helpful to study
the evolution of the scale factor for a flat geometry in greater detail. Since κ = 0 for a flat universe, the

25
 Bachelor of Science Thesis in Physics B. Cavcic

a(t)
κ = −1

κ=0

κ=1
t
Figure 8: Qualitative scale factor evolution without a cosmological constant for different geometries.
Closed geometry (k = 1) leads to a “big crunch“ scenario. Open geometry (k = −1) leads to eternal
expansion. Flat geometry (k = 0) leads to a universe that expands and comes to a halt.

first Friedmann equation (4.15) reduces to

 2
2ȧ 8πG
H = = ρ. (4.28)
a 3

Using (4.25), we get

 2
ȧ
∝ a−3(1+w) (4.29)
a
1
⇔ ȧ ∝ a− 2 (1+3w) (4.30)

Using that ȧ = da/dt, we find

1
a 2 (1+3w) da ∝ dt, (4.31)
which upon integrating gives
a ∝ t2/3(1+w) . (4.32)

Evidently, the singularity at w = −1 needs special treatment; from (4.25) we see that ρ ∝ 1 in the case
of w = −1, therefore the Hubble parameter H is constant according to the Friedmann equation (4.15).
This also means that the constant Λ in (4.27) is equivalent to a fluid with w = −1. Using the definition

26
 Bachelor of Science Thesis in Physics B. Cavcic

H := ȧ/a, we get
1
Hdt = da. (4.33)
a
Integrating, we find
a ∝ eHt . (4.34)
We conclude that a universe dominated by a cosmological constant leads to exponential expansion. This
solution is generally known as de Sitter expansion.
We may also consider an empty universe for which ρ = 0. In that case the Friedmann equation is
simply ȧ = 0, i.e. the scale factor a is constant.
Summarizing our results, the scale factor for a flat universe evolves in the following way:

2/(3(1+w))
t
 w ̸= −1
a(t) ∝ eHt w = −1 . (4.35)

1 ρ=0


We have collected all results for different domination in Table 1. In Figure 9 all scale factors that appear
in the table are plotted. As we can see, only a universe dominated by a cosmological constant leads
to accelerated expansion. This has lead to the assumption that the “dark energy“ considered to be
responsible for the acceleration is caused by the cosmological constant Λ [35].

a(t)
w ρ(a) a(t)
ΛD
MD
MD 0 a−3 t2/3

RD 1/3 a−4 t1/2 RD

a0
ΛD −1 a0 eHt Empty

Empty 0 t
t
t0

Table 1: Scale factor evolution for a mat- Figure 9: Plot of scale factor evolution
ter (MD), radiation (RD) and Λ-dominated for different dominations in a flat universe
(ΛD) flat universe (κ = 0), respectively. (k = 0). In all models, the scale factor
must coincide with the one measured in the
present universe, a(t0 ) = a0 .

27
 Bachelor of Science Thesis in Physics B. Cavcic

4.4 Standard Cosmology

The current model that best fits observational data and describes the creation of the universe is known
as the Standard Model of Cosmology. It is also known as the Concordance Model [2] or Benchmark Model
[15]. The standard model can be seen as a combination of two models: ΛCDM (lambda-Cold Dark Mat-
ter) and the Hot Big Bang, although these are sometimes used interchangeably and even as synonyms
for the standard model itself. In this report, we will consider these two models separate and as a subset
of the standard model (See Figure 10). ΛCDM describes the distribution of fluids in the universe while
the latter is the theory that the universe evolved from a hot, dense state from which matter was syn-
thesized. In this section, we will review these in more detail, and in §5 some problems of the standard
model will be highlighted which will lead to the inflation proposal.

ΛCDM

STANDARD COSMOLOGY

HOT BIG BANG

Figure 10: The standard model of cosmology consists of the ΛCDM and Hot Big Bang model.

4.4.1 Hot Big Bang Model

Since this report is largely concerned with the dynamics of inflation, not much time will be allocated to
describing the details of nucleosynthesis. We will therefore only provide the historical context of the
big bang model and how it was developed, starting with Olbers’ paradox.
In 1823, astronomer and member of the celestial police7 Heinrich Wilhelm Olbers stated a problem
with the assumption of an infinite, homogeneous and eternally static universe [18]. The argument was
that the night sky should be dark in that case, as it would be constantly illuminated by the infinite
number of stars occupying each point.
The paradox remained a mystery until it was resolved as evidence for the expansion of the universe
was found; in a paper published by Hubble and Humason [19] in 1934, it was revealed that the redshift
of galaxies were proportional to their distance from earth. Therefore, at some distance, light emitted
from galaxies is redshifted beyond the visible part of the electromagnetic spectrum, leading to the dark
appearance of the night sky. In addition to Olbers’ paradox being resolved, the results from the 1934
7
Formally the United Astronomical Society, was a group of astronomers with the goal of finding new planets in the solar
system [17].

28
 Bachelor of Science Thesis in Physics B. Cavcic

paper also suggested that the universe is expanding, since higher redshift also implies a larger receding
velocity. Today, this is known as Hubble’s law, and was the first evidence of an expanding universe [1].
Two years after Einstein’s last paper on general relativity, he released another paper that marked
the beginning of physical cosmology [20] in which the cosmological principle8 of a homogeneous and
isotropic universe was used to solve the field equations. Assuming a closed geometry, the equations
led to a non-static universe which Einstein initially thought was erroneous. In order to have a universe
that does not change with time, Einstein introduced the cosmological constant, often denoted Λ. Later
in 1922, Alexander Friedmann generalized the situation to closed, flat and open geometries while also
allowing for dynamical universes, leading to a prediction of Hubble’s law [21].
Following cosmological models of an expanding universe, extrapolating backwards in time sug-
gested an initial singularity which Friedmann calculated must have happened 1010 years into the past.
In 1931, Georges Lemaître suggested that the initial state must have been dense and have left an after-
glow which he worded as the “vanished brilliance of the origin of the worlds“ [1]. Later in 1948, Ralph
Alpher, Hans Bethe* and Georges Lemaître proposed in the famous “αβγ“ paper [22] that the early
universe was responsible for the nucleosynthesis of the elements, forming as the universe cooled due
to expansion. Following their research on nucleosynthesis, Alpher and Gamow, together with Robert
Herman, predicted the Cosmic Microwave Background (CMB), a black-body radiation encompassing the
entire universe created in the instant of baryon-photon decoupling [1].
The collection of these results led to the hot big bang model, in which the universe evolved from a
hot, dense state that created the entire observable universe.

4.4.2 ΛCDM
When describing the distribution of matter in the universe, it is common to work with density parame-
ters, which can be defined from the Friedmann equations if recasted in a different form. The Friedmann
equation in a flat universe is
4πG
H2 = ρ. (4.36)
3
For current time t = t0 we have
3H02
ρ= =: ρcrit . (4.37)
4πG
ρcrit is known as the critical current density. Historically, it was defined as the density at which the scale
factor reaches an extrema, which is a maxima in the case of a closed universe (see §4.3.2).
We can split the energy density and pressure into different contributions,
X X
ρ := ρi p := pi . (4.38)
i i

8
Coined by Edward Milne [1].

29
 Bachelor of Science Thesis in Physics B. Cavcic

For example, ρΛ and ρm denote the energy density contribution from dark energy and matter, respec-
tively. Together with the definition of critical density (4.37), we can now define the density parameter,

ρi (t0 )
Ωi = . (4.39)
ρcrit

With this definition, we can cast the Friedmann equations in a different form,
 2
H X
= Ωi a−3(1+wi ) + Ωκ a−2 . (4.40)
H0 i

Evaluating the Friedmann equation today, t = t0 , we have H = H0 and a = a0 := 1, which leads to the
cosmic sum,
X
Ωi + Ωκ = 1 . (4.41)
i

All density parameters measured today should satisfy the above relation. In Table 2, we have collected
all density parameters as measured from the CMB by the Planck satellite. Note that we have not included
contributions from radiation or neutrinos, or any other fluid, as the vast majority is comprised of matter
and dark energy. In fact, most of the matter contribution comes from dark matter, an unknown entity
known to cause flat rotation curves in galaxies [15]. This collection of results is known as lambda-
cold dark matter, or ΛCDM, where the Λ alludes to the assumption that the cosmological constant is
responsible for dark energy.

Ωm ΩΛ Ωκ Sum H0 (km/s/Mpc)

0.3111 ± 0.0056 0.6889 ± 0.0056 0 ± 0.011 1 67.66 ± 0.42

Table 2: Latest data release from the Planck satellite [23]. H0 : Hubble constant, Ωm : matter (baryonic +
dark matter), ΩΛ : dark energy, Ωκ : curvature Parameter.

Another important observation is that Ωκ ≈ 0, i.e. the universe is very flat. If dark energy is indeed a
positive cosmological constant, we therefore conclude that the fate of the universe is eternal expansion,
see §4.3.2 for more details.
Assuming that the universe has expanded since the beginning of time, we can determine the age of
the universe by extrapolating backwards until the scale factor is equal to zero. Having evaluated the

30
 Bachelor of Science Thesis in Physics B. Cavcic

density parameters, this becomes a suprisingly simple task. To begin with, (4.40) can be rewritten as
!1/2
X
ȧ = H0 Ωi a−(1+wi ) + Ωκ , (4.42)
i

and using that ȧ = da/dt, we can integrate to obtain

!−1/2
Z a
1 X
t= Ωi a−(1+wi ) + Ωκ da. (4.43)
H0 0 i

From this equation we can obtain the age of the universe at any given scale factor. In particular, the
age of the universe today is given by a = a0 := 1. Using the best estimates from Table 2 and assuming
a flat universe, we can numerically integrate the above equation to determine the current age,
Z 1
1
−1/2
tage = ΩΛ + Ωm a−1 da ≈ 13.79 ± 0.03 Gyrs . (4.44)
H0 0

Although this result may be familiar to the reader, there is a potential systematic error to H0 , the
Hubble constant. In the section that follows, we will see how two independent methods give rise to
contradicting values of this constant, and how it may be resolved.

4.4.3 Hubble Tension

The value of the Hubble constant that is used in Table 3 was obtained by observing the oldest source
of light, the cosmic microwave background. Late-universe observations, however, consistently point to
a larger value. For example, in 2021 and using the ultra-precise Gaia space telescope, Riess et al [24]
observed 75 Milky Way Cepheids to find a value of H0 = 73.2±1.3 km/s/Mpc. The discrepancy between
this measurement and the one from the Planck satellite exceeds 5 standard deviations, rendering it a
statistically significant difference. This is known as the Hubble tension, and the cause of it is unknown.
Current attempts are being made to explain the Hubble tension, one of which relies on testing
independent methods to evalute the Hubble constant. These include using gravitational lensing of
quasars, baryon acoustic oscillations and gravitational waves [25]–[27]. The more independent methods
that are being used, the easier it will be to point to methods with systematic errors. On the other hand,
some research suggests that the Hubble tension may point to new physics beyond the ΛCDM model [28].
An example is Early Dark Energy (EDE) models, where the introduction of a dark energy component
in the early universe has been shown to alleviate the Hubble tension [29].

31
 Bachelor of Science Thesis in Physics B. Cavcic

5 Introduction to Inflation
Following the advent of standard cosmology, several problems [30] were raised connected to the early
universe, by the names of
1 the horizon problem,
2 the flatness problem,
3 unwanted relics problem (includes the monopole problem).
It turns out that all three of these problems can be solved by introducing a period of accelerated ex-
pansion in the early universe known as inflation. In the following section, we will describe the horizon
problem in greater detail, and see inflation solves it, while overlooking the flatness problem and un-
wanted relics. Before doing so, we will have to introduce two distinct but related concepts known as
the particle horizon and Hubble radius.
For two points in spacetime to be in thermal equilibrium, they must be in causal contact, which
means that there must have been enough time in the universe for light to travel from one point to the
other. Since the age of the universe is finite (see §4), the distance a light signal could have traveled since
the big bang is also finite. As a result, there is a sphere of causality for every point in spacetime beyond
which the point loses causal contact, known as the particle horizon, see Figure 11.
Pa
rticle horizo

dhor
n

Causally disconnected region

Figure 11: For each point in space, there is an associated particle horizon which is defined as the maximum
distance a light signal could have traveled since the big bang. Regions outside the horizon are causally
disconnected from the point.

In a non-expanding universe, the size of the particle horizon would simply be dhor = ct. Since we
allow for an expanding universe, we need to consider a distance that is comoving with the expansion. To

32
 Bachelor of Science Thesis in Physics B. Cavcic

calculate the size of horizon in an expanding universe, we consider the path of a light signal travelling
radially in a flat universe9 . As light travels along a null geodesic, we have ds2 = 0 so the FLRW metric
(4.11) reduces to
dt
dt2 = a(t)2 dr2 =⇒ dr = . (5.1)
a(t)
Integrating the above equation from some initial time ti to a future time t, we obtain the comoving
distance, Z t Z ti
dt
d(t) := dr = c . (5.2)
ti t a(t)
This is the distance that is comoving with the expansion. Using the definition of the Hubble parameter,
H := (ȧ/a)2 , we can make the following substitution:
Z a Z ln a
da
d(a) = c = (aH)−1 d ln a. (5.3)
ai a2 H ln ai

The factor (aH)−1 is known as the comoving Hubble radius. It is the distance at which the expansion
rate is faster than the speed of light. In the limit that ai → 0, we obtain the comoving particle horizon,
Z a Z a
da
dhor (t) = 2
= (aH)−1 d ln a. (5.4)
0 a H 0

The comoving particle horizon is the maximum distance a signal traveling at the speed of light could
have traversed since the singularity. Hence, any two points separated by a distance larger than the
comoving particle horizon are causally disconnected, i.e. they have never been in contact with each
other. This is the language needed to formulate the horizon problem.
Finally, note that (5.3) and (5.4) are coordinate distances. In order to obtain the physical distance, we
simply multiply the equations by the scale factor a, see §4.1 for further explanation.

5.1 The Horizon Problem and Inflation

It was noted even before detailed measurements by WMAP and the Planck satellite that the the temper-
ature of the CMB should be isotropic to a great extent [31]. This poses a problem in standard cosmology,
as many regions in the CMB did not have time to be in thermal equilibrium at the time of decoupling,
as we will see in this section.
9
We make these assumption since observations point to a flat universe (see §4.4), while the radial assumption is just to
simplify the equations, although a non-radial path virtually follows the same procedure.

33
 Bachelor of Science Thesis in Physics B. Cavcic

Starting off, we use (4.32) to write

a(t) = (t/t0 )2/3(1+w) , (5.5)

keeping in mind that w ̸= −1 and that t0 is the present time for which a0 := 1. From this equation we
see that H := ȧ/a = t−1 , hence
1
(aH)−1 = H0−1 a 2 (1+3w) . (5.6)
We can now compute the comoving distance (5.3),

2H0−1 h 1 (1+3w)
Z a i
1/2(1+3w)
d(a) = a−1 (aH)−1 da = a2 − ai . (5.7)
ai 1 + 3w

To compute the size of the particle horizon, we let ai → 0, which gives

2H0−1 1 (1+3w) 2
dhor (a) = a2 = (aH)−1 . (5.8)
1 + 3w 1 + 3w
From (5.7), we find that the comoving distance between two antipodal points on the CMB is

2H0−1 h 1
(1+3w)
i
d(a0 ) = 2 · 1 − aCMB
2
. (5.9)
1 + 3w
On the other hand, the comoving horizon distance (5.8) for any point on the CMB is

2H0−1 12 (1+3w)
dhor (a) = a . (5.10)
1 + 3w CMB
For all points on the CMB to be causally connected, the particle horizon must be larger than the distance
between two antipodal points, i.e.
dhor > d, (5.11)
or
2H0−1 12 (1+3w) 2H0−1 1 (1+3w)
aCMB > a2 , (5.12)
1 + 3w 1 + 3w
which gives the condition that
2 1
(1+3w)
< aCMB
2
. (5.13)
3
According to Carroll [4], aCMB ≈ 1200. For a matter or radiation-dominated universe (w = 0, w = 1/3
respectively), the above expression will not hold true. This means that there inevitably will be causally
disconnected regions in the CMB. In particular, there should be over 1000 disconnected patches, but

34
 Bachelor of Science Thesis in Physics B. Cavcic

comoving scales

(aH)−1

Density Fluctuation

In
fl ang
at B
i ig
on
otB
H

time
horizon reheating horizon CMB
exit re-entry

Figure 12: Before inflation, the Hubble sphere is larger than the temperature fluctuations, allowing all
points to thermally equilibriate. During inflation, the Hubble sphere shrinks to a size smaller than the
temperature fluctuations and starts expanding once inflation ends.

the temperature of the CMB is uniform to a remarkable 4 decimals [1]. This is the horizon problem of
cosmology.
To satisfy the inequality, we must have a fluid such that 1 + 3w < 0. This implies that w < −1/3,
and since the equation of state parameter is defined by w := ρ/P , the fluid will inevitably need to have
a negative pressure. The second Friedmann equation (4.17) then implies that

ä > 0. (5.14)

In other words, solving the horizon problem requires an accelerated expansion prior to the CMB known
as inflation. The implication of an accelerating universe allows for a simple interpretation; if the uni-
verse had a period of accelerated expansion before the CMB, it could have started from smaller scales
where all points could be in causal contact.
A more powerful way of describing inflation is in terms of the comoving Hubble sphere; accelerated
expansion implies a smaller comoving Hubble sphere, whereas decelerating expansion is associated
with a growing one10 . Without inflation, the comoving Hubble sphere grows monotonically towards
the value it obtains at the time of decoupling. On the other hand, accelerated expansion such as inflation
would give rise to a shrinking comoving horizon, that will later expand according to the standard model
10
Recall that the Hubble radius is defined as the distance at which the expansion rate equals the speed of light.

35
 Bachelor of Science Thesis in Physics B. Cavcic

after inflation ends. This allows the comoving Hubble sphere to have been initially larger than the
temperature fluctuations, allowing all points to be in causal contact. This is illustrated in Figure 12. From
the figure, we can also see how inflation explains the observable perturbations in the CMB; the initial
quantum fluctuations will freeze as the Hubble sphere shrinks to a size smaller than the fluctuations.
This fact can be used to predict what the CMB will look like, although we will not discuss this more in
the thesis.

5.2 The Inflaton Field

As we saw in §4.3, introducing a constant to the Einstein field equations leads to accelerated expansion.
In principle, we could associate inflation with a constant in the Einstein field equations, but this does
not allow for a braking mechanism, among other things. A more general description can be obtained by
introducing a scalar field in the Einstein-Hilbert action, ϕ(t, r), known as the inflaton with a potential
V (ϕ) [2]. This leads to
Z  
4
p 1 1 µν
S = d x −|g| R − g ∇µ ϕ∇ν ϕ − V (ϕ) = S0 + Sϕ , (5.15)
16πG 2

where
Z p 1
S0 := d4 x −|g| R, (5.16)
16πG
Z
1
Sϕ := − d4 x g µν ∇µ ϕ∇ν ϕ + V (ϕ). (5.17)
2
The dynamics of the field are described by the equations of motion, which can be obtained using the
Euler-Lagrange (E-L) equations11 [4],
 
δS ∂L ∂L
= − ∂µ = 0, (5.18)
δϕ ∂ϕ ∂(∂µ ϕ)

where L is the integrand of the action. For Sϕ , the two terms are given by

∂L dV ∂L
=− , = −g µν ∂ν ϕ. (5.19)
∂ϕ dϕ ∂(∂µ ϕ)
11
We briefly switch to notation in terms of partial derivatives for convenience.

36
 Bachelor of Science Thesis in Physics B. Cavcic

To obtain the rightmost expression, the partial derivative of the E-L equations has to be taken with care,
see Appendix A for a detailed walkthrough. Equation (5.18) now reads

dV
g µν ∇µ ∇ν ϕ − = 0, (5.20)
dϕ
or contracted,
dV
∇µ ∇µ ϕ − = 0. (5.21)
dϕ
Since ∇µ ϕ is a (1,0) tensor, we can use (3.14) to obtain

∇µ ∇µ ϕ = ∂µ (∇µ ϕ) + Γµµν (∇µ )ϕ = ∂µ ∂ µ ϕ + Γµµν ∂ ν ϕ. (5.22)

In the last equality we used that the covariant derivative of a scalar is defined to be equal to the partial
derivative, see Appendix A.
Equation (5.21) is therefore
dV
∂µ ∂ µ ϕ + Γµµν ∂ ν ϕ − = 0. (5.23)
dϕ
We now let the field be spatially homogeneous, i.e. ϕ(t, r) = ϕ(t), and assume the FLRW metric for
which the corresponding Christoffel symbols are given in Appendix B. Upon doing this, we obtain the
following equation of motion:
dV
ϕ̈ + 3H ϕ̇ + =0 . (5.24)
dϕ

Next, we want to compute the energy-momentum tensor. Taking the variations of Sϕ , we find
Z     
4
p 1 µν p 1 µν
δSϕ = d x δ( −|g|) − g ∇µ ϕ∇ν ϕ − V (ϕ) − −|g| δg ∇µ ϕ∇ν ϕ
2 2
Z    
4 1p µν 1 µν 1p µν
= dx − −|g|gµν δg − g ∇µ ϕ∇ν ϕ − V (ϕ) − −|g|δg (∇µ ϕ∇ν ϕ)
2 2 2
Z    
4 1 1 ρσ
µν
p
= d x δg −|g| gµν g ∇ρ ϕ∇σ ϕ + V (ϕ) − ∇µ ϕ∇ν ϕ .
2 2
(5.25)

37
 Bachelor of Science Thesis in Physics B. Cavcic

The definition of the energy-momentum tensor (3.58) leads to

 
2 δSϕ 1 ρσ
Tµν := − p = −gµν g ∇ρ ϕ∇σ ϕ + V (ϕ) + ∇µ ϕ∇ν ϕ . (5.26)
−|g| δg µν 2

Along with the assumption of a perfect fluid, we get

1 1
T00 = ρ = ϕ̇2 + V (ϕ) , Tii = P = ϕ̇2 − V (ϕ) , (5.27)
2 2

and the equation of state parameter

1 2
P ϕ̇ − V (ϕ)
w := = 21 . (5.28)
ρ 2
ϕ̇2 + V (ϕ)

Equation (5.27), together with (5.24) and the Friedmann equations (4.15, 4.17), give enough information
to describe the evolution of the inflaton.
Of course, introducing a scalar field to the Einstein-Hilbert action does not imply accelerated ex-
pansion; it is a condition that has to be imposed. This will be the topic of the next section.

5.3 Slow-roll Inflation

In the previous section, we derived the equations necessary for knowing how the inflaton will evolve.
However, we need to impose the following conditions to actually ignite inflation and account for astro-
nomical observations:

1 Accelerated expansion; ä > 0,

2 de Sitter inflation; 12 ϕ̇2 ≪ V ,

3 Slow-roll; |ϕ̈| ≪ |3H ϕ̇|, |V ′ |.

We impose the first condition because accelerated expansion is the definition of inflation (see §5.2). The
latter two are connected to observational constraints; according to (5.27), it is only possible to secure a
negative pressure if 12 ϕ̇2 < V . In addition, the equation of state parameter w needs to be close to −1
in order to account for the power spectrum observed in the CMB [32]. From (4.26), we see that this is
equivalent to having 21 ϕ̇2 ≪ V . The third condition comes from the inflaton equation of motion (5.24)

38
 Bachelor of Science Thesis in Physics B. Cavcic

and is there in order to have an inflationary period that does not end too soon, which will prove to be
important.
From the assumptions above, we wish to derive potential-dependent parameters such that we can
determine the conditions for slow-roll inflation for any potential. These are known as slow-roll param-
eters [33]. In this report, will derive and use the two lowest-order parameters, commonly denoted ϵ and
η.
Using conditions 2 and 3 from above, the equation of motion (5.24) and first Friedmann equation
(4.15) reduce to

3H ϕ̇ + V ′ = 0, (5.29a)
 2
2 ȧ 1
H = = V, (5.29b)
a 3MPl2
√
where we have defined the Planck mass12 , MPl := 1/ 8πG. Now, with the first condition the second
Friedmann equation reads
ä
= H 2 + Ḣ > 0, (5.30)
a
i.e.
Ḣ
− 2 < 1. (5.31)
H
Combining (5.29a) and (5.29b), the condition becomes

2
M2 V′

Ḣ
ϵ := − 2 = Pl <1 . (5.32)
H 2 V

This is the first slow-roll parameter, and as promised it is only dependent on the potential. Inflation
occurs as long as ϵ < 1 and ends when ϵ = 1.
We can define another parameter that must be small. Based on condition 3, we define

ϕ̈
η := − ≪ 1. (5.33)
H ϕ̇
12
Here we assume ℏ = 1.

39
 Bachelor of Science Thesis in Physics B. Cavcic

To find η as a function of the potential, we differentiate (5.24),

3(Ḣ ϕ̇ + H ϕ̈) + V ′′ ϕ̇ = 0
!
Ḣ ϕ̈ V ′′
⇔ 3 + + 2 = 0.
H 2 H ϕ̇ H

Using the definition of ϵ and (5.29b) to replace the Hubble parameter, we get

V ′′
3(−ϵ − η) + 3MPl2 =0
V

V ′′
⇔ |η| = MPl2 −ϵ ≪1 . (5.34)
V

For prolonged inflation, we therefore require that the expression above is kept small.
There is yet another parameter that is important for comparison with astrophysical observations,
namely the number of e-folds, a parameter relating the scale factor before and after inflation. From the
Hubble parameter, we have

Z tf
ȧ 1 af
H := =⇒ da = Hdt =⇒ N := ln = Hdt . (5.35)
a a ai ti

where N is the number of e-folds. Ideally, we would want to recast this integral such that it is potential-
dependent. To begin with, we make a change of coordinates, which leads to
Z tf Z ϕf
H
N= Hdt = dϕ. (5.36)
ti ϕi ϕ̇
Using (5.29a) and (5.29b), we find
Z ϕf Z ϕf
1 V 1 V
N =− 2 dϕ = − dϕ. (5.37)
Mpl ϕi V′ ϕi Mpl2 V ′

40
 Bachelor of Science Thesis in Physics B. Cavcic

Finally, using the definition of the slow roll parameter (5.32), we get

Z ϕf
1 dϕ
N =− √ . (5.38)
ϕi 2ϵ Mpl

ϕf is uniquely defined by the condition for the end of inflation, ϵ = 1.

To solve the horizon problem, the number of e-folds has to exceed [2]

N ≳ 60 . (5.39)

Alas, we have gathered everything needed to explore the rich catalogue of inflationary models whose
number exceeds 110 in the ASPIC library of inflatons [33].

6 Case Studies
In the previous section we saw how inflation, a period of accelerated expansion previous to CMB decou-
pling, can solve the horizon problem. We also saw that inflation can be described by a scalar inflaton
field. The explicit form of the field, however, is generic; by just changing the potential, we can get
different models for inflation.
In this section, we will take a look at three different models of inflation and see how the slow-roll
conditions constrain the field. In particular, we will be interested in calculating the initial value of the
inflaton.

6.1 Chaotic Inflation

Up to this point, we have not discussed whether inflation occurs simultaneously everywhere or is
position-dependent. Linde [34] proposed in 1983 that the inflaton field may satisfy the slow-roll con-
ditions in certain patches, triggering inflation and creating a bubble universe which an observer would
see as homogeneous and isotropic. This model of inflation is called chaotic inflation, also known as
eternal inflation or large-field inflation (LFI).
Large-field models have a potential of the form V (ϕ) ∝ ϕp , where p is an integer or in some models
a rational number [33]. We will take a look at the p = 2 scenario, which is also known as massive scalar
field inflation. The potential is then
1
V (ϕ) = m2 ϕ2 , (6.1)
2
where m is the mass of the field. We let the field roll down to the origin, see Figure 13. The slow-roll

41
 Bachelor of Science Thesis in Physics B. Cavcic

parameters (5.32 and 5.34) are then

 2
Mpl
ϵ=2 , η = 0. (6.2)
ϕ

Inflation ends when ϵ = 1, hence

2
√

Mpl
ϵ=2 = 1 =⇒ ϕf = 2Mpl . (6.3)
ϕf

Next, we can compute the number of e-folds through (5.38),

√
Z 2Mpl
1 dϕ 1 2 1
N =− √ = ϕ − . (6.4)
ϕi 2ϵ Mpl 4Mpl2 i 2

Requring that N = 60, we find the condition

ϕi ≈ 16Mpl . (6.5)

In addition to the above requirement, an upper limit on the mass of the scalar field m can be obtained by
considering perturbations of the inflaton. These are perturbations such that ϕ → ϕ + δϕ, which causes
different parts of the universe to end inflation at different times. These provide density fluctuations
observable in the CMB. The theory behind cosmological perturbations is beyond the scope of this thesis,

V (ϕ)

ϕ̇

ϕ
ϕf ϕi

Figure 13: Slow-roll with a massive scalar field. The inflationary phase takes place in the shaded region.

42
 Bachelor of Science Thesis in Physics B. Cavcic

but we cite the result obtained when normalizing these fluctuations [2]:

m < 10−6 MPL . (6.6)

Since M Pl ≪ 1, m is a very small number. In general, LFI’s have small proportionality constants
[4], which is known as weak coupling. In the second case study, we will take a closer look at this
phenomenon.

6.1.1 The Vacuum Catastrophe

There are many appealing aspects of eternal inflation, one of them being its relation to a currently
unresolved problem in physics; the vacuum catastrophe.
In the ’90s, evidence was put forward that the expansion of the universe is accelerating (see §4.4).
The cause of the acceleration, often referred to as dark energy, has been proposed to be the cosmological
constant [35]. As seen in Figure 9, a universe dominated by Λ leads to accelerated expansion in a flat
universe. In §4.3.2 we also showed that a cosmological constant is equivalent to a constant vacuum
energy that permeates all of space. This vacuum energy can be independetly calculated from quantum
field theory, and it turns out to be 120 orders of magnitude larger than the value of the cosmological
constant [36]. This is known as the vacuum catastrophe, or cosmological constant problem .
Several attempts have been made to explain this discrepancy, one of which was proposed by Linde
[37], related to eternal inflation. Initially, the universe is driven by an inflaton ϕ and a subsequent slow
roll of the so-called “quintessence“ field Φ. Perturbations of the field Φ are generated by ϕ, leading to
different field values in each universe. This leads to a multiverse in which each universe is characterized
by a different cosmological constant. In this scenario, the universe in which we inhabit happens to be
one of a small cosmological constant. This anthropic solution to the vacuum catastrophe has been
further developed in the context of string theory, were models of stable string theory vacua are being
investigated. Upon calculating the number of possible vacua, the multiverse is found to consist of 10500
universes, known as the “string theory landscape“ [37].
Theories of the multiverse and anthropic solutions have draw significant critique, in particular with
regards to the philosophy of science. According to G.F.R Ellis [38], a good scientific theory is constructed
with the smallest set of elements (Occam’s razor), is empirically testable, shows explanatory power and
is connected to the rest of science. A common argument against the multiverse theory is therefore
related to its empirical testability; universes external to our own are in principle unobservable, hence
no direct verification of the multiverse theory is possible and can not be justified on scientific grounds.
Additionally, it may seem that the establishment of a large number of universes to anthropically explain
the vacuum catastrophe is an overly complicated solution to a single problem [39].
On the contrary, proponents of the multiverse theory have called to a modification of the norms
of science [39]; for example, despite a lack of empirical testability, the existence of a multiverse can be
indirectly inferred from observations that the inflaton is given by a large field [37].

43
 Bachelor of Science Thesis in Physics B. Cavcic

6.2 Natural Inflation

In the previous section, we discussed how large-field inflatons have weak coupling constants. This has
posed a problem in particle physics, as these constans have to be fine-tuned in order to stay small in the
case of quantum corrections. This lead to the proposal of natural inflation by Freese et al. [40] in 1990.
The fine-tuning problem was solved by showing how a type of particle known as the Nambu-Goldstone
boson naturally gives rise to inflation while keeping the coupling constants small. These particles have
a potential of the form   
4 ϕ
V (ϕ) = M 1 + cos , (6.7)
f
where M is a normalization factor and 1/f a frequency parameter. We let the inflaton roll down the
slope given by ϕ ∈ (0, ϕf ), see Figure 14. The slow-roll parameters (5.32, 5.34) are

Mpl2 sin2 (ϕ/f ) Mpl2

ϵ= , η=− . (6.8)
2f 2 [1 + cos(ϕ/f )]2 2f 2

Requiring that |η| ≪ 1, we get the following condition:

√
f ≫ MPl / 2. (6.9)

V (ϕ)

2M 4

ϕ̇

ϕ
πf

Figure 14: Slow-roll with a natural inflaton field. The inflationary phase may take place in the shaded
region.

44
 Bachelor of Science Thesis in Physics B. Cavcic

ϕi

cos
−1 (1
− 2e
−N M P l
2 /f

l
MP
2
)

Trans-
Sub-Planck
Planck

MPl
M0

Figure 15: The initial value of the inflaton is trans-planckian in the red regime and sub-planckian in the
green.

Setting ϵ = 1 we obtain the final value of the inflaton,

ϕf −1
1 − 2f 2 /Mpl2
= cos , (6.10)
f 1 + 2f 2 /Mpl2

The number of e-folds are given by

ϕf
f2 1 − cos(ϕf /f )
Z
1 dϕ
N =− √ = 2 ln . (6.11)
ϕi 2ϵ Mpl Mpl 1 − cos(ϕi /f )

Inverting to get ϕi , we find

ϕi = f cos−1 1 − (1 − cos(ϕf /f )) exp −N MPl2 /f 2 .



(6.12)

Replacing with ϕf /f ,
" #
4f 2
ϕi = f cos−1 exp −N Mpl2 /f 2 .


1− 2 (6.13)
Mpl + 2f 2

45
 Bachelor of Science Thesis in Physics B. Cavcic

Using condition (6.9), we get

ϕi ≈ f cos−1 1 − 2 exp −N MPl2 /f 2 .



(6.14)

Now the question whether this initial value can be trans-planckian or not remains. In Figure 15, we
have plotted ϕi against the Planck mass. Although the function is dependent on the parameters N and
f , the function will retain its shape, hence there is no loss of generality. We find that there is both a
trans-planckian and sub-planckian regime.

6.3 Valley Hybrid Inflation

There are models of inflation that include more than one field, known as multifield inflation [41]. A
well-known two-field model is the valley hybrid model, with the following potential [33]:
1 1 1
V (ϕ, ψ) = λ′ (ψ 2 − M 2 )2 + m2 ϕ2 + λϕ2 ψ 2 . (6.15)
4 2 2
The potential can look very different depending on the parameters, but the general idea is one that
looks like Figure 16. The field is assumed to roll along the valley where ψ = 0, and later fall into one
of the two minima clearly visible in the figure. In this regime, the potential reduces to the following
ϕ˙

Figure 16: Valley hybrid inflation potential with λ′ = 2, M = 1, m = 0.2 and λ = 0.64.

46
 Bachelor of Science Thesis in Physics B. Cavcic

single-field scenario: "

 2 #
ϕ
Veff (ϕ) = ∆4 1 + . (6.16)
µ
√ √
where we have defined ∆ := M λ′1/4 / 2 and µ := 2M 2 /m.
The slow roll parameters are

2MPl2 ϕ2 2MPl2 µ2
ϵ= , η= . (6.17)
µ4 (1 + ϕ2 /µ2 )2 (µ2 + ϕ2 )2

The number of e-folds is

Z ϕf  
1 dϕ 1 2 ϕi 1 2
N =− √ = µ ln + (ϕi − ϕf ) . (6.18)
ϕi 2ϵ MP l 2MPl2 ϕf 2

To find ϕf , we set ϵ = 1 and get

2MPl2 ϕ2f 2
√
1=
2 =⇒ ϕf − 2MP l ϕf + µ2 = 0. (6.19)
µ4 1+ ϕ2f /µ2

Solving the above equation leads to

r
MP l MPl2
ϕf = √ ± − µ2 . (6.20)
2 2

Consequently, we get the following condition:

MPl
µ≤ √ . (6.21)
2
At this point, we would like to invert (6.18) to obtain the value of the initial field ϕi . Unfortunately, this
can not be done analytically. From our analysis, we can therefore only conclude the existence of the
above condition.

7 Conclusion
In this thesis, we have explored the foundation of general relativity by motivating each component in
the Einstein field equations and found the corresponding Einstein-Hilbert action. From this, we saw
how the cosmological principle and assumption that matter is a perfect fluid leads to the FLRW metric,

47
 Bachelor of Science Thesis in Physics B. Cavcic

and how the Einstein field equations under this metric yield the Friedmann equations that govern the
evolution of the universe. Within this framework, we detailed the standard model of cosmology, and
referred to observations that point to a spatially flat universe dominated by a cosmological constant.
Additionally, we used astrophysical observations to compute the age of the universe and found a value
of 13.79 ± 0.03 Gyrs.
Next, we saw how standard cosmology predicts a particle horizon size that is smaller than the CMB,
disallowing for the observed homogeneity and isotropy. We saw how this naturally leads to the need
for an accelerated expansion of the universe known as inflation.
Using the Lagrangian approach to general relativity that was outlined in §3.7, we assumed a scalar
field known as the inflaton with a slowly rolling potential as a driving mechanism. This lead to a
general equation of motion for the inflaton, and we defined two parameters that characterize this type
of inflation. Under these circumstances, we studied three types of inflation in order to see if the field
exceeds the Planck mass. We found that the first scenario, chaotic inflation, is always trans-planckian,
whereas the second field, natural inflation, has parameter-dependent regimes that are either trans-
planckian or sub-planckian. Finally, we tested a multifield scenario known as valley hybrid inflation,
and found that this leads to an expression that can not be inverted analytically.
In this thesis, we have restricted ourselves to the discussion of classical inflation. In reality, the the-
ory can be expanded to encompass cosmological perturbations. These lead to two additional parameters,
the scalar spectral index ns , which describes how density fluctuations vary with scale, and the tensor-
to-scalar ratio r, defined as the ratio of amplitude of gravitational waves generated by inflation to the
amplitude of scalar perturbations. In principle, these parameters can be measured, and the models that
have been discussed in this thesis can be further investigated to give observable predictions.

7.1 Ethical Considerations

In §6 we have explored models of inflation that make extraordinary predictions; chaotic inflation was
the first theory to suggest the existence of a multiverse. As we have already touched upon in §6, it is
unclear whether the existence of such a multiverse is empirically testable or can even be considered as a
good scientific theory, should it provide an anthropic explanation to the vacuum catastrophe. Further-
more, natural inflation was developed as an attempt to solve the fine-tuning problems of other models.
However, it is unclear whether the fine-tuning problem is real or of anthropic nature. If there is a pos-
sibility for the latter case, is it then worth to allocate time and resources to such scientific endeavours?

48
 Bachelor of Science Thesis in Physics B. Cavcic

Appendix A Mathematical Derivations

General Covariant Derivative
We will prove the equation shown in (3.14). Consider the covariant derivative of a dual vector,

∇µ ων = ∂µ ων + Γ̃ρµν ωρ . (A.1)

We want to relate Γ̃ρµν to Γρµν . The covariant derivative of a scalar k is equal to the partial derivative,

∇µ k = ∂µ k, (A.2)

hence the inner product reads

∇µ (ων V ν ) = ∂µ (ων )V ν + ων ∂µ V ν . (A.3)

But the above equation is also equal to

∇µ (ων V ν ) = ∇µ (ων ) V ν + ων (∇µ V ν )

(A.4)
= ∂µ (ων )V ν + Γ̃ρµν ωρ V ν + ων ∂µ V ν + ων Γνµρ V ρ .

Comparing (3.8) and (3.9) we find

Γ̃ρµν = −Γρµν . (A.5)
We can now extend the covariant derivative for an arbitrary (n,m) tensor,

∇ρ T µ1 µ2 ...µn ν1 ν2 ...νm
. (A.6)
= ∂ρ T µ1 µ2 ...µn ν1 ν2 ...νm + Γµρσ1 T σµ2 ...µn ν1 ν2 ...νm + · · · − Γσρν1 T µ1 µ2 ...µn σν2 ...νm − . . .

Variation of |g| and R

Variation of the Determinant
For an nxn matrix A, the following holds true:

ln |A| = Tr ln A, (A.7)

as can be shown by diagonalizing the matrix A, and showing that the equation is true for the diagonal-
ized matrix where A → diag(λ1 , λ2 , ...λn ).

49
 Bachelor of Science Thesis in Physics B. Cavcic

Varying the equation above we obtain:

1
δ|A| = Tr A−1 δA =⇒ δ|A| = |A|T r(A−1 δA).


(A.8)
|g|

Hence for the determinant of the metric g 13 ,

δ|g| = gg µν δgµν . (A.9)

We want to express all terms in δS as variations of the inverse of the metric, g µν . For matrices, we know
that δA−1 = A−2 δA, hence in index notation

δgµν = −gµλ gνρ δg λρ , (A.10)

so
δ|g| = −gg µν gµλ gνρ δg λρ = −ggλρ g λρ = −ggµν g µν . (A.11)

p the last step we relabeled λ and ρ to µ and ν, respectively. Finally, we can compute the variation of
In
−|g|,
p 1 1 1 g 1p
δ −|g| = − p δg = p gµν δg µν = − −|g|gµν δg µν . (A.12)
2 −|g| 2 −|g| 2

Variation of the Ricci Scalar

Recall the definition of the Riemann tensor,
ρ
Rσµν = ∂µ Γρνσ + Γρµλ Γλνσ − (µ ↔ ν). (A.13)

Varying, we find
ρ
δRσµν = ∂µ (δΓρνσ ) + δ(Γρµλ )Γλνσ + Γρµλ δ(Γλνσ ) − (µ ↔ ν). (A.14)
There is a way of rewriting the equation above in terms of covariant derivatives of the connection, which
will prove to be useful. Since the variation of a connection is the difference between two connections,
we have that δΓρµν is a (1,3) tensor. Hence, from (3.14) we have that

∇µ (δΓρνσ ) = ∂µ (δΓρνσ ) + Γρµλ δ(Γλνσ ) − Γλµν δ(Γρλσ ) − Γλµσ δ(Γρνλ ). (A.15)

Recognizing equal terms from A.14 leads to

ρ
δRσµν = ∇µ (δΓρνσ ) − ∇ν (δΓρµσ ). (A.16)
13
This equation is known as Jacobi’s formula.

50
 Bachelor of Science Thesis in Physics B. Cavcic

Now we can contract by letting ρ → µ to obtain the variation of the Ricci tensor,

δRσν = ∇µ (δΓµνσ ) − ∇ν (δΓµµσ ). (A.17)

Finally, we find the variation of the Ricci scalar by varying the trace R = g µν Rµν :

δR = Rµν δg σν + g σν ∇µ (δΓµνσ ) − ∇σ (δΓµµσ ) = Rσν δg σν + ∇µ (g σν δ(Γµνσ ) − g σµ δ(Γµµσ )). (A.18)

In the last step, we used that ∇σ = g σµ ∇µ .

The Second Term in the Euler-Lagrange Equations

In §4.2.1, we presented the following result:

dL
= g µν ∂ν ϕ, (A.19)
d(∂µ ϕ)

for the Lagrangian density

1
L = − g µν ∇µ ϕ∇ν ϕ − V (ϕ). (A.20)
2
The lagrangian density is scalar, hence there are no free indices. When differentiating it with respect to
something with a lower index, however, the result should be something with an upper index. A simple
way of making sure this happens is to relabel all indices in the expression to be differentiated, and then
take the derivative. Furthermore, we have the general fact that

∂(∂µ A)
= δνµ . (A.21)
∂(∂ν A)

With all of this in mind, we can now differentiate the first term of the Lagrangian:
 
∂ 1 ρσ 1 1 1 1
g ∂ρ ∂σ ϕ = g ρσ δρµ ϕ∂σ ϕ + g ρσ ∂ρ ϕδσµ = g µσ ∂σ ϕ + g ρµ ∂ρ ϕ = g µν ∂ν ϕ. (A.22)
∂(∂µ ϕ) 2 2 2 2 2

In the last step, we relabeled the dummy indices ρ and σ to ν.

Appendix B Tensor Components for the FLRW Metric

In this section we compute the desired tensor components for the FLRW metric, useful for §4.2. Prior to
this, we will present a quick method for finding the Christoffel symbols and Riemann tensor components

51
 Bachelor of Science Thesis in Physics B. Cavcic

for a given metric. To this end, we define the following isotropic metric:

ds2 = −f (r, t)dt2 + g(r, t)dr2 + h(r, t)(dθ2 + sin2 θ dϕ2 ). (B.1)

Christoffel Symbols
To quickly find non-zero Christoffel symbols for this metric, we can employ the proper time integral to
obtain the geodesic equation. To begin with, consider the definition of proper time,

dτ 2 := −gµν dxµ dxν . (B.2)

Squaring and differentiating the proper time with an arbitrary parameter λ, we obtain
1/2
dxµ dxν

dτ
= −gµν , (B.3)
dλ dλ dλ

which leads to the proper time functional

1/2
dxµ dxν
Z  Z
τ= −gµν dλ = (−F )1/2 dλ, (B.4)
dλ dλ

where F := gµν (dxµ /dλ)(dxν /dλ). Varying this integral leads to

Z Z
1
1/2
δτ = δ(−F ) dλ = − (−F )−1/2 δF dλ. (B.5)
2
Let λ = τ . Then
dxµ dxν dxµ dxν
F = gµν = gµν = gµν U µ U ν = −1, (B.6)
dλ dλ dτ dτ
where U is standard notation for the four-velocity U µ = dxµ /dτ . The last equality can be quickly
checked by dividing (B.2) by dτ 2 . With F = −1, (B.5) now equals
Z
1
δτ = − δF dτ. (B.7)
2
Hence, finding stationary points given by δτ = 0 are equivalent to finding stationary points of the
integral
dxµ dxν
Z Z
I = F dτ = gµν dτ, (B.8)
dτ dτ

52
 Bachelor of Science Thesis in Physics B. Cavcic

under the parametrization λ = τ .

Finding stationary points δI = 0 leads to the geodesic equation, as outlined in Carroll [4]. Therefore,
the Christoffel symbols can be obtained by using the following technique: let xµ → δxµ and demand
that δI vanishes for each coordinate. In the case of the isotropic metric (B.1), we have
Z
I = −f (r, t)ṫ2 + g(r, t)ṙ2 + h(r, t)(θ̇2 + sin2 θ ϕ̇2 ) dτ. (B.9)

For example, letting t → t + δt leads to f (r, t) = f + ft δt, where ft denotes the time derivative of f .
The first term of I is then
d(δt)
f ṫ2 → f ṫ2 + 2f ṫ + ft ṫ2 δt, (B.10)
dτ
and similarly for g and h. Varying I and equating to zero we get
Z  
d(δt) 2 2 2 2 2
δI = −2f ṫ − ft ṫ δt + gt ṙ + ht (θ̇ + sin θ ϕ̇ ) dτ = 0. (B.11)
dτ

Using partial integration on the first term leads to

Z h i
2f ẗ + 2f ṙt + ft ṫ2 + gt ṙ2 + ht (θ̇2 + sin2 θ ϕ̇2 ) δt dτ = 0. (B.12)

Requiring that the expression holds true for all δt, we are left with

ft 2 fr gt 2 ht 2
ẗ + ṫ + ṙt + ṙ + (θ̇ + sin2 θ ϕ̇2 ) = 0. (B.13)
2f f 2f 2f
Comparing with the geodesic equation (3.18), we identify the following components:

Γ000 = ft /f Γ011 = gt /2f

Γ022 = ht /2f Γ033 = Γ022 sin2 θ
Γ001 = Γ010 = fr /2f

Proceeding in the same way for coordinates r, θ and ϕ we can get the rest of the non-zero Christoffel
symbols:

Γ100 = fr /2g Γ111 = gr /2g

Γ122 = −hr /2g Γ133 = Γ122 sin2 θ

53
 Bachelor of Science Thesis in Physics B. Cavcic

Γ101 = Γ110 = gt /2g Γ212 = Γ221 = hr /2h

Γ202 = Γ220 = ht /2h Γ233 = − sin θ cos θ
Γ303 = Γ330 = ht /2h Γ313 = Γ331 = hr /2h
Γ323 = Γ332 = cot θ

Riemann Tensor and Ricci Tensor/Scalar

The Riemann tensor is given by
ρ
Rσµν = ∂µ Γρνσ − ∂ν Γρµσ + Γρµλ Γλνσ − Γρνλ Γλµσ . (3.27)

In four dimensions, there are 44 = 256 Riemann components, but not all of them are independent
thanks to the three symmetries of the Riemann tensor: two antisymmetries,

Rρσµν = −Rσρµν , (B.14)

Rρσµν = −Rρσνµ , (B.15)

and skew symmetry,

Rρσµν = Rµνρσ . (B.16)
This reduces the amount of independent components in the Riemann tensor to

n4 − 2n3 + 3n2 − 2n
    
1 n n
· · +1 = . (B.17)
2 2 2 8

0101 0202 0303 1212 1313 2323 0221

0331 0112 0332 0113 0223 1002 1332

1003 1223 2003 2113 0123 0213 0312

Table 3: The 20 independent components in a four-dimensional Riemann tensor are given by the above
indices.

54
 Bachelor of Science Thesis in Physics B. Cavcic

Using this equation, we find that there are only 21 independent components in four dimensions14 . The
indices of these are listed in Table 3. Computing all independent components for the isotropic metric
(B.1) we find the following non-zero terms:

frr − gtt ft gt fr gr f2 g2 ft ht fr hr h2 htt

R0101 = + − − r + t R0202 = + + t −
2 4f 4g 4f 4g 4f 4g 4h 4h
gt ht gr hr h2 hrr
R0303 = R0202 sin2 θ R1212 = + + r −
4f 4g 4h 2
 2
h2

ht
R1313 = R1212 sin2 θ R2323 = − r + h sin2 θ
4f 4g
htr fr ht gt hr ht hr
R0221 = − − − R0331 = R0221 sin2 θ
2 4f 4g 4h
The computation of the Ricci tensor and Ricci scalar are straightforward from here, but we will refrain
from presenting the final expressions as they are rather messy.
We now consider the case of the FLRW metric (4.11), for which

a(t)2
f = 1, g= , h = a(t)2 r2 . (B.18)
1 − κr2
The non-zero components of the Christoffel symbols are then
κr
Γ111 = Γ122 = −r(1 − κr2 )
1 − κr2
ȧ
Γ133 = Γ122 sin2 θ Γ101 = Γ110 =
a
1 ȧ
Γ212 = Γ221 = Γ202 = Γ220 =
r a
ȧ
Γ233 = − sin θ cos θ Γ303 = Γ330 =
a
1
Γ313 = Γ331 = Γ323 = Γ332 = cot θ
r
14
There is actually one less independent component due to the Bianchi identity, but we will ignore this.

55
 Bachelor of Science Thesis in Physics B. Cavcic

The corresponding non-zero components of the Riemann tensor are

äa
R0101 = − R0202 = −äar2
1 − κr2
a2 r2 (ȧ2 + κ)
R0303 = R0202 sin2 θ R1212 =
1 − κr2
R1313 = R1212 sin2 θ = ȧ2 + k a2 r4 sin2 θ


R2323

Contracting, we get the Ricci components,

ä äa + 2ȧ2 + 2κ
R00 = −3 R11 =
a 1 − κr2
R22 = r2 (äa + 2ȧ2 + 2κ) R33 = R22 sin2 θ

and the Ricci scalar, " #

 2
ä ȧ κ
R=6 + + 2 .
a a a

56
References
[1] G. Montani, M. V. Battisti, R. Benini, and G. Imponente, Primordial cosmology. World Scientific,
2011.
[2] D. Baumann, “Tasi lectures on inflation,” arXiv preprint arXiv:0907.5424, 2009.
[3] Nine year microwave sky, https://wmap.gsfc.nasa.gov/media/121238/index.html,
Accessed: 2023-05-17.
[4] S. Carroll, Spacetime and geometry: Pearson new international edition, 2014.
[5] J. M. Lee, Introduction to smooth manifolds (Graduate texts in mathematics, 218), eng. New York:
Springer, 2002, isbn: 0387954953.
[6] B. F. Schutz, A first course in general relativity, eng, 2nd ed. Cambridge: Cambridge University
Press, 2009, isbn: 9780521887052.
[7] W. Rindler, Introduction to special relativity, eng, 2. ed. Oxford: Clarenden Press, 1991, isbn:
0198539533.
[8] U. J. Le Verrier, “Theorie du mouvement de Mercure,” Annales de l’Observatoire de Paris, vol. 5,
p. 1, Jan. 1859.
[9] D. S. L. Soares, Newtonian gravitational deflection of light revisited, 2009. arXiv: physics/0508030
[physics.gen-ph].
[10] S. T. Thornton, Classical dynamics of particles and systems, eng, 5. ed. Belmont, Calif: Brooks/Cole
- Thomson learning, 2004, isbn: 0534408966.
[11] A. Teimouri, S. Talaganis, J. Edholm, and A. Mazumdar, “Generalised boundary terms for higher
derivative theories of gravity,” Journal of High Energy Physics, vol. 2016, no. 8, Aug. 2016. doi:
10.1007/jhep08(2016)144. [Online]. Available: https://doi.org/10.1007%2Fjhep08%
282016%29144.
[12] H. Khodabakhshi, A. Shirzad, F. Shojai, and R. B. Mann, “Black hole entropy and boundary con-
ditions,” Phys. Rev. D, vol. 101, p. 124 007, 12 Jun. 2020. doi: 10.1103/PhysRevD.101.124007.
[Online]. Available: https://link.aps.org/doi/10.1103/PhysRevD.101.124007.
[13] S. Ghosh, P. Jain, R. Kothari, M. Panwar, G. Singh, and P. Tiwari, “Probing cosmology beyond
ΛCDM using SKA,” eng, Journal of astrophysics and astronomy, vol. 44, no. 1, 2023, issn: 0973-
7758.
[14] J. Colin, R. Mohayaee, M. Rameez, and S. Sarkar, “Evidence for anisotropy of cosmic acceleration,”
Astronomy & Astrophysics, vol. 631, p. L13, 2019.
[15] B. Ryden, Introduction to cosmology. Cambridge University Press, 2017.
 Bachelor of Science Thesis in Physics B. Cavcic

[16] F. Mandl, Statistical physics (The Manchester physics series), eng, 2. ed. Chichester: Wiley, 1987,
isbn: 0471915327.
[17] H. W. Olbers, “Olbers, heinrich wilhelm (1758–1840),” 2001.
[18] M. Arpino and F. Scardigli, “Inferences from the dark sky: Olbers’ paradox revisited,” eng, Euro-
pean journal of physics, vol. 24, no. 1, pp. 39–45, 2003, issn: 0143-0807.
[19] E. Hubble and M. L. Humason, “The velocity-distance relation for isolated extragalactic nebulae,”
eng, Proceedings of the National Academy of Sciences - PNAS, vol. 20, no. 5, pp. 264–268, 1934, issn:
0027-8424.
[20] “Kosmologische betrachtungen zur allgemeinen relativitätstheorie,” ger, Die Naturwissenschaften,
vol. 7, no. 14, pp. 232–232, 1919, issn: 0028-1042.
[21] A. Friedmann, “Über die Krümmung des Raumes,” Zeitschrift fur Physik, vol. 10, pp. 377–386, Jan.
1922. doi: 10.1007/BF01332580.
[22] R. A. Alpher, H. Bethe, and G. Gamow, “The origin of chemical elements,” Physical Review, vol. 73,
no. 7, p. 803, 1948.
[23] N. Aghanim et al., “Planck 2018 results. VI. Cosmological parameters,” Astron. Astrophys., vol. 641,
A6, 2020, [Erratum: Astron.Astrophys. 652, C4 (2021)]. doi: 10.1051/0004-6361/201833910.
arXiv: 1807.06209 [astro-ph.CO].
[24] A. G. Riess, S. Casertano, W. Yuan, et al., “Cosmic distances calibrated to 1% precision with Gaia
EDR3 parallaxes and Hubble Space Telescope photometry of 75 Milky Way Cepheids confirm
tension with ΛCDM,” eng, Astrophysical journal. Letters, vol. 908, no. 1, pp. L6–, 2021, issn: 2041-
8205.
[25] J. L. Miller, “Gravitational-lensing measurements push Hubble-constant discrepancy past 5σ,”
Physics Today, vol. 73, no. 3, pp. 14–16, Mar. 2020, issn: 0031-9228. doi: 10.1063/PT.3.4424.
eprint: https://pubs.aip.org/physicstoday/article-pdf/73/3/14/10123805/14\
_1\_online.pdf. [Online]. Available: https://doi.org/10.1063/PT.3.4424.
[26] A. Cuceu, J. Farr, P. Lemos, and A. Font-Ribera, “Baryon Acoustic Oscillations and the Hubble
Constant: Past, Present and Future,” JCAP, vol. 10, p. 044, 2019. doi: 10 . 1088 / 1475 - 7516 /
2019/10/044. arXiv: 1906.11628 [astro-ph.CO].
[27] E. Trott and D. Huterer, Challenges for the statistical gravitational-wave method to measure the
Hubble constant, 2022. arXiv: 2112.00241 [astro-ph.CO].
[28] “In the realm of the hubble tension—a review of solutions,” eng ; nor, Classical and quantum
gravity, vol. 38, no. 15, pp. 153 001–, 2021, issn: 0264-9381.

58
 Bachelor of Science Thesis in Physics B. Cavcic

[29] T. Karwal and M. Kamionkowski, “Dark energy at early times, the hubble parameter, and the
string axiverse,” Phys. Rev. D, vol. 94, p. 103 523, 10 Nov. 2016. doi: 10.1103/PhysRevD.94.
103523. [Online]. Available: https : / / link . aps . org / doi / 10 . 1103 / PhysRevD . 94 .
103523.
[30] A. R. Liddle, “An introduction to cosmological inflation,” High energy physics and cosmology,
p. 260, 1998.
[31] A. H. Guth, “The Inflationary Universe: A Possible Solution to the Horizon and Flatness Problems,”
Phys. Rev. D, vol. 23, L.-Z. Fang and R. Ruffini, Eds., pp. 347–356, 1981. doi: 10.1103/PhysRevD.
23.347.
[32] L. Ackerman, W. Fischler, S. Kundu, and N. Sivanandam, “Constraining the inflationary equation
of state,” eng, Journal of cosmology and astroparticle physics, vol. 2011, no. 5, pp. 024–24, 2011,
issn: 1475-7516.
[33] J. Martin, C. Ringeval, and V. Vennin, “Encyclopædia inflationaris,” Physics of the Dark Universe,
vol. 5, pp. 75–235, 2014.
[34] A. D. Linde, “Chaotic inflation,” Physics Letters B, vol. 129, no. 3-4, pp. 177–181, 1983.
[35] S. M. Carroll, “The cosmological constant,” eng, Living reviews in relativity, vol. 4, no. 1, pp. 1–1,
2001, issn: 1433-8351.
[36] R. J. Adler, B. Casey, and O. C. Jacob, “Vacuum catastrophe: An elementary exposition of the
cosmological constant problem,” eng, American journal of physics, vol. 63, no. 7, pp. 620–626,
1995, issn: 0002-9505.
[37] A. Linde, “A brief history of the multiverse,” eng, Reports on progress in physics, vol. 80, no. 2,
pp. 022 001–022 001, 2017, issn: 0034-4885.
[38] G. F. R. Ellis, “The unique nature of cosmology,” eng, in Revisiting the Foundations of Relativistic
Physics, Dordrecht: Springer Netherlands, pp. 193–220, isbn: 9781402012853.
[39] H. Kragh, “Contemporary history of cosmology and the controversy over the multiverse,” eng,
Annals of science, vol. 66, no. 4, pp. 529–551, 2009, issn: 0003-3790.
[40] K. Freese, J. A. Frieman, and A. V. Olinto, “Natural inflation with pseudo nambu-goldstone bosons,”
Physical Review Letters, vol. 65, no. 26, p. 3233, 1990.
[41] J.-O. Gong, “Multi-field inflation and cosmological perturbations,” International Journal of Modern
Physics D, vol. 26, no. 01, p. 1 740 003, Jan. 2017. doi: 10.1142/s021827181740003x. [Online].
Available: https://doi.org/10.1142%2Fs021827181740003x.

59